Small Animal Imaging: Basics and Practical Guide

Small Animal Imaging Fabian Kiessling • Bernd J. Pichler (Editors) Peter Hauff (Co-Editor) Small Animal Imaging Ba...

Author: Fabian Kiessling | Bernd J. Pichler

110 downloads 1744 Views 31MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!

Report copyright / DMCA form

DOWNLOAD PDF

Small Animal Imaging

Fabian Kiessling • Bernd J. Pichler (Editors) Peter Hauff (Co-Editor)

Small Animal Imaging Basics and Practical Guide

Editors Prof. Dr. Fabian Kiessling Chair of Experimental Molecular Imaging University of Aachen (RWTH) Pauwelsstraße 20 52074 Aachen Germany [email protected] Prof. Dr. Bernd J. Pichler Laboratory for Preclinical Imaging and Imaging Technology of the Werner Siemens-Foundation Department of Radiology Eberhard Karls University of Tübingen Röntgenweg 13 72076 Tübingen GERMANY [email protected]

Co-Editor PD Dr. Peter Hauff Bayer Schering Pharma AG Global Drug Discovery TRG Diagnostic Imaging Müllerstr. 178 13353 Berlin Germany [email protected]

ISBN 978-3-642-12944-5 e-ISBN 978-3-642-12945-2 DOI 10.1007/978-3-642-12945-2 Springer Heidelberg Dordrecht London New York Library of Congress Control Number: 2010936069 © Springer-Verlag Berlin Heidelberg 2011 This work is subject to copyright. All rights are reserved, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilm or in any other way, and storage in data banks. Duplication of this publication or parts thereof is permitted only under the provisions of the German Copyright Law of September 9, 1965, in its current version, and permission for use must always be obtained from Springer. Violations are liable to prosecution under the German Copyright Law. The use of general descriptive names, registered names, trademarks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. Product liability: The publishers cannot guarantee the accuracy of any information about dosage and application contained in this book. In every individual case the user must check such information by consulting the relevant literature. Cover design: eStudioCalamar, Figueres/Berlin Printed on acid-free paper Springer is part of Springer Science+Business Media (www.springer.com)

Preface

During the last decade there have been tremendous advances in molecular biology and many important regulatory pathways of diseases have been identified. Along with these, genomics and proteomics are currently being implemented as important tools in the clinical workflow. On the other hand, there has been significant progress in non-invasive imaging technologies. Nowadays it is possible to scan an entire patient by CT and MRI with high spatial resolution and with exquisite tissue contrast within seconds or minutes. Contrast agents can be applied and their accumulation monitored dynamically to gain functional data about tissue vascularisation, perfusion and permeability. In addition, imaging modalities that are highly sensitive to administered radiolabelled probes like PET and SPECT enable us to elucidate changes in metabolism and proliferation as well as in molecular profiles of tissues with high sensitivity. Besides these clinically established methods there are novel promising imaging tools and applications which are currently in the stage of development. These include, for example, molecular ultrasound, high field MRI as well as photoacoustic and optical imaging. Beyond this, imaging modalities have been developed further to such a high degree that they are now able to be applied to very small animals like mice and rats for diagnostic purposes. Current dedicated small animal imaging modalities allow the in vivo assessment of morphological structures or functional, metabolic and molecular processes in mice and rats as in humans. Utilizing these tools in the preclinical arena can also significantly improve the identification and development of novel diagnostic or therapeutic drugs and facilitate the translation of preclinical findings to the clinics and vice versa. Important surrogate markers and imaging strategies can be developed and tested along with novel therapeutic drugs. Longitudinal data can be obtained from the same animal, which means that the animal can serve as its own control. In this manner the disease progression or the pharmacological effect of a drug can be monitored much more effectively. As a result high statistical power can be achieved with a reduced number of animals, which lowers costs and recognizes ethical considerations on animal protection. Non invasive imaging also has the potential to identify therapeutic drugs with limited effectiveness at a very early stage of its development. Therefore it can be used as a preclinical screening tool to boost the clinical drug success rate of currently one in five to, for example, one in three which would significantly lower the development cost for a new drug. Nevertheless, although there is no doubt about the potentially beneficial role of small animal imaging in preclinical research it has not been broadly established.

v

vi

Preface

Many imaging applications have never exceeded the Proof of Principle status and are so time consuming that it is not realistic to use them in preclinical research routinely. This often goes in line with limited data reproducibility. Besides this, in many publications non-invasive imaging acts as an appealing embellishment without having evident impact on its scientific gist. These current obstacles for the implementation of non-invasive small animal imaging are aggravated by failing studies where either a suboptimal imaging modality or contrast was chosen or where failures were made in statistical study planning and animal handling. Thus, this book aims to be a guide for all who intend to implement small animal imaging in their routine research. It provides concrete hints on how an effective small animal unit can be built up, how the personnel should be trained, where pitfalls in study planning are and which imaging modalities should be used for different purposes. Also, basic problems like the choice of the correct anesthesia and its influence on animal physiology as well as techniques of catheterization for drug administration are considered. Finally this book specifically serves as a guide for the correct and comprehensive quantification and interpretation of imaging data. We very much hope that this book will be of significant value for our readers in their daily work and we wish every success in the exciting and creative field of preclinical imaging. Aachen, Germany Tuebingen, Germany Berlin, Germany

Fabian Kiessling Bernd J. Pichler Peter Hauff

Contents

Part I Role of Small Animal Imaging 1 Noninvasive Imaging for Supporting Basic Research . . . . . . . . . . . . . . Pat Zanzonico

3

2 Non-Invasive Imaging in the Pharmaceutical Industry . . . . . . . . . . . . Sally-Ann Ricketts, Paul D. Hockings, and John C. Waterton

17

3 How to Set Up a Small Animal Imaging Unit . . . . . . . . . . . . . . . . . . . . . David Stout

29

4 Noninvasive Small Rodent Imaging: Significance for the 3R Principles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Nicolau Beckmann and Peter Maier

47

Part II Study Planning and Animal Preparation 5 Institutional Preconditions for Small Animal Imaging . . . . . . . . . . . . . René H. Tolba

61

6 Statistical Considerations for Animal Imaging Studies . . . . . . . . . . . . Hannes-Friedrich Ulbrich

69

7 Animal Anesthesia and Monitoring . . . . . . . . . . . . . . . . . . . . . . . . . . . . Marc Hein, Anna B. Roehl, and René H. Tolba

83

8 Drug Administration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Klaus Nebendahl, and Peter Hauff

93

Part III Imaging Modalities and Probes 9 How to Choose the Right Imaging Modality . . . . . . . . . . . . . . . . . . . . . 119 Fabian Kiessling, Bernd J. Pichler, and Peter Hauff 10 X-Ray and X-Ray-CT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125 Willi A. Kalender, Paul Deak, Klaus Engelke, and Marek Karolczak 11 CT Contrast Agents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141 Hubertus Pietsch

vii

viii

12 Small Animal Magnetic Resonance Imaging: Basic Principles, Instrumentation and Practical Issue . . . . . . . . . . . . . 151 Peter Jakob 13 MR Contrast Agents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165 Eliana Gianolio, Alessandra Viale, Daniela Delli Castelli, and Silvio Aime 14 MR Spectroscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195 Markus Becker 15 Ultrasound . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207 Stuart Foster and Catherine Theodoropoulos 16 Ultrasound Contrast Agents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219 Michel Schneider 17 PET and SPECT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231 Sibylle I. Ziegler 18 Radiotracer I: Standard Tracer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 237 Walter Mier and Uwe Haberkorn 19 Radiotracer II: Peptide-Based Radiopharmaceuticals . . . . . . . . . . . . . 247 Roland Haubner and Clemens Decristoforo 20 Optical Imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 267 Ralf B. Schulz and Vasilis Ntziachristos 21 Optical Probes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 281 Kai Licha 22 Multi-modal Imaging and Image Fusion . . . . . . . . . . . . . . . . . . . . . . . . 293 Vesna Sossi Part IV Ex Vivo Validation Methods 23 Ex Vivo and In Vitro Cross Calibration Methods . . . . . . . . . . . . . . . . . 317 Albertine Dubois, Julien Dauguet, and Thierry Delzescaux 24 In Vitro Methods for In Vivo Quantitation of PET and SPECT Imaging Probes: Autoradiography and Gamma Counting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 347 David Stout Part V Data Postprocessing 25 Qualitative and Quantitative Data Analysis . . . . . . . . . . . . . . . . . . . . . . 363 Felix Gremse and Volkmar Schulz 26 Guidelines for Nuclear Image Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . 379 Martin S. Judenhofer, Stefan Wiehr, Damaris Kukuk, Kristina Fischer, and Bernd J. Pichler

Contents

Contents

ix

27 Kinetic Modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 387 Jörg van den Hoff 28 Data Documentation Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 405 Sönke H. Bartling and Wolfhard Semmler Part VI Special Applications 29 Imaging in Developmental Biology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 417 Katrien Vandoorne, Stav Sapoznik, Tal Raz, Inbal Biton, and Michal Neeman 30 Imaging in Gynecology Research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 437 Matthias W. Laschke and Michael D. Menger 31 Imaging in Cardiovascular Research . . . . . . . . . . . . . . . . . . . . . . . . . . . 449 Michael Schäfers, Klaus Tiemann, Michael Kuhlmann, Lars Stegger, Klaus Schäfers, and Sven Hermann 32 Imaging in Neurology Research I: Neurooncology . . . . . . . . . . . . . . . . 473 Yannic Waerzeggers, Parisa Monfared, Alexandra Winkeler, Thomas Viel, and Andreas H. Jacobs 33 Imaging in Neurology Research II: PET Imaging of CNS Disorders . 499 Gjermund Henriksen and Alexander Drzezga 34 Imaging in Neurology Research III: Focus on Neurotransmitter Imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . 515 Wynne K. Schiffer 35 Imaging in Oncology Research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 543 Wolfgang A. Weber and Fabian Kiessling 36 Imaging in Immunology Research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 565 Jason T. Lee, Evan D. Nair-Gill, Brian A. Rabinovich, Caius G. Radu, and Owen N. Witte Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 585 Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 589

List of Abbreviations

µCT Micro-computed tomography µMRI Micro-magnetic resonance imaging AAALAC Association for Assessment and Accreditation of Laboratory Animal Care AC Air conditioning ACH Air changes per hour ADA American Disabilities Act ADME Adsorption, distribution, metabolization and excretion ALARA As low as reasonably achievable ALS Amyotrophic lateral sclerosis AMPH Amphetamine AMPT Alpha-methyl-para-tyrosine AOI Area of interest APCs Antigen presenting cells APD Avalanche photodiode APP Amyloid precursor protein ASL Arterial spin labeling BAC Bacterial artificial chromosome BAL Bronchoavealoar lavage BBB Blood–brain barrier BCNU 1,,3-bis(2-chloroethyl)-1-nitrosourea BET Big endothelin BGO Bismuth germinate BLI Bioluminescence imaging BMD Bone mineral density BMS Bulk magnetic susceptibility BOLD Blood oxygenation level dependent BP Binding potential BPNP Bismuth sulfide polymer coating BSL Bio safety level CA Contrast agents CAR Coxsackie and adenovirus receptor

xi

xii

CCAC Canadian council of animal care CCD Charge coupled device CEA Carcinoembryonicantigen CEST Chemical exchange saturation transfer CG Chrysamine G CM Contrast media CMC Carboxymethylcellulose CMOS Complementary metal oxide semiconductor CNR Contrast-to-noise-ratio CNS Central nervous system COMT Catechol-O-methyltransferase COPD Chronic obstructive pulmonary disease COV Coefficient of variation CR Congo red CRF Corticotropin releasing factor CSI Chemical shift imaging CSS Cage changing station CT Computed tomography CTDI CT dose index CTLS Cytotoxic T lymphocytes DCE Dynamic contrast enhanced DCIS Ductal carcinoma in situ DCT Discrete cosine transform DECT Dual energy CT DFO Desferoxamine DICOM Digital imaging and communications DILI Drug induced liver injury DLB Dementia with lewy bodies DLNs Draining lymph nodes DMEM Dulbecco’s modified Eagle medium DNP Dynamic nuclear polarization DOI Depth of interaction DOT Diffuse optical tomography DOTA 1,4,7,10-tetraazacyclododecane-N,N¢,N, N¢-tetraacetic acid DOTATOC DOTA-Tyr3-octreotide DSA Digital subtraction angiography DTI Diffusion tensor imaging DTPA Diethylenetriamine-tetraacetic acid DV Distribution volume DVR Distribution volume ratio DWI Diffusion-weighted imaging

List of Abbreviations

List of Abbreviations

xiii

DXA Dual-energy X-ray absorptiometry EAE Experimental autoimmune encephalomyelitis EB-CCD Electron-bombarded CCD EBV Epstein–Barr virus ECF Extracellular fluid ECG Electrocardiography EDDA Ethylendiamin-N,N`diacetic acid EGF Epidermal growth factor EGFR Epidermal growth factor receptor EGR Epidermal growth factor EHS Environmental health and safety EM Electromagnetic EMCL Extramyocellular lipids EPSI Echo planar spectroscopic imaging ER Eendoplasmic reticulum ETS European convention for the protection of vertebrate animals used for experimental and other scientific purpose FACS Fluorescence-activated cell sorting FAZA Fluoroazomycin arabinoside FBP Filtered back projection FDG Fluorodeoxyglucose FDHT 16beta-18F-fluoro-5alpha-dihydrotestosterone fDOT Fluorescence DOT FES Fluorine-18 fluoroestradiol FFD Free form deformation FID Free induction decay FITC Fluorescein-isothiocyanate FLT 18F-3¢-deoxy-3¢-fluorothymidine FMISO Fluorine-18 labeled fluoromisonidazole fMRI Functional MRI FMT Fluorescence-mediated molecular tomography FMT l-[3-18F]fluoro-a-methyl tyrosine FOV Field-of-view FSH Follicle stimulating hormone FTLD Frontotemporal lobar degeneration FWHM Full-width at half-maximum Gd-DTPA Gadolinium-diethylenetriaminepentaacetate GEM Genetically engineered mouse GFP Green fluorescent protein GIST Gastrointestinal stromal tumor GLUT-1 Glucose transporter 1

xiv

GSO GV-SOLAS H&E HFC HK-II hNET hNIS HPGe HPLC HRE HSA HSP90 HSV-tk HTOS HU HYNIC IACUC IAPs IATA ICAM-1 ICCD ICG ICMIC ILAR IMCL IMT ISIS IT IVC LGSO LH LN LO LPS MAC MAG3 MAP MAR MB MCAo MCP MEMRI

List of Abbreviations

Germanate oxyorthosilicate Society of laboratory animals Hematoxylin and eosin High fat/high cholesterol Hexokinase II Human norepinephrine transporter Human sodium iodide symporter High purity germanium High-pressure liquid chromatography Hypoxia-response element Human serum albumin Heat shock protein 90 Herpes simplex thymidine kinase High throughput organic synthesis Hounsfield unit 2-hydrazinonicotinic acid Institutional animal care and use committee Inhibitors of apoptosis proteins International air transport association guidelines Intercellular adhesion molecule-1 Intensified CCD Indocyanine Green In Vivo Cellular and Molecular Imaging Center Institute for Laboratory Animal Research intramyocellular lipids l-[3-123I]iodo-a-methyl tyrosine Image-selected in vivo spectroscopy Information technology Individually ventilated cages Lutetium germanate oxyorthosilicate Luteinizing hormone Lymph node Lead optimization Lipopolysaccharide Minimum alveolar concentration Mercaptoacetyltriglycine Maximum a posteriori Micro-autoradiography Microbubble Middle cerebral artery occlusion Microchannel plate Manganese-enhanced MRI

List of Abbreviations

xv

METH Methamphetamine MHC Major histocompatibility complex MI Mechanical Index MLEM Maximum likelihood expectation maximization MMP Matrix metallo proteinase MRI Magnetic resonance imaging MRM MR microscopy mRNA Messenger ribonucleic acid MRS Magnetic resonance spectroscopy MRSI Magnetic resonance spectroscopic imaging MSA Multiple system atrophy MSCT Multislice/dual source computed tomography MTP Medial tibial plateau MTT Mean transit time NAS Network attached storage NCI National Cancer Institute NDT Nondestructive testing NET Norepinephrine transporter NFAT Nuclear factor of activated T cells NHL Non-Hodgkin’s lymphoma NIBIB National Institute for Biomedical Imaging and Bioengineering NIH National Institutes of Health NIR Near-infrared NIS Sodium iodide symporter NK Natural killer NMRS Nuclear magnetic resonance spectroscopy NMV Net magnetization vector NOTA 1,4,7- triazacyclononane-1,4,7-triacetic acid NPC Neural progenitor cell NSC Neural stem cell NSP Nucleoside salvage pathway OAT Optoacoustic tomography OCT Optical coherence tomography OI Optical imaging OPO Optical parametric oscillator OPT Optical projection tomography OSEM Ordered subsets expectation maximization OT Optical tomography OVA Ovalbumin PAT Photoacoustic tomography PBBR Pheripheral benzodiazepine receptor

xvi

PBS PD PDGF PDH PEG PEPE PET PFC Pgp PHIP PKC PLC PMT PnAO PPE PRESS PS PSCA PSF PSMA PSPMTs PTX PVE QDs QM QWBA RAID rBV RC rCBF REM RES RF RFP RGD ROI RSO RTK SA SAA SAI SAIRP

List of Abbreviations

Phosphate buffered saline Proton density Platelet-derived growth factor Pyruvate dehydrogenase Polyethylene glycol Perfluoropolyether nanoparticle Positron emission tomography Perfluorocarbon P-glycoprotein Para-hydrogen induced polarization Protein kinase C Phospholipase C Photomultiplier tube Propylenediamine-dioxime Porcine pancreatic elastase Point resolved spectroscopy Phosphatidylserine Prostate stem cell antigen Point spread function Prostate specific membrane antigen Position sensitive PMTs Pertussis toxin Partial volume effect Quantum dots Quantum mechanics Quantitative whole-body autoradiography Redundant drive array Relative blood volume Recovery coefficient Regional cerebral blood flow Resource equation method Reticuloendothelial system Radiowave Frequency Red fluorescent protein Arginine-glycine-aspartate Region-of-interest Radiation safety offices Receptor tyrosine kinase Serum albumin Sugar amino acids Small animal imaging Small-animal imaging research program

List of Abbreviations

xvii

SAR Structure activity relationship SI Spectroscopic imaging siPM Silicon photomultiplier SNR Signal-to-noise ratio SOP Standard operating procedures SPAQ Sensitive particle acoustic quantification SPECT Single photon emission computed tomography SPIO Superparamagnetic iron oxide SPM Statistical parametric mapping SR Synchrotron radiation SR Shift reagent SRTM Simplified reference tissue model SSD Sum of the squared intensity differences ST Saturation transfer STEAM Stimulated echo acquisition mode SUV Standardized uptake value TAC Thoracal Aortic Constriction TAC Time activity curve TCA Tricarboxylic acid cycle TCEP Triscarboxyethylphosphine TCR T-cell receptor TDI Tissue-Doppler imaging TE Echo time TETA 1,4,8,11-tetraazacyclotetradecane-1,4,8,11-tetraacetic acid 3D Three-dimension(al) TK1 Thymidine kinase-1 TKIs TK inhibitors TLD Thermoluminescent dosimeter TRAIL Tumour necrosis factor-related apoptosis-inducing ligand TSTU Tetramethyluronium tetrafluoroborate 2D Two-dimension(al) UBM Ultrasound biomicroscopy UCA Ultrasound contrast agents US Ultrasound USPIO Ultrasmall particles of iron oxide VAP Vascular access port VCAM-1 Vascular cell adhesion molecule-1 VCT Volumetric CT VEGF Vascular endothelial growth factor VEGF-R2 Vascular endothelial growth factor receptor-2 WAG Waste anesthetic gas

Part Role of Small Animal Imaging

I

1

Noninvasive Imaging for Supporting Basic Research Pat Zanzonico

1.1 Introduction

Contents 1.1 Introduction ...........................................................

3

1.2 State-of-the-Art Technologies................................ 1.2.1 Imaging Modalities . ................................................ 1.2.2 Imaging Probes/Contrast Agents ............................. 1.2.3 Multimodality Imaging . ..........................................

5 5 8 8

1.3 Outlook ................................................................... 11 1.4 Laboratory Protocols: An Illustrative Example......................................... 11 References............................................................................ 15

P. Zanzonico Memorial Sloan-Kettering Cancer Center, 1275 York Avenue, New York, NY, USA e-mail: [email protected]

Imaging has long been indispensable in clinical practice. Increasingly, in vivo imaging of small laboratory animals (i.e., mice and rats) has emerged as a critical component of preclinical biomedical research as well (Beckman et al. 2007; Cherry 2006; Gross and PiwnicaWorms 2006; Pomper 2005). Small-animal imaging provides a noninvasive means of assaying biological structure and function in vivo, yielding quantitative, spatially and temporally indexed information on normal and diseased tissues such as tumors. Importantly, because of its noninvasive nature, imaging allows serial (i.e., longitudinal) assay of rodent models of human cancer and cardiovascular, neurological, and other diseases over the entire natural history of the disease process, from inception to progression, and monitoring of the effectiveness of treatment or other interventions (with each animal serving as its own control and thereby reducing biological variability). This also serves to minimize the number of experimental animals required for a particular study. With the ongoing development of genetically engineered (i.e., transgenic and knock out) rodent models of cancer and other critical diseases (Malakoff 2000), such models are increasingly more realistic in recapitulating the natural history and clinical sequelae of the corresponding human disease and the ability to track these disease models long-term is therefore invaluable. Importantly, in contrast to cell or tissue culture-based experiments, studies in intact animals incorporate all of the interacting physiological factors – neuronal, hormonal, nutritional, immunological, etc. – present in the complex in vivo milieu. Intact whole-animal models also facilitate investigation of systemic aspects of disease such as cancer metastasis, which are difficult or impossible to replicate in ex vivo systems. Further, because many of the same imaging

F. Kiessling and B.J. Pichler (eds.), Small Animal Imaging, DOI: 10.1007/978-3-642-12945-2_1, © Springer-Verlag Berlin Heidelberg 2011

3

4

P. Zanzonico

modalities – magnetic resonance imaging and spectroscopic imaging (MRI and MRSI), positron emission tomography (PET), single-photon emission computed tomography (SPECT), transmission computed tomography (CT), and ultrasound (US) – used in the clinic are also used in the laboratory setting, the findings of smallanimal imaging are readily translatable to patients. Prior to the inception of “small-animal imaging,” experimental animals were generally imaged using clinical instrumentation (during off-hours, of course), and many useful studies have been performed in this way. In many instances, however, the performance (most notably, the spatial resolution) of clinical imaging devices is inadequate (i.e., the spatial resolution is prohibitively coarse) for scientifically useful imaging of tumors and organs in mice and rats, as illustrated in Fig. 1.1. In addition to the need for better spatial resolution, the development of dedicated small-animal imaging instruments, a

b

and of centralized facilities to house these instruments, has been motivated by a number of practical considerations. First, by incorporating invasive and therefore clinically impractical corroborative assays (e.g., interstitial probe measurements, histology, immunohistochemistry, etc.) into small-animal imaging studies, new and/ or existing clinical imaging paradigms can be clarified, validated, and/or improved in the laboratory and then translated back to the clinic. Second, biosecurity (i.e., protection from transmission of infectious and other diseases among experimental animals and between animals and humans) of immunodeficient and other genetically engineered animal models requires that such animals remain within a “clean” barrier facility and are not, for example, transported out of such a facility to a clinical imaging area and then back to the facility. Third, in certain institutions and jurisdictions, experimental animals are prohibited by regulation from entry into clinical c

Tumor allograft

=

viable rim + Necrotic core

Fig. 1.1 Comparative PET images of tumor-bearing mice acquired with a clinical and a small-animal PET scanner. (a) A photograph (not to scale) showing the orientation of the animals in the PET images. (b) A coronal PET image of a mouse with a Lewis Y antigen-expressing HCT15 human colorectal carcinoma xenograft in each of its two hindlimbs. The image was acquired at ~2 days postinjection of an yttrium-86 (86Y)-labeled humanized anti-Lewis Y antibody, hu3S193, using a clinical PET scanner, the GE Advance™ (General Electric Medical Systems, Pewaukee, WI), with a full-width half-maximum (FWHM) spatial resolution of 6 mm and volume resolution of 216 mm3. (c) A coronal PET image with an FSA II murine fibrosarcoma allograft in the right hindlimb. The images were acquired at ~1 h postinjection of fluorine-18 (18F)-labeled fluorodeoxyglucose (FDG) using a dedicated rodent PET scanner, the R4 microPET™ (Concorde Microsystems, Knoxville, TN),

with a FWHM spatial resolution of 2.2 mm and volume resolution of 10.6 mm3. All three tumors were comparable in size (1–1.5 cm in the largest dimension). Although the image acquired on the clinical scanner (b) clearly demonstrates highcontrast uptake of the radiotracer by the two tumors, it does not show any heterogeneity of uptake within the tumors. If any such heterogeneity is present, any parameter derived from the measured uptake will reflect some ill-defined value of any such parameter averaged over the entire tumor. In contrast, the image acquired on the small-animal scanner (c) distinguishes the differential uptake of FDG between biologically distinct cell subpopulations within the tumor, namely, high uptake in a viable rim and much lower uptake in a largely necrotic core. For any parameters derived from the tracer uptakes in (c), therefore, distinct, and more meaningful, parameter values can be derived for the viable rim and for the necrotic core.

5

1 Noninvasive Imaging for Supporting Basic Research

areas. Fourth, the limited and, at times, unpredictable availability (i.e., at night, overnight, and/or on weekends and holidays) of clinical imaging instrumentation makes it very difficult to plan and perform experiments, especially experiments involving large numbers of animals and/or multiple imaging sessions. Small-animal imaging is the preclinical component of the burgeoning field of “molecular imaging” (i.e., noninvasive visualization of normal as well as abnormal processes at a molecular genetic or cellular level) (Pomper 2001). This new field of investigation has expanded rapidly following designation of “cancer imaging” as one of six “extraordinary scientific opportunities” by the National Cancer Institute (NCI) in the United States in the late 1990s. Major funding allocations to the research community followed, including the NCI Small-Animal Imaging Research Program (SAIRP) and In Vivo Cellular and Molecular Imaging Centers (ICMIC) program. Additionally, a new US National Institutes of Health (NIH) institute, the National Institute for Biomedical Imaging and Bioengineering (NIBIB), was formed. The clinically translatable, noninvasive, and quantitative nature of small-animal imaging makes it an invaluable component of modern biomedical research. And the availability of dedicated small-animal imaging devices and facilities has resulted in wider and more scientifically useful application of imaging in preclinical experimentation.

1.2 State-of-the-Art Technologies 1.2.1 Imaging Modalities “Reverse translation” of clinical imaging modalities to small-animal research requires substantial improvement in performance (most notably, in spatial resolution)

of the respective modalities and has involved reengineering of many aspects of their imaging hardware, firmware, and software (see Table 1.1) (Beckman et al. 2007; Cherry 2006; Gross and Piwnica-Worms 2006; Pomper 2005; Rowland and Cherry 2008). In addition, radiations in the optical and nearinfrared (NIR) regions of the electromagnetic (EM) spectrum, which are highly attenuated by tissue and therefore generally not useful for in vivo imaging of subjects as large as humans,1 have been successfully adapted to imaging of small animals (Contag et al. 1998; Ntziachristos et al. 2005). Specialized technologies have led to widespread and very productive use of such radiations – both bioluminescent and fluorescent – in imaging studies of rodents (Fig. 1.2). Because light emitted at any depth of tissue is scattered and otherwise dispersed as it passes through overlying tissue before emanating from the surface of the animal, the apparent size of the light source is considerably larger than its actual size (Fig. 1.2c). Despite the excellent spatial resolution of the CCDs themselves, the effective resolution of optical and NIR imaging is, therefore, rather coarse. Further, for planar optical and NIR imaging, the resulting images are only semiquantitative: attenuation, scatter, and dispersion of the emitted light as it passes through overlying tissue make the measured signal highly depth-dependent. Thus, a focus of cells lying deep within tissue may appear less luminescent or fluorescent than an identical focus of cells at a more shallow depth; if excessively deep, such a focus of cells may be undetectable altogether. NIR radiations, however, have a substantially higher penetrance through tissue than blue to green radiations. Importantly, therefore, Fluorescence imaging is being pursued clinically in intraoperative, laparoscopic, and endoscopic settings, where there is no over lying tissue.

1

Table 1.1 Comparative spatial resolution of clinical and preclinical imaging modalities and associated design refinements Modality Spatial resolution (mm) Clinical-to-preclinical design refinement(s) Clinical Preclinical MRI

~1 mm

£100 mm

Higher field-strength magnets, improved gradient fields and coils

MRSI

~1 cm

~2 mm

Higher field-strength magnets, improved gradient fields and coils

PET

~5 mm

1–2 mm

Reduced detector-element size, smaller-diameter detector rings

SPECT

~1 cm

0.5–2 mm

Pinhole collimation (and resulting magnification)

CT

1–2 mm

£200 mm

Higher X-ray flux, smaller focal spot, and higher magnification

US

1–2 mm

£100 mm

Higher-frequency scan heads

6

a

P. Zanzonico

Bioluminescence

b

Fluorescence

c

Fig. 1.2 Bioluminescent and fluorescent optical or NIR in vivo imaging. In both bioluminescent and fluorescent imaging, the animals are placed in a light-tight imaging enclosure and the emitted optical or NIR signal is imaged by a charge-coupled detector (CCD). (a) In bioluminescent imaging, cells (e.g., tumor cells) which are to be localized or tracked in vivo must first be genetically transduced ex vivo to express a so-called “reporter gene,” i.e., the luciferase gene. After the cells have been implanted, infused, or otherwise administered to the experimental animal, the luciferase substrate (e.g., d-luciferin in the case of firefly luciferase) is systemically administered. Wherever the administered substrate encounters the luciferaseexpressing cells, the ensuing d-luciferin-luciferase reaction emits light, which is detected and localized by the imaging system. The CCD in bioluminescent imaging is maintained at a very low temperature (−100°C), thereby ensuring that any electronic output it produces results from light striking the CCD rather than the background “dark current.” In this way, the otherwise undetectably small signal originating in vivo and escaping from the surface of the animal can produce an image. (b) In fluorescence imaging, cells to be imaged may either be

genetically transduced ex vivo to express a fluorescent molecule such as green fluorescent protein (GFP) or a fluorescent molecular probe targeting the cells of interest may be systemically administered. In either case, the animal is then illuminated with light at an appropriate wavelength (obtained by filtration) to excite the fluorophore in situ and the resulting emitted light (which has a slightly different wavelength than the excitation light) is itself filtered and detected by the CCD. The excitation light may be provided by reflectance (or epi-) illumination or by transillumination of the animal. Further, by computer processing, the abundant spontaneous fluorescence of the animal’s tissues as well as of foodstuffs in the gut must be mathematically separated, or “de-convolved,” from the overall fluorescence to yield an image specifically of the fluorophore. (c) In practice, the luminescence or fluorescence image is superimposed on a conventional (i.e., white-light) photograph of the animal as positioned for imaging to provide some orientation as to the anatomic location of the signal(s) in vivo. The sample image shows a pseudocolor bioluminescence image superimposed on a gray-scale photograph of a mouse.

by using laser transillumination for excitation of administered NIR molecular probes in situ, tomographic fluorescence images can be mathematically reconstructed (Ntziachristos et al. 2005). The resulting three-dimensional images – in contrast to planar images – are quantitative: the signal intensity thus reconstructed is directly related to the local concentration of the fluorophore.

Salient features of the various small-animal imaging modalities are summarized in Table 1.2. For each modality, current commercially available instruments provide largely turn-key operation. However, MRI/ MRSI may require custom-fabricated coils for specialized applications as well as shimming and other tuning adjustments prior to imaging. And, US is rather userdependent, relying on the manual skill of the operator.

Up to 5 1

5–60 min

30–90 min

10–15 min

Up to ~60 min with set-up

~5 min

~5 min

~10 min

PET

SPECT

CT

US

Optical: bioluminescence

Optical: fluorescence

NIR: fluorescence

Low; high between cystic and solid structures

£100 mm

<5 mm

<5 mm None

Variable

None

High among soft tissue, lung, and bone; none among soft tissues

£200 mm

~1 cm

None

None

Variablec

High

0.5–2 mm

1–2 mm

~2 mm

£100 mm

pM

nM

nM

?

mM

Sub pM

No

No

Yes

Yes

No

No

Yes

No

mM-mMc Sub pM

Yes

Dynamic imaging

mM-mM

Probe or contrastagent sensitivityb

0

0

0

0

10–20

10–100

10–100

0

0

Radiation dose (cGy)

$200–300K

$100–200K

$200–400K

$200K

$200–400K

$600–800K

$600–800K

~$1M

Equipment cost

a

Image contrast among tissues in the absence of any administered probe or contrast agent b Imageable concentration in vivo of administered probe or contrast agent or, in the case of MRSI, endogeneous metabolites c Natural-abundance nuclei such as hydrogen-1 and phosphorus-31 in endogeneous metabolites present at concentrations in the mM-mM range provide intrinsic contrast

1

1

1

1

1 or 2

1

Up to ~60 min with set-up

MRSI

Up to ~10

Up to ~60 min with set-up

MRI

Table 1.2 Performance parameters and logistical features of small-animal imaging modalities Spatial Intrinsic Modality Time per Number resolution contrasta study of animals per study

1 Noninvasive Imaging for Supporting Basic Research 7

8

Although US is the only real-time imaging modality (and therefore can be used for image-guided injections, for example), set-up and acquisition per image can take up to 5 min or longer. The radiation doses to experimental animals in PET, SPECT, and CT studies are considerably – one to two orders of magnitude – higher than those encountered in the corresponding clinical studies (Cherry 2006). Indeed, at absorbed doses of the order of 100 cGy, they approach singlefraction doses used in fractionated external-beam radiation therapy. Investigators should be aware of the magnitude of absorbed doses encountered in smallanimal PET, SPECT, and CT studies and potential radiogenic perturbation of their experimental system. MRI, CT, and US, because of their high-resolution visualization of anatomy, are often characterized as “structural” imaging modalities. MRSI, PET, SPECT, and optical imaging, on the other hand, are generally characterized as “functional” (or “molecular”) imaging modalities. Such a distinction, however, is increasingly arbitrary and inaccurate, as dynamic and/or static MR, CT, and US imaging may be performed following administration of a blood-flow or molecularly targeted contrast agent and functional images derived.

P. Zanzonico

small-animal experimentation (Pomper 2005; Serganova et al. 2008). These include: (1) “biomarker” or “surrogate” imaging – related to some physiological process (e.g., blood flow) or to some downstream effects of one or more endogenous molecular/genetic processes (e.g., FDG PET imaging reflecting upregulation of glucose transporters and/or glycolytic metabolic pathways in many tumors); (2) “direct” imaging of specific molecules – based on binding of radiolabeled ligands (e.g., imaging of the a5b3 integrin, commonly overexpressed in tumor vasculature, with radiolabeled glycosylated RGD (arginine–glycine–aspartate)-containing peptides); and (3) “indirect” reporter gene imaging. The reporter gene imaging paradigm (Fig. 1.3), representing a convergence of molecular and cell biology and the imaging sciences, is providing new insights into signal transduction pathways, oncogenesis in genetic mouse models, endogenous molecular genetic/biological processes, and response to therapy in animal models of human disease. A general summary of modality-specific imaging probes/contrast agents is presented in Table 1.3.

1.2.3 Multimodality Imaging 1.2.2 Imaging Probes/Contrast Agents Several different classes of imaging probes/contrast agents and imaging strategies are currently employed in

Since information derived from multiple modalities is often complementary (e.g., localizing the site of an apparently abnormal metabolic process to a pathologic structure such as a tumor), integration of this

Fig. 1.3 Design of a reporter gene construct and the indirect reporter imaging paradigm. (a) (1) The basic structure of a reporter gene complex is shown, expressing either herpes simplex virus 1 thymidine kinase (HSV1-tk) and/or luciferase. (Other reporter genes include the human norepinephrine transporter (hNET) and the human sodium iodide symporter (hNIS).) Control and regulation of gene expression is accomplished through promoter and enhancer regions that are located at the 5¢ end (“up-stream”) of the reporter gene. These promoter/enhancer elements can be “constitutive” and result in continuous gene expression (“always on”) or “inducible” (“conditionally on”) and sensitive to activation by endogenous transcription factors and promoters. Following the initiation of transcription and translation, the gene product accumulates. (2) In this case the reporter gene product is the enzyme HSV1-tk, which phosphorylates selected radiolabeled thymidine analogs; these probes are not phosphorylated by endogenous mammalian tk. The phosphorylated probe does not cross the cell membrane readily and is effectively “trapped” and accumulates selectively within transduced cells. Probe (activity) accumulation thus reflects the

level of HSV1-tk enzyme activity and of HSV1-tk reporter gene expression. (3) In this case, luciferase is the reporter gene product and expression is detected via its catalytic action on the administered d-luciferin substrate resulting in production of bioluminescence. (b) Different reporter gene constructs are transfected into target cells by a vector (e.g., a virus). The reporter gene may or may not be integrated into the host-cell genome; transcription of the reporter gene to (messenger ribonucleic acid) mRNA is initiated by “constitutive” or “inducible” promoters, and translation of the mRNA to a protein occurs on the ribosomes. The reporter gene product can be a cytoplasmic or nuclear enzyme, a transporter in the cell membrane, a receptor at the cell surface or part of cytoplasmic or nuclear complex, an artificial cell surface antigen, or a fluorescent protein. Often, a complimentary reporter probe (e.g., a radiolabeled, magnetic, or bioluminescent molecule) is administered and the probe signal is directly related to the level of reporter gene product, thus reflecting levels of transcription, modulation and regulation of translation, protein–protein interactions, and/or posttranslational regulation of protein conformation and degradation

9

1 Noninvasive Imaging for Supporting Basic Research mRNA transcription

Regulatory Elements

a

mRNA translation

5¢ end

Reporter Protein (enzyme)

3¢ end

TATA (1)

DNA Reporter Gene HSVI-IK/ Luciferase

Poly-A transiation

+ PET

cGy dose/G 2.4 Radiolabeled Probe “Trapped” Probe 0.0

+

(2) Bioluminescence 14¢/10²

Photons

12 Luciferin Probe

8 6 4

Proton corts

10

O2 + ATP

2 (3)

b

Vector Reporter Gene

Antigen Receptors Fluorescent protein

Radiolabeled, magnetic,or flurescent probe

Nucleus

Gene Product mRNA

Transporter Enzyme

Regulation of reporter gene expression: Constitutive promoter (”always on”) or inducible promoter (”conditionally on”)

10

P. Zanzonico

Table 1.3 Probes and contrast agents for small-animal imaging modalities Modality Type of probe or contrast agent MRIa

Tl relaxation agents such as gadolinium-diethylenetriaminepentaacetate (Gd-DTPA) T2 relaxation agents such as superparamagnetic iron oxide (SPIO) particles (ferumoxides)

MRSIa

Agents (e.g., a drug) containing a natural-abundance magnetic isotope (e.g., fluorine-19) of a nonphysiologic element (e.g., fluorine) or an enriched magnetic isotope (e.g., carbon-13) of a physiologic element (e.g., carbon)

PET

Radiotracers labeled with a positron-emitting radionuclide

SPECT

Radiotracers labeled with a single-photon (i.e., X- and g-ray)-emitting radionuclide

CT

Molecules or particles containing a high atomic-number element (e.g., iodine)b

USa

Echogenic agents such as microbubbles (Lindner 2004)

Optical: bioluminescence

Substrates such as d-luciferin for enzyme-catalyzed bioluminescent (e.g., luciferase-luciferin) reactions

Optical: fluorescencea

Optical fluorophores such as quantum dots, nanoparticles, fluoroproteins, etcc

NIR: fluorescence

NIR fluorophores such as quantum dots, nanoparticles, fluoroproteins, etcc

a

MRI, MRSI, CT, US, and optical fluorescence imaging are often performed without administering any probe or contrast agent. For MRI, MRSI, CT, and US, contrast results from intrinsic differences among tissues. For fluorescence imaging, contrast results from administered cells genetically transduced ex vivo to express a reporter fluorophore such as green fluorescent protein (GFP) b Dedicated small-animal CT scanners are commonly not spiral scanners, and therefore, require ~10 min to acquire a study. As a result, water-soluble radiographic contrast agents, which are rapidly cleared from blood through the kidneys, are not suitable for small-animal devices – nearly all of such contrast agents are cleared well before a 10-min scan is completed. Specialized contrast agents which have long residence times in blood, such as polyethylene glycol-coated (i.e., PEGylated) chylomicron remnants impregnated with iodine, must be used instead (Wisner et al. 2002) c Fluorescent probes may be characterized as “active” or “activatable.” Active probes yield a fluorescent signal whenever illuminated by the light of an appropriate excitation wavelength; in other words, they are always “on.” “Activatable” probes contain multiple fluorophores covalently linked to a molecular backbone and the intact probe is thus self-quenching. Only in the presence of enzymes that specifically cleave the fluorophore-backbone covalent bonds are the fluorophores dispersed, self-quenching eliminated, and a signal obtained. Depending on the particular covalent bond, fluorescence imaging with activatable probes thus provides a means of detecting and assaying the activities of specific enzymes in situ (Ntziachristos et al. 2002) a

information may be helpful and at times critical. In addition to anatomic localization of signal foci, image registration and fusion provide intra as well as intermodality corroboration of diverse images and more accurate and reliable diagnostic and treatmentmonitoring information. The problem, however, is that differences in image size and dynamic range, voxel dimensions and depth, image orientation, subject position and posture, and information quality and quantity make it difficult to unambiguously colocate areas of interest in multiple image sets. The objective of image registration and fusion, therefore, is (a) to appropriately modify the format, size, position, and even shape of one or both image sets to provide a point-to-point correspondence between images and (b) to provide a practical integrated display of the images thus aligned (Zanzonico 2008). This process entails spatial registration of the respective images in a common coordinate system based on optimization

of some “goodness-of-alignment,” or “similarity,” criterion (or metric). There are two practical approaches to image registration and fusion, software and hardware approaches. In the software approach, images are acquired on separate devices, imported into a common image-processing computer platform, and registered and fused using the appropriate software. In the hardware approach, images are acquired on a single, multimodality device (such as a PET–CT or SPECT–CT scanner) and transparently registered and fused with the manufacturer’s integrated software; a number of such multimodality devices, clinical as well as preclinical, are commercially available. Among multimodality imaging studies, PET–MRI (or SPECT–MRI) may ultimately provide the greatest yield of scientific information by combining the quantitative molecular imaging capabilities of PET (and the large number and variety of current PET radiotracers) with the excellent anatomic resolution,

11

1 Noninvasive Imaging for Supporting Basic Research

marked soft-tissue contrast, and functional imaging capabilities (e.g., perfusion by dynamic contrastenhanced (DCE) imaging) provided by MRI (Pichler et al. 2008). A unique, potentially important advantage of PET–MRI over other multimodality imaging studies is that PET and MR data can be acquired simultaneously. In contrast, for PET–CT and SPECT–CT, the respective images are acquired sequentially. Implicit in the registration of such sequentially acquired images is the assumption that the subject is morphologically and functionally stable over the time interval between and during the image acquisitions. Otherwise, one may be attempting to register images of what are effectively different subjects. Intuitively, registration of functional and structural images (e.g., PET and CT images) or of different structural images (e.g., CT and MR images) is rather forgiving of the time interval between sequential image acquisitions, as anatomy generally changes quite slowly (typically over days to weeks). Importantly, however, this may not be the case in tumor xenograft models in rodents. Such xenografts may grow significantly over a matter of hours, with size-dependent alteration of their functional as well as structural properties; for example, as tumor mass increases, the intratumoral number and distribution of hypoxic cells may also vary. For multimodality imaging devices, where one image is acquired immediately following the other, the time interval between acquisitions is typically of the order of only several minutes, with no significant structural alterations expected. Registration of sequentially acquired functional images, however, is potentially more problematic, as functional properties, such as blood flow, hypoxia, etc., may change transiently over time frames of minutes and even seconds. Smallanimal MRI scanners with PET inserts have been developed and simultaneous MRI and PET studies successfully performed (Pichler et al. 2008).

1.3 Outlook Imaging-based experimentation in small-animal models is now an established and widely used approach in basic and translational biomedical research and will no doubt remain an important component of such research. With small-animal imaging resources already

widely available among academic medical centers, research institutes, and industrial laboratories, its growth is likely to slow, however, compared to that seen over the last decade. Nonetheless, several areas – drug development, treatment monitoring, and novel therapeutic strategies such as adoptive immunotherapy and gene therapy – are likely to be particularly productive in the application of small-animal imaging (Beckman et al. 2007; Gross and Piwnica-Worms 2006). In drug development, imaging-based assays are particularly amenable to quantitative characterization of pharmacokinetics and pharmacodynamics of new therapeutics and may accelerate the drug discovery process. Transgenic and knock-out mouse models of human disease may be used for identification and validation of “drug-able” molecular targets. Clinically translatable imaging paradigms developed and validated in animal models may also provide earlier and more clinically meaningful assays of therapeutic response, enabling clinicians to rapidly distinguish “responders” from “non-responders” and promptly switch patients from ineffective to potentially more effective therapies, thereby avoiding unnecessary toxicities, expense, and loss of time; this paradigm is illustrated in Fig. 1.4 (Smith-Jones et al. 2004). Clinically applicable gene imaging techniques will likely be essential if adoptive immunotherapies and gene therapies are to progress. Tumor targeting of immune effector cells and their progeny and delivery and expression of therapeutic genes can likely only be assessed by stable incorporation of a suitable reporter gene into the genomes of effector cells and of therapeutic viral vectors, respectively. Further, in order to maximize the yield of functional information, it is likely that multimodality imaging as well as dynamic imaging (with appropriate kinetic analyses) will become more commonplace.

1.4 Laboratory Protocols: An Illustrative Example Space does not allow for presentation of detailed protocols in the current chapter, and these are addressed for the various modalities in the subsequent chapters of this volume. Rather, a case study (Cho et al. 2009) is presented which illustrates several of the salient features of small-animal imaging experimentation, including

12

P. Zanzonico

17-AAG-Treated Baseline

Control

24 hr

Bladder

Baseline

24 hr

Coronal Sections

Tumor

Heart Western blots

Kidneys

Control

Transverse Sections

Tumor

17-AAG

HER2 p85 P13k

a

b

c

d

e

Fig. 1.4 Coronal (top row with head at bottom) and transverse (bottom row with dorsal surface at top) R4 microPET™ (Concorde Microsystems) images at 3 h postinjection of the gallium-68-labeled F(ab¢)2 fragment of Herceptin™ (68Ga-DOTA-F(ab¢)2 Herceptin) into athymic nude mice with BT474 breast tumor xenografts on the right flanks; DOTA is the metal chelator 1,4,7,10-tetraazacyclodecane-N,N¢,N²,N²¢tetraacetate. Herceptin™, used in the treatment of some breast cancers, is an antibody directed against the HER2n tyrosine kinase, which is overexpressed in BT474 and other breast tumors. HER2n is a client protein of the heat shock protein 90 (HSP90) chaperone protein. (a) One mouse underwent a baseline study followed by treatment with the geldanamycin derivative 17-AAG (an inhibitor of HSP90), followed 24 h later by (b) a second scan with 68Ga-DOTA-F(ab¢)2 Herceptin. The

tumor uptake of 68Ga-DOTA-F(ab¢)2 Herceptin decreased 50% between the pre and posttreatment scans. (c, d) The control (i.e., untreated) mouse underwent two scans 24 h apart, with no significant change in tumor uptake. (e) As corroborated by the Western blots shown, 17-AAG induced degradation of HSP90 and, in turn, HER2n. This study illustrates the application of imaging to characterization of the pharmacodynamics of molecularly targeted anticancer therapy. Hypothetically, for example, breast cancer patients can undergo pre and posttherapy scans to identify responders (i.e., having scan results analogous to those in (a, b)) and nonresponders (i.e., having scan results analogous to those in (c, d)) to HSP90 inhibitors. Responders would then be effectively treated with such inhibitors, while nonresponders would be switched to alternative treatment. (Adapted from Smith-Jones et al. 2004)

noninvasive quantitative characterization of structure and function in situ, multimodality imaging, and corroboration of imaging results with invasive assays (i.e., histology and immunohistochemistry). This example focuses on the development and validation of an imaging paradigm for detection and localization of tumor hypoxia and, specifically, on the relationship between DCE-MRI using Magnevist® (Gd-DTPA) and dynamic PET using 18F-fluoromisonidazole (FMiso). This multimodality study, performed using a Bruker 7T BioSpin™ imaging spectrometer (Bruker, Germany) and a Focus 120 microPET™ (Concorde Microsystems, Knoxville, TN USA), utilized stereotactic fiduciary markers in a heterotopic Dunning R3327-AT rat prostate tumor model. A notable component of this work was registration of not only the multiple in vivo images but also of the in vivo and ex vivo images. Specifically, in vivo imaging results were corroborated by

comparison to registered tumor sections stained for hematoxylin and eosin (H&E) to identify viable and necrotic tissue areas and pimonidazole to identify hypoxic tissue areas. As shown in Fig. 1.5, a Helmholtz MRI coil was constructed in two parts, with the upper part of the coil and a fiduciary-marker assembly initially detached to facilitate positioning of the animal bearing a tumor xenograft in its hindlimb. The fiduciary-marker assembly is shown in Fig. 1.5a attached to the coil-marker system. The marker assembly was composed of two cylindrical disks and one flat plate. The top disk, or marker holder (or “button”), had three holes for Gd (for MRI)- or 18F (for PET)-filled vertical markers. A bottom disk with identically aligned holes was separated from the marker holder disk by a flat plate. Separate release screws fastened each disk to the flat plate, which was fixed to the top of the MRI coil. At the end of the

13

1 Noninvasive Imaging for Supporting Basic Research

a

Top view

Side view

Top coil Marker holder disk

Three vertical markers Top horizontal marker

Flat plate “Button” Histology marker disk Top coil

Bottom horizontal marker Marker holder not shown

Bottom coil

Release screw for a histology marker disk

c MRI fiduciary marker

Release screw for marker holder disk

b

w/o Coil-Marker assembly Mold

Tumor

w/ Coil-Marker assembly Marker assembly

Top coil

c

d

PET fiduciary marker Coronal view

Fig. 1.5 (a) Magnetic coil-fiduciary marker assembly for threedimensional registration of MR, PET, and histological images. (b) Dorsal view of tumor-bearing caudal end of rat. (c) Registered (by rigid transform) MR and PET images of the three vertical

fiduciary markers. For MR imaging, the fiduciary markers are filled with an aqueous Gd solution, and for PET imaging, with an aqueous 18F solution. (Adapted from Cho et al. 2009)

imaging experiments, the bottom marker disk was detached. The flat plate had one side-hole for a horizontal marker. An additional holder was placed in the center of the bottom coil with a grooved disk for holding another horizontal marker. The three vertical markers (forming an oblique triangle) and the two horizontal markers were composed of 22-gauge catheter tubing which were heat-sealed once filled. The vertical markers allowed in-plane registration of the PET and MR images (see Fig. 1.5c) and the two horizontal markers depth-wise registration. The three vertical markers were pushed into the 0.9-mm holes through the two separate cylindrical disks. The separation of the top and bottom coils was adjustable over a range of 2–3 cm to accommodate tumors of various sizes. As illustrated in Fig. 1.5b, the magnetic coil-fiduciary-marker assembly was placed in the animal’s custom-fabricated whole-body mold, which was fabricated in a semicylindrical Lucite™ holder compatible with both the MR and PET scanner palettes. Close-up views of the top

horizontal-marker holder (button) and the three angiocatheters in place are shown in Fig. 1.6a and b, respectively; the procedure for registration of the in vivo (PET and MR) and ex vivo (histological) images is explained in Fig. 1.6c. As detailed in Fig. 1.7, the comparison of the registered DCE-MRI “kep” parametric maps (Kep is a parameter related to blood flow and vascular permeability and therefore considered a measure of tissue perfusion (Hoffmann et al. 1995)) and the H&E-stained sections and of the registered 18F-FMiso PET late-slope maps and pimonidazole-stained sections confirms the capability of combined MRI–PET measurements to functionally image the tumor microenvironment and distinguish regions of necrosis, hypoxia, and normoxia. This multimodality approach to functional tissue segmentation, validated by registration with histological images, should minimize the probability of false-positive findings of tumor hypoxia that are likely to result if one only were to perform static PET imaging of 18F-FMiso

14

P. Zanzonico

a

b Top horizontal marker holder

Button removed to show catheters

Button 10 mm

c

Angiocathters (needles removed) Template Button

Probe holes

Skin overlying tumor Tumor

y

Cryofixative

X

Microtome blade

Portion of tumor already microsected away

Z

Tumor section i on slide

10mm

Fig. 1.6 (a) Close-up view of the top horizontal-marker holder (or “button”) showing the three holes for the vertical catheters; the multiple smaller holes (not used in the study being described) are for various interstitial probes. (b) A postsacrifice photograph of the tumor in situ with the overlying skin and fiduciary-marker assembly removed but with the three plastic angiocatheters, used for registration of the histological images with the in vivo (i.e., PET and MR) images, in place. The tumor is removed with the horizontal-marker holder (button) and angiocatheters in place and the entire tumor-button-angiocatheter assembly is frozen in cryofixative for sectioning. (c) The procedure for the depth (or z-axis) registration of the in vivo and histological images. The

button, frozen held in place with the angiocatheters against the shallow surface of the excised tumor, is placed against the microtome chuck. As a result, the cryosections thus cut are parallel to the in vivo (PET and MR) coronal images. The first (i.e., the deepest) cryosection in which tumor is visible is matched with the first PET and MR coronal images in which the tumor is also visible, thus achieving depth-wise alignment of the cryosections and the PET and MR images. Histological sections and PET and MR images at the same depth are then aligned in the coronal (or xy-plane) by rigidly transforming the three angiocatheter holes visible in the section with the fiduciary markers visible in the PET/MR images. (Adapted from Cho et al. 2009)

15

1 Noninvasive Imaging for Supporting Basic Research

a

b

c

d

e

Fig. 1.7 (a) A T2-weighted coronal MR image of the tumor xenograft in the animal shown in Fig. 1.5b. (b) The registered H&E-stained section of the same tumor showing the channels created by the three angiocatheters and the necrotic and viable areas of the tumor-stained purple and blue, respectively (left panel). The necrotic area of the tumor outlined (middle panel). A mask overlying the necrotic area of the tumor as outlined in the middle panel (right panel). (c) The registered DCE-MRI-derived kep parametric map of the same tumor without (left panel) and with (right panel) the necrotic area of the tumor segmented out using the “necrosis mask” created in (b). (d) The registered

dynamic 18F-FMiso PET late-slope parametric map also without (left panel) and with (right panel) the necrotic area of the tumor segmented out. (e) The registered pimonidazole-stained section of the same tumor showing the hypoxic areas of the tumor-stained green. Note the spatial concordance of the hypoxic tumor areas as identified by the segmented 18F-FMiso parametric map and the pimonidazole-stained section, with the elimination of possible false-positive hypoxic areas corresponding to necrosis, perhaps due to FMiso diffusing from viable but hypoxic areas of tumor into nearby necrotic areas. (Adapted from Cho et al. 2009)

or other “hypoxia” radiotracers. More generally, this study illustrates the power of small-animal imaging in quantitatively characterizing in vivo biology in ways that are otherwise difficult or impossible.

Cherry S (2006) In-vivo whole-body imaging of the laboratory mouse. In: Fox J, Barthold S, Davisson M, Newcomer C, Quimby F, Smith A (eds) The mouse in biomedical research, vol 3 (Normative biology, husbandry, and models), 2nd edn. Elsevier Academic, San Diego Cho HJ, Ackerstaff E, Carlin S et al (2009) Noninvasive multimodality imaging of the tumor microenvironment: registered dynamic MRI and PET studies of a preclinical tumor model of tumor hypoxia. Neoplasia 11:247–259 Contag PR, Olomu IN, Stevenson DK (1998) Bioluminescent indicators in living mammals. Nat Med 4:245–247 Gross S, Piwnica-Worms D (2006) Molecular imaging strategies for drug discovery and development. Curr Opinion Chem Biol 10:334–342

References Beckman N, Kneuer R, Gremlcih HU et al (2007) In vivo mouse imaging and spectroscopy in drug discovery. NMR Biomed 20:154–185

16 Hoffmann U, Brix G, Knopp MV et al (1995) Pharmokinetic mapping of the breast: a new method for dynamic MR mammography. Magn Reson Med 33:506–514 Lindner J (2004) Microbubbles in medical imaging: current applications and future directions. Nature Rev Drug Discov 3:527–532 Malakoff D (2000) Suppliers: the rise of the mouse, biomedicine’s model mammal. Science 288:248–253 Ntziachristos V, Tung CH, Bremer C et al (2002) Fluorescence molecular tomography resolves protease activity in vivo. Nat Med 8:757–760 Ntziachristos V, Ripoli J, Wang LV et al (2005) Looking and listening to light: the evolution of whole-body photonic imaging. Nat Biotechnol 23:313–320 Pichler BJ, Wehrl HF, Kolb A et al (2008) Positron emission tomography/magnetic resonance imaging: the next generation of multi-modality imaging. Semin Nucl Med 38:199–208 Pomper MG (2001) Molecular imaging: an overview. Acad Radiol 8:1141–1153 Pomper MG (2005) Translational molecular imaging for cancer. Cancer Imaging 5:S16–S26

P. Zanzonico Rowland DJ, Cherry SR (2008) Small-animal nuclear medicine: Instrumentation and related considerationsin pre-clinical research. Semin Nucl Med 38:208–222 Serganova I, Mayer-Kukuck HH et al (2008) Molecular imaging: reporter gene imaging. In: Semmler W, Schwaiger M (eds) Molecular imaging II. Handbook of experimental pathology 185/II. Springer, Berlin Smith-Jones PM, Solit DB, Akhurst T et al (2004) The pharmacodynamics of HER2 degradation in response to HSP90 inhibitors can be imaged in animals with 68Ga labeled F(ab¢)2 fragments of Herceptin. Nat Biotechnol 22:701–706 Wisner ER, Weichert JP, Longino M et al (2002) A surfacemodified chylomicron remnant-like emulsion for per cutaneous computed tomography lymphography: synthesis and preliminary imaging findings. Invest Radiol 37: 232–239 Zanzonico P (2008) Multimodality image registration and fusion. In: Dhawan AP, Huang HK, Kim DS (eds) Principles and advanced methods in medical imaging and image analysis. World Scientific, Singapore

2

Non-Invasive Imaging in the Pharmaceutical Industry Sally-Ann Ricketts, Paul D. Hockings, and John C. Waterton

Contents

2.1 Introduction

2.1 Introduction............................................................. 17

Over recent years, whilst it is taking increased time and cost to get a new drug to market, productivity is declining. It is critical for the survival of the pharmaceutical industry, to both improve efficiency of the drug discovery and development process and to increase success rates, with better predictivity from preclinical to clinical investigation (Kola. 2008). Imaging provides unique tools to improve efficiency and predictivity. Imaging uniquely provides non-invasive, localized and quantitative information on focal disease and allows follow-up studies to be carried out within the same animal or human subject. Although animal imaging has been utilized in drug discovery and development since the 1970s, there has been much increased interest over recent years. This has been driven by the ever-expanding roles that imaging can play within drug discovery, by advances in imaging technologies and also by the increased availability and reliability of preclinical scanners for all modalities. The overall aim for the use of imaging within the pharmaceutical industry is to accelerate drug discovery by improving the speed and quality of decisionmaking. This can be achieved throughout the drug discovery pipeline, from increasing target confidence, decreasing time taken within the make-test cycle, better understanding of preclinical pharmacology and providing toxicology and efficacy biomarkers. Each of these will be explored in more detail, together with examples, throughout this chapter. The whole range of imaging modalities can be utilized within drug discovery, from optical (bioluminescence and fluorescence), ultrasound, X-ray technologies (radiography, CT and DEXA), magnetic resonance (MRI, MRS) and nuclear medicine (SPECT, PET). Again examples of these will be given throughout this chapter.

2.2 Target Confidence................................................... 18 2.3 The Design-Make-Test Cycle................................. 20 2.4 Preclinical Pharmacology...................................... 21 2.5 Preclinical Toxicology............................................. 23 2.6 Biomarkers.............................................................. 25 2.7 Conclusions.............................................................. 27 References............................................................................ 27

S.-A. Ricketts (*) and J.C. Waterton Translational Sciences, AstraZeneca, Alderley Park, Macclesfield, UK e-mail: [email protected] P.D. Hockings Translational Sciences, AstraZeneca, Mölndal, Sweden

F. Kiessling and B.J. Pichler (eds.), Small Animal Imaging, DOI: 10.1007/978-3-642-12945-2_2, © Springer-Verlag Berlin Heidelberg 2011

17

18

Imaging can bring additional ethical benefits in reducing the number of animals (Hultin et al. 2004; McIntyre et al. 2004; Bowyer et al. 2009; Waterton et al. 2000) because the same subject can be followed throughout the study; sample sizes may also be reduced because of better exclusion criteria and more reproducible endpoints. In addition, refinement of animal experimentation can often be achieved since the sensitivity of imaging endpoints may allow use of milder models or even use of animals that are clinically normal (Tessier et al. 2003; Bowyer et al. 2009). Many imaging studies within drug discovery are aimed at providing translatable data that will be used within the drug’s clinical development, for example, influencing the clinical imaging schedule to look at drug efficacy with an imaging biomarker. Equally, there are examples of where clinical imaging can influence the preclinical imaging approach (Bowyer et al. 2009). Despite these benefits, it should be recognized that few animals, perhaps only 1–2% of the total used in pharmaceutical drug discovery projects, ever receive an imaging measurement. Most are assessed using other, equally valid, techniques such as histology, molecular biomarkers or behavioural assessment. The decision to use imaging should be made on a case-by-case basis and will be dependent upon disease area, drug mechanism of action, questions that need to be addressed and the imaging technology within the specific field. We now briefly describe some examples from our own and other laboratories where imaging has helped decision-making at different points in the drug discovery process.

2.2 Target Confidence Of all the genes in the human genome, by some estimates, about 10% offer potential targets for pharmacologic intervention in disease, while perhaps only 1%

S.-A. Ricketts et al.

have been exploited to date (Hopkins and Groom 2002). Imaging can play a valuable role in selecting the target, understanding the relationship between target distribution and activity and measuring focal progression of pathology. It is therefore natural that imaging plays an important role in building target confidence. Formerly, some drug researchers regarded target validation as a task to be carried out when a novel target was identified, and then might neglect the quality of the target again until early clinical trials. Today, all companies recognize that it is critical to build target confidence as drug projects progress through the pharmaceutical pipeline in order to reduce the unacceptable levels of attrition seen in late stage clinical trials. Target validation prior to initiating a chemistry campaign: In the early phase of drug discovery, the key issue is speed and agility, so there is usually no time to develop new tracers or animal models. Target activity can be modulated by, for example, either knocking out or over-expressing the target in transgenic animals or using existing small molecules or biologics known to modulate target activity. Target confidence in man: Target validation in man means establishing that target activity is causally related to the symptoms of the disease under investigation. Clearly, one can use the same positive control approach that is used preclinically if clinically approved substances are available. Another approach is to label the drug candidate with a positron-emitting radioisotope to assess target engagement. If, for example, the drug candidate can be shown to engage the target but the symptoms of the disease are not changed in a meaningful way, then the clinical hypothesis has been de-validated and resources can be deployed in other drug projects. In conclusion, imaging is a useful tool to build target confidence with the potential to reduce late phase attrition due to lack of efficacy. Importantly, it must be recognized that drug projects can be killed even in a relatively advanced phase if the target is de-validated.

19

2 Non-Invasive Imaging in the Pharmaceutical Industry

Example 1: MRI of Lung Oedema in Endothelin Receptor Antagonist Studies Positive control compounds can be used to test the link between the target and disease modulation. In this example, we investigated the link between inhibition of the endothelin receptor and lung oedema. MRI was selected as the imaging technique of choice because signal from normal lung parenchyma is effectively invisible, thus highlighting areas of oedema in the lung (Beckmann et al. 2001). Lung oedema was induced by injecting big endothelin-1 (BET) i.v. In vehicle-treated animals, oedema was observed to spread from the vasculature into the lungs whereas in animals pretreated with an endothelin receptor antagonist, oedema formation was effectively blocked (Bell et al. 2007). Chronic obstructive pulmonary disease (COPD) patient symptoms, lung function and quality of life may be improved by reducing extravasation and development of lung oedema. This work shows that blocking endothelin activity impacts lung oedema formation and thus may be a suitable COPD target. a

b

c percent change in lung signal intensity

80

vehicle

70

ET-R antagonist

60 50 40 30 20 10 0

0

5

10 15 20 time after iv BET (min)

25

30

(a) MR image through the thorax of a vehicle-treated rat after injection of BET. Note the presence of oedema in the lung. (b) Rat pretreated with an endothelin receptor antagonist displays no obvious lung oedema. (c) Time course of lung oedema formation after injection of BET

20

S.-A. Ricketts et al.

Example 2: PET Receptor Occupancy for Target Validation PET was used to show that the NPY Y5R antagonist MK-0557 achieved 90% receptor occupancy, even though there was no clinically meaningful weight loss in obese humans after 1 year of treatment. The PET data gave confidence that the maximal weight loss had been reached for MK-0557 and that this was not a sub-optimal dose or brain penetration issue. As a result, this target could be de-validated (Erondu et al. 2006). The same compound produced significant weight loss in a mouse obesity model, so it was hypothesized that the endogenous NPY tone in humans is much lower than that in rodents, highlighting the need for target validation studies in man (Fong 2008).

2.3 The Design-Make-Test Cycle Once the drug developer has selected a credible drug target and has found a starting point in chemical space (assuming the aim is to produce a small-molecule drug), the process of lead optimization (LO) begins. The structure of the chemical starting point has to be iterated to optimize the in vitro efficacy against target, to optimize pharmacokinetics compatible with oral dosing and to achieve an acceptable toxicological profile. This iterative process is sometimes described as the design-make-test cycle, since the chemists cannot

design their next iteration until they have the result of the biological test on the previous iteration. Although the design-make-test cycle sits at the very heart of drug discovery, in vivo studies play a relatively minor part at this stage. Most tests are performed in vitro, since these tend to be inexpensive, rapid and amenable to automation; moreover, it is likely to be unethical to perform an in vivo experiment where an in vitro alternative exists. Rarely, the complexity of the biology makes it impractical to model the disease adequately in vitro and in vivo models may be required at the later stages of LO.

21

2 Non-Invasive Imaging in the Pharmaceutical Industry

Example 3: MRI to Support Lead Optimisation for Acute Stroke An example of in vivo imaging in LO is in focal cerebral ischaemia (stroke) research, where the rat middle cerebral artery occlusion model has been used to choose between compounds with similar in vitro properties. An example is shown below where 13 compounds with good in vitro potency were compared in vivo, using an innovative group-sequential design to reduce animal numbers with no loss in statistical power. In this case, three out of 13 compounds were found inactive in vivo, and ten were found active: one of these ten was eventually progressed into preclinical development. 10−4

P value in Student's t-test

10−3

active

10−2

10−1

inactive 100

0

6 9 number of rats

12

15

MRI in rat permanent focal cerebral ischaemia model: 16 slices through the brain are shown. The area of infarct (arrow) can be easily be segmented to derive the infarct volume in the territory of the middle cerebral artery (Waterton et al. 2000). Inset shows the operation of the group-sequential design: significance testing was performed after 6, 9, 12 and 15 rats were entered for each LO compound, and if the calculated p value was very small (upper grey zone), a stopping rule was invoked and the compound declared active; conversely, a high p value (lower grey zone) invoked a second stopping rule and the compound was declared inactive

2.4 Preclinical Pharmacology Many preclinical models demonstrate aspects of a clinical disease but do not represent the full disease spectrum or time to disease progression. In order to understand drug pharmacological effects, it is important to have good characterization and validation of the preclinical model and how this relates to the clinical disease state. Typically, model validation studies will be performed using histology and/or other pharmacodynamic endpoints. Additionally,

anatomical, functional and molecular imaging techniques can all play roles in aiding preclinical model characterization and the changes that ensue following drug administration. Understanding and being able to quantify changes that result following drug administration in the preclinical setting are obviously of great importance within drug discovery. Such information will be utilized in decisions on the progression of the drug and may be incorporated into the Investigators’ Brochure and supporting publications. These imaging techniques

22

S.-A. Ricketts et al.

and methodologies may be specific only for the preclinical setting due to the type of imaging modality and/or availability/licencing of the contrast agent, or may be translatable to the clinical arena. Below are examples of the use of imaging in preclinical model characterization and changes following drug administration. Imaging may also be employed to evaluate drug distribution, with drug molecules labelled with a detectable probe and distribution monitored throughout time post injection. This type of approach has been used for

many years within the neurology disease area, with drugs typically labelled with 11C and brain uptake and localization evaluated using PET imaging. With the increased emergence of biological drug compounds, there has been a trend to understand their distribution and areas of accumulation. Due to the long blood halflife of biological compounds, short-lived isotopes such as 11C are not suitable, and therefore, biologics are typically labelled with either a fluorescent tag for optical detection or longer-lived radioisotope such as 64Cu or 111 In for PET and SPECT imaging, respectively.

Example 4: Hyperpolarised 3He to Monitor Drug Effects in a Murine COPD Model Hyperpolarised 3He MRI is an emerging respiratory imaging technology with key advantages that it can measure gas distribution, and importantly functional endpoints, which appear to be more sensitive than conventional 1H MRI (Olsson et al. 2009). This imaging methodology can also potentially be translatable to human proof-of-principle studies. We have used this technology to assess treatment effects of drugs using a lipopolysaccharide (LPS) murine model of lung inflammation to mimic the inflammatory response of COPD. Administration of LPS causes formation of lesions and ventilation defects within the lungs, which can be easily visualized using hyperpolarised 3He. The example shown (courtesy of Lars E. Olsson) compares lung lesion volumes in untreated mice with those exposed to LPS and either treated with vehicle or budesonide. Exposure to LPS induced significant inflammation and large ventilation defects. LPS exposure together with budesonide treatment resulted in a significantly smaller volume of ventilation defects. Cross validation of the MRI data was performed using bronchoalveolar lavage (BAL), with inflammatory markers supporting the findings from MRI.

Control

LPS

LPS + Budesonide

Volume lesion/animal (ml)

0.15

0.10

***2 0.05 ***1 0.00 CONTROL

VEHICLE

BUDESONIDE

23

2 Non-Invasive Imaging in the Pharmaceutical Industry

Example 5: MRI to Evaluate Compound Efficacy in a Guinea Pig Osteoarthritis Model One of the applications of MRI is to provide high, quantifiable anatomical information. We developed methods to enable high-resolution 3D MRI to characterize cartilage loss in a model of spontaneous knee osteoarthritis in the Dunkin Hartley guinea pig (Tessier et al. 2003). Using a non-invasive longitudinal technique to monitor disease progression or modification with therapy over time has real cost benefits when dealing with more complex, longer duration disease models. In this model, we have used MRI to evaluate the effects of doxycycline treatment and compared the results to matrix metalloproteinase (MMP) activity (Bowyer et al. 2009). Guinea pigs (9 months old) were dosed with vehicle or doxycycline (0.6 or 3.0 mg⁄kg⁄day) for 66 days. Fat-suppressed 3D gradient-echo MRI of the left knee was acquired pre and post dosing. Change in medial tibial plateau (MTP) cartilage volume (MT.VC) was determined using image analysis. At termination, MTP cartilage was removed from knees and proteolytic MMP activity determined using a fluorescent peptide substrate assay. Results showed that vehicle-treated animals lost 20.5% MT.VC. The 0.6 mg⁄kg⁄day doxycycline group lost 8.6% whilst the 3.0 mg⁄kg⁄day group lost 10.0%. Endogenous levels of active MMPs were below limits of detection in all samples. However, doxycycline treatment ablated amino phenyl mercuric acid activated MMP-13 and MMP-8 levels and reduced MMP-9 levels by 65% and MMP-1 levels by 24%. Doxycycline treatment thus resulted in partial protection from MT.VC loss although it was associated with complete reduction in MMP-13 and MMP-8 and partial reduction in MMP-9 activity. Doxycycline has only limited efficacy in man, and these data suggest that compounds providing better chondroprotection need to be sought.

Example 6: 111In-DTPA-Pertuzumab Sensitively Detects Trastuzumab-Mediated HER2 Downregulation Pertuzumab is a HER2 dimerization inhibitor that binds to an epitope unique from that of trastuzumab. SPECT studies both in vitro and in vivo demonstrated that 111In-DTPA-pertuzumab could sensitively detect HER2 downregulation after 3 days of treatment with trastuzumab and detected a reduction in viable HER2-positive tumour cells after 3 weeks of therapy in MDA-MB-361 human breast cancer xenografts (McLarty et al. 2009).

2.5 Preclinical Toxicology Safety assessment of candidate drugs involves taking the data from model biological systems (e.g. cell cultures, laboratory animals) and using these results to try to predict toxicity in humans. The continued attrition of late stage compounds due to clinical safety issues indicates that our ability to predict toxicity can be further optimized, and additional tools to provide improved data on which to base predictions would be most welcome. Drug developers need to show that drugs are both efficacious and have acceptable toxicity: while in vivo imaging has been used extensively in both preclinical and clinical efficacy studies, it has not as yet seen widespread use in preclinical toxicology (Hockings 2006). The ability to image the same animal at multiple timepoints suggests that imaging is likely to make its greatest contribution in chronic investigational safety studies. There are also clear advantages such as

bridging the gap between animals and humans by using the same non-invasive endpoints and utilizing functional safety biomarkers not readily available otherwise, e.g., cardiac function. Despite these examples of safety imaging biomarkers that are ready to use, there are relatively few imaging studies with a safety endpoint. The key blockers for imaging studies tend to be lack of knowledge about in vivo imaging by preclinical toxicologists because of the sparse literature, and a lack of time to conduct appropriate validation studies with positive and negative control substances as safety issues arise unexpectedly and then often become critical path activities for drug projects. Other issues can be concerns about confounds from anaesthesia for restraint (not always necessary), logistics, and a concern that imaging studies are not usually conducted under (or close to) the exacting regulatory standards of Good Laboratory Practice. However, none of these problems are completely insurmountable.

24

S.-A. Ricketts et al.

Example 7: Cholestasis The need for better biomarkers of drug-induced liver injury (DILI) has been identified (Holt and Ju 2006). In particular, drug-induced cholestasis is an area where MR imaging can make a contribution as new clinically approved contrast agents such as gadoxetate (Gd-EOB-DTPA, Bayer Schering) have become available. Gadoxetate is taken up into hepatocytes and subsequently excreted into bile, and hence, cholestasis can be detected by the accumulation of gadoxetate in the liver with resultant enhancement on the MR image. Ulloa et al. (2010) gave a single i.v. bolus dose of b-Estradiol 17-(b-d-glucuronide) (E217Bg) to induce an acute but transient cholestasis in rats by inhibiting the multidrug resistance protein 2 (Mrp2) and the bile salt export pump (Bsep). Dynamic contrast-enhanced magnetic resonance imaging (DCE MRI) was used to assess clearance of gadoxetate by monitoring changes in liver contrast over a 2 hour period. E217Bg was found to significantly increase the half-life of gadoxetate clearance from the rat liver from 30 to 180 min presumably due to blockage of Mrp2 and reduction in bile flow. Gadoxetate DCE MRI may therefore be a suitable biomarker of cholestasis. Furthermore, cholestasis imaging has the potential to be of use as a safety biomarker in clinical studies. a

b

c

d

e

f

g

h

Examples of dynamic MR images through the liver of vehicle (top row) and treated (bottom row) rats at t = 0 (a, e), 10 (b, f), 40 (c, g), 70 (d, h) min after contrast injection. Signal intensity in the liver was normalized to the signal of the phantom (white arrow in (a))

Example 8: Gastric Emptying

vehicle 10 25 50

6

Gastric emptying is not always studied as part of the safety pharmacology core battery but can be examined if appropriate. The standard assay is to dose a dye such as charcoal and then kill the animal after a fixed period of time and examine intestinal transport. Obviously, this method is limited to examinations at a single timepoint. Hultin et al. (2004) used functional X-ray of conscious rats after a barium meal to examine the timecourse of gastric emptying and the effect of various doses of corticotropin releasing factor (CRF). They showed that functional X-ray measurements provided a reproducible and quantifiable method to evaluate gastric motility and emptying.

5 4 3 2 1

−5

0 0

5

10

15

20

Timecourse of gastric motility effects after i.v. CRF (mg/kg) in conscious rats

2 Non-Invasive Imaging in the Pharmaceutical Industry

2.6 Biomarkers In the last few years, the regulatory, academic and drug development communities have developed an increasingly sophisticated understanding of the nature of biomarkers and their use in drug development. A biomarker is now understood very generally (Atkinson et al. 2001), as any “characteristic that is objectively measured and evaluated as an indicator of normal biological processes, pathogenic processes or pharmacologic responses to a therapeutic intervention”. Imaging measurements are included within this definition, as are the more familiar molecular biomarkers and even physiologic tests such as electrocardiography or blood pressure. Importantly, from the regulatory perspective, a biomarker is not necessarily seen as a surrogate endpoint. Thus, a biomarker can still be useful in drug development, even when it is not exhaustively validated to eliminate all possible risk of a false-positive or false-negative outcome. The process of qualification, i.e. understanding the value of a biomarker, involves establishing the relationship between the measured biomarker and the underlying biology, as well as understanding what a biomarker measurement implies for the patient’s ultimate clinical outcome. Imaging biomarkers are used in drug development in many diseases, in particular in inflammation, degeneration, ischaemia and neoplasia. In phase 1 and phase 2 trials, imaging biomarkers are used to establish that the drug reaches its target, has some pharmacologic effects on the cell phenotype, causes some local physiologic change such as a change in perfusion or local inflammation and ultimately modifies the structural progression of the disease. The actual use of imaging biomarkers in phase 1 and phase 2 drug development obviously involves human clinical studies and is therefore outside the scope of this chapter. However, before such an imaging biomarker is used with investigational agents (particularly with first-in-class drugs against targets not

25

previously addressed in man), animal studies using the same imaging biomarker/drug combination are often critical, both in the design of the clinical trial, and in interpreting its outcome. Ideally, the imaging biomarker has a similar presentation in the animal model and in the human patient. Examples in which imaging biomarkers have been translated between animal and man include cancer (FDG-PET, FLT-PET, DCE MRI, ADC MRI, phosphomonoester MRS), psychiatry (functional MRI, pharmacoMRI), myocardial or cerebrovascular ischaemia (ADC MRI, perfusion), atheroma (ultrasound, CT, MRI and PET measurements of plaque size, composition and inflammation), respiratory diseases (CT and MRI measurements of emphysema, perfusion and ventilation ), arthritis (MRI and X-ray measurements of bone lesions, cartilage and synovitis), osteoporosis (bone mineral density measurements) diabetes, obesity, thrombosis, neurodegeneration and many others. Imaging biomarkers are also increasingly used in safety assessment, for example cardiac ultrasound to assess and monitor the safety of cardiotoxic anticancer drugs. What measurements are important using imaging biomarkers in animals? Firstly, we can ask whether the new drug affects the imaging biomarker and the effect size. Using knowledge of the relationship between the animal model and human disease, we can use this information to predict the likely size of the drug effect on the imaging biomarker in a clinical trial, which is of course essential to sample size calculations. We can measure the duration of the effect and the dose response. Importantly, we can verify that the change in the imaging biomarker following drug treatment parallels changes in the underlying histopathology: these measurements are usually much easier to perform in animals than in humans but are essential in qualifying the biomarker. Finally, we may be able to use our imaging biomarker studies in animals to inform our thinking about whether a given change in the imaging biomarker will predict a specific clinical benefit to the patient.

26

S.-A. Ricketts et al.

Example 9: DCE MRI Proof of Principal Biomarker for a Vascular Disrupting Agent

10

0

−20 −30 −40 −50 −60 −70 *

−80 −90 *

Change in DCE MRI enhancing fraction (EHF60), 24 hours after treatment with increasing doses of ZD6126. Insets show H&E histopathology and corresponding DCE MRI Ktrans maps, which can be used to help interpret the changes seen in human tumours (Bradley et al. 2007)

a

AZD8055

Vehicle

b 2.5 2.0

1.0 0.5 0.0

*** 1.41± 0.10

1.5

2.09± 0.19

mTOR is a serine/threonine kinase belonging to the P13K superfamily of kinases, and the P13K-AKTmTOR pathway is involved in glucose uptake and metabolism. Using a human glioma xenograft model (U87-MG), we evaluated whether FDG-PET could be used as a direct mechanistic readout of TOR kinase inhibition. To evaluate this, U87-MG tumour-bearing mice were administered with either a single dose of mTOR kinase inhibitor (AZD8055) or vehicle, 15 minutes later administered with FDG-PET and imaged 1 hour after drug or vehicle dosing. Results showed there was a significant decrease in tumour FDG uptake in the animals treated with AZD8055 compared to vehicle group. Imaging data correlated with histological pharmacodynamic readouts, with marked reductions of both pS6 and pAKT immuno staining with treatment. Therefore, this data suggested that FDG uptake could be used as an early sign of metabolic response to mTOR inhibition (Keen et al. 2010).

57.510

−10

MaxSuV

Example 10: FDG-PET as a Proof of Mechanism Biomarker for an mTOR Kinase Inhibitor

%age change in EHF

An example of animal imaging to qualify a biomarker for use in human drug development is the use of DCE MRI in the development of ZD6126. DCE MRI was used in this case as a proxy for tumour perfusion. ZD6126 is a vascular destructive agent: it targets immature tumour vasculature causing massive intratumoral infarction. Studies in xenografts in mice and rats in three different laboratories established the dose response, speed of onset and the persistence of the infarction (Bradley et al. 2007; McIntyre et al. 2004; Evelhoch et al. 2004), as well as optimizing image analyses at intermediate doses where complete infarction was not achieved. In addition, the relationship between the imaging change and histopathological change was established. These data were then used to design and interpret the phase 1 trials of the compound in man.

60

0

ZD6126 (mg/kg)

2.5

Vehicle

AZD8055

(a) Single slice of FDG-PET images of U87-MG xenograft bearing nude mice; (b) tumour FDG MaxSUV (mean ± SEM). Triple asterisk p = 0.009

2 Non-Invasive Imaging in the Pharmaceutical Industry

2.7 Conclusions As discussed at the beginning of this chapter, the aim for any science or technology used within drug discovery and development is to accelerate the drug discovery process by improving the quality of decision-making. The challenge is to provide data in a timely fashion to impact decision-making, and therefore, there is a preference for ex vivo imaging techniques such as histopathology, immunohistochemistry and autoradiography, which have excellent spatial resolution and exquisite molecular sensitivity and specificity. However, although in vivo imaging is generally more costly and can lack sensitivity and specificity in comparison with the ex vivo modalities, hopefully throughout this chapter, we have highlighted the range of drug discovery activities and areas of impact and value that in vivo imaging can have within preclinical drug discovery and development.

References Atkinson AJ Jr, Colburn WA, DeGruttola VG, DeMets DL, Downing GJ, Hoth DF, Oates JA, Peck CC, Schooley RT, Spilker BA, Woodcock J, Zeger SL (2001) Biomarkers and surrogate endpoints: preferred definitions and conceptual framework. Clin Pharmacol Ther 69:89–95 Beckmann N et al (2001) Pulmonary edema induced by allergen challenge in the rat: non-invasive assessment by magnetic resonance imaging. Magn Reson Med 45:88–95 Bell JP, Chevalier E, Tessier J, Bradley D, Mather ME, Young SS, Bahl AK, Escott KJ, Young A (2007) Use of magnetic resonance imaging (MRI) to track big endothelin-1-induced edema in the rat lung. Am J Respir Crit Care Med 175:A533 Bowyer J, Heapy CG, Flannelly JK, Waterton JC, Maciewicz RA (2009) Evaluation of a magnetic resonance biomarker of osteoarthritis disease progression: doxycycline slows tibial cartilage loss in the Dunkin Hartley guinea pig. Int J Exp Pathol 90:174–181 Bradley DP, Tessier JJ, Ashton SE, Waterton JC, Wilson Z, Worthington PL, Ryan AJ (2007) Correlation of MRI biomarkers with tumor necrosis in Hras5 tumor xenograft in athymic rats. Neoplasia 9:382–391 Erondu N et al (2006) Neuropeptide Y5 receptor antagonism does not induce clinically meaningful weight loss in overweight and obese adults. Cell Metab 4:260–262 Evelhoch JL, LoRusso PM, He Z, DelProposto Z, Polin L, Corbett TH, Langmuir P, Wheeler C, Stone A, Leadbetter J,

27 Ryan AJ, Blakey DC, Waterton JC (2004) Magnetic resonance imaging measurements of the response of murine and human tumors to the vascular-targeting agent ZD6126. Clin Cancer Res 10:3650–3657 Fong TM (2008) Development of anti-obesity agents: drugs that target neuropeptide and neurotransmitter systems. Expert Opin Investig Drugs 17:321–325 Keen H, Ricketts SA, Bales J, Shannon A, Logie A, Odedra R, Wedge SR, Guichard SM (2010) The mTOR kinase inhibitor AZD8055 modulates 18F-FDG uptake in vivo in the human glioma xenograft model U87-MG. In: Proceedings AACRNCI-EORTC Molecular Targets and Cancer Therapeutics, Boston, p A225 Kola I (2008) The state of innovation in drug development. Clin Pharmacol Ther 83:227–230 Hockings PD (2006) Magnetic resonance imaging in pharmaceutical safety assessment. In: Vogel HG, Hock FJ, Maas J, Mayer D (eds) Drug discovery and evaluation: safety and pharmacokinetic assays. Springer, Heidelberg, pp 385–393 Holt MP, Ju C (2006) Mechanisms of drug-induced liver injury. AAPS J 8:E48–E54 Hopkins AL, Groom CR (2002) Opinion: the druggable genome. Nat Rev Drug Discov 1:727–730 Hultin L, Hyberg G, von Mentzer B, Martinez V (2004) Effects of corticotropin releasing factor and urocortins on gastric emptying and motility in rats assessed by functional X-ray. In: XII European Symposium on Neurogastroenterology and Motility, Cambridge, UK, p T150 McIntyre DJ, Robinson SP, Howe FA, Griffiths JR, Ryan AJ, Blakey DC, Peers IS, Waterton JC (2004) Single dose of the antivascular agent, ZD6126 (N-acetylcolchinol-Ophosphate), reduces perfusion for at least 96 hours in the GH3 prolactinoma rat tumor model. Neoplasia 6:150–157 McLarty K, Cornelissen B, Cai Z, Scollard DA, Costantini DL, Done SJ, Reilly RM (2009) Micro-SPECT/CT with 111In-DTPA-Pertuzumab sensitively detects trastuzumabmediated HER2 downregulation and tumor response in athymic mice bearing MDAMB-361 HUMAN breast cancer xenografts. J Nucl Med 50:1340–1348 Olsson LE et al (2009) 1H and hyperpolarized 3He MR imaging of mouse with LPS-induced inflammation. J Magn Reson Imaging 29:977–981 Tessier JJ, Bowyer J, Brownrigg NJ, Peers IS, Westwood FR, Waterton JC, Maciewicz RA (2003) Characterisation of the guinea pig model of osteoarthritis by in vivo three-dimensional magnetic resonance imaging. Osteoarthritis Cartilage 11:845–853 Ulloa J et al (2010) Effects of a single intravenous dose of estradiol 17b-glucoronide on bile flow: assessment with gadoxetate DCEMRI. Proceedings ISMRM 18, 2593 Waterton JC, Middleton BJ, Pickford R, Allott CP, Checkley D, Keith RA (2000) Reduced animal use in efficacy testing in disease models with use of sequential experimental designs. Dev Anim Vet Sci 31:737–745

3

How to Set Up a Small Animal Imaging Unit David Stout

3.1 Overview

Contents 3.1 Overview.................................................................. 3.1.1 Site Requirements..................................................... 3.1.2 Animal Use Issues.................................................... 3.1.3 Anesthesia................................................................. 3.1.4 Radiation Use........................................................... 3.1.5 Biosafety................................................................... 3.1.6 Infrastructure Requirements..................................... 3.1.7 Ergonomic Flow....................................................... 3.1.8 Training and Education............................................. 3.1.9 Security..................................................................... 3.1.10 Budgeting.................................................................. 3.1.11 Fire and Safety Issues...............................................

29 30 32 33 35 37 37 41 42 43 44 45

3.2 Summary................................................................. 45 References............................................................................ 45

D. Stout UCLA Crump Institute for Molecular Imaging, Preclinical Imaging Center, 570 Westwood Plaza, CNSI room 2151, Los Angeles, CA 90095, USA e-mail: [email protected]

Creation of an animal imaging center is rarely as simple as purchasing a system, placing it in a room somewhere, and turning people loose to use it. While there are a few imaging devices that may be appropriate to place on a laboratory bench top and use right out of the box without further integration, for example, a microscope, most systems require training of users and a specialized environment for handling data and animals. The use of animals requires approval of physical spaces, imaging systems, and protocols by Institutional Animal Care and Use Committees (IACUC), radiation use requires approval from Radiation Safety Offices (RSO), local institutional veterinary considerations may apply, and usually biosafety approval is required. There may also be other Environmental Health and Safety (EHS) issues to consider, such as magnetic field safety and air flow requirements. This process of comprehensive planning and oversight is now commonplace to house, use, and experiment with animals (Stout et al. 2005; Klaunberg and Davis 2008). The first step in designing an imaging center is to determine what types of research will be supported, a process that then leads to what imaging modalities and specific imaging systems are required. There are many different modalities and vendors selling various imaging systems, most of which are quite expensive, so careful selection is required to devise the optimal mix of systems (Stout and Zaidi 2009). Imaging modalities are now often combined, such as PET/CT, PET/MR, and optical/X-ray, so consideration must be given to whether the combination is useful or limiting with respect to the intended workflow of the center (Table 3.1). The support infrastructure for various imaging systems varies in terms of space, power, air conditioning,

F. Kiessling and B.J. Pichler (eds.), Small Animal Imaging, DOI: 10.1007/978-3-642-12945-2_3, © Springer-Verlag Berlin Heidelberg 2011

29

30 Table 3.1 Design considerations for a small animal imaging center Purpose of the imaging facility Types of imaging equipment: PET, CT, ultrasound, MRI, optical, etc Ancillary equipment: gamma counter, cryostat, biosafety cabinet, anesthesia Ancillary spaces: surgical area, hot lab, supply storage, animal housing, cage cleaning Barrier conditions for immune-compromised animals (SCID, nude animals) Security requirements Radiation use and electromagnetic safety requirements Biohazard requirements Facilities planning: hoods, air, vacuum, tanks Room layouts: space, work path, supply delivery path Animal use issues: gowning, access control, waste management, pathogen control Vivarium support: cage racks, supplies, cage cleaning, access control Anesthesia: vaporizers, induction boxes, waste management Mouse handling: prep area, injection techniques Mouse imaging chambers: positioning, multimodality, sterility Hypothermia prevention: heating systems for before, during, and after imaging Personnel: locations, break area, training, staffing requirements

and ancillary equipment necessary. For example, a cyclotron and radiochemistry lab or isotope delivery arrangement is necessary for PET and SPECT systems, magnetic field restrictions are required for MR, and a low radiation background location is needed for gamma counters. The use of immune-compromised animals and human xenograft tumors may lead to biosafety requirements that specify usage conditions such as gowning chambers, barrier facilities, and controlled access. Often it is advisable to have a dedicated vivarium to house the animals, which might also include a separate room for surgical or other interventions. Personnel support will depend again on the systems being used and the level of training and support planned for the experiments. Imaging systems capable of examining small animals are a relatively recent development; thus, new systems and methods for imaging are frequently being created and brought to market. These new tools, techniques, and protocols may need to be included or

D. Stout

exchanged into the center in the future. Either a center employee or a faculty member associated with the center will need to keep informed about imaging systems, methodology, and the changing regulatory environment associated with the field. The lifetime and depreciation schedule of these systems may be shorter than similar clinical systems. An example of the changing requirements for imaging is given in Figs. 3.1–3.3, showing how the imaging center at the Crump Institute at UCLA has evolved over the past 20 years. At first, a relatively small room to house one or two systems was sufficient. As more systems were added, it became clear that a consolidated location was needed to bring the systems into one area instead of being spread between several buildings. As more systems and support areas were required, eventually a new building was created where the layout could be expanded and optimized, instead of working within the confines of an existing building. A little extra space and planning ahead for possible increases in workflow may save time and money when expansion is needed. When designing a center, the process can be divided into several categories: vendor specific site requirements (cooling, power, space), computer infrastructure requirements for archiving and database information, on-site animal use and housing requirements, radiation and biosafety requirements, and human ergonomic planning to make sure people and equipment can easily and safely coexist. Optimal center design often includes the combined expertise of architects, members of the oversight committees, animal care personnel, and people experienced and familiar with the usage requirements of the imaging systems. Working as a team and getting the center open quickly will reduce the chance of unforeseen obstacles and delays.

3.1.1 Site Requirements Equipment manufacturers normally provide site-planning guides with equipment specifications and space requirements. Typically, there are power and air conditioning requirements that specify dedicated power circuits and appropriate temperature and humidity ranges. Additional space around the systems is not usually specified but is often necessary for good access during normal use and may be required for servicing and good airflow around the systems. Requirements vary widely

31

3 How to Set Up a Small Animal Imaging Unit Door

Door

Door

Hood

Lead cave

6' Biosafety Cabinet

Potent a m croPET

Countertop

Door

O2

sink

Computers

Nearby Computer server room

microCAT

Chair

microPET/VT

Chair

02 02

Computers

Air Air

Investigator

Optical

Storage

Optical Surgical Area

S

Animal Prep

Optical

Biosafety Cabinet

Storage

Dose Station

S

Optical

S

Hot Lab

S

Lead Cave

Viewing Area

Cart Space

microPET Prototype

Countertop

Chair Window

2 computers Window 2 computers

2 computers Window

14x15

Display computers

Investigators & Staff Only

14x9

14x16

Countertop

microPET

Door

O2

P4 micro PET

Countertop

Fig. 3.1 UCLA small animal imaging center from 1990 to 2003. converted from an old fluoroscopy imaging suite. Originally, only the left and center rooms were used to house a clinical PET system solely used for animal work. In 1997, it was expanded to include the right side room for the prototype microPET, then in 2000, the clinical system was exchanged for a commercial microPET system and a biosafety cabinet (hood) added

People Animals

Radiolabeling Laboratory

Air Air

Freezer

Autoradiography

Supplies Autorad

Adjacent Vivarium

Cryotome

Door

Radiation, supplies

Fig. 3.2 UCLA Crump imaging center from 2003 to 2009. This location was renovated within an existing building. The computer cluster room and radiolabeling lab were added in 2005 to

accommodate expanding support requirements for radiolabeling I-124 and Cu-64 and the need to shift image processing away from the host computers

based on the systems, from a simple space on a counter to specialized buildings. Large scale installations such as cyclotrons and MRI imaging systems require significant space and additional magnetic and radiation shielding requirements. For projects of this scale, manufactures often require specific construction plans and usually work closely with sites to create the appropriate building systems and floor plans. The precise requirements are specific to the system being installed; thus, these types of site plans are highly customized to match the exact requirements for the end user.

Most imaging systems such as PET, CT, SPECT, optical, and autoradiography require a dedicated power circuit, rely on room air circulation for heat removal, and need space on either the floor or the counter for use. Except for MR, these systems can be located in almost any room or together in various combinations. Systems using radiation need to have a separate area for holding and dispensing radioactive doses, and consideration is necessary for placement with respect to personnel exposure. It may be required to have dosimetry estimates for both personnel within the center and any adjacent surrounding areas that

32

Maestro

Dose Drawing

Optical

Display Computers

chair

Dose Drawing

Display Room

Gamma Counter

CryT WB

cryostat

Counter

−20

sink

WB cryostat Lead Fume hood Cave

AutoradRadiolabeling

Hood

counter

Supply Cabinets

Fuji

Counter

microPET-CT

Optical 1

mPET

Island Countertop

−80 Doorway

BSC

Lead Cave

Optical 2

BSC

Island Countertop

Computer

chair

PETbox Computers mCT

BSC

chair

mPET

Supplies Supplies

D. Stout

P−32

Vivarium immediately adjacent to imaging facility

Fig. 3.3 UCLA Crump imaging center as of 2009. The new location increased space ~50% over the previous design, with the vivarium 10× larger than before. The added space lowers radiation exposure and allows more people to work easily at the

same time. A third PET imaging system was added (PETbox), and two spaces remain open in the optical room for future expansion

may be close enough to have public radiation exposure. Fortunately, the amount of radioactivity required for most small animal work is fairly small and represents minimal exposure risks. The physical requirements for imaging systems are often fairly simple to meet; the animal use requirements and biosafety considerations are usually more challenging. Once purchased and installed, imaging systems need to be evaluated and a determination made whether they are performing as specified. Acceptance testing is an important step that requires personnel to have some training and expertise concerning how the systems should function. This expertise may come from vendor training or experience gained from the personnel who trained on similar systems at another site. Proper training is essential; otherwise, the investment of time, money, and resources might not be used properly or appropriately.

animals, including primates, pigs, and dogs. Often, the optimal species is based on the availability of a suitable model of human disease, the need to see small structures (resolution and sensitivity limits of systems are what determine what can be measured), and perhaps the importance of behavioral or biochemical measurements. Design specifications of an imaging center require knowledge of the species intended for use, since there may be restrictions on mixing species in the same physical spaces. There are specific requirements for vivarium and animal use area described by the Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC). Accreditation by this group, although voluntary, is often a requirement by various funding agencies. Each institution also has an IACUC that oversees all animal work. The variety of research projects will also place demands on design, such as the use of biohazardous materials and carcinogens, difference species and the need to provide support for surgical, delivery of biological agents using viral vectors or other kinds of interventions. Perhaps the best way to meet the animal use constraints is to have a dedicated vivarium associated with the imaging center, where the appropriate physical spaces are provided to enable the intended animal use requirements. The exact requirements depend on the

3.1.2 Animal Use Issues Preclinical imaging work can utilize a variety of animal species, which have different housing and isolation requirements. While mice and rats are now the most frequently used species, there are still many projects that are best conducted with other

33

3 How to Set Up a Small Animal Imaging Unit

requirements of the institution, planned research, and AAALAC guidelines. There are some procedures such as tumor measurements and drug dosing that may be required frequently, thus are best served within the vivarium rather than in the imaging spaces. The use of immune-compromised animals might require separate housing of animals within dedicated barrier conditions and the use of microisolator cages. Retrofitting or reconfiguring an existing space for proper animal housing may be very difficult, so it is important to determine the needs and requirements in order to create the appropriate vivarium space (Fig. 3.4). Within the imaging center, the need for proper environmental support is crucial to create accurate and reproducible data (Fueger et al. 2006). This is particularly true for imaging research that involves measuring of in vivo metabolism, where the physiology of the animal can have profound effects upon the results. Equipment vendors vary widely with their environmental support efforts, with some systems providing nothing and others providing heating, anesthesia, and perhaps some positioning mechanism. A great deal of

Procedure room Animals

Animal Holding

Hot rat

Hot ms Barrier Area

Mop

People

Cage Wash Area

Bedding

Fig. 3.4 A full vivarium was added to the center in 2009, complete with cash-washing facilities, barrier areas (in blue), and separate housing for radioactive animals (hot rat and mouse racks). A procedures room was added with an anesthesia system and ducted biosafety cabinet, suitable for work with biohazardous and carcinogenic materials. Animals enter through a passthrough biosafety cabinet, while people go through the gowning chamber

Fig. 3.5 Heating plate used to preheat animals before anesthesia induction. Similar plates are used under all locations where animals are held during and after the imaging process. Also on the plate is an imaging chamber that when connected provides heating, anesthesia delivery, pathogen control and reproducible positioning

attention is often placed on center layout and equipment specifications; however, the environmental support is equally important and often far less expensive to properly implement. Good support includes welldesigned anesthesia systems to both provide and exhaust anesthetic agents, heating options, barrier containment capabilities, easy-to-use and clean imaging support devices or chambers, and a good way to reproducibly and optimally position animals within the imaging systems. Various options for imaging chambers are now available to provide barrier containment, gas anesthesia, heating, and positioning (Suckow et al. 2008) (Fig. 3.5). Any work involving hands-on animal handling ideally has a dedicated location with all the supplies and equipment close at hand. For surgical work, this might include a microscope, heating plate, fiber optic lighting (incandescent lighting can overheat animals), an anesthetic induction box, and nose cone assembly, along with any other tools or supplies. Waste disposal should also be convenient, such as regular and biohazardous waste containers, sharps containers, or shielded areas for radioactive trash. Frequently used supplies such as absorbent pads, gloves, needles, and syringes need a place ideally suited for easy access, along with cabinets for supply storage.

3.1.3 Anesthesia The use of anesthetic agents to immobilize animals is necessary for most experiments; otherwise, measurement

34

D. Stout

of where and how much signal is present is not possible due to blurring or motion artifacts. Most commonly used with preclinical imaging systems are gas inhalant anesthetics such as isoflurane, halothane, or sevoflurane, which provide a safe, effective, and constant depth of anesthesia. Injected agents can be used alone or are often in combination cocktails most suitable for the species being used. There are also an increasing number of options for imaging without anesthesia. Inhalant anesthetics are the anesthetic of choice for most experiments and are preferred by most investigators and veterinary staff. Induction and recovery of anesthesia are fast, typically a few minutes at most. The depth of anesthesia remains constant, and it can be administered continuously for as long as necessary. Symptoms of overdosing are easy to observe and readily reversed by lowering the amount of anesthetic agent being supplied. Gas anesthetics are not controlled substances in many countries, making procurement and supply storage simple, unlike injected anesthetic agents. The combination of ease of use, safety for the animals, and consistency make gas the ideal anesthetic agent for most preclinical research. If gas anesthetics are to be employed, then air or oxygen will be needed as a carrier gas. The choice of both carrier gas and anesthetic agent will affect the physiology and resulting image data (Flores et al. 2008). The placement of the supply tanks can be either near the vaporizer or in a specific location ideally suited to delivery and removal of the tanks. Locating the tanks near a door minimizes the need to move bulky and

potentially dangerous heavy objects near expensive and sometimes delicate imaging systems. If multiple vaporizers are to be used, a wall-mounted gas manifold, switching valve, and low-pressure alarm are worth considering (Fig. 3.6). For safety reasons, tanks and manifolds cannot be located close to any electrical panels. Oxygen lines require special installation and soldering to avoid explosions. To save valuable counter top space, vaporizers can be wall-mounted, ideally near where animals are used so that access to adjust anesthetic levels is within easy reach. With gas anesthetics, the waste anesthetic gas (WAG) stream must be addressed to avoid exposure to personnel in the center. A common method for WAG management is the use of F-air charcoal filters; however, this is not a good method for dealing with waste gasses. Charcoal quickly forms saturated channels, and it only takes a few minutes until the anesthetic gas starts to leak through the filter and back into the room. Charcoal also only absorbs the gas, later to release it. There is no binding or irreversible trapping of the agent. In a highuse facility, these canisters can add unwanted cost and complexity to the operations and would require frequent changes throughout the day. Destroying the anesthetic agent also presents problems by creating toxic liquids or gasses. The best option to date is to either remove gasses using a vacuum system or dilute and remove the gasses using airflow, similar to how noxious gasses are dealt with using a fume hood. Animal use facilities typically require that all room air be completely exhausted to the

Fig. 3.6 Wall-mounted gas anesthesia is easy to use and provides a constant depth of anesthesia. This single cost effective system provides anesthetic gas to multiple induction boxes, a nose cone, imaging chamber, microPET and microCT systems.

Wall mounting keeps valuable countertop space clear. Using a constant pressure system avoids cumbersome flow meters. Potential drawback is only one anesthesia gas concentration for all points of use

35

3 How to Set Up a Small Animal Imaging Unit

outside, with no recirculation of any air. One can utilize this exhaust by ducting the anesthetic gasses into the return air duct, where WAG is highly diluted before being released. Either in-house vacuum or a dedicated vacuum pump can also be used to pull waste gasses away from investigators; however, these systems need to be carefully designed to avoid pulling room air into pathogen-free barrier locations or interfering with delivery of the anesthetic gas to the animals. It is important to consider how WAG can be managed and to have EHS approve of the design before implementation, as what is allowable varies considerably between different locations. Injected anesthetics have often been used in the past and are still widely used for larger species where safe induction requires the animal to be asleep before it is safe to use gas anesthesia, as in work with primates. Injected anesthetics are frequently used in a combination of both an immobilizing and dissociative agent, such as ketamine and midazolam. Pentobarbital has often been used; however, it typically lasts much longer than imaging experiments require and can easily lead to overdose and death if not correctly administered. The primary advantage to injected anesthetics is that the equipment required is simple: a needle and syringe along with a bottle of anesthetic agent. The drawback is that these agents are not easily reversed in the event of an overdose, nor is it easy to titrate the dose. Stress from poor handling can cause adrenaline to temporarily override the anesthetic effect, leading investigators to sometimes inject a second dose, which can often lead to death by overdose. Animals are constantly metabolizing or excreting the anesthetic agent; thus, unless constantly infused, there is a variable depth of anesthesia, which may have unwanted effects on the data. The recovery times for injected anesthetics are also much longer than inhalant anesthetics. Injectable agents are also drugs of abuse, requiring a prescription, drug logs, and double lock and key storage, complicating the experimental procedure. There are some imaging systems capable of imaging live awake animals, using video and postprocessing software to remove movement from the data. Conscious imaging has many advantages, primarily the ability to look at “normal” activity, either behavior or biochemical. There is some question as to what constitutes normal, since a laboratory environment and handling techniques could induce stress and other unwanted changes in animal metabolism.

Acclimatization and good training of personnel handling the animals can make a big difference in creating a stable reproducible environment for conscious experiment. Other reasons to use conscious imaging include the need to measure normal blood pressure, behavior, and heart rates or to avoid other confounding effects of anesthesia on animal physiology. Regardless of what type of anesthetic agent is used, heating the animals is essential (Suckow et al. 2008). Especially with mice, where the mass is tiny, these animals will quickly become the temperature of the surrounding environment and likely die from hypothermia if not properly warmed. Even mild hypothermia can cause unwanted changes, such as the brown fat activation easily seen with FDG imaging using PET (Baba et al. 2007). Heating ideally starts immediately when the animals enter the imaging center, even before anesthesia induction. All places where animals are kept should be kept warm, including cage storage spaces, induction chambers, surgical areas, injection locations, imaging chambers, and recovery locations. The choice of anesthetic agent and when to deliver it depends on the specific experiment. In some cases, it might be best to inject imaging agents while animals are conscious, and then wait for uptake and nonspecific clearance before inducing anesthesia and imaging. This conscious injection method allows the use of conscious physiology, while preserving the use of anesthesia to keep animals immobile for imaging. There can be complications using this method, such as increased muscle uptake or activation of brain areas due to exposure to light and sound. Anesthetic agents have different effects on physiology, so the best agent to use may depend on what physiological information is under investigation. Other factors, such as circadian rhythm (Collaco et al. 2005), position of the signal within the animal (Virostko et al. 2004), and fasting state (Suckow et al. 2008), can alter the physiology and resulting imaging probe distribution.

3.1.4 Radiation Use The use of ionizing radiation requires training, licensing, and application of the radiation safety principle of ALARA: as low as reasonably achievable. X-ray and CT systems produce ionizing radiation; however, the

36

energies are low compared to PET, and the radiation is not present when the source is not turned on. Shielding is fairly simple with only a few millimeters of lead required; thus, these kinds of systems are typically self-shielded and pose little or no exposure risk to nearby personnel. It is worth pointing out that animals injected with SPECT or PET isotopes placed inside these systems may still create measurable exposure rates, since the energies of these isotopes are higher than the X-ray energies. SPECT and PET imaging use radioactive isotopes that create ionizing radiation, with half lives ranging from minutes to days. The principles of time, distance, and shielding apply to using these isotopes, so consideration of all the handling and storage aspects from delivery to final disposal is necessary. Isotopes may be delivered ready for use or may require radiochemistry to prepare the desired radiolabeled compound. If the center is to support radiochemistry, a fume hood and lead shielding may be required within a dedicated room for handling moderate amounts of activity. Metabolically interesting molecules often can contain iodine atoms, and iodine isotopes are readily available for either PET or SPECT imaging. To prevent exposure to personnel and unwanted uptake in the thyroid, usually a special insert chamber with a charcoal filter is required for any radioactive work using iodine. Once prepared, or if delivered ready to use, radioactive compounds need to be safely stored in a shielded area and dispensed using a dose calibrator. A lead shielding configuration is often used in combination with an L-shaped shield and lead glass window for dose dispensing. Injections into mice are limited to a small volume, usually less than 250 mL, so very small amounts of liquid are handled. A 1 or 0.5-mL tuberculin syringe that has a fixed needle is ideal for this work, since there is very little wasted dead volume compared to a needle with a hub attached to a syringe using a Luer fitting. A dose carrier, either lead or tungsten, can be used to transport the dose to the injection area. Once injected, the syringe can be assayed for residual activity and disposed of in a shielded sharps container. The vial or syringe with the bulk dose can be held until no longer radioactive before disposal. The injection site varies depending on the species, probe, and experimental design. Larger animals are often injected into paw or tail vessels; rabbits have easy access via the ear vein; rats and mice can be injected via intraperitoneal (ip), tail vein (iv) or using surgically implanted catheters. For mice, most often

D. Stout

either ip or tail vein injections are used. Using a large and heavy shielding cover on the syringe makes injections nearly impossible, so usually the syringe is used quickly with no shielding for injections. An experienced person can inject either ip or iv in only a matter of a few seconds; thus, radiation protection is accomplished using the shortest possible exposure time. Another alternative is to first place a catheter in the blood vessel, reducing the radioactivity handling time to a minimum of just making the tubing connection and injection. Once injected, the animals are radioactive and need to be kept away from personnel if possible. During imaging, typically shielding is not an option, so the imaging systems ideally should be located away from where people are likely to sit or stand. After imaging, animals can be put back into cages, kept in a shielded location. Once the experiment is over, the animal, cage, bedding material, and any absorbent padding used for imaging along with any urine or feces should be considered radioactive and disposed of accordingly. For short-lived isotopes, the trash can be held overnight or until surveyed and verified as no longer radioactive before disposal. If longer-lived isotopes are planned for use, consideration of where wastes and cages can be stored is necessary. Storage of radioactive animals may need to be considered, as vivarium personnel are often not trained as radioactivity workers nor is it wise to expose other animals or personnel in the vivarium to radioactivity. Fortunately, the amount of radioactivity required for imaging and the often short half-life mean that there is very little activity in most animals when they are returned to the vivarium. For PET experiments, activity levels are low enough that radioactive animals can be stored in a dedicated radioactive cage rack located within the main vivarium room, or perhaps in a separate room. One option is to make the investigators responsible for cage care until the isotope has decayed to background. The specific protocol for handling activity and animals depends on the isotope and half-life involved. The radiation dose to the animals may be of concern, especially with imaging of xenograft tumors. Frequent SPECT or PET imaging and high-resolution CT imaging can lead to substantial radiation exposure (Taschereau and Chatziioannou 2007; Taschereau et al. 2006), which may alter tumor kinetics (Pan et al. 2008). Care should be taken to ensure the radioactivity used is safe and both reasonable for the imaging system characteristics and

37

3 How to Set Up a Small Animal Imaging Unit

for the animal. A low-resolution, low-dose CT may be all that is necessary for anatomical localization with PET or SPECT imaging.

3.1.5 Biosafety Biosafety concerns play a paramount role in the design of an imaging center. Typically, the requirements will dictate what can and cannot be done in a center and also may require certain physical and procedural conditions that may constrain how and what experiments are conducted. There may be restrictions about mixing different biosafety level animals within the vivarium or even within the imaging center, so it is vital that these issues be explored before a plan is created. Requirements vary widely and change with time, so consultation with biosafety personnel is essential. Biologically interesting events that initiate immune response, cancer progression, or even administration of carcinogenic or biohazardous substances usually require a biocontainment environment. There are four biosafety levels (BSL) used to define the severity of potential problems, ranging from I (no hazard) to IV (severe hazard). In the past, most preclinical work was considered to be of little or no hazard; however, that is no longer true. The need for biosafety protection has grown due to the use of viral agents for genetic manipulations and the widespread use of immune-compromised animals that require protection from environmental pathogens. If one wants to study a communicable disease such as tuberculosis, then there are serious steps required to protect both personnel and the public from inadvertent exposure. The advent of immune-compromised animals capable of sustaining implanted human tumor xenografts has greatly aided our understanding of cancer; however, these tumors represent a risk of exposure to personnel since they may harbor hazardous agents such as viruses or bacteria. For this reason, xenograft work is now considered as BSL II, which requires protection of personnel using gloves and disposable gowns and proper biohazardous trash disposal of anything coming into contact with the animals. There is an increasing need and interest to study disease processes and communicable airborne agents that require BSL III conditions; however, these represent a much more difficult challenge to study, since

any equipment used for this type of work must be sterilized between use. Typically, imaging systems are not designed for this environment or the sterilization process. A better option may be to isolate the animals into sealed chambers and move them out of the BSL III zone for imaging.

3.1.6 Infrastructure Requirements Imaging systems require a variety of support systems for efficient use. These range from personnel for operations, training, and supply stocking to counter space, anesthesia systems, and surgical areas to radiochemistry labs, computer networks, database software, archival systems, and websites for scheduling and retrieving data. Often multiple imaging systems are grouped together as a shared resource or core facility that is used by multiple investigators; thus, much of the infrastructure can be shared by the various systems. Lack of comprehensive planning during the design phase can result in budget problems, overtaxed IT support, and other problems with efficient and optimal use of the imaging center.

3.1.6.1 Personnel Staffing support will most likely be the greatest budget cost after the center is equipped and operational. There are usually two types of personnel involved with imaging centers: dedicated staff and peripheral support personnel. Dedicated staff are those who work in the center full time as the primary support people for experiments. They typically operate equipment, conduct routine quality control measurements, order and stock supplies, and ensure that everything is ready for any experiments. Peripheral staff are those who may assist part time or who are part of a different group, such as the information technology (IT) computer support, veterinary staff, administrative support, purchasing, and building facilities maintenance people. These people perform vital functions, yet their role may not be very obvious or may not be frequently required. Sometimes, these peripheral people may be supported by departmental, institutional, or overhead budgets; other times, it may be necessary to include them as part of the budget plan.

38

D. Stout

The number and type of people working within or as part of the center vary widely with different equipment and between established imaging centers. Some systems lend themselves to easy use by investigators with little or no oversight, such as optical, microtomes, and autoradiography systems. Other systems such as PET, MR, and ultrasound may operate best with experienced personnel who can ensure the proper use and have the experience and skills necessary to create reproducible and optimal imaging results. In some locations, only trained personnel can operate devices that use or produce radiation, which is where dedicated staff are useful, especially in a center used by multiple people and projects. Data analysis may be part of the center support through dedicated personnel, or this function may rest in the hands of the investigators. If not part of the center support, then training about how to work with and quantify image data is needed. A range of support can be found in existing centers. One option is a hands-off low cost approach, where systems are available for use with very minimal staff support; mainly just keeping everything working and supplies stocked. The other end of the spectrum is full service center, which may help with the design of the experiment and conducts all the work, analyzes the data, and produces a final set of measurements. Obviously the full service route requires more people and cost, with the benefit of more consistent results. Most common is some sort of middle ground level of support, where dedicated personnel are present to help with experiments, but the investigators are still very involved with acquiring and processing of data. Any of these support levels can work well, though good training and educated users are essential when there is less staffing support.

Fig. 3.7 Computer infrastructure layout for a multisystem center using a central network attached storage (NAS) with a redundant drive array (RAID), computer cluster, online archival storage location and a public access website. The website can only copy data from the secured online archive into a network available folder. Center systems are kept on a private subnet to manage network traffic and for added security

3.1.6.2 Computer Infrastructure The computer infrastructure consisting of wiring, switches, network drives, and perhaps a computer cluster for image processing is literally the backbone upon which the center relies for data movement, archiving, and retrieval of information. Creating images and any associated data requires entering in formation into a software program and database, computational processing either on the host computer associated with the imaging system or on a remote computer, followed by file transfers. Automating this process is often well worth the minimal cost, to reduce staffing requirements and the potential for mistakes (Taschereau et al. 2008). Archiving data first requires making sure all the necessary information has been entered and the appropriate files, images, or records are suitably created and configured, followed by moving the data to a storage location and clearing space on the imaging system computer. There are several data retrieval methods, including using portable hard drives, burning DVD’s, USB data sticks, or file transfers through a website interaction. These interactions can be manual (person to person) or automated through the use of an online database (Fig. 3.7). Data and computer security are serious concerns. It may be advisable to operate using private subnets for the imaging center, to be less exposed to outside hacking attacks and less subject to data traffic flow restrictions from other activities. Virus software and other IT support activities, such as software updates, usually must be restricted to prevent interference with data collection activities. IT services can be scheduled and run at predetermined times when they will not interfere with experiments. Regular disk defragmentation is usually a

Database & Display Computers

Multiple Imaging Systems

Host Computers

NAS RAID

Cluster

Public Web Server

Online Archive

Image Retrieval Public FTP Site

DVD Tape

Copy data to Investigator computer

Private Subnet

Public Access

39

3 How to Set Up a Small Animal Imaging Unit

good idea and is best done after data have been cleared to archiving and time permits. Some imaging devices can both acquire and process data at the same time. This may cause disk access problems if too many processes are attempted at the same time. For this reason, and for system security, it is best not to use the host computer attached to the imaging system for any web browsing or other activities. A separate computer for database and other uses can be used to minimize any disruptions to the imaging process. Extensive processing required for some imaging modalities may make use of dedicated hardware, computer clusters, or other computers nearby through distributed processing software. In some cases, the resulting improvement in image resolution is well worth the additional computation cost in order to see small structures or changes (Qi et al. 1998).

3.1.6.3 Scheduling, Database and Usage Tracking The best method for scheduling usage depends on the systems involved and the volume of studies planned for center operations. Some systems such as optical, microtome, surgical areas, and procedure rooms are well suited to using an online calendar for reserving

time. Other systems that require staffing or more complex support, such as radiochemistry and cyclotron time, may be best suited to email-based requests followed by scheduling meetings of the personnel involved. Whichever scheduling method is used, the amount of time and resources required needs to be coordinated with supplies, support, and other related activities. The scheduling process is also a good time to ensure there are proper protocols and approvals in place and that any information needed for usage and billing reports is collected. The online calendar approach enables people to easily see when systems are available for use. For experiments where staffing, isotope, and radiochemistry support or coordination with other imaging systems (PET-CT) are necessary, a schedule where various demands upon the resources are sorted out needs to be created and sent to the people involved. Often this is done on a weekly basis for most imaging centers, so that demands of both clinical and preclinical centers are met and investigators are given sufficient notice about when they can conduct their experiments. A shared resource where multiple users and perhaps outside contract work is conducted needs to have a way to prioritize projects, based in part on how flexible the various experiments are to moving imaging times or dates (Fig. 3.8).

Fig. 3.8 Scheduling website (top) and data retrieval and search website

40

A well-designed database can make it much easier for investigators to track usage and for the center staff to manage the facility (Truong et al. 2008). Perhaps the most advantageous use of a centralized database is to create unique session identification numbers to both identify a particular experiment and track the usage for billing and usage reports. The archival process can be linked to the database by placing all the information related to a single experiment into a folder named using the session ID number. Any information associated with the experiment can be either entered into the database or added to the folder for archival. Since most grant

D. Stout

funding agencies require both archiving of data and public access to at least some of the data, it is often required to have an archival and data tracking system in place. An added benefit is that an online depository of information is much easier to search to find information, compared to having information either only in someone’s memory or in a hard-to-find laboratory notebook. As staff, students, and postdocs rotate through labs, sometimes, it can be hard for the faculty or lab manager to find older experimental information and to know if something has been tried or worked in the past (Fig. 3.9).

Fig. 3.9 Database software used to generate session identification numbers, track animals, authorizations, experimental information, generate billing and usage reports

3 How to Set Up a Small Animal Imaging Unit

In addition to usage and archival support, a database can be used to track information not otherwise collected by vendor software. While some systems have simple requirements and perhaps software that collects any necessary data, many systems have associated information not directly part of the imaging process. The database might be used to collect information such as metabolite or blood measurements, anesthesia details, physiological monitoring data or investigation specific information related to approval licenses, billing information, and usage statistics. The manager or director of the imaging center will almost certainly be required to know who is using the center and to what extent each investigator is using the facility. There may be regulatory oversight committees that require information reports on biosafety, radiation usage, and animal use. Most likely this person will also write grants, need to create usage and progress reports, and be responsible for billing; thus, an easy-to-use comprehensive database is invaluable for center operations. Along with the computer systems, software will need to be available to investigators to view and analyze the image data. Depending on the vendor, software may be provided that can be freely distributed, or there may be substantial licensing costs. A number of image analysis software programs are also freely available online (OsiriX, AMIDE, ImageJ) that can often read a range of data. Since there is no standardized format for the wide range of preclinical data types, multiple programs may be necessary depending on the imaging system mix within the center. Data analysis may require extensive user training (see below); however, there are tools online that are specifically designed to help with image analysis (Huang et al. 2005).

41

best use. In an imaging center, you want to have the animal holding, preparation, and imaging areas well laid out. Moving animals frequently can induce stress, so it is worthwhile to consider how animals can be brought into the facility and placed in a holding area to minimize the need for further movement. Since animals are commonly transported on carts, a location for the cart is needed and a path suitable for easy transport (Fig. 3.10). Radiation use for PET and SPECT requires a holding area and dose-drawing location and involves the use of lead shielding and somewhat heavy lead dosecarrying holders (pigs). When there are many animals being imaged using short static imaging scans, it may be advantageous to draw multiple doses of radiation at the same time and put them into holders on a cart for transport to the injection location. A path will be needed that is cart friendly, much like that needed for animal entry and exiting from the center. Supplies require both a suitable path for delivery and storage locations. Some items, such as gas tanks, should be located near doors to avoid having to move heavy bulky items near expensive imaging equipment that might be damaged by inadvertent contact. Other supply items will need to be stored nearby in cabinets or shelving for easy access. Items used frequently need to be easily at hand and trash bins located where needed. The ability to store things openly may be restricted by fire codes and biosafety protocols. Trash may need to be segregated based on biosafety and radiation safety concerns. Suitable containers need

3.1.7 Ergonomic Flow A useful approach to center design is to consider the various types of activities and consider their movement or flow throughout the center: primarily people, animals, radiation, supplies, and data. It is fairly obvious that people move throughout the imaging center, and one would like to minimize the need to move back and forth needlessly. This can be compared to good kitchen design, where a work triangle consisting of a sink, stove, and refrigerator are optimally placed for

Fig. 3.10 The biosafety cabinet, PET scanner (right) and CT system (left) are located within easy reach for the investigator. Animals waiting to be injected and imaged are located on a central island on a heating plate. Computers for the imaging systems are located to the left of the CT, where exposure to radiation is lowest

42

to be located close to where the trash is generated. The right side of Fig. 3.6 shows a biosafety cabinet where animals are anesthetized, injected, placed in imaging chambers, and moved into and out of cages. There are both regular and biohazardous trashcans located under the cabinet in easy reach for the investigator. There is also a shielded location nearby for sharps (not shown). Short-lived isotopes can be put into regular or biohazardous trash and held overnight before disposal. Longerlived isotopes need to be kept in separately labeled bags and held for suitable decay time. It may be necessary to have the imaging center janitorial staff educated on what they can and cannot remove, or perhaps not allow trash removal at all except by knowledgeable staff. The data flow is quite important and the least visible type of movement critical to imaging center function. A well-defined plan is needed to move data from the imaging systems to any processing locations and to the archival system. Ideally, the imaging systems begin each day with empty hard drives. This ensures that space will be available for new data and makes it less like data will be lost due to communication failures when drives nearly full are fragmenting the incoming data. One additional consideration is the inclusion of a nonanimal, nonradiation area with a window for visibility into the center. This location can be where staff are allowed food and drink, where image verification for archival or analysis can be conducted, and visitors can see operations without interfering with the process or being exposed to radiation. Often times, the imaging systems can be fairly loud due to cooling fans, so this room provides a respite from the noise and demands of the imaging work. It is good for moral of both staff and investigators who may spend long hours conducting experiments. It also allows a safe way to let people rest while still be close to their animals.

3.1.8 Training and Education Imaging systems are becoming an essential tool for biologists and other scientists to evaluate molecular processes in vivo. As they leave the labs where they were developed, these systems are now more automated and less overseen by physicists and programmers; thus, it becomes a challenge to train personnel and service and operate the systems using staff not familiar with the inner workings of the software and hardware. There are also a wide range of factors where

D. Stout

physics meets biology that must be understood, measured, or controlled for the research to create useful information. These factors include controlling temperature, anesthesia, positioning, contrast agents, injection methods, data collection, and animal handling conditions. Given that the imaging systems are increasingly a tool used by biologists, it is important to recognize that there is a substantial turnover in the people who use these systems, typically staff, students, or postdocs. Education is thus an ongoing requirement, not just a one-time event. Another area where ongoing education is necessary is the area of new and improved imaging systems, data analysis techniques, and experimental methods. New imaging systems are still being created, and older ones improved through advancements in technology. Careful consideration of these new systems and potential replacement systems is needed given the disruption and cost of adding or changing systems in an operational imaging center. Data analysis tools and optimal methods for conducting research are still very active areas of research, thus to stay current in the field, someone needs to evaluate and teach these methods to ensure a uniform application within the center. In theory, all investigators can be taught how to operate imaging systems, and there would be no need for oversight and management of the systems. There is no inherent reason why this cannot be done, however, in practice; this method often proves unfeasible for a number of reasons. First is that without common oversight, regulatory compliance becomes exceedingly difficult to monitor, and accidents with radiation or biohazardous materials can more readily occur. It also becomes harder to ensure that established standard operating procedures (SOP) for data collection and processing are followed. The cost of the equipment and supporting environment usually means that an imaging center is established and used by numerous investigators, which can lead to conflicts and a messy environment when people do not properly clean up after themselves. Inclusion of a manager and/or staff to the center creates the ability to more effectively operate the systems and verify compliance. This leads to two types of education required for operations; training for the staff and education for the users.

3.1.8.1 Staff Training In-depth training is necessary for staff operating imaging systems compared to users who may not need to

43

3 How to Set Up a Small Animal Imaging Unit

know the inner workings of the devices. Staff need to perform routine calibrations, quality control measurements and may be required to repair or troubleshoot systems when problems crop up. If staff members frequently operate the imaging systems, then they are ideally suited to spotting any problems and ensuring any database and usage information are properly collected. Staff should be knowledgeable about any common or potential errors in image generation and possible solutions. Ideally, the staff members have the knowledge to carry out all aspects of the imaging experiment, thus are available to users who need training or help with their experiments. They may even actively participate in experiments by helping with injections, animal care, and even data analysis. Ironically, as imaging systems become more automated and less prone to problems, it becomes more difficult to develop the skills and knowledge needed to diagnose and repair these systems. While reliability and ease of use are certainly good things, they mean that there is less opportunity to learn about repair and problems that might be found in images. A good training program with examples of problems therefore becomes essential. Some problems can be intentionally created for teaching purposes, such as no normalization in a PET scan, center of rotation artifacts in SPECT or CT scans, and metal objects in MR and CT images. Other examples of problems may be available from vendors or other sites that have similar equipment. If available, physicists with expertise in the field are often the best resource for training and troubleshooting help. Often systems are kept under service contracts for repair; however, the time spent waiting for service may cause the experiment already in process to be lost, as biology cannot be put on hold. There may have been months of preparation and considerable time and effort put into animals, along with a time course of data collection needed at specific time points. If the problem is simple or readily solved over the phone, staff members may be able to prevent the loss of valuable data, time, and resources.

3.1.8.2 User Education In general, it is always useful to have very knowledgeable users of imaging equipment who can perform all the necessary steps to acquire and process data. There is however a distinction to be made between what is necessary to know to do the experiment properly, and what

is beyond necessary and perhaps not relevant to the task at hand. The process of learning the skills required to create the animal models and experimental details being investigated is not trivial; thus, the users are already burdened with a great deal of learning and concern for the experiment at hand and may not be interested in knowing more than what is necessary to acquire image data. A brief overview of the imaging modality and the underlying principles of how signals are generated and measured can be quite useful, whereas a comprehensive explanation of the hardware or software is usually not of much value to the end user. There are a number of essential things that users must learn and adhere to with any imaging work, primarily related to animal handling and injections of imaging agents. It is imperative that animals be anesthetized, heated, injected, positioned, and handled correctly in order to collect useful data. The precise details are usually specific to the experiment; however, there are certain aspects that are typically the same for all experiments and are suited for creation of an SOP manual. Often, the anesthesia, positioning, and heating methods are uniformly applied, as well as injection methodology. Timing, amount, and route of injection may vary but are typically not widely variable. Of particular note is the proper training of users to handle animals in a practiced manner. This helps lower the stress to the animals and can be critically important if injected anesthetics are used. Animals highly stressed by poor handling can die or override anesthesia with high adrenaline levels, only to die when injected with additional anesthetics. For this and other reasons related to ease of use, gas anesthesia is highly recommended where experimentally appropriate. Users of the imaging center often are the people who analyze the data, so there must also be training in the proper techniques to create meaningful information. Image analysis can be simple or complex, depending on the system and level of quantitation desired. Most important is to train people to make appropriate measurements and to do so consistently.

3.1.9 Security Preclinical imaging is a euphemism for animal imaging, a name used both to make the title less controversial and to emphasize that the research is aimed at producing scientific discoveries with clinical relevance. Unfortunately,

44

some people ignorant of the value of research conducted using animals have become violent and dangerous. This means a restricted access environment is required for the safety and security of both animals and people working with animals. Simply locking the doors is insufficient, as people can come in behind careless people, doors may be left propped open or the key, password, or access code might fall into the wrong hands. Ideally, the animal use areas are restricted to only those people who need to be there, with the use of photo identification badges that can be used as access cards to open doors. This is more secure than the use of passwords or keys, and makes it easier to turn off individual access once a person no longer needs to be in the area. A video surveillance system is nearly essential, in part to ensure people using the center are doing so properly, and to ensure safety and accountability in the event of an accident or dispute. There may be times when the biology under investigation requires imaging at certain times not conforming to staffed hours of operation, so a video system enables usage of the facility with at least the option to review what was done if necessary. Access control can be particularly important for biosafety reasons and for high-use centers that may have hundreds of authorized users.

3.1.10 Budgeting No description of an imaging center would be complete without a budget plan. The primary costs involved are personnel, service contracts, supplies, and depreciation. Staffing support depends on both the number of people and the level of expertise required. The skills needed for imaging research may come from hiring experienced staff or more likely will be created through trial and error coming from learning on the job. The greatest benefit may come from having good veterinary support, since the health and well-being of the animals play a crucial role in all imaging experiments. Service contracts often come with choices for the cost and level or timing of support. Centers with in-house expertise may be happy with minimal support, whereas others may want a fast response or more hands-on help. Supply costs are fairly predictable with time and experience. The costs of imaging systems are substantial, and replacement expenses are a large budget item best

D. Stout

suited to parsing out over time using a depreciation schedule. This is not always possible with certain funding sources, so care is needed in applying any charges towards depreciation. For many institutions, a shared resource instrumentation grant is the best or only option to replacing or acquiring new instruments. This is not ideal, since if the grant is not funded, then the equipment cannot be purchased. A depreciation plan enables equipment purchases to be less dependent on grants and more suitable to expected replacement timing; however, the costs involved may be more than users can afford. If all the costs associated with the imaging center are factored into any pricing for research, the amount may well be outside of what investigators can afford, especially for modalities requiring extensive support such as PET imaging. Some centers operate on governmental or institutional support and do not charge their users; however, this is rare and normally some sort of recharge pricing is used to recover some of the costs. The budget may be in part supplied from imaging center grants, core facilities budgets on research grants, use of clinical income, or perhaps contract work with outside companies. With institutional support and due to restrictions on some grants, a subsidized group rate is often needed in addition to nonsubsidized and outside user rates. Many institutions require a detailed budget and sales and service contract in order to charge for services. The creation and maintenance of the documentation and budget items can be taxing and require the help of administrative personnel. Invariably, someone will be responsible for oversight of the imaging center. This will require a substantial commitment of time and effort towards budget, training, and reporting issues. Grant writing to describe the center and progress reports on usage are common requirements. Depending on how the data analysis will be handled, there may be a significant need to train and educate people who are new to using the center, either through teaching courses, seminars, or interactive workshops. These important activities are sometimes delegated to junior faculty, who are already challenged with getting their own research plans going and trying to establish tenure. The person in charge of the center will need to have a good understanding of the imaging systems, how to design experiments, budgeting and business skills and be good at educating new users.

45

3 How to Set Up a Small Animal Imaging Unit

3.1.11 Fire and Safety Issues The American Disabilities Act (ADA) requires that spaces be wheelchair accessible in the United States, which means wide doors and specific heights and fixtures that must be used. There are also fire safety codes that require rooms over a certain size to have more than one means of egress. These two items can dominate the floor plan determination of an imaging center. The addition of fire sprinklers, eye washes, emergency showers, fire extinguishers, and alarm boxes all have to be made part of any plan. Wide and recessed doorways secondary doors can take up considerable space where equipment cannot be located. Access may become a critical issue, especially if there are gowning chambers required to enter a space. Fire and EHS regulations require a hazardous materials diamond sign, signifying health, fire and reactivity labeling covering any materials located within each room. A list of any agents is usually required to be maintained, and limits are often placed on what can be used, depending on the room designation (office, lab, vivarium, etc.). Laboratory space is built to certain standards for fire and safety, which is the type of room most suitable for an animal use imaging center. Due to both the animal use requiring routine disinfection and the use of radiation, all materials used in flooring, wall coverings, and furniture need to be nonabsorbent materials. Cloth or fabric materials are not suitable, especially for chairs, as these can absorb and retain spills or aerosolized agents. Tile floors are also not advised, as the cracks begin to separate and accumulate material as they age. An epoxy or sealed vinyl floor is the best option. Routine cleaning is needed for all surfaces, especially within the biosafety cabinet. Janitorial support may need to be trained in proper cleaning methods, and specific timing may be necessary to avoid disrupting experiments. In the event of a fire alarm or other declared emergency, a plan should be considered for what to do about animals under anesthesia. Usually, personnel are required to evacuate immediately and are not allowed back into the building for minutes to hours. Unfortunately, false alarms are all too frequent; however, they must be taken seriously, and leaving people behind is not usually an option. Personnel may want to consider having a plan to take animals out of chambers and anesthesia boxes, quickly putting them into cages. Leaving animals behind under anesthesia may result in their loss if prolonged

time is required to clear the problem. The use of oxygen as an anesthetic carrier gas may also be a problem if there is in fact a fire.

3.2 Summary Defining the types of research to be conducted within a small animal imaging center enables the selection of appropriate equipment and support requirements. The use of animals requires comprehensive planning to house, handle, and image animals in a safe and optimal environment. This includes the appropriate level of biosafety equipment and containment areas, good design for handling and storage of radiation, magnetic field safety, good lighting, power, space, and support equipment for the imaging systems, as well as anesthesia and heating methods. The building, staff, and computer infrastructure can be sized and configured to match the imaging systems, necessary space requirements, and level of support to be provided, given the resources and budget available to the center. A knowledgeable person to oversee operations and to help with training, grants, and budget issues is an important necessity. Data, personnel, and animal security require systems be in place to ensure safety if or when problems might arise. Careful planning for the center layout, some space set aside for future expansion, and planning ahead on how information will be handled and disseminated will help make the center function well into the future. Automation of routine processes, the use of online databases, and scheduling software will reduce staffing requirements and save costs while at the same time making operations easy and efficient. A wide range of expertise is needed to design and build a small animal imaging center; thus, a team of people with various skills is needed to create the center and make sure the functions required are put into place and supported.

References Baba S, Engles JM, Huso DL, Ishimori T, Wahl RL (2007) Comparison of uptake of multiple clinical radiotracers into brown adipose tissue under cold-stimulated and nonstimulated conditions. J Nucl Med 48:1715–1723 Collaco AM, Rahman S, Dougherty EJ, Williams BB, Geusz ME (2005) Circadian regulation of a viral gene promoter in

46 live transgenic mice expressing firefly luciferase. Mol Imaging Biol 7:342–350 Flores JE, McFarland LM, Vanderbilt A, Ogasawara AK, Williams SP (2008) The effects of anesthetic agent and carrier gas on blood glucose and tissue uptake in mice undergoing dynamic FDG-PET imaging: sevoflurane and isoflurane compared in air and in oxygen. Mol Imaging Biol 10:192–200 Fueger BJ, Czernin J, Hildebrandt I, Tran C, Halpern BS, Stout DB, Phelps ME, Weber WA (2006) Impact of animal handling on the results of FDG-PET studies in mice. J Nucl Med 47(6):999–1006 Huang SC, Truong D, Wu HM, Chatziioannou AF, Shao W, Wu AM, Phelps ME (2005) An internet-based kinetic imaging system (KIS) for microPET. Mol Imaging Biol 7:330–341 Klaunberg BA, Davis JA (2008) Considerations for laboratory animal imaging center design and setup. ILAR J 49(1):4–16 Pan MH, Huang SC, Liao YP, Schaue D, Wang CC, Stout DB, Barrio JR, McBride WH (2008) FLT-PET imaging of radiation responses in murine tumors. Mol Imaging Biol 10:325–334 Qi J, Leahy RM, Cherry SR, Chatziioannou A, Farquhar TH (1998) High-resolution 3D Bayesian image reconstruction using the microPET small-animal scanner. Phys Med Biol 43:1001–1013 Stout D, Zaidi H (2009) PET preclinical multimodality imaging in vivo. PET Clin 3(3):251–273

D. Stout Stout DB, Chatziioannou A, Lawson T, Silverman R, Phelps ME (2005) Small animal imaging center design: the Crump Institute at UCLA. Mol Imaging Biol 1:1–10 Suckow CE, Kuntner C, Chow PL, Silverman RW, Chatziioannou AF, Stout DB (2008) Multimodality rodent imaging chambers for use under barrier conditions with gas anesthesia. Mol Imaging Biol 10:1007 Taschereau R, Chatziioannou AF (2007) Monte Carlo simulations of absorbed dose in a mouse phantom from 18-fluorine compounds. Med Phys 34(3):1026–36 Taschereau R, Chow PL, Chatziioannou AF (2006) Monte Carlo simulations of dose from microCT imaging procedures in a realistic mouse phantom. Med Phys 33(1): 216–224 Taschereau R, Stout DB, Chatziioannou AF (2008) Preclinical imaging facility data flow integration and automation. Society Nuclear Medicine, New Orleans Truong DC, Chatziioannou AF, Stout DB, Czernin J, Phelps ME, Huang SC (2008) A data portal for molecular imaging. Society Nuclear Medicine, New Orleans Virostko J, Chen Z, Fowler M, Poffenberger G, Powers AC, Jansen ED (2004) Factors influencing quantification of in vivo bioluminescence imaging: application to assessment of pancreatic islet transplants. Mol Imaging 3(4): 333–342

4

Noninvasive Small Rodent Imaging: Significance for the 3R Principles Nicolau Beckmann and Peter Maier

4.1 Introduction

Contents 4.1 Introduction ........................................................... 47 4.2 The Contributions of Imaging to the 3R’s . ......... 4.2.1 Reduction . ............................................................... 4.2.2 Refinement . ............................................................. 4.2.3 Replacement ............................................................

48 48 49 49

4.3 Experimental Lung Research in Small Rodents: Nonimaging Techniques . ...................... 49 4.4 Imaging in Experimental Lung Research: Basic Considerations for Animal Welfare ........... 50 4.5 Magnetic Resonance Imaging in Experimental Lung Research: Impact on the 3Rs Leads to More Relevant Data ............................... 4.5.1 Technical Requirements .......................................... 4.5.2 Detection and Quantification of the Inflammatory Response: Allergen-Induced Lung Inflammation .................................................. 4.5.3 Drug Effects are Detected Earlier Using MRI . ....... 4.5.4 Structural Changes are Detectable by MRI: Elastase-Induced Emphysema.................................. 4.5.5 Functional Information can be Derived Noninvasively...........................................................

51 51 51 52 53 54

4.6 Final Remarks . ...................................................... 54 References ........................................................................... 56

N. Beckmann (*) Novartis Institutes for BioMedical Research, Global Imaging Group, 4056 Basel, Switzerland e-mail: [email protected] P. Maier 3R Research Foundation, 3110 Muensingen, Switzerland

Small rodents (mice and rats) are often used in biomedical research and during drug development. Accordingly, the development and use of methods which improve the welfare and reduce the number of animals including the reduction of pain and stress in small rodents during experiments is a challenge for reaping scientific, humanitarian, and economic benefits. At the same time, the quality or the readout of the scientific experiment should not be compromised. In the book, The Principles of Humane Experimental Technique (Russell and Burch 1959), William Russell and Rex Burch stated in 1959 their 3R concept (replacement, reduction, and refinement) in relation to the humane treatment of experimental animals. The authors set the frame in the search for alternatives and changes of experimental procedures that could result in (1) the replacement of animals, (2) a reduction in the number used, or (3) a refinement of techniques that alleviate potential pain and distress of the animals. Replacement often means the use of an inanimate system as an alternative (e.g., a computer model or program, a mannequin). It can also mean the replacement of sentient animals (usually vertebrates) with less sentient animals (usually invertebrates such as flies, worms, bacteria, etc.). It includes as well the use of cell and tissue cultures, whenever possible of human origin (Guillouzo and Guguen-Guillouzo 2008; Spiel mann et al. 2008). Reduction means an appropriate experimental design and an improved data analysis leading to less animals being used in the experiments without loss of useful information. This may be achieved by reducing the number of variables through good experimental design, by using genetically homogeneous animals

F. Kiessling and B.J. Pichler (eds.), Small Animal Imaging, DOI: 10.1007/978-3-642-12945-2_4, © Springer-Verlag Berlin Heidelberg 2011

47

48

(inbred strains), or by ensuring that the conditions of the experiment are rigorously controlled. Furthermore, the use of in vitro data might give hints for improving the design of the experiments. Refinement means a change in some aspects of the experiment leading to a minimization of pain and distress experienced by the animals. Refinement should result in an overall increase of animal welfare during experimentation and is part of good science. Often, refinement leads to a partial reduction or replacement of animals. Examples include the establishment of humane endpoints for intervention in a study using earlier endpoints derived from noninvasive methods or the use of effective preemptive, peri-, or postoperative analgesic protocols. Today, the 3Rs concept has become the generally accepted scientific basis of institutions serving the development of alternatives to animal experiments. Governmental institutions such as the European Centre for the Validation of Alternative Methods (ECVAM, http://ecvam.jrc.it/, Ispra, Italy), the Interagency Center for the Evaluation of Alternative Toxicological Methods (NICEATM/ICCVAM, http://iccvam.niehs.nih.gov/, Research Triangle Park, North Carolina, US), the UK National Centre for the Replacement, Refinement and Reduction of Animals in Research (NC3Rs, http:// www.nc3rs.org.uk, London, UK), or the Zentralstelle zur Erfassung und Bewertung von Ersatz- und Ergänzungsmethoden zum Tierversuch (ZEBET, http:// www.bfr.bund.de/cd/1433, Berlin, Germany) and charitable trusts or scientific foundations such as the Fund for the Replacement of Animals in Medical Experiments (FRAME, http://www.frame.org.uk, UK) or the Swiss 3R Research Foundation (http://www.foschung3r.ch, Münsingen, Switzerland) actively support the development, validation, and acceptance of methods which could replace, refine, or reduce the use of laboratory animals. Of note, the 3Rs concept has been put forward in Europe in the field of regulatory toxicology and by the EPAA initiative (http://ec.europa.eu/enterprise/ epaa/3_3_research.htm). Some of the assays were successfully introduced and replace animal tests in hazard assessments to identify the toxicological properties of chemicals to which humans or the environment are exposed. In contrast, it is a more challenging task to replace a given animal test that is used for drug safety and drug regulatory purposes. Difficulties to substitute animal experiments with in vitro or in silico methods arise when complex regulatory processes, e.g., of the cardiovascular, metabolic, or neuronal systems, are to be analyzed. This is

N. Beckmann and P. Maier

particularly the case when the disease mechanisms are not well understood. Accordingly, at present, animal experimentation remains central to investigate diseaserelated processes and symptoms. Traditional experiments rely heavily on invasive techniques to monitor changes in blood biochemistry, tissue structure, or function in numerous animal disease models. In some cases, a proportion or all of the animals used during the course of a study are sacrificed for histopathological assessment, with the aim to track the progression or regression of a disease over time, or to determine the levels of efficacy or toxicity evident in specific organs or tissues. However, many of these invasive techniques fail to identify crucial steps and details of how a disease develops or how a substance elicits effects.

4.2 The Contributions of Imaging to the 3R’s Noninvasive imaging techniques provide potential to detect and monitor anatomical, functional, metabolic, or molecular changes within the body of animals with minimal pain, distress, or premature termination. Imaging techniques allow models of diseases, and response to exogenous substances, to be monitored in a temporal and spatial manner, therefore allowing a greater amount of information to be derived from smaller number of animals, which in turn increases the statistical validity of the data by reducing the level of experimental variation (Beckmann 2006; Ripoll et al. 2008; Rudin and Weissleder 2003). Noninvasive imaging allows more informative and humane endpoints to be used and, perhaps most importantly, functional details to be studied in the context of a living animal. In the present contribution the significance of imaging for animal welfare and the 3Rs concept is discussed and underlined with data from experiments actually applied in the drug development process. The basic contributions that in vivo imaging can make to the 3Rs are:

4.2.1 Reduction • Imaging can use each animal as own control, allowing paired comparisons, thereby increasing the statistical power of experiments. • Imaging studies are inherently sequential (only one animal can be scanned at any one time), lending

4 Noninvasive Small Rodent Imaging: Significance for the 3R Principles

themselves to sequential experimental designs, which use fewer animals to achieve the same statistical power as compared to conventional designs. • Use of imaging can improve decisions within pharmaceutical research (Beckmann 2006; Ripoll et al. 2008; Rudin and Weissleder 2003) and lead, for example, to an early closure of programs which would otherwise continue to use animals for safety and efficacy testing, yet ultimately not deliver a useful medicine. One typical example is a rat model of prolactinoma. Pituitary hyperplasia is induced by chronic stimulation with estradiol. The variability in pituitary volume before treatment, however, is large, as reflected in a coefficient of variation (COV) of 80%. In order to detect a 50% volume increase with a statistical significance of p = 0.05, group sizes of 35 animals are required. However, monitoring the relative volume increase in each animal by imaging results in a COV of 12%, and a sample size of n = 4 is sufficient to reach the same level of statistical significance (Rudin et al. 1988).

4.2.2 Refinement • Imaging allows the measurement of discrete morphological/functional changes which substitute clinical, often painful, signs detectable by visual inspection. This is of specific importance when nondiseased animals are used. • When diseased, clinically abnormal animals are used, it is often possible to use milder disease models than with conventional structural and functional assessment protocols. • Noninvasive imaging of animal disease models can provide earlier humane endpoints and rigorous inclusion/exclusion criteria, because of the “online” assessment of the injury progression in longterm disease models.

4.2.3 Replacement • Imaging biomarkers provide better and earlier decisions in human clinical trials in Phase I/II (Beckmann 2006). Increased use of imaging biomarkers has the potential to decrease reliance on animal disease model efficacy data, and eventually,

49

will permit replacement of a proportion of animal experimentation by improved clinical trials (indirect replacement). In the sections to come, attention is devoted to applications of magnetic resonance imaging (MRI) in the context of experimental lung research and drug development in small rodents to further illustrate the value of imaging in animal welfare.

4.3 Experimental Lung Research in Small Rodents: Nonimaging Techniques Respiratory diseases such as asthma and chronic obstructive pulmonary disease (COPD) involve a complex interaction of many different inflammatory and structural cell types, all of which release a multitude of inflammatory mediators. Activated eosinophils are considered to play a major role in asthma because they contribute to epithelial cell damage, bronchial hyperresponsiveness, plasma exudation, and edema of the airway mucosa, in addition to smooth muscle hypertrophy and mucus plugging through the release of enzymes and proteins (Hamid et al. 2003; Hogan et al. 2008). COPD is associated with an abnormal inflammatory response of the lungs to noxious particles and gases and encompasses chronic obstructive bronchitis and obstruction of small airways reflecting mucous hypersecretion. Emphysema, a critical component of COPD, is defined as enlargement of air spaces as a result of destruction of the lung parenchyma, leading to loss of lung elasticity and closure of small airways. Most patients with COPD have all three pathological mechanisms (chronic obstructive bronchitis, emphysema, and mucus plugging) (Barnes 2002; Barnes et al. 2003). Small-rodent models have been established in an attempt to mimic and study specific aspects of human respiratory disease, and to evaluate new drugs (see Canning 2003; Brusselle et al. 2006; Lloyd 2007; Martin and Tamaoka 2006 for reviews). The inflammatory status of the lungs in such models is usually inferred from the analysis of broncho-alveolar lavage (BAL) fluid or is determined histologically. The terminal nature of these procedures precludes repeated assessments in the same animal. Pulmonary function tests in rodents are also important for experimental research of various respiratory disorders (Glaab et al. 2007; Hoymann 2007). The techniques used vary from

50

noninvasive barometric plethysmography in unrestrained (“Penh system”) or in restrained animals (headout plethysmography) to invasive methods requiring anesthesia such as orotracheally intubated animals or the use of tracheotomized and ventilated animals (Glaab et al. 2007; Hoymann 2007). The main concern with Penh approaches is that due to the lack of pressure signal, no actual changes in lung resistance and compliance can be determined. An added complication in restrained animals is the fact that they are stressed during measurements. On the other hand, invasive methods are performed under anesthesia, are normally terminal, and complications can arise following repeated intubation (Glaab et al. 2007; Hoymann 2007). Taking these points together, an ideal examination method should detect and follow the inflammatory response, the site and type of lung damage in living animals as well as altered lung function, and the time and dose response effect of potential drugs in an animal disease model. Imaging has the potential to fulfill these combined requirements and to provide further knowledge which contributes to the understanding of lung biology in intact animals.

4.4 Imaging in Experimental Lung Research: Basic Considerations for Animal Welfare An ample variety of techniques including “micro” X-ray computed tomography (micro-CT), position emission tomography (PET), bioluminescence, fluorescence imaging, MRI, and ultrasound can be used to study various processes such as ventilation, perfusion, pulmonary hypertension, lung inflammation, and gene transfer (Beckmann et al. 2007; Driehuys and Hedlund 2007; Johnson 2007; Schuster et al. 2004). Images from more than one modality can also be fused, allowing structure–function and function–function relationships to be studied on a regional basis. All these techniques offer translational potential from animal to human studies (Schuster et al. 2004). Every technique has its own merits when imaging the lungs of small rodents, and most of the applications published up to now demonstrate that the approaches are basically complementary. Based on the noninvasive nature of imaging, repetitive measurements are a fundamental asset when adopting it in experimental studies.

N. Beckmann and P. Maier

However, taking the animal welfare into account, some practicalities need to be carefully considered when performing studies in small rodents, which may dictate the choice in favor of a certain technique: 1. Dose of radiation when performing repetitive studies. Accuracy of micro-CT images is determined by the X-ray dose given to the animal. Thus, one concern when using micro-CT is the radiation dose, which, despite not being lethal, may be high enough to induce changes in the immune response and other biological pathways, so that experimental outcomes could be affected (Boone et al. 2004). For an ideal scanner, a COV of 1% in the linear attenuation coefficient can be expected for an image with an isotropic voxel size of 135 mm of a mouse exposed to 0.25 Gy. If the same COV is to be achieved in a 65-mm isotropic voxel, a dose of 5.0 Gy would be necessary. 2. Acquisition conditions in repeated measurement experiments should interfere minimally with the physiology and the well-being of the animals can be performed. This becomes crucial in the context of drug testing in models of lung diseases. Unexpected results may occur when disregarding the pathophysiological conditions of the animals. For instance, one needs to carefully consider possible interferences between the pathophysiology of the disease models and lung injury complications that might potentially be caused by mechanical ventilation, especially if this is applied repeatedly. Indeed, it has been reported that mechanical ventilation of healthy rats can cause an increase of neutrophils in BAL fluid, pulmonary edema, and even hypoxemia which may lead to progressive circulatory failure and death (Walder et al. 2005). Consequently, mechanical ventilation should be avoided whenever possible for the longitudinal investigation of lung disease models with expected inflammatory responses. 3. Respiration and cardiac rates may vary following a given insult or stimulus. Thus, it needs to be verified whether cardiac and/or respiratory gating may not influence the acquired data. Whenever possible, nongated acquisitions are preferred. 4. The influence of anesthesia needs to be carefully addressed, especially when performing functional studies. However, repeated anesthesia may be an issue even in anatomical studies, in particular in the context of drug testing. The same holds true in

4 Noninvasive Small Rodent Imaging: Significance for the 3R Principles

case analgesics or sedatives are used after surgical interventions.

4.5 Magnetic Resonance Imaging in Experimental Lung Research: Impact on the 3Rs Leads to More Relevant Data 4.5.1 Technical Requirements The living lung is one of the most challenging organs to image by MRI. Because of the low water density of approximately 20–30%, the proton density is significantly lower than in other tissues. Moreover, due to the difference in magnetic susceptibilities between lung tissue and the air that comprises about 80% of the pulmonary volume, local inhomogeneities in the magnetic field are produced, resulting in very short T2 and T2* relaxation times. Using a gradient-echo sequence, mean T2* values of the order of 500 ms have been measured in vivo at 4.7 T in normal rat and mouse lungs (Beckmann et al. 2001a). In order to detect a signal from lung parenchyma, especially at high fields, nonstandard techniques need to be used to reduce considerably the echo time (TE). Different protocols have been developed to overcome some of the inherent difficulties associated with the detection of parenchymal signal. For rodent imaging, the projection reconstruction method with self-refocused radio-frequency (RF) pulses has been used to achieve TE <300 ms and thus enhance the visibility of lung parenchyma (Gewalt et al. 1993). A further challenge in lung MRI is that cardiac and respiratory movements can cause marked image artifacts. These problems are more evident in small rodents, because of their higher cardiac and respiratory rates. To address this issue, scan-synchronous artificial ventilation has been developed (Hedlund et al. 2000), in which breathing is restricted to the recovery period after data acquisition, thereby minimizing breathing-related motion artifacts. Breathing rate is kept constant and controlled by initiating a breath after the appropriate number of pulse sequence repetitions. In addition, imaging acquisition is triggered by the electrocardiogram. Although the combination of projection reconstruction methods with synchronous ventilation and cardiac

51

gating enables high resolution images to be obtained from the rat lung (Gewalt et al. 1993), the acquisition time is long, amounting to about 30–40 min per image. From the point of view of animal welfare and for drug testing in vivo in animal models of airways diseases, one needs to carefully consider possible interferences between the pathophysiology of the disease models and lung injury complications that might potentially be caused by mechanical ventilation (Peng et al. 2005; Walder et al. 2005). It has been shown that a standard gradient-echo sequence produces sharp proton images from the rat and mouse chest, in which potential artifacts caused by cardiac and respiratory movements are suppressed by image averaging (Beckmann et al. 2001b; Blé et al. 2008), despite the fact that neither respiratory nor cardiac gating is applied. Animals breathe spontaneously during image acquisition. Under the conditions chosen for acquisition (TE of the order of 3 ms), the signal from the lung parenchyma itself is too weak to be detected at 4.7 T. However, the absence of any detectable parenchymal signal in combination with a background devoid of artifacts provides high contrastto-noise ratio for the detection of fluid signals associated with inflammatory processes in rodent models of airways diseases (Beckmann et al. 2007).

4.5.2 Detection and Quantification of the Inflammatory Response: Allergen-Induced Lung Inflammation A characteristic feature of respiratory diseases such as asthma is edema of the airways due to an increase in the permeability of lung microvessels to plasma proteins. The resulting effect is the leakage of fluid from the microvascular circulation into the surrounding tissue. Assessment of this fluid can be important for diagnostic purposes and for planning and guiding treatment, e.g., with drug candidates. In rats actively sensitized to ovalbumin (OVA) and challenged with the antigen, an intense, even, edematous signal has been detected in the lungs 24 h after challenge (Fig. 4.1a). By acquiring 20 images displaced from each other by an amount corresponding to a single slice thickness, the whole thorax of the animal was scanned and the total volume of the fluid signals determined (Beckmann et al. 2001b). The volume of fluid signal

52

a

b

vehicle (n=6) budesonide (n=6)

1.4

MRI signal volume (mI)

Fig. 4.1 Allergen challenge. (a) Axial images through the chest of an actively sensitized BN rat, acquired before (left) and 24 h after intratracheal (i.t.) OVA instillation (right). Note that the lung parenchyma appears dark; however, following allergen, a prominent MRI fluid signal is seen (arrow). This signal is due to edema. The acquisition time for one image was of 1 min. Animal respired spontaneously during image acquisition. (b) Volume of fluid signals detected by MRI in the lungs for an OVA dose of 3 mg/kg (i.t.). Administration of the corticosteroid budesonide (1 mg/kg i.t.) 24 h after OVA (arrow) leads to an acceleration of the resolution of the MRI signals, suggesting a rapid effect of the compound. This effect is not detectable by conventional BAL fluid analysis. Data are expressed as means ± SEM (n = 6 rats per group). For details, see Beckmann et al. (2001b)

N. Beckmann and P. Maier

1.2 1.0 0.8 0.6 0.4 0.2 0.0 –0.2 0 6 12 18 24 30 36 42 48 54 60 66 72 78 84 90 96

Time after OVA (h)

was dependent on the dose of allergen and reached a maximum between 24 and 48 h after challenge. Despite the extensive presence of fluid in the lungs, no abnormal behavior of the rats has been noticed. The MRI signal was highly significantly correlated with a variety of inflammatory parameters determined in the BAL fluid recovered from the same animals (Tigani et al. 2002). Of special interest, the strongest correlations were with the eosinophil numbers, eosinophil peroxidase activity (a marker of eosinophil activation), and the total protein content (a marker for plasma extravasation). Importantly, the signal detected by MRI correlated significantly with the perivascular edema assessed by histology (Tigani et al. 2003). After about 100 h following the challenge, the signal had been completely resolved.

4.5.3 Drug Effects are Detected Earlier Using MRI Following the validation of the MRI signal against the traditional, terminal techniques, BAL fluid analysis

and histology, it can be used to study the effects of anti-inflammatory compounds in the animal model. In one experimental paradigm the drugs were given 24 h after the challenge with OVA, a time point when an extensive MRI signal is present in the rat lung. For instance, treatment with the steroid, budesonide, accelerated the rate of resolution of the MRI signal (Fig. 4.1b). A clear trend toward a reduction in the fluid signal was observed as early as 3 h after drug administration, and the effect was statistically significant from 6 to 72 h. The decline in the edematous signal correlated significantly with the reduction in perivascular edema quantified by histology (Tigani et al. 2003). This suggests that an effect to suppress perivascular edema underlied the decrease in the MRI signal following treatment with budesonide. By contrast, BAL fluid markers of inflammation were not affected at 6 h after treatment (Tigani et al. 2003). Only at 48 h following steroid administration, an effect was detected in the BAL. Let us compare from the perspective of animal welfare the imaging approach with the traditional BAL fluid analysis, which necessitates six rats per time

53

4 Noninvasive Small Rodent Imaging: Significance for the 3R Principles

point. If BAL analysis were to be performed at the same time points as imaging had been carried out, 60 rats would have been necessary in both the compound group and vehicle group, resulting in a total of 120 animals. For the imaging analysis in contrast, only 12 rats were necessary. Moreover, the terminal BAL analysis was not able to detect the rapid effects of the steroid on the established lung inflammation revealed by MRI. Obviously, histology could have been performed; however, besides being time-consuming, only information from a very limited region can be obtained. This example demonstrates the clear advantage provided by imaging in reducing considerably the number of animals used in the study while at the same time refining the experiment.

4.5.4 Structural Changes are Detectable by MRI: Elastase-Induced Emphysema A second example concerns the use of MRI in a rat model of emphysema (Quintana et al. 2006). For animals treated with porcine pancreatic elastase (PPE), the parenchymal signal intensity was decreased in the first 6 weeks following PPE (Fig. 4.2). Consistent with this, extensive enlargement of the alveoli was observed in alveoli rich sections of histological slices. A tendency toward recovery of the MRI signal intensity was apparent at week 8, which correlated with a reduction of the emphysematous damage assessed histologically by point morphometry. Related to this reduction in

0.30 **

0.25

***

***

0.20

***

0.15 0.10

bas

w2

w4 MRI

w6

w8

% Alveolar parenchyma

Parenchymal signal intensity (a.u.)

0.35

60 50 40 30 20 10 0

Fig. 4.2 Elastase-induced emphysema. Coronal MRI sections from a rat at baseline and at several time points following i.t. administration of PPE (75 U/100 g body weight). Animal respired spontaneously during image acquisition. Note the decrease in signal intensity from lung parenchyma as denoted by

bas

w2 w4 w6 Histology

w8

the arrows and the corresponding graph from the mean signal intensity in the lower left side of the lung. Histology from the same anatomical location demonstrated a decrease in the number of alveoli in elastase-treated animals. For details, see Quintana et al. (2006)

54

damage could be the fact that following PPE the elastin content initially decreases, but appreciable elastic fiber deposition, and granulation of the alveolar airspaces containing fibroblasts, endothelial cells, and a provisional collagen matrix are observed weeks after injury. The significant correlation between the MRI signal intensity of lung parenchyma and the percentile alveolar content determined by histology indicates that proton MRI is sensitive to noninvasively detect structural changes of the lung parenchyma related to the development of emphysema in this model (Quintana et al. 2006). However, it is clear that the MRI readouts have to be validated previously against the histopathological detectable structural changes. In this study, six rats have been measured using MRI up to week 8 after elastase. With histology, 30 animals were necessary to characterize the temporal effect of elastase at the same time points in which MRI had been performed. Moreover, whereas MRI provided a spatially resolved measure over the lung, histology did only provide punctual information. Obtaining volumetric information using histology would have been a very difficult task. Finally, histology was timeconsuming, whereas MRI provided information readily after image acquisition. This example shows that MRI again clearly contributed to reduce the numbers of animals used in the study and to refine the experiment.

4.5.5 Functional Information can be Derived Noninvasively The availability of fast gradients makes short TEs possible without the use of nonstandard RF pulses and spatial encoding techniques, thereby avoiding difficulties associated with them. For instance, signal of lung parenchymal tissue from spontaneously breathing rats and mice can be detected at 4.7 T, a field strength commonly encountered in animal scanners, with a reasonable signal-to-noise ratio using a gradient-echo sequence with TE of the order of 500 ms within about 1 min measurement time (Beckmann et al. 2001a). This approach enables one to derive ventilation-related information exploring the weakly paramagnetic character of molecular oxygen. Indeed, a highly significant negative correlation was found between the parenchymal signal in the rat lung and the partial pressure of oxygen in the blood, for different amounts of oxygen administered

N. Beckmann and P. Maier

(Beckmann et al. 2004). Following this reasoning, an increased parenchymal signal should be consistent with a reduced oxygen level or vice-versa. Proton MRI has been used to detect the effects of bronchoconstrictor and bronchodilator compounds in spontaneously breathing rats (Beckmann 2006). For instance, a significant increase of the parenchymal signal intensity was observed in the upper airways, from the first minutes following intravenous administration of a compound eliciting bronchoconstriction (Fig. 4.3). The long-lasting signal increase was reversed by application of a bronchodilator agent, consistent with an increase in oxygenation. Airway resistance measures derived invasively in anesthetized, paralyzed, and artificially ventilated rats showed the same time profile as that of the MRI signal. These observations suggest that the MRI signal changes in the upper airways are due to bronchoconstrictor effects. The refinement provided by the method is evident: while the airway resistance assessments were invasive and terminal, MRI has been performed in anesthetized, spontaneously breathing rats. Repeated assessments should also be feasible using MRI, further contributing to reduce the number of animals used in the experiment.

4.6 Final Remarks The main message of this contribution is that, generally speaking, better science can be achieved when following the 3R principles, e.g., by introducing noninvasive imaging into in vivo experimentation. Using examples from our own experience in adopting MRI in the area of small rodent models of airways diseases, we have shown the win–win situation between animal welfare and the relevance of data obtained from animal studies. We illustrated how noninvasive imaging results in a significant reduction in the number of animals used for experimentation. Depending on the application, a reduction between 80 and 90% is estimated. Since repeated measurements are feasible, each animal can serve as its own control, thereby reducing the variability of the data. Mo reover, as exemplified by the study of anti-inflammatory compounds on established inflammation, imaging provides potential to obtain information not accessible to conventional, post mortem assessments. Such an experimental refinement is often the consequence of the fact that imaging provides information of the organ in situ in

55

4 Noninvasive Small Rodent Imaging: Significance for the 3R Principles

Terminal airway resistance assessment

MRI parenchymal signal Normalized MRI signal

125

B

120

A

115

105

150 0

100 95

0

90 85

Airway resistance [cmH2O/(L.s-1)]

600 450 300

110

15 min

A –5

0

5

10

15

20

25

30

35

30 min

B

40

Time (min)

Fig. 4.3 Effects of a bronchoconstrictive (A) and a bronchodilating compound (B) administered intravenously. For the MRI experiment, the anesthetized rat could respire spontaneously. After positioning the animal in the magnet, images at the same anatomical location were acquired sequentially with a time resolution of 1 min per image. Following the acquisition of baseline images, the bronchoconstrictive agent was injected at time point 0 and the bronchodilator was injected at time point 25 min. The curve on the left shows the signal intensity evaluated in the same region in the upper airways, normalized to the mean signal

intensity in the baseline images. The increase of signal after administration of compound A signal was consistent with its bronchoconstrictive effects, while the decrease around time 25 min was consistent with the bronchodilating effects of compound B. On the right side, the airway resistance measured in another rat is shown, for the same compounds administered. For this terminal experiment, the animal has been tracheotomized and artificially ventilated. Notice the similar profiles of the MRI parenchymal signal and the airways resistance measures. For details, see Beckmann (2006)

the intact organism. Thus, data that are more relevant to address therapeutic effects can be obtained using imaging. Besides therapy testing, in vivo toxicology, encompassing the identification of adverse effects of air pollutants, drugs, or nanoparticles for medical use, will largely profit from imaging (Reid 2006; Wang and Yan 2008). Indeed, noninvasive small rodent imaging is in line with the concept and strategy of toxicity testing in the twenty-first century developed by the US National Academy of Sciences, the US National Academy of Engineering, the Institute of Medicine, and the US National Research Council (http://dels.nas.edu/dels/ rpt_briefs/Toxicity_Testing_final.pdf). Some targeted testing in animals will be needed to sufficiently understand how chemicals are broken down in the body. Combined with imaging techniques, such assays will ensure an adequate toxicological evaluation of chemicals. As an example, it has been recently shown that the presence and biological effects of carbon nanotubes can be detected in vivo by noninvasive MRI techniques (Al Faraj et al. 2009). Some points need to be kept in mind, however, when adopting imaging in animal studies. Validation against established methods: Before imaging is able to make a contribution to address a given question, the readouts need to be carefully validated

against standard, often terminal measures. During this phase, the value of imaging is unknown. However, wellvalidated readouts are likely going to be more relevant and thus better suited to have an impact also on animal welfare. Anesthesia and interaction with readouts: Anesthesia can become a limiting factor of small animal imaging: it may not only influence the results of functional acquisitions, but also interfere with drug studies. Although healthy small rodents easily recover from gas anesthesia, this may represent an additional burden when the animals have been subjected to, for example, a chemical stimulus to elicit a pathology and are in addition treated with drugs. It is recommended to carefully conceive the studies in order to achieve a balance between the gain in information and the number of times an animal is anesthetized. Also, the minimum interval between sequential anesthesias needs to be considered. As a pragmatic way, acquisition times under 30 min and a minimum interval of 3 h between sequential anesthesias are recommended. Also, gas anesthesia is preferable, since it is easier to control and recovery time is only a few minutes. Imaging acquisition guided by physiological requirements: It is important to adapt the imaging acquisitions to the biology of the species and not vice-versa. In the case of lung studies, for instance, it is recommended to

56

perform acquisitions on spontaneously breathing animals, as artificial ventilation may induce lung injury. Overall speaking, it is our experience that imaging can make substantial contributions to the 3Rs concept. As imaging systems are largely available in hospitals for diagnostic purposes, there is potential to address translational aspects from the animal models to the human situation and vice-versa. In this sense, a more humane treatment of experimental animals and more relevant animal studies become feasible. It remains the responsibility of the investigator to conceive the experiments in a way that their nature and performance are continuously refined to ensure that the 3Rs concept is respected in every biomedical activity addressed using imaging.

References Al Faraj A, Cieslar K, Lacroix G, Gaillard S, Canet-Soulas E, Crémillieux Y (2009) In vivo imaging of carbon nanotube biodistribution using magnetic resonance imaging. Nano Lett 9:1023–1027 Barnes PJ (2002) New treatments for COPD. Nat Rev Drug Discov 1:437–446 Barnes PJ, Shapiro SD, Pauwels RA (2003) Chronic obstructive pulmonary disease: molecular and cellular mechanisms. Eur Respir J 22:672–688 Beckmann N (2006) In vivo MR techniques in drug discovery and development. Taylor & Francis, New York Beckmann N, Tigani B, Mazzoni L, Fozard JR (2001a) MRI of lung parenchyma in rats and mice using a gradient-echo sequence. NMR Biomed 14:297–306 Beckmann N, Tigani B, Ekatodramis D, Borer R, Mazzoni L, Fozard JR (2001b) Pulmonary edema induced by allergen challenge in the rat: noninvasive assessment by magnetic resonance imaging. Magn Reson Med 45:88–95 Beckmann N, Cannet C, Zurbruegg S, Rudin M, Tigani B (2004) Proton MRI of lung parenchyma reflects allergen-induced airway remodeling and endotoxin-aroused hyporesponsiveness: a step toward ventilation studies in spontaneously breathing rats. Magn Reson Med 52:258–268 Beckmann N, Cannet C, Karmouty-Quintana H, Tigani B, Zurbruegg S, Blé FX, Crémillieux Y, Trifilieff A (2007) Lung MRI for experimental drug research. Eur J Radiol 64:381–396 Blé FX, Cannet C, Zurbruegg S, Karmouty-Quintana H, Bergmann R, Frossard N, Trifilieff A, Beckmann N (2008) Allergen-induced lung inflammation in actively sensitized mice assessed with MR imaging. Radiology 248:834–843 Boone JM, Velazquez O, Cherry SR (2004) Small-animal X-ray dose from micro-CT. Mol Imaging 3:149–158 Brusselle GG, Bracke KR, Maes T, D’hulst AI, Moerloose KB, Joos GF, Pauwels RA (2006) Murine models of COPD. Pulm Pharmacol Ther 19:155–165

N. Beckmann and P. Maier Canning BJ (2003) Modeling asthma and COPD in animals: a pointless exercise? Curr Opin Pharmacol 3:244–250 Driehuys B, Hedlund LW (2007) Imaging techniques for small animal models of pulmonary disease: MR microscopy. Toxicol Pathol 35:49–58 Gewalt SL, Glover GH, Hedlund LW, Cofer GP, MacFall JR, Johnson GA (1993) MR microscopy of the rat lung using projection reconstruction. Magn Reson Med 29:99–106 Glaab T, Taube C, Braun A, Mitzner W (2007) Invasive and noninvasive methods for studying pulmonary function in mice. Respir Res 8:63 Guillouzo A, Guguen-Guillouzo C (2008) Evolving concepts in liver tissue modeling and implications for in vitro toxicology. Expert Opin Drug Metab Toxicol 4:1279–1294 Hamid Q, Tulic’ MK, Liu MC, Moqbel R (2003) Inflammatory cells in asthma: mechanisms and implications for therapy. J Allergy Clin Immunol 111(suppl):S5–S12 Hedlund LW, Cofer GP, Owen SJ, Johnson GA (2000) MR-compatible ventilator for small animals: computer-controlled ventilation for proton and noble gas imaging. Magn Reson Imaging 18:753–759 Hogan SP, Rosenberg HF, Moqbel R, Phipps S, Foster PS, Lacy P, Kay AB, Rothenberg ME (2008) Eosinophils: biological properties and role in health and disease. Clin Exp Allergy 38:709–750 Hoymann HG (2007) Invasive and noninvasive lung function measurements in rodents. J Pharmacol Toxicol Methods 55: 16–26 Johnson KA (2007) Imaging techniques for small animal imaging models of pulmonary disease: micro-CT. Toxicol Pathol 35:59–64 Lloyd CM (2007) Building better mouse models of asthma. Curr Allergy Asthma Rep 7:231–236 Martin JG, Tamaoka M (2006) Rat models of asthma and chronic obstructive lung disease. Pulm Pharmacol Ther 19:377–385 Peng X, Abdulnour RE, Sammani S, Ma SF, Han EJ, Hasan EJ, Tuder R, Garcia JG, Hassoun PM (2005) Inducible nitric oxide synthase contributes to ventilator-induced lung injury. Am J Respir Crit Care Med 172:470–479 Quintana HK, Cannet C, Zurbruegg S, Blé FX, Fozard JR, Page CP, Beckmann N (2006) Proton MRI as a noninvasive tool to assess elastase-induced lung damage in spontaneously breathing rats. Magn Reson Med 56:1242–1250 Reid D (2006) MRI in pharmaceutical safety assessment. In: Beckmann N (ed) In vivo MR techniques in drug discovery and development. Taylor & Francis, New York, pp 537–554 Ripoll J, Ntziachristos V, Cannet C, Babin AL, Kneuer R, Gremlich HU, Beckmann N (2008) Investigating pharmacology in vivo using magnetic resonance and optical imaging. Drugs R D 9:277–306 Rudin M, Weissleder R (2003) Molecular imaging in drug discovery and development. Nat Rev Drug Discov 2:123–131 Rudin M, Briner U, Doepfner W (1988) Quantitative magnetic resonance imaging of estradiol-induced pituitary hyperplasia in rats. Magn Reson Med 7:285–291 Russell WMS, Burch RL (1959) The principles of humane experimental technique. Methuen, London Schuster DP, Kovacs A, Garbow J, Piwnica-Worms D (2004) Recent advances in imaging the lungs of intact small animals. Am J Respir Cell Mol Biol 30:129–138

4 Noninvasive Small Rodent Imaging: Significance for the 3R Principles Spielmann H, Grune B, Liebsch M, Seiler A, Vogel R (2008) Successful validation of in vitro methods in toxicology by ZEBET, the National Centre for Alternatives in Germany at the BfR (Federal Institute for Risk Assessment). Exp Toxicol Pathol 60:225–233 Tigani B, Schaeublin E, Sugar R, Jackson AD, Fozard JR, Beckmann N (2002) Pulmonary inflammation monitored noninvasively by MRI in freely breathing rats. Biochem Biophys Res Commun 292:216–221 Tigani B, Cannet C, Zurbrugg S, Schaeublin E, Mazzoni L, Fozard JR, Beckmann N (2003) Resolution of the oedema

57

associated with allergic pulmonary inflammation in rats assessed noninvasively by magnetic resonance imaging. Br J Pharmacol 140:239–246 Walder B, Fontao E, Totsch M, Morel DR (2005) Time and tidal volume-dependent ventilator-induced lung injury in healthy rats. Eur J Anaesthesiol 10:785–794 Wang YX, Yan SX (2008) Biomedical imaging in the safety evaluation of new drugs. Lab Anim 42:433–441

Part Study Planning and Animal Preparation

II

5

Institutional Preconditions for Small Animal Imaging René H.Tolba

Contents 5.1 Background............................................................... 61 5.2 Laboratory Animal Facility.................................... 61 5.2.1 Temperature and Humidity Control........................... 62 5.2.2 Ventilation and Room-Pressurization......................... 62 5.3

Power and Lighting.................................................. 63

5.4 Noise Control + Vibration....................................... 63 5.5

Animal Husbandry................................................... 63

5.6 Institutional Requirements (Diets, Water, etc.)..... 64 5.7 Occupational Health and Safety of Personnel...... 64 5.8 Animal Transport..................................................... 66 5.9 Classifications of Rodents (Based on Microbiological Status).......................... 66 5.9.1 General Considerations for an Animal Facility......... 66 5.9.2 Housing of Animals................................................... 67 References............................................................................ 68

5.1 Background Research with laboratory animal models, in particular, genetically modified rodents like mice and rats, is increasingly recognized as a powerful discovery tool in medical research. One major limitation in the use of experimental animals is often the need to sacrifice the animals to perform blood, tissue, or molecular analysis. This is a major obstacle to observe the biological process under investigation in vivo. Functional, molecular, as well as morphologic quantitative imaging techniques are an important tool for providing data about biochemical, genetic, or pharmacological processes in vivo and repetitively in the same animal. Therefore, the same animals are used as intra-individual controls; this will reduce the standard deviation and will further reduce the number of animals needed per study. This is in line with the three R principles: “Replacement of animal experiments, Reduction of animal experiments, and Refinement of experiments” first described in 1959 by W.M.S. Russell and R.L. Burch (The Principles of Humane Experimental Technique). • • • • • •

Laboratory animal facility, laboratory animal housing Housing systems (Microisolator, IVC, isolator) Laws, regulations Institutional preconditions (diets, water, hypoxia etc.) Occupational Health and Safety of Personnel Animal transport

5.2 Laboratory Animal Facility R.H.Tolba Institute for Laboratory Animal Science and Experimental Surgery, RWTH Aachen University, Pauwelsstr. 30, 52074 Aachen, Germany e-mail: [email protected]

The Guide for the Care and Use of Laboratory Animals published in 1996 by the Institute for Laboratory Animal Research (ILAR) states, “A well-planned, welldesigned, well-constructed, and properly maintained

F. Kiessling and B.J. Pichler (eds.), Small Animal Imaging, DOI: 10.1007/978-3-642-12945-2_5, © Springer-Verlag Berlin Heidelberg 2011

61

62

R.H. Tolba

facility is an important element of good animal care and use, and it facilitates efficient, economical, and safe operation.” There are several factors influencing the design and size of the laboratory animal facility. The determining criteria are given by the institutional research topics, the number, and species of animals to be housed as well as the location of the facility (e.g., on campus, in an existing building or off campus). The animal facility has to be designed and constructed in accordance with all applicable state and local building codes. Moreover, there are several international, national, and local laws regarding biosafety, hazardous substances, controlled substances, occupational health, and animal welfare as well as laws governing the use of genetically modified organisms important to know and to implement. Knowledge of these rules and regulations is essential already at the design phase of the future facility. The planning phase of the facility is allocated with approximately 10% of the building costs, but responsible for approximately 80% of the running costs of the facility. The flexibility of the building for future aspects of biomedical research should also be accounted for. Therefore, the facility design is a crucial step for creating an optimal and economic facility. Functional areas of the facility: The size and the location (central vs. de-central or multi-site) will determine whether areas for separate service functions are needed or necessary. Sufficient space is required to ensure: • Separation of animal species or accommodation of individual projects. • Receive, quarantine, and send animals. • Provide space for animal housing. In some very small facilities, directly to or nearby an experimental laboratory or imaging unit, some functional areas listed below could be unnecessary or included in the central animal facility. • Space for administration, supervision, and direction of the facility. • Staff quarters, showers, sinks, lockers, and toilets. • Highly specialized laboratories (e.g., microsurgical operation theaters). • Individual areas for premedication and preparation, surgery, intensive care, wake-up rooms, necropsy, behavior testing, etc. • Hazardous areas for special biological, chemical, or radioactive agents.

• Receiving and storage areas for food diets, bedding, and cage equipment. • Washing area for cage equipment. • An autoclave for equipment and, a disinfectionchamber or pass-through-box. • Area for holding soiled and unclean equipment, waste disposal. • Area for storing waste prior to incineration or removal (if necessary). Operative framework according to accepted standards: • Guide for the Care and Use of Laboratory Animals (NRC 1996) • European Community directive EEC 86/609 • AAALAC standards for accreditation • National Legislation and References (e.g., AVMA Panel on Euthanasia, etc.)

5.2.1 Temperature and Humidity Control Air conditioning (AC) is a prerequisite for regulating environmental parameters for laboratory animal facilities. Temperature and humidity control is necessary to ensure laboratory animal well-being as well as to prevent variations due to changing climatic conditions. In general, the AC systems should be able to provide the following conditions throughout the year for rodents within the range of approximately 20–23° Celsius or C (69.8–73.4 F). The relative humidity should be controllable within the range of 30–70% throughout the year with a mean of 50–60%. According to the geographical location of the facility, it is sometimes necessary to define a wider range (energy saving).

5.2.2 Ventilation and Room-Pressurization Ventilation of the animal rooms and the rest of the facility is necessary to ensure a balance of air quality (remove CO2 and NH3, provide fresh air and O2), animal comfort, and energy efficiency to provide cage environments that optimize animal welfare and research efficiency.

63

5 Institutional Preconditions for Small Animal Imaging

The AC systems should provide a healthy and comfortable environment for researchers and facility personnel. According to the number of animals, personnel, and equipment used in the room, 15–18 air changes per hour (ACH) are needed. Ventilation with at least 10 ACH is defined as minimum by AAALAC. Pressurization of the rooms with +50 Pascal for the prevention of contamination of the single animal rooms is often used as a safety measure. The animal room is then in positive pressure in relation to the inner floor (passage corridor), and the inner floor again in +50 Pascal overpressure to the exterior floor. This will ensure a pressure gradient (+150 Pascal overall) from the animal rooms, to the inner corridor to the exterior and provide maximum protection against airborne contamination. Heating, ventilating, and AC systems should be designed so that normal operation can be continued with a standby system. In brief, a 100% redundancy is needed. The normal systems should run with 80% (additional capacity for weather extremes etc.), and the second system has to back up the entire first system. The animal facility and human occupancy should be ventilated with a separated system.

devices should be avoided, because they might create odious noises.

5.4 Noise Control and Vibration The facility must be provided with noise-free or maximally noise-reduced environment. Noise control is a very important – and still underestimated – factor in the design and the construction of an animal facility. This starts with the selection of the material for walls and ceilings as well as the installation of plants and machinery (e.g., covered or vibration isolated like in a submarine!). No ultrasonic sources, e.g., for alarm systems!

5.5 Animal Husbandry

5.3 Power and Lighting

The caging system is an important factor in the social and physical well-being of laboratory animals. The housing system has to provide adequate space according to species, the size, and number of animals as well as for the needs of the experiments. We must meet the needs of the laboratory animals regarding maintenance of body temperature, urination, defecation (in combination with bedding material), and reproduction.

The electrical system should provide appropriate lighting and sufficient number of power outlets with a matching power supply. Lighting: Dual light sources should be installed in all animal rooms. Both light sources are needed during working of animal technicians only. For well-being of the animals, only one light source is sufficient. The light intensity in the room should be 300–450 lx (measured 1 m above the floor). The light intensity in the animal cage is, depending on cage material, much lower (ca. 60 lx). This will ensure that no negative effects, even for albinotic animals, are triggered by the lightning. A time-controlled lighting system should be used to ensure a regular diurnal time/night cycle. In general, an emergency (standby power) system is needed for critical system components. For other systems, a power backup emergency system (15 s start time) is needed. Ultrasonic light-dimming

• Provide easy and safe access to food and water from young to adult animals and provide adequate ventilation on cage level. Nowadays, cages are constructed from durable, sturdy materials such as polycarbonate, polysulphone, polyphenylsulphone, or etherimed. Three principal cage designs are available, namely, open cages (USA: static), filter top cages (microisolater), or individually ventilated cages (Figs. 5.1–5.4). • Laws, Regulations In order to meet all laws concerning an animal facility, the management should contact as early as possible the local authorities and discuss specific items directly. According to our experience, this is the easiest and safest way to prevent costly mistakes and foster good communications and understanding between the parties. A brief description of some laws and regulations can be found in the references (e.g., Law regarding genetically modified organism, Animal protection act, Occupational Health and Safety regulations etc.)

64

R.H. Tolba

a

b

Fig. 5.1 (a) Euro-Type 1 Super Long, Makrolon Plastic, open cage, Tecniplast, Italy. (b) Euro-Type 1 Super Long, Polysulphone Plastic, Filter Top, Allentown, USA

a

b

Fig. 5.2 (a) Euro-Type 2 Long Cage, Polysulfon plastic, Filter Top cage, Tecniplast, Italy. (b) Euro-Type 2 Long Cage, Polysulfon plastic, Filter Top cage, Allentown, USA

5.6 Institutional Requirements (Diets, Water, etc.) Many experimental purposes require using preconditioned animals especially for imaging. In most instances, protocols approval by the Institutional Animal Care and Use Committee (IACUC) or the responsible authorities before the start of the experimental procedure is required. This is particularly needed if the animal is restricted in the use of food or

water or any other environmental deprivation like low oxygen tension etc. is intended.

5.7 Occupational Health and Safety of Personnel Philip B. Crosby (born 1926), one of the pioneers in Quality management, said: “Quality begins with people not with things.”

65

5 Institutional Preconditions for Small Animal Imaging

a

b

Fig. 5.3 (a) Green Line Cage, Polysulfon plastic, Individually Ventilated Cage, Tecniplast, Italy. (b) Allentown XJ Cage, Polysulfon plastic, Individually Ventilated Cage, Allentown, USA

a

b

Fig. 5.4 (a) Individually Ventilated Cages, Green Line, Polysulfon plastic, Racks and Ventilation Unit, Tecniplast, Italy. (b) Individually Ventilated Cages, Allentown XJ-System, Polysulfon plastic, Racks and Ventilation Unit, Allentown, USA

66

The selection of appropriate staff for the laboratory animal facility is the key component in the management of an animal facility. The other critical factor is the initial in-house training of staff at all levels. During this training, the Occupational Health and Safety Program has to be conducted to all personnel to become familiar. This is a continuous effort with annual refresher courses. Only through continuous training efforts, the facility will be able to ensure a program that has a life and does not exist only in written form. The program has to establish and maintain, e.g., chemical, biological, physical safety, safety equipment and “Safe Work Practices,” “Personal Protective Equipment” etc. The program should be reviewed periodically and adjusted to the facility and program needs.

5.8 Animal Transport The transport of laboratory animals from one place in the facility to another is very important and must be undertaken with care. Consideration for the transport of animals is according to factors like the mode of transport, the species (always only one species in one transport box), the transport container or cage, the animal density in the box, food and water during transportation, protection from contamination from outside, protection of the environment if the animals are infectious, as well as transport injuries and transport-stress. Especially when animals are being used for imaging purposes, the transport stress can have a major influence on the quality of the imaging process or the quality of the data measured (e.g., transport stress, higher blood pressure, and higher microcirculation!). The mode of transport of animals depends also on the distance (in house, shipping to other locations) as well as seasonal and climatic conditions. To minimize the stress for the animals, special commercially available HEPA-filtered transport containers should be used with a special food (e.g., jelly food) to provide enough food and water during the transport period. Before starting the experiment or imaging, the animals should acclimatize at least 5–7 days to the new surroundings. Most countries have implemented special laws regarding the transport of living animals (USA, Canada, European Community etc.). For air

R.H. Tolba

transport, the International Air Transport Association Guidelines (IATA, www.iata.org) are mandatory. Some countries, e.g., Great Britain, still require quarantine of animals after shipping from other countries before handing them over to the recipient.

5.9 Classifications of Rodents (Based on Microbiological Status) Axenic (germ free) – Animals completely free of all microorganisms. • Bred, reared, and maintained under sterile conditions (e.g., Isolator). Gnotobiotic – Animals with a defined microbial flora. • Germ-free animals deliberately administered certain varieties of harmless bacteria to aid in digestion. • Bred, reared, and maintained under sterile conditions (e.g., Isolator). Specific pathogen free (SPF) – Animals that are free of disease-causing microbes or pathogens. • Founder animals of an SPF colony are cesareanderived. • Bred, born, reared, maintained in environments that prevent exposure to or transmission of pathogens (e.g., Barrier or Individually Ventilated Cages, IVC). Conventional – Animals with an undetermined microbiological status. • Raised and maintained under standard conditions

5.9.1 General Considerations for an Animal Facility It is desirable to operate a central animal facility with several units (barriers) for different animal species and different hygiene levels. The other options are several smaller units close to the laboratories (satellite facilities, multi-site, de-centralized). The animal facility has a core breeding unit (repository) in an SPF status. Specific Pathogen Free is a term used for laboratory animals that are free of particular pathogens that are tested every 3 months according to specific recommendations (Recommendations

67

5 Institutional Preconditions for Small Animal Imaging

for the health monitoring of rodent and rabbit colonies in breeding and experimental units, FELASA Working Group on Health Monitoring of Rodent and Rabbit, Colonies: W. Nicklas (Convenor), P. Baneux, R. Boot, T. Decelle, A. A. Deeny, M. Fumanelli & B. Illgen-Wilcke. Laboratory Animals (2002) 36, 20–42) It has to be always accompanied by a list of the tested and absent or found pathogens. The health monitoring is assured through sentinel mice. Analogous to the “canary in the coal mine,” sentinel rodents monitor the pathogen status of investigators or breeding rodents. A specific number of animals or cages are monitored by sentinel animals. It can be a section/area in a room or a rack that is monitored with two sentinel animals. Every time a rodent cage is changed (usually weekly), about a tablespoon of soiled bedding from that cage is transferred to the sentinel cage. In this way, sentinels are exposed to whatever pathogens may be present in the urine, feces, fur, saliva, dander etc. from 100% of the cages on the rack. Because of this, investigating personnel are not allowed to handle or move sentinels or sentinel cages. At the end of the monitoring period (usually one quarter or 12 weeks), the sentinels are sampled and tested for a specific list of infectious agents. Sentinels may be tested more or less frequently for selected pathogens or prior to shipping or experimental needs. Only when no pathogens on the list are found (NO positive results), the animals are specific pathogen-free. To assure this hygienic level, in general, no researcher has access to the core breeding unit. This minimizes the risk of infection for this unit. New animals are transferred only via embryo transfer to this core breeding unit. Embryo transfer refers to a step in the process whereby several embryos from donors (the transferred line) are placed into the uterus of a foster mouse with the intent to establish a pregnancy. By using this technique, it is possible to transfer a line with a minimal risk of bacterial or viral contamination to the barrier unit. From the core breeding unit, the animals are then transferred to an experimental barrier according to the researchers’ need, where several personnel have access to the animals. It is possible to have several units according to different hygienic levels or experimental needs. It is also possible to build one unit outside of the barrier. It is sometimes essential to transport animals within the building or between

campus buildings in order to facilitate research. To ensure the integrity and well-being of the animals being transported into the imaging facility, a system has to be setup that will facilitate the research while minimizing the risk of spreading a contaminant. Once the animals have left the barrier, they never go back to this unit. Following imaging, radiation, e.g., the animals are returned to a different unit or quarantine room. Employing this one-way policy will minimize the risk of infections for animals in other units.

5.9.2 Housing of Animals Animals can be housed in open caging, plastic “shoe box” cages from macrolone, polysulfon- or etherimide plastic with stainless steel wire-bar lid. Additional items also include bedding, water bottle, food, and cage card holder (see picture of Allentown and Tecniplast). Open caging is the easiest way to house rodents, but there is a greater risk of contaminating the animals as well as causing allergies to the personnel. Currently, the microisolator technique is the system of choice. In addition, there is a plastic top that holds a permeable filter. It is also possible to use individually ventilated cages (IVC) in general or for special procedures where the whole cage rack (and the cages in the rack) is separately ventilated with HEPA-filtered air and no connection with the outer space of the room. Each single cage is separately ventilated and maintained with a small overpressure to protect the animals inside. The inside of the cage, filter top, wire lid, water bottle, bedding, and food are considered “clean,” that is, free of particles that may contain disease-causing pathogens. The outside surfaces of the cage and top are potentially contaminated or “dirty.” A transfer of animals is never performed “open” in the room. For this purpose, the closed IVC is transferred to a cage-changing station (-CCS), a mobile microbiology safety cabinet. In this laminar air flow-protected environment, the IVC can be opened and animals can be transferred into the new, clean cage without contact to possible contaminated air or persons in the room. After the experimental procedure or the imaging process, animals can be transferred into an IVC rack near the imaging facility or a special designated room in the facility. When there is a known risk of infection,

68

a higher biosafety level (e.g., BSL-2) etc., it is also possible to have this cage in a permanent negative pressure to ensure biosafety.

References “Replacement of animal experiments, Reduction of animal experiments and Refinement of experiments” first described in 1959 by W.M.S. Russell and R.L. Burch (The Principles of Humane Experimental Technique) Animal Welfare Act – 9 CFR Chapter 1, Subchapter A, Animal Welfare, USA GV-SOLAS Gelbes HEFT Nr. 1: Planung, Struktur von Versuchstierbereichen tierexperimentell tätiger Institutionen. Hilken G, Haemisch A, Kluge R, Tolba R (eds) 5. Auflage in Bearbeitung, www.gv-solas.de GV-Solas gelbes Heft Tiergerechte Haltung: Labormäuse http:// www.gv-solas.de/auss/hal/maushaltung.pdf Hessler JR, Noel DM (eds) Lehner planing and designing research animal facilities. American College of Laboratory Animal Medicine Series, AP Biosafety in microbiological and biomedical laboratories, 4th edn. 1999, HHS Publication No. (CDC) 93-8395

R.H. Tolba Guide to the Care and Use of Experimental Animals (1993) Canadian Council on Animal Care. Vol. 1 CPCSEA Guidelines For Laboratory Animal Facility (2003) Indian J Pharmacol 35:257–274 (special) Guide for the Care and Use of Laboratory Animals (1996) National Research Council, National Academy of Sciences, USA Council Directive on the Approximation of Laws, Regulations and Administrative Provisions of the Member States Regarding the Protection of Animals Used for Experimental and Other Scientific Purposes (1986) European Union (Directive 86/609/EEC) European Convention for the Protection of Vertebrate Animals Used for Experimental and Other Scientific Purposes (1985) Council of Europe (Convention ETS 123) Recommendations for the health monitoring of rodent and rabbit colonies in breeding and experimental units, FELASA Working Group on Health Monitoring of Rodent and Rabbit, Colonies: Nicklas W (Convenor), Baneux P, Boot R, Decelle T, Deeny AA, Fumanelli M, Illgen-Wilcke B (2002) Lab Anim 36:20–42 Occupational Health and Safety in the Care and Use of Research Animals (1997) National Research Council, National Academy of Sciences, USA Guide for the Care and Use of Agricultural Animals in Agricultural Research and Teaching, Federation of Animal Science Societies, 1st revised edn (1999) USA

6

Statistical Considerations for Animal Imaging Studies Hannes-Friedrich Ulbrich

Contents 6.1 Design of Experiments in Animal Imaging.......... 69 6.2 Comparing Interventions....................................... 70 6.3 Measuring the Desired Effect: Definition of Endpoints ......................................... 72 6.4 Sources of Variability ............................................ 73 6.5 Avoiding Bias . ........................................................ 73 6.6 Blocking .................................................................. 74 6.7 Drop-Outs................................................................ 74 6.8 Regression, Repeated Measures, Experimental Units, and Other Topics................. 75

“If an experiment is well designed, it is relatively easy to get help with the statistical analysis; but if it is incorrectly designed, it may be impossible to extract any useful information from it.” Festing (2002, p. 191) with or without help, designing an experiment is a challenging task for the experimenter. Based on some examples, crucial steps for designing animal imaging studies will be explained, accompanied by some hints for efficient study conduct and data evaluation. Statistics provides such techniques and procedures, therefore, it can be seen as an essential component of empirical research, in biology as a whole as well as in animal imaging.

6.9 Measurement Schedules . ...................................... 77 6.10 Animal Allocation and Randomization ............... 77 6.11 Sample Size Estimation.......................................... 78 6.12

6.1 Design of Experiments in Animal Imaging

Study Design and Research Plan........................... 79

6.13 Practical Considerations for Conduct and Evaluation of Experiments in Animal Imaging.................................................. 6.13.1 Data Structure........................................................... 6.13.2 Document the Unexpected . ..................................... 6.13.3 Changes in Study Design . .......................................

79 79 80 80

6.14 Statistical Evaluation of Animal Imaging Experiments............................................. 80 6.15 Outliers ................................................................... 80 6.16 Hints on Evaluating the Data ............................... 81 References ........................................................................... 82

H.-F. Ulbrich GDD Statistics, Global Drug Discovery, Bayer Schering Pharma AG, Müllerstrasse 178, 13353 Berlin, Germany e-mail: [email protected]

In order to separate the signal from the background noise, an experiment can be conducted. An experiment is understood as a procedure for collecting scientific data on the response (effect) to an intervention. In this chapter, the terms experiment and study will be used synonymously describing controlled experiments as opposed to observational studies. Since observational studies (also named surveys) investigate mostly responses on preexisting “conditions”, they are a kind of scientific investigation less likely to be performed as animal imaging study. Experiments must be planned carefully and the planning process should be completed before the experiment begins. This planning should also include the statistical methods intended for assessing the results. Changes after the start of the experiment are indeed sometimes necessary; they shall be kept to an absolute minimum. Changes after the experiment has

F. Kiessling and B.J. Pichler (eds.), Small Animal Imaging, DOI: 10.1007/978-3-642-12945-2_6, © Springer-Verlag Berlin Heidelberg 2011

69

70

Terminology • Experiment – procedure for collecting scientific data on the response (effect) to an intervention • Experimental unit – the smallest unit to which an intervention can be assigned to • Intervention – the entire description of what an experimental unit is assigned to • Observational unit – the smallest unit on which a response will be determined • a – the (nominal) significance level • (1-b) – the (nominal) power • D – the (relevant) difference of interest • s – standard deviation (square root of the variance) begun are usually very complex to implement and have statistical as well as credibility consequences to be considered carefully when changing the experimental plan. Therefore, along with the clear objective of the study, essentials of the experimental design shall be fixed in advance in a research plan or study protocol. The study plan shall be accessible for and understood by everybody (lab technician as well as researcher) involved in conducting an experiment. Designing the experiment is one of the most im portant tasks during the planning phase. As stated by Festing (2002, p. 191), “if an experiment is well designed, it is relatively easy to get help with the statistical analysis; but if it is incorrectly designed, it may be impossible to extract any useful information from it.” An experimental design is the manner in which the experimental units – most likely the individual animals in an animal imaging study – are arranged in accordance with the different interventions and the effects that are considered to influence the experiment. The principles of good experimental design are universal and have been known for more than 50 years (Cox 1958). Applying them to animal experiments and animal imaging experiments, in particular, requires some additional thoughts.

6.2 Comparing Interventions Example 1 – “treatment” vs. “control”: Arguably, the simplest design is to work with two groups – “treatment” and (vehicle) “control” as the interventions – measuring the response (e.g., diameter of a lesion) of

H.-F. Ulbrich

each group member once; the difference between the average group response shall be the effect of interest. There will always be variability between the responses of the members of each group. Taking this variability into account, the difference between the groups can be used to determine whether the disparity between groups is a real one. If this difference is not completely obvious, statistical analysis (e.g., Student’s t-test) has to be employed. The objective of that experiment shall be to show the superiority of the treatment over vehicle control. Even in that simple case, the following conditions must be clearly established before the experiment gets started. 1. How certain has the result to be? (Rather confirmation or exploration, delivering test results or estimations) 2. What are the different interventions of interest? 3. What will be measured – once or repeatedly – to assess the results of the experiment? (Identification of experimental units, measurement schedule, to be measured items) 4. What results are to be expected under the different interventions? (Determination of experimental variation, their different sources, how to limit it, or allow for it in a fair manner) 5. How can the experiment be organized and conducted that most information is obtainable from the resources used? 6. How large will the experiment be? (The “right” number of experimental units – neither too much nor too few – taking scientific and ethical considerations into account) 7. How will the results be analyzed, again to gain as much information as possible If based on profound prior information, e.g., in a later stage of development, an experiment is planned to confirm that “treatment” is really superior to “control,” then null hypothesis, alternative hypothesis, as well as the (biological) relevant difference D between the populations’ average responses are known beforehand. The researcher is familiar with the amount of biological (among animals) and measurement variability for each group. So, let us assume that this total variability s is equal for each group. The groups in the experiment are expected to be samples from the respective populations. The test statistic (e.g., Student’s t-test) to be used for deciding the relevant question has been fixed. It is now the researcher’s obligation to decide on both the significance level (the nominal a-level) and

6 Statistical Considerations for Animal Imaging Studies

the power (1-b) for the intended test statistic. The significance level a indicates the likelihood that the two groups will be found to be different, although there is no difference between the respective populations. Therefore, the significance level has to be sufficiently low (common values are 5 or 1%, meaning that if there were 100 experiments done with samples from populations without any difference, five or one respectively of these experiments would show a significant result – just by chance alone). Statistical power (1-b) states how large the likelihood has to be that we correctly reject the null hypothesis (in favor of the alternative one is aiming for) with the intended significance level a (as above) if the distance between the underlying populations is equal to the (biological) relevant difference D. The power has to be set by the researcher. A power of let say 80% tells us that eight of ten experiments done with samples from populations different by the (biological) relevant difference will let us reject the null hypotheses at the preset significance level a. (Two out of ten potential experiments will fail to show a significant result.) For confirmatory experiments a and (1-b) are clearly the measures for how certain the results have to be. With a chosen test statistic and the decisions made how large a, (1-b), D and s shall be, the size of this experiment (sample size) can be determined (see below). As long as D and s are not yet known (e.g., in earlier stages of development), experiments are rather of exploratory than confirmatory nature. Exploratory analysis is an approach to analyzing data for the purpose of suggesting (new) hypotheses about the causes of observed phenomena, assessing assumptions on which statistical data modeling and evaluation will be based as well as providing a better basis for further data collection through experiments. An exploratory approach, therefore, should collect evidence for more specific hypotheses to be followed during further research and development. Example 2 – two different treatments: Let us imagine a slight change, instead of testing “treatment” against (vehicle) “control,” one might be interested in comparing two absolutely new substances called A and B for the purpose of getting information on whether the one or the other is more promising for further development. Since this will be the first experiment with interventions applying A and B nothing is known in advance on whether A is better than B or vice versa, nothing on the variability s let alone a difference D considered to be relevant.

71

Rather than comparing A and B by using a statistical test, an estimate of the difference m between A and B should be derived along with an appropriate (1-a) confidence interval, (1-a) conventionally being 95 or 99%. Beyond comparing different substances, the interventions of interest in an experiment comparing A with B might also be: • Low dose of substance X (A) – high dose of substance X (B). • Once per day application (A) – twice per day application (B) of the same substance X and dose. • Once per day application of a full dose (A) – twice per day application (B) of half the full dose of the same substance X. If, for example, contrast media (CM) were used and an interaction between CM and substance X (drugcontrast medium interaction as a special case of drug– drug interaction) could not be excluded, then the interventions A and B might actually become: • Low dose of X under CM (A) – high dose of X under CM (B). Whatever the results of the latter comparison, conclusions drawn from such an experiment are valid only for the combination of X and CM as applied; whether the observed differences between A and B are the same as would have been between low and high dose of X alone is beyond the results of this particular experiment. While comparing devices and/or specific adjustments of the same technical device (e.g., different scanning modes), interventions A and B translate to two different device adjustments. Example 3 – more than two treatments: Let us again change the above experiment a little and extend it by some more substances of interest, making it a comparison of four interventions called (by substances used) A, B, C and D. Here, a lot of different questions might be of interest, e.g.: • Are there any differences at all between the interventions A to D? • Are there any pairwise differences – between A and B, or A and C, or A and D, or B and C, or B and D, or C and D? • Are there any differences between a particular (reference) intervention, say D and any of the others? In each case a different number of questions is to be addressed – one, six, or three, respectively – by the otherwise same experimental setting.

72

In each of the above mentioned examples, it is assumed that an animal within the experiment belongs to exactly one particular intervention group and that an animal’s response to the intervention is measured once. The measurements are independent of each other; each animal serves as both experimental and observational unit. Neither this independence nor the unity of experimental and observational unit has always to be the case. Example 4 – intraindividual differences, more observational than experimental units: The following two experiments explore the effect of one and only one intervention by measuring each animal twice in either of the following manners: (4a) Before and after the intervention – the animal is measured at two distinct time points (4b) Both an area of interest (say, tumor) and a appropriately chosen reference area (tumor free) – the animal is measured twice at the same time point, but on two different locations (4c) Both an area of interest (say, tumor) and a appropriately chosen reference area (tumor free) – as in 4b but the animal being completely scanned, area of interest and reference area measured by defining them as part of one and the same picture or sequence. The different animals in the experiment are still independent of each other; the two measurements taken from the same animal are not. Each animal is an experimental unit, each animal’s measurement time (4a) or location (4b, 4c) an observational unit (regardless of how many variables are actually to be measured, how many values therefore to be taken). These experiments are one-group experiments where the experimenter is interested in whether the intraindividual differences are merely in favor of the intervention.

6.3 Measuring the Desired Effect: Definition of Endpoints Comparison of interventions “better,” “worse,” or even “different” does not have any meaning without the reference to a (biologically) meaningful endpoint, a quantity measured, a property or characteristic determined, or a value derived from a combination of former. Measurable

H.-F. Ulbrich

quantities might be the length, width, height, or body temperature of an animal; the animal’s sex can be determined; tumor volume can be derived from its length, width, and depth. For mammals, only sex seems to be well defined, all others need further specification. Example 5 – Endpoint specification: Let us consider Beagle dogs to be the experimental animal, the following endpoints determined for each animal might be of interest in the experiment: • Length: from tip of the nose to base of the tail, i.e., excluding tail and (hind) legs • Width: hip width – as distinguished from shoulder width or maximum overall width • Height: height of the withers • Body temperature: taken in the ear – as opposed to taken in the mouth or the anus Since not directly measurable, some quantities might be derived: • Tumor volume: calculated as (π / 6) · length ·width height assuming the shape of an ellipsoid – rather than assuming a rectangular prism length·width·height • Tumor volume (if height/depth was not easily measurable): calculated as ·

·

(p / 6)· length · width2

assuming the shape of an ellipsoid and height to be equal to width of the tumor • Lesion area: circumscribing the lesion at a monitor, the value then be given by the machine Since different measurement devices (carpenter’s rule vs. Vernier caliper; different CT scanner modes) as well as different measurement points (ear or anus body temperature) and different biological conditions (calmness or stress; different moments of the day) affect the values obtained, consistent endpoint definition includes the specification of the measurement devices and their parameters, measurement areas, time points of reading, among others. Defining an endpoint less precisely than possible does make an experiment less conclusive by adding to the noise or generating bias (see below). For each endpoint, one has to know its type: • Nominal – values are clearly distinct categories without any internal order, e.g., colors like black, red, green; or gender

73

6 Statistical Considerations for Animal Imaging Studies

• Ordinal – values are distinct categories that can be (completely) ordered from low to high, e.g., quality of a signal: very poor, poor, moderate, good, very good • Cardinal – counts, i.e., nonnegative integer numbers, e.g., number of liver lesions • Interval scaled – continuous measurable quantities for which the precision of the measurement is an important issue, e.g., height (measured in mm or inch or any other unit of length) The type of the endpoint is an important parameter for deciding on an appropriate statistical evaluation strategy: starting with an adequate data description (e.g., rates, proportions, medians and interquartile ranges, means, and standard deviations), useful graphs (e.g., pie charts, bar charts or histograms, box plots), appropriate tests of significance (e.g., c2 tests for contingency tables, t-tests or ANOVA for Gaussian distributed data), and regression models (like logistic regression or multivariate linear regression). The answer to the question whether interval-scaled variables can be fairly well described by a Gaussian (“normal”) distribution has a deep impact on the above mentioned decision as well. Body length of adult animals of either sex (but not for both sexes together!) is a good example where a Gaussian distribution can be assumed. For others: If there is no published consensus on the matter, deciding if the data can be described by a particular distribution can be a problem. Transforming the data (e.g., by a log transformation, see Keene 1995) might be an option. The answer to the question is also important for sample size estimation.

6.4 Sources of Variability Variability of all biological phenomena is a major cause of the complexity of biological experiments. Animals are different even within the range of the same species. Avoiding as much variability as possible is certainly a good way to increase the signal-to-noise ratio. Genetically, monozygotic littermates are more similar to each other than dizygotic, which are more similar than offspring of two different dams. Animals of the same strain but different breeders might be more different than animals of the same breeder. Animals of different ages – especially juveniles – are likely to be more diverse than ones of the same age. Since body

weight of male as well as female juveniles is highly correlated with age, it can be concluded that juveniles of the same sex and weight are more similar to each other than juveniles of either different sex or weight. Differences in the animals’ environment before and during the experiment might make them more or less similar. Different feeding schemes or diets during upbringing, different cages in multianimal caging conditions, different locations of cages within stables (leading to different microclimate conditions: light, temperature, draft, frequency of human contact) are likely to cause additional variability among animals. Any measurement procedure itself carries more or less variability. This variability might be caused by the device, device parameters differently set, the experimenter’s ability to handle the device, her or his ability to handle the animals, or the experimenter’s reading. In animal imaging studies, anesthetization of the animal, the animal’s position toward a scanner, and chosen reference areas for background signal measurement are important sources of (undesired) variability. If each specific variability is fairly spread over all animals and measurement occasions during the experiment, it simply adds up to the overall noise impairing the signal-to-noise ratio. Otherwise, variability might be related to bias.

6.5 Avoiding Bias Example 6 – Assessment bias: Consider an experiment comparing a traditional treatment scheme (intervention A) with a new one (B) developed and proposed by the experimenter him/herself. The number of clearly visible lesions (within a specific organ) was chosen as primary endpoint; the less the number of lesions, the better the treatment scheme. The experimenter’s decision whether to consider a lesion as clearly visible might subconsciously be associated with the interest to show the superiority of the new scheme; the decision might no longer be inherently fair for both interventions. The change in results (number of lesions recorded) caused by the subjective element to assessing observations leads to bias – this particular bias is called assessment bias. Any systematic difference between intervention groups apart from the intervention itself might cause the experiment’s results to be biased, e.g.,

74

• Heavier animals allocated to a particular intervention group – allocation bias. • Different devices (or device modes) used for measuring different groups. • Different animal housing conditions (climate, light, multicaging conditions) for different groups. • Different observers (e.g., lab technicians of different expertise) for different groups – observer bias. • Different times of the day for measuring different groups. Being aware of potential bias while designing the experiment helps avoiding it. In animal imaging studies assessment bias might be important. Arguably, the best idea to prevent assessment bias is to keep the observer “blind” toward the intervention group an animal belongs to. This might be accomplished by: • Different individuals as facilitator (handling and treating the animals) and observer (taking the measurements). • Identification of the animals by a haphazard numbering scheme not known to the observer. • Letting the observer see the animals for inspection in random order.

6.6 Blocking There is another important way to increase by design the signal-to-noise ratio of an experiment: blocking. By blocking, one splits the animals into homogeneous subsets in such a way that variation within each subset is considerably lower than variation between subsets. Example 7 – Blocking by sex: While designing an animal experiment, it is always recommended to think about potentially different responses by animals of different sex. Such differences might be caused by the body size or hormone configuration or some other trait different between the sexes. Splitting the experiment virtually into two otherwise completely identical – one based on the females, the other based on the males – would make the experiment one with block design. Complete blocking means splitting the experiment in such a way that within each block all the interventions take place, while one and the same animal number ratio between the interventions is maintained.

H.-F. Ulbrich Table 6.1 Complete blocking by sex and weight and breeder Block 1

5 heaviest female animals of breeder 1

... Block n1

5 lightest female animals of breeder 1

Block (n1 + 1)

5 heaviest male animals of breeder 1

... Block (n1 + n2)

5 lightest male animals of breeder 1

Block (n1 + n2 + 1)

5 heaviest female animals of breeder 2

...

Example 8 – Complete blocking by sex and weight and breeder: Let us consider an experiment where the experimenter wants to get more information on the new intervention A as compared to the well-known intervention B by planning one and a half as much animals for A as for B, a three to two animal number ratio between interventions is needed. For scientific reasons, the experiment has to be a large one, based on more animals than one breeder can supply by a certain date. It is assumed that the interventions themselves might affect heavier animals less than lighter ones, supposedly females will react (slightly) differently from males. Since one and a half as many animals are planned for A as compared to B, the smallest possible block size is five animals, therefore each block has to have a size of a multiple of five. Blocking then might look like Table 6.1.

6.7 Drop-Outs An animal not adhering to its appointed intervention any longer is called a drop-out. Leaving the experiment prematurely might happen for quite different reasons, e.g.: • Death (without or with – in multianimal caging – cannibalism) • Escape • Infection • Premature stop of intervention for animal welfare reasons (e.g., an observed tumor volume of 2000 mm³ or above, like in Hanfelt 1997, p 298) • Early sacrifices (e.g., every other animal after half experiment time for histological examination)

75

6 Statistical Considerations for Animal Imaging Studies

Early sacrifices can and shall be planned in full – including the consequences for sample size and evaluation. Escaping an unlocked cage hopefully never happens; nevertheless, it is the only one of the above examples that might be considered as completely independent of the intervention. Loosing animals that way is a waste of resources, anyway. All other examples given must be thought of as potentially intervention related, and therefore, likely to generate bias, i.e., causing an unjust change in observed effect(s). Having to take out animals for welfare reasons has to be considered like any other premature termination while designing the experiment. Besides recording the drop-out time (and reason), any measurement to be taken (e.g., body weight and histological examination) has to be decided on during design phase. Example 9 – Design with drop-out: Intervention A shall be the placebo in a xenograft experiment to compare different doses of a novel substance (say, interventions B, C, D and E being increasing doses) for inhibition of tumor growth by limiting angiogenesis. Substances will be applied on day 0, animals will be checked daily for body weight and tumor length and width by an experienced lab technician; every seventh day up to day 28, the animals have to undergo an imaging procedure for measuring angiogenesis and tumor volume. Midterm, on day 14, after the imaging procedure, one third of the animals shall be sacrificed for histological examination. All others are to be histologically examined after day 28. It is expected for certain that animals of the placebo group have to be taken out earlier because of unbearable tumor load, 2000 mm³ being set as limit. Nobody knows when this will happen to the first animal or whether this will affect any animal of the interventions B to E. The study protocol has to make clear how the tumor volume is to be determined from length and width (as taken daily by the lab technician), who is responsible for the calculations and the timely decision, and whether to stop the experiment for that particular animal. It has to be decided who takes this responsibility for the days when the main lab technician will not be available because of weekends, holidays, or sick-leave. Furthermore, it has to be defined what happens to an animal reaching the tumor burden limit on a day without scheduled imaging procedure (say, on day 6) – to be imaged or not before leaving is here a decent question. And what about the histological examination?

It is extremely important to have the animals foreseen for midterm sacrifices identified before the experiment actually begins. Although it might sound appealing to some, sacrificing on day 14 merely animals with already larger tumor volumes (and therefore more likely to be excluded the following days) would generate bias – most probably of a quite remarkable amount. This has to be avoided.

6.8 Regression, Repeated Measures, Experimental Units, and Other Topics Example 9 (continued) – Another look: Intervention A more precisely defined as vehicle control might be considered as a treatment with a dose of 0 units of the new substance. Such an experiment can be planned not only for comparing the different doses to the control but also for establishing a dose–response relationship by fitting a regression model. The type of the regression (linear or nonlinear, e.g., logistic) suitable depends on a lot of additional knowledge or assumptions like type of the response variable, monotonicity of responses – to name a few. Example 9 (continued) – Yet another look: As stated in the study protocol, each animal has to be repeatedly examined – daily (starting on day 0) for body weight and tumor volume, and weekly by the expensive (device availability and cost of use) and animal stressing (anesthesia) imaging procedure. Let us suppose tumor volume determined by the imaging procedure is the experiment’s primary endpoint. Then one gets up to five measurements (on day 0, 7, 14, 21, 28) per animal. Although this approach increases the number of measurement values remarkably for each intervention group, it does not increase the number of experimental units; values taken of the same animal cannot be considered independent of each other (as values of different animals are). The use of a significance test based on the assumption of independent experimental units (like t-tests, U-tests, or classical ANOVA) would be inappropriate here. The observational unit here would be each animal’s daily check, whereupon values on the primary endpoint (volume determined by imaging) are only to be taken at some of them. The proper identification of the experimental unit is a crucial issue for the validity of both the chosen design and the related statistical evaluation strategy.

76

H.-F. Ulbrich

Check Box: Designing an Experiment in Animal Imaging First and foremost, a clear objective must be defined. Following is a (nonexhaustive) list of intertwined questions to be considered designing an animal imaging experiment: • Shall the experiment be designed in collaboration between experimenter and (bio)statician in plenty of time before the experiment starts? • What is the main (biological) effect of interest, are there any secondary effects to be considered as well? • Is the experiment part of either exploratory or confirmatory research? • What tentative design supports the main question (randomized group comparison, intraindividual comparison, more advanced or specialized designs, etc.)? • How is the experimental unit defined (e.g., species, strain, knock out, immune deficient, sex, etc.)? • What quantities shall describe the desired effect(s) – the primary (and further) endpoint(s)? • What are the measurable quantities and how do they relate (e.g., by judging them using a score or by using a particular mathematical formula) to the quantities describing the effect? • Measurement schedule (and observational unit) – what measurements are to be carried out, when, how often, by how many observers? • Will the observers be “blind” regarding the intervention each particular experimental unit undergoes? • What are the major sources of to-be-expected variability (animals, measurement devices, observers, environment, etc.)? • What measures can and will be taken to limit any of these variabilities (single strain, unisex, single observer, etc.), what’s not feasible? • How are the different interventions defined, what are the relations between any of them (e.g., multipled dose)? • How is it assured that the (environment) conditions (i.e., all conditions besides the intervention itself) are kept equal for each experimental unit during the whole experiment? • How are the data to be collected, kept, and checked for plausibility? • How shall the endpoint(s) be summarized within each intervention? • How can the overall effect be expressed in comparative terms between interventions (signal)? • What, if any, is the minimum magnitude of the signal one is interested in? • Assigning the experimental units – is there any blocking factor (necessary)? • What, if any, is the likely drop-out rate; might it be intervention-specific? • How many experimental units (animals) are needed – overall, per intervention – taking into account the to-be-expected drop-out rate and pattern? • How will the animals be randomized (allocated) to the different interventions? • What kind of graphical technique (scatter plot, box plot, etc.) will be used to support: −− Checking the raw data, −− Presenting the (main) results? • What kind of statistical procedures will eventually be used for data evaluation? • Checking the assumption(s) for the chosen statistical evaluation strategy – will it be done at all, how, and to what consequences? Wanting the results to be published, the following topics (likely to be requested by the editors) might also be of interest from the beginning: • How to describe the used statistical methods (especially if uncommon methods were to be used), what references are needed? • How to describe design changes – if there were any unavoidable – and checking procedures related to the chosen statistical evaluation strategy?

77

6 Statistical Considerations for Animal Imaging Studies

Rodents are highly social animals. Therefore, (moderate) multianimal caging might be considered for preventing the animals from isolation (and the stress caused by crowding) for a long-term experiment. Since multianimal caging influences the single animal’s stress level and body weight, animals within the same cage can rarely be considered as independent to each other. The cage becomes the experimental unit – with consequences on sample sizes, drop-out patterns, and statistical evaluation strategy. The normal biological diurnal rhythms of an animal’s biochemistry and physiology alter its responses depending on the time of day that treatments are applied, samples are taken, or an anesthesia is administered before imaging. Example 10 – Time of the day as a potential for bias: Imagine an experiment on tumor vascularization where always starting at seven in the morning all animals of intervention A will be imaged, followed by B, C, D, and finally E. All the animals of intervention E are imaged late in the afternoon. Whatever the interactions of the circadian rhythm and the interventions are, such a design does not level them out fairly. It is, therefore, recommended to think about alternatives such as a random order of animals. On the other hand, if a sequence that is too complicated to be handled correctly makes the whole process error-prone, one should rather apply a less complicated order. Example 11 – A warning: Occasionally, experiments are conducted under multianimal caging conditions without single animal identification. If there were only one observational unit (measurement time point) per animal, the experiment could still be well designed taking advantage from the unity of observational and experimental unit. A similar experiment where animals are to be measured on more than one occasion would be questionable by design, since intraindividual changes over time (see example 4a) are not determinable. The cage would be the observational unit, and therefore, the experimental unit as well, a scenario likely to be less efficient with regard to the number of animals to be included.

6.9 Measurement Schedules Animals have a circadian rhythm, humans as creatures have it as well. Beyond that, as social beings, most humans are used to living according to a hebdomadal

rhythm. This has to be considered as an additional challenge for the design of experiments lasting several days or even weeks. Daily examination means including weekends and holidays. Example 12 – Tumor doubling time and weekend measurements: Imagine an experiment with (nonshrinking) tumors and tumor doubling time (the time it takes the tumor from observation start to grow to twice its volume) as primary endpoint. Precision of time measurement shall be 1 day. It is quite a natural choice to evaluate tumor doubling times by a time-toevent analysis. Not determining tumor volume on weekends would mean that Monday’s reading finds tumors that already had doubled their volume either by Saturday, Sunday, or Monday. Since it is not possible to decide, which tumors had doubled by Saturday, Sunday, or Monday, respectively, the actual time measurement precision is three (!) days (instead of the intended one). For any time-to-event analysis, time measurement precision is of particular importance. Sometimes a biweekly measurement (e.g., on Monday evenings and Friday mornings for measures not influenced by any circadian rhythm) might be sufficient and the best choice – having only one observer (lab technician) involved.

6.10 Animal Allocation and Randomization Designing an experiment is like gambling with the devil: only a random strategy can defeat all his betting systems. RA Fisher (quoted in Box et al. 2005) Randomized experimentation is not haphazard. Randomized allocation of animals (or, more generally, experimental units) to the different interventions should be based on random digits generated by a (pseudo-) random number generator. Random digits can be found in Machin et al. (1997, p. 300). In simpler cases tossing a fair die or (for a two-intervention experiment with equal group size) a coin can be used. More complex cases can be addressed by using software, although there seems to be no randomization software covering all possible designs. SAS® proc PLAN might be a good start; for a list of available randomization software see Bland (2008).

78

H.-F. Ulbrich

Without a true random process as independent basis, animal allocation is prone to bias. Except for random deviances, randomization balances groups with respect to all factors other than intervention. Differences between the responses are, therefore, due to the different interventions. Complete randomization would be a valid option, but with the risk of ending up with only roughly the intended number of animals per group. Blocked randomization is a better choice since it balances the animals within each block as intended. Blocked randomization is a natural complement for blocked designs (see above), covering also in the simplest case: a whole experiment understood as one block.

6.11 Sample Size Estimation The goal of estimating sample size in the planning of experiments is not specifically to reduce the number of animals used in the experiment. Rather, it is to estimate the minimum number of animals required to accomplish the research goals (i.e., to reliably determine whether an important effect exists). Khamis (1997, p. 55). Sample size – in case of an animal imaging experiment the number of animals needed for the experiment – has to be decided on before the experiment starts. Two main strategies for determining sample size have been described (Festing and Altman 2002), one often labeled resource equation method (REM), the other power analysis method. For determining sample size by the power analysis method, one has to have chosen a design and the primary endpoint. Depending on the design, the intended ratios between the sample sizes for each group have to Table 6.2 Sample size for a one-sided t-test A B

be set. One has to know the relevant difference D considered to be what Khamis calls the “important effect” and the to-be-expected variability s. Both the intended significance level a and power (1-b) must be fixed. The sample size necessary for the experiment can then be estimated by using either tabulated values (like in Machin et al. 1997), published sample size formulas (like in Horn and Vollandt 1999), or specialized software like SAS® proc POWER or nQuery Advisor®. Example 12 – Sample size for a one-sided t-test: Let us consider a two-group comparison for a Gaussian distributed endpoint by assuming the same variability within both groups. Both groups shall have the same number of animals (sample size ratio = 1). The comparison could then be conducted by using Student’s t-test. Suppose intervention A would give a mean response of 12 units and one were interested if intervention B could increase the response up to 20 units, for both interventions a (within-group) variability (standard deviation) of s = 8 units can be realistically assumed based on prior knowledge. Then the effect of interest D would become D = mB – mA = 20–12 = 8 units, an effect of the same size as the standard deviation s =8 units (commonly expressed as an effect size of s = 1). Asking for a nominal power (1-b) of 80% and a nominal significance level a of 5% for groups of the same size would give a minimum of 14 animals per group, i.e., 28 animals overall. Some alterations of the request can be found in Table 6.2. As one can see, increasing power, lowering the significance level, an increased variability, or a smaller effect of interest require larger sample sizes. If drop-out has to be assumed for the experiment, the sample size has to be increased by the reciprocal of the anticipated drop-out rate. It might be most common to have all groups of the same size (all sample size ratios being one), but for many-to-one comparisons (two or more different

Effect D

Standard dev s

Effect size d

Significance level a

Power (1-b)

n per group

As described

12

20

8

8

1

0.05

80

14

Increased power

12

20

8

8

1

0.05

90

18

Lower significance level

12

20

8

8

1

0.01

80

22

Smaller effect D

12

16

4

8

0.5

0.05

80

51

Larger variability

12

20

8

16

0.5

0.05

80

51

79

6 Statistical Considerations for Animal Imaging Studies

interventions all compared to one control) a larger number of animals for the control group might be advisable (Horn and Vollandt 1999, p. 36). Compared with the power analysis, the REM is simple but quite crude. The REM depends on the law of diminishing returns and is based on the suggestion of Mead (1988, p. 587). One has to know the number of different interventions (T), the number of blocks (B), and a third number E recommended by Mead to be between 10 and 20 representing the variability. The number of animals needed should be calculated by N = (T−1) + (B−1) + E + 1. Dividing N by the number of interventions might give a noninteger number to be allocated to each group. These nonintegers should be rounded up. Before applying this approach, it might be useful to consider that REM-based sample size estimations remain the same regardless of any changes in • • • •

Desired significance level Intended power of the experiment Variability Desired important effect to be shown

Therefore, it might be worthwhile to compare REMbased sample size estimations with power analysis ones even in cases when the latter can be based only on very weak assumptions on relevant difference D or expected variability s. Sample sizes determined have to be checked for compliance with any legal and animal welfare regulations (e.g., Australian Government 2008), available facilities, and costs before actually starting the experiment. “The number of animals used in an experiment should be the minimum sufficient to answer the question posed.” (NC3RS 2008).

6.12 Study Design and Research Plan Whatever the decisions on the particular design issues are, all of them shall be addressed in the research plan (study protocol). Many of the aforementioned issues might be difficult to decide on. It is always a good start to check the literature for similar experiments and the design used, effect and variability estimates, as well as any hints about the difficulties of using that design. Depending on the particular scientific question, there are some recommendations worthwhile to consider on

how to deal with factors that might have influence but one is not interested in: • Leave out factors by keeping them constant, if possible; e.g., a male-only experiment, based on the same number of animals, is likely to show an effect more clearly – if there is one – than a mixed malefemale experiment. • If one cannot leave a factor out, block for it, if the factor is measurable before the experiment starts. • Randomize the experiment, randomize in blocks, describe exactly in advance how animals will be randomized. • Decide on an adequate sample size, i.e., taking into account the experiment’s objective, resources needed, and lab animal welfare. In principle, a well-designed experiment has to be repeatable under exactly the same (as well as modified) conditions. The conditions are set in the research plan. Every experiment is to more or lesser extent prone to both error (e.g., application of the wrong substance) and unforeseeable incidents (e.g., animals catching or already bearing an infection). Design as complex as necessary, as simple as possible. Depending on experience and prior knowledge as well as the complexity of the intended experiment, developing the appropriate design takes considerable time. The more complex the design (and therefore the experiment), the more consulting a biostatistician during design phase is recommended.

6.13 Practical Considerations for Conduct and Evaluation of Experiments in Animal Imaging 6.13.1 Data Structure Setting up the experiment logistics shall include the development of an appropriate data structure. Valuable criteria for appropriateness, besides the self-evident correctness, are convenience for data input and usefulness for statistical evaluation. Each record (row in a data table) has to be un equivocally identified (by animal number, time point of measurement, etc.).

80

Convenience for data input during study conduct inhibits the error rate. Codes instead of lengthier text entries support transparency of data handling. Entering the data in the same sequence as the measurements are to be taken prevents from assignment mistakes; eventually the unequivocally identified records can be reordered by any sorting procedure. Since it is always possible that a measurement cannot be taken (because of measurement device failure or early drop-out), codes have to be foreseen for “missing” values. (A period [.] is usually a better choice than a particular number like 0 or −9 because it cannot be confused with a valid value while generating means or the like.) Usually there is more than one possibility to structure data of a particular experiment correctly. Transforming data from one into another structure is possible but laborious. Depending on the statistical evaluation strategy and the statistics software intended for use, there are more as well as less favorable among all available data structures. For both data clarity and smooth transfer into statistics software, it is strongly recommended to keep the raw data (measured data and record identification) separate from any evaluation such as means, counts, or graphs. Using a spreadsheet program, this can simply be accomplished by reserving one sheet for the raw data alone; graphs and means (and virtual copies of the raw data) can then be compiled in any other sheet by referencing to the raw data sheet.

6.13.2 Document the Unexpected While conducting an experiment, it is always possible to encounter a situation not foreseen in the planning phase: allocation errors might occur; animals might react strangely, get eczema, or move because of premature recovery from anesthesia. It is recommended to document these events immediately, since they might occur again and become a remarkable (interesting although unwanted) result of the experiment.

6.13.3 Changes in Study Design If the number of unexpected occurrences increases, one might be inclined to change the study design. In

H.-F. Ulbrich

general that is not recommended. During the experiment there is no time for either sound analysis of reasons for the unexpected or solid redesign. Changes must be documented in both the study plan and the report. Taking the efforts of the design stage into account it is clear that changes must be the exception, not the rule, because they are complex to implement, have statistical consequences requiring careful consideration, and may have an impact on the credibility of results, as it is difficult to prove that a change has not been made to favor the results hoped for by researchers (see assessment bias). Almost certainly it is better to conduct the welldesigned experiment as planned to get as consistent as possible information including about what went wrong than to end up with a lot of information generated under nonrepeatable conditions.

6.14 Statistical Evaluation of Animal Imaging Experiments Results of an experiment should be assessed by an appropriate statistical analysis reflecting the purpose of the experiment, even if in some cases the results of the experiment appear to be so clear-cut that they seem obvious (in either or the other direction). Evaluating the animal imaging experiment should be quite easy if the experiment was well planned, the data adequately recorded, and the experiment went fully according to that plan without any unexpected irregularities. Any analysis shall begin with examining the raw data (as well as the data derived by a formula) for consistency, easily and most efficiently performed by using graphical methods.

6.15 Outliers Looking at the data one might find some data points outside the range of prior expectation and, therefore, be tempted to declare them “outlier.” Simply excluding them is a highhanded step, it reduces the number of valid values and is likely to change the effect (by generating bias) in an unjust manner. Therefore, any exceptional looking value should be declared “outlier” if and only if there were independent

81

6 Statistical Considerations for Animal Imaging Studies

evidence that the information is incorrect (e.g., because of measurement device failure or concomitant animal condition like an infection). These exceptional looking values shall be documented including the reason for declaring them “outlier;” if it is clear that the reason is independent of the interventions, these values should be excluded from the analysis.

6.16 Hints on Evaluating the Data Evaluating the experiment and describing its results shall consist of • Description and presentation of the raw data, i.e., the data as measured or obtained, and the derived data as well, depending on the amount of data and the type of each endpoint in either single case or aggregated way (e.g., by the number of values, minimum, maximum, mean, standard deviation, median, or mode). • Estimation of the effect(s) and its variability, presented by the estimates and confidence intervals. • Calculation of the planned test statistics and related p-values. • Formulation of the statistical models built, reporting the units the models, and therefore, the coefficients are based on. • The measures of model checking. • Interpretation of the statistical results from the experimenter’s perspective to answer the main as well as other scientific questions the experiment was actually designed for. The extent to which each of the topics applies depends on both the design of the experiment and the intended audience for the report. It is rarely possible to omit the first and last of the above topics. Graphing the data whenever you can is a good advice. Graphs or, more specifically, charts are one of the primary means of exploration and communication in the practice of science. Charts have advantages for communicating experimental and statistical findings. They make it easy to observe magnitude and direction of differences or trends, to assess the importance of effects, or to check for deviances from the assumptions made. First addressee of the charts shall be the experimenter or the experimental team itself. To develop appropriate charts, one needs to take into account the type and number of endpoint(s) to be

displayed. Whether the chart’s objective is composition (univariate) or comparison of values across groups is another crucial point for selecting a graph. It is recommended to sketch the graphs needed for evaluation by hand and to check them beforehand for completeness and consistency. Completeness covers topics such as titles, axis labels including units, legends, and the information how many data points contribute to the chart; consistency includes whether different charts use the same axis scale or whether a unique color scheme is applied across graphs. Finding software capable of producing the chosen graphs and charts might be a challenge. Translating statistical findings into biological meaning is the first step of reporting results. Tables, graphs, and the narrative shall support each other. The Chicago Guide to Writing about Multivariate Analysis by Miller (2005, the outstandingly extended version of Miller 2004) gives a lot of advice on how to organize data and findings for presentation. While interpreting statistical significance, one should again be aware of the strong relationship between a-level, power (1-b), sample size, variability s, and the effect D considered relevant. Biological (or substantial) relevance and statistical significance are no synonyms; the latter should undoubtedly support to establish the former. Convincing others of the importance of the experiment’s findings should be based on one’s own confidence that a replication of the experiment would allow for the same conclusions, although the magnitudes will be slightly different. This confidence is to largest extent based on the experiment’s design. And therefore, in other words: Experiments using laboratory animals should be well designed, efficiently executed, correctly analyzed, clearly presented, and correctly interpreted if they are to be ethically acceptable. Festing and Altman (2002)

Reference and further Reading There are at least half a dozen good textbooks available on experimental design in general (e.g., Atkinson et al. 2007), some of them being updated by new editions every couple of years (e.g., Montgomery 2005). Many of them take their illustrative examples from a particular field of science (from animal science

82 to industrial quality control). Since there is no textbook (yet) directly dedicated to the design of experiments in animal imaging studies, Chow and Liu (1999), Festing (2002b) or Festing et al. (2002) might be a good choice.

References Atkinson A, Donev A, Tobias R (2007) Optimum experimental designs, with SAS. OUP, Oxford. ISBN 978-0-19-929660-6 Australian Government (2008) Guidelines to promote the wellbeing of animals used for scientific purposes – The assessment and alleviation of pain and distress in research animals. http://www.nhmrc.gov.au/publications/synopses/_files/ ea18.pdf. Accessed 31 Mar 2009 Bland M (2008) Directory of randomisation software and services – Randomisation programs. http://www-users.york. ac.uk/~mb55/guide/randsery.htm. Last updated: 19 Mar 2008, Accessed 27 Sept 2009 Box GEP, Hunter JS, Hunter WG (2005) Statistics for experimenters, 2nd edn. Wiley, New York Chow S-C, Liu J-P (1999) Design and analysis of animal studies in pharmaceutical development. Chapman & Hall, London Cox DR (1958) Planning experiments. Wiley, New York Festing MFW (2002a) Introduction. The design and statistical analysis of animal experiments. ILAR J 43:191–193 Festing MFW (ed.) (2002b) Experimental design and statistics in biomedical research (Special Issue). ILAR J 43(4): 191−258 Festing MFW, Altman DG (2002) Guidelines for the design and statistical analysis of experiments using laboratory animals. ILAR J 43:244–258

H.-F. Ulbrich Festing MFW, Overend P, Gaines Das R, Cortina Borja M, Berdoy M (2002) The design of animal experiments: Reducing the use of animals in research through better experimental design. Royal Society of Medicine, London Hanfelt JJ (1997) Statistical approaches to experimental design and data analysis of in vivo studies. Breast Cancer Res Treat 46:279–302 Horn M, Vollandt R (1999) A survey of sample size formulas for pairwise and many-one multiple comparisons in the parametric, nonparametric and binomial case. Biom J 42: 27–44 Keene ON (1995) The log transformation is special. Stat Med 14:811–819 Khamis HJ (1997) Statistics and the issue of animal numbers in research. Contemp Top Lab Anim Sci 36:54–59 Machin D, Campbell M, Fayers P, Pinol A (1997) Sample size tables for clinical studies, 2nd edn. Blackwell, Oxford Mead R (1988) The design of experiments: Statistical principles for practical application. Cambridge University, Cambridge Miller JE (2004) The Chicago guide to writing about numbers. Chicago University, Chicago Miller JE (2005) The Chicago guide to writing about multivariate analysis. Chicago University, Chicago Montgomery DC (2005) Design and analysis of experiments, 6th edn. Wiley, Hoboken NC3RS – National Centre for the Replacement, Refinement and Reduction of Animals in Research (2008) Responsibility in the use of animals in bioscience research. http://www.nc3rs. org.uk/page.asp?id=871. Accessed 07 June 2009

7

Animal Anesthesia and Monitoring Marc Hein, Anna B. Roehl, and René H. Tolba

Contents 7.1 Anesthesia................................................................. 83 7.1.1 Definition.................................................................... 83 7.1.2 Legal Basis in the European Union (EU) and the United States of America (USA)................... 83 7.1.3 General Considerations............................................... 84 7.1.4 Anesthetic Methods.................................................... 85 7.1.5 Volatile Anesthetics in Rodent Anesthesia................. 85 7.1.6 Anesthetics.................................................................. 88 7.2 Analgetics................................................................... 89 7.2.1 Introduction................................................................. 89 7.2.2 Definition.................................................................... 89 7.2.3 Sedation...................................................................... 90 References............................................................................ 91

7.1 Anesthesia 7.1.1 Definition Anesthesia is a pharmacologically induced reversible state of amnesia, analgesia, loss of consciousness, loss of skeletal muscle reflexes, and decreased stress response.

7.1.2 Legal Basis in the European Union (EU) and the United States of America (USA)

Suggested Reading.............................................................. 91

7.1.2.1 European Union Directive 86/609/EEC on the protection of animals used for experimental and other scientific purposes (27.11.1986) Article 8 1. All experiments shall be carried out under general or local anesthesia. 2. Paragraph 1 above does not apply when: (a) Anesthesia is judged to be more traumatic to the animal than the experiment itself; (b) Anesthesia is incompatible with the object of the experiment. In such cases, appropriate legislative and/or administrative measures shall be taken to ensure that no such experiment is carried out unnecessarily.

M. Hein, A.B. Roehl, and R.H. Tolba (*) Universitätsklinikum Aachen, Pauwelsstrasse 30, 52074 Aachen, Germany e-mail: [email protected]

nesthesia should be used in the case of seriA ous injuries, which may cause severe pain. 3. If anesthesia is not possible, analgesics or other appropriate methods should be used in order to ensure as far as possible that pain, suffering,

F. Kiessling and B.J. Pichler (eds.), Small Animal Imaging, DOI: 10.1007/978-3-642-12945-2_7, © Springer-Verlag Berlin Heidelberg 2011

83

84

distress, or harm is limited and that in any event, the animal is not subject to severe pain, distress, or suffering. 4. Provided such action is compatible with the object of the experiment, an anesthetized animal that suffers considerable pain once anesthesia has worn off, shall be treated in good time with pain-relieving means or, if this is not possible, shall be immediately killed by a humane method.

7.1.2.2 The United States of America USDA Animal Welfare Act Regulations §2.31(d)(1) (iv)(A) and (ix): Procedures that may cause more than momentary or slight pain or distress to the animals will be performed with appropriate sedatives, analgesics, or anesthetics, unless withholding such agents is justified for scientific reasons, in writing, by the principal investigator and will continue for only the necessary period of time. Activities that involve surgery include appropriate provision for preoperative and postoperative care of the animals in accordance with established veterinary medical and nursing practices. PHS Policy IV.C.1.a-b: Procedures with animals will avoid or minimize discomfort, distress, and pain to the animals, consistent with sound research design. Procedures that may cause more than momentary or slight pain or distress to the animals will be performed with appropriate sedation, analgesia, or anesthesia, unless the procedure is justified for scientific reasons in writing by the investigator. Guide for the Care and Use of Laboratory Animals, p. 64: An integral component of veterinary medical care is prevention or alleviation of pain associated with procedural and surgical protocols…. Pain is a stressor and, if not relieved, can lead to unacceptable levels of stress and distress in animals. The proper use of anesthetics and analgesics in research animals is an ethical and scientific imperative… In general, unless the contrary is known or established, it should be assumed that procedures that cause pain in humans also cause pain in animals.

M. Hein et al.

7.1.3 General Considerations Anesthesia in laboratory animals is depending on: • Species • Body weight • Application route of the anesthetic drugs (i.p., i.m., s.c. etc.) • Physical examination with or without surgical procedure • Duration of the procedure (short term vs. long term) • Survival experiment vs. final experiment • Pain exposure • Compatibility of the anesthetic drugs with the experimental goal • Personal and technical equipment available Starvation or deprivation of food and water is not recommended routinely in laboratory animals below 2 kg of body weight. Small laboratory animals like rodents develop easily a hypoglycaemia or a metabolic acidosis. A hypoglycaemia slows down the metabolism of anesthetic drugs and prolongs recovery after anesthesia (Ref. 3 S. 143). If the experimental setting is depending on an empty gastric and bowel lumen, full absorbable diets are recommended. In special field of research, where a mild hypoglycaemia is needed, food deprivation can be performed with close surveillance for 8–12 h before induction of anesthesia.

7.1.3.1 Hypothermia Small laboratory animals develop very easy hypothermia. Hypothermia is defined as body temperature below 35°C (95°F). Reasons are reduced muscle activity, low temperature in the laboratory, alcoholic disinfection, anesthetic side effects (inhibition / interaction with the central nervous thermoregulation system), and peripheral vasodilatation. Hypothermia is killer no 1 in small laboratory animals. Animals will develop hypoventilation and hypercapnia, followed by a metabolic acidosis, lowering of blood pressure, bradycardia, shock etc. Preventive measures: • Check the temperature of the animal • Usage of warming pads with feedback control mechanism (avoid overheating), covering with blankets

85

7 Animal Anesthesia and Monitoring

• Usage of prewarmed fluids • Keep the anesthesia as long as needed and as short as possible

7.1.4.2 Pharmacological Anesthesia

7.1.4 Anesthetic Methods

Any anesthesia administered of permitting performance of a diagnostic or therapeutic procedure with loss of consciousness and loss of skeletal muscle reflexes. This requires Stadium III1 according to Guedel and does not allow a surgical procedure. Nevertheless, in most cases, this is the anesthetic approach of choice in imaging laboratory animals.

7.1.4.1 Surgical Anesthesia

Mode of Administration

Any anesthesia administered of permitting performance of an operative procedure, as differentiated from diagnostic and therapeutic anesthesia, leads to loss of sensation with muscle relaxation adequate for an operative procedure. This requires Stadium III2 according to Guedel in humans, which is difficult to adapt to small animals. Alternatively, responses to different stimuli could be used to judge anesthetic depth. With increased anesthetic depth, stimuli with increasing intense (noise, pain, intubation) are needed to provoke responses, which were attenuated at different levels (Fig. 7.1). In the majority of cases, the absence of limb movement after tail pressure should provide a sufficient anesthetic level for most procedures. The loss of hemodynamic responses (increase of blood pressure or heart rate) or apnea in spontaneous breathing animals might indicate a level, which is too deep. On the other hand, the return of righting reflex will indicate a good recovery from anesthesia.

Drugs given to induce or maintain general anesthesia are either given as: Gases (inhalational anesthetics) Injections anesthetics

Keeping the animal warm is much better and easy in contrast to warm it up after the procedure.

It is possible to combine these two forms, with an injection given to induce anesthesia and a gas used to maintain it (or vice versa), although it is possible to deliver anesthesia solely by inhalation or injection.

7.1.5 Volatile Anesthetics in Rodent Anesthesia Volatile anesthetics (VA) provide main advantages for rodent anesthesia compared to intravenous or intraperitoneal applications of hypnotic drugs. In duction and emergence from anesthesia and delivered concentration can easily be controlled, resulting in Responses

Positioning/ noise

Fig. 7.1 Matrix of relevant stimuli and responses to judge anesthetic depth. Stimuli are in approximately increasing order of noxiousness and responses in increasing difficulty of suppression

Pain Cornea Intubation

tail pressure tail clamp

increasing anesthetic depth

heart rate

blood pressure

corneal reflex

contraction

Hemodynamics

abdominal

Stimuli

limb movement

righting reflex

Movement

86

M. Hein et al.

low cardiopulmonary complications and mortality, especially if only limited monitoring of the animal is available (Hacker et al. 2005). VA are lipophilic halogenated ethers, which are liquid at room temperature and thus require special vaporizers for safe application in spontaneous breathing or mechanically ventilated animals. VA differ in their hypnotic potency, which correlates linear with their fat solubility (Meyer-Overton hypothesis). This correlation is des cribed by the minimum alveolar concentration (MAC) that is needed to prevent movement in 50% of the subjects in response to a painful (surgical) stimulus (Table 7.1). The MAC value is different between species, strains, and dependent of age and gender. Even the inter-individual variability ranges between 4 and 10% and should be specified in case of doubt if different drugs are compared at equipotent anesthetics levels. Furthermore, body temperature and duration of anesthesia and repetitive application will influence the MAC (Stratmann et al. 2009). The end expiratory gas concentration, which corresponds to the alveolar and thus blood partial pressure, can be taken as a measure for brain partial pressure under steady state conditions. VA concentrations can be measured by infrared absorption spectroscopy with analyzers in the main or side stream of the respiratory circuit. Thus, the target drug dosage can be controlled

during anesthesia and easily adjusted if needed. The loss of consciousness normally occurs at 0.5 MAC and can be judged by the loss of the righting reflex. For mono-anesthesia with VA, twofold higher values are needed if 95% of the animals should not respond to painful stimuli or righting. The combination with other anesthetics like opioids, barbiturates, ketamine, midazolame, or a-agonists might reduce the MAC values by 50–90% and should be very carefully adjusted. The uptake of VA depends on differences in partial pressure between lung and blood, solubility, and cardiac output. The blood/gas partition coefficient (lBG), or “blood solubility” correlates inversely with the induction rate (Table 7.1). The high cardiac output related to body weight of rodents accelerates the uptake compared with large animals or humans. Induction and emergence are additionally influenced by the fat/blood (lFB) and brain/blood (lBB) partition coefficients and continuous duration of application. High values delay especially the recovery from anesthesia (Table 7.1). VA demonstrates additional analgesic and muscle relaxant properties with different characteristics. All VA depress the pulmonary and cardiovascular function in a dose-dependent manner, with qualitative and quantitative differences. Respiratory depth

Table 7.1 Properties of commonly used volatile anesthetics in rodents Halothane Isoflurane

Sevoflurane

Desflurane

lblood-gas

2.5

1.43

0.7

0.4

lbrain-blood

1.9

1.6

1.7

1.3

lfat-blood

51

45

47.5

27.2

Metabolism (%)

10–20

0.2

2

0.02

MAC (%)

1.1–1.2

1.2–2.4

1.9–3.7

5.7

LORR (%)

0.6–0.7

0.6–1

1.1–1.4

3.7

LD50/ED50

2.2

4.3

–

2.5

Apnoec Index

2.3

3.1

2.5

1.6

Induction (%)

5

6–8

Maintenance (%)

1.5–2.5

3–4

Recovery (1.6 MAC, 2 h) RORR [min]

25

10

6.2

4.4

ROMC [min]

56

36

15

9.8

l partition coefficients; MAC minimum alveolar concentration; LORR loss of righting reflex; LD50/ED50 ratio of median lethal and effect dosage; RORR return of righting reflex; ROMC return of muscle coordination (Komatsu et al. 1997; Petrenko et al. 2007; Pancaro et al. 2005; Kissin et al. 1983; Eger and Johnson 1987)

87

7 Animal Anesthesia and Monitoring

(tidal volume) is commonly reduced and partially compensated by an increase in respiratory frequency (RF). At higher concentrations (>1.5 MAC), RF decreases and the response to carbon dioxide tension is suppressed, as a consequence of the inhibition of brain stem neurons. This leads to a reduction of minute volume resulting in hypercapnia. The respiratory depression can partly be antagonized by surgical stimulation. Respiratory complications include apnoea at higher concentrations (>1.6–3.1fold of MAC, or “apnoeic index”) and during induction, coughing, salivation, and broncho or laryngospasm (Table 7.1). In general, concentrations of VA for relevant cardiovascular depression are about 50% higher than for respiratory depression. Thus, surveillance of respiration in spontaneous breathing animals could avoid cardiovascular complications. No increase of heart rate to a noxious stimulus appears first at 2.1–2.9 MAC. This characteristic explains the good safety margins of VA as judged by the ratio of median lethal to the effective dosage (LD50/ED50, Table 7.1). Despite this, the drugs differ in their negative inotropic and vasodilatory potency. All of them lead to a dosedependent reduction of myocardial contractility, which reaches a significant level above 1.3–1.5 MAC (Preckel et al. 2002; Heerdt et al. 1998; Hettrick et al. 1996). No differences between the drugs exist, if con centrations are adjusted for the corresponding MAC values (Kanaya et al. 1998). Vasodilatation increases up to 0.9 MAC while maintaining atrioventricular coupling and cardiac output. All substances increase brain perfusion, which could lead to an increase of intracranial pressure. VA are triggers of malignant hyperthermia, but to our knowledge only described for special inbred strains in rodents. At least VA provide organ protective properties. They are able to increase tolerance to ischemia and reduce reperfusion injury if applicated before (preconditioning) or at time of reperfusion (postconditioning). This phenomenon is described in the heart, brain, kidney, liver, and lung (Weber et al. 2005; Matchett et al. 2009; Schmidt et al. 2007; Reutershan et al. 2006).

7.1.5.1 Halothane Halothane is a core medicine in the WHO’s “Essential Drug List” but, is superseded by newer inhalational

anesthetic agents for human use. It provides good hypnotic but minor analgetic and muscle relaxant properties. It is the only drug sensibilizing the heart for catecholamines and may induce complex ventricular extrasystoles. Due to its high solubility, induction and emergence from anesthesia are slow. Halothane shows the highest metabolism rate (ca. 30%) and thus the highest toxicity, among other hepatitis.

7.1.5.2 Isoflurane Isoflurane is the most commonly used VA in veterinarian and human anesthesia. It is characterized by a good hypnotic, analgesic, and relaxant action. It is the drug with the highest safety margin. Induction and emergence from anesthesia are faster compared to halothane. In mice, an opisthotonus could occur during the onset and offset of anesthesia. Isoflurane induces bronchodilatation but increases pulmonary vascular resistance. The high apnoeic index demonstrates the low depression of the respiratory function.

7.1.5.3 Sevoflurane Sevoflurane is a sweet-smelling VA, which is frequently used for induction in pediatric anesthesia. The analgesic and muscle relaxant component of the drug is less compared to isoflurane. No hepatic toxicity was found. Fluorides from metabolism and compound A, which develops in dry and warm carbon dioxide absorbents, are able to induce renal injury. Vasodilation is less expressed than during isoflurane application (Conzen et al. 1992). Respiratory depression is less pronounced compared with isoflurane, although the apnoeic index is lower (Groeben et al. 2004). The incidence of breath holding during induction correlates with the applicated concentration (Pancaro et al. 2005).

7.1.5.4 Desflurane Desflurane has the lowest hypnotic potency and thus the highest MAC value. Therefore, Desflurane has the most rapid onset and offset of the VA, which is nearly regardless of duration and concentration. Thus, it is an ideal agent for repeated use, especially in functional brain

88

M. Hein et al.

imaging. Despite its low toxicity, it induces the highest respiratory depression, with breath holding during induction, bronchi and laryngospasm, coughing, and salivation. Reduction of myocardial contractility and vasodilation is higher than during isoflurane or sevoflurane anesthesia (Hettrick et al. 1996).

7.1.5.5 Anesthetic Management In rodent anesthesia, isoflurane or sevoflurane, which is higher in price, should be preferred related to its low respiratory and cardiovascular depression. Halothane should not be used any more due to its high toxicity and low safety margins. Less experience exists with desflurane in rodents, which is the most expensive drug among the VA. The short recovery from anesthesia, as documented by return of righting reflex and recovery of muscle coordination (Table 7.1) marks the only advantage. Comparable recovery times could be observed after sevoflurane application. For induction anesthesia, chambers with tubing connectors on opposite ends of the box are recommended, which could be connected to an anesthesia machine. Direct application of the liquid agent should be avoided. Chambers should not be larger than needed. The time, which is needed to reach 95% of the concentration set on the vaporizer, could be calculated as follows: T95 [min] = 3 · box volume [1] / gas flow [1 / min]

A fast onset of inhalational anesthesia can be obtained with a fresh gas flow of 5 L/min (100% oxygen) with an inspiratory concentration of 5% isoflurane or 6–8% sevoflurane. For maintenance, concentrations should be reduced to 1.5–2.5% for isoflurane and 3–4% for sevoflurane in spontaneous breathing or mechanical-ventilated rodents. Actively scavenged chambers and coaxial mask with connection to scavenging systems should be used to reduce waste gas and thus exposure of the operator.

7.1.5.6 Airway Management Most studies are performed in spontaneous breathing rodents. This setting requires face masks with a tight seal and minimal dead space to impede rebreathing and

thus hypercapnia. Ideally, coaxial masks are used to scavenge waste gas. Procedures needing a depth anesthesia or including thoracotomy require an intubation of the animal and controlled mechanical ventilation. Oral intubations as well as tracheotomy are feasible procedures. The use of translumination (indirect light from the neck), intubation guides, and an intubation speculum in an upright position (60°) of the animal provides good conditions. Standard intravenous cannulas with 20–14 gauge sizes can be used. Blind intubation should be avoided, because already a marginal mucosal edema might produce a relevant obstruction of the upper airway. Repeated traumatic attempts will increase salivation, which further complicates the procedure. Different rodent anesthesia machines are available. The simplest design contains a flow meter for oxygen and air, a vaporizer, a breathing circuit, a mask, and a scavenging system. It could be expanded with an oxygen flush valve and a pressure control unit. Parallel connection of additional breathing circuits enables the simultaneous use in several animals. For controlled ventilation, special ventilators for small animals as well as human ventilators may be used. Long breathing tubes with a low compliance should be avoided especially if a volume controlled modus is used. Tidal volume (Vt) and RF for rats and mice could be calculated as follows: Vt [ml] = 6.2 · body weight [kg] 1.01, RF [1/min] = 53.5 bodyweight [kg] -0.26.

A pressure-controlled ventilation with a maximum pressure of 10–15 mbar and an end expiratory pressure of 5 mbar will avoid mechanical lung injury. Surveillance of the animal includes inspection of respiratory excursion, capnometry, and pulsoxymetry.

7.1.6 Anesthetics Ketamine (Ketavet®) is an N-Methyl-d-Aspartate (NMDA) antagonist and induces a dissociative anesthesia. It elevates the sympathicus tonus and the muscle tonus. It is not recommended that the substance alone be used (psychedelic during wake up phase, no surgical tolerance in visceral surgery alone). It should be used in combination with other anesthetic drugs (Sedatives, Neuroleptika, and Hypnotics etc.).

89

7 Animal Anesthesia and Monitoring

Ketamine + Xylazine

Ketamine + Medetomidin

Mouse

Rat

Hamster

Guinea pig

80–100 mg/kg (K)

60–90 mg/kg (K)

80–100 mg/kg (K)

40 mg/kg (K)

5–10 mg/kg (X) i.p.

5–10 mg/kg (X) i.p. Duration: 30–45 min

5–10 mg/kg (X) i.p.

5 mg/kg (X) i.p. Duration: 60 min

75 mg/kg (K)

75 mg/kg (K)

80–100 mg/kg (K)

40 mg/kg i.p. oder i.m.

1.0 mg/kg (M) i.p.

0.5 mg/kg (M) i.p oder s.c.

0.25 mg/kg (M) i.p.

0.5 mg/kg (M) i.p.

Sedation 44 mg/kg s.c. 50–100 mg/kg i.p. 40–80 mg/kg i.p. Ketamine application is possible to nearly every route (i.v., s.c., i.p., i.m., oral, rectal)

Pentobarbital (Release ad us. vet; Nembutal®) and thiopental are barbiturates and used either intraperitoneally or intravenously. The therapeutic window is very narrow, and even in the therapeutic range, side effects like respiratory and/or cardiac arrest can occur.

Pentobarbital

Duration: 40 min 22–44 mg/kg i.m.

7.2.2 Definition “Pain” is derived from the Latin “poena,” and has been defined by the Committee on Taxonomy for the International Association for the Study of Pain as “an unpleasant sensory and emotional experience associated with actual or potential tissue damage.”

Mouse

Rat

Hamster

Guinea pig

30–70 mg/kg i.p.

30–55 mg/kg i.p.

40–90 mg/kg i.p.

15–40 mg/kg i.p. Duration: 90 min

Thiopental

25–50 mg/kg i.p.

20–40 mg/kg i.p.

7.2 Analgetics 7.2.1 Introduction The anatomic structures and neurophysiologic mechanisms leading to the perception of pain are similar in humans and non-human animals.Therefore, it is re asonable to assume that if a procedure is painful to humans, is damaging or potentially damaging to tissues, or induces escape and emotional responses in an animal, it must be considered to be painful to that animal.

Carprofen

20–40 mg/kg i.p.

15–20 mg/kg i.v. 15–20 mg/kg i.p.

According to international recommendations, it is a good scientific practice that all procedures that may cause more than momentary or slight pain or distress to the animals will be performed with appropriate sedatives, analgesics, or anesthetics. Carprofen (Rimadyl®) is a non-steroidal antiphlogistic drug with analgetic and antipyretic potency in combination with anti-inflammatory properties. For postoperative pain management, the drug should be given during surgery, because the drug is released with some delay. Meloxicame (Metacam® 5 mg/mL ad us. vet) is also a non-steroidal antiphlogistic drug. It is used for the peri and postoperative pain management and rheumatoid diseases.

Mouse

Rat

Hamster

Guinea pig

4–5 mg/kg s.c.

4–5 mg/kg s.c.

4–5 mg/kg s.c.

4–5 mg/kg s.c.

Duration: 24 h

Duration: 24 h

Duration: 24 h

Duration: 24 h

90

Meloxicame

M. Hein et al. Mouse

Rat

Hamster

10 mg/kg p.o., s.c. (1–2 mg/kg s.c.)

0.2–1 mg/kg p.o., s.c. (1–2 mg/kg s.c)

0.2 mg/kg p.o., s.c. first day

Guinea pig

Application: 1×/day

Application: 1×/day

Application: 1×/day 0.5 mg/kg p.o., s.c.

0.1 mg/kg p.o., s.c. from second day Application: 1×/day

Fentanyl (Fentanyl-Janssen®, Fentanyl-Curamed®) is a short-term Opioid-Analgetic drug 100 times stronger than morphine. In combination with a hypnotic drug, it is possible to reduce the fentanyle dosage by ca. 50%.

Fentanyl

is not recommended for diabetics, it is contraindicated in patients with cardiac disease. Following administration, marked peripheral vasoconstriction and bradycardia are noted. Dosage of induction

Mouse

Rat

Hamster

Guinea pig

0.025–0.6 mg/kg s.c.

0.01–1.0 mg/kg s.c.

k.A.

n.a.

2.0–4.0 g/day p.o.

Buprenorphine (Temgesic®) is a semi-synthetic opiate with partial agonist actions at the m opioid receptor and antagonist actions at other opioid receptors. Buprenorphine is a thebaine derivative with analgesia approximately 25–40 times as potent as morphine.

Buprenorphine

agents (i.e. propofol) can be drastically reduced, as may the volumes of anesthetic gases (i.e. halothane, isoflurane, sevoflurane) used to maintain general anesthesia.

Mouse

Rat

Hamster

Guinea pig

0.05–0.1 mg/kg s.c., i.p.

0.01–0.05 mg/kg s.c., i.p.

0.05–0.1 mg/kg s.c., i.p.

0.05 mg/kg s.c.

Duration: 8–12 h

Duration: 6–12 h

Duration: 8–12 h

1.1 mM in DMSO topic

0.4 mg/kg p.o.

Dosage interval: 3×/day s.c.

High first-pass effect; dosage is ten times higher given orally. Local Anesthetics: Infiltration of wounds, intercostal spaces etc. would provide long-term analgesia with minimal systemic side effects. Bupivacaine or ropivacaine is an ideal drug for this purpose. Beyond a cumulative dosage of 5 mg/kg bupivacaine, systemic side effects should be considered, whereas no side effects could be observed after s.c. application of 25 mg/kg ropivacain in rats (Danielsson et al. 1997).

7.2.3 Sedation Medetomidine (Domitor®) Medetomidine (active form medetomidine hydrochloride) is a specific a2adrenergic agonist. Noradrenaline release is reduced, and this results in sedation and anxiolyses. It can be given by intramuscular injection (IM), subcutaneous injection (SC), or intravenous injection (IV). It

Duration: 6–12 h

Combination with Ketamine is very common. Medetomidine

Mouse

Rat

Hamster

Guinea pig

30–100 mg/kg s.c.

30–100 mg/kg s.c.

n.a.

n.a.

The effects of Medetomidine can be reversed using Atipamezole (Antisedan®) 1.0 mg/kg s.c., i.p. or i.v. Xylazine (Rompun®) is an agonist at the a2-adrenergic receptor. It is ten times less selective then Medetomidine. As with other a2 agonists, adverse effects include bradycardia, conduction disturbances, and myocardial depression. The effects of Xylazien can be reversed using Yohimbine (Yobine) or atipamezole (Antisedan). In laboratory animal anesthesia, xylazine is often used in combination with ketamine. Mouse

Rat

Hamster

Guinea pig

Xylazine

5–10 mg/kg i.m., i.p., s.c.

5–10 mg/kg i.m., i.p., s.c.

n.a.

5 mg/kg i.m., i.p.

Xylazine + Ketamine

See Ketamine

See Ketamine

See Ketamine

See Ketamine

n.a. not available

91

7 Animal Anesthesia and Monitoring

Midazolam (Dormicum®) is a benzodiazepine, which provides dose-dependent sedation, hypnosis, and muscle relaxation without analgesia. It is commonly used for premedication (1–2 mg/kg i.m.). If it is combined with ketamine (5–10 mg/kg i.m.), the dosage should be reduced to 0.5–1 mg/kg i.m. Its action could be antagonized with flumazenil. Modified according to: Duke University and Medical Centre: Guidelines for Rodent Analgesia Guidelines for the Assessment and Management of Pain in Rodents and Rabbits, July 2006 Guidelines for Anesthesia and Analgesia in Rodents and Rabbits; April 2009 University of South Florida: Guidelines on Anesthesia and Analgesia in Laboratory Animals University of Pennsylvania: IACUC Guideline Rodent Anesthesia & Analgesia Formulary, May 2008 CliniPharm Wirkstoffprofile, Stand 09.12.09 GV-SOLAS, Ausschuss für Anästhesie und Analgesie: Schmerztherapie bei Versuchstieren http://www.iasp-pain.org/trms-p.html.

References Conzen PF, Vollmar B, Habazettl H, Frink EJ, Peter K, Messmer K (1992) Systemic and regional hemodynamics of isoflurane and sevoflurane in rats. Anesth Analg 74:79–88 Danielsson BR, Danielson MK, Boo EL, Arvidsson T, Halldin MM (1997) Toxicity of bupivacaine and ropivacaine in relation to free plasma concentrations in pregnant rats: a comparative study. Pharmacol Toxicol 81:90–96 Eger EI II, Johnson BH (1987) Rates of awakening from anesthesia with I-653, halothane, isoflurane, and sevoflurane: a test of the effect of anesthetic concentration and duration in rats. Anesth Analg 66:977–982 Groeben H, Meier S, Tankersley CG, Mitzner W, Brown RH (2004) Influence of volatile anaesthetics on hypercapnoeic ventilatory responses in mice with blunted respiratory drive. Br J Anaesth 92:697–703 Hacker SO, White CE, Black IH (2005) A comparison of targetcontrolled infusion versus volatile inhalant anesthesia for heart rate, respiratory rate, and recovery time in a rat model. Contemp Top Lab Anim Sci 44:7–12 Heerdt PM, Gandhi CD, Dickstein ML (1998) Disparity of isoflurane effects on left and right ventricular afterload and hydraulic power generation in swine. Anesth Analg 87:511–521

Hettrick DA, Pagel PS, Warltier DC (1996) Desflurane, sevoflurane, and isoflurane impair canine left ventricular-arterial coupling and mechanical efficiency. Anesthesiology 85:403–413 Kanaya N, Kawana S, Tsuchida H, Miyamoto A, Ohshika H, Namik A (1998) Comparative myocardial depression of sevoflurane, isoflurane, and halothane in cultured neonatal rat ventricular myocytes. Anesth Analg 87:1041–1047 Kissin I, Morgan PL, Smith LR (1983) Comparison of isoflu rane and halothane safety margins in rats. Anesthesiology 58:556–561 Komatsu H, Nogaya J, Kuratani N, Ueki M, Yokono S, Ogli K (1997) Psychomotor performance during initial stage of exposure to halothane, enflurane, isoflurane and sevoflurane in mice. Clin Exp Pharmacol Physiol 24:706–709 Matchett GA, Allard MW, Martin RD, Zhang JH (2009) Neuroprotective effect of volatile anesthetic agents: molecular mechanisms. Neurol Res 31:128–134 Pancaro C, Giovannoni S, Toscano A, Peduto VA (2005) Apnea during induction of anesthesia with sevoflurane is related to its mode of administration. Can J Anaesth 52:591–594 Petrenko AB, Tsujita M, Kohno T, Sakimura K, Baba H (2007) Mutation of alpha1G T-type calcium channels in mice does not change anesthetic requirements for loss of the righting reflex and minimum alveolar concentration but delays the onset of anesthetic induction. Anesthesiology 106:1177–1185 Preckel B, Ebel D, Mullenheim J, Frassdorf J, Thamer V, Schlack W (2002) The direct myocardial effects of xenon in the dog heart in vivo. Anesth Analg 94:545–551 Reutershan J, Chang D, Hayes JK, Ley K (2006) Protective effects of isoflurane pretreatment in endotoxin-induced lung injury. Anesthesiology 104:511–517 Schmidt R, Tritschler E, Hoetzel A, Loop T, Humar M, Halverscheid L, Geiger KK, Pannen BHJ (2007) Heme oxygenase-1 induction by the clinically used anesthetic isoflurane protects rat livers from ischemia/reperfusion injury. Ann Surg 245:931–942 Stratmann G, Sall JW, Eger EI II, Laster MJ, Bell JS, May LDV, Eilers H, Krause M, Heusen FVD, Gonzalez HE (2009) Increasing the duration of isoflurane anesthesia decreases the minimum alveolar anesthetic concentration in 7-day-old but not in 60-day-old rats. Anesth Analg 109:801–806 Weber NC, Preckel B, Schlack W (2005) The effect of anaesthetics on the myocardium–new insights into myocardial protection. Eur J Anaesthesiol 22:647–657

Suggested Reading A Reference Source for the Recognition & Alleviation of Pain & Distress in Animals Animal Welfare Information Center, United States Department of Agriculture National Agricultural Library (2003) http://www.nal.usda.gov/ awic/pubs/awic200003.htm

8

Drug Administration Klaus Nebendahl† and Peter Hauff

8.1 General

Contents 8.1 General.................................................................... 93 8.2 Principles of Administration.................................. 8.2.1 Handling and Restraining......................................... 8.2.2 Site of Administration.............................................. 8.2.3 Preparation of the Site.............................................. 8.2.4 Safety and Solubility of Substances......................... 8.2.5 Concentration of Substances.................................... 8.2.6 pH of the Injected Solution....................................... 8.2.7 Volume and Frequency of Administration................ 8.2.8 Rate of Absorption/Distribution of Administered Substances.....................................

94 94 94 94 94 95 95 95 95

8.3 Enteral Administration.......................................... 96 8.3.1 Oral Administration (per os)..................................... 96 8.3.2 Intragastric Administration by Gavage..................... 98 8.4 Parenteral Administration..................................... 99 8.4.1 Topical Application.................................................. 100 8.4.2 Inhalation.................................................................. 101 8.5 Administration by Injection.................................. 101 8.5.1 Intradermal Injection (i.d.)........................................ 101 8.5.2 Intramuscular Injection (i.m.)................................... 101 8.5.3 Subcutaneous Injection (s.c.).................................... 103 8.5.4 Intraperitoneal Injection (i.p.)................................... 103 8.5.5 Intravenous Injections (i.v.)...................................... 105 8.5.6 Other Methods for Parenteral Administration.......... 111 References............................................................................ 114

K. Nebendahl† University of Göttingen, 37075 Göttingen, Germany P. Hauff () Global Drug Discovery, Bayer Schering Pharma AG, 13353 Berlin, Germany e-mail: [email protected]

For biomedical research purpose the most commonly used laboratory animals are mice, followed by rats. All over the world the percentage usually varies around 80–90% of the total number of animals used (Tierschutzbericht 2007). For this reason this chapter goes in detail only on these two species. Drugs and other test substances such as chemical elements, compounds, antibodies, cells etc. can be administered to small laboratory animals through a variety of routes. Numerous routes have been well documented in the literature. As every route has both, advantages as well as disadvantages and as, for instance, the absorption, bioavailability and metabolism of the substance are factors which should be considered carefully, a knowledge of available methods and techniques of administration and of the disposition and fate of the administrated substance will aid the scientist to choose the most suitable route for his/her purpose. Therefore, the exact details of any administration must be checked prior to any experiment. A complete review of all methods used for administration would go far beyond the scope of this chapter. For a full discussion of all administration routes available, the reader is referred to the periodical literature. Only some administration routes, techniques and guidelines for sites of administration, preparation of sites, injection techniques, and safe injection volumes will be described here. As a number of questions often arise concerning the suitability of solutions for injection and unfortunately precise answers are often not forthcoming (Waynforth and Flecknell 1992a, b), some remarks are made in respect to volume and rate of injection, absorption of the administrated substance, bioavailability and distribution of substances in the body as well as factors that may modify the dosage etc.

F. Kiessling and B.J. Pichler (eds.), Small Animal Imaging, DOI: 10.1007/978-3-642-12945-2_8, © Springer-Verlag Berlin Heidelberg 2011

93

94

An introduction to this topic would be incomplete without a reference to humane treatment of mice and rats which are the most frequently used animals in experimental studies. The German “Animal Welfare Act” (Tiersch 2006) claimed that pain, suffering and harm may be inflicted on animals only in so far as unavoidable to attain the purpose of the experiment and where unavoidable, they shall be limited to the absolute minimum. The “European convention for the protection of vertebrate animals used for experimental and other scientific purpose” (ETS 123 1986) stated, that persons carrying out procedures shall have had appropriate education and training in handling and restraining experimental animals, should have the foundation for responsible use of the animals and should attain a scientifically high standard, in order to protect the animals used in procedures which may possibly cause pain, suffering or distress and to ensure that, where unavoidable, they shall be kept to a minimum. First and foremost of moral reasons and not because of legal reasons, it must kept in mind that nearly every administration can be performed relative painless, if attention is given to proper restraint of the animal and adequate technical skill is employed.

8.2 Principles of Administration 8.2.1 Handling and Restraining Nearly every animal can be trained to accept handling and restraining and can become familiar with the attending persons. Though time consuming, it is essential to minimise distress. For this reason all procedures should be carried out by persons well known to the animal. Sedation or general anaesthesia is generally only required if the technique involved is more than a pinprick or the administered substances have side effects like causing pain etc.

8.2.2 Site of Administration Numerous sites have been described in the literature to administer substances to small laboratory animals but some of these sites are unacceptable nowadays, for example foot-pad injection of Freund’s complete adjuvant.

K. Nebendahl and P. Hauff

8.2.3 Preparation of the Site Sometimes the area must be clipped or cleaned with warm water. Afterwards, the skin should be swabbed with disinfectants or alcohol. To prevent pain it is sometimes necessary to apply local analgesics to the site before administration.

8.2.4 Safety and Solubility of Substances All parenteral administration must be done using an aseptic technique and the substances or injection solutions must be sterile and free from pyrogenous material. In order to administer an accurate dosage and to avoid causing tissue damage due to the toxicity of the substance, the volume and the way of administration has to be considered. If the substance has to be diluted the selected thinning agent must be safe. Physiological saline (0.9% sodium chloride) or other physiological solvents like phosphate buffered saline (PBS) or various culture media are suitable vehicles. Although distilled water can be used sometimes, saline should be preferred because water ad injectionem injected subcutaneously causes pain and injected intravenously produces haemolysis. For reasons of solubility or rate of absorption some substances require a more complex solvent to render them suitable for administration. Many solvents, for example water, water with 0.85% sodium chloride, 60% polyethylene glycol, 10% Tween 80, 0.5% methylcellulose, have been found suitable in most instances and do not greatly affect the activity of interest of the substances to be investigated due to their own inherent properties (Woodard 1965). All of these vehicles can be administered by any of the injection routes available, but the concentrations mentioned are the maximum practicable, and in many cases it is possible and indeed desirable that lower concentrations should be used (Waynforth and Flecknell 1992a, b). When administrating drugs, the solvent should ideally be the same as the one in which the drug is normally formulated. Lipid-soluble substances can only be dissolved in oil, but their absorption is delayed when administered. As oil can not be injected intravenously, lipophilic substances should be injected in a 15% oil–water emulsion. Oil-based adjuvants or oil–water emulsions given intraperitoneal injected may cause acute peritonitis.

95

8 Drug Administration

Therefore, they should only be administered by this mode when all other routes have proved ineffective. Substances can be injected in the form of a suspension. Since suspended particles have the tendency to sediment, these particles should be evenly distributed before the suspension is injected intravenously. If injected intravenously it should be noted that the particles will be filtered out in the capillary beds of the extremities and the lung, modifying the distribution of the injected material and sometimes causing pulmonary distress to the animal (Waynforth and Flecknell 1992a, b).

8.2.5 Concentration of Substances The concentration can vary over a fairly wide range without greatly influencing the results of the experiments. Lower concentrations are clearly desirable (Waynforth and Flecknell 1992a, b). Factors limiting the use of aqueous solutions for parenteral administration are probably related to their osmotic pressure. Low concentrations can be corrected by the addition of sodium chloride but ought not be so high as to materially exceed the osmotic pressure of 0.15 M sodium chloride (Woodard 1965). Highly concentrated solutions can be administered intravenously provided the rate of injection is kept slow and precaution are taken to avoid getting the solution outside the vein (Shimizu 2004).

8.2.6 pH of the Injected Solution Because extremes in pH cause local tissue damages, solutions should be neutral (pH 7.0–7.3) ideally. But rodents, like humans, tolerate injections of solutions within a fairly wide range of pH. For all routes of administration, a working range is in the range of pH 4.5–8.0 (Woodard 1965). The widest tolerance to pH is shown by the intravenous route because of the buffering capacity of the blood and of the very quick dilution through the side-on flow of venous blood, followed by the intramuscular and then the subcutaneous route. Nevertheless, especially when low or high pH solutions are intravenously injected, the rate of injection must be kept slow and the administration should be done extremely cautious to avoid solutions getting outside the vein.

8.2.7 Volume and Frequency of Administration Though volume and frequency of administration is mainly determined by the requirements of the experiment, the animal should not be strained or stressed. The volume of substances given is limited by their toxicity and by the size of the animal and should be as small as possible. Likewise the frequency of administration should be restricted to a minimum, to avoid unnecessary stress. Volume and rate of injection have to be considered, particularly if solutions are given intravenously, because haemodynamic changes and pulmonary oedema may occur and very rapid injections can produce cardiovascular failure and be lethal. Some recommendations regarding the maximum injection volumes of the Society of Laboratory Animals (GV-SOLAS, August 2010) are listed in the chapter Appendix of this book (pp. 587-588).

8.2.8 Rate of Absorption/Distribution of Administered Substances The rate of absorption is influenced by the blood flow to the site of administration, the nature of the substances and the manner and concentration in which they are presented (Wolfensohn and Lloyd 1994). It influences the time-course of substance effect and is an important factor in determining substance dosage (Waynforth and Flecknell 1992a, b). Only in a few cases is it possible to inject substances at the place where they should be effective. Normally, they must be absorbed from the site of administration into the blood. Therefore, the size of the absorbing surface, the blood flow to the site of administration and the solubility of the substance in the tissue fluids are important factors which determine the rate of absorption. While endogenous compounds normally pass through biological membranes by special transport mechanisms, the penetration of xenobiotics follows physical principles, such as passive diffusion according to the concentration gradient (Frimmer and Lämmler 1977). Lipid solubility, physicochemical properties, degree of ionisation and molecular size of substances are important factors in respect to the rate of absorption. Substances which are highly soluble in body fluids will be absorbed quickly. Substances which are ionised and are not lipid soluble will only be absorbed if a specific carrier exists

96

(Wolfensohn and Lloyd 1994). As absorption of administered substances mostly happens by diffusion down a concentration gradient, the absorption of substances is dependent on the dosage.

8.3 Enteral Administration Enteral administration involves the introduction of substances into the gastrointestinal tract via the mouth or through the anus using a suppository. The latter method is not very practical when working with small laboratory animals (Baumans et al. 1993). The advantage of enteral administration is the fact that it is possible to give comparatively large amounts of non-sterile substances or solutions. For oral preparations, a pH value as low as 3 can be tolerated for a solution of fairly high buffer capacity (Woodard 1965). On the other hand, alkaline solutions are very poorly tolerated by the mouth. In principle, gastrointestinal absorption takes place over the whole length of the digestive tract. Absorption of most orally given substances occurs by diffusion of their non-ionised forms, since the mucosal lining of the gastrointestinal tract is almost impermeable for ionised molecules. Consequently, absorption of substances will be enhanced in the acid stomach or in the nearly neutral intestine depending on the ionic character of the compound. The far larger surface area of the intestinal villi, however, makes intestinal absorption dominant in most cases (Claassen 1994a). Buccal or sublingual administration would avoid the destructive effects that may be encountered in the stomach following oral administration (Woodard 1965). Consequently, this route is sometimes required in animal experimentation. The absorption rate of substances has not been investigated to any considerable extent and is probably insignificant in small rodents. It is known that the mucosa of the mouth cavity can only absorb hydrophobic, non-ionised substances. For substances given orally the stomach is a significant site of absorption for many acidic or neutral compounds, whereas only the weakest bases are absorbed to any appreciable extent at normal gastric pH values (Baggot 1977). All other substances are absorbed only at a very small rate. Using the oral administration route, it should be known that substances can be destroyed by the gastric juice, that the food content of the stomach influences both the rate and order of gastric emptying, and that

K. Nebendahl and P. Hauff

the rate of substance absorption is markedly influenced by its residence time in the stomach and is directly related to the rate at which substances are passed from the stomach into the intestine (Levine 1970). Because of its extensive surface area and rich blood supply, the upper small intestine of rodents is the major site of absorption for all substances after oral administration. Absorption in the intestine is dependent on: (1) the physicochemical state of the substances, (2) the non-absorptive physiological function and state of the intestine, (3) the metabolic activity and function of the absorbing cells, and (4) the structure of the absorbing surface (Levine 1970). Through the efflux of intestinal juice, pancreas juice and bile the intestinal content becomes nearly neutral. This will reduce the degree of ionisation and the absorption rate of substances will increase. Lipid-soluble substances of the digestive content are rapidly absorbed, whereas, dependent on size, the absorption of solids is moderate. Using the oral administration routes, it has to be kept in mind that enzymes of the microflora of the digestive tract can metabolise substances. Under physiological conditions, microorganisms are only to be found in the large intestine and normally such enzymatic metabolism applies only to those substances which are not yet absorbed in the upper tract. Therefore, oral administration can have the disadvantage that enzymes of microorganisms can alter substances before they are absorbed in the bowel. If the metabolites are not biologically active, the administration of these substances may be without any effect. On the other hand, some insoluble substances become soluble as the result of enzyme activity during their passage through the stomach and the small intestine and hence their absorption than become possible in the large intestine. Excluding those absorbed in the mouth and rectum, all substances given and absorbed through the gastrointestinal tract are transported by the portal vein to the liver. Some substances can be metabolised there to a large extent, before reaching the systemic circulation and/or the site of action. This phenomenon is called “first pass effect” (Löscher et al. 2006) and has to be considered in selecting routes of administration.

8.3.1 Oral Administration (per os) The simplest method to administer substances is by mixing them with food or drinking water. However,

97

8 Drug Administration

this method is neither very accurate nor it is practicable with substances which are unpalatable, insoluble or chemically unstable in drinking water or when they irritate the mucosa of the gastrointestinal tract. Giving substances with food or drinking water is an easy method of administration. For example, training of the animal is not necessary because no handling or restraining is required. Furthermore, as the animal is not disturbed, the administration takes place without any stress, and eating, drinking or digestion happens under normal physiological conditions. Thereby, it is secured that only the substances are effective. Mixing substances into meal food must be done carefully and decomposition must be prevented otherwise the animal might refuse to eat the substances or incalculable substance doses will be taken up. As decomposition of the food is only prevented by feeding pellets, this feeding method is recommended. But to prepare an accurate mixture and to pelletise the meal food some technical equipment is necessary. In most cases, it is best to involve a food-stuff company to produce an exact mixture pressed in pellets. The daily food and water intake of the animal must be known before starting so that the amount of substances to be mixed with food or drinking water can be calculated (Table 8.1). Other factors may also come in play. If – for instance – the circadian rhythm influences the effect of a substance, the feeding behaviour of the animal should be known. Mice and rats are not nibblers but spontaneous meal eaters. As they are nocturnally active animals they ingest most of their food during the dark period (Ritskes-Hoitinger 2004). For instance, rats eat only small parts of their food during the day and consume 80% of their daily food intake at night taking 5–8 meals in this time (Claassen 1994b). Fasting is often used, as an empty stomach is considered prerequisite for bioavailability and drug absorption. An empty stomach is as well often required when applying substances per gavage because it is sometimes

Table 8.1 Estimated average fooda intake of mice and rats Growing Adult Mouse Rat Mouse Rat g/day

3–5

8–25

5–7

25–30

considered that the content of the stomach could dilute the test substances or even interact with it. But fasting causes severe changes in physiologic and biochemical processes of the animal, which become more severe with longer duration of food withdraw (Claassen 1994c) and leads to increased locomotory and grooming behaviour (Vermeulen et al. 1997). In consideration of these facts, it must be known how long the animal must or can be starved without any harm to its welfare. For instance, Jeffrey et al. (1987) observed variable amounts of food in the stomachs of male rats following an overnight fast and noted that the diet type and diet regimen can result in variable quantities of food being retained in the stomach. Schlingmann et al. (1997) found that the stomach of rats fasting 6, 12 and 18 h was almost empty and that only food deprivation for 6 h did not cause any distress. As food deprivation will result in a decreased need for water, this feeding schedule leads to a gradual increase of the haematocrit value, with a corresponding decrease in plasma volume. A decrease of the plasma and interstitial volume may result in a significant decrease of substance distribution volume. Various investigators reported, for example that after overnight food restriction, the glutathione content per gram liver diminished by 45–57% in mice and that fasting effects may vary to a great extent between strains and individuals. Therefore, Claassen (1994c) advanced the view that the use of fasting animals in experiments seems only be permitted when the feeding condition functions as a factor in the experimental design. Oral administration by diet or drinking water is not suitable when the exact amount of substance intake is required. Because food and water wastage happens at all time, it is, for instance, impossible to determine the exact amount of diet or water intake of rodents. Therefore, neither the precise food or water intake nor the correct intake of substances with food or water is normally feasible. The only way this can be done is by keeping the animals in metabolic cages and recording the wastage.

Pregnant Mouse

Rat

Lactating Mouse

Rat

6–8

25–35

7–15

35–65

Source: Ritskes-Hoitinger and Chaewalibog (2003) ME content: 8–12 MJ/kg diet. ME metabolizable energy a Food intake can vary dependent on strain, body weight, acivity, condition of holding (temperature and humidity) and other factors

98

The feeding or drinking intake is influenced by conditions of housing and especially by the environment too. Normally, the temperature in rooms for laboratory animals is well-controlled by air conditioning. It should be noted that high temperatures will decrease the intake of food and increase the consumption of water. For instance, the water intake of rats at 22°C is about 20% higher than the food intake, while at 30°C water intake is more than double (Weihe 1987). For this reason the temperature and humidity has to be recorded during the experiment. The substances in food or drinking water can cause the animal to refuse to eat or drink and the animal may lose weight or becomes dehydrated. To stimulate the consumption of rodents the substances can be administered in fat dripping and, if it does not interfere with the experiment, sucrose cubes or a 10% glucose solution can be given. If the substances increase the metabolic rate, overeating must be prevented to avoid an overdose. For these reasons, the animal and its food and water intake must be observed careful. Another disadvantage is that, for observing or measuring food or water intake the animal must be individually housed. Numerous reports about mice and rats refer that this condition provoke ad-libitum-fed rodents to eat more than group-housed animal and that this housing condition may cause stress to animals and will reduce the success of oral administration (Claassen 1994b).

8.3.2 Intragastric Administration by Gavage This route of administration is carried out using a special steel needle. This bulb-tipped gastric gavage needle helps to prevent damaging the oesophagus or stomach and from passing through the glottal opening into the trachea. The diameter of the bulb and the length of the needle will depend on the size of the animal to be gavaged. Before starting administration, the feeding needle is checked to be of suitable length, i.e. reaching from the mouth to the caudal end of the breastbone (outside of the animal) and this should be noted on the tube (Fig. 8.1). The chosen needle and the appropriate syringe filled with the desired amount of solution are linked together and the needle can be moistened with water or oil to make

K. Nebendahl and P. Hauff

Fig. 8.1 Measuring the suitable length of a probe for intragastric administration in mice, i.e. reaching from the mouth to the end of the breastbone

it slippery. For administration, the conscious animal must be restrained very firmly by gripping a fold of skin from the scruff of the neck down the back so that the head of the animal is kept immobile. Firm restrain with complete immobilisation of the head of the animal is essential for the following procedure. Its position is vertical with the head tipped slightly forward. No mouth gag is necessary. To force the jaw open pressure can be applied to the mandible. The ball tip is inserted behind the incisors into the back of the mouth. Using the needle as a lever, the head of the animal can be tipped back. As soon as a straight line is formed, the needle is passed gently through the mouth and pharynx into the oesophagus (Shimizu 2004). The animal usually swallow as the feeding needle approached the pharynx, these swallowing movements can help so that the probe slips through the oesophageal opening. If any obstruction is felt, no force must be exerted and another try must be made to find the oesophagus opening. As it is important to prevent the tube from entering the trachea the animal must be well observed. For example the needle can usually be seen passing down the oesophagus on the left side of the neck (Wolfersohn and Lloyd 1994). Being in the trachea the animal will cough, and it is possible to feel the tube touching the cartilage rings of the trachea (Baumans et al. 1993). Once in the oesophagus the tube is gently pushed down into the stomach. The passage may be obstructed at the sphincter to the entrance of the stomach. Manipulation of the syringe to produce a gentle thrusting movement combined with a gentle backward and forward movement will often overcome this difficulty.

99

8 Drug Administration

Once the needle is in place the solution is administered slowly. If any obstruction is felt, if the animal coughs, chokes or begins to struggle vigorously, or if the injected solution is seen coming out through the mouth or nose, these may indicate that the needle entered the respiratory tract and the injection is being made erroneously into the lunge. Any of these signs would necessitate immediate stop of the administration and withdrawal of the needle. The animal must be observed very carefully. If there is any sign of lung damage, the animal should be euthanized humanely. Once it has been ascertained that the needle is in the right position the solution can be given fairly rapidly. As soon as the administration is finished, the tube must be withdrawn. Administering a corrosive solution the needle must be flushed with saline before pulling back to make sure that the mucosa of the upper gastric tract will not be irritated or even damaged (Fig. 8.2). Using this technique there could be a certain risk of introducing trauma or a fatal accident as mentioned before. For this reason, only well trained, skilled personnel should carry out intragastric administration. First training should be done with dead animals. When training with anaesthetised animals it is important to choose an anaesthetic that allows the animals to retain its swallowing reflex (e.g. ketamine/xylazine). However, intragastric administration by gavage can be performed very rapidly. A very well trained person can feed as many as 120 well trained animals per hour as example (Ferrill and Hill 1943). When properly performed, accidental tracheal infusion is rare (Kraus 1980).

Knowledge of small rodent’s drinking or feeding habits will help to determine whether or not a substance is given on a full stomach or an empty one. When fed ad libitum, animals will never have an empty stomach at any time of the day. The stomach of a mouse or rat is maximal packed at the end of the dark period and a minimum of content is found at the end of the light period. Therefore, if no overnight food restriction is done, the animals should receive only a very small volume. Otherwise, the animals have to be starved prior to the administration of larger volumes. By this method substances can administered directly into the stomach with accurate dosages and reliable timing. When a liquid volume is administered by gavage a substantial amount of the dose passes rapidly through the stomach to the small intestine. This occurs both in fasted and in fed animals (Claassen 1994a) and results in the substances being absorbed very rapidly. For example within one minute after quinine solution administration to the fasting animal 30% (mice) and 24% (rat) of the dose is transferred to the small intestine (Watanabe et al. 1977). Absorption is much faster than that with dietary administration, leading to a peak and not a slow and prolonged gradient in the blood. Using a special feeding needle for capsules and filling the substances into commercially available gelatine miniature capsule this technique can also be used for intragastric administration of solid materials (Lax et al 1983, Waynforth and Flecknell 1992).

8.4 Parenteral Administration

Fig. 8.2 Radiograph of gavage in a rat with a straight, bluntended stainless steel needle. The stomach is filled with X-ray contrast medium

Application of substances other than via the gastrointestinal tract to the body is called parenteral administration. Methods of parenteral administration without penetrating the skin are topical applications and inhalation. Other methods are implantation of catheters and subcutaneous or intraperitoneal implantation of pellets, osmotic pumps or vascular access ports (VAPs). But the most common parenteral routes are injection of substances, solutions or suspensions into the mouse or rat subcutaneous, intraperitoneal or intravenous. Small amounts of solutions or suspensions are injected, and large volumes are infused. The rate of absorption is dependent on the route of administration. The most rapid absorption is achieved by intravenous injection. The large surface area of the

100

abdominal cavity and its abundant blood supply also facilitate rapid absorption after intraperitoneal injection. But the absorption from this route is usually onehalf to one-fourth as rapid as it is from the intravenous route (Woodard 1965). Some injected fluids cause tissues irritation. Therefore, they should only be used after being diluted with solutions (for example: sodium chloride solution) and should, if possible, be given directly into the blood stream where the solution becomes quickly diluted. Animals must be properly handled and restrained in order that injection procedures can be carried out without endangering either the animal or the operator. Various proper restraining techniques and many commercially available mechanical restraining devices have been described in literature. But it is outside the scope of this chapter to discuss in detail this issue now or later. The reader is pointed out to the special literature about restrain. The duration and extent of handling or restraining will depend, among other factors, on the application method chosen and the requirements of the experiment. Over the years many different routes and methods of injection techniques have been used and documented. Some of these recommended techniques are the same as for venous blood collection or bleeding procedures. Although injections should normally be given without anaesthesia, this is not a general rule. For instance, if the person carrying out the injection is not well trained, or if the substances are irritating, it is better to anaesthetise the rat or mouse with a shortacting narcotic. When using injection or infusion techniques, several points deserve special attention. This method requires strict asepsis. Sharp and sterile needles of a size appropriate to the size of the animal and, in particular for intravenous injection, by the size of the selected vessel, must be used. The required needle gauge depends on the viscosity and volume as well as the speed of injection of a fluid. To assure short injection and restrain time larger diameter needles are necessary for injection of large volumes or viscous fluids. On the other hand, using too thin needles there is the risk that the needle may snap from the syringe. But as a thin needle causes less pain and prevents the fluid from flowing back, the smallest needle should be always selected. The Injection of cold fluids is painful (Baumans et al. 1993), so the fluid must be brought at least to room temperature or better still up to body temperature before used. Substances administered into

K. Nebendahl and P. Hauff

the blood circulation must be soluble in suitable solvents and, because of the risk of embolism, air bubbles in the fluid have to be avoided. As the physiological principles of fluid administration can not be discussed here in depth, only some recommendations are given with respect to administration volume and needle sizes.

8.4.1 Topical Application The skin is a convenient site for the administration of drugs. The absorption of substances through the skin is an area of research that has been extensively studied for a number of years. Dermal absorption represents a pathway for substances to enter the body, particularly in cases of occupational and environmental exposure. Numerous factors can affect the extent of percutaneous absorption of a substance. These include the physicochemical properties of the substance, the attributes of the vehicle and the permeability of the skin (Franklin et al. 1989; Wester and Maibach 1986). Important factors affecting dermal absorption are the water and lipid solubility properties of the substances used for topical application. The outermost layer of the skin is essentially a lipid barrier, whereas the viable epidermis, which lies below the stratum corneum, is basically of an aqueous environment (Guy and Hadgraft 1991). Thus, the ability of a substance to be absorbed through skin and enter the systemic circulation is determined by its ability to partition into both lipid and water phases. The most usual sites are the skin covering the back or the abdomen. When not using hairless or nude animal, the hair must be removed. This can be done either with a depilatory or with an electric clipper or wet shaver. As depilatories are chemical substances, they could interfere with the study. Their use should therefore be considered carefully. After clipping hair, the hairless area should be cleaned of any fat and grease and other debris. If possible the animal should be prepared several days in advance so as not to affect the experiment. Shaving should be done at least 1 day ahead because of microinjuries. The substances should be dissolved in a volatile solvent or mixed in a suitable cream before application and then applied with a dropper or smeared onto the skin with a swab.

101

8 Drug Administration

Some precautions are usually necessary to prevent the animal from licking or scratching the application sites (Woodard 1965). As active hair growth or the age of the skin will influence absorption of topically applied substances, the state of the hair cycle should be taken into consideration (Waynforth and Flecknell 1992a).

8.4.2 Inhalation The inhalation route, necessary for some drugs or vaccines, is the nearest akin to an intravenous injection because of the relatively large lung alveolar surface presented for absorption by a membrane that is separated from the blood by one or two cell layers. Consequently, absorption of gases or aerosols that reach the alveoli is virtually complete. The greatest problem in using this method is the generation of a suitable aerosol of the test substances (Shimizu 2004) and to accurately determine its dosage (Waynforth and Flecknell 1992a).

8.5 Administration by Injection 8.5.1 Intradermal Injection (i.d.) The intradermal (intracutaneous) injection is not recommended in general and should be restricted to cases of absolute necessity (CCAC 2002). It is a tricky procedure in any species, but especially difficult in the mouse because the skin is very thin (Suckow 2001). The usual sites and the preparation of the skin are the same as for topical applications. Before starting the injection the site must be clean and swabbed with an antiseptic. Local or general anaesthesia is generally required for accurate intradermal injection techniques and for humane considerations (Pekow and Baumans 2003). Only very small quantities of solutions – per injection site 0.02– 0.05 mL – can be deposited intradermal. For injection, a special short-bevelled hypodermic needle or an ordinary 27 G needle is held bevel down almost parallel to the skin and is pierced only within 1 mm or 2 mm into the skin of mice or rats respectively. Resistance should be felt both as the needle advance and as the fluid is injected.

Fig. 8.3 Intradermal injection in a mouse

A successful intradermal injection of even a small quantity of fluid results in a hard bleb at the injection site, while no bleb is formed when the injection is done by mistake subcutaneously (Fig. 8.3).

8.5.2 Intramuscular Injection (i.m.) This route usually results in more rapid absorption than from the subcutaneous route (Woodard 1965). Absorption usually takes 45–60 min for most fluids. Repository forms are available which remain for days and weeks (Moreland 1965). Intramuscular injections are usually avoided in the mouse because of its small muscle mass and as muscle masses of the rat are not very large, the number of practical injection sites and the volume that can safely be given by the intramuscular routes are restricted Intramuscular injections are frequently painful, due to the distension of muscle fibres which occurs, and therefore good technique and restraint are required (Wolfensohn and Lloyd 1994). Recommended injection volumes and suggested hypodermic needle sizes are given for mice in Table 8.2 and for rats in Table 8.3. The intramuscular route should only be used if a less painful alternative is impossible, or if it is clinically required (Working Party Report 2001). If necessary, the injection may be given into the quadriceps muscle of the mice. In rats, the quadriceps feels like a small peanut on the front of the thigh, and can be immobilised with the thumb and forefinger of one hand while injecting with the other (Wolfensohn and Lloyd 1994). The muscles of the posterior thigh area

102

K. Nebendahl and P. Hauff

Table 8.2 Recommended administration volumes and suggested needle sizes for mice s.c. mL/kg

Gauge

Wolfensohn and Lloyd (1994)

0.5

25 G

Diel et al (2001)

5

Pekow and Baumans (2003)c

0.25

25 G

0.05

27 G

0.5

25 G

0.3d

26 G

Bihun and Bauck (2004)

2–3

21–26 G

0.03b

23–26 G

10

21–26 G

0.5–3d

23–26 G

2–3

<25 G

0.05

25–27 G

2–3

<23 G

0.2

<25 G

Shimizu (2004) Weiss et al (2009)

e

a

i.m. mL/kg

Gauge

i.p. mL/kg

Gauge

27 G

2–3

27 G

10

0.1b

0.5

b

0.03

i.v. mL/kg

Gauge 25–28 G

5

1.5

0.15

2–4 Sites b Per site c Weight of the animal: 25 g d Slow (1–3 min) e Weight of the animal: 30 g a

Table 8.3 Recommended administration volumes and suggested needle sizes for rats s.c. mL/kg

Gauge

i.m. mL/kg

Waynforth and Flecknell (1992a, b)

Up to 5

21–25 G

0.2

Wolfensohn and Lloyd (1994)

1–2b

25 G

0.1

Diel et al (2001)

5

Pekow and Baumans (2003)d

1

25 G

0.1

25 G

2

25 G

4a

25 G

Bihun and Bauck (2004)

5–10

21–26 G

0.2–0.3c

22–26 G

10

22–26 G

0.5–3a

22–26 G

Weiss et al (2009)e

2c

Gauge

0.25c

Gauge

Up to 10 25 G

5–10

5

i.v. mL/kg

Gauge

2

23–25 G

1

25–27 G

a

23–25 G

10

0.1c c

i.p. mL/kg

5

1.25a

Slow (1–3 min) 2–4 Sites c Per site d Weight of the animal: 200 g e Weight of the animal: 250 g a

b

of mice and rats should be avoided because the sciatic nerve runs along the back of the femur. Irritant substances that are inadvertently injected in close proximity to this nerve may result in lameness or in the animal’s self-mutilation of the affected limb. Intramuscular injections for rats can also be given into the gluteus muscles of the hind leg. During the injection procedure the animal must be anaesthetised or manually restrained by another person. Penetration by the needle of only few mm is sufficient for a deep intramuscular injection and

avoids the risk of damaging the periost of the femur. The tip of the needle should be directed away from the femur always. Care is taken to visualise the depth at which the needle is placed, ideally centering the tip near the middle of the muscle group. Use of a short needle gives better control (Pekow and Baumans 2003). Sometimes an intramuscular injection will fail, even though the needle is felt to be in the muscle mass, because it actually lies in one of the fascial planes (Waynforth and Flecknell 1992a). Prior to injection of the fluid, aspiration by pulling

8 Drug Administration

Fig. 8.4 Intramuscular injection into the area of the gluteal muscles of the hind leg in a mouse using a 26 G needle

back lightly on the plunger of the syringe must be done to rule out accidental injection into a blood vessel. After injection, the site should be massaged to disperse the dose because of the difficulties described, the intramuscular administration should only be used if there is no alternative and should only be performed by well-trained persons (Wolfensohn and Lloyd 1994) (Fig. 8.4).

8.5.3 Subcutaneous Injection (s.c.) In comparison with other routes of administration, the subcutaneous injection has several advantages. The subcutaneous area is well supplied with capillaries but their number may differ at various sites of injection and this may lead to differences in the rate of substance absorption. Nevertheless, as this method of administration will produce a substance depot, subcutaneous injection is the best option when a relative long period of substance absorption from a repository injection site is desired. This technique is painful if the pH or osmolarity is wrong, or if the material is irritant or cytotoxic. Tissue necrosis can occur (Working Party Report 2001). Substances or solutions which have such adverse characteristics should be diluted with suitable fluids before being subcutaneously administered. On the other hand, oil suspensions can be given subcutaneously. An increased rate of absorption can be achieved by the additional use of hyaluronidase injected at the same site (Woodard 1965).

103

Subcutaneous injection is the preferred method for the administration of substances into small laboratory animals. This is due to the simplicity of the injection technique, greater choice of injection sites and the possibility to deposit large volumes. Normally, the scapular region is the preferred site for a single subcutaneous injection. Other recommended sites are the back or the ventral abdomen. As subcutaneous injections of compatible solutions are rarely painful (Wolfensohn and Lloyd 1994), a conscious animal can usually be used. Recommended injection volumes and suggested hypodermic needle sizes for aqueous solutions are given for mice in Table 8.2 and for rats in Table 8.3. Viscous solutions or suspensions require a needle of 2–4 wire gauge numbers smaller. For subcutaneous injection, a fold of loose skin is lifted between the thumb and the forefinger and at the base of the fold the needle, attached to the syringe, is passed in an anterior direction through the skin parallel to the body of the animal, to avoid penetrating deeper tissues. Ideally, the whole of the needle shaft should lie subcutaneously as this prevents leakage of the injected fluid. When in position, the tip of the needle should be moved up and down to reveal its whereabouts and also to ascertain that the needle is truly subcutaneous. If the tip cannot be discerned then the needle could be in an intraperitoneal or intramuscular position and must be slightly withdrawn to lie subcutaneously. Before injecting the substance, aspiration has to be done to ensure that the needle has not entered a blood vessel or has moved out of the skin again. (Fig. 8.5).

Fig. 8.5 Subcutaneous injection in a mouse at the base of a fold of loose skin (area at the back) using a 26 G needle

104

K. Nebendahl and P. Hauff

8.5.4 Intraperitoneal Injection (i.p.) Also the Joint Working Group of Refinement (Working Party Report 2001) stated that this route should not be used routinely, (except for administration of certain anaesthetics) this application procedure is one of the most frequently used parenteral routes of administration in small laboratory animals. But this technique is not recommended for pregnant animals since the needle can puncture the gravid uterus. The large surface area of the abdominal cavity and its abundant blood supply facilitate rapid absorption. Absorption from this rote is usually one-half to one-quart as rapid as that from the intravenous route (Woodard 1965). However, for long-term studies, repeated injections may lead to tissue reaction and adhesions. As relative large volumes can be given intraperitoneal, potentially irritant substances can be generously diluted. When using this method it has to be kept in mind, that substances given intraperitoneal are first absorbed into the portal circulation. Biotransformation of the injected substances may take place in the liver before they reach the general circulation, so that their bioavailability is quite different from that of an intravenous injection. Intraperitoneal injections are generally undertaken without anaesthesia. Recommended injection volumes and suggested hypodermic needle sizes are given for mice in Table 8.2 and for rats in Table 8.3. The abdomen can be divided into four quadrants by the midline and a line perpendicular to it passing through the umbilicus. The upper two quadrants are a hazardous area to inject because the liver, stomach and spleen are situated here. For this reason, the intraperitoneal injection should be placed into one of the lower quadrant of the abdomen. To avoid injection onto the cecum at the right, the injection into mice should be done into the caudal left abdominal quadrant (Suckow 2001) and while using rats, the application should be done into the right quadrant to avoid punctuation of vital organs. Mice and small rats can be properly restrained and injected by one person. When injecting larger rats, it is advisable to have an assistant. The hindquarter and tail are restrained by the assistant, and the operator extends one of the animal’s hind legs and carries out the injection (Waynforth and Flecknell 1992a) (Fig. 8.6). At the start of the administration, some investigators tilt the animal so the head is lower than the abdomen in an attempt to slide the viscera cranially and away from the needle. However, the viscera are quite

Fig. 8.6 Intraperitonael injection into the lower left quadrant of a mouse using a 26 G needle

immobile because of the slight vacuum in the abdomen and this manipulation is of questionable value (Fallon 1996). Because of the risk that the injection is made between the skin and the abdomen muscles or the risk of accidental penetration into the viscera, the short needle should enter the skin neither parallel nor vertical but rather at an angle of 10–20°. With a quick, firm motion, the tip of the needle is inserted through the skin subcutaneously. To avoid leakage from the puncture point, the needle is run through subcutaneous tissue in a cranial direction for 2–3 mm and then a final short thrust is made through the abdominal wall, holding the needle nearly vertical (Shimizu 2004). It is necessary to insert only the tip of the needle into the peritoneal cavity otherwise, the intestine can be punctured (Waynforth and Flecknell 1992a). Prior to administration of the compound, aspiration must be done. If any kind of fluid – urine, intestinal content or blood – is pulled back into the syringe, the needle is probably in one of the abdominal organs and should be repositioned. This may have no serious consequences for the animal but may be associated with slow and erratic absorption of the compound. This is a particular problem in the case of anaesthetics as the animal will not respond as expected to the drug. Redosing under these circumstances is tricky as the initial dose will eventually be absorbed and, combined with the second dose, may prove fatal (Suckow 2001). As there are only few references to incorrect intraperitoneal injections, it is obviously often assumed that an intraperitoneal injection always delivers the substances to the peritoneal cavity. But, a lack of

8 Drug Administration

references to the possible occurrence of injection errors makes it likely that it is not always realised that such errors may be a source of experimental variation. Nevertheless, a significant part of the injections are actually made intragastrically, intraintestinally, subcutaneously, retoperitoneally or intracystically (Claassen 1994e). The frequency of erroneous injections by skilled investigators has been reported by Lewis et al. (1966), who stated that in nearly 20% of the animals, parts of the material were not injected into the peritoneal cavity. But even the well controlled use of a standardised injection technique can only reduce the number of erroneous injections to 5.5% (Schneider and Schneider 1970). Sometimes it is possible to recognise the error, for instance, when the injection is made into the intestine, fluid will be seen often issuing from the rectum immediately after the injection (Waynforth and Flecknell 1992a, b), or the animal will defecate. If an injected fluid needs to be quickly diluted in the blood, then the intravenous instead of the intraperitoneal route should be given preference, thus also avoiding the risk of peritonitis.

8.5.5 Intravenous Injections (i.v.) Intravenous administration offers various advantages over other routes of injection. For example, it gives control over the rate of introduction into the general circulation, rapid response etc. and it provides the most complete availability of substances with minimal delay. By controlling the administration rate, constant plasma concentrations can be obtained at the required level. Unexpected side-effects during administration can be halted by stopping the injection. Compounds that are poorly absorbed by the digestive tract or are unacceptably painful when given intramuscularly or subcutaneously may be administered intravenously when given carefully into the vein without leaks into the surrounding tissues. Several general points on intravenous injection or infusion deserve special attention. Except in terminal experiments, reasonable aseptic techniques with sterile equipment must be employed, particularly when animals are being used in long-term studies and frequent injections are required. The syringe plus needle must first be filled with the solution for injection so that no air bubbles are administered. When using large veins it should be easy to aspirate blood if the needle lies correctly but it is not always possible to do so with small

105

veins. After injecting a small amount of the solution into a small vein, the injected fluid should be washed away by the blood in the vessel. If this does not happen the position of the needle is doubtful. As soon as a bleb should arise, the position of the needle is certainly not in the vein but in the surrounding tissue. A fresh attempt must be made or the needle should be moved in the surrounding tissue in such a way that it then enters the vein. When finishing the intravenous injection, a swab always has to be pressed on to the injection site while pulling out the needle to prevent backflow of administered fluid and/or blood. Once the vessel has to be used several times, the first injection should be made as distal as possible in relation to the heart and subsequent administrations should be placed progressively more proximal. This proceeding is necessary because multiple puncture of the same vein and frequent injections of substances can damage and/ or block the vein and the distal part of the vessel can no longer be used for subsequent administrations. While selecting the site of injection, consequences of a possible intravascular thrombosis or a possible extravascular administration of substances, etc., have to be considered. Mice and rats have no readily accessible veins of sufficient size for venipuncture. Therefore, many methods for vascular access have been described in the literature for mice (Salem et al. 1963; Suckow 2001; Shimizu 2004), for rats (Kraus 1980; Petty 1982; Cocchetto and Bjornsson 1983; Waynforth and Flecknell 1992a; Nebendahl 2000) and for mice and rats (Moreland 1965; Pekow and Baumans 2003). Percutaneous injections are made with conscious animals into the lateral tail vein, lateral marginal vein (v. saphena) or dorsal metatarsal vein, and with anaesthetised animals into the sublingual vein or penile vein. Some techniques are difficult to be performed, therefore, many methods have been developed to make this task easier, including the use of a wide range of hypodermic needle sizes, improving visibility of the injection site by magnification, transillumination, shaving, surgical incision and the application of heat, tourniquets, and chemicals for vasodilatation. After making a skin incision and surgical exposition, administrations can be made into the external jugular vein or femoral vein. Both routes require the use of anaesthesia. Different restraint devices and other equipment have been recommended to immobilise the mouse or rat and thus facilitate injection. Anaesthetising the

106

animal is considered helpful, but may be contraindicated for some experiments. Use of these methods or equipment cannot guarantee a successful injection. For instance, the vein can be deflected by the needle, an inserted needle can dislodge or perforate the vein whenever the syringe is manipulated or when the restrained animal flinches (Nachtman et al. 1988). As some routes require special technical expertise and skill, they are not advisable for persons who rarely use them. Such persons should confine themselves to methods they can perform or to methods that are easy to learn. To avoid pain and shock, injection must always be given slowly, especially when administering large volumes.

8.5.5.1 Lateral Tail Vein The most common sites for intravenous injection in mice or rats are the lateral tail veins. These vessels are readily visualised, but are quite small in diameter and injection into them requires considerable practise and skill. From the tip to the root of the tail the lateral vein lies immediately beneath the skin but the vein narrows from the root to the tip. Because of the very small diameter at the tip of the tail, the vessels are not visualised at the tail top even in albino animals. Therefore, the whole length of the vein cannot be used for this administration technique. Videm (1980) presented a simple and relatively rapid method using the lateral vein of a rat at the root of the tail where the skin, after being shaved, is thin and smooth and where the vessels are superficial and accessible for intravenous injection. As anaesthesia causes decrease in blood pressure, this treatment is normally not used, but, because the tail is sensitive, the use of a restraint device is necessary. Several types of clear plastic restraint devices which allow tail access are commercially available for small laboratory animals (Waynforth and Flecknell 1992a; Fallon 1996; Pekow and Baumans 2003). Since the tail is a major thermoregulatory organ with a large surface available for heat loss, an enhanced blood flow in the dilated veins and hence a successful venepuncture can be ensured by warming prior to injection. Warming the whole animal by placing the animal into a thermostatically warmed “hot-box” (Conybeare et al. 1988), warming the tail under a heating lamp (Shimizu 2004) or holding the tail in warm water for 1–2 min (Barrow 1968; Fallon 1996) can

K. Nebendahl and P. Hauff

induce tail veins dilation. When placing animals in a warmed box, the chamber temperature should not exceed 37°C and the animal must be observed constantly for signs of distress (Working Party Report 2001). Another method to obtain a good venous filling is constriction. Several methods are described using finger pressure (Barrow 1968) and/or a tourniquet (Videm 1980; Petty 1982; Waynforth and Flecknell 1992a). The pressure or the tourniquet must be released just before the injection is made. For injection, the tail should be bent down with one hand while the vein is punctured at the angle of the bend with the needle and syringe held in the other hand. The vessel must be entered at a very small angle almost parallel to the vein. The tip of the needle is then advanced only few millimetres. Aspiration of blood to confirm correct needle placement can be done with rats, but is not productive in mice (Pekow and Baumans 2003). Starting the injection, no resistance will be felt as the fluid is administered and the vein will appear to “blench out” (Suckow 2001). Failure of insertion is identified by swelling of the tail or blanching of the skin. For slow intravenous injections, a butterfly needle (needles with attached flexible tubing) should be used, because as soon the needle has been seated in the vessel and taped in place, the flexible tubing permits the animal some freedom from complete restraint and the tip of the needle cannot be pulled out or penetrate through the vessel during the injection period. The injection is carried out using a 23–25 G needle or better a 27 G butterfly set (Fig. 8.7).

Fig. 8.7 Lateral tail vein injection in a rat using a 24 G “overthe-needle” catheter

8 Drug Administration

107

8.5.5.2 Lateral Marginal Vein (Large Saphenous Vein) of Rats In the opinion of Grice (1963) injection into this vein or saphenous vein of a rat provides the method of choice. The animal may be anaesthetised or placed in a restrainer, leaving one hind-limb free. The posterior and lateral surface of the thigh and leg of the hind limb are shaved. The rat is held firmly by an assistant placing the right hand over the hips of the animal, with the free limb positioned between the first and second fingers and applying sufficient pressure to cause this vein to become quite prominent without the use of any form of tourniquet. The left hand of the assistant holds the foot firmly (Pearce 1957). Pearce (1957) recommended a 26 or 27 G hypodermic needle for injection (Fig. 8.8).

Fig. 8.8 Insertion of a 29 G needle into the lateral marginal vein of a mouse

8.5.5.3 Dorsal Metatarsal Vein (Small Saphenous Vein) Everett and Sawyer (1956) first described this injection technique for rats. The vein is found on the dorsal surface of the foot. The skin over the lateral plantar surface is shaved and swabbed with antiseptic. The operator holds the foot by the toes during the procedure, and an assistant aids by holding the knee in such a manner that sudden flexion and withdrawal of the foot are prevented. A 26-gauge hypodermic needle kept almost horizontal to the surface and directed towards the ankle is inserted through the skin and into the vein at a point where it just starts to travel up the foot after first crossing the foot and supplying the toes with blood. Everett and Sawyer (1956) recommended a 26 G hypodermic needle for injection (Fig. 8.9).

8.5.5.4 Penis Vein (Penile Vein) The advantage of this method is in its applicability to animals – for instance mice – having no easily accessible site to intravenous injection (Salem et al. 1963). Petty (1982) maintained that the dorsal penis vein injection is much simpler, more rapid, more

Fig. 8.9 Dorsal metatarsal vein injection in a rat using a 26 G needle (source Nebendahl (2000)

reproducible and easier to accomplish than tail vein injection and the technique can be learned easily. But Waynforth and Flecknell (1992a) stated that this route should only be used under special circumstances, because of the consequences of damage to the vein. Nightingale and Mouravieff (1973) investigated whether the penile vein is part of the portal or general circulation. Based on these experiments, they concluded that injection into this vein leads to the general circulation and that a first-pass effect on metabolism is not to be expected. The anaesthetised animal is placed on its back on the table. The glans penis is exposed by sliding the prepuce downwards while pressing at the base of the penis. It is

108

K. Nebendahl and P. Hauff

Fig. 8.10 Dorsal penis vein injection in a rat using a 26 G needle (source Nebendahl (2000)

Fig. 8.11 Sublingual vein injection in a rat using a 24 g “overthe-needle” catheter

held at the very tip with the thumb and the forefinger. If it is not placed under too much tension, the penile vein is easily seen as a distinct central vein. Once the vein is pierced, aspiration of blood is virtually impossible, therefore only a very small volume has to be injected first to see if it flows freely. After the injection, the injection site is pressed with a swab for a few seconds and the gland is encouraged to retract to prevent further bleeding (Waynforth and Flecknell 1992a). For rats from weaning age and older, Waynforth and Flecknell (1992a) recommended a 24-G hypodermic needle and a 30-G for smaller animals and Salem et al. (1963) a 26–30 G needle for mice (Fig. 8.10).

for rats a 23–25 G hypodermic needle for injection (Fig. 8.11). Without special experience, access to the jugular vein or the femoral vein is only possible after surgical exposition of these vessels. As injections via a needle or infusions via a catheter require a similar preparation more details in respect to these injection techniques are given later.

8.5.5.5 Sublingual Vein (Tongue Vein) As the entire procedure, namely, anaesthesia, suture, and injection, can be done in less than 5 min, Petty (1982) recommended the sublingual vein for rats, which is frequently overlooked as an injection site. As the injection technique is described in detail for rats by Petty (1982), Waynforth and Flecknell (1992a) or Diehl et al. (2001) and for adult mice by Waynforth and Parkin (1969), the reader is referred to these articles. Also, this technique is easy to perform and any differences in the eating or drinking habits of the animals were noted after injection, the necessary of repeated anaesthesia limited frequent injections by this method. Waynforth and Flecknell (1992a) recommended for rats and mice a 25 or 30 G and Diel et al. (2001)

8.5.5.6 Intravenous Injection or Infusion by Catheter Chronic venous cannulation , implantation of indwelling vascular catheters or VAPs in small laboratory animals are an accepted and extremely useful experimental technique for repeated injections or permanent infusions, since they reduce the stress of multiple injection associated with, for example, restraint and discomfort due to repeated needle pricks into vessels. Short-term catheterization of a few days is relative easy to accomplish, for instance using an over-the-needle catheter, however, complications increased with the amount of time required for the catheter to be maintained and are related to the side of placement. A number of authors have described techniques for chronic catheterization of different blood vessels of mice or rats (Desjardins 1986; Joint Working Group on Refinement 1993). Each technique has its own individuality and the investigator must adapt to meet the

109

8 Drug Administration

needs of the experimental design (Petty 1982). The procedures described differ also in the various protection devices employed to prevent the animal from manipulating the catheter through pushing, pulling away or biting, in the methods of maintaining proper catheter placement and of exteriorizing the catheter, in the use of different surgical techniques and in the methods used to maintain the patency of the catheter. The procedures used to fit animals with an indwelling vascular catheter fall into two general categories: 1 Direct access is accomplished by attaching a syringe or a piece of tubing to the exteriorized distal end of the catheter just before injection or infusion. This requires some form of handling or restraint of the animal. The duration and extent of these manipulation will of course depend on the method chosen for implantation, on the type of catheter, on the requirement of the study and in particular on the experience of the operator. 2 Remote access involves tethering the animal, usually by a protecting device, as the catheter is extended beyond the animal’s home cage. This method permits access to the vascular system of otherwise undisturbed, freely moving animals housed singly. As advances in the polymer industry have led to the production of many synthetic materials with acceptable biocompatibility, catheters for nearly every vessel of small laboratory animals are available nowadays. Also various methods have been used to minimise the incidence of thrombotic occlusion of the intravascular catheter or to remove existing thrombotic obstructions to prolong the patent lifetime of catheters, standardised instructions for routine catheter care have not been created till now. To prevent clotting in the catheter, it should be flushed with heparinized saline or another anticoagulant after placement and between infusions or injections, if not daily, at least twice a week and the dead space in the cannula is then replaced by a carefully calculated amount of fresh anticoagulant. A concentration of 10–1,000 IU heparin per millilitre saline is recommended (Joint Working Group on Refinement 1993). If the catheter still becomes blocked, it may be possible to dissolve the thrombotic occlusion by filling the catheter with a solution of urokinase or streptokinase (Hurtubise et al. 1980). Cannulas without heparin form small clots at the tip that, if dislodged, can cause pulmonary, renal or heart infarcts.

Percutaneous Insertion of the Catheter Percutaneous methods involve the implantation of the catheter into a blood vessel by piercing the vessel with a needle and pushing the catheter through the needle into the vessel. In a non surgical approach, catheters are inserted into a vessel of an animal. For instance Little et al. (1962) and Rhodes and Patterson (1979) inserted the needle through the intact skin into the tail vein of a rat and Marini and Garlick (2006) in the same vessel of a mouse. When blood flowed freely, the catheter was guided through the trough of the needle into the vein. Nachtman et al. (1988) or Waynforth and Flecknell (1992a) used a commercially available over-the-needle catheter, comprising a short cannula fitted over-theneedle which is only a few mm longer than the cannula. This unit has two distinct advantages over hypodermic needles. It provides a visual check that the vein has been entered as blood fills the needle and once the cannula is established in a vessel, any movement by the animal or the operator does not lead to penetration or laceration of the vessel wall as easily as a hypodermic needle because its cannula is pliable and blunt. After positioning the tip of the needle and cannula into the vein, the needle is withdrawn while holding the cannula firmly in place. Once the cannula has been filled with blood the needle can be withdrawn completely.

Practical Hint Another easy and cost effective way of catheterization of the lateral tail vein of mice is used in non-invasive imaging experiments with MRI and CT frequently. A tail vein is catheterized using a 30 G needle connected to a 10 cm PE 10 polyethylene catheter (Portex) filled with 0.9% NaCl. Successful puncture of the mouse tail vein is controlled by blood reflux into the catheter and by the injection of 20 mL 0.9% NaCl. Ten centimetre of this catheter contains a volume of 10 mL. Then, the needle is fixed on the mouse tail with a superglue. The distal end of the catheter can be connected with a PE 50 polyethylene catheter (Portex) containing the contrast agent and a 1 mL tuberculin syringe. After each examination, the needle can be removed from the tail vein.

110

Surgical Implantation of the Catheter Numerous methods are available for implantation of catheters in different veins and/or arteries of small laboratories rodents (Petty 1982). Cocchetto and Bjornsson (1983) reviewed many articles on arterial and venous implantations, gave methodological notes as well as procedural comments. Detailed procedures of preparing catheter equipment and inserting the catheter have been described, for instance, by Weeks and Davis (1964) and Harms and Ojeda (1974), who gave details about an easy-to-prepare cannula and a simple procedure for cannulation. Frequently used vessels for chronic catheterization are the tail vein, extern and intern right jugular vein, femoral vein and artery or the left carotid artery and the aorta. Surgical techniques for the permanent catheterization of the jugular vein of rats (Remie et al. 1990) or of mice (Bardelmeijer et al. 2002), of the tail vein (Born and Moller 1974), of the jugular vein, femoral and carotid artery (Yoburn et al. 1984) and of the carotid artery, dorsal aorta, jugular and tail vein (Waynforth and Flecknell 1992b) are comprehensively described in the mentioned reports. The reader is pointed to these articles. As it is not easy to get a tube into a surgically exposed, small vessel of a small rodent, a lien vessel cannulator is sometimes needed. There are various techniques to fix the extravascular or intravascular portion of a catheter with the vessel and the surrounding tissue or within the lumen of the blood vessel. Details can be found in the article by Cocchetto and Bjornsson (1983). The intravascular catheters are commonly exteriorized by subcutaneous tunnelling from the vascular incision site to the back of the neck or between the shoulder blades (Cocchetto and Bjornsson 1983), and sometimes also to the root of the tail (Jones and Hynd 1981) The various procedures for protecting the catheter differ in the protective device used to prevent the animal from manipulating the free end of the catheter and/ or the connection to the infusion pump. Many authors have addressed the problem of the protection of the catheter from kinking and chewing by the animal or described techniques for proper catheter placement. Other writers have directed their attention towards minimal restraint or towards methods of compensation for rotational movement of the animal in the cage

K. Nebendahl and P. Hauff

(References see Nebendahl 2000). If catheters are implanted for continuous infusion over long periods and the use of a portable pump is not practical, the tube must be connected with an infusion pump. Therefore, it is necessary to take special precautions to prevent twisting and kinking of the delivery tubing through movements of the animal. The freedom of the animal must be inhibited as little as possible by the protective device and must allow the animal to move in an unrestricted manner within its cage. If the implanted tube is to be used for multiple injections, the free end of the tube must also be protected from biting or pulling out. For this purpose, the same protective device as for long-term infusion can be used. But, if the catheter is exteriorized in the neck area and only a short piece is outside the body, it is not necessary to protect the catheter, if the animal is housed singly. Animals are not only stressed by the surgical procedure but also by the fact that they have to wear a protecting device and are housed singly or are limited in their movement (Birkhahn et al. 1976; O’Neill and Kaufman 1990). In general, the animals require a period of several days to recover from the procedures for the implantation of chronic catheters. For instance, during the first four postoperative days the normal weight gain is disturbed and even weight loss can occur (Popovic and Popovic 1960). As a general rule, infusion of about 1% of the blood volume per hour will not affect fluid disposition. Infusion of larger volumes should be based on preliminary studies designed to determine whether the cellular and ionic components of blood are maintained in a normal range (Desjardins 1986). For example, Steiger et al. (1972) developing a long-term, unrestrained intravenously fed rat model infused 30–60 mL/day of a specially formulated solution into rats weighing 140–250 g for weeks. However, at low flow rates less than 0.5 mL/h, catheter occlusion by thrombosis may be observed (Cox and Beazley 1975). As it is well established that an exteriorised vascular catheter is a source for the introduction of infection into the body, it is particularly important to avoid this as an infected animal will be useless for experimentation apart from the welfare aspects in the suffering it may cause. Therefore, it is essential to look for further methods for parenteral administration which does not have the known liability of exteriorised vascular catheter to become infected.

111

8 Drug Administration

8.5.6 Other Methods for Parenteral Administration 8.5.6.1 Pellets The delivery of substances at a slow, steady rate over a period specified can be supplied by using controlled – drug delivery releasing pellets. The pellets – available via www.innovrsrch.com – release the active product in the animal effectively and continuously. They are intended for simple subcutaneous implantation (Shimizu 2004).

8.5.6.2 Osmotic Minipumps Implantable osmotic minipumps are commercially available and are designed to deliver as a continuous infusion of precise volumes for a long period of up to 6 weeks without the need for external connection or frequent animal handling. This small device has been used very successfully in small laboratory animals and numerous reports about the use of osmotic minipumps can be ordered from the bibliography of the manufacturer (www.alzet.com). The minipump system is composed of three concentric layers – the substance reservoir, the osmotic sleeve and the rate-controlling, semipermeable membrane. An additional component, called the “flow moderator” is a 21 G stainless steel tube with a plastic end-cup. For more details and the principles of operation, see the technical information manual of the manufacturer (Fig. 8.12). Meanwhile, twelve different pumps are available for rats and mice varying in size from 1.5 × 0.6 cm to 5.1 × 1.4 cm, with a fixed volume delivery rate between 0.11 and 10.0 mL/h, a nominal duration from 1 day to 6 weeks and a reservoir volume from 100 mL to 2 mL (Fig. 8.13). Because of the mechanism by which the pumps operate, their delivery profile is independent of the chemical and physical properties of the agent dispensed. Successful delivery of an enormous range of compounds has been reported in the literature (www.alzet.com/pump selection/compound characteristics). The average pumping rate should be calibrated and correct performance should

be checked before implantation. The minimal animal sizes for subcutaneous or intraperitoneal implantation of pumps are 10 g for mice and 20 g for rats. Implantation is a surgical procedure and should be done using aseptic technique. In addition, all pump models are easily attached to a catheter, such that a pump, implantated either subcutaneously or intaperitoneally, is used to infuse into a vessel, organ or tissue. Full instruction for the correct use of the minipumps can be found in the Alzet® technical information manual. The compelling advantage of these minipumps are that they can be placed in situ without further need for infusion equipment and that a large number of animals can be treated effectively with a uniform infusion rate. However, infusion is limited to specific volumes for restricted time periods (Desjardins 1986). The disadvantage can perhaps be solved by implanting a new pump.

8.5.6.3 Vascular Access Port (VAP) A VAP is a subcutaneously implanted miniaturised device which allows long-term vascular access without the mentioned disadvantages of exteriorised vascular cannula. They are commercially available for mice and rats. A VAP permits continuous access to the venous or arterial system for compound administration. Using strict aseptic technique, the cannula of the VAP is inserted into a vein or artery (femoral artery or vein, jugular vein, and carotid artery are common sites but the portal vein (Gervaz et al. 2000) and other vessels are used too) and secured in place. The other end of the cannula is attached to the port that is secured in a subcutaneous location, most often over the shoulders. The surgical technique for implantation is described in detail by Waynforth and Flecknell (1992b). When the surgical incisions are closed, the entire system, VAP and cannula, is internal. When the animal is to be used, it is either restrained manually or in a suitable restraint apparatus. To access the port for injection, the skin over it must be shaved and disinfected. A needle is inserted through the skin and the diaphragm into the reservoir of the VAP, stopping when it hits the hard reservoir base. Administration of fluids can then be made. Special Huber point needles may be used to prolong the life of the port (Desjardins 1986; Suckow 2001) (Fig. 8.14).

112

K. Nebendahl and P. Hauff

Principle P ncip of Operation Op atio

Fig. 8.12 Principles of operation of an ALZET® osmotic pump

ALZET pumps have 3 concentric layers: Semipermeable Membrane

Rate-controlling, semi-permeable membrane Osmotic layer Impermeable drug reservoir

Flow Moderator

Osmode Layer

Impermeable reservoir

Test Agent

ALZET pumps work by osmotic displacement. Water enters the pump across the outer, semi-permeable membrane due to the presence of a high concentration of sodium chloride in the osmotic chamber. The entry of water causes the osmotic chamber to expand, thereby compressing the flexible reservoir and delivering the drug solution through the delivery portal.

Release Rates and Durations 1 3 day days

1 weak

2 weak

4 weak

6 weak

1003D 1 0 µl/hr

1007D 0 5 µl/hr

1002

0 25 µl hr

1004

0 11 µl hr

2001D 8 0 µl/hr

2001

1 0 µl/hr

2002

0 5 µl/hr

2004

0 25 µl hr

2006

0 15 µl hr

2ML1

10 0 µl hr

2ML2

5 0 µl/hr

2ML4

Fig. 8.13 Available ALZET® osmotic pump models

8.5.6.4 Oro-Endotracheal Intubation With an expanding use of small laboratory animals in research involving surgery of increasing complexity, there is an increasing need for the improvement of anaesthetic techniques for these animals. A free airway is useful in reducing mortality during and after operations. Inhalation anaesthetics provide excellent

2 5 µl/hr

control over induction and maintenance of anaesthesia but are difficult to use because of the mode of delivery and miniaturising the standard anaesthetic protocol. Numerous reports using a mask (Dudley et al. 1975; Levy et al. 1980; Norris and Miles 1982; Szczesny et al. 2003) or a self-inflating bag-mask (Schumacher et al. 2003) for administration of volatile anaesthetics in spontaneously breathing small laboratory animal

113

8 Drug Administration Fig. 8.14 Vascular access port. Fluid pathway through the port and advantages of a port

VASCULAR ACCESS PORT FLUID PATHWAY THROUGH THE PORT

ADVANTAGES OF THE PORT • low maintenance

SYRINGE WITH HUBER POINT NEEDLE

• infection rates reduced • animal stress minimized • permits common housing • exteriorized components unnecessary SELF-CLOSING MEMBRANE

SKIN PORT CHAMBER DRUG FLOW

CATHETER BLOOD FLOW BLOOD VESSEL

have been published. However, there is the problem of adequately fitting masks, lack of control of ventilation, waste of anaesthetic gases and hazardous pollution of the operating room. However, intubation procedure for small laboratory animals requires special miniaturised equipment and skill to perform safely oro-endotracheal intubation. Numerous acceptable techniques for this application procedure of rats including blind intubation are reported. More information about many of these methods with rats can be found by Nebendahl (2000); Heard (2004) or Henke and Erhardt (2004). Also, the majority of animal research is based on mouse models, but there are only few reports about endotracheal intubation of mice. All these reports stressed that direct visualisation of the glottis is indispensable. For example Costa et al. (1986) developed a fiberoptic laryngoscope for transoral tracheal intubation technique of mice and Brown et al. (1999) described a simple constructed metal blade used as a laryngoscope to facilitate oropharyngeal exposure and transillumination of the neck to facilitate visualising of the trachea. Vergari et al. (2003) used to facilitate the intubation tube entering the trachea a small bore, straight fibre-optic arthroscope and Spoelsta et al. (2007) a purpose-made laryngoscope and a self-built plastic support. Hamacher et al. (2008) described in detail a microscopic wire guide-based orotracheal

mouse intubation. A corresponding didactic video of this method is available at: www.ltk.uzh.ch (animal welfare/techniques/intubation mice). For starting intubation of small laboratory animals, an endotracheal cannula of suitable size must be selected. For instance, Proctor and Fernando (1973) used for the intubation of rats a 16 G nylon intravenous cannula. Flecknell (1996) recommended the use of a 12 or 16 G arterial cannula and Remie et al. (1990) a modified 14–18 G Haemalock arterial stick and Costa et al. (1986) a 16 G Teflon tubing for rats. Some authors recommend modification of the introductory end of the tube to facilitate penetration of the tracheal lumen or to prevent unintentional intubation of a bronchus and the Luer fitting to provide connection to an anaesthetic circulation (Proctor and Fernando 1973; Flecknell 1996) Prior to intubation, the animal is anaesthetised to a sufficient depth to abolish the cough and swallowing reflex. For premedication, sedatives and tranquilisers together with injectable anaesthetics are administered or a face mask/nose cone connected to the anaesthesia machine or a transparent induction chamber, which should have both an inlet for delivery a fresh gas and an outlet for effective removal of waste anaesthetic gases (Otto 2004) can be used. Atropin administration (0.04 mg/kg) is useful in reducing mucus secretion and helps to prevent tube blockage (Flecknell 1996). Laryngospasm may be prevented or be alleviated by

114

K. Nebendahl and P. Hauff

ventilation and causes respiratory difficulties, the cannula must be exchanged with a new one.

8.5.6.5 Intratracheal/Intrabronchial Administration Some experimental protocols require that substances be given intratracheal or intrabronchial, for these studies intubation techniques can be used too (Lany et al. 1980; Costa et al. 1989; Smith 1991). The administered volume for a rat (bodyweight: 200 g) should be not larger than 0.04 mL. Fig. 8.15 Insertion of an endotracheal tube (over-the-needle catheter) into the trachea of a mouse using the Seldinger technique

spraying a local anaesthetic solution on the cannula or on the vocal cords before intubation. A careful examination of the pharynx and larynx must be carried out and if the laryngeal region is covered with mucus, the area must be cleared with a cotton – tipped applicator to allow visualisation. Techniques of intubation are described in detail for rats by Flecknell (1996), Heard (2004) and Henke and Erhardt (2004). In the following, an intubation method for mice under direct vision using adopted Seldiger vessel cannulation technique is described in detail. The anaesthetised mouse is positioned in dorsal recumbence on a slant board. An elastic band is than affixed to the upper incisors and fastened to the board, to extend the head and neck. The end of a flexible fibre-optic rod of a powerful light source is positioned so that it contacts the surface of the skin in front of the neck near the pharyngoepiglottis region. The light penetrating the tissue will illuminate the larynx and the focal cords can be located. The tongue pulled gently forward and to one side. A slight pressure exerted on the ventral surface of the neck may further facilitate visualisation of the glottis. Now a tube – Costa et al. (1986) recommend a 19 G cannula for mice and a 16 G cannula for rats and Spoelsta et al. (2007) a 24 G intravenous catheter – must be brought via the pharynx alongside the epiglottis, through the laryngeal opening into the trachea (Fig. 8.15). After connecting the tube to the ventilator the animal must be ventilated and chest movement and air exchange must be checked for correct cannula placement. If the tube becomes plugged with mucus secretion during

References Bardelmeijer HA et al (2002) Cannulation of the jugular vein in mice: a method for serial withdraw of blood samples. Lab Anim 37:181–187 Barrow RH (1968) Modified intravenous injection technique in the rat. Lab Anim Care 18:570–571 Baumans V et al (1993) Experimental procedures. In: van Zutphen LFM, Baumans V, Beynen AC (eds) Principles of laboratory animal science. Elsevier, Amsterdam, pp 299–318 Bihun C, Bauck L (2004) Basic anatomy, physiology, husbandry, and clinical techniques. In: Quesenberry KE, Carpenter JW (eds) Ferrets, rabbits, and rodents: clinical medicine and surgery. W.B. Saunders, Philadelphia, pp 286–315 Birkhahn RH et al (1976) The stress response in the rat from harnessing for chronic intravenous infusion. J Surg Res 21:185–190 Born CT, Moller ML (1974) A simple procedure for long-term intravenous infusion in rats. Lab Anim Sci 24:355–358 Brown RH et al (1999) A method of endotracheal intubation and pulmonary functional assessment for repeated studies in mice. J Appl Physiol 87:2362–2365 CCAC (Canadian Council on Animal Care) (2002) Guidelines on acceptable immunological procedures. CCAC, Ottawa. http://www.ccac.ca/gui_pol/policies Claassen V (ed) (1994a) Oral drug administration. In: Neglected factors in pharmacology and neuroscience research. Elsevier, Amsterdam, pp 59–115 Claassen V (ed) (1994b) Food and water intake. In: Neglected factors in pharmacology and neuroscience research. Elsevier, Amsterdam, pp 267–286 Claassen V (ed) (1994c) Fasting. In: Neglected factors in pharmacology and neuroscience research. Elsevier, Amsterdam, pp 290–320 Claassen V (ed) (1994d) Food restriction. In: Neglected factors in pharmacology and neuroscience research. Elsevier, Amsterdam, pp 321–342 Claassen V (ed) (1994e) Intraperitoneal drug administration. In: Neglected factors in pharmacology and neuroscience research. Elsevier, Amsterdam, pp 46–56

8 Drug Administration Cocchetto DM, Bjornsson TD (1983) Methods for vascular access and collection of body fluids from the laboratory rat. J Pharm Sci 72:465–492 Costa DL et al (1986) Transoral trachea intubation of rodents using a fiberoptic laryngoscope. Lab Anim Sci 36: 256–261 Conybeare G et al (1988) A improved simple technique for the collection of blood samples from rat and mice. Lab Anim 22:177–182 Cox CE, Beazley RM (1975) Chronic venous catheterization: a technique for implanting and maintaining venous catheters in rats. J Surg Res 18:607–610 Desjardins C (1986) Indwelling vascular cannulas for remote blood sampling, infusion, and long-term instrumentation of small laboratory animals. In: Gay WJ, Heavner JE (eds) Methods of animal experimentation, vol VII, part A. Academic, Orlando, pp 143–194 Diel K-H et al (2001) A good practice guide to the administration of substances and removal of blood, including routes and volumes. J Appl Toxicol 21:15–23 Dudley W et al (1975) An apparatus for anesthetizing small laboratory animals. Lab Anim Sci 25:481–482 ETS 123 (1986) European convention for the protection of vertebrate animals used for experimental and other scientific purpose. Council of Europe, Strasbourgh Everett JW, Sawyer CH (1956) The small saphenous vein as a route for intravenous injection in the white rat. Nature 178:268–269 Fallon MT (1996) Rats and mice. In: Label-Laird K, Swindle MM, Flecknell P (eds) Handbook of rodent and rabbit medicine. Pergamon, Oxford, pp 1–38 Flecknell P (ed) (1996) Anaesthesia. In: Laboratory animal anaesthesia, 2 edn. Academic, London, pp 15–73/159–237 Ferrill HC, Hill C (1943) A simplified method for feeding rats. J Lab Clin Med 28:1624–1625 Franklin CA et al (1989) Use of percutaneous absorption data in risk assessment. J Am Coll Toxicol 8:815–827 Gervaz P et al (2000) Permanent access to the portal vein in rats: an experimental model. Eur Surg Res 32:203–206 GV–SOLAS Recommended maximum injection volumes in lab animals of the Society for Laboratory Animal Science (GV-SOLAS), August 2010, http://www.gv-solas.de/auss/ tie/Injektionsvol_August_2010.pdf Grice HC (1963) Methods for obtaining blood and for intravenous injections in laboratory animals. Lab Anim Care 14:483–493 Guy RH, Hadgraft J (1991) Principles of skin permeability relevant to chemical exposure. In: Hobson DW (ed) Dermal and ocular toxicology. Fundamentals and Methods. CRC, Boca Raton, pp 221–246 Hamacher J et al (2008) Microscopic wire guide-based orotracheal mouse intubation: description, evaluation and comparison with transillumination. Lab Anim 42:222–230 Harms PG, Ojeda SR (1974) A rapid and simple procedure for chronic cannulation of the rat jugular vein. J Appl Physiol 36:391–392 Heard DJ (2004) Anestesia, analgesia, and sedation of small mammals. In: Quesenberry KE, Carpenter JW (eds) Ferrets, rabbits, and rodents: clinical medicine and surgery. W.B. Saunders, Philadelphia, pp 356–369

115 Henke J, Erhardt W (2004) Nager. In: Henke J et al (eds) Anästhesie & Analgesie beim Klein und Heimtier. Schattauer, Stuttgart, pp 642–663 Hurtubise MR et al (1980) Restoring patency of occluded central venous catheters. Arch Surg 115:212–213 Jeffrey P et al (1987) Does the rat have an empty stomach after an overnight fast? Lab Anim 21:330–334 Joint Working Group on Refinement (1993) Removal of blood from laboratory mammals and birds. Lab Anim 27:1–22 Kraus AL (1980) Research methodology. In: Baker HJ HJ et al (eds) The laboratory rat, vol 2, Research applications. Academic, New York, pp 19–23 Lany J et al (1980) A simple method for exact intratracheal installation in small laboratory rodents. Cancer Lett 10:91–94 Lax ER et al (1983) A simple method for oral administration of drugs in solid form to fully conscious rats. Lab Anim 17:50–54 Levine RR (1970) Factors affecting gastrointestinal absorption of drugs. Am J Dig Dis 15:171–188 Levy DE et al (1980) A mask for delivery of inhalation gases to small laboratory animals. Lab Anim Sci 30:868–870 Lewis R et al (1966) Error of intraperitoneal injections in rats. Lab Anim Care 16:505–509 Little JR et al (1962) Determination of lymphocyte turnover by continuous infusion of H 3-Thymidine. Blood 19: 236–242 Löscher W et al (2006) Allgemeine Pharmakologie. In: Löscher W, Ungemach FR, Kroker R (eds) Pharmakotherapie bei Haus- und Nutztieren, 7th edn. Parey, Berlin, pp 4–7 Marini JC, Garlick PJ (2006) Non-surgical alternatives to invasive procedures in mice. Lab Anim 40:275–281 Moreland AF (1965) Collection and withdraw of body fluids and infusion techniques. In: Gay WJ (ed) Methods of animal experimentation, vol 1. Academic, New York, pp 1–42 Nachtman RG et al (1988) Commercial over-the-needle catheters for intravenous injections and blood sampling in rats. Lab Anim Sci 38:629–630 Nebendahl K (2000) Routes of administration. In: Krinke GJ (ed) The laboratory rat. Academic, San Diego, pp 463–483 Nightingale CH, Mouravieff M (1973) Reliable and simple method of intravenous injection into the rat. J Pharm Sci 62:860–861 Norris ML, Miles P (1982) An improved, portable machine designed to induce and maintain surgical anaesthesia in small laboratory animals. Lab Anim 16:227–230 O’Neill PJ, Kaufman LN (1990) Effects of indwelling arterial catheters or physical restraint on food consumption and growth patterns of rats: advantages of non-invasive blood pressure measurement techniques. Lab Anim Sci 40:641–643 Otto K (2004) Anesthesia, analgesia and euthanasia. In: Hedrich H (ed) The laboratory mouse. Elsevier, San Diego, pp 555–569 Pearce KA (1957) A route for intravenous injection in the albino rat. Nature 178:709 Pekow CA, Baumans V (2003) Common nonsurgical techniques and procedures. In: Hau J, van Hoosier GL (eds) Handbook of laboratory animal science, vol II, 2nd edn. CRC, Boca Raton, pp 351–390

116 Petty C (1982) Blood. In: Petty C (ed) Research techniques in the rat. Charles C. Thomas, Springfield, pp 66–107 Pope RS (1968) Small vessel cannulator. J Appl Physiol 24:276 Popovic V, Popovic P (1960) Permanent cannulation of aorta and vena cave in rats and squirrels. J Appl Physiol 15:727–728 Proctor E, Fernando AR (1973) Oro-endotracheal Intubation in the rat. Br J Anaesth 45:139–142 Remie R et al (1990) Anaesthesia of the laboratory rat. In: van Dongen JJ et al (eds) Manual of microsurgery on the rat. Elsevier, Amsterdam, pp 159–169 Rhodes ML, Patterson CE (1979) Chronic intravenous infusion in the rat: a nonsurgical approach. Lab Anim Sci 29: 82–84 Ritskes-Hoitinger M, Chaewalibog A (2003) Nutrition requirements, design, and feeding schedule in animal experimentation. In: Hau J, van Hoosier GL (eds) Handbook of laboratory animal science, vol II, 2nd edn. CRC, Boca Raton, pp 351–390 Ritskes-Hoitinger M (2004) Nutrition of laboratory mice. In: Hedrich H (ed) The laboratory mouse. Elsevier Academic, London, pp 463–479 Salem H et al (1963) Mico-method for intravenous injection and blood sampling. J Pharm Sci 52:794–795 Schlingmann F et al (1997) Food deprivation: how long and how. In: O’Donoghue PN (ed) Harmonization of laboratory animal husbandry. Proceedings of the Sixth FELASA Symposium. Royal Society of Medicine Press, London, pp 89–92 Schneider G, Schneider G (1970) Zur Technik der intraperitoneal Injektion bei Ratten. Z Versuchstierk 12:16–19 Schumacher J et al (2003) A miniature self-inflating bag-mask ventilator for rats. Lab Anim 37:360–362 Shimizu S (2004) Routes of administration. In: Hedrich H (ed) The laboratory mouse. Elsevier, San Diego, pp 527–542 Smith G (1991) A simple non-surgical method of intrabronchial installation for the establishment of respiratory infections in the rat. Lab Anim 25:46–49 Spoelsta EN et al (2007) A novel and simple method for endotracheal intubation of mice. Lab Anim 41:128–135 Steiger E et al (1972) A technique for long-term intravenous feeding in unrestrained rats. Arch Surg 104:330–332 Suckow MA (2001) Experimental methodology. In: Suckow MA et al (eds) The laboratory mouse. CRC, Boca Raton, pp 113–134 Szczesny G et al (2003) Long-term anaesthesia using inhalatory isoflurane in different strains of mice – the haemodynamic effects. Lab Anim 38:64–68 Tiersch G (2006) “Animal Welfare Act”, Tierschutzgesetz vom 18 .Mai. 2006 (BGBL.I S.1206,1313) www.bmelv.de TierschutzTiergesundheit/Tierschutz/Tierschutzrecht/ Tierschutzgesetz

K. Nebendahl and P. Hauff Tiierschutzbericht (2007) www.bmelv.de TierschutzTiergesundheit /Tierschutz/ Tierschuzberichte Vergari A et al (2003) Video-assisted orotracheal intubation in mice. Lab Anim 37:204–206 Vermeulen JK et al (1997) Food deprivation: common sense or nonsense? Anim Technol 48:45 Videm S (1980) A method for blood sampling and intravenous injections in rats. Z Versuchstierk 22:101–104 Waynforth HB, Parkin R (1969) Sublingual vein injection in rodents. Lab Anim 3:35–37 Waynforth HB, Flecknell PA (1992a) a) Administration of substances. In: Waynforth HB, Flecknell PA (eds) Experimental and surgical technique in the rat, 2nd edn. Academic, London, pp 1–67 Waynforth HB, Flecknell PA (1992b) Specific surgical operations. In: Waynforth HB, Flecknell PA (eds) Experimental and surgical technique in the rat, 2nd edn. Academic, London, pp 203–310 Watanabe J et al (1977) Gastric emtying rate constants after oral administration of drug solutions to mice, rats and rabbits. Chem Pharm Bull (Tokyo) 25:2147–2155 Weeks JR, Davis JD (1964) Chronic intravenous cannulas for rats. J Appl Physiol 19:540–541 Weihe WH (1987) The laboratory rat. In: Poole TB (ed) The UFAW handbook on the care & management of laboratory animals, 6th edn. Longman, Essex, pp 309–330 Weiss J et al (2009) Verabreichung (Applikation) von Substanzen und Probeentnahme. In: Weiss J, Becker K, Bernsmann K, Dietrich H, Nebendahl K (eds) Tierpflege in Forschung und Klinik. Enke, Stuttgart, pp 362–370 Wester RC, Maibach HI (1986) Dermatopharmakinetics: a dead membrane or a complex multifunctional viable process. In: Bridges JW, Chasseaud LF (eds) Progress in drug metabolism, vol 9. Taylor & Francis, London, pp 95–109 Wolfensohn S, Lloyd M (1994) Procedural data. In: Wolfensohn S, Lloyd M (eds) Handbook of laboratory animals management and welfare. Oxford University Press, Oxford, pp 143–147 Working Party Report (2001) Refining procedures for the administration of substances. Lab Anim 35:1–41 Woodard G (1965) Principles of drug administration. In: Gay WJ (ed) Methods of animal experimentation, vol 1. Academic, New York, pp 343–359 Yoburn BC, Morales R, Inturrisi CE (1984) Chronic vascular catheterization in the rat: comparison of three techniques. Physiol Behav 33:89–94

Part Imaging Modalities and Probes

III

9

How to Choose the Right Imaging Modality Fabian Kiessling, Bernd J. Pichler, and Peter Hauff

Contents References............................................................................ 124

F. Kiessling (*) University of Aachen (RWTH), Pauwelsstrasse 20, 52074 Aachen, Germany e-mail: [email protected] B.J. Pichler Department of Radiology, Laboratory for Preclinical Imaging and Imaging Technology of the Werner Siemens-Foundation, University of Tuebingen, 72076 Tuebingen, Germany P. Hauff Global Drug Discovery, Bayer Schering Pharma AG, Müllerstr. 178, 13353 Berlin, Germany

For noninvasive imaging of small animals, a bunch of different modalities are available, about which a detailed introduction will be given in the following chapters. All imaging modalities have strengths and limitations and a noninvasive imaging study will be most effective if the choice of the imaging modality depends on the parameter of interest under consideration of costs and measurement time per animal (Fig. 9.1 and Table 9.1). Certainly, one frequent demand is the determination of the size of a single lesion – for example of a subcutaneous tumor. Previously, size determination was mostly done using a caliper. Caliper measurements, however, only provide two dimensions, are user-dependant, and afflicted with large measurement errors for polymorphic lesions. In addition, reproducibility of the data is limited. For lesion size determination, ultrasound, CT, and MRI can be considered to be the most suitable imaging modalities. Ultrasound is most cost- and time-effective and measurements can be performed in the laboratory environment. However, ultrasound is user-dependent unless an automated scan method is applied, which is implemented in recent high-frequency ultrasound devices (Foster et al. 2002). Unfortunately, it is difficult to perform automated whole body ultrasound scans. Manual ultrasound scans, however, are marked by a lower reproducibility and reliability, which limits the suitability of ultrasound for whole mouse and whole rat imaging, respectively, and for screening for multiple lesions. Scanning with CT and MRI has the advantage that an entire animal can be investigated easily and that in clinical scanners (or scanners with a large gantry opening/bore) more than one animal can be examined at the same time (Bock et al. 2003). However, when using CT, the X-ray dose deposited at the animal has to be

F. Kiessling and B.J. Pichler (eds.), Small Animal Imaging, DOI: 10.1007/978-3-642-12945-2_9, © Springer-Verlag Berlin Heidelberg 2011

119

120

F. Kiessling et al. orthotopic

CT, MRI

subcutaneous

US, MRI, CT, (OI)

abdomen

MRI, CT

lung

CT, MRI

orthotopic

MRI, (CT, US)

subcutaneous

US, MRI, (CT)

single lesion Morphology multiple lesions

single lesion Vascularisation + Morphology multiple lesions

MRI, (CT)

single lesion Molecular markers multiple lesions

angiogenesis, inflammation

US, OI, PET, SPECT

other applications

OI, PET, SPECT

PET, SPECT, OT (+ morphology from CT/MRI)

proof of principle

OI, PET, SPECT

drug biodistribution

PET, SPECT, (+ ex vivo γ-counting)

drug labeling Drug Development evaluation of drug efficacy

see above

Fig. 9.1 Sketch illustrating major indications for small animal imaging and the recommended imaging modality. The order of the imaging modalities represents their priority for the particular imaging indication. Please note that this recommendation has

been composed based on the author’s experiences and focuses on broad and high-throughput implementation of small animal imaging. OI optical imaging; OT optical tomography; VCT volumetric CT

considered. It can easily exceed 250 mGy for a whole body mouse scan and significantly accumulate in repetitive scans (Bartling et al. 2007). In this case, a relevant influence on the biological system and the pathology cannot be excluded anymore. Nevertheless, when scanning with considerably low resolution (>150 mm voxel side length), a higher throughput can be achieved with CT (less than 1 min measurement time/mouse for some dedicated mCT scanners) as compared with MRI (usually 5–20 min for a morphologybased scan protocol). On the other hand, it is an advantage of MRI that sequences with different weighting can be combined, which enables to depict tissue edema or necrotic areas easily. The excellent soft tissue contrast can also be important to localize and delineate lesion of interest.

This particularly holds true for pathologies that are not below the skin such as pancreatic or prostate tumors (Hamzah et al. 2008; Kiessling et al. 2004). In CT and ultrasound contrast material may be applied to enhance the contrast, but carries the risk of inducing side effects and infections on the animal. Especially for CT, dedicated contrast agents are required ensuring a long blood retention rate. For mouse CT imaging, the volume of contrast agent to be applied can easily reach 300– 400 mL, which is not only questionable concerning animal use and care guidelines, but also cause physiological effects. Furthermore, contrast agent administration is time-consuming and cost-intensive. The size of the lesion also has to be considered. Relevant measurement errors may derive from partial volume effects and thus small lesions should be imaged

121

9 How to Choose the Right Imaging Modality

Table 9.1 This table gives an overview on approximated scan times for a mouse/rat using different imaging devices and for different applications Typical measurement times per animal MRI

Morphologic scan

10–30 min

+Vascularizationa/spectroscopy

+20 min

Molecular imaging (including relaxometry)

30 min and more

Morphologic scan

20 s to 1 h (depending on spatial resolution and scanner type)

+ Vascularization

+1–15 min

Morphologic scan

1–5 min

Morphologic scan + Doppler

1–5 min

Contrast-enhanced scana

5–15 min

Molecular imaging

~5–15 min

OI/OTb

All applicationsa

Few seconds to minutes

PET/SPECT

All applicationsa

10–30 min

mCTb

US

Please note that the scan times can highly vary depending on the scanner type as well as on the required image resolution and contrast-to-noise ratio. This particularly holds true for mCT and high-field MRI. Our data refer to standard applications with a moderate image resolution. OI: Optical imaging, OT: Optical tomography a Five to ten additional minutes are required for contrast agent injection/animal catheterization b Consider that long image reconstruction times may occur

with high spatial resolution. This, however, would demand for higher X-ray exposure in CT. As a consequence, for size determination of small orthotopic lesions outside the lung, MRI using dedicated small animal scanners is the method of choice. In oncology research optical imaging of rodents carrying transplants of cells that are transfected with fluorescent proteins like green fluorescent protein (GFP) and red fluorescent protein (RFP) may also be applied to estimate tumor size (Choy et al. 2003; Hoffman 2005). However, it is important to note that the transfection of cells with fluorescent proteins can alter their biological properties and that the fluorescence will only be present in viable parts of tumors. In addition, one has to consider that two-dimensional optical imaging is surface-weighted and that the lesion size may be under or overestimated depending on the distance of the tumor from the skin surface. Phenotyping of transgenic animal is another field of broad interest. In this context, it is often necessary to scan a high number of animals and to characterize several organs simultaneously including the skeleton and the vessels. MRI and mCT may both be used. If skeletal abnormalities or calcifications are expected mCT may be preferential, while abnormalities in abdominal organs often are better seen in MRI. For the determination of

body fat – an important biomarker for animal models of diabetes – specialized commercial systems are available that either base on NMR (Nixon et al. 2010) or X-rays (Stiller et al. 2007). Clinical studies showed that a pathologic process often is not described sufficiently by morphological information and that lesion size can have limited impact on the patient’s prognosis for many diseases. In addition, the evaluation of novel therapeutic drugs in the clinics can be facilitated significantly if the preclinical and clinical imaging protocols are adjusted to each other. Therefore, besides the assessment of tumor size, there is increasing demand for a more sophisticated imaging beyond morphology. In this regard, MRI offers a wide range of applications to supplement the morphological information. Some of them can be applied without the need to inject contrast agents and thus can easily be integrated in the scan protocols with only moderate additional measurement time requirement. Information about metabolism can be obtained by MR-spectroscopy; vessels morphology and tissue perfusion can be assessed by time of flight angiography or arterial spin-labeling (ASL) techniques. Diffusion-weighted imaging provides an important measure of cell density (Semmler et al. 2007).

122

For a detailed vascular characterization of tissues, however, contrast material has to be administered. Dynamic contrast-enhanced (DCE) MRI can provide valuable information about relative blood volume, vascular perfusion, and permeability and can be combined with relaxometric measurements of the relative blood volume and with sophisticated imaging techniques like “vessel size imaging” providing a measure of the mean microvascular diameter in the tissue (Troprès et al. 2001). However, when intending to use these techniques in animal experiments, one should consider that approximately 5–10 min are additionally required for animal preparation (particularly for catheterization) and, that depending on the scan protocol, the measurement time will increase by 10 to >30 min. Thus, at least 30 min examination time should be scheduled per animal. Another important point is that these contrastenhanced examinations more frequently fail due to dislocation of the catheter or motion of the animal. Thus, when planning a study that includes the analysis of tissue vascularization by contrast-enhanced scans, the number of animals should be chosen higher as for a purely morphology-based examination. For vascular characterization also, ultrasound offers some interesting options (Kiessling et al. 2009). Without the need of contrast agent administration, Doppler techniques can be used to visualize vessels by their flow and thus to assess the relative blood volume in the tissue. It can also be used to estimate blood flow velocities and determine cardiac data (echocardiography). However, it has to be considered that the sensitivity of Doppler for blood flow depends on the pulse frequency, and with devices operating between 3 and 40 MHz, slow blood flow in microvessels may not be detected sufficiently. Microbubbles can be administered as contrast agent and sensitive Doppler- and B-mode scan techniques are available to assess microvascular characteristics (Kiessling et al. 2009). An extended explanation of the ultrasound techniques for the determination of relative blood volume, blood velocity, and perfusion is given in Chaps. 15 and 16. Vessel permeability cannot be determined since mic robubbles do not leave the vascular bed. Microbubbles can also be loaded with specific ligands to act as molecular probes (Kiessling et al. 2009; Kaufmann and Lindner 2007). While it is certainly a limitation that only intravascular targets can be addressed, the high sensitivity of ultrasound for these microbubbles and the option of scanning with

F. Kiessling et al.

exquisite spatial resolution make ultrasound highly attractive for research on angiogenesis and inflammation. In this regard, it might be of interest to know that in mice and rats the method also works reliably in the brain without the need to open the skull (Reinhardt et al. 2005). In conclusion, ultrasound is highly suited for functional and molecular vascular imaging of superficially localized tissues like skeletal muscle, subcutaneous tumors, and the brain. Cardiac functional imaging is another field where ultrasound can compete with other imaging modalities. In the abdomen, however, identification of the lesion can be difficult and the optimal access way may often be occupied by bowel gas, which gives advantage to other imaging modalities. Here, to the author’s experience, DCE MRI and DCE CT may be chosen for functional and PET, SPECT, or optical imaging for molecular imaging demands. Although in preclinical research molecular imaging has rather exceeded proof of principle status, it can be expected to gain striking importance in future for drug development, the evaluation of drug efficacy, and therapy response. PET and SPECT may be preferred for testing drug biodistribution and assessing target binding kinetics because these modalities have the highest sensitivity for contrast agents. In addition, as compared with signaling molecules of other modalities like dyes or nanoparticles, the conjugation of a radioisotope less affects the properties of a biomolecule. The possibility of absolute quantification and the option of a relatively easy clinical translation are further advantages of nuclear imaging methods. However, the need of radioactive tracers raises special demands on animal handling, staff education, laboratory safety, and expensive infrastructure. Particularly, for studies with large animal numbers, the total amount of radioactivity handled can become significant. Thus, for initial experiments on the binding of many new drugs, optical methods may be preferred since optical imaging is cheap and the probes are easy to handle. Nevertheless, it should be kept in mind that two-dimensional optical imaging is not quantitative, surface-weighted and that even in optical molecular tomography quantification is difficult. In addition, for small molecules like glucose, amino acids and nucleic acids coupling of a dye will most probably change its pharmacokinetics dramatically or inhibit its uptake by cellular transporters. The same holds true for RNA

9 How to Choose the Right Imaging Modality

chains, which should be radiolabeled if it is the aim to study its natural biodistribution. The situation is different if it is not the aim to accurately assess the biodistribution and the accumulation of a biomolecule of a therapeutic substance, but to monitor its effect on tissue metabolism, cellular proliferation, angiogenesis, apoptosis, or on the general molecular tissue profile with targeted probes. These targeted probes just have to work reliably and must not be close in its binding characteristics and biodistribution to the natural ligand. In this context, it can be reasonable to supplement data about lesion size with clinical methods like [18F] FDG-PET quantifying glucose metabolism and [18F] FLT-PET indicating cell proliferation. Using these tools, therapy response and the development of therapy resistance can often be detected earlier as compared with lesion size determination alone. Furthermore, metabolic imaging allows to directly assess pathway properties using specific probes. Alternatively, one may intend to develop novel specific probes to more directly and more effectively monitor therapy response by a decrease in target density or by the change of a molecular marker profile. If it is the intention to solely investigate the molecular regulation on preclinical level, one is considerably free in the choice of the imaging modality. In this context the strengths and limitations of the available imaging modalities are summarized in Fig. 9.1. However, if it is intended to develop noninvasive imaging procedures for monitoring and personalizing the use of novel therapeutic drugs in later clinical studies, the imaging modality should be chosen depending on the clinical need. Here, PET may be superior in most cases that go along with whole body imaging. Optical imaging may be used if intraoperative imaging is intended or if endoscopic imaging (like bronchoscopy, gastroscopy, or colonoscopy) will be applied. Molecular ultrasound may also gain clinical importance for imaging that goes along with interventions like needle biopsy of breast and prostate lesions, for therapy response control between whole body staging intervals, and for cardiovascular analysis. Some readers may wonder that in regard of molecular imaging MRI was not mentioned so far, although some exciting studies were published using targeted paramagnetic liposomes and USPIO (Kiessling et al. 2007). However, these studies report more on successful proof of principle studies than on the use of MRI to

123

answer a complex biological question. Before the latter can be achieved, some significant improvements are required that are not easy to achieve. Firstly, the sensitivity of MRI to contrast agents is several magnitudes lower than that of optical and nuclear imaging methods. Thus, targeted drugs have to be administered in high dosages, which carry the risk of inducing biological effects on the target tissue. Furthermore, large signaling molecules have to be coupled to the ligand in order to make it detectable. This makes the optimization of the pharmacokinetic properties of the compound difficult. Thus, a significant percentage of the targeted probes may accumulate unspecifically in the interstitial space or in necrotic tissue or is removed from the blood by the reticulo endothelial system. These limitations may be overcome in future and there might be high potential for more recent molecular MRI techniques like CEST and PARACEST (Modo and Bulte 2007) or for MRI of nuclei like fluorine. However, to date, it is too early to recommend MRI as a routine molecular imaging tool in small animal research. Considering all these different demands on preclinical imaging, it becomes clear that an all-around imaging modality is not available and that for a complex scientific question the different imaging modalities have to be combined. As a consequence, the manufacturers offer tools for retrospective fusion of CT and MRI data with PET, SPECT, and optical imaging data. Even some dual modality imaging devices (e.g., mSPECT–CT and mPET–CT) were commercialized for small animal imaging recently and PET inserts for small animal MRI scanners were constructed at different places (Judenhofer et al. 2008). In this regard, there is an intense debate in the scientific community if these dual modality scanners really improve imaging efficacy. Certainly, for some demands it is highly attractive to have the option of assessing different information simultaneously. Particularly, the combination of MRI with PET, SPECT, and optical imaging would enable an ideal combination of functional and targeted imaging. Very often, high resolution morphological images from CT or MRI are required to accurately quantify the molecular in formation from PET, optical, or SPECT by providing landmarks for defining the region of interests or correcting partial volume effects. In addition, the implementation of dual modality devices saves laboratory space and thus can reduce costs. On the other hand, it can hamper high-throughput imaging significantly when only having a dual modality scanner instead of two

124

separated devices since, for example, long SPECT scans can consume the measurement time availability required for a routine mCT study.

References Bartling SH, Stiller W, Semmler W, Kiessling F (2007) Small animal computed tomography imaging. Curr Med Imaging Rev 3:45–49 Bock NA, Konyer NB, Henkelman RM (2003) Multiple-mouse MRI. Magn Reson Med 49(1):158–167 Choy G, Choyke P, Libutti SK (2003) Current advances in molecular imaging: noninvasive in vivo bioluminescent and fluorescent optical imaging in cancer research. Mol Imaging 2:303–312 Foster FS, Zhang MY, Zhou YQ et al (2002) A new ultrasound instrument for in vivo microimaging of mice. Ultrasound Med Biol 28:1165–1172 Hamzah J, Jugold M, Kiessling F, Rigby P, Manzur M, Marti HH, Rabie T, Kaden S, Gröne H-J, Hämmerling GJ, Arnold B, Ganss R (2008) Vascular normalisation in RGS5–deficient tumours promotes immune destruction. Nature 453:410–414 Hoffman RM (2005) The multiple uses of fluorescent proteins to visualize cancer in vivo. Nat Rev Cancer 5:796–806 Judenhofer MS, Wehrl HF, Newport DF, Catana C, Siegel SB, Becker M, Thielscher A, Kneilling M, Lichy MP, Eichner M, Klingel K, Reischl G, Widmaier S, Rocken M, Nutt RE, Machulla HJ, Uludag K, Cherry SR, Claussen CD, Pichler PBJ (2008) Simultaneous a new approach for functional and morphological imaging. Nat Med 14:459–465

F. Kiessling et al. Kaufmann BA, Lindner JR (2007) Molecular imaging with targeted contrast ultrasound. Curr Opin Biotechnol 18:11–16 Kiessling F, Huber PE, Grobholz R, Heilmann M, Meding J, Lichy MP, Fink C, Krix M, Peschke P (2004) Schlemmer HP dynamic MR-tomography and proton MR spectroscopy of prostate cancers in rats treated by radiotherapy. Invest Radiol 39:34–44 Kiessling F, Jugold M, Woenne EC, Brix G (2007) Non-invasive assessment of vessel morphology and function in tumors by magnetic resonance imaging. Eur Radiol 17:2136–2148 Kiessling F, Huppert M, Palmowski M (2009) Functional and molecular ultrasound imaging: concepts and contrast agents. Curr Med Chem 16:627–642 Modo MJ, Bulte JMW (2007) Molecular and cellular MR imaging. CRC, Boca Raton Nixon JP, Zhang M, Wang C, Kuskowski MA, Novak CM, Levine JA, Billington CJ, Kotz CM (2010) Evaluation of a quantitative magnetic resonance imaging system for whole body composition analysis in rodents. Obesity 18:1652–1659 Reinhardt M, Hauff P, Linker RA, Briel A, Gold R, Rieckmann P, Becker G, Toyka KV, Maurer M, Schirner M (2005) Ultrasound derived imaging and quantification of cell adhesion molecules in experimental autoimmune encephalomyelitis (EAE) by Sensitive Particle Acoustic Quantification (SPAQ). Neuroimage 27:267–278 Semmler W, Reiser MF, Hricak H (2007) Magnetic resonance tomography. Springer, Berlin Stiller W, Kobayashi M, Kazuhiko K, Stampfl U, Richter G, Wolfhard S, Kiessling F (2007) Initial experience with a novel low-dose micro-CT system. Röfo 179:669–675 Troprès I, Grimault S, Vaeth A et al (2001) Vessel size imaging. Magn Reson Med 45:397–408

X-Ray and X-Ray-CT

10

Willi A. Kalender, Paul Deak, Klaus Engelke, and Marek Karolczak

Contents 10.1 2D X-Ray Projection Imaging............................. 126 10.1.1 Radiography............................................................ 126 10.1.2 Angiography........................................................... 126 10.1.3 Bone Densitometry................................................. 126 10.2 3D X-Ray CT Imaging Principles (Micro-CT)............................................................ 127 10.2.1 In Vitro Micro-CT Imaging.................................... 128 10.2.2 In Vivo Micro-CT Imaging..................................... 129 10.3 Quality Control and Dose Considerations.......... 131 10.3.1 Image Quality......................................................... 131 10.3.2 Dose Assessment.................................................... 132 10.3.3 Dose Optimization.................................................. 133 10.4 Advanced Micro-CT Techniques......................... 134 10.4.1 Dynamic Micro-CT................................................ 134 10.4.2 Dual-Energy CT...................................................... 136 10.4.3 Image Postprocessing and Rendering..................... 136 10.5

Conclusions and Recommendations.................... 137

References............................................................................ 138

W.A. Kalender (*), P. Deak, K. Engelke, and M. Karolczak Institute of Medical Physics, University Erlangen-Nuernberg, Henkestrasse 91, 91052 Erlangen, Germany e-mail: [email protected]

Since their discovery in 1895, X-rays have been widely used for imaging humans. Recently, they have also gained an importance in small animal imaging (SAI). Most techniques known from clinical medicine, including single- and dual-energy X-ray imaging, have been successfully ported to SAI and are the subject of this chapter. As trivial as it is, simple X-ray examinations may bring diagnostically valuable information in a variety of applications. Unenhanced radiography reveals skeletal anatomy, contrast-enhanced imaging allows improved visualization of the vasculature and strongly vascularized areas, and dedicated methods such as bone densitometry deliver quantitative information. In analogy to clinical X-ray imaging, we will separately describe standard two-dimensional (2D) projection imaging and the more advanced three-dimensional (3D) computed tomography (CT) imaging techniques. Also in analogy to clinical applications, CT is considered to be of significantly higher importance as it provides more information and possibilities than conventional 2D approaches. It will therefore be covered in much more detail. The X-ray instrumentation used for small animal examinations does not differ, in principle, from that known from clinical medicine. The main difference lies in the required higher resolution levels, which implies the use of special components: X-ray tubes with very small focal spots, so-called micro-focus tubes, and image detectors with small detector elements in order to produce sharp images. Respective micro-focus tubes offer spot sizes down to a few micrometers (mm) but can only provide very limited power values. Typical focal spot sizes are 5–50 mm, and the typical power values are of the same order of magnitude with 5–50 Watts (W). Nowadays, X-ray imaging is done almost exclusively using digital image capture; respective detectors with small pixel sizes and large matrices record the X-ray intensity and convert

F. Kiessling and B.J. Pichler (eds.), Small Animal Imaging, DOI: 10.1007/978-3-642-12945-2_10, © Springer-Verlag Berlin Heidelberg 2011

125

126

the detected intensity directly into an electric signal. In clinical radiology, pixel sizes of typically 150–300 mm are in use; for SAI, pixel sizes of 50–100 mm are preferred due to the relatively small size of mice and rats. It is technologically feasible to provide the necessary detectors of typically 5 by 5 cm2 covering the entire animal in one projection. Typical pixel sizes of about 50 mm and matrix sizes of 1,000 by 1,000, i.e., a “1k × 1k” matrix, result in an active sensor area of 5 by 5 cm2. Respective detector technology has become available during the past decade. Further developments presently aim at higher frame rates, i.e., to capture more images per second; further reduction of the pixel size is not a topic due to technical and cost reasons. Whenever significantly higher resolution is desired in spite of this, the necessary smaller effective pixel sizes can be provided by magnification techniques. In practice, this means that the object is placed close to the source and the smallest focal spot size is selected. In this way, resolution values of a few micrometers can be reached although pixel sizes are 50 mm or more. The magnification approach can be used in equal fashion for 2D and 3D high-resolution imaging. In general, the X-ray components for 2D projection and for 3D CT imaging are very similar in most respects and often identical. The purpose of this chapter is to give an overview on technical and methodological aspects and developments of X-ray imaging for SAI and to provide general recommendations regarding the responsible use of X-rays. Readers interested in more details, in particular with respect to the physics and technology involved, are referred to standard textbooks on imaging, e.g., (Kalender 2005) as an introduction and overview on CT. More materials on micro-computed tomography (micro-CT, mCT) imaging can, for example, be found in (Holdsworth and Thornton 2002; Paulus et al. 2000; Ritman 2004; Badea et al. 2006; Bartling et al. 2007; Del Guerra et al. 2008; Engelke et al. 2008).

10.1 2D X-Ray Projection Imaging

W.A. Kalender et al.

overlying structures, which overshadow organs or structures of interest. But they do provide high resolution and low-noise images at acceptably low-radiation dose. Also, imaging is performed very fast since a single X-ray pulse is sufficient. Figure 10.1 shows examples of SAI projection images. Unenhanced 2D radiographs are typically used for imaging the skeleton since the soft tissue contrast is very limited.

10.1.2 Angiography To enhance contrast of vessels and of strongly vascularized areas, X-ray contrast media are applied by intravenous or intraarterial injection. In most cases, the iodinated X-ray contrast media used in clinical radiology are applied. However, alternative X-ray tracers are of great topical interest; the results of respective development efforts are primarily tested and evaluated in small animal studies (see Chap. 11). The main area of application is direct angiography, the examination of vascular structures (Fig. 10.1b); subtracting pre- and postinjection images after contrast medium administration yields the typical digital subtraction angiography (DSA) images, which are established in clinical radiology (Almajdub et al. 2008; Badea et al. 2008). Vascular studies may also reveal functional information to some degree when series of images are taken to assess contrast medium kinetics indicating perfusion or other tissue or lesion characteristics. Although simple in concept, angiographic studies are challenging when applied to SAI. The application is not trivial due to the small size of the animal. The rapid contrast flow in small animals, due to their high heart rates (e.g., up to 600 beats per minute for mice), is significantly higher than in humans and demands fast imaging capabilities, in particular high detector frame rates. This in turn requires high X-ray photon flux and the respective high X-ray power in order to acquire the necessary data in acceptably short time. High power, however, is in conflict with the desired small focus size, so the two have to be matched.

10.1.1 Radiography In many anatomical studies, valuable information can be obtained from simple 2D projection images as known from classical radiography. They suffer in quality with respect to contrast and general visibility from

10.1.3 Bone Densitometry In addition to imaging skeletal structures by standard radiography, there often is an interest to determine

127

10 X-Ray and X-Ray-CT

a

b

c

Fig. 10.1 2D X-ray projection imaging. Examples of (a) radiography, (b) angiography, and (c) bone densitometry of a rat

bone mineral density (BMD) values quantitatively. For this purpose, the same techniques are employed as in clinical medicine; for 2D projection measurements, dual-energy X-ray absorptiometry (DXA) (Kastl et al. 2002; Libouban et al. 2002; Soon et al. 2006; Nazarian et al. 2009) is most commonly used. During the ex amination, two measurements with X-ray beams of different energies, i.e., different high voltage settings and filtration, are obtained. In most cases, this is done by scanning the complete body or excised bones with a fan beam instead of a cone beam. Since soft tissues and bones exhibit different absorption behavior at the two energy levels, they can be differentiated from each other (more details on this technique are offered in Sect. 10.4). With proper calibration and subtraction of soft tissue, skeletal structures can be isolated and BMD can be determined. Standard clinical bone densitometers with special software are commonly used today for BMD and body composition measurements of small animals. The investigation of mice is not recommended because partial volume artifacts caused by the limited spatial

resolution of these scanners result in high accuracy errors. An example is shown in Fig. 10.1c. Some devices allow both in vivo and ex vivo measurements and whole-body or selected region-of-interest examinations within a few minutes.

10.2 3D X-Ray CT Imaging Principles (Micro-CT) Micro-CT, (µCT), has been promoted for various applications for about three decades by now. The focus was originally on very high-resolution imaging. Technical applications were most important initially and are often referred to as “nondestructive testing” (NDT); they also bear relevance for research in medicine as indicated by the image of a stent in Fig. 10.2d below. Medical mCT applications initially focused on preclinical research such as the in vitro evaluation of biopsies or small specimens. In vivo imaging of small animals became feasible in the last decade due to the advent of

128

W.A. Kalender et al.

new detector technology and is growing in importance. Small animals can, in principle, be examined on typical in vitro scanners with the object or animal rotating. Vice versa, specimens can be examined on in vivo scanners. It appears both logical and instructive nevertheless to clearly distinguish these two scanner types and to describe them separately as follows below. Flat-detector CT scanners recently introduced in clinical medicine, which offer larger fields of measurement, are sometimes also used for SAI applications (Bartling et al. 2007). However, they typically offer only spatial resolution of about 200–300 mm, which is only marginally superior as compared to clinical CT. Their use is likely to diminish since dedicated SAI mCT scanners have become more widely available. Therefore, they are not considered further in the

discussions below. We will here use the arbitrary, but accepted definition that the term mCT is applicable only if spatial resolution of 100 mm or better can be achieved (Kalender 2005; Bartling et al. 2007).

10.2.1 In Vitro Micro-CT Imaging The aim of most in vitro applications is to investigate high-contrast object structures at very high spatial resolution, typically at 5–50 mm. For this purpose, X-ray sources with very small focus sizes of also 5–50 mm are used. A typical scanner design is shown in Fig. 10.2a. The magnification can be easily varied by translating the sample between X-ray source and 0

a

20

60

40

80

100

1.0

6mm resolution @ 10% MTF

0.8 MTF

b 1.0

0.6

0.8 0.6

0.4

0.4

0.2

0.2 0.0

0.0 0

20

40

60

80

100

LP/mm

c

Fig. 10.2 mCT imaging in vitro. (a) Typical experimental scanner setup for very high resolution. (b) Measured modulation transfer function demonstrating resolution down to 6 mm. (c) 3D

d

display of a bone biopsy sample. (d) 3D display of a bifurcating vascular stent imaged for controlling production quality

129

10 X-Ray and X-Ray-CT

detector. The sample to be scanned is rotated by 360° during the scan. Depending on the resolution requirements, scan times of typically 10–300 min may result. Even higher spatial resolutions of below 1 mm can be achieved by the use of synchrotron radiation (SR) sources, which offer high-intensity monochromatic radiation in parallel-ray geometry. As a consequence, the spatial resolution is determined by the X-ray detector, and therefore, SR mCT set-ups are typically equipped with detectors specifically developed for these applications. They are not commercially available. SR mCT is the tool of choice for specific research questions but is not of practical importance for typical SAI. Since micro-focus X-ray tubes are limited by very low power levels, long scan times result. This is inconvenient but not very problematic for biopsy and specimen work although, for example, soft tissue samples may suffer deformations during long scan times, in particular if not embedded properly. Of course, for biopsy and specimen imaging, the radiation dose received by the sample is not a matter of concern. One typical application of mCT in preclinical research is the 3D quantification of trabecular bone structure of rats and mice, which nowadays is an integrated part of the development of bone-protective agents used against osteoporosis. Human bone biopsies (e.g., Fig. 10.2c) are also routinely investigated with mCT. With the exception of metal objects with high-Z materials, these scanners are also used for the inspection of technical objects such as the vessel stent (Fig. 10.2d). Such examinations serve to control and to optimize the manufacturing process of the respective materials and for quality control.

10.2.2 In Vivo Micro-CT Imaging In vivo imaging of small animals poses several demands, which are different from the in vitro imaging task. They are very similar in most respects to those relevant for clinical imaging: speed is of high importance, and dose has to be kept “as low as reasonably achievable,” i.e., the ALARA principle is valid for SAI in the same way as it is for clinical imaging. This is not common practice and general knowledge in the field yet. Anecdotal reports, mostly “private communication,” stated that animals suffered acute radiation damage; precise dose values for the CT protocols in use

Table 10.1 Some typical features characterizing and distingui shing in vitro and in vivo mCT In vitro In vivo scanning scanning Focal spot size (mm)

1–50

50–200

X-ray power (W)

1–50

10–300

Spatial resolution (mm)

5–100

50–200

Scan times (min)

10–300

0.1–30

Field of measurement (mm)

1–100

30–100

Dose

Not important

ALARA

were generally not available, however. This situation has to be improved. Therefore, the aspects of quality control and dose assessment will be stressed below. Some features that distinguish in vivo from in vitro imaging with mCT are summarized in Table 10.1. Again, X-ray sources with a focus size smaller than used in clinical imaging are demanded; however, typically 50–200 mm are accepted to allow for higher power levels. It is understood that area detectors – as opposed to a single linear array – are the component of choice to use a full cone beam and to thereby make efficient use of the available X-ray power. Modern SAI scanners allow for scan times of a few seconds up to a few minutes whereas scan times of typically 20–30 min were the standard for many years. A practical demand in SAI is that the animal remains stationary and that it can be supplied with anesthetizing gas, electrodes etc. during the course of the examination. Therefore, it is mandatory that the gantry rotates and the animal remains stationary at rest on the bed. Scanners vary in size significantly; desktop scanners can be accommodated relatively easily in an animal laboratory environment. In addition to the mCT scanner itself, technical facilities to monitor, to prepare, and to support the animal are necessary; this may include cardiac and respiratory monitors, warming pads, and anesthetizing gas. A respective scanner and a scan preparation scene are shown in Fig. 10.3. Some typical imaging results are shown in Fig. 10.3c–e, all derived from a single scan after contrast medium injection. Image planes can be placed arbitrarily in the scanned volume for all CT examinations; here, a sagittal (Fig. 10.3c) and a coronal view (Fig. 10.3d) were chosen to give an overview of the complete anatomy. Using appropriate software for postprocessing, which is available for most scanners

130 Fig. 10.3 Micro-CT in vivo imaging. (a, b) Typical desktop mCT scanner for small animal imaging (courtesy of Kießling, Aachen). (c–e) Scan of a mouse after contrast medium injection showing sagittal and coronal views (c, d) and rendering the kidney vasculature in very fine detail (e) (courtesy of Hauff, Berlin)

W.A. Kalender et al.

a

c

e

b

d

131

10 X-Ray and X-Ray-CT

today, details can be rendered in a quasi isolated fashion (e.g., Fig. 10.3e).

be ensured. To a certain degree, SAI has to adhere to the same principles that are valid for clinical imaging.

10.3.1 Image Quality

10.3 Quality Control and Dose Considerations It is understood that animal experiments have to be justified ethically in every respect. All precautions have to be taken that the animal is not submitted to an undue burden. The consequence of these considerations is that control of image quality and dose has to Table 10.2 Important image quality parameters and typical phantoms for their assessment Parameter Phantom CT values

Water or water-equivalent plastic

Noise

Water or water-equivalent plastic

Homogeneity

Water or water-equivalent plastic

Spatial resolution

Wire, typical 10 mm diameter tungsten in air

Spatial resolution

Bar pattern

Contrast resolution

Low-contrast patterns

Calibration

Materials at defined concentration, e.g., Ca or I

Dose

Cylinder with bore hole

4-1

4-2

4-3

4-6

4-7

4-8

Fig. 10.4 Quality assurance phantoms (see list in Table 10.2) shown by photos (top row) and by CT images (bottom row). From left to right: water phantom, wire resolution phantom, bar

Image quality and dose are measured during acceptance testing when a new scanner is installed and should be tested at regular intervals thereafter or whenever problems arise. The phantoms necessary for measurements are mostly cylinders. With typical diameters of 20 and 32 mm, they mimic well the attenuation presented by mice and lean rats. They also correspond to the 20 and 32 cm diameters of phantoms used for clinical CT mimicking the typical attenuation for head and body scans. A list of the most important image quality parameters to be checked is given in Table 10.2. The necessary phantom types are listed there as well; examples are shown in Fig. 10.4. The accuracy and stability of CT values are important to allow for follow-up studies and for comparability to other studies. Water cylinders are commonly used to check these two parameters. They also allow determining image noise and image homogeneity at the same time; this is important to monitor scanner performance over time and to check dose efficiency. Spatial resolution can be assessed in two ways: by objective quantitative measurement such as a point response function and modulation transfer function calculation or by visual inspection of regular patterns. 4-4

4-9

4-5

4-9

resolution phantom, low-contrast resolution phantom, and hydroxyapatite calibration phantom (courtesy of QRM GmbH, Möhrendorf, Germany)

132

W.A. Kalender et al.

Axial

Sagittal

Coronal

a

b

c

d

e

f

Fig. 10.5 3D dose distributions can be calculated accurately using Monte Carlo methods. Transverse (a, b), sagittal (c, d), and coronal views (e, f) of the CT and the dose values are shown,

respectively. Dose to the mouse is relatively homogeneous and was kept well below 50 mSv in this examination

The first is necessary for acceptance testing and reliable longitudinal control; the latter is more intuitive and allows a direct comparison to the imaging task at hand such as an assessment of trabecular bone structures. The capability of discerning low-contrast structures such as typical soft tissue details, often referred to as low-contrast resolution, can be tested with respective patterns in a low-contrast phantom. If reliable quantitative measurements of bone density or other tissue parameters are desired, calibration phantoms containing samples of the respective material at defined concentrations are necessary. For bone density measurements, for example, phantoms with different concentration of hydroxyapatite are in use.

10 cm long, averages dose over its complete length and does not give a direct measure of the dose to the mouse or to single organs. It is possible to estimate these values by using the concept of the CT Dose Index (CTDI), which is commonly used in clinical CT (Kalender 2005). The alternative is using small thermoluminescent dosimeters (TLD), which provide direct point measurements and can even be used in cavities of the animal or a specimen. TLD measurements are a very cumbersome and error-prone procedure, however. Calculations provide a practical and more elegant alternative. The calculation of 3D dose distributions is possible for mCT in a manner analogous to methods used in general radiology and in radiation therapy treatment planning. For this purpose, the so-called Monte Carlo methods are used: based on the CT image set that reflects the X-ray attenuation of the animal or object in question, the path of thousands of photons through the animal is simulated for each single projection, and the deposition of energy due to each photon interaction with tissue is recorded and assigned exactly to the volume element where this occurred. The method can take arbitrary scanner geometries and scan protocols into account, which makes it very flexible; results are very reliable, can be achieved more easily than by measurements, and can be carried out retrospectively for any image data set (Deak et al. 2008). An example

10.3.2 Dose Assessment The X-ray dose to the animal under examination depends on many variables. It is hard to determine experimentally with high accuracy in small animals, but the order of magnitude should be known. Measurement concepts similar to those used in clinical CT involve a simple ionization chamber measurement in a phantom of a diameter similar to the animal in question. The chamber, typically

133

10 X-Ray and X-Ray-CT

is shown in Fig. 10.5. Here the 3D dose distribution was calculated for a whole-body scan of a mouse carried out with three 360° rotations at 50 kV and a total collimation of 40 mm. Dose values ranged between 8 and 32 mGy in this case as indicated by the color scale for soft and bone tissue, respectively. The dose distributions resemble the original CT images as structures, which absorb more energy, show up as high-contrast structures in the CT images and in the dose images in a similar manner. For a 360° scan, the dose distribution is rather homogeneous over the complete volume; for projection radiography, dose is always very high at the entrance and low at the exit point.

While this may appear trivial, it is not trivial to decide on the level of image quality needed and the implication of this decision on dose. The two most important image quality features, spatial resolution and image noise, are linked to dose by the following two basic relations: 1. Noise variance changes inversely proportional to the dose D. σ2 µ

2. Noise variance changes inversely proportio nal to the 4th power of the spatial resolution element d. σ2 µ

10.3.3 Dose Optimization Optimization of dose simply means that a given task should be achieved at the minimum dose necessary, i.e., the ALARA principle has to be applied. It does not mean that dose reduction is the sole purpose, definitely not to levels where image quality, in particular the visibility of soft tissue structures, is impaired. In addition to the general optimization principle which aims at avoiding or limiting damage to the animal, there are also practical considerations to be taken into account. The need and demand for repeated studies on one and the same animal over an extended time span, e.g., to measure tumor volume growth rates, make it a necessity to limit dose and to ensure that no radiation effects interfere with the study objectives. It is not generally known which dose levels are critical and which are not. The “low-dose range” for humans is mostly specified up to 200 mSv and is defined as the range where no detrimental effects associated with radiation have been observed (UNSCEAR 2000; Kalender 2005). It appears reasonable to try and limit dose in SAI to this low-dose range whenever possible. In order to be able to do so, dose values associated with all scan protocols available to the user have to be known. The respective data should be specified by the manufacturer; they are directly related to the tube voltage, the tube current, and the scan time. This allows for a first orientation. Whenever parameters are changed, e.g., by doubling the scan time at otherwise unchanged parameters, the dose will also double.

1 . D

1 . d4

The basic message of formula 1 is that low image noise, which is desirable in many cases, is associated with high dose and vice versa. Dose optimization here means to accept that noise level where diagnosis is still possible, but not to aim for “brilliant, nearly noise-free” images as this would entail unnecessarily high dose. In any case, the noise level will only change slowly with changes in dose; an increase in dose by a factor of 4, for example, will result in a reduction of noise by a factor of 2 only. The effect is demonstrated in Fig. 10.6a–c using a low-contrast phantom comparing low, medium, and high dose acquisition. The basic message of formula 2 is that the noise levels can be influenced much more efficiently by the choice of the resolution level in the images. This depends both on the acquisition and on the image reconstruction parameters. The user can influence it most easily by choosing “smooth,” “standard,” or “sharp” reconstructions after the scan; the choice will depend on the task. When imaging high-contrast structures such as bone or contrast-filled vessels, the signal is high and high resolution is desired. A sharp reconstruction is indicated, and the high noise can be tolerated since the contrast is high. When imaging low-contrast structures such as abdominal organs or other soft issues, the signal is low. To provide an adequate contrast-to-noise ratio for diagnosis, noise has to be kept low. Smoothed images with reduced spatial resolution are appropriate in such cases and preferred to increasing dose. The effect is demonstrated in Fig. 10.6d–f using the data of a single scan

134

W.A. Kalender et al.

a

b

c

d

e

f

Fig. 10.6 Noise is a function of the dose and of the level of spatial resolution. (a–c) Noise is reduced when increasing dose as demonstrated by scanning a low-contrast phantom with 18,

55, and 185 mAs at 40 kV. (d–f) Noise is decreased more efficiently when reconstructing images with softer kernels, however, at the expense of reduced spatial resolution

reconstructed with sharp, standard, and smooth re construction. There are further possibilities to limit or to reduce dose which, again, are similar to the general radiation protection principles applied in clinical radiology. The most basic is the recommendation to employ X-rays only when there is a clear indication. Repetition of examinations should be limited to the necessary minimum. The examination range should be limited to the field of interest; this simply means, for example, that whole-body examinations should be avoided when only a single organ is of interest. A more difficult topic is the choice of the “optimal” energy as it will depend on the equipment at hand and also on the goal, i.e., the question if soft tissue (density differences), skeletal structures (calcium), or vessels (iodine or other contrast materials) are to be imaged (Kalender et al. 2009). Tube voltages from 40 to 80 kV are indicated as opposed to the tube voltages of 80–120 kV used in clinical radiology.

10.4 Advanced Micro-CT Techniques There are quite a number of CT applications that go beyond the routine tasks of 3D imaging of the anatomy with or without contrast media. The range of “advanced” applications is wide and not well defined. We here focus on three topical examples but hope that many more applications will evolve and gain in importance.

10.4.1 Dynamic Micro-CT Static imaging with mCT does not always deliver sufficient diagnostic results. Dynamic imaging may be conducted for singular, for example, contrast agent first pass studies, or repeated periodic and nonperiodic events. In the latter case, acquisition can be synchronized with physiological signals, for example, breathing or ECG signals. Tomographic images can then be

135

10 X-Ray and X-Ray-CT

reconstructed for selected phases of the physiological cycle. In any case, temporal resolution of the scanner needs to be high enough to differentiate between phases of a given function. Although the principle may sound simple, it is a challenge in SAI due to the fact that animal physiology differs from that of humans. Table 10.3 summarizes some of the basic dynamic properties, the breathing, and heart rates for humans, rats, and mice. Small animals breathe significantly faster than humans; their blood circulation and typical heart rates are also higher. Both facts mean that in order to acquire “still” images of an animal without motion artifacts, the acquisition system must be able to record views (projections) in time intervals of only a few milliseconds. This is not possible in mCT at present.

Table 10.3 Dynamic parameters for humans and small animals (see Chap. 37) Human Rat Mouse Breathing rate (min−1)

12–20

80–150

80–150

Heart rate (min−1)

60–80

300–500

300–800

Circulation time (s)

6

1

1

However, remarkable capabilities for fast and repeated scanning have become available recently. The area detectors, which almost all modern mCT scanners are equipped with, have relatively low frame rates. A high-end scanner equipped with an acquisition system delivering 25 frames (projections) per second can resolve at best 40 ms time intervals of a dynamic cycle. This might be just sufficient to conduct breath synchronization but is not sufficient for cardiac imaging. Nevertheless, dynamic imaging of contrast medium passage, i.e., the recording of concentrationtime curves, is well possible. The assessment of tissue or organ perfusion is one of the goals. Typically, temporal resolution of a few seconds would be necessary to conduct such examinations based on the observation of the first pass of a contrast agent in the vascular system. Using a model test setup, it can be shown that temporal resolution is sufficient (Fig. 10.7). In this case, the perfusion model (Fig. 10.7a) was scanned with a dual-source mCT of the type shown in Fig. 10.3a with 4 s rotation time and 1.4 s effective scan time per image. Apparently, it can match the temporal resolution of fast clinical CT as shown by the arterial concentration curves (Fig. 10.7b).

a

b

1900 Artery Tissue Perfusion (max HU) (max HU) (ml/min/100 ml)

1500

Clinical CT

Fig. 10.7 Dynamic micro-CT using fast dual-source mCT. A perfusion model (a) was used to show that mCT can provide information on contrast medium kinetics similar to clinical CT (b)

CT-value / HU

Micro-CT

1807

428

229

1650

346

220,5

1100

Artery - Clinical CT Artery - Micro-CT Tissue-Clinical CT

700

Tissue-Micro-CT

300

−100

6

10

14

18

22 time / s

26

30

34

136

W.A. Kalender et al.

10.4.2 Dual-Energy CT In the concept of dual-energy CT (DECT) imaging, the object under investigation is scanned with two different energy spectra, typically by using two different voltage settings and filtration. DECT exploits the dependence of the X-ray attenuation properties of matter on energy; different materials, for example, water (soft tissue) and calcium (bones), exhibit different relationship between X-ray energy and attenuation (Fig. 10.8a). Based on this, images resulting from the two scans can be combined in order to extract additional information about the object composition (Alvarez and Macovski 1976; Kalender et al. 1986). The two scans can either be conducted simultaneously if the scanner is equipped with two tubedetector systems or sequentially one after another with the tube voltage altered between scans.

Mass attenuation coefficent, cm2 /g

a

10.4.3 Image Postprocessing and Rendering The task of “reading” the images and to document the findings plays an important role and should not be underestimated with respect to the necessary expertise and time required. Quantitative evaluation often

100 I Ca H2o

10

-1

0,1 0

b

DECT is already established in clinical radiology (Flohr et al. 2006; Johnson et al. 2007; Schlomka et al. 2008); mCT applications are also under development. Both improved visualization and differentiation of different tissues, such as bone and materials, (Fig. 10.8b–d), and possibilities for selective quantification, such as bone mineral independent of the soft tissue and fat marrow components, are of high interest.

40

80 120 Energy, keV

160

200

c

Fig. 10.8 Dual-energy mCT. The differences in the energy dependence of different materials (a) are exploited by scanning the animal at two different energies. This allows viewing high

d

atomic number materials such as calcium and iodine only (b) and even to separate these into calcium (c) and iodine images (d)

137

10 X-Ray and X-Ray-CT

necessitates automated evaluation procedures to ensure objective and reproducible results. For example, the quantification of BMD requires the transformation of the measured CT values into BMD values. This can be solved using a calibration phantom simultaneously scanned with the sample or by a calibration performed before or after the sample scan. In addition, a segmentation step is required to select the desired analysis volume of interest, for example, the vertebrae or the knee. In order to minimize the effects of noise and partial volume artifacts, 3D segmentation algorithms with local adaptive thresholds should be preferred to global threshold-based segmentation approaches that are often applied slice by slice. Figure 10.9 shows an example of advanced 3D segmentation in which the same algorithm was used for the thin trabeculae in the metaphysis and the thicker ones in the epiphysis. Visualization in appropriate or, simply, in an adequate and convincing manner is often worth the effort. Different visualization or rendering techniques such as 3D surface-shaded displays, maximum intensity projections, volume rendering, and virtual endoscopy are routinely used in clinical radiology and should be considered for SAI using mCT. A short introduction to such techniques can be found, for example, in (Kalender 2005). Figure 10.10 shows screenshots of a respective example; a movie showing various perspectives of the study and a virtual flight through the spinal canal (Fig. 10.10d) can be accessed on the institute’s homepage (www.imp.unierlangen.de/Medical Imaging).

10.5 Conclusions and Recommendations SAI using X-rays bears many similarities to imaging humans with respect to the technology and scan approaches, but also with respect to the general philosophy. Scanning shall and must not mean an unnecessary burden for the animals. Dose values are known! To try and limit levels of organ dose values to below 100 mGy per examination appears to be a possible consensus (Ford et al. 2003; Boone et al. 2004; Kalender 2005). Unless there is a good reason, we should not expect or demand spatial resolution much below 100 mm for SAI in view of scan times, noise, and dose. Accordingly, the most recent scanner designs are optimized to provide short scan times of about one or only a few minutes at resolution levels just below 100 mm with exposure levels in the range of 10–50 mGy or even below. Depending on the desired low-contrast resolution and the resulting signal-to-noise levels different settings may be required; higher dose levels have to be accepted in that case. Advanced techniques, similar to the special applications available in clinical CT, are partly available today and will become more elaborate. Future developments will aim at higher speed, higher dose efficiency, and, possibly most important, supporting multimodality imaging. Combination scanning with CT and positron emission tomography or CT and fluorescence imaging are primary examples for this trend.

Fig. 10.9 Example for a fully automated analysis of trabecular bone structures of a rat femur scanned by mCT

138

W.A. Kalender et al.

Fig. 10.10 3D data sets can be viewed in arbitrary perspectives and display schemes. (a, b) Overviews in volume rendering technique. (c) Shaded surface display of the spine. (d) Virtual flight through the spinal canal. Please also see www.imp. uni-erlangen.de. (processing of images a courtesy of Fovia Inc., Palo Alto, USA)

References Almajdub M, Magnier L, Juillard L, Janier M (2008) Kidney volume quantification using contrast-enhanced in vivo X-ray micro-CT in mice. Contrast Media Mol Imaging 3(3):120–126 Alvarez RE, Macovski A (1976) Energy-selective reconstructions in x-ray computerized tomography. Phys Med Biol 21(5):733–744 Badea CT, Hedlund LW, Lin MD, Boslego Mackel JF, Johnson GA (2006) Tumor imaging in small animals with a combined micro-CT/micro-DSA system using iodinated conventional and blood pool contrast agents. Contrast Media Mol Imaging 1(4):153–164 Badea CT, Drangova M, Holdsworth DW, Johnson GA (2008) In vivo small-animal imaging using micro-CT and digital subtraction angiography. Phys Med Biol 53(19):R19–R50 Bartling SH, Stiller W, Semmler W, Kiessling F (2007) Small animal computed tomography imaging. Curr Med Imaging Rev 3:45–59 Boone J, Velazquez O, Cherry SR (2004) Small-animal X-ray dose from micro-CT. Mol Imaging 3(3):149–158 Deak P, van Straten M, Shrimpton PC, Zankl M, Kalender WA (2008) Validation of a Monte Carlo tool for patient-specific dose simulations in multi-slice computed tomography. Eur Radiol 18:759–772 Del Guerra A, Belcari N, Llacer GL, Marcatili S, Moehrs S, Panetta D (2008) Advanced radiation measurements techniques in diagnostic radiology and molecular imaging. Radiat Prot Dosimetry 131(1):136–142 Engelke K, Prevrhal S, Genant HK (2008) Macro and micro imaging of bone architecture. In: Bilezikian J, Raisz L, Martin TJ (eds) Principles of bone biology, vol II. Academic, San Diego

Flohr T, McCollough CH, Bruder H et al (2006) First performance evaluation of a dual-source CT (DSCT) system. Eur Radiol 16:256–268 Ford NL, Thornton MM, Holdsworth DW (2003) Fundamental image quality limits for microcomputed tomography in small animals. Med Phys 30(11):2869–2877 Holdsworth DW, Thornton MM (2002) Micro-CT in small animal and specimen imaging. Trends Biotechnol 20(8):S34–S39 Johnson TRC, Krauß B, Sedlmair M et al (2007) Material differentiaiton by dual energy CT: initial experience. Eur Radiol 17:1510–1517 Kalender WA (2005) Computed tomography. fundamentals, system technology, image quality, applications, 2nd edn. Publicis, Erlangen Kalender WA, Perman WH, Vetter JR, Klotz E (1986) Evaluation of a prototype dual-energy computed tomographic apparatus. I. Phantom studies. Med Phys 13(3):334–339 Kalender WA, Deak P, Kellermeier M, Van Straten M, Vollmar S (2009) Application- and patient size-dependent optimization of x-ray spectra for CT. Med Phys 36(3):993–1007 Kastl S, Sommer T, Klein W, Hohenberger W, Engelke K (2002) Accuracy and precision of bone mineral density and bone mineral content in excised rat humeri using fan beam dualenergy X-ray absorptiometry. Bone 30(1):243–246 Libouban H, Simon Y, Silve C et al (2002) Comparison of pencil-, fan-, and cone-beam dual X-ray absorptiometers for evaluation of bone mineral content in excised rat bone. J Clin Densitom 5(4):355–361 Nazarian A, Cory E, Muller R, Snyder BD (2009) Shortcomings of DXA to assess changes in bone tissue density and microstructure induced by metabolic bone diseases in rat models. Osteoporos Int 20(1):123–132 Paulus MJ, Gleason SS, Kennel SJ, Hunsicker PR, Johnson DK (2000) High resolution X-ray computed tomography: an

10 X-Ray and X-Ray-CT emerging tool for small animal cancer research. Neoplasia 2(1–2):62–70 Ritman EL (2004) Micro-computed tomography – current status and development. Annu Rev Biomed Eng 6:185–208 Schlomka JP, Roessl E, Dorscheid R et al (2008) Experimental feasibility of multi-energy photon-counting K-edge imaging in preclinical computed tomography. Phys Med Biol 53(15):4031–4047

139 Soon G, Quintin A, Scalfo F et al (2006) PIXImus bone densitometer and associated technical measurement issues of skeletal growth in the young rat. Calcif Tissue Int 78(3): 186–192 UNSCEAR (2000) Annex D: medical radiation exposures. United Nations Publications, New York

CT Contrast Agents

11

Hubertus Pietsch

Contents 11.1 Introduction........................................................... 141 11.2 X-Ray Contrast Media for Small Animal Imaging.................................................... 142 11.2.1 Iodinated Extracellular CM.................................... 143 11.2.2 Vascular Space CM: Blood Pool Imaging Agents....................................................... 143 11.2.3 Hepatocellular CM.................................................. 144 11.3 Pharmacokinetic Properties of CM for X-Ray or mCT........................................................ 145 11.3.1 Iodinated Extracellular CM.................................... 145 11.3.2 Vascular Space CM: Blood Pool Imaging Agents....................................................... 145 11.3.3 Hepatocellular CM.................................................. 148 11.4 Indications for CM Used for X-Ray and mCT Imaging.................................................. 148 References............................................................................ 149

H. Pietsch TRG Diagnostic Imaging, Global Drug Discovery, Bayer Schering Pharma AG, Müllerstrasse 178, 13353 Berlin, Germany e-mail: [email protected]

11.1 Introduction The contrast in X-ray images is typically produced by attenuation of the X-rays in the examined tissues or samples. Without contrast, only skeletal structures of sacrificed or live animals can be visualized in detail. The transit and accumulation of contrast media (CM) in different organs (enhancement) improve the differentiation of morphological structures, particularly between normal and pathological tissue. Contrast enhancement for computed tomography (CT) imaging is primarily achieved in practice by the use of water-soluble contrast agents that distribute exclusively in the extracellular fluid (ECF) space. A major problem with these agents, however, is their short duration of contrast enhancement, which has been partially solved by the use of a bolus or continuous bolus infusions in conjunction with rapidsequence scanning techniques and significant efforts to create agents for opacification of specific organs, such as the liver and spleen. The CM used for X-ray or micro-CT (mCT) imaging differ in their patterns of biodistribution after intravenous injection. The most suitable classification system for CM is to distinguish between extracellular fluid CM (ECF-CM), vascular space CM (blood pool enhancement) and hepatocellular or tissue-specific CM (Table 11.1). Furthermore, with oral administration of contrast agents such as iodine or barium, the gastrointestinal tract can be visualized (gastrointestinal CM).

F. Kiessling and B.J. Pichler (eds.), Small Animal Imaging, DOI: 10.1007/978-3-642-12945-2_11, © Springer-Verlag Berlin Heidelberg 2011

141

142

H. Pietsch

Table 11.1 Classification of X-ray contrast media for small animal CT imaging Iodinated extracellular fluid CM (urographic water-soluble CM)

Vascular space CM (blood pool agent)

Hepatocellular CM (tissue-specific CM)

Gastrointestinal CM

Monomeric nonionic CM

Macromolecular CM

Iodine-containing lipid

Iodinated CM

Iopromide

Dysprosium-DTPA-dextran CM

Iohexol Iopamidol

Oil emulsions of ethiodol and iodine

Iopromide liposomes

Diatrizoate (ionic CM)

Iodixanol liposomes

Iopromide-370 (nonionic CM)

Cholesteryl ioponate

Ioversol

Polyiodinated triglyceride emulsion (ITG-LE)

Barium sulfate

Fenestra DC Iopentol Iomeprol

Iodine-containing micelles MPEG-iodolysine micelles Dysprosium EOB DTPA

Iobitridol Ioxilan

Nanoparticles Bismuth-sulfide polymer coating (BPNP)

Dimeric nonionic CM

Gold nanoparticles

Iotrolan Iodixanol

Liposomal encapsulation of iodinated CM Polyethylene glycol-coated (PEG) iopromide liposomes PEG iohexol liposomes Lipid mixture with Iodixanol Polyiodinated triglyceride emulsion (ITG-LE) Fenestra VC

11.2 X-Ray Contrast Media for Small Animal Imaging 11.2.1 Iodinated Extracellular CM Iodinated extracellular CM are either negatively charged ionic molecules (e.g., diatrizoate) or nonionic molecules (e.g., iopromide, iomeprol, and iodixanol). In this review, we will only discuss nonionic iodinated contrast agents (low-osmolar CM), which were introduced more than 20 years ago and have become standard contrast agents in clinical CT. Nonionic CM can be divided into two classes with regard to their chemical structures: nonionic

monomers and nonionic dimers. All of these agents are based on a benzene ring with three iodine atoms attached (2,4,6-triiodobenzoic acid). The monomers contain one tri-iodinated benzene ring, while the dimers contain two tri-iodinated benzene rings. The benzene ring of nonionic CM lacks an ionizable carboxyl group and is made soluble by including several hydrophilic groups as side chains. Furthermore, the carboxyl group is reacted with an amine to make the molecules electrically neutral in solution. Both monomeric and dimeric nonionic compounds are low-osmolar CM. The nonionic monomers have an osmolality that is almost twice that of plasma, and the ratio between the number of iodine atoms and the number of particles in the solution is 3:1. The nonionic dimers are iso-osmolar

143

11 CT Contrast Agents

with plasma (300 mOsm/kg) and are present in a 6:1 ratio.

Nonionic CM differ from one another with regard to important physicochemical properties, including water solubility, viscosity, osmolality, and electrical charge. The water solubility of these contrast agents must be very good as highly concentrated CM solutions are used; for example, up to 800 mL is a typical dose of iopamidol or iopromide administered in adults, the equivalent of approximately 240 g of iodine (Rau et al. 1997; Rosovsky et al. 1996). As with sugars or peptides, the solubility of nonionic CM is mediated by hydrolytic groups (–OH, –CONH–). Viscosity is a measure of the flow properties of solutions depending on temperature and is measured as millipascals per second. It increases with both higher CM concentrations and falling temperature. The viscosities of the various CM are different at the same iodine concentration and temperature. The viscosity of a contrast medium is of considerable practical importance. When thin-bore needles, cannulae, and catheters are used, high viscosity preparations can only be administered with considerable effort. Rapid administration of large volumes, as is necessary in mCT, and injection through thin-bore cannulae and catheters is greatly facilitated by the use of low viscosity CM such as iopromide. When prewarmed to body temperature, the viscosity of CM further decreases (Krause 1994; Hughes and Bisset 1991). The osmotic pressure of a solution is determined by the concentration of dissolved particles. Normally, osmotic pressure is calculated in the unit osmolality, which describes the concentration of solute (osmoles) per kg of water (mOsm/kg H2O). The osmolality of CM depends mainly on the concentration and only slightly on the temperature, and different CM can display greatly divergent osmolalities at the same concentration. Greater tolerability of low-osmolality agents has been demonstrated in many animal experiments and several clinical studies. The low osmolality causes less pain at the injection site and reduced heat sensation during power injection (Kim et al. 1990). Furthermore, hemodynamic and cardiovascular changes such as vasodilatation, bradycardia, and pulmonary hypertension are greatly reduced. At comparable iodine concentrations, the nonionic monomers (low-osmolar CM)

Viscosity mPa*s

11.2.1.1 Comparison of Properties Among Nonionic Contrast Media

Iomeprol 300

14

Iopromide 300

12

Ioversol 300 Iohexol 300

10

Iodixanol 320

8 6 4 0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

Osmolality mOsm/kg H20

Fig. 11.1 Osmolality and viscosity of non-ionic monomeric CA (Iomeprol, Iopromide, Ioversol, Iohexol) and non-ionic dimeric CA (Iodixanol) as formulated for injections into humans or animals

have in vitro a lower viscosity but a higher osmolality compared with nonionic dimeric CM (iso-osmolar CM) (Fig. 11.1). An increase of the iodine concentration of CM solution increases exponentially its viscosity (Jost et al. 2009).

11.2.2 Vascular Space CM: Blood Pool Imaging Agents For mCT, a concentration of intravascular CM that enables the enhancement of 50–60 Hounsfield units (HU) over a period of at least 15 min would be desirable to allow efficient blood pool imaging, which would be especially useful for imaging of cardiovascular diseases (e.g., stenosis, ischemia, atherosclerosis, capillary permeability, and neovascularity). Three points seem to be necessary for a blood pool contrast agent: 1. The size of the CM molecules must be larger than the capillary fenestrations. 2. Phagocytosis by the RES should be avoided as long as possible. 3. A radiopaque moiety must be structurally incorporated within the particle. There are two different methods used to ensure the prolonged circulation of imaging agents in an effort to make them suitable for blood (vascular) imaging. The first uses in vivo labeling of blood cells or plasma components with diagnostic labels, while the second uses synthetic macromolecules, colloidal solutions, and/or liposomes carrying diagnostic labels.

144

H. Pietsch

Macromolecules include proteins (e.g., bovine serum albumin), synthetic polypeptides (e.g., dextran), or polyamino acids (e.g., poly l-lysine). By the middle of the 1900s, however, methods for modifying liposome surfaces were introduced, resulting in liposomes that could avoid the RES for prolonged periods of time (Allen 1994; Senior 1987). The half-life of conventional liposomes in the blood stream after intravenous application can be increased as follows: 1. 2. 3. 4. 5.

A reduction in liposome size to below 100 nm An increase in cholesterol content in the capsule The use of saturated (gel-state) phospholipids The use of uncharged lipids The use of higher liposome doses

Alternatively, the surfaces of liposomes can be altered by the inclusion of glycolipids (e.g., monosialoganglioside GM) or polymers (typically polyethylene glycol (PEG)). This alteration results in a sterically stabilized liposome with a more hydrophilic surface that is less able to bind plasma opsonins, resulting in decreased RES uptake and a prolonged circulation half-life (Allen 1994). Surface-modified liposomes are able to avoid the mononuclear phagocytic system MPS, thus allowing targeting to non-MPS organs. Furthermore, the inclusion of lipid derivates of PEG in the liposome membrane has been shown to very potently increase the liposomal circulation times of iohexol, liposomes, and iopromide (Koa et al. 2003; Sachse et al. 1997). For mCT imaging, a new family of CM has been developed. The Fenestra family is comprised of iodinated lipids that provide contrast enhancement integrated into a novel oil-in-water lipid emulsion. Fenestra VC, a 20% (wt/ol) oil-in-water lipid emulsion formulation containing an iodinated triglyceride (glyceryl-2-oleoyl-1,3-di-7-(3-amino-2,4,6-tri iodophenyl)-hepanoate) (ITG-DHOG), provides prolonged vascular contrast for the visualization of cardiac, abdominal, tumor, and peripheral vasculature. It is hepatotropic and is taken up by hepatocytes via an energy-dependent and ApoE receptor-mediated pathway. The surface of the vehicle is modified by the addition of methoxypolyethylene glycol 1-2-distearyl phosphatidylethanolamine, which enables long-lasting blood pool contrast enhancement. The formulation is iso-osmolar with plasma and has a particle size of

150 nm and an iodine concentration of 50 mg/mL (Choukèr et al. 2008).

11.2.3 Hepatocellular CM A number of approaches have been investigated in an effort to identify radiopaque imaging agents that selectively accumulate in the liver and spleen. These include, among others, radiopaque liposomes, polyiodinated triglyceride, and metal complexes of EOB DTPA such as the dysprosium complex. In the 1990s, various preparation methods for liposomes carrying CT radiopaque molecules were available, which allowed iodine encapsulation with efficiencies up to 50% and corresponding encapsulated iodine-to-lipid ratios (mg/mg) between 0.4 and 4.7. Liposomes with narrow size ranges and maximum particle sizes well below 5 m can be produced with these methods. When injected intravenously, these particles are rapidly taken up by macrophages localized in the liver, spleen, and bone marrow. In various animal experiments, liver opacification with density increases at or above 50 DHU has been demonstrated, allowing the delineation of small focal lesions (Seltzer et al. 1988). An iodinated triglyceride emulsion (IT-LE) packed into the lipophilic core of a synthetic chylomicron remnant has been described by Bakan. This emulsion can be used to image the liver parenchyma as it is internalized by the hepatocytes, whereas liver tumor cells have less functional lipoprotein receptors and, therefore, show little enhancement. Fenestra LC contrast medium contains the emulsion described by Bakan and provides extended hepatobiliary contrast enhancement, which is useful for the visualization of abdominal anatomy and function. It is comprised of iodinated lipids that provide contrast enhancement and a novel oil–water lipid emulsion that serves as a hepatocyte-selective delivery system. This formulation in a synthetic oil-in-water lipid emulsion particle resembles a chylomicron remnant, which helps in selective localization of the lipids to various sites within the body. Fenestra LC, properly termed 1,3-bis-[7-(3amino-2,4,6-triiodophenyl)-heptanoyl]-2-oleoyl glycerol (DHOG), is a commercially available

145

11 CT Contrast Agents

hepatocyte-selective iodinated contrast medium for preclinical use only. Its structure and particle size of 90–180 nm are similar to endogenous chylomicron remnants (75–400 nm), allowing for fast receptormediated uptake in the liver. DHOG is supplied as an opaque solution at a concentration of 50 mg I/mL. Furthermore, it is iso-osmotic with plasma at 300– 350 mOsm/kg and has a viscosity of 3–4 cP at 37°C (Choukèr et al. 2008).

11.3 Pharmacokinetic Properties of CM for X-Ray or mCT The CM used for X-ray and CT imaging differ in their pharmacokinetic behaviors and patterns of distribution following intravenous injection.

11.3.1 Iodinated Extracellular CM High water solubility, low distribution coefficients between butanol and buffer, and low binding to plasma proteins (<5%) are all properties of nonionic ECF-CM that contribute to their pharmacokinetic behavior, enabling them to be rapidly cleared from the blood within minutes following intravenous injection. The first pass of the CM with the highest vascular contrast is only retained for approximately 3 min. Following intravenous administration, nonionic CM are rapidly distributed between the vascular and interstitial spaces with a distribution half-life of about 25 min. They are then cleared from the blood with an elimination half-life of about 20 min. in rats and excreted without tubular reabsorption or secretion, mainly via glomerular filtration. In patients with normal renal function, the extrarenal elimination of ECF-CM is minimal (<2%). These agents are not metabolized after intravenous injection (Krause 1994). The pharmacokinetic behavior of nonionic CM after intravenous injection in small animals is comparable to that seen in human patients. After intravenous injection of 125I-labeled iopromide (60 mg I/kg body weight) in rats, 82% of the dose is excreted via the kidneys within 3 h, and the

elimination half-life in the blood and urine is about 20 min. In rats, about 90% of the dose is recovered from the urine, and 10% from the feces, regardless of the dose administered (Mützel et al. 1983). Hydrophilicity prevents ECF-CM from crossing cell membranes. After oral or intraduodenal administration of 125I-labeled iopromide to rats, less than 2% of the dose is absorbed (Mützel et al. 1983).

11.3.2 Vascular Space CM: Blood Pool Imaging Agents CM that remain in the vasculature with adequate enhancement for at least 15 min are called blood pool contrast agents. A number of different approaches have been taken to produce longer-lasting blood pool contrast agents for CT. Some important examples include: • Water-soluble macromolecular agents such as dysprosium-DTPA-dextran, which have shown blood pool contrast enhancement for up to 45 min in rabbits (Vera and Mattrey 2002). • Iodinated oil emulsions of ethiodol, which remain intravascular for 20 min after injection in phantom studies of excised canine hearts and in rabbits (Cassel et al. 1982). • Iodine-substituted poly-l-lysine micelles (MPEGiodolysine), which cause noticeable enhancement in the blood, liver, and spleen for more than 3 h following intravenous injection into rats (Torchilin et al. 1999). • Nanoparticles, which can act as blood pool CT-CM (e.g., nanosized bismuth-sulfide with a polymer coating (BPNP) and gold nanoparticles). Rabin et al. (2006) described a polymer-coated Bi2S3 with excellent stability at high concentrations (0.25 M Bi3+) and high X-ray absorption at fivefold greater than iodine. After intravenous administration of 250 I, BPNPs are distributed with a half-life of 140 ± 15 min in mice. Delayed imaging at 12–24 h after injection shows that the BPNPs are distributed to organs containing phagocytic cells (e.g., the liver, spleen, and lymph nodes). For example, in one study the liver, signal intensity increased from −22 ± 77 to

146

H. Pietsch

740 ± 210 HU at 24 h (Rabin et al. 2006). This enhancement likely reflects the uptake of BPNPs into Kupffer cells and hepatocytes. Animal studies have demonstrated that gold nanoparticles (250 gold atoms per molecule) are useful as X-ray contrast agents. Gold provides about 2.7 times greater contrast per unit weight than iodine, and its greater absorption of X-rays in comparison to iodine enables good contrast at lower X-ray doses. Gold nanoparticles (1.9 nm in diameter) have been injected intravenously into mice at a concentration of 270 mg Au/cm3 and a volume of 0.01 mL/g. In this experiment, the blood gold concentration decreased in a biphasic manner with a 50% drop between 2 and 10 min, followed by a slower decrease by another 50% between 15 min and 1.4 h. The highest tissue gold concentration 15 min after injection was in the kidney (10.60 ± 0.2% of the injected dose per gram of measured tissue). Whole body gold clearance was 77.5 ± 0.4% of the total injected after 5 h, and retention in the liver and spleen was minimal with elimination via the kidneys (Fig. 11.2).

Given that CT liposomes with extended circulation times may be useful as blood pool (vascular) imaging agents, some groups have evaluated the imaging properties of various surface-modified, contrast-carrying liposomes (Sachse et al. 1997; Schmiedl et al. 1999). In a study by Sachse et al., conventional and surface-modified liposomes containing the nonionic monomeric CM iopromide were tested as potential blood pool agents. In a biodistribution study using rats, no significant differences in blood concentration were found 1 h after injection between two different PEG coating agents (DSPE-PEG2000- or CHHS-PEG2000) and unmodified conventional liposomes at a dose of 250 mg total I/kg body weight (approximately 500 mg total lipid/kg). At 4 h after injection, however, the DSPE-PEG liposomes displayed a significantly higher blood level compared to the other liposome formulations, amounting to approximately 28% of the dose (1.1 mg I/g wet weight) compared to 16% for the unmodified liposomes (Fig. 11.3).

12.0%

Percent injected dose per gram

10.0%

8.0% Kidney

6.0% er

Liv

4.0% Tumor 2.0%

Blood

Muscle 0.0% 0

5

10

15

20

24

Hours post injection

Fig. 11.2 Pharmacokinetic data for the biodistribution of gold in mice following intravenous injection of 0.2 mL gold nanoparticles (percent injected dose per gram in the blood, kidneys, liver, muscle, and tumor over a 24 h period) (Hainfeld et al. 2006)

147

11 CT Contrast Agents A

blood

Unmodified liposomes CHHS-PEG DSPE-PEG

mg I/g w.w.

1,50

1,00

0,50

0,00

15 min

1h

4h

24 h

time

Fig. 11.3 Biodistribution of unmodified and modified iopromide liposomes in the blood of rats at a dose of 250 mg total I/kg body weight. Data are the mean values from four animals (Sachse et al. 1997)

The new pool-selective emulsion (ITG-PEG) CM, including Fenestra VC, differ from the previous approaches in that their mean particle size is less than 150 nm in diameter. Furthermore, they remain in the blood for several hours after injection and are cleared within 24 h. Ford et al. and Graham et al. investigated the pharmacokinetic behavior of Fenestra VC (concentration 50 mg I/mL) in mice. After a single injected dose of 0.5 mg I/g body weight, the peak enhancement in the right ventricle and vena cava occurred at 0.25 h, providing an enhancement of 340 HU over the baseline values. This vascular enhancement persisted for 2–4 h after injection before returning to near baseline values by 24 h after injection. The half-life in blood was approximately 8 h. When the PEG moieties were eliminated from these particles, the formulation was eliminated by the hepatobiliary system. Subsequent enhancement occurred in the liver (290 HU over baseline) and spleen (410 HU over baseline) at 24 h after injection (Ford et al. 2006). Fenestra requires 48 h to be cleared; however, some researchers have observed spleen enhancement weeks after a single injection (Fig. 11.4).

500 Liver Blood

CT Numbet [HU]

400

300

200

100

0

0

5

10

15

20

25

Time [h]

Fig. 11.4 Fenestra VC is cleared from the blood into the liver parenchyma with equal enhancement of the blood and liver after injection at a dose of 0.015 mL/g body weight (0.75 mg I/g body weight) in mice (Graham et al. 2008)

148

H. Pietsch

11.3.3 Hepatocellular CM 1000 Density [HU]

In this section, we will only describe liposomes and an iodinated triglyceride emulsion as they have been most successful as hepatocyte-selective CM for CT. After intravenous injection, liposomes are mainly taken up by the phagocytic cells of the RES, which is comprised of circulating blood macrophages and localized macrophages in the liver (Kupffer cells), spleen, and bone marrow. This biodistribution pattern can be used for passive targeting of the liver and spleen as they are the main sites of liposome uptake (Poste 1983). Liposome uptake can be increased by the incorporation of negatively charged lipids (e.g., phosphatidic acid, phosphatidyl glycerol, and phosphatidyl serine), the effects of which are more pronounced for smaller liposomes (Senior 1987). The incorporation of cholesterol into the liposome membrane results in increased plasma stability, prolonged blood circulation, and decreased intracellular degradation. The organ distribution of liposomes is a function of liposome dose, which is reflected by reduced liver uptake at higher doses and seems to be due to saturation of the endocytotic uptake mechanism into Kupffer cells (Senior 1987). In one study, iopromide-carrying liposomes with a diameter of approximately 0.5 m and an encapsulation efficiency ranging from 30 to 40% showed dose-dependent pharmacokinetics in rats after an intravenous injection. The terminal half-life in blood increased from 0.8 h for a 250 mg I/kg dose to 2.9 h for a 1,000 mg I/kg dose. The elimination of iodine occurred mainly via renal clearance and was complete within 7 days. Liver enhancement was observed at a dose of 200 mg I/kg, which is equivalent to the clinically relevant value of 30 HU (Krause et al. 1993). After intravenous injection of Fenestra LC, an iodinated triglyceride emulsion, it distributes to the vasculature of the liver and passes through the endothelial fenestrations into the space of Disse. There it binds to the apoE receptor on hepatocytes and is subsequently internalized into these cells and excreted via the biliary system. In one study, a dose of 1 g I/kg body weight was injected intravenously into mice. The liver showed steadily increasing enhancement with a peak enhancement of 300% at 300 min postinjection compared to

liver aorta kidney spleen

750 500 250 0

0

200

Time [min]

400

24h

48h

Fig. 11.5 Organ densities of the liver, aorta, spleen, and kidney parenchyma in Hounsfield units measured before and at various points after intravenous injection of Fenestra LC at a dose of 1 g I/kg body weight in mice (Henning et al. 2008)

the baseline value. At 48 h post injection, the enhancement was still at 50%. The contrast agent then underwent primary elimination via biliary excretion (Fig. 11.5) (Henning et al. 2008).

11.4 Indications for CM Used for X-Ray and mCT Imaging There are numerous applications for X-ray CM in small animal CT (Table 11.2), including the characterization of structures, vessels (e.g., tumor vascularization), tumors, and soft tissue morphology and function. Nonionic, extracellular, water-soluble agents are generally applied in CT, primarily due to their relative safety and availability. A major problem with these agents, however, is their short duration of contrast enhancement. This problem has been partially solved by using bolus injections or bolus infusions of CM in conjunction with rapid-sequence scanning techniques. In addition, many agents have been generated for specific opacification of the blood pool, liver, and spleen. Blood pool imaging can aid in the detection of structural and functional abnormalities such as those caused by thrombi or atherosclerotic lesions. Furthermore, it would be of particular interest to evaluate the current state of blood flow and investigate irregularities caused by pathological changes.

149

11 CT Contrast Agents Table 11.2 Overview of the applications of contrast media in CT after intravenous administration in small animals Contrast medium Application Extracellular fluid CM Head imaging for stroke assessment (perfusion CT, CT angiography) Heart imaging Abdomen imaging (kidney, urinary tract) Liver imaging (tumor detection) Angiography (alternative to catheter angiography) Vascular space CM (blood pool CM)

Tumor imaging Early detection of tumors by angiogenesis Cardiovascular imaging Myocardial infarction Imaging of structural and functional abnormities, including thrombi and atherosclerotic lesions

Hepatocellular or tissue-specific CM

Liver imaging Detection of focal liver lesions Tumor characterization Hepatic parenchyma imaging

The combined use of mCT and hepatocellular CM allows for detailed noninvasive imaging of hepatic tumors, the hepatic parenchyma, and other liver lesions that in the past would have required euthanization of the animal or placement of the tumor in a flank location or someplace other than the liver.

References Allen TM (1994) The use of glycolipids and hydrophilic polymers in avoiding rapid uptake of liposomes by mononuclear phagocyte system. Adv Drug Delivery Rev 13:285–309 Bakan DA, Doerr-Stevens JK, Weichert JP et al (2001) Imaging efficacy of a hepatocyte-selective polyiodinated triglyceride for contrast-enhanced computed tomography. Am J Ther 8:359–365 Cassel DM, Young SW, Brody WR et al (1982) Radiographic blood pool contrast agents for vascular and tumor imaging with projection radiography and computed tomography. J Comput Assist Tomogr 6(1):141–146 Choukèr A, Lizak M, Schimel D et al (2008) Comparison of Fenestra VC contrast-enhanced computed tomography imaging with gadopentetate dimeglumine and Ferucarbotran magnetic resonance imaging for the in vivo evaluation of murine liver damage after ischemia and reperfusion. Invest Radiol 43:77–91 Ford NL, Graham KC, Groom AC et al (2006) Time-course characterization of the computed tomography contrast enhance-

ment of an iodinated blood-pool contrast agent in mice using a volumetric flat-panel equipped computed tomography scanner. Invest Radiol 41(4):384–390 Graham KC, Ford NL, MacKenzie LT et al (2008) Noninvasive quantification of tumor volume in preclinical liver metastasis models using contrast-enhanced X-ray computed tomography. Invest Radiol 43:92–99 Hainfeld JF, Slatkin DN, Focella TM et al (2006) Gold nanoparticles: a new X-ray contrast agent. Br J Radiol 79: 248–253 Henning T, Weber AW, Bauer JS et al (2008) Imaging characteristics of DHOG, a hepatobiliary contrast agent for preclinical microCT in mice. Acta Radiol 15:342–349 Hughes PM, Bisset R (1991) Non-ionic contrast media: a comparison of iodine delivery rates during manual injection angiography. Br J Radiol 64(761):417–419 Koa C-Y, Hoffman EA, Beck KC et al (2003) Long-residencetime nano-scale liposomal iohexol for X-ray-based blood pool imaging. Acta Radiol 10:475–483 Kim SH, Park JH, Kim YI et al (1990) Experimental tissue damage after subcutaneous injection of water soluble contrast media. Invest Radiol 25:678–685 Krause W, Leike J, Sachse A et al (1993) Characterization of iopromide liposomes. Invest Radiol 28(11):1028–1032 Krause W (1994) Preclinical characterization of iopromide. Invest Radiol 29(1):S21–S32 Mützel W, Speck U, Weinmann H-J (1983) Pharmacokinetics of iopromide in rat and dog. In: Taenzer V, Zeitler E (eds) Contrast media in urography, angiography and computed tomography. Thieme, Stuttgart, pp 85–90 Poste G (1983) Liposome targeting in vivo: problems and opportunities. Biol Cell 47:19–38 Rabin O, Perez JM, Grimm et al (2006) An X-ray computed tomography imaging agent based on long-circulating bismuth sulphide nanoparticles. Nat Mater 5(2): 118–122 Rau T, Mathey D, Schofer J (1997) High-dose tolerance of iodinated x-ray contrast media. New developments in x-ray and MR angiography symposium CIRSE, 9.9.96, Funchal, Madeira. Cardiovasc Intervent Radiol 20:8–9 Rosovsky M, Rusinek H, Berenstein A et al (1996) High-dose administration of non-ionic contrast media: a retrospective review. Radiology 200:119–122 Sachse A, Leike JU, Schneider T et al (1997) Biodistribution and computed tomography blood-pool imaging properties of polyethylene glycol-coated iopromide-carrying. Invest Radiol 32(1):44–50 Seltzer SE, Gregoriadis G, Dick R (1988) Evaluation of the dehydration-rehydration method for production of contrast-carrying liposomes. Invest Radiol 23(2):131–138 Senior JH (1987) Fate and behaviour of liposomes in vivo: a review of controlling factors. Critic Rev Ther Drug Carrier Syst 3:123–193 Schmiedl UP, Krause W, Leike J et al (1999) CT blood pool enhancement in primates with iopromide-carrying liposomes containing soy phosphatidyl glycerol. Acad Radiol 6:164–169 Torchilin VP, Frank-Kamenetsky MD, Wolf GL (1999) CT visualization of blood pool in rats by using long-circulating, iodine-containing micelles. Acad Radiol 6:61–65 Vera DR, Mattrey RF (2002) A molecular CT blood pool contrast agent. Acad Radiol 9:784–792

Small Animal Magnetic Resonance Imaging: Basic Principles, Instrumentation and Practical Issue

12

Peter Jakob

Contents 12.1 Introduction............................................................. 151 12.2 The Basic Operating Principles of Magnetic Resonance................................................................ 152 12.3 Basic Operating Principles of MR Imaging......... 155 12.4 Advanced Reader Section: Advanced Operating Principles of MR Imaging................... 156 12.5 MR-Facility and MR-Hardware............................ 157 12.6 Basic Information Content of MR Microscopy.............................................................. 160 12.7 The Bermuda Triangle of Small Animal MR Microscopy....................................................... 160 12.8 Animal Handling..................................................... 161 12.9 MR Microscopy: Imaging Examples.................... 162 12.9.1 Tumor Characterization via MR Microscopy........... 162 12.9.2 MR Microscopy in Fixed Specimens........................ 163 12.10 Summary................................................................... 163 References............................................................................ 164

P. Jakob Department of Experimental Physics 5, University of Würzburg, Am Hubland, 97074 Würzburg, Germany e-mail: [email protected]

12.1 Introduction In biomedical research, many experimental study designs require the application of small animal models for complex human diseases, where the underlying disease mechanisms can be explored in vivo by collecting anatomical, functional, and metabolic information noninvasively and simultaneously. Therefore, there is clearly a great demand in animal research for imaging modalities that can provide this information in vivo in a longitudinal and three-dimensional (3D) fashion without limitations by depth resolution. Among the different noninvasive imaging techniques available, MRI (magnetic resonance imaging) is currently one of the modalities of choice, since MRI has a number of characteristics that make it an ideal candidate for a broad range of imaging tasks: (1) extraordinary 3D capabilities, (2) superb contrast abilities and spatial resolution combined with a high temporal resolution, and (3) a high safety profile since no harmful ionizing radiation is applied. Since the acquisition of the first human magnetic resonance (MR) images in the early 1980s, MRI has gained an ever-increasing role not only in medical diagnosis, but also in biomedical research. This level of success is largely due to its noninvasive nature and its high inherent tissue contrast capable of detecting subtle changes in soft tissue. Because of this abundance in contrast possibilities and richness in information content compared to, for example, X-rays (where tissue absorption is the only contrast parameter), the MR community has spent tremendous efforts to extend the possibilities of MRI for small animal investigations. MRI at this more microscopic level (also called MR microscopy, MRM) is based on the same physical principles as its clinical counterpart MRI: Nuclear

F. Kiessling and B.J. Pichler (eds.), Small Animal Imaging, DOI: 10.1007/978-3-642-12945-2_12, © Springer-Verlag Berlin Heidelberg 2011

151

152

spins (e.g., the protons of a water molecule) possess tiny magnetization, which precess at a very specific frequency when placed in a strong magnetic field. These spins can interact efficiently with an additional applied oscillating magnetic field at the same (resonance) frequency, absorbing and reemitting energy that can be spatially encoded via magnetic field gradients to generate a digital image data set. MRM, which can loosely be defined as MRI in the spatial resolution regime beyond 100 mm, is now increasingly exploited to provide anatomical and functional information in (transgenic) animal models, with the focus of revealing anatomical and physiological changes during disease progression or in response to potential therapeutic treatment. Depending on the technique used, slice-based as well as true 3D image sets can be obtained, with resolution of 100 mm in all dimensions in living animals and 10 mm isotropic in fixed specimens, with the possibility to view images of whole animals or organs in arbitrary planes. This extension of MRI to MRM, however, required a tenfold increase in spatial resolution in all three dimensions, resulting in signal reductions of at least a factor of 1,000. Overcoming this deficit required the development of technology specific to MRM – strong magnets, dedicated receiver coils and chains, strong gradients sets, image acquisition sequences, and finally biological support for small animals in high magnetic fields. The focus of this chapter is (a) to describe the most fundamental aspects of MR and MRM and its limitation, (b) to describe the corresponding instrumentation with a focus on small animal imaging, (c) to give some practical advice for performing animal studies.

12.2 The Basic Operating Principles of Magnetic Resonance The purpose of the following section is to describe the most fundamental aspects of MR (Abragam 1961; Ernst et al. 1986). As the term “magnetic resonance” already indicates, MR is a resonance phenomena which takes place in the world of nuclear magnetic moments. Any precise and detailed description of MR phenomena requires the use of “hard core physics = quantum mechanics (QM),” which we for clarity reason avoid in favor of a more approximate, and obviously not fully correct, description.

P. Jakob

First, it is important to understand the origin of the MR signal, which has its source in the nuclear magnetic moments of the atoms and molecules that constitute the biological object to be imaged. In principle, many nuclei exhibit the property of possessing a magnetic moment, including 1H, 31P, 13C, 15N, and 19F, but in the following we consider only the signal source in conventional MRI which is actually water or, more specifically, the tiny magnetic moments of the protons that make up the nuclei of the hydrogen atoms in water. This so-called hydrogen or 1H MR utilizing the hydrogen nucleus is the most important one for common small animal applications. This importance stems from the fact that the 1H nucleus is highly naturally abundant (110 mol/L concentration in pure water and a little less in biological tissue) and has an associated large magnetic moment compared to other nuclei. To understand the origin of the 1H-MR-signal, one needs to have a closer look at the basic properties of the atomic nucleus itself. The nucleus of a hydrogen atom can be envisioned as a positively charged proton of spherical shape, which in addition is spinning around one axis (Fig. 12.1a). According to the fundamental principles of electromagnetism, a charged spinning particle produces a magnetic field in its vicinity and thus possesses a magnetic moment. Therefore, these protons can be considered as miniaturized magnets or tiny compass needles, which we call in the following “spins.” If we consider now a sample containing water and thus a plenty of hydrogen nuclei in an absolutely magnetic field free environment, we observe that the orientation of the spins is random in space (Fig. 12.1b). However, when the same water sample is placed in a homogeneous magnetic field B0, a very different situation arises, which in the following will be explained in a two-step fashion. Step 1: The magnetic field B0 provides an axis (referred to as the longitudinal axis) for the nuclei to align along or against. Spin physics (Abragam 1961; Ernst et al. 1986) now dictates that there is an imbalance in the aligned vs. anti-aligned fraction and that a small surplus amount of the spins will align with B0. Or, in other words, the amount of spin alignment along the B0 direction is slightly higher than the alignment against B0 (this behavior is somewhat confusing for nonphysicists, since they would assume that all spins would align along B0 like compass needles in the earth magnetic field). This surplus amount of spins and thus magnetic moments aligned in the same direction as B0 now sum up to a larger total magnetic moment resulting in a net

153

12 Small Animal Magnetic Resonance Imaging: Basic Principles, Instrumentation and Practical Issue

a

b

with magnetic field B0

magnet nuclear spin

+

compass needle

N no magnetic field

B0

~ = longitudinal direction

S transverse direction

c

d

e

N

N

N

RF-coil NMV T1 -recovery

B0

T2 -decay

S

S

S

Fig. 12.1 Basic description of MR. (a) A proton – the positively charged nucleus of a hydrogen atom – spinning around one axis can be envisioned as a tiny compass needle (=spin). (b) The spins have random orientation in space in the absence of an external magnetic field. If an external magnetic field, B0, is applied, the spins align along or against the B0 direction with a small surplus amount of the spins aligning with B0 (please note, that this is a schematic drawing only which (1) exaggerates spin alignment along B0, and (2) for clearness spatially separates these two spin alignment groups). (c) If the surplus amount of

spins and thus magnetic moments aligned with B0 are added together, it results in net magnetization vector (NMV), which precesses around B0. (d) Applying a short radio frequency pulse generated by a RF coil creates a “magnetic” torque, causing the NMV to tilt by 90° from the longitudinal direction to the transverse direction. (e) Once the RF pulse is turned off, the tilted magnetization will come full circle, by recovering or “relaxing back” to its equilibrium state (T1 relaxation), while the transverse component decays (T2 relaxation)

magnetization vector (NMV), which still points in the direction of B0. According to the rules of QM, this net magnetization can be calculated and it can be shown that the NMV ~ B0 (Abragam 1961; Ernst et al. 1986).

This is an important result, since it tells us that in order to achieve a high net magnetization (and thus MR signal, which is directly proportional to the magnitude of NMV) one needs a high magnetic field strength.

154

P. Jakob

However, it is important to stress at this stage that even at very high magnetic fields such as that of a 17.6 T MRM system and at rodent temperature, the small surplus amount of spins aligned with B0 is only on the order of a few parts per million, which explains the rather low sensitivity of MR and consecutively MRI and MRM. Step 2: In addition to the alignment of the individual magnetic moments described in step 1, the arising net magnetization shows the unexpected behavior of precessing (rotating) around the longitudinal axis at a certain frequency w0, the so-called Larmor frequency (Fig. 12.1c). The arising precessional frequency can be calculated according to the so-called Larmor equation (Abragam 1961; Ernst et al. 1986):

ω 0 = γ · B0 .

(12.1)

The constant g in this equation is a nucleus-specific constant known as the gyromagnetic ratio. For hydrogen g is 42.58 MHz/T, and typical values for B0 for small animal imagers range from 4.7 T to as high as 20 T. Thus, correspondingly protons precess at frequencies that fall into the radiowave frequency (RF) range of 200–850 MHz. The Larmor frequency is very important for the following considerations, as it denotes the frequency of radio waves with which the precessing magnetization can interact. In summary, if one puts a sample of water into a homogeneous magnetic field, one observes spin alignment accompanied by a precession motion of the resulting magnetization, which can be seen as the important starting point of the following considerations. So far, we have only discussed what happens if we put a hydrogen containing material in a static magnetic field. In order to generate a measureable MR signal, a further step that constitutes the “resonance part” of our MR experiment is needed (Fig. 12.1d). This crucial step, also known as “RF excitation,” compromises a short radio frequency (RF) pulse which is applied for several milliseconds to the precessing magnetization NMV. This RF pulse can be generated by a RF coil which is different in shape and size compared to RF antennas used in radio stations and which actually has to be placed in very close proximity to the sample. The RF field produced can be viewed as an alternating magnetic field in the MHz range, which can interact with the precessing magnetization by creating a “magnetic” torque. As a result of applying this RF pulse, the

precessing magnetization will absorb energy, change its direction, and can be tilted away from its equilibrium state (which is originally along the longitudinal direction). In other words, the RF pulse can be seen as an external disturbance of the precessing magnetization in the equilibrium state. This energy absorption and tilting process is most efficient, when the radiofrequency magnetic field is applied at exactly the Larmor frequency ω 0 = γ · B0 , which reflects the resonance aspect of our MR experiment. In most of the MR experiments in use today, the RF applied is designed to tilt the precessing magnetization by exactly 90° in the so-called transverse direction (Fig. 12.1d). As soon as the RF pulse is switched off, three different processes take place all at the same time: Process 1: The magnetization continues to precess in the transverse plane: While the magnetization is continuing its precessing motion (@ Larmor frequency), it is changing the magnetic flux presented to a surrounding receiver coil placed in the transverse plane, which can be actually the same coil used beforehand as an excitation tool. The changing magnetic flux from the spinning nuclear magnetic moments induces a voltage via electromagnetic induction in the receiver coil (= dynamo principle). This induced voltage, which can be further amplified and digitized, represents the MR signal created by our MR experiment. This process describes how the MR signal is actually generated. Process 2: Once the RF pulse is turned off, the tilted magnetization will come full circle, by recovering or “relaxing back” to its equilibrium state, which is along the longitudinal direction (Fig. 12.1e). This longitudinal recovery process is known as spin–lattice relaxation with a tissue characteristic time constant T1. This T1 recovery is due to nuclei releasing energy to the surrounding, which is called the “lattice” for historical reasons. Process 3: The T1 recovery process is accompanied by a decay of the transverse magnetization which is known as spin–spin relaxation with the tissue characteristic time constant T2. This T2 process originates from magnetic interaction between the magnetic moments of the individual spins. In the case of water protons, after a few tenth to hundred milliseconds of precessing, the protons return back to being aligned with the magnetic field. Afterward, the experiment can be repeated and the MR signal can be detected again. It should be noted that if all tissue type would have the same water density and identical relaxation times T1 and T2, they all

12 Small Animal Magnetic Resonance Imaging: Basic Principles, Instrumentation and Practical Issue

would produce the identical MR signal intensity and thus it would be impossible to obtain contrast between the different tissue types. Fortunately, different tissues relax at different rates (i.e., have different relaxation times T1 and T2), which can be exploited to create image contrast. As relaxation is one of the fundamental aspects of MR, an extensive literature on theoretical and experimental aspects of relaxation has developed since the earliest days of MR (Abragam 1961; Cowan 1997; McConnell 1987). Theoretical treatments (Abragam 1961; Cowan 1997; McConnell 1987) emphasize the role of spontaneous and stimulated emission processes from excited states back to the ground state of a spin. Generally speaking, the release of energy involved in longitudinal relaxation (T1 relaxation) can be either “spontaneous” or “stimulated.” However, theoretical considerations predict that spontaneous emission is highly unlikely in the (relatively low) frequency range of MR, e.g., an excited proton in a static magnetic field of, for example, 1.5 T ( 64 MHz), has a transition probability of only W » 10−24/s and thus would require more than »1024 s to return spontaneously to its stable ground state, the longitudinal direction; thus, spontaneous emission is a completely ineffective relaxation mechanism for magnetic resonance. Instead, spin transition is a consequence of a stimulated emission process originating from time-varying magnetic fields via magnetic coupling of the spin system to the surrounding. Thus, for efficient stimulated transition to occur, magnetic interactions between excited spins and local magnetic fields, which are timevarying and at the appropriate frequency, the Larmor frequency, are mandatory (Abragam 1961; Cowan 1997; McConnell 1987). In the case of 1H (spin 1/2) MR in solution (i.e., in vivo), one predominant source of local time-varying magnetic fields that are efficient in stimulating relaxation are e.g., dipole–dipole through-space interactions. In this type of interaction, a spin and its corresponding magnetic dipole (producing a magnetic dipole field around), which is subject to random Brownian motion, creates on the site of neighboring spins fluctuating magnetic fields randomly with time. These magnetic field fluctuations are vital for stimulated emission and thus relaxation and are most efficient when they produce oscillations exactly at the Larmor frequency. In summary, the fluctuating magnetic fields from nearby tumbling magnetic moments promote proton relaxation.

155

12.3 Basic Operating Principles of MR Imaging In the basic MR experiment described above, where an ensemble of spins (e.g., our rodent) gets placed in a homogeneous magnetic field B0 and excited via a RF pulse at Larmor frequency, the entire sample would produce one and only one large MR signal precisely at the Larmor frequency. In other words, with the main magnet and radio-frequency coil alone, one is not able to perform an MR imaging experiment, since so far we have absolutely no means to distinguish MR signals from different locations of the rodent. The imaging task now consists of spatially resolving the MR signal by assigning each voxel a corresponding MR signal via a process called spatial encoding. The key of this encoding process are additional magnetic fields inside the magnet, which can be purposely switched on and off (Lauterbur 1973; Mansfield and Grannell 1975; Mansfield 1977). For this procedure, an additional hardware element, a so-called gradient coil, is required, which can produce inhomogeneous magnetic fields along all three spatial dimensions (x, y, z) inside the sample, superimposed on B0. The gradient coils are on purpose designed to generate linear varying magnetic fields, e.g., a gradient field G, where the magnetic field strength varies in a welldefined manner with the position along each spatial dimension. The gradient G (dimension: change of magnetic field in mT/unit change of spatial coordinate in m) has three components, Gx, Gy, and Gz, each associated with its respective spatial axis x, y, or z. The application of such a gradient field reduces the main magnetic field strength at one end of the animal and increases it at the other end (Fig. 12.2). In response, the Larmor frequency is linearly varying along space, because the precession frequency is proportional to the magnetic field strength, as can be seen by the following modified Larmor equation for a gradient G applied in x-direction:

ω 0 ( x) = γ ·( B0 + G · x).

(12.2)

This one-to-one relationship between magnetic field, frequency, and space expressed by (12.2) can now be used to perform a simple one-dimensional (1D) MRI experiment called “frequency encoding”: For this imaging experiment, we apply this well-defined magnetic field gradient during MR signal acquisition. In this case spins in different locations inside the gradient field experience

156

P. Jakob

a

b

magnetic field/ resonance frequency

magnet

magnetic field/ resonance frequency

ω 0 +∆ ω ω0

B0

ω0

magnet & gradient coil

G B0

ω 0 –∆ω Space x

Space x

Fig. 12.2 Principle of spatial localization using an additional gradient field. (a) In the presence of the main magnetic field only, one resonance signal at the corresponding Larmor frequency can only be detected. (b) The application of a gradient field reduces the main magnetic field strength at the left end of the animal and increases it at the right end. Thus, the gradient

creates a frequency dispersion across the animal, where the Larmor frequency is linearly varying along space. Via Fouriertransform, which extracts the frequency content and thus spatial information out of the MR signal, a one-dimensional image of the animal can be obtained

different magnetic fields and have associated with different Larmor frequencies, which again are directly related to their spatial position via (12.2). Since the detected MR signal contains now all the frequency components created by the spin ensemble in the gradient field, the corresponding signal amplitudes represent a 1D image of the animal. For this reason, this technique of spatial localization is called “frequency encoding.” The spatial information can be extracted from the acquired frequency-encoded MR signal mathematically via a computer performing a so-called Fourier-transform (FT), which extracts the frequency and thus spatial information out of the MR time signal. This concept can be extended easily to 2D and 3D imaging by employing additional gradient fields. Further, it is important to note that the basic of all spatial encoding steps in any conventional MR imaging experiment, including the advanced concepts of phase encoding or slice selection (see the following Sect. 12.4, relies on the simple relationship between magnetic field, frequency, and space).

line methods, reconstruction from projections, and Fourier imaging. For a more detailed discussion and a good illustrated treatment about the development and historical significance of each of these, the reader is referred to a number of general publications (Callaghan 1991; Ernst et al. 1986; Gadian 1982; Haacke et al. 1999; Morris 1986; Wehrli et al. 1988). In general, MRI sequence is a collection of RF and gradient events designed to create an MR image with particular characteristics. Most present-day MRI is based on the principles of FT. 2D FT MRI, which is the working horse for MRM, achieves spatial localization using selective excitation and making use of the frequency and the phase of the MR data simultaneously. The spatial distribution of protons within a sample, r(x,y), is mapped by rendering their Larmor precession frequency spatially dependent with magnetic field gradients. In this 2D-FT imaging approach three fundamental steps are combined to localize points spatially along the three axes in space (x, y, z) using three orthogonal magnetic field gradients. The first step is slice selection using selective irradiation to produce MR signals from only a relatively narrow “planar” volume. Selective irradiation (Fig. 12.3) that prepares the magnetization of interest for subsequent spatial encoding uses a frequency-selective, band-limited RF pulse (e.g., a RF pulse which excites only certain frequencies inside a limited frequency band) in conjunction with a selection gradient GSlice. The next fundamental process is the in-plane localization of the received MR

12.4 Advanced Reader Section: Advanced Operating Principles of MR Imaging Throughout the development of MRI techniques, there have been many methods for data acquisition and image reconstruction, including the sensitive point and

12 Small Animal Magnetic Resonance Imaging: Basic Principles, Instrumentation and Practical Issue

signal with additional gradients. Two fundamental approaches are available here which exploit the frequency and the phase of the received signal. Frequency encoding uses a gradient GRead during signal reception (also known as read-out period), so that spins at different positions along this gradient generate signals at different frequencies (see preceding section and Fig. 12.3). The detected time MR signal S(t) is digitized as a series

a

1

Slice selection

NPhase

Ph

as

e

en

co

di

ng

2

NRead

3

Frequency encoding

b selective RF pulse

1

signal reception & data acquisition NRead

157

of NRead brief samples using an analog-digital-converter (ADC), where NRead determines the number of pixels along the read-direction. In the phase-encoding process, a gradient pulse GPhase of duration T is used between excitation and signal reception. While the gradient during signal recording has a direct effect on the frequency of the signal, the gradient pulse before reception affects the phase of the detected signal. By this phase-encoding pulse, a spatially dependent phase shift j in the direction orthogonal to the read-direction is introduced. A series of such phase-encoded signals forms the basis for spatial encoding in the second dimension. The complete phase-encoding procedure requires repeated measurements for NPhase different values of the phase-encoding gradient, where NPhase determines the number of pixels along the phase-direction (Fig. 12.3). It is important to note that the gradients applied during the spatial encoding experiments physically perform a 2D FT on the spin density. Hence, the set of received time signals acquired during the imaging experiment represents the 2D spatial frequency spectrum of the object. The most noteworthy feature of this frequency spectrum is the property of its Fourier inverse yielding the image. In other words, MRI is compared to optical imaging by a two-step procedure: Step 1 comprises the Fourier-transform of the object via the application of spatially encoding gradients; Step 2 comprises the reconstruction of the MR image via performing an inverse Fourier-transform of the acquired MR data using a computer.

RF/ADC MR-signal S(t)

12.5 MR-Facility and MR-Hardware

GSlice

3 GRead NPhase

2 GPhase

NPhase repetitions T

Fig. 12.3 Principle of three-dimensional spatial localization using three orthogonal gradients. (a) Schematic summary of the spatial encoding procedure. (b) Corresponding pulse sequence diagram (oversimplified version). Step 1 compromises the slice selection procedure using a frequency-selective RF pulse in combination with a slice selection gradient. During step 2, the so-called phase encoding procedure, a gradient pulse Gphase of duration T is applied to produce a position-dependent phase shift j. Finally, in step 3 frequency encoding is performed during the detection of the MR signal S(t)

The MR system (shown schematically in Fig. 12.4) that detects and spatially localizes the hydrogen nuclei usually comprises three different key elements. (a) The main magnet, which is the most costly element of the system, creates the timely constant and spatially homogeneous magnetic field B0 inside the sample. The magnets in widespread use for animal research today are powerful cryo-based superconducting magnet system with excellent field stability and magnetic field strength up to 21 T. The low temperature required for superconductivity is achieved with the help of liquid helium. The magnetic field strength is used to align the nuclei, thus

158

P. Jakob

a

b

main magnet: 7-21T

c

RF coil

Gradient coil

Fig. 12.4 MR-hardware components: (a) the main magnet creates the timely constant and spatially homogeneous magnetic field B0 inside the sample. (b) The radio-frequency antenna

(in this particular case a solenoid coil) which surrounds the animal. (c) The gradient fields are generated by current carrying coils

producing the net magnetization, and at the same time, determines the excitation and detection frequency of the MR experiment. Since the signal-tonoise ratio (SNR) increases as B0 increases (see later), it is best to minimally have a B0 of at least 4.7 T and ideally at least 7.0 T or higher for small animal MR imaging. In order to put these field strength numbers into a context, it is important to know that one Tesla represents a field strength that is approximately 10,000 times stronger than the earth magnetic field. In addition, horizontal scanners are preferable to vertical scanners for small animal applications for the obvious reason that it is easier to have the animal in prone position rather than have it in a “standing position” suspended by its teeth. The bore size of the magnet is “animal size adapted” for cost efficiency and typically on the order of 16–30 cm. The typically usable bore size that remains after inserting the animal in the inevitable additional RF- and gradient coils is on the order of only 5–15 cm.

small animal MRI. However, these systems are relatively expensive and not disposable. Preexisting clinical whole-body MRI systems offer a more cost-effective (of course only when used during nights and weekends) and widely available alternative to dedicated small animal systems (Brockmann et al. 2007). Compared to dedicated small bore scanners, most clinical scanners operate at lower field strengths between 1.5 and 3.0 T, which limits the achievable SNR. Thus, according to (12.3), these scanners have limited spatial and temporal resolution compared to dedicated animal scanners. However, besides their availability, clinical MR imaging systems also offer advantages, such as image contrasts which directly match the image contrasts familiar from patient studies (since relaxation times are fieldstrength dependent). Thus, a direct transfer of the knowledge from mouse to man and vice versa is possible. Small animal MRI with clinical 1.5–3.0 T MRI systems can deliver sufficient image quality in applications which do not require extremely high temporal or spatial resolution and offers many diagnostic options for applications such as brain-, abdominal-, spinal cord-, and cardiac-imaging (Brockmann et al. 2007). Whenever no extraordinary spatial resolution is

Small bore MR scanners operating at field strengths between 4.7 and 21 T are the systems of choice for

12 Small Animal Magnetic Resonance Imaging: Basic Principles, Instrumentation and Practical Issue

required, small animal imaging protocols can be realized on clinical MR scanners in scan times comparable to those used for human examinations using optimized RF coils and dedicated pulse sequences. Nevertheless, many animal studies are exclusively limited for SNR reasons to dedicated small bore animal scanners. In this context, the possibility of scanning several animals (in particular mice) at a time in a whole body human system with field strengths between 1.5 and 3 T and with bore sizes of approximately 60 cm should be mentioned (Dazai et al. 2004). This approach facilitates high throughput, concurrent imaging of multiple mice (e.g., 6–8 mice simultaneously) for certain imaging tasks. This increased throughput, however, comes at the cost of certain limitations including significant loading times, the reduced ability to individually control each independent mouse experiment, to efficiently perform triggering and navigation, and finally, reduced image quality due to the lower field strength. (b) The radio-frequency antenna also called RF coil (in the most simplest implementation just a loop of copper wire), which is in close proximity or surrounds the animal, often works, as already described above, both as an RF transmitter and a RF receiver for radio-frequency fields. Thus, RF coils are the key elements for initiating the resonance experiment and detecting the response MR signal of the precessing magnetization and finally also key players in the “SNR-game” (see (12.3)). Various RF coil types are commonly used in small animal MRI such as surface coils, volume coils, and phased array coils (Callaghan 1991; Gareis et al. 2007; Haase et al. 2000). A surface coil is essentially a loop of wire which may form various shapes and can be placed on or around the surface of the animal. Traditionally, the utility of the surface coil arises from its simple design and inherent spatial selectivity, allowing localized observations of specific tissue. Additionally, and significantly, surface coils have a much higher sensitivity and thus deliver higher SNR than homogeneous volume coils for regions near the sample surface. However, as a rule of thumb, single surface coils can only effectively image a limited region whose dimensions are comparable to the diameter of the surface coil, since the sensitivity drops rapidly with distance from the loop center. Thus, due to the rapid fall-off of the sensitivity of the surface coil, large-volume coils outperform surface coils for regions deeper within the sample. Volume coils such

159

as the birdcage or solenoid coils (Callaghan 1991; Haase et al. 2000) instead cover the whole animal and provide a very homogeneous sensitivity profile over the whole animal, so that whole-body mouse imaging becomes possible, however at the price of lower SNR values at the surface of the animal. Thus, surface coils are advantageous for MR when the region of interest is very near the surface, and volume coils are advantageous at greater depths, where the volume coil has a higher sensitivity in addition to its higher homogeneity. In practice, simple animal experiments can be performed on clinical systems using already existing small surface coils which are available in various diameters (typically 4–10 cm). A costly, but advanced possibility for optimizing the sensitivity performance in small animal studies is the concept of phased array coils (Gareis et al. 2007), which has become increasingly important when using clinical scanners. Using phased array coils, the limited fieldof-view of a single coil loop can be increased without significant loss in sensitivity when several decoupled surface coils are used, each with its own signal acquisition path (phased array). In this way, several independent images are acquired at the same time, and these are combined into a single image after dedicated postprocessing schemes. This allows substantially increased overall SNR. A general feature of phased array coils is the flexibility to adapt the coil geometry more closely to the region of interest than classical designs. The main purpose of such array systems is to distribute the high SNR performance of their small component surface coils, over a larger area covered by the entire array, with no increase in imaging time. Thus, this phased array technology can be used to improve image quality within the same acquisition time. Today, phased array coils for mice and rats with 4–16 coil elements are commercially available for clinical MR systems and are the “coilsof-choice” whenever extended field-of-view applications are of importance. (c) Unlike the strong field generated by the cryomagnet, the gradient fields are much weaker and are generated by current-carrying coils designed to produce a desired magnetic field gradient. For 2D or 3D MRI, three pairs of orthogonal current-carrying coils located within the magnet are needed for spatial encoding. These coils and thus their magnetic field action can be turned on and off during the course of an imaging experiments

160

P. Jakob

(= sequence) leading to significant acoustic noise, since physical forces, the so-called Lorentz forces, produce bulk vibrations of the gradient coil. Since the spatial resolution of the encoded image is determined by the time-amplitude product of the gradient pulses used to encode the image, high resolution microscopy requires extraordinary strong gradients. Theses gradients are needed to create a sufficient frequency dispersion across the animal. The gradients for MRM (200 mT/m-1T/m) are typically 10–50 times those of a clinical system (20 mT/m) to achieve the same “anatomical resolution” in rodents compared to humans. Depending on the MR imaging technique used, slice-based 2D image data sets with arbitrary slice orientation or full 3D image data can be obtained.

12.6 Basic Information Content of MR Microscopy In the proceeding section we have discussed the most important physical and hardware aspects of an MR imaging experiment, at least superficially, and in the following we want to have a closer look at the corresponding information content of the generated MR images. Commonly, the MR images acquired are based on the radio frequency (RF) signals emitted from hydrogen nuclei (=protons) contained in the body, where the image contrast is based primarily on the inherent MR tissue properties. The most fundamental inherent tissue properties, which can be exploited as a natural source of image contrast by appropriate MR pulse sequence design, are in detail (a) the local concentration of hydrogen (= magnetization or spin density or even simpler water density) and (b) the spin relaxation parameters (T1, T2). Additional contrast sources such as flow, diffusion, temperature, pH, etc. are also available, but are beyond the scope of this section. Contrast in MR images arises now from the fact that different tissue types have different water densities which can relax at different relaxation rates (i.e., have different T1 and T2 values). In standard MR imaging techniques, the MR signal intensity increases with the local spin density and spin–spin relaxation time T2 and decreases with the spin–lattice relaxation time T1. In more sophisticated MR experiments, for example, the intrinsic tissue magnetic susceptibilities or the tissues’

flow and diffusion characteristics can be exploited as a natural source of image contrast. Despite MRI’s various endogenous sources of contrast, which can be emphasized by appropriate MR sequence design, there is still the option/need for additional exogenous contrast materials which artificially enhance contrast. The driving force behind the use of additional contrast agents in animal models is, for example, (a) to enhance the contrast between normal and pathological tissue, (b) to better delineate regions otherwise invisible to standard MRI techniques, and (c) to obtain functional tissue information (e.g., perfusion). In contrast to conventional radiography and computed tomography contrast agents (iodine, barium), where the contrast media are visualized themselves due to a greater attenuation of the X-ray beam in agent-containing tissues, in MR imaging, the effect of the contrast agents is indirect. Here, no signal is derived from the MR contrast agents themselves, but they work indirectly by altering the relaxation times of hydrogen nuclei that come in close proximity to the contrast agent (Hendrick and Haacke 1993; Lauffer 1987; Muller 1996; Wood and Hardy 1993). Depending on the field of application, different types of agents can be used to cause either positive (increased MR signal intensity) or negative (decreased MR signal intensity) contrast enhancement in MRI images of tissues where the agent has been accumulated. Some prominent examples of contrast agents for animal MRI include paramagnetic substances, such as molecular oxygen, complexes of lanthanides (Gd3+) (Lauffer 1987), or superparamagnetic substances, such as small crystals (<100 nm) of magnetite (Fe3O4) (Bulte and Kraitchman 2004).

12.7 The Bermuda Triangle of Small Animal MR Microscopy Already the pioneers of MR imaging realized that the fundamental limits of spatial resolution were in the microscopic domain. As one scales proton MR imaging from the human dimension down to the microscopic dimension of a rodent, several challenges and limitations emerge (Benveniste and Blackband 2002; Callaghan 1991; Eccles and Callaghan 1986). To appreciate this special situation, it is instructive to briefly review some basic MR signal-to-noise

12 Small Animal Magnetic Resonance Imaging: Basic Principles, Instrumentation and Practical Issue

considerations. One major goal in rodent MRM is to achieve a very high spatial resolution in order to efficiently resolve the most tiny structures of the animal under investigation. In this context, however, it is important to realize that the rodent image must be acquired with a voxel volume approximately 1,000– 3,000 times smaller than that of a human, in order to be able to resolve the same relative anatomical details. A typical voxel in human MRI is 1.5 × 1.5 × 10 mm3, comprising a voxel volume of »23 mL. To resolve the same anatomical details in a mouse, the typical voxel is 100 × 100 × 500 mm3 which represents a voxel volume of only 5 nL. These simple facts bring us to one of the most important aspects of MRM: The SNR which can be achieved in a certain imaging time. The SNR is one key parameter describing the quality of MRI images comparing the level of the desired MR signal to the level of background noise corrupting the MR signal. The SNR for a given 3D pixel element (voxel) in the image can be described as a first approximation by the following (oversimplified) SNR equation:

SNR ∼ ρ · µ · B0 ·

1 · Vvox · Taq . d

(12.3)

According to (12.3), the SNR depends on several factors, including 1. The net nuclear magnetization NMV: The density of water (or other examined nuclei) r in the object, the strength m of the spins nuclear magnetic moments, and the amount of spin alignment which is proportional to B0 collectively constitute the net nuclear magnetization and thus the signal source in MRI. In summary, the MR signal increases in a first approximation linearly with B0 which explains the continuing development of stronger magnets for MRI. 2. The size d of the reception coil: As a rule of thumb, the SNR increases linearly as the coil size de creases. In other words, the smaller the coil and the better it encompasses a given sample, the better the coil performs and thus the SNR improves. 3. The volume of image voxel Vvox: The larger the voxel volume, the more spins contribute to the MR signal; however, increasing the voxel volume is not the method of choice to increase the SNR, since this comes along with a reduction in spatial resolution. 4. The total acquisition time Taq needed to take an

161

image, which can be prolonged by signal averaging (whenever the SNR of a single experiment appears to be insufficient). The square root dependence in this equation tells us that doubling the SNR by averaging would require 4-times longer Taq. However, extending imaging times introduces the need for precise physiological control to prevent rodent motion from causing artifacts. Equation (12.3) applied to the scenario of small animal imaging rules that the MR signal from the “tiny mouse voxel” is approximately 3,000 times smaller than that from the “big human voxel.” This signal deficit can be partially compensated by going to higher magnetic field strength, designing size-adapted, more efficient receiver coils, and finally by increasing the total acquisition time Taq via signal averaging. Given these issues, once the magnet has been selected and a dedicated RF coil has been designed, the achievable spatial resolution is mainly determined by total acquisition time Taq available. The fact that the SNR increases only with the square root of Taq while decreasing linearly with Vvox has a huge impact on imaging time. For example, halving the voxel volume reduces the SNR by a factor of 2, and thus to regain the SNR of the original voxel will require a fourfold increase in the total imaging time. Thus, an original 10 min lasting mouse scan would require 40 min with the voxel volume being halved. This simple relationship is of paramount importance when planning imaging studies in rodents, since the animals may not be able to tolerate prolonged imaging sessions. Rodents can typically be anesthetized for 1–3 h, which actually sets an upper limit for the MR imaging protocols to be used.

12.8 Animal Handling In general, MRM requires the rodents to be immobilized and thus anesthetized. To avoid negative chronotropic and inotropic effects, inhalative anesthetics such as isoflurane or IV anesthetics such as propofol are commonly used and given via nose cone or tail vein catheter and allow for short induction and recovery times. While under anesthesia, the temperature of the animals must be kept constant at 37°C using either warm airflow and circulating water or electrical heating pads, since the temperature inside the magnet core is only in the 15–20°C range. In general, the vital functions of the animals

162

P. Jakob

during anesthesia have to be carefully controlled. In addition, to minimize motion artifacts, ECG and respiratory gating or navigation is mandatory to obtain highquality time-resolved cardiac images.

12.9 MR Microscopy: Imaging Examples In the following section, we do not review all the possibilities that MRM offers, but instead highlight the state-of-the-art use of MRM in two examples which were chosen to demonstrate the major capabilities of in vivo MRM.

12.9.1 Tumor Characterization via MR Microscopy In general, the histopathological characterization of tumors requires the use of biopsy, which is an invasive surgical procedure. However, a biopsy does not always show the real grade of the tumor because of tumor Reference

T1 [ms]

muscle

heterogeneity. MRM in tumor models now offers the major advantage of noninvasively delivering detailed anatomical, functional, and physiological information in one imaging session. This information, which also can be obtained in a longitudinal fashion, can then be exploited for a significantly improved tumor characterization. The in vivo measurements presented here were performed on a human xenograft tumor model UT-SCC-14 (human squamous cell carcinoma from the mobile part of the tongue) (Melkus et al. 2008). The tumor cells were transplanted subcutaneously onto the lower leg of the outbred nude mice. The tumorbearing mice was investigated with a multiparameter MR-protocol at a 17.6 T MRM system when the tumor was approximately 7 mm in diameter. This multiparameter protocol was only timely feasible, because the SNR provided by the 17.6 T system was sufficient to keep the imaging times for the individual MR sequences at a tolerable level (see also (12.3)). Figure 12.5a shows an axial T2-weighted reference MR image of the mouse’s lower leg, depicting the “big sized” tumor next to the mouse’s leg muscle. Figure 12.5b, c show corresponding T1 and T2 parameter maps of the same slice. These color-coded parameter maps provide quantitatively the local relaxation times on a pixel-by-pixel 2000

T2 [ms]

100 80

1500

60 1000 40 500 tumor

20 0

0 −3

Perfusion [% muscle]

400 ADC [mm/s2]

x 10

Lactate [a.u.]

350 300

1.5

250 200

1

150 100

0.5

50 0

0

Fig. 12.5 Tumor characterization via MR microscopy. (a) Structural imaging of a tumor-bearing mouse leg. Corresponding (b) T1-map, (c) T2-map, (d) perfusion map, (e) diffusion map, (f) lactat map of the same slice, for more details please see text

12 Small Animal Magnetic Resonance Imaging: Basic Principles, Instrumentation and Practical Issue

basis and thus high contrast parameter images, which allow for an improved tissue characterization in vivo. The tumor shows increased T1 and T2 relaxation times compared to the muscle tissue. Figure 12.5d provides spatially resolved perfusion information obtained from a so-called spin labeling experiment, demonstrating that the perfusion of the tumor tissue is almost equal to the muscle tissue. Diffusion, another quantity which can be measured in vivo by MRM (Cercignani and Horsfield 2001; Stejskal and Tanner 1965), is defined as the process of random molecular thermal motion occurring at a microscopic scale. Figure 12.5e depicts a quantitative diffusion map, yielding the apparent diffusion coefficient (ADC) on a pixel-by-pixel basis. A close look at this parameter map tells us that the ADC inside the tumor tissue is significantly reduced compared to the muscle tissue. The aforementioned imaging protocols typically employed a field-of-view of 16 × 16 mm2, a matrix size of 128 × 128, and a slice thickness of 0.5–1 mm. With these imaging parameters, an in-plane spatial resolution of 125 × 125 mm2 within a total imaging time of 8–16 min for each protocol was achieved. Finally, Fig. 12.5f shows the result from a magnetic resonance spectroscopic imaging (MRSI) experiment (Melkus et al. 2008). MRSI is a powerful tool which provides a spatially resolved manner-specific chemical information about the biochemistry of low-molecular-mass metabolites and polypeptides. In this particular application the distribution of lactat, which is an end product of glycolysis and which increases rapidly during hypoxia and ischemia, is shown. The lactat map demonstrates significantly increased lactat concentrations in the upper portion of the tumor, which is a strong hint for tumor necrosis. In summary, this imaging example was chosen in order to demonstrate the plethora of information as well as the quantitative nature of MRM, which can be exploited for better tissue characterization in animal models.

12.9.2 MR Microscopy in Fixed Specimens MRM with the ability of full volume encoding allows to view the complex anatomy and has raised the possibility for studying fixed specimens, e.g., one can view an entire embryo from different perspectives with arbitrary view orientations without physical destruction of the specimen. The example shown in Fig. 12.6 depicts one sagital slice from a 3D data set of a fixed

163

Fig. 12.6 One sagital slice from a 3D data set of a fixed mouse embryo obtained within a total imaging time of 4h

mouse embryo. This 3D data set was obtained with a spatial resolution of 20 × 39 × 39 mm3 within a total imaging time of approximately 4 h employing signal averaging for SNR improvement. This kind of spatial resolution would have been impossible in a living rodent because of physiological constraints. This imaging example was chosen as a teaching example in order to demonstrate one fundamental limit of MRM expressed quantitatively in (12.3): high spatial resolution requires long data acquisition times, which fundamentally limit the spatial resolution that can be achieved in living rodents.

12.10 Summary MRM has now advanced to a stage where it can be practically applied to questions of fundamental importance to basic scientists. Today, MRM allows

164

to follow serially through disease progression and regression in the same animal over extended time periods. The foregoing introduction to MR and MRI lays the groundwork to appreciate the possibilities and challenges encountered when performing MRM in rodents, which are further explored in the following book sections. Acknowledgment This work was supported by the Deutsche Forschungsgemeinschaft with grants Ad 132/3-1, Ba 1433/4-1, Fl 225/2-1 and Mu 576/14-1 and the Graduiertenkolleg 1048 “Organogenesis.” The author would like to thank Esra Lang, Gerd Melkus, Philipp Mörchel, Simon Triphan for contributing to Figs. 12.5 and 12.6.

References Abragam A (1961) Principles of nuclear magnetism. Clarendon, Oxford, pp 1–599 Benveniste H, Blackband S (2002) MR microscopy and high resolution small animal MRI: applications in neuroscience research. Prog Neurobiol 67:393–420 Brockmann MA, Kemmling A, Groden C (2007) Current issues and perspectives in small rodent magnetic resonance imaging using clinical MRI scanners. Methods 43:79–87 Bulte JW, Kraitchman DL (2004) Iron oxide MR contrast agents for molecular and cellular imaging. NMR Biomed 17:484–499 Callaghan PT (1991) Principles of magnetic resonance microscopy. Oxford University Press, Oxford Cercignani M, Horsfield MA (2001) The physical basis of diffusion weighted MRI. J Neurol Sci 186:11–14 Cowan B (1997) Nuclear magnetic resonance and relaxation. Cambridge University Press, Cambridge Dazai J, Bock NA, Nieman BJ, Lorinda M, Davidson R, Henkelman M, Chen XJ (2004) Multiple mouse biological loading and monitoring system for MRI. Magn Res Med 52:709–715 Eccles CD, Callaghan PT (1986) High resolution imaging: the NMR microscope. J Magn Reson 68:393–398 Ernst RR, Bodenhausen G, Wokaun A (1986) Principles of nuclear magnetic resonance in one and two dimensions. Clarendon, Oxford

P. Jakob Gadian DG (1982) Nuclear magnetic resonance and its applications to living systems. Clarendon, Oxford Gareis D, Wichmann T, Lanz T, Melkus G, Horn M, Jakob PM (2007) Mouse MRI using phased-array coils. NMR Biomed 20(3):326–334 Haacke EM, Brown RW, Thompson MR, Venkatesan R (1999) Magnetic resonance imaging: physical principles and sequence design. Wiley-Liss, New York Haase A, Odoj F, von Kienlin M, Warnking J, Fidler F, Weisser A, Nittka M, Rommel E, Lanz T, Kalusche B, Griswold M (2000) NMR probeheads for in vivo applications. Concepts Magn Res 12(6):361–388 Hendrick RE, Haacke EM (1993) Basic physics of MR contrast agents and maximization of image contrast. JMRI 3: 137–148 Lauffer RB (1987) Paramagnetic metal complexes as water proton relaxation agents for NMR imaging: theory and design. Chem Rev 87(5):901–927 Lauterbur PC (1973) Image formation by induced local interactions – examples employing nuclear magnetic resonance. Nature 242:190–191 Mansfield P (1977) Multi-planar image formation using NMR spin echoes. J Phys C 10:L55–L58 Mansfield P, Grannell PK (1975) Diffraction in microscopy in solids and liquids by NMR. Phys Rev B 12:3618 McConnell J (1987) The theory of nuclear magnetic relaxation in liquids. Cambridge University Press, New York, pp 1–196 Melkus G, Mörchel P, Behr VC, Kotas M, Flentje M, Jakob PM (2008) Short-echo spectroscopic imaging combined with lactate editing in a single scan. NMR Biomed 21: 1076–1086 Morris PG (1986) Nuclear magnetic resonance imaging in medicine and biology. Clarendon, Oxford Muller R (1996) Contrast agents in whole body magnetic resonance: operating mechanism. In: Encylopedia of nuclear magnetic resonance, D. M. Grant and R. K. Harris, vol 3. Wiley, New York Stejskal EO, Tanner JE (1965) Spin diffusion measurements: spin echoes in the presence of a time-dependent field gradient. J Chem Phys 42:288–292 Wehrli FW, Shaw D, Kneeland JB (1988) Biomedical magnetic resonance imaging. Weinheim, VCH Wood ML, Hardy PA (1993) Proton relaxation enhancement. JMRI 3:149–156

MR Contrast Agents

13

Eliana Gianolio, Alessandra Viale, Daniela Delli Castelli, and Silvio Aime

Contents

13.1 Introduction

13.1 Introduction........................................................... 165

The superb spatial resolution and the outstanding capacity of differentiating soft tissues have determined the widespread success of magnetic resonance imaging (MRI) in clinical diagnosis (Rinck 2003). The contrast in an MR image is the result of a complex interplay of numerous factors, including the relative T1 and T2 relaxation times, proton density of the imaged tissues, and instrumental parameters. The MR image contrast can be further enhanced by administration of suitable MRI contrast agents (CAs). Unlike CAs used in X-ray computed tomography and in nuclear medicine, the currently used MRI CAs are not directly visualized in the image. Only their effects are observed as the contrast is affected by the variation that the CA causes on water protons relaxation times and consequently on the intensity of the NMR signal (Brücher and Sherry 2001). Generally, the purpose of a CA is to reduce T1 or T2 in order to obtain a hyper or hypointense signal in short times and a better signal-to-noise ratio with the acquisition of a higher number of measurements. CAs that predominantly reduce T1 are called positive, whereas those that mainly affect T2 are called negative. The search for positive MRI CAs was oriented towards paramagnetic metal complexes because unpaired electrons display remarkable ability to reduce T1 and T2 of their solutions, (Engelstadt and Wolf 1988; Rinck 1993), whereas iron oxide particles are mostly used as negative agents. Another class of CAs is represented by CEST agents. They act through the transfer of saturated magnetization to the “bulk” water signal. Such a saturation transfer (ST) leads to a decrease of the MR signal intensity, and therefore, these species too act as negative agents.

13.2 T1–Paramagnetic Mn- and Gd-Based Relaxation Agents................................................. 166 13.2.1 Theory of Paramagnetic Relaxation....................... 167 13.2.2 Blood Pool Agents.................................................. 169 13.2.3 Gd-Based Probes for Molecular Imaging............... 171 13.2.4 Gd-Complexes as Responsive Agents.................... 172 13.3 T2–Susceptibility Agents....................................... 174 13.3.1 Iron Oxide Particles................................................ 174 13.3.2 Paramagnetic Particles............................................ 175 13.4 CEST Agents......................................................... 176 13.4.1 DiaCEST................................................................. 177 13.4.2 ParaCEST................................................................ 178 13.4.3 LipoCEST............................................................... 180 13.5 Hyperpolarized Molecules................................... 182 13.5.1 Hyperpolarization Techniques................................ 183 13.6 19F-containing Species........................................... 188 13.7 Conclusions............................................................ 190 References............................................................................ 190

E. Gianolio, A. Viale, D. Delli. Castelli, and S. Aime (*) Department of Chemistry IFM and Molecular Imaging Center, University of Torino, Via.P. Giuria 7, 10125 Torino, Italy e-mail: [email protected]

F. Kiessling and B.J. Pichler (eds.), Small Animal Imaging, DOI: 10.1007/978-3-642-12945-2_13, © Springer-Verlag Berlin Heidelberg 2011

165

166

E. Gianolio et al. Paramagnetic Metal Chelates Iron Oxide Particles

CEST Agents

MRI Contrast Agents

F-19 Containing Molecules

Hy erp Hyperpolarized Molecules

Fig. 13.1 Classes of MRI contrast agents

In recent years, much attention has been devoted to chemicals that are sensitive enough to be directly visualized in MR images. For instance, 19F-containing systems have been proposed. The required local concentration is of course much higher than that one used in the case of paramagnetic agents. Nevertheless it has been shown that aggregates loaded with a large number of fluorine-containing molecules may yield 19 F-MR images. One advantage deals with the possibility of exploiting different chemical shifts of the F-19 resonances, thus allowing the detection of more than one agent in the same image. Another emerging class of frequency-encoded agents is represented by 13C-hyperpolarized molecules. These molecules display a high sensitivity that allows their direct visualization in an MR image; moreover, it has been possible to acquire images of the hyperpolarized products of their metabolism (metabolic imaging). The use of these hyperpolarized systems opened a new horizon in the study of abnormalities of cellular metabolism and, if successfully translated to clinics, represents a real breakthrough in the field of medical imaging (Fig. 13.1).

13.2 T1–Paramagnetic Mn- and Gd-Based Relaxation Agents Paramagnetic substances are under intense scrutiny as MRI CAs since the early days of NMR tomography. Although stable organic radicals, such as nitroxides and O2, were also considered, it was immediately clear

that paramagnetic metal complexes are the candidates of choice for this application. On this basis, the metal ions more suitable for this application have been identified among the ones having the highest number of unpaired electrons, namely Mn(II) and Fe(III) (5 unpaired electrons) in the transition metals series and Gd(III) (7 unpaired electrons) among the Lanthanides. Currently, several Gd-based and one Mn-based agents are available for clinical applications. Their use has led to remarkable improvements in medical diagnosis in terms of higher specificity, better tissue characterization, reduction of image artifacts, and functional information. Besides acting as catalyst for the relaxation of water protons, a paramagnetic MRI CA has to possess several additional properties in order to guarantee the safety issues required for “in vivo” applications at the administered doses, namely high thermodynamic (and possibly kinetic) stability, good solubility, and low osmolality (Brücher and Sherry 2001; Tweedle 1989). The first CAs approved for clinical use was Gd-DTPA (Magnevist®, Schering AG, Germany) that, in more than ten years of clinical experimentation, has been administered to many millions of patients. Other Gd(III)-based CA similar to Magnevist® became soon available: Gd-DOTA (Dotarem®, Guerbert SA, France), Gd-DTPA-BMA (Omniscan®, GE Health, USA), and Gd-HPDO3A (Prohance®, Bracco Imaging, Italy) (Weinmann et al.2000). These CAs have very similar pharmacokinetic properties because they distribute in the extracellular fluid and are eliminated via glomerular filtration. In neurological diseases, they are particularly useful to delineate lesions as a result of the disruption of the blood–brain barrier. Two derivatives of Gd-DTPA have been successively introduced, Gd-EOB-DTPA (Schmitt-Willich et al. 1999) (Eovist®, Schering AG, Germany) and Gd-BOPTA (Uggeri et al. 1995) (Multihance®, Bracco Imaging, Italy). They are characterized by an increased lipophilicity due to the introduction of an aromatic substituent on the carbon backbone of the DTPA ligand. This modification significantly alters the pharmacokinetics and the biodistribution of these CAs as compared to the parent Gd-DTPA making them hepatospecific agents (Fig. 13.2). Paramagnetic chelates of Mn(II) (five unpaired electrons) have also been considered. The main drawback appears to be related to the stability of these complexes. Mn(II) is an essential metal; therefore, the

167

13 MR Contrast Agents Fig. 13.2 Structures of the Gd(III)-based MRI contrast agents currently used in the clinical practice

-OOC

COO-

COON

N Gd3+ N N

-OOC

-OOC

COO-

N

N

Gd3+

-OOC

COO-

N

COO-

Gd-DOTA

Gd-DTPA

DOTAREM®

MAGNEVIST ®

COO-OOC

COON

N Gd3+ N N

-OOC OH

-OOC

N

N

Gd3+

H3CHNOC

COO-

N

CONHCH3

Gd-DTPA-BMA OMNISCAN ®

Gd-HPDO3A PROHANCE ®

O

O -OOC -OOC

COO-

N

N Gd3+ Gd-BOPTA MULTIHANCE®

evolution has selected biological structures that are able to sequester Mn(II) ions with high efficiency. Combined with the fact that Mn(II) forms highly labile coordination complexes, it has been difficult to design Mn(II) chelates that maintain their integrity when administered to living organisms. Actually, MnDPDP has entered the clinical practice and is recommended as a hepatotropic agent (Rummeny 1997). It is the only agent that does its job by releasing metal ions to endogenous macromolecules. The huge proton relaxation enhancement brought about by the resulting Mn(II) protein adducts is responsible for the MRI visualization of hepatocytes also at low administered doses of MnDPDP (Fig. 13.3).

COO-

N

COOCOO-

-OOC

N

N

N

COO-

Gd3+ COO-

-OOC Gd-EOB-DTPA EOVIST ®

13.2.1 Theory of Paramagnetic Relaxation The efficiency of a relaxation agent is commonly evaluated in vitro by the measure of its relaxivity (r1) that, for commercial CAs as Magnevist, Dotarem, Prohance, and Omniscan, is around 3.4–3.5 mM−1 s−1 (at 20 MHz and 39°C). It represents the relaxation enhancement of water protons in the presence of the paramagnetic complex at 1 mM concentration. The observed longitudinal relaxation rate (R1obs) of the water protons in an aqueous solution containing the paramagnetic complex is the sum of three contributions (13.1) (Banci et al. 1991): (1) the diamagnetic one (R1o), whose value

168

E. Gianolio et al.

COO-

-OOC = O PO 3

N

N Mn2+

OPO = 3

OHHO

Mn-DPDP TESLASCAN®

Fig. 13.3 Structure of the Mn(II)-based MRI contrast agent used in the clinical practice

corresponds to proton relaxation rate measured in the presence of a diamagnetic (La, Lu, Y) complex of the same ligand; (2) the paramagnetic one, relative to the exchange of water molecules from the inner coordination sphere of the metal ion with bulk water (R1pis); and (3) the paramagnetic one, relative to the contribution of water molecules that diffuse in the outer coordination sphere of the paramagnetic center (R1pos). Sometimes, also a fourth paramagnetic contribution is taken into account that is due to the presence of mobile protons or water molecules (normally bound to the chelate through hydrogen bonds) in the second coordination sphere of the metal ion (Botta 2000). R1obs = R1o + R1pis + R1pos .

(13.1)

The inner sphere contribution is directly proportional to the molar concentration of the paramagnetic complex [C], to the number of water molecules coordinated to the paramagnetic center, q, and inversely proportional to the sum of the mean residence lifetime, tm, of the coordinated water protons and their relaxation time, T1M (13.2). R1pis =

q[C ] . 55.5(T1M + t m )

(13.2)

The latter parameter is inversely proportional to the sixth power of the distance between the metal center and the coordinated water protons (r) and depends from the molecular reorientational time, tR, of the chelate, from the electronic relaxation times, TiE (i = 1,2), of the unpaired electrons of the metal (which depend on the applied magnetic field strength) and from the applied magnetic field strength itself (13.3 and 13.4). 2

3τ c ö 1 2 æ µ ö γ 2 g 2 µ 2 S (S + 1) æ 7τ c = ç ο ÷ I e rB6 çè 1 + ω 2τ 2 + 1 + ω 2τ 2 ÷ø T1M 15 è 4π ø H S c I c

(13.3)

τ c-1 = t -R1 + t -m1 + Ti -E1 .

(13.4)

The outer sphere contribution depends on TiE, on the distance of maximum approach between the solvent and the paramagnetic solute, on the relative diffusion coefficients, and again on the magnetic field strength. (Aime et al. 2003b) The dependence of R1pis and R1pos on magnetic field is very important because the analysis of the magnetic field dependence allows the determination of the principal parameters characterizing the relaxivity of a paramagnetic chelate. This information can be obtained through an NMR instrument in which the magnetic field is changed (Field-Cycling Relaxometer) to obtain the measure of r1 on a wide range of frequencies (typically 0.01–50 MHz). At the frequencies most commonly used in commercial tomographs (20–63 MHz), tR of the chelate is the most important determinant of the observed relaxivity. A quantitative analysis of r1 dependence on the different structural and dynamic parameters shows that, for systems with long tR (e.g., protein-bound complex), the maximum attainable r1 values can be achieved through the optimization of tM and TiE (Banci et al. 1991). All the available commercial CAs are monohydrated (q = 1) systems with a molecular weight of ca. 600– 800 Da that yields to rotational correlation times tR of about 60–80 ps. For the class of polyaminocarboxylate Gd(III) complexes, the exchange lifetime tM is typically found to be in the range of few hundreds of ns and T1E » 1 ns at 0.5 T, and thus, the inner sphere relaxivity, ripIS, assumes a value of ca. 2.5–3.5 mM−1 s−1, at 25°C. Therefore, as it was early recognized, it is evident that at 0.5 T, the overall correlation time is largely dominated by the rotational correlation time, whereas the contribution of both the exchange lifetime and the electronic relaxation plays an almost negligible role. An important structural parameter that influences the inner sphere relaxivity is the hydration number q. This represents a scaling factor in the equation that defines inner sphere relaxivity, and a higher number of coordinated water molecules (q > 1) provide a clear advantage in terms of efficiency. The use of hepta or hexa-dentate ligands would in principle result in Gd(III) complexes with 2 and 3 coordinated water molecules, respectively, but the decrease of the denticity of the ligands is likely to be accompanied by a decrease of their thermodynamic stability and an increase of their toxicity. Furthermore, systems with q = 2 may suffer a “quenching” effect upon interacting with endogenous anions or with proteins, as donor

169

13 MR Contrast Agents

atoms from lactate or Asp or Glu residues may replace the coordinated water molecules (Aime et al. 1998, Aime et al. 2000).

13.2.2 Blood Pool Agents Several systems have been studied over the last two decades for the design of macromolecular Gd(III)complexes as MRI blood pool CAs. These paramagnetic macromolecules do not diffuse across healthy vasculature and remain intravascular, thus reporting on the anatomy of the vessels bed (Fig. 13.4). Behind a higher vascular retention time, macromolecular systems are often endowed with sensibly higher relaxivities (at 0.5–1.5 T), thanks to the elongation of reorientational correlation times of slowly tumbling systems. The most straightforward method to increase the plasma circulation time and promote vascular confinement of a contrast agent is to covalently bind it to biomacromolecules such as proteins, polysaccharides, and dextrans. An alternative to covalent binding has been pursued by exploiting the noncovalent reversible interaction between low molecular weight hydrophobic Gd(III)complexes and proteins.

Pre contrast

Human serum albumin (HSA) has been by far the most investigated protein for binding of Gd(III) chelates. Besides the attainment of high relaxivities, a high binding affinity to HSA enables the Gd(III) chelate with a long intravascular retention time, which is the property required for a good blood pool agent for MR angiography. In blood, HSA has a concentration of about 0.6 mM, and its main physiological role deals with the transport of a huge number of substrates (Carter and Ho 1994). In Table 13.1, a list of Gd(III)-complexes and the relevant parameters of their binding interaction to HSA are reported. Although the theory of the paramagnetic relaxation foresees relaxivities up to 100–120 mM−1s−1 (at 20 MHz) for complexes bound to macromolecular systems, the data listed in Table 13.1 show r1b values significantly lower than the predicted ones. The primary reason for the quenching of the relaxation enhancement is often associated to the occurrence of a long exchange lifetime, tM, of the coordinated water (Aime et al. 1999a). Recently, Caravan and coworkers (Zhang et al. 2005) tackled the problem of the residual mobility of a Gd(III) complex bound to HSA by designing a system containing two anchoring sites on the protein. Interestingly, the observed relaxivity for such adduct, though high (60 mM−1s−1), is still significantly lower than that foreseen by the paramagnetic relaxation theory.

5 min after 0.1 mmol/kg i.v. of extracellular CA

5 min after 0.015 mmol/kg i.v. of angiographic ca

Fig. 13.4 3D contrast-enhanced MRA of the Rat Head at 2 T. Frontal Maximum Intensity Projection GRE 39/8/60°/1 (TR/TE, ms/a/NEX)

170

E. Gianolio et al.

Table 13.1 Affinity constant and relaxivity of the adducts of some Gd(III) complexes whit HSA, calculated at 0.47 T and 298 K. The coordinated water exchange life time of free complexes is also reported Metal complex n*KA(M−1) R1b(mM−1s−1) Reference tM (ns) [Gd-BOPTA (H2O)]2−

O

33.0

280

Aime et al. (2001)

2.0 × 104

44.0

80

Aime et al. (1996)

9.0 × 104

37.2

250

Lauffer et al. (1998)

COON

N

-OOC

4.0 × 102

-OOC

COO-

N

Gd3+

COO-

[Gd-DOTA(BOM)3(H2O)]−

O

O -OOC

COON

N Gd3+ N N

-OOC

COO-

O

MS-325

Aime et al. (1999a)

-O P O O

O -OOC N

-OOC

N

-OOC

COO -

N

Gd3+

COO -

B22956

2.6 × 104 OH

O N H

H

-OOC -OOC -OOC

N

N Gd3+

N

COOCOO-

COO-

27.6

250

Aime et al. (1999a)

171

13 MR Contrast Agents -OOC

COON

H3C-(H2C)16

Gd3 + N

COO-

N COO-

Gd-AAZTAC17 Fig. 13.5 Structure of the lipophilic Gd-AAZTAC17 complex

On our side, a derivative of Gd-AAZTA containing a long aliphatic chain (Gd-AAZTAC17 – Fig. 13.5) has been synthesized and its relaxometric properties investigated in detail (Gianolio et al. 2007). The complex showed to have the outstanding properties of the parent complex, namely: (1) two inner sphere water molecules in fast exchange with the bulk, (2) high thermodynamic stability in aqueous solution, and (3) a nearly complete inertness towards the influence of bidentate endogenous anions. Already at submillimolar concentrations (cmc 0.1 mM), the presence of the hydrophobic chain induces the formation of micelles, with a relaxivity (per Gd(III) ion) of 30 mM−1s−1 at 20 MHz and 298 K. At concentration lower than cmc, Gd-AATZA-C17 displays a high affinity binding to HSA Ka = 2.4 × 104 M−1 giving a macromolecular adduct endowed with the higher relaxivity value (84 mM−1s−1) up to now reported for HSA-bound Gd-complexes.

13.2.3 Gd-Based Probes for Molecular Imaging Molecular Imaging (Weissleder and Mahmood 2001) deals with in vivo characterization and measurement of biological processes at the cellular and molecular level. With Molecular Imaging, early diagnosis of disease will become possible as the detection of altered biochemical process largely anticipates the anatomical changes that are at the basis of current diagnostic modalities. MRI-Gd(III)-based agents are much less sensitive than radionuclear and optical imaging probes. Therefore, MR-Molecular Imaging invariantly involves the need of accumulating a high number of contrast enhancing units at the site of interest. The basis for the design of a Gd-containing imaging probe is first dictated from the

concentration and localization (vascular, extracellular matrix, on the cellular membrane, intracellular) of the target molecule (Aime et al. 2002a). Of course, the most accessible targets are those present on the surface of endothelial vessels. In principle, they can be visualized by a number of macromolecular conjugates containing many Gd(III) complexes endowed with the proper vector recognizing the given target. A nice example of targeting an endothelial site has been reported some years ago by Sipkins et al. (1998) in the targeting of a specific angiogenesis marker, the endothelial integrin avb3, whose presence has been shown to correlate with tumor grade. The imaging probe used in this work is a Gd-containing polymerized liposome. The target is first recognized by a biotinilated antibody against avb3, which is then the site of binding for an avidin moiety on the surface of the liposome. Each liposome has a mean diameter of 300–350 nm, which appears suitable to avoid the uptake by the reticuloendothelial system (RES) (Fig. 13.6). This approach provided enhanced and detailed detection of rabbit carcinoma through the imaging of the angiogenetic vasculature. The same avb3 target has been addressed by means of lipidic nanoparticles containing a huge number of Gd-chelated units (94.400 Gd/particle characterized by r1 = 19.1 s−1mM−1 (per Gd) and r1 = 1.800.000 per particle). The targeting has been pursued by introducing an avb3 integrin peptidomimetic antagonist on the surface of the lipid nanoparticle (Winter et al. 2003). In a recent work, avb3 epitopes have been targeted by quantum dot-based nanoparticles coated with a paramagnetic micellar shell (Fig. 13.7): the use of such dual probe allows visualizing ongoing tumor angiogenesis by using fluorescence imaging and MRI Gd Monoclonal Antibody Biotin

Gd

Gd Liposome

Avidin Gd

Gd Gd

Gd Endothelial ανβ3 receptor

Fig. 13.6 Targeting of the endothelial integrin avb3 as a specific angiogenesis marker. The receptor is recognized by a biotinylated antibody that is then bound by an avidin moiety bearing a Gd(III)-loaded liposome

172

E. Gianolio et al.

RGD peptide

Paramagnetic lipid

Pegylated lipid

Quantum dot

Fig. 13.7 The system consists of a quantum dot core covered by a micellar shell composed of pegylated phospholipid (PEG-DSPE) and Gd-DTPA-based lipids (Gd-DTPABSA). RGD peptides are conjugated to the nanoparticle to provide specificity for the avb3 integrin. (Reprinted from Babincova et al. 2000 with permission from Springer)

(Mulder et al. 2009). Upon intravenous injection of RGD-pQDs in tumor-bearing mice, intravital microscopy allowed the detection of angiogenically activated endothelium at cellular resolution with a small scanning window and limited penetration depth, while MRI was used to visualize angiogenesis at anatomical resolution throughout the entire tumor. The large size of these constructs limits their delivery to targets on the endothelial walls. To target receptors in solid tissues, other routes have to be followed. Bhujwalla and coworkers (Artemov et al. 2003) have developed and applied a two-component Gd-based avidin–biotin system for the visualization of HER-2/ scan receptors. The latter is a member of the epidermal growth factor family, and it is amplified in multiple cancers. Their approach consisted of addressing the extracellular domain of the receptors by means of a biotinilated mAb. After clearance of the unbound mAb, Gd-labeled avidin is administered and binds, with high affinity, to the biotinilated mAb. The expression level of the receptor was estimated at 7 × 105 receptors/cell, and the average number of Gd-DTPA units per avidin molecule was 12.5. The method has been successfully applied in an experimental mouse model of breast carcinoma. An interesting route to MR signal amplification has been developed by Weissledder and coworkers

(Bogdanov et al. 2002). Their approach is based on enzyme-mediated polymerization of paramagnetic substrates into oligomers of higher relaxivity. As substrates, they used Gd-chelates functionalized with phenolic substituents, which undergo rapid condensation in the presence of H2O2 and peroxidase. The increased molecular size of the oligomeric structures causes an increase of the molecular reorientational time, which, in turn, results in an increase of the observed relaxivity. This approach has been applied to the imaging of E-selectin-peroxidase conjugate. As far as the cell internalization of Gd-chelates is concerned, several routes have been explored, namely pinocytosis, phagocytosis, receptors, receptor-mediated endocytosis, etc. (Aime et al. 2002a).

13.2.4 Gd-Complexes as Responsive Agents The term “responsive” refers to diagnostic agents whose contrasting properties are sensitive to a given physico-chemical variable that characterizes the microenvironment in which the probe distributes. Typical parameters of primary diagnostic relevance include pH, temperature, pO2, enzymatic activity, redox potential, and concentration of specific ions and lowweight metabolites. Unfortunately, their clinical use is still uncertain mainly because an accurate measurement of one of the above-mentioned parameters with a given relaxing probe requires a precise knowledge of the local concentration of the contrast medium in the region of interest. Only if the actual concentration is known, the observed change in the relaxation rate of water protons can be safely attributed to a change of the parameter to be measured.

13.2.4.1 pH Sensitive The design of a Gd(III)-based complex whose relaxivity is pH-dependent requires that at least one of the structural or dynamic parameters determining its relaxivity is made pH-dependent. In most of the examples so far reported, the pH dependence of the relaxivity reflects changes in the hydration of the metal complex.

173

13 MR Contrast Agents

O

8

Basic pH q=0 O

O

6

O O

O N

N O

Gd

O N

N

NHR

H2O

CH2

O

Gd

O

N

N O

O

N

N

r1(mM−1s−1)

Acid pH q=2

4

2

N R

3

4

5

6

7

8

9

10

11

12

pH

Fig. 13.8 pH dependence of the relaxivity (20 MHz, 25°C) of a Gd(III) complex bearing an arylsulfonamide group. The protonation/deprotonation process of the arylsulfonamide group determines a variation in the hydration of the complex

For instance, we found that the relaxivity of a series of macrocyclic Gd(III) complexes bearing b-arylsulfonamide groups is markedly pH-dependent (Fig. 13.8) on passing from about 8 s−1mM−1 at pH <4 to ca. 2.2 s−1mM−1 at pH >8 (Lowe et al. 2001). It has been demonstrated that the observed decrease (about fourfold) of r1 is the result of a switch in the number of water molecules coordinated to the Gd(III) ion from 2 (at low pH values) to 0 (at basic pHs). This corresponds to a change in the coordination ability of the b-arylsulfonamide arm that binds the metal ion only when it is in the deprotonated form. In some cases, the pH dependence of the relaxivity is associated with changes in the structure of the second hydration shell. Two such systems have been reported by Sherry’s group. The first case deals with a macrocyclic tetramide derivative of DOTA (DOTA4AmP) that possesses an unusual r1 vs. pH dependence (Zhang et al. 1999). The relaxivity of this complex increases from pH 4 to pH 6, decreases up to pH 8.5, remains constant up to pH 10.5, and, then, increases again. The authors suggested that this behavior is related to the formation/disruption of the hydrogen bond network between the pendant phosphonate groups and the water bound to the Gd(III) ion. The deprotonation of phosphonate occurring at pH >4 promotes the formation of the hydrogen bond network that slows down the exchange of the metalbound water protons. On the contrary, the behavior observed at pH >10.5 was accounted for in terms of a shortening of tM catalyzed by OH− ions. It has been demonstrated that this complex can be successfully used “in vivo” for mapping renal and systemic pH (Raghunand et al. 2003).

pH-dependent probes can also be obtained when the proton concentration is able to affect the structure of a macromolecular substrate that, in turn, results in changes of its dynamic properties. An interesting example is represented by a macromolecular Gd(III) construct formed by 30 Gd(III) units covalently linked, by a squaric acid moiety, to a polyornithine (114 residues) (Aime et al. 1999b). At acidic pH, the unreacted amino groups of the polymer are protonated and, therefore, tend to be localized as far apart as possible, whereas at basic pH, the progressive deprotonation of the NH3+ groups determines an overall rigidification of the polymer structure owing to the formation of intramolecular hydrogen bonds between adjacent peptidic linkages. As expected, the reduced rotational mobility of the polymeric backbone upon increasing pH enhances the relaxivity of the system. Nevertheless, even if the enhancement is not particularly remarkable (ca. 40% in the 3–8 pH interval), it is worth to note that the relaxivity of this system is considerably higher than the previous examples over the entire pH range, thus allowing, in principle, the detection of pH changes at lower concentration of the responsive probe. Moreover, it has successively been shown that the pH assessment with this system can be made independent of the knowledge of its actual concentration through a ratiometric approach that relates T1 and T2 changes to the pH of the solution.

13.2.4.2 Agents Sensitive to the Redox Potential A diagnostic MRI probe sensitive to the in vivo redox potential would be very useful for detecting regions

174

with a reduced oxygen partial pressure (PO2), in several pathologies including strokes and tumors. Very few Gd(III) chelates sensitive to the tissue oxygenation have been so far reported. Our group has investigated the potential ability of GdDOTP to act as an allosteric effector of hemoglobin (Aime et al. 1995). In fact, it has been observed that this chelate binds specifically to the T-form of the protein that is characterized by a lower affinity towards oxygen. The interaction is driven by electrostatic forces and leads to a significant relaxivity enhancement (ca. fivefold) owing to the restricted molecular tumbling of the paramagnetic complex once it is bound to the protein. Although hemoglobin can be considered as an excellent indirect target for detecting PO2, the practical applicability of the method suffers for the inability of GdDOTP to enter red blood cells. Another approach deals with the use of liposomes built up with an amphiphilic Gd(III) complex containing a radical-sensitive disulfide bridge (Glogard et al. 2003). The relaxivity (at 20 MHz and 25°C) of the liposomal paramagnetic agent is 13.6 s−1mM−1, i.e., twofold higher than the free Gd(III) complex (r1 of 6.5 s−1mM−1). Likely, the limited relaxivity enhancement is due to the rotational flexibility of the complex bound to the liposome. This system was tested “in vitro” by inducing the cleavage of the S-S bond chemically (by dithiothreitol) or physically (by g-rays). In both cases, the relaxivity decreased from the value of the liposomal agents to that one of the free Gd(III) complex.

13.2.4.3 Enzyme Responsive One possible route to design enzyme responsive agents is to synthesize paramagnetic inhibitors, whose binding to the active site of the protein can be signaled by the consequent relaxivity enhancement. An example of this approach has been provided by Anelli et al., who synthesized a linear Gd(III)-complex bearing an arylsulfonamide moiety that is a well-known inhibitor of Carbonic Anhydrase (Anelli et al. 2000). In vitro experiments demonstrated that this complex binds quite strongly (KA of about 1.5 × 104) to the enzyme and its relaxivity in the bound form being about fivefold higher than the free complex (27 s−1 mM−1 vs. 5 s−1 mM−1 at 20 MHz). Unfortunately, an attempt to test the validity of this “in vivo” approach failed, likely owing to the small amount of enzyme circulating in the blood.

E. Gianolio et al.

An alternative approach is to design Gd(III) complexes acting as substrate for a specific enzyme. Along this direction, an example has been provided by Lauffer et al. who prepared a Gd(III) chelate containing a phosphoric ester sensitive to the attack of serum alkaline phosphatase (Lauffer et al. 1997). The hydrolysis yields the exposure of a hydrophobic moiety well suitable to bind to HSA. Upon binding, there is an increase of the relaxivity as a consequence of the lengthening of the molecular reorientational time. This approach was used by the same research group for designing Gd(III)complexes sensitive to TAFI (thrombin-activatable fibrinolysis inhibitor), a carboxypeptidase B involved in the clot degradation (Nivorozhkin et al. 2001).

13.3 T2-Susceptibility Agents 13.3.1 Iron Oxide Particles These materials consist of superparamagnetic iron oxide (SPIO) that is made of ferric (Fe3+) and ferrous (Fe2+) ions (Muller et al. 2005). Their peculiar magnetic behavior is associated with the occurrence of domains in their microcrystalline structure that alter markedly the characteristic properties of ferrimagnetism and ferromagnetism such as spontaneous magnetization, residual magnetization, and magnetic hysteresis. These particles (that per se are a kind of ferrite and, therefore, ferrimagnetic) are magnetically transformed into a paramagnetic substance. Inside each domain, the magnetic moments of the various ions sum up to yield large magnetic moments, in principle analogous to a single paramagnetic ion but of much larger strength, for which they are called superparamagnetic. The distortion of the magnetic field caused by the superparamagnetic core induces large changes in mag netic susceptibility which, in turn, leads to hypointensities in T2- or T2*-weighted images. Thus, the areas containing particles display fast transverse relaxation rates and low signal intensity (“negative contrast”). Due to the large magnetic susceptibility of an iron oxide particle, the signal void is much larger than the particle size, enhancing detectability at the expenses of resolution. These particles have been used to track “in vivo” different cell types (Cunningham et al. 2005),

175

13 MR Contrast Agents

2. In terms of relaxation enhancement properties, the larger magnetic susceptibility of SPIO yields to larger R2/R1 relaxation rates and higher T2-shortening effects than those brought about by the smaller USPIO particles.

Fig. 13.9 Schematic view of the composition of superparamagnetic iron oxide particles

including T lymphocytes (Kircher et al. 2003), macrophages (Engberink et al. 2007), and stem cells (Frangioni and Hajjar 2004). Iron oxide particles for MRI applications are classified into two classes, namely SPIO and ultrasmall superparamagnetic iron oxide (USPIO) on the basis of the overall size of the protective cover on the surface of the magnetic particle (Corot et al. 2006). In fact, in both types of particles, the iron oxide colloid particles are encapsulated by organic material such as dextran and carboxydextran to improve their compatibility with the biological systems (Fig. 13.9). The difference in the overall size of the particles causes marked changes in two important properties as far as their use as MRI CAs is concerned: 1. Upon intravenous administration, their blood pool lifetime is strongly affected by their dimensions. USPIO has longer half-life in the blood vessels than SPIO because their smaller size (<50 nm) makes their uptake from macrophages more difficult. On the contrary, SPIO (with diameters of the order of 100–200 nm) is very rapidly removed from the circulation by the RES. For instance 70–80% of injected AMI-25 particles (size 57–250 nm) are taken up by liver, 6% by the spleen, and a small amount by bone marrow (Schumann Giamperi 1993). USPIO remains in blood for longer time and more readily accumulates in the lymph nodes and bone marrow.

SPIOs and USPIOs (and also MPIOs, the micron-sized iron oxide particles) are under intense scrutiny as tools for cellular labeling for tracking cells’ migration, in particular into the nervous system. SPIOs appear to work better than USPIOs in labeling experiments first of all because the phagocytic uptake increases with the particle size. Cells incubated with SPIOs result in an increase of R2 upon increasing the local cellular concentration. Typically, R2 of monocytes labeled with SPIO at the concentration of 1.0 mg Fe/mL is 13.1 s−1. This has lead to a detection limit of 58 labeled monocytes per a voxel volume of 0.05mL (Engberink et al. 2007). Although SPIOs are more suitable for in vitro cellular labeling, USPIOs are frequently preferred for in vivo studies, thanks to their longer blood pool half lifetime. Thus, in vivo labeling by USPIO has been pursued in several diseased states, namely stroke (Saleh et al. 2004) and multiple sclerosis (Dousset et al. 2006). In humans, the lifetime of blood pool is more than 34 h whereas the t1/2 of SPIO appears to be shorter than 6 min, and such a half lifetime markedly limits their possibility to label endogenous monocytes in circulation.

13.3.2 Paramagnetic Particles In the late 1980s, it was reported that paramagnetic low molecular weight Dy(III)-complexes can act as T2-susceptibility agents in MRI images when they are unequally distributed in vessels and in the surrounding tissues (Rosen et al. 1991). The effect can be further enhanced when the paramagnetic complexes are entrapped in vesicles such as liposomes (Fossheim et al. 1997). The observed behavior is well accounted for in terms of the field gradients created by the compartment containing the paramagnetic ions that induce the spin dephasing of the water protons diffusing in the outer region of the compartment. It is straightforward to note that any nanosized system containing paramagnetic metal ions would act as a T2-susceptibility agent.

176

E. Gianolio et al.

In this context, paramagnetic liposomes have a high potential owing to the high payload of paramagnetic complexes that can be either entrapped in their inner cavities or, upon suitable functionalization with lipophilic substituents, be incorporated in their membrane bilayer (Terreno et al. 2007a). The use of paramagnetic liposomes has been demonstrated advantageous for developing highly sensitive T2-susceptibility agents as possible alternative to the well-established class of iron oxide nanoparticles. The compartmentalization effect, which is proportional to the magnetic field strength, is clearly illustrated in Fig. 13.10. The sensitivity of such systems is dependent on the intrinsic paramagnetism (described by the effective magnetic moment, meff) of the Ln ion (Dy(III) is the most effective one) and on the overall concentration of the paramagnetic centers compartmentalized in the vesicle. Hence, the incorporation of amphiphilic Dy(III) complexes in the liposome bilayer, in addition to the encapsulation of huge amounts of a hydrophilic Dy(III) agent in the aqueous cavity, yields to a marked sensitivity enhancement that makes these systems interesting agents for high field applications. Another class of liposome-based T2-susceptibility agents is represented by magnetoliposomes, in which iron oxide particles are encapsulated in liposomes (Bulte and De Cuyper 2003). Upon their interaction with externally applied magnetic fields, these systems have been exploited in novel magnetic targeting experiments (Babincova et al. 2000), as well as drug delivery carriers (Babincova et al. 2004), and in hyperthermia therapeutic (Ito et al. 2005). As far as their applications in MR-Molecular Imaging is concerned,

magnetoliposomes have been successfully used as blood pool agents (Martina et al. 2005), for targeting bone marrow (Bulte et al. 1999), and solid tumors (Fortin-Ripoche et al. 2006). Another class of very promising paramagnetic nanoparticles to be used as T2-susceptibility MRI CAs is represented by Lanthanide oxides particles. In a recent study, Peters and coworkers (Norek et al. 2008) reported the characterization of Dy2O3 dextran-coated particles endowed with very high transverse relaxivities. They found the relaxivity to be dependent on the particle radius being the optimal value around 70 nm. Under those conditions, the r2 value is about 190 s−1 mM−1 at 7 T, and 25°C, which is comparable with the relaxivity of, for example, PEGylated magnetoliposomes (r2 = 240 s−1 mM−1) or protein-coated magnetoferritin (r2 = 218 s−1 mM−1) at B = 1.5 T.35 The efficiency is even higher at higher magnetic fields, as the proton relaxivity r2 in the presence of Dy2O3 nanoparticles is increasing with B2, in contrast to superparamagnetic entities, which, due to their constant magnetic moment, produce no r2 enhancement at B > 1.5 T.

Fig. 13.10 Left: schematic view of the interactions between the local magnetic fields generated by a paramagnetic nanoparticle and the water molecules diffusing around it. Middle: effect of the

liposome encapsulation of a paramagnetic complex (Dy-HPDO3A) on the transverse relaxation rate at 7 T and 312 K. Right: In vitro RARE T2-weighted MR image obtained at 7 T and 312 K

13.4 CEST Agents The novel landscape of Molecular Imaging prompted the search for new paradigms in the design of MR-imaging reporters. A general insight deals with the exploitation of the frequency, the key parameter of the NMR phenomenon.

177

13 MR Contrast Agents

A convenient way to generate a “frequency- encoding” contrast is to exploit the well-known (either in NMR or in MRI fields) saturation transfer (ST) experiment (Henkelman et al. 2001). The basic rationale is to deal with a system containing at least one set of protons whose exchange with the bulk, kex, is slow on the NMR/MRI timescale, i.e., kex has to be smaller than the difference in the resonance frequency between the two exchanging pools (kex < Dw). Upon irradiation of the mobile protons by using a proper radiofrequency field (of intensity B2), saturated magnetization is transferred, via chemical exchange, to the bulk water resonance, which will decrease, the MRI signal in the image voxels containing the agent. The resulting contrast is usually called CEST (chemical exchange ST) contrast, and the same acronym is also used for defining the class of chemicals that are able to generate it. The basic CEST experiment is sketched in Fig. 13.11. With respect to the conventional MRI agents, CEST probes allow the design of MRI protocols in which the contrast can be generated “at will” only if the proper frequency corresponding to the exchangeable protons of the CEST agent is saturated. As a consequence, the precontrast image can be recorded almost simultaneously

E

CEST agent

b

to the postcontrast one as their acquisition simply differs from the on/off switch of the irradiation field. Even more important, this new approach offers the intriguing possibility of visualizing more than one agent in the same region, thus opening new exciting perspectives for the MRI applications in the biomedical field. In addition to this, CEST agents were very soon identified as candidates for designing concentrationindependent responsive probes. In fact, the presence of two magnetically nonequivalent “CEST-active” proton sites whose ST efficiency is differently dependent on the physico-chemical variable of interest allows the exploitation of a ratiometric approach that makes the MRI response independent of the concentration of the imaging probe (Ward and Balaban 2000).

13.4.1 DiaCEST The most critical issue for the translocation of CEST agents to “in vivo” applications is represented by their low sensitivity. Theoretically, the ST effect is dependent upon several parameters, among which, particularly E

Saturation Pulse B2

a

a

Chemical exchange E bulk water

b

a

Fig. 13.11 Schematic picture illustrating the ST process mediated by chemical exchange. The application of the saturation field B2 at the resonance frequency of the spins of the CEST agent equals their populations. The chemical exchange perturbs

b

Saturation transfer E

b

a

the Boltzmann equilibrium of the bulk water protons due to the transfer of an excess of CEST spins from the high energy level. The CEST contrast arises from the consequent decrease in the spin population difference between a and b states

178

E. Gianolio et al.

a

b

100

100 2.35 T 11.75 T 23.5 T 117.5 T

ST%

80

80

60

60

40

40

20

20 0

0 101

102

103 Kex

104

105

106

− S −1

0

50

100

150

200

250

[mobile protons] - mM

Fig. 13.12 (a) Simulated kex dependence of ST% at different B1 field intensity. (The data were calculated at 7 T, Dw = 50 ppm, concentration of mobile protons 50 mM, irradiation time 4 s);

(b) Simulated dependence of ST% on the concentration of saturated mobile protons. (The data were calculated at 7 T, Dw = 50 ppm, kex 3,300 s−1, B1 11.75 mT, irradiation time 4 s)

relevant are kex and the number of mobile protons per molecule. The simulated curves reported in Fig. 13.12a, calculated by using the theoretical model recently described by Woessner et al. (2005a), show the kex dependence of ST % at different B1 field intensities. For a given set of parameters (Dw, T1, and T2 for the exchanging proton sites, concentration of mobile protons, magnetic field strength, irradiation time), the increase of kex leads to a ST enhancement up to a specific kex value, after which the efficiency of the ST drops off. Such a decrease can not be ascribed to the approaching of the coalescence condition, because, being Dw the same for all simulations, it occurs at different kex values. Rather, it is the result of the reduced number of mobile spins effectively saturated by the B1 field owing to the broadening of their resonance associated to the increase of kex. The simulated profiles suggest that kex can not be increased at will and, furthermore, that high B1 intensity is required for maximizing the ST effect. It is important to recall here that the largest B1 value for an in vivo MR scan is limited, for safety purposes, by the SAR (specific absorption rate) value. For this reason, the design of highly sensitive CEST agents has to be pursued by increasing the number of mobile protons. Figure 13.12b reports a theoretical curve describing the dependence of the ST effect on the concentration of the irradiated protons. By considering an ST detection threshold of about 10%, the required concentration of mobile protons is in the order of few mM. Of course,

the corresponding minimum concentration of CEST agent is dependent upon the number of mobile protons present per molecule. Actually, the detection limit for small-sized CEST agent (containing less than 10 protons/molecule), like aminoacids, heterocyclic compounds, sugars, or paramagnetic metal complexes, is in the mM range (Ward et al. 2000; Aime et al. 2002a). A significant sensitivity improvement was achieved by investigating the ST properties of macromolecular agents, both diamagnetic (polyaminoacids, dendrimers, RNA-like polymers) (Goffeney et al. 2001) and paramagnetic ones (Aime et al. 2003a). Such systems contain thousands of mobile protons, and consequently, their detection limit is the mM range. Although the sensitivity displayed by the macromolecular CEST agents is close to the current limit for the clinically approved Gd(III)-based agents, the possibility to exploit the peculiar properties of CEST media in molecular imaging MR protocols requires a further step forward.

13.4.2 ParaCEST As said above, a good CA for CEST applications has to possess mobile protons whose exchange rate with water (kex) is as high as possible before their broadening makes the RF irradiation ineffective. Roughly, this condition occurs when kex approaches the separation

179

13 MR Contrast Agents

(in Hz) between the chemical shift values, Dw, of the two exchanging sites (kex » Dw). Therefore, larger Dw values enable the exploitation of higher kex values, thus resulting in an enhanced CEST effect. In 2001, Zhang et al. (2001) showed that a particularly useful source of highly shifted exchangeable protons can be provided by the slowly exchanging water protons bound to a paramagnetic Eu(III) chelate. In this complex, the irradiation of such protons, which resonate at ca. 50 ppm downfield from the bulk water signal, determined a significant CEST effect in the images obtained at 4.7 T. An alternative source for highly shifted exchangeable protons may be provided by amide protons of the ligand that forms the paramagnetic complexes. This approach might offer a better control of the basic requisites of a CEST CA as, in principle, one may design the molecular structure of the complex in order to pursue optimal values for the chemical shifts and exchange rates of the amide protons. Moreover, other advantages of this approach may be represented by the increase in the number of labile protons to be irradiated (with consequent reduction of the administered dose) and by the extension of the CEST application to lanthanide complexes endowed with fast-exchanging metal-bound water or, even better, without any coordinated water molecule. Macrocyclic tetramide derivatives of DOTA represent the primary ligand choice for PARACEST agents. In fact, this coordination cage ensures sufficiently slow water exchange rates (and also a high chemical inertness of the metal complex) in order to fulfill the Dw > kex condition. The first example (DOTAMGly, elsewhere named DOTA-4AmC – Fig. 13.13) of the detection of a CEST contrast upon irradiation of amide protons in a PARACEST agent was reported in 2002 (Aime et al. 2002b) Among the different Ln(III) ions investigated, the Yb(III) complex was the most efficient agent when the amide protons were saturated (kex 180 s−1, Dw offset = −16 ppm), with an observed effect of 52%, at pH 7.4, in a 25-mM solution of the agent (7 T, B2 intensity

-OOCH CHNOC 2

N

N

CONHCH2COO-

Ln -OOCH CHNOC 2

N

N

CONHCH2COO-

Fig. 13.13 Structure of [Ln-DOTAMGly]− complexes

25 mT). Other additional important findings of this work were as follows: (1) the observation that the Dw offsets for the same proton site are strongly dependent on the magnetic property of the Ln(III) ion, (2) the demonstration of the remarkable pH dependence of the CEST contrast upon irradiation of the amide protons, and (3) the exploitation of the two different proton pools for detecting a pH-dependent and concentrationindependent ST response. The CEST sensitivity displayed by [Ln-DOTAMGly]− complexes is dependent on the irradiated proton site: for amide protons, a CEST effect of 5% requires a complex concentration of few millimolar, whereas for the metal coordinated water protons, the threshold lowers down to 0.5–1 mM (Terreno et al. 2004). More recently, Woods et al. demonstrated that also OH groups can be exploited in a CEST experiment with a PARACEST probe, at least under certain experimental conditions (Woods et al. 2006). Figure 13.14 reports the Z-spectrum of a 36-mM solution of [Eu-CNPHC]3+ (see the figure inset) acquired in d3-acetonitrile containing small amount of water (25°C, 23°C, B2 intensity 2 mT). The three distinct CEST peaks corresponding to the three nonequivalent coordinating OH groups can be easily detected along with a fourth peak attributable to the metal-bound water protons. A significant sensitivity enhancement for PARACEST agents could be attained by designing polymeric systems containing a high number of PARACEST units. A couple of examples have been presented so far, involving the linkage of PARACEST agents, similar to those discussed in this section, to macromolecules like dendrimers (Pikkemaat et al. 2007), or to nanosized systems like perfluorocarbon (PFC) nanoparticles (Winter et al. 2006). A route for improving the sensitivity of paramagnetic CEST agents without recurring to polymeric PARACEST systems has been proposed in 2003(Aime et al. 2003a). The idea was to promote the formation of noncovalent supramolecular adducts (named SUPRACEST) between a paramagnetic shift reagent (SR) and a substrate, possibly macromolecular, acting as source of mobile protons. The proof-of-concept for this approach was obtained by using [TmHDOTP]4−, a well-known SR for cations, and polyarginine, a cationic polyaminoacid containing two different sets of mobile protons potentially CESTactive (the amide protons of the backbone, and the guanidine protons of the side chains). The Z-spectrum

180

E. Gianolio et al. 100 HO

N

N

OH 80

Eu HO

N

N

CONH2

NO2

Eu-OH2

60 M1/M0%

Fig. 13.14 Z-spectrum (right) of a 35-mM solution of [Eu-CNPHC]3+ (left) (6.3 T, 25°C). (Modified from Hunjan et al. 2001 with permission from American Chemical Society)

40 OH OH OH

20 0 30

of the polymer alone at pH 7.4 (0.11 mM, B2 intensity 25 mT) did not show any CEST effect (Fig. 13.15, open squares) because the exchange rate of the guanidine protons is too fast. Likely the CEST detection upon saturation of the amide protons is prevented by the large spillover effect caused by the relatively intense B2 field (actually, a CEST effect was detected for both proton sites at pH 5 and B2 intensity of 1 mT). By adding the paramagnetic SR (18-fold molar excess), a supramolecular adduct is formed, and a CEST peak at an offset of about 30 ppm appears in the Z-spectrum (Fig. 13.15, solid squares). The measurement of the ST effect in solutions containing different concentrations of the interacting 100 80 60 I/ %

40 20 0 60

40

20

0 −20 v/ppm

−40

−60

Fig. 13.15 Z-spectra of a solution of a 0.11-mM solution of polyarginine alone (open squares) and in the presence of 2 mM of [TmHDOTP]4− (solid squares) (7T, 312 K, pH 7.4, saturation scheme: 2 s single cw pulse, B2 intensity 25 mT). The CEST peak at 30 ppm corresponds to the guanidine protons of the polymer shifted upon the interaction with the paramagnetic shift reagent. (Reprinted from Hall et al. 1997 with permission from Wiley VCH Verlag GmbH & Co)

20

-10 0 10 Saturation Offset / PPM

-20

species (maintaining the SR/Polyarg ratio of 18:1) highlighted the noticeable sensitivity improvement of this system. A CEST effect of 5% required sub-mM concentrations of polymer (1.7 mM) and SR (30 mM).

13.4.3 LipoCEST Nanovesicles able to entrap solvent in their cavity, like liposomes, represent an ideal system for developing highly sensitive CEST agents. In fact, the number of water protons inside the liposome cavity is several order of magnitude larger (106–109 depending on the size of the vesicle) than macromolecular systems. In addition, liposome membranes are water permeable, and the exchange rate of the intraliposomal water protons (i.e., the kex of the CEST pool) can be properly modulated by changing the lipid composition. For instance, it has been demonstrated that membranes based on saturated phospholipids, like DPPC (dipalmitoyl-phosphatidylcholine), are much less permeable than bilayers containing partially unsaturated phospholipids like POPC (palmitoyloleyl-phosphatidylcholine) (Terreno et al. 2008a). In addition, the exchange rate of the water molecules across the membrane is not very fast (102–103 s−1), and therefore, high ST efficiency can be attained without using high intensity B2 fields. Of course, in order to act as a CEST agent, the resonance frequency of the intraliposomal water protons must be different from the bulk water protons in which the liposomes are suspended, and this task was successfully accomplished by encapsulating a paramagnetic SR in the liposome cavity (Aime et al. 2005).

181

13 MR Contrast Agents Fig. 13.16 Macrocyclic Ln(III)-based complexes typically used as shift reagents for liposome-based CEST agents

−OOC

N

N

COO−

−OOC N

Ln −OOC

N

N

COO−

−OOC

Ln N

COO−

−OOC

N

N

COO−

Ln N

COO−

[LnDOTMA]−

[LnDOTA]−

N

−OOC

N

N

OH

[LnHPDO3A]

Paramagnetic lanthanide(III) complexes are recognized as the best class of shift reagents, especially those having a high symmetry macrocyclic structure, like [LnDOTMA]−, [LnDOTA]−, or [LnHPDO3A] (Fig. 13.16), where one fast-exchanging water molecule is axially coordinated to the metal centre, thus maximizing the induced paramagnetic shift that is dominated by the dipolar contribution. Tm(III) and Dy(III) are the most used Ln(III) ions in virtue of their high values (of opposite sign) of the dipolar constant, CD, that directly affects the dipolar contribution. Interestingly, the paramagnetic shift induced by these ions, when they are complexed by the same ligand, has an opposite sign, and this behavior expands the range of the saturation frequency offset values that are accessible for these nanoprobes. The liposome encapsulation of high amount (ca. 200 mM) of the SRs induces paramagnetic shift offset of about ±4 ppm. The sensitivity of these nanosystems, dubbed LipoCEST, is very high: subnanomolar amounts of vesicles are sufficient to generate an ST of ca. 10% in vitro. Since the efficiency of the ST can be limited by the occurrence of spillover effects that are expected for saturation frequency offsets close to the resonance of bulk water (associated with the use of high intensity B2 fields), the development of LipoCEST agents with highly shifted intraliposomal water protons could be very beneficial for improving the potential of such systems. Two approaches have been adopted to achieve this task: (1) to exploit shift contribution arising from bulk

magnetic susceptibility (BMS) effects and/or (2) to increase the concentration of paramagnetic centers encapsulated in the liposome cavity. The former route requires the compartmentalization of the paramagnetic agent in nonspherical vesicles, whereas traditional liposomes adopt a spherical shape (Fig. 13.17a). Liposomes are flexible objects, and due to the water permeability of their membrane, they are sensitive to osmotic forces. When suspended in a hyperosmotic medium, liposomes react by shrinking themselves, leaking water, and changing shape. Hence, a liposome encapsulating an ipotonic solution of an SR (better if neutral as [LnHPDO3A]) and, then, dialyzed against an isotonic buffer (step necessary for the separation of the not encapsulated SR) will not be spherical anymore (Fig. 13.17b). As a consequence, the BMS contribution will be operative, and the Dintralipo value will increase. In addition, the osmotic shrinkage concentrates the SR in the liposome aqueous core, thus also enhancing the extent of the dipolar contribution. A nice demonstration of this effect was obtained by encapsulating a Gd(III) complex ([GdHPDO3A]), for which the dipolar contribution is null (CD for Gd is zero). The observation of the signal of the intraliposomal water protons at ca. 7 ppm from the bulk and the detection of a CEST effect upon saturation highlighted the validity of this approach (Aime et al. 2007). Of course, when the BMS contribution is added to the dipolar one (for instance encapsulating [TmHPDO3A]), a further increase in the Dintralipo values can be obtained (Terreno et al. 2007b).

Fig. 13.17 Picture of the several class of liposomebased CEST agents developed so far. (a) Spherical LipoCEST; (b) non spherical LipoCEST; (c) non spherical LipoCEST with incorporation of an amphiphilic SR; (d) non spherical LipoCEST with incorporation of neutral polynuclear hydrophobic SRs.

b

a

c

d

182

E. Gianolio et al.

A significant shift increase can also be achieved by exploiting one of the advantages of using liposomes, i.e., the possibility to incorporate molecules in their bilayer. The incorporation of an amphiphilic SR in a nonspherical LipoCEST (Fig. 13.17c) has two main advantages: (1) it increases the overall amount of paramagnetic centers in the liposome cavity (at least for the portion pointing inwards) with the consequent increase of both dipolar and BMS contributions and (2) it strongly affects the magnetic anisotropy of the liposome membrane, thus allowing the switch from the parallel to the perpendicular alignment of the vesicles within the static magnetic field of the NMR/MRI spectrometer. Since the orientation of the vesicles can modulate the sign and the magnitude of the paramagnetic shift, the Dintralipo values for these systems are strongly dependent on the magnetic properties of the incorporated SR (Delli Castelli et al. 2008). The amount of paramagnetic centers encapsulated in the aqueous core of a liposome can be further increased by using neutral polynuclear hydrophilic SRs (Fig. 13.17d) (Terreno et al. 2008b) In fact, as the maximum concentration allowed is limited by osmotic rules, neutral multimers should increase the overall payload of paramagnets inside the vesicle. Actually, the spectra reported in Fig. 13.18 indicate that the use of bi or trinuclear SRs increases the Dintralipo values with respect to a mononuclear (here [TmHPDO3A]) agent. By using all the strategies mentioned above, it is now possible to prepare liposome-based CEST agents with Dintralipo values in the interval ± 60 ppm. This result is relevant for the sensitivity issue, even if for osmotically shrunken liposomes, the advantage of saturating

far from the bulk water resonance is partly counterbalanced by the decreased number of intraliposomal water protons that can be saturated due to the osmotic shrinkage. However, the current range of available Dintralipo values makes these nanosystems promising candidates for the multiple detection of CEST agents, as it has been already reported as proof-of-concept on an ex vivo model (Terreno et al. 2008c). One possible pitfall related to the in vivo use of LipoCEST agents is represented by the required high stability of the vesicle that can be detected in CEST-MR experiment as long as they are intact. A possible route to overcome this issue without losing sensitivity relies on the use of different nanosystems (e.g., nanoemulsions) (Winter et al. 2006) or viral capsides (Vasalatiy et al. 2008) that may carry a very high payload of paramagnetic CEST agents providing the pool of mobile protons to be saturated.

13.5 Hyperpolarized Molecules The energy difference between the nuclear spin states involved in the NMR transitions is very low, and therefore, these states are almost identically populated. The NMR signal intensity is proportional to polarization (P), which depends upon the difference in the spin levels population according to (13.5):

P=

N + - N-

æ g ·h· B0 ö = tanh ç , è 2·kB ·T ÷ø

N + + N-

(13.5)

where N+ and N− represent the numbers of spins in the two different orientations with respect to the applied

TmHPDO3A −OOC N

N

COO−

−OOC N

Tm −OOC

40

20

0

−20

N

N

N +

N

N

Tm

−OOC

CONH

HNOC

N

−OOC

CONH HNOC COO−

COO−

Tm N

ppm −OOC

N

N

N

COO−

HNOC N

N COO−

Tm+

N

CONH COO−

−OOC

N

N +

COO−

Tm

N

COO−

Fig. 13.18 1H-NMR spectra (14.1 T, 25°C) of aqueous suspensions of osmotically shrunken liposomes incorporating the same Tm-based amphiphilic SR and encapsulating the same amount of the hydrophilic SRs shown on the right

183

13 MR Contrast Agents

magnetic field B0, g is the gyromagnetic ratio of the nucleus, kB is the Boltzmann constant, and T is the absolute temperature. Thus, NMR and MRI result to be rather insensitive techniques in respect with other analytical and imaging methodologies. However, it has been shown that a nonequilibrium state, characterized by a larger difference in the spin levels populations (hyperpolarized state), can be achieved, thus providing a route of raising the sensitivity of the NMR experiment. Hyperpolarized molecules yield very strong signal enhancements and make the observation of low g heteronuclei possible, which usually are not detectable due to their low natural abundance and gyromagnetic ratios. Since they do not show any background signal in the corresponding images, it is possible to obtain very clear images where only the regions reached by the agent are visible on a black background. This has been exploited in many MRI experiments, ranging from vascular to perfusion studies, up to new and promising applications in molecular and metabolic imaging. Hyperpolarized agents generate the contrast in a different way with respect to standard MRI CAs. In fact, while the latter act on the relaxation of water protons causing a positive (T1-CA) or a negative (T2-CA) effect on the water signal intensity, hyperpolarized molecules are themselves source of the NMR signal. In this case, signal intensity and SNR linearly depend upon their concentration and polarization level. The main drawback of hyperpolarization is its short lifetime. In fact, relaxation processes re-equilibrate the spin populations on the T1 timeframe, destroying hyperpolarization. As a consequence, the type of molecules that can be used as hyperpolarized probes is limited to small molecules containing at

least one long-relaxing nucleus, possibly isolated from dipolarly coupled proton nuclei. Furthermore, images of hyperpolarized substances must be acquired very quickly after administration, by using either pulse sequences consisting of small flip angles, in order to preserve magnetization or single shot experiments. The mostly used sequences are RARE, EPI, and true FISP (Mansson et al. 2006). For molecular imaging applications, in which the metabolic transformations are monitored by the detection of the metabolites NMR signals after administration of a hyperpolarized substrate, CSI sequences are used, and a number of studies aimed at optimizing the pulse sequences and the image reconstruction methods have been reported (Levin et al. 2007; Larson et al. 2008).

13.5.1 Hyperpolarization Techniques According to (13.5), hyperpolarization can be obtained by simply maintaining the sample at very high magnetic field and ultralow temperature for a time sufficient to force the system into a nonequilibrium state in which the spin populations are consistently altered (Fig. 13.19). This has been sometimes called the “brute force” approach. Due to very long relaxation times in the ultralow temperature range, “relaxation switches” are necessary to decrease the polarization time. In spite of the potential general applicability of the method, this approach is not easily applicable due to difficulties in reaching very high magnetic field strengths and very low temperature conditions. Furthermore, the

1 0,9 Nuclei Polarization

0,8 0,7 0,6 0,5 0,4 0,3 0,2 0,1

Temperature (mK)

96

90

84

78

72

66

60

54

48

42

36

30

24

18

12

0 6

Fig. 13.19 Temperature dependence of polarization at 20 T (deep blue), 16 T (blue), and 12 T (cyan)

184

E. Gianolio et al.

relaxation switches, if not completely removed or inactivated after the completion of the polarization process, may cause relaxation and polarization loss. For these reasons, other hyperpolarization techniques have been developed, namely laser-induced hyperpolarization of noble gases, parahydrogen-induced polarization (PHIP), and dynamic nuclear polarization (DNP).

13.5.1.1 Hyperpolarization of He-3 and Xe-129 by Optical Pumping When a noble gas (He-3 or Xe-129) is mixed with a vapor of an alkali-metal or of metastable atoms (usually 3He atoms), and the mixture is irradiated with circularly polarized light from lasers at suitable frequency, the angular momentum of the laser light is adsorbed by the alkali or metastable atoms, which result to be polarized. Then, interatomic collisions allow polarization to be transferred to the noble gas atoms via a “spin exchange” process (Fig. 13.20) (Albert et al. 1994; Moller et al. 2002; Altes and Salerno 2004). Hyperpolarized noble gases have found applications in MRI of the lungs, and in a limited number of perfusion studies (Albert et al. 1994; Moller et al. 2002; Altes and Salerno 2004).

He-3 has been used in ventilation and diffusion studies applied to emphysema diagnosis (Altes and Salerno 2004); Xe-129 solubility and lipophilic properties have been exploited for functional MRI, and its binding to hemoglobin has been used to measure blood oxygenation (Swanson et al. 1997; Wolber et al. 2000). Methods for intravenous injection of both Xe and He have been developed: Xe can be dissolved in some biocompatible carriers (Wolber et al. 1999), while 3He has been loaded in microbubbles suitably functionalized with peptides (Callot et al. 2001) (Fig. 13.21). Finally, HP-Xenon-based biosensors, which trap Xe atoms in molecular cages functionalized with suitable vectors, have been proposed in order to target specific biomolecules. In these systems, Xe has been encapsulated in a criptophane cage, bound to a biotin unit for protein recognition (Spence et al. 2001). As the cage-encapsulated Xe is in exchange with free Xe, and the two corresponding NMR signals are well shifted, an interesting development has been introduced to enhance the sensitivity of the sensor. In fact, by irradiating the Xe resonance in the targeted cage, saturated magnetization is transferred to the signal of free Xe, which is exchanging with it, in a way similar to what occurs for CEST agents. Therefore, this class of agents has been named HYPER-CEST (Schroder et al. 2006).

hv

Rb Rb

Rb Hyperpolarization of Rb atoms

Rb Rb

Rb

Collisions between Rb and Xe Rb Xe

(Van der Waals molecule)

Fig. 13.20 Schematic representation of the laser-induced hyperpolarization of noble gases

Polarization transfer to Xe atoms via “spin exchange”

Xe

Xe Xe

13 MR Contrast Agents

Fig. 13.21 Coronal HP 3He image of the lungs of a normal healthy human volunteer. (Reprinted from Merritt et al. 2007a with permission from Wiley)

13.5.1.2 Parahydrogen-Induced Polarization The hydrogen molecule can be easily hyperpolarized by changing the relative populations of its nuclear spin states, without the need for complex hyperpolarization devices. In fact, nuclear spins in the H2 molecule can be either aligned or opposed: three degenerate levels correspond to the nuclear isomer named ortho-hydrogen (o-H2, S = 1, 75% natural abundance), while only one nuclear state corresponds to the nuclear isomer named parahydrogen (p-H2, S = 0, 25% abundance). The parahydrogen state has lower energy, and it is therefore possible to enrich hydrogen in the para form by keeping the H2 gas at low temperature. A paramagnetic catalyst is required for the enrichment because the ortho-para transition is otherwise forbidden by symmetry rules. The hydrogen hyperpolarization process thus simply consists in maintaining hydrogen gas at low temperature in the presence of charcoal or iron oxide or other paramagnetic substances, and the polarization degree depends on the temperature, being about 50% at 73 K and almost 100% at 4 K. Parahydrogen polarization can be transferred to other molecules by chemical reaction. In hydrogenation reactions carried out with H2 enriched in the paraform, the product molecules present NMR spectra, which are characterized by strongly enhanced absorption/emission signals which, in theory, may be up to 105 times higher than those observed when normal

185

hydrogen is used. This phenomenon has been called PHIP (Bargon and Natterer 1997; Duckett and Wood 2008). Such polarization can then be transferred from H atoms to neighboring heteronuclei in the hydrogenated products via scalar couplings or nuclear Overhauser effect (nOe) (Barkemeyer et al. 1995; Aime et al. 2003b). For MRI applications, two operations must be carried out after the parahydrogenation has taken place: (1) the antiphase pattern of the heteronucleus signal obtained after polarization transfer must be converted into an in-phase signal in order to allow the image acquisition (this can be achieved by a field cycling procedure (Johannesson et al. 2004) or the application of a suitable pulse sequence (Goldman and Johannesson 2005)); (2) the catalyst used for parahydrogenation and (if present) the organic solvent must be removed before injection. The catalyst (usually cationic Rh(I) complexes) may be removed by cation exchange, and the solvent by spray-distillation. The fulfillment of these operations requires time that is at the expenses of the attainable signal intensity, as polarization decay occurs on the T1 timescale. An alternative method for the attainment of pure aqueous solutions of hyperpolarized molecules consists of carrying out the parahydrogenation reaction in an organic solvent not miscible with water, and then quickly extracting the hyperpolarized water-soluble molecule by phase transfer (Aime et al. 2008). The first example of MRI of a PHIP hyperpolarized 13 C agent was reported by Golman et al. in 2001. A solution of hyperpolarized dimethylmaleate derived by parahydrogenation of dimethylacetylenedicarboxylate was injected into the tail vein of a rat, and a clear picture of the distribution of the molecule in the major and minor vessels of the animal was obtained in a subsecond time frame. Successively, hyperpolarized hydroxyethylpropionate (derived by parahydrogenation of hydroxyethylacrylate) has been used to obtain the angiography of a guinea pig head (Goldman et al. 2005), coronary angiographic images, and images of the pulmonary arterial circulation in pigs (Ishii et al. 2007) (Fig. 13.22). Hyperpolarized molecules can find application also in perfusion investigations, as shown in ref (Ishii et al. 2007), where perfusion of the lungs and of the myocardium, respectively, has been clearly visualized. For quantitative determinations of perfusion parameters, an adaptation of the currently used Bolus Tracking Theory (which correlates the signal intensity with the

186

E. Gianolio et al.

Fig. 13.22 HP 13C MRI angiograms depicting pulmonary blood flow through a coronal slice just posterior to the heart of a pig (images taken at 1-s intervals). (Reprinted from Schlemmer et al. 1999 with permission from Wiley)

contrast agent concentration in the tissue) has been suggested, taking into account the loss of signal intensity due to polarization decay. This has allowed obtaining cerebral perfusion maps in rats and parametric maps for the assessment of renal cortical blood flow in rabbits (Johansson et al. 2004). Furthermore, hyperpolarized hydroxyethylpropionate has been used in interventional MRI, as the catheter used for injection can be clearly visualized (Magnusson et al. 2007). Recently, it has been shown that hyperpolarized succinate can be obtained by parahydrogenation of fumarate (Chekmenev et al. 2008), and it has been used to perform in vivo CSI (Bhattacharya et al. 2009). As it is based on the addition of a parahydrogen molecule to an unsaturated substrate, the PHIP method appears to be strongly limited in terms of the number of candidate molecules. Nevertheless, it has been recently shown that a good signal enhancement can be detected for the 15N resonances of N-containing substrates reversibly coordinated to an Ir complex that has been added of a parahydrogen molecule (Adams et al. 2009). The observed behavior opens interesting perspectives to polarization via para-H2, as it shows that enhanced NMR signals can be obtained without hydrogenation of the substrate as previously assumed.

13.5.1.3 Dynamic Nuclear Polarization DNP is a method that consists in irradiating electrons to promote the transfer of thermal polarization (which

is much higher for electrons due to their high g value) to nuclei in solids via “flip-flop” transitions (Comment et al. 2007). Nuclei are polarized by two different mechanisms, which can operate separately or in parallel depending on the conditions: i) the solid effect causes nuclei to be polarized in a single step by means of a two quantum transition; ii) thermal mixing, on the contrary, is a two-step process where the first step is a single quantum transition. Thermal mixing is generally the experimentally dominant mechanism when the width of the ESR line is comparable to or larger than the NMR frequency, and the concentration of electron spins is high (Comment et al. 2007). In practice, the material to be polarized is dissolved in a glass-forming solvent, doped with a stable radical species and placed into the magnetic field. The solution is frozen and irradiated. At the end of the polarization process, the sample is raised above the liquid Helium level and is rapidly dissolved in hot water, still inside the magnetic field. It is then quickly transferred for the MRI acquisition. It has been demonstrated that by using this dissolution method, a good level of polarization is maintained for the time necessary for the acquisition of the MR image (Ardenkjaer-Larsen et al. 2003). In principle, every nucleus in every molecule can be hyperpolarized by DNP. Examples include DNPhyperpolarization of 15N in urea (Ardenkjaer-Larsen et al. 2003), carbazole (Hu et al. 2000), aminoacids (Hall et al. 1997) and choline (Gabellieri et al. 2008), and of 89Y in Y(III) chelates (Merritt et al. 2007a). Among these, choline appears very promising since its

187

13 MR Contrast Agents

N nucleus is characterized by a particularly long T1 (285 s at 11.7 T and 298 K), and this allowed following its fate in vitro in the presence of purified human choline kinase, which promotes its conversion to phosphocholine, opening the way to metabolic imaging of 15 N-choline (Gabellieri et al. 2008). Nevertheless, no in vivo applications have been reported up to now. In fact, at the moment, only 13C has been used for in vivo MRI experiments. This is due to the higher gyromagnetic ratio of 13C with respect to other low-relaxing nuclei such as 15N, which increases its sensitivity, and to the availability of dedicated rf coils tuned at the 13 C frequency. Many examples of 13C-containing substrates hyperpolarized by DNP have been reported. The most challenging application of these hyperpolarized 13C tracers is their use in molecular/metabolic imaging, which exploits the high signal intensity for visualizing the products of metabolism few seconds after administration of the agent. The first studies of metabolic pathways by 13C-hyper polarized substances have been performed by using hyperpolarized 13C-pyruvate. Pyruvate is a key molecule in major metabolic and catabolic pathways in the mammalian cells, as it is converted to alanine, lactate, or carbonate to a different extent depending on the status of the cells. A number of papers dealing with metabolic imaging by 13C-hyperpolarized pyruvate have appeared in the literature since 2006 (Golman et al. 2006; Merritt et al. 2007b; Albers et al. 2008; Day et al. 2007; Schroeder et al. 2008, 2009). Its first application is the tumor imaging in vivo (Golman et al. 2006): the maps of distribution of pyruvate, alanine, and lactate 15

Anatomical 1H image

13C-pyruvate

after pyruvate injection show that tumors can be localized by the highest NMR 13C signal from lactate produced few second after injection. The various tumor histologic grades can also be differentiated on the basis of lactate levels (Albers et al. 2008). The procedure is currently under intense scrutiny, and the clinical phase of trial has been undertaken in order to develop a method for monitoring the therapeutic treatment of prostate cancer. In fact, a reduction of the conversion of pyruvate to lactate in tumor has been observed within 24 h of chemotherapy in the treatment of lymphoma-bearing mice (Day et al. 2007) (Fig. 13.23). Hyperpolarized 13C-pyruvate has found application also in the assessment of cardiac metabolism, both in isolated rat hearts and in vivo. In this case, the pyruvate transformation into bicarbonate promoted by pyruvate deyhdrogenase (PDH) is monitored: the bicarbonate level in the myocardium is lower or absent in the ischemic or postischemic area due to a diminished activity of PDH (Schroeder et al. 2008, 2009). Besides hyperpolarized pyruvate, other DNP hyperpolarized molecules have been tested for molecular/metabolic imaging, ranging from lactate (Chen et al. 2008) to glutamate (Gallagher et al. 2008a), to fumarate (Zandt et al. 2009), and to bicarbonate (Gallagher et al. 2008b). The latter one is particularly interesting because it allows determining the extracellular pH in vivo from 13C MR measurements of the relative signal intensities of bicarbonate and carbon dioxide, since their concentrations determine local pH according to the equation: pH = pKa + log([HCO3−]/ [CO2]) (Fig. 13.24). 13C-alanine

13C-lactate

30

8

20

15

4

10

0

0

0

Fig. 13.23 Metabolic maps of a rat bearing P22 tumor. The tumor is clearly evidenced by the highest lactate signal. (Reprinted from Golman et al. 2006 with permission from the American Association for Cancer Research)

188

E. Gianolio et al.

a

b

HCO3

7.3

6.9

13C-MRS

6.7

pH

7.1

6.0

c

60 40

180

160 140 Chemical shift (p.p.m)

120

20 0

100 80 60 40 20 0

(% of maximum)

CO2

d

13CO 2

80

H13CO3 (% of maximum)

100

Fig. 13.24 Left: 13C spectrum from a murine lymphoma in vivo, acquired 10 s after injection of hyperpolarized 13C-bicarbonate. Right: metabolic and pH maps of the mouse. (a) 1H image

(the tumor is outlined in red in the 1H image; (b) pH map; (c) bicarbonate map; (d) CO2 map (Reprinted from Weinmann et al. 2000 with permission from Nature Publishing Group)

13.6 19F-containing Species

accurate quantification using spectroscopy at 4.7 T, as corroborated with proton relaxation rate measurements at 1.5 T and trace element (gadolinium) analysis. Finally, the extension of these techniques to a clinically relevant application, the evaluation of the fibrin burden within an ex vivo human carotid endarterectomy sample, demonstrates the potential use of these particles for uniquely identifying unstable atherosclerotic lesions in vivo. Previous studies have reported that the minimum detectable limit of 19F at 1.5 T is 30 mM (Schlemmer et al. 1999). This concentration is difficult to achieve with a circulating agent that might passively accumulate at a selected site in vivo. Targeted PFC nanoparticles, however, might overcome this limitation, because the specific binding permits local accumulation of nanoparticles at the site of pathology (up to 1 nM concentration of particles). As each individual nanoparticle carries an extremely high payload of fluorine (12–13 M), the detection threshold can be attained. Actually, it has been shown that it was possible to accumulate 19F at the plaque surface to a concentration >0.1 M, which is well above the reported minimum detection limit. Recently, this approach has been extended to cellular labeling for MRI-based tracking. In 2007, Ahrens et al. (Srinivas et al. 2007) described an in vivo imaging method for visualizing and

Given its high gyromagnetic ratio, fluorine-19 is almost as sensitive as protons, but the very low abundance of fluorine-containing compounds in vivo has ruled out any clinical applicability of this technique. However, 19 F MRI has been used for a variety of research applications including studying tumor metabolism (Schle mmer et al. 1999) and mapping tissues physiologic pO2 (Noth et al. 1999). To overcome the detection limitations of fluorine-containing compounds in vivo, it has been suggested either to increase the applied magnetic field (Hunjan et al. 2001; Fan et al. 2002) or to increase the concentration of 19F at the observation site. Actually, it has been shown that aggregates (e.g., PFC nanoparticles) loaded with a large number of fluorine-containing molecules may allow the detection of 19 F-MR images. In the work of Morawski et al. (2004), the PFC content of fibrin-targeted nanoparticles is exploited for the purposes of 19F imaging and spectroscopy of atherosclerotic plaques (Fig. 13.25). Additionally, the quantity of bound nanoparticles formulated with different PFC species was calculated using spectroscopy. Results indicate that the high degree of nanoparticle binding to fibrin clots and the lack of background 19F signal allow

189

13 MR Contrast Agents Fig. 13.25 (a) 19F image (4.7 T) of a single slice through a clot treated with crown ether emulsion. (b) 1H image (4.7 T) of the same slice showing the anatomy of the clot. (c) False color overlay of 19F signal onto 1H image clearly showing localization of 19F signal to the clot surface. (Reprinted from Wolber et al. 1999 with permission from Wiley)

a

19F

b

MRI

clot surface

c

1H

1F

MRI

agarose

MRI with 19F color overlay

19F

Signal High

Low

quantifying a specific cell population. T Cells were labeled ex vivo with a perfluoropolyether nanoparticle (PEPE) tracer agent and then detected in vivo using 19F MRI following cell transfer. 19F MRI selectively visualizes only the labeled cells with no background, and a conventional 1H image taken in the same imaging session provides the anatomical context. Using the nonobese diabetic mouse, an established model of type 1 diabetes, 19F MRI data were acquired showing the early homing behavior of diabetogenic T cells to the pancreas. A computational algorithm provided T cell counts in the pancreas. Approximately 2% of the transferred cells homed to the pancreas after 48 h. The technique allows for both unambiguous detection of labeled cells and quantification directly from the in vivo images. The in vivo quantification and cell trafficking patterns were verified using 19F spectroscopy and fluorescence microscopy in excised pancreata. Another example is provided by Wickline and coworkers (Partlow et al. 2007) who demonstrated that fluorine MRI of stem/progenitor cells labeled with different types of liquid PFC nanoparticles yields unique and sensitive cell markers distinct from any tissue

background signal. Mononuclear cells harvested from human umbilical cord blood were labeled with nanoparticles composed of two distinct PFC cores (perfluorooctylbromide and perfluoro-15-crown-5 ether). The sensitivity for detecting and imaging labeled cells was defined on 11.7T (research) and 1.5T (clinical) scanners. Stem/progenitor cells readily internalized PFC nanoparticles without aid of adjunctive labeling techniques, and cells remained functional in vivo. PFClabeled cells exhibited distinct 19F signals and were readily detected after both local and intravenous injection. Advances in the design of fluorinated nanoparticles for molecular MRI were reported by Wickline and coworkers (Neubauer et al. 2008) who showed that remarkable gain in signal enhancement can be obtained by directly incorporating a large number of relaxation agents (50.000–100.000 Gd-complexes), into the lipid monolayer that surrounds the iPFC: in formulation of this kind of nanoparticles, they used Gd-complexes functionalized with bis-oleate (BOA) in order to locate the paramagnetic centers at the nanoparticle surface such that the separation from the PFC core is no more

190 30

4.5

Gd in Solution Gd on NP

Gd in Solution Gd on NP

25

3.5 R2 (1/s)

20 R1 (1/s)

Fig. 13.26 19F relaxation rates (R1 and R2) as a function of Gd concentration (millimoles Gd per liter of PFC) for nanoparticles either formulated with different amounts of Gd on their surface (black squares) or mixed with an aqueous chelate of Gd (gray diamonds). (Reprinted from Zandt et al. 2009 with permission from Wiley)

E. Gianolio et al.

2.5

15 10

1.5 5 0.5

0

100

200

300

[Gd] (mM) in 19F volume

than the length of the lipid monolayer components (»15 Å). This proximity condition influences 19F relaxation by reducing its relaxation times with a linear dependence from Gd(III) concentration (Fig. 13.26). This design increases the magnetic relaxation rate of the 19F nuclei fourfold at 1.5 T and effects a 125% increase in signal – an effect that is maintained when they are targeted to human plasma clots. By varying the surface concentration of Gd, the relaxation effect can be quantitatively modulated to taylor particle properties. This novel strategy dramatically improves the sensitivity and range of 19F MRI/MRS and forms the basis for designing CAs capable of sensing their surface chemistry.

13.7 Conclusions The field of MRI CA is still continuing to grow, and new ideas further widen their range of applications. The limited sensitivity of MRI probes is the major limitation for a widespread diffusion of MRI in molecular imaging studies. Therefore, much work is currently focussed on the task of accumulating a sufficient number of contrast units at the targeting sites. This task is pursued mainly through i) exploitation of cellular uptake pathways that use the cell itself as a container for the accumulation of the imaging payload or ii) the set-up of nanosized (or even microsized) particles containing a huge amount of the imaging reporting units. The latter approach has opened new horizons that conjugate imaging science and nanotechnology. An important feedback from this merging is, for instance, the development of new protocols of

0 0

100 [Gd] (mM) in

200 19

300

F volume

imaging-guided therapy that may find great interest in the domain of personalized medicine.

References Adams RW, Aguilar JA, Atkinson KD, Cowley MJ, Elliott PIP, Duckett SB, Green GGR, Khazal IG, López-Serrano J, Williamson DC (2009) Science 323:1708–1711 Aime S, Ascenzi P, Comoglio E et al (1995) J Am Chem Soc 117:9365 Aime S, Botta M, Fasano M, Geninatti Crich S, Terreno E (1996) J Biol Inorg Chem 1:312–319 Aime S, Botta M, Fasano M, Terreno E (1998) Chem Soc Rev 27:19 Aime S, Chiaussa M, Digilio G et al (1999a) J Biol Inorg Chem 4:766–774 Aime S, Botta M, Geninatti S et al (1999b) Chem Comm 1577 -1578 Aime S, Gianolio E, Terreno E et al (2000) J Biol Inorg Chem 5:488 Aime S, Fasano M, Terreno E, Botta M (2001) In: Merbach AE, Tóth E (eds) The chemistry of contrast agents in medical magnetic resonance imaging. Wiley, Chichester Aime S, Cabella C, Colombatto S, Geninatti Crich S, Gianolio E, Maggioni F (2002a) JMRI 16:394–406 Aime S, Barge A, Delli Castelli D, Fedeli F, Mortillaro A, Nielsen FU, Terreno E (2002b) Magn Reson Med 47: 639 Aime S, Delli Castelli D, Terreno E (2003a) Angew Chemie Int Ed 42:4527 Aime S, Gobetto R, Reineri F (2003b) J Chem Phys 119: 8890–8897 Aime S, Delli Castelli D, Terreno E (2005) Angew Chem Int Ed 44:5513–5515 Aime S, Delli Castelli D, Lawson D, Terreno E (2007) J Am Chem Soc 129:2430–2431 Aime S, Dastrù W, Giovenzana GB, Gobetto R, Reineri F, Santelia D, Viale A (2008) J Am Chem Soc 130: 15047–15053

13 MR Contrast Agents Albers MJ, Bok R, Chen AP, Cunningham CH, Zierhut ML, Zhang VY, Kohler SJ, Tropp J, Hurd R, Yen Y-F et al (2008) Cancer Res 68:8607–8615 Albert MS, Cates GD, Driehuys B, Happer W, Saam B, Springer CS Jr et al (1994) Nature 370:199–201 Altes TA, Salerno M (2004) J Thoracic Imag 19:250–258 Anelli PL, Bertini I, Fragai M et al (2000) Eur J Inorg Chem: 625–630 Ardenkjaer-Larsen JH, Fridlund B, Gram A, Hansson G, Hansson L, Lerche MH, Servin R, Thaning M, Golman K (2003) Proc Natl Acad Sci USA 100: 10158–10163 Artemov D, Mori N, Ravi R, Bhujwalla ZM (2003) Cancer Res 63:2723–2727 Babincova M, Altanerova V, Lambert M, Altaner C, Machova E, Sramka M, Babinec P (2000) Naturforsch C 55:278 Babincova M, Leszczynska D, Sourivong P, Babinec P, Leszczynski J (2004) Med Hypotheses 62:375 Banci L, Bestini I, Luchinat C (1991) Nuclear and electronic relaxation. VCH, Weinheim Bargon J, Natterer J (1997) Progr Nucl Magn Reson Spect 31:293–315 Barkemeyer J, Haake M, Bargon J (1995) J Am Chem Soc 117:2927–2928 Bhattacharya P et al (2009) Abstract n. 461, 50th Experimental Nuclear Magnetic Resonance Conference, Pacific Grove, California, March 29–April 3 Bogdanov A, Matuszewski L, Bremer C, Petrovski A, Weissleder R (2002) Mol Imaging 1:16–23 Botta M (2000) Eur J Inorg Chem 3:399 Brücher E, Sherry AD (2001) In: Merbach AE, Tóth E (eds) The chemistry of contrast agents in medical magnetic resonance imaging. Wiley, Chichester Bulte JW, De Cuyper M (2003) Methods Enzymol 373:175 Bulte JW, De Cuyper M, Despres D, Franck JA (1999) J Magn Reson Imaging 9:329 Callot V, Canet E, Brochot J, Viallon M, Humblot H, Briguet A, Tournier H, Cremillieux Y (2001) Magn Res Med 46: 535–540 Carter D, Ho JX (1994) Adv Prot Chem 45:153–203 Chekmenev EY, Hovener J, Norton VA, Harris K, Batchelder LS, Bhattacharya P, Ross BD, Weitekamp DP (2008) J Am Chem Soc 130:4212–4213 Chen AP, Kurhanewicz J, Bok R, Xu D, Joun D, Zhang V, Nelson SJ, Hurd RE, Vigneron DB (2008) Magn Res Imag 26:721–726 Comment A, van den Brandt B, Uffmann K, Kurdzesau F, Jannin S, Konter JA, Hautle P, Wenckebach WTh, Gruetter R, van der Klink JJ (2007) Conc Magn Reson Part B 31B: 255–269 Corot C, Robert P, Idee JM, Port M (2006) Adv Drug Deliv Rev 58:471 Cunningham CH, Arai T, Yang PC, McConnell MV, Pauly JM, Conolly SM (2005) Magn Reson Med 53:999–1005 Day SE, Kettunen MI, Gallagher FA, Hu D-E, Lerche M, Wolber J, Golman K, Ardenkjaer-Larsen JH, Brindle KM (2007) Nat Med 13:1382–1387 Delli Castelli D, Terreno E, Carrera C, Giovenzana GB, Mazzon R, Rollet S, Visigalli M, Aime S (2008) Inorg Chem 47: 2928–2930

191 Dousset V, Brochet B, Deloire MSA, Lagoarde L, Barroso B, Caille JM, Petri KG (2006) Am J Neuroradiol 27:1000 Duckett SB, Wood NJ (2008) Coordination Chem Rev 252:2278–2291 Engberink RDO, van der Pol SMA, Döpp ED, de Vries HE, Blezer ELA (2007) Radiology 243:467–474 Engelstadt BL, Wolf GL (1988) In: Stark DD, Bradley WG Jr (eds) Magnetic Resonance Imaging. V.C. Mosby, St. Louis Fan X, River JN, Zamora M, Al-Hallaq HA, Karczmar GS (2002) Int J Radiat Oncol Biol Phys 54:1202–1209 Fortin-Ripoche JP, Martina MS, Gazeau F, Menager C, Wilhelm C, Bacri JC, Lesieur S, Clement O (2006) Radiology 239:415 Fossheim S, Kellar KE, Fahlvik AK, Klaveness J (1997) Magn Reson Imaging 15:193-202 Frangioni JV, Hajjar RJ (2004) Circulation 110:3378–3383 Gabellieri C, Reynolds S, Lavie A, Payne GS, Leach MO, Eykyn TR (2008) J Am Chem Soc 130:4598–4599 Gallagher FA, Kettunen MI, Day SE, Lerche M, Brindle KM (2008a) Magn Res Med 60:253–257 Gallagher FA, Kettunen MI, Day SE, Hu D-E, Ardenkjær-Larsen JH, Zandt R, Jensen PR, Karlsson M, Golman K, Lerche MH, Brindle KM (2008b) Nature 453:940–943 Gianolio E, Giovenzana GB, Longo D, Longo I, Menegotto I, Aime S (2007) Chem Eur J 13:5785–5797 Glogard C, Stensrud G, Aime S (2003) Magn Res Chem 41:585 Goffeney N, Bulte JWM, Duyn J, Bryant LH Jr, van Zijl PCM (2001) J Am Chem Soc 123:8628 Goldman M, Johannesson H (2005) CR Phisique 6:575–581 Goldman M, Johannesson H, Axelsson O, Karlsson M (2005) Magn Res Imag 23:153–157 Golman K, Axelsson O, Johannesson H, Månsson S, Olofsson C, Petersson JS (2001) Magn Reson Med 46:1–5 Golman K, Zandt R, Lerche M, Perhson R, Ardenkjaer-Larsen H (2006) Cancer Res 66:10855–10860 Hall DA, Maus DC, Gerfen GJ, Inati SJ, Becerra LR, Dahlquist FW, Griffin RG (1997) Science 276:930–932 Henkelman RM, Stanisz GJ, Graham SJ (2001) NMR Biomed 14:57–64 Hu JZ, Solum MS, Wind RA, Nilsson BL, Peterson MA, Pugmire RJ, Grant DM (2000) J Phys Chem A 104:4413–4420 Hunjan S, Zhao D, Canstandtinescu A, Hahan EW, Antich PP, Mason RP (2001) Int J Radiat Oncol Biol Phys 49: 1097–1108 Ishii M, Emami K, Kadlecek S, Petersson JS, Golman K, Vahdat V, Yu J, Cadman RV, MacDuffie-Woodburn J, Stephen M et al (2007) Magn Res Med 57:459–463 Ito A, Shinkai M, Honda H, Kobayashi T (2005) J Biosci Bioeng 100:1 Johannesson H, Axelsson O, Karlsson M (2004) CR Physique 5:315–324 Johansson E, Mansson, S, Wirestam R, Svensson J, Petersson JS, Golman K, Stahlberg F (2004) Magn Reson Med 51: 464–472 Kircher MF, Allport JR, Graves EE, Love V, Josephson L, Lichtman AH, Weissleder R (2003) Cancer Res 63:6838–6846 Larson PEZ, Kerr AB, Chen AP, Lustig MS, Zierhut ML, Hu S, Cunningham CH, Pauly JM, Kurhanewicz J, Vigneron DB (2008) J Magn Res 194:121–127

192 Lauffer R, Mc Murry TJ, Dunham SO et al (1997) PCT Int Appl WO9736619 Lauffer R, Parmelee D, Dunham S et al (1998) Radiology 207:529–538 Levin YS, Mayer D, Yen Y-F, Hurd RE, Spielman DM (2007) Magn Res Med 58:245–252 Lowe MP, Parker D, Reany O et al (2001) J Am Chem Soc 123:7601 Magnusson P, Johansson E, Månsson S, Petersson JS, Chai C-M, Hansson G, Axelsson O, Golman K (2007) Magn Res Med 57:1140–1147 Mansson S, Johansson E, Magnusson P, Chai C-M, Hansson G, Petersson JS, Stahlberg F, Golman K (2006) 13C imaging – a new diagnostic platform. Eur Radiol 16:57–67 Martina MS, Fortin JP, Menager C, Clement O, Barratt G, Grabielle-Madelmont C, Gazeau F, Cabuil V, Lesieur S (2005) J Am Chem Soc 127:10676 Merritt ME, Harrison C, Kovacs Z, Kshirsagar P, Malloy CR, Sherry AD (2007a) J Am Chem Soc 129: 12942–12943 Merritt ME, Harrison C, Storey C, Jeffrey FM, SherryAD MCR (2007b) Proc Natl Acad Sci USA 104:19773–19777 Moller HE, Chen XJ, Saam B, Hagspiel KD, Johnson GA, Altes TA, de Lange EE, Kauczor HU (2002) Magn Res Med 47:1029–1051 Morawski AM, Winter PM, Yu X, Fuhrhop RW, Scott MJ, Hockett F, Robertson JD, Gaffney PJ, Lanza GM, Wickline SA (2004) Magn Reson Med 52:1255–1262 Mulder WJM, Castermans K, van Beijnum JR, oude Egbrink MGA, Chin PTK, Fayad ZA, Löwik CWJM, Kaijzel EL, Que I, Storm G, Strijkers GJ, Griffioen AW, Nicolay K (2009) Angiogenesis 12:17–24 Muller RN, Vander Elst L, Roch A, Peters JA, Csajbok E, Gillis P, Gossuin Y (2005) Adv Inorg Chem 57:239 Neubauer AM, Myerson J, Caruthers SD, Hockett FD, Winter PM, Chen J, Gaffney PJ, Robertson JD, Lanza GM, Wickline SA (2008) Magn Reson Med 60:1066–1072 Nivorozhkin AL, Kolodziej AF, Caravan P et al (2001) Angew Chemie Int Ed 40:2903 Norek M, Kampert E, Zeitler U, Peters JA (2008) J Am Chem Soc 130:5335–5340 Noth U, Grohn P, Jork A, Zimmermann U, Haase A, Lutz J (1999) Magn Reson Med 42:1039–1047 Partlow KC, Chen J, Brant JA, Neubauer AM, Meyerrose TE, Creer MH, Nolta JA, Caruthers SD, Lanza M, Wickline SA (2007) FASEB J 21:1647–1654 Pikkemaat JA, Wegh RT, Lamerichs R, van de Molengraaf RA, Langereis S, Burdinski D, Raymond AY, Janssen HM, de Waal BF, Willard NP, Meijer EW, Grull H (2007) Contrast Media Mol Imaging 2:229–239 Raghunand N, Howison C, Sherry AD et al (2003) Magn Reson Med 49:249 Rinck PA (1993) Magnetic resonance in medicine. Blackwell, Oxford Rinck PA (2003) Magnetic resonance in medicine. ABW Wissenschaftsverlag GmbH, Berlin Rosen BR, Belliveau JW, Aronen HJ, Kennedy D, Buchbinder BR, Fischman A, Gruber M, Glas J, Weisskoff RM, Cohen MS, Hochberg FH, Brady TJ (1991) Magn Reson Med 22:293 Rummeny EJ (1997) Marchal Acta Radiol 38:626

E. Gianolio et al. Saleh A, Schroeter M, Jonkmanns C, Hartung HP, Modder H, Jander S (2004) Brain 127:1670 Schlemmer HP, Becker M, Bachert P, Dietz A, Rudat V, Vanselow B, Wollensack P, Zuna I, Knopp MV, Weidauer H, Wannenmacher M, van Kaick G (1999) Cancer Res 59:2363–2369 Schmitt-Willich H, Brehm M, Evers CLJ et al (1999) Inorg Chem 38:1134 Schroder L, Lowery TJ, Hilty C, Wemmer DE, Pines A (2006) Science 314:446–449 Schroeder MA, Cochlin LE, Heather LC, Clarke K, Radda GK, Tyler DJ (2008) Proc Natl Acad Sci USA 105:12051–12056 Schroeder MA, Atherton EJ, Cochlin LE, Clarke K, Radda GK, Tyler DJ (2009) Magn Res Med 61:1007–1014 Schumann Giamperi G (1993) Invest Radiol 28:753 Sipkins DA, Cheresh DA, Kazemi MR, Nevin LM, Bednarski MD, Li KCP (1998) Nat Med 4:623–626 Spence MM, Rubin SM, Dimitrov IE, Ruiz EJ, Wemmer DE, Pines A, Yaoi SO, Tiani F, Schultzi PG (2001) Proc Natl Acad Sci USA 98:10654–10657 Srinivas M, Morel PA, Ernst LA, Laidlaw DH, Ahrens ET (2007) Magn Reson Med 58:725–734 Swanson SD, Rosen MS, Agranoff BW, Coulter KP, Welsh RC, Chupp TE (1997) Magn Reson Med 38:695–698 Terreno E, Delli Castelli D, Cravotto G, Milone L, Aime S (2004) Invest Radiol 39:235–243 Terreno E, Cabella C, Carrera C, Delli Castelli D, Lanzardo S, Mazzon R, Rollet S, Visigalli M, Aime S (2007a), Proc Joint Annual Meeting ISMRM-ESMRMB, Berlin, Germany, p. 248 Terreno E, Cabella C, Carrera C, Delli Castelli D, Mazzon R, Rollet S, Stancanello J, Visigalli M, Aime S (2007b) Angew Chem Int Ed 46:966–968 Terreno E, Sanino A, Carrera C, Delli Castelli D, Giovenzana GB, Lombardi A, Mazzon R, Milone L, Visigalli M, Aime S (2008a) J Inorg Biochem 102:1112–1119 Terreno E, Barge A, Beltrami L, Cravotto G, Delli Castelli D, Fedeli F, Jebasingh B, Aime S (2008b) Chem Commun 7:600–602 Terreno E, Delli Castelli D, Milone L, Rollet S, Stancanello J, Violante E, Aime S (2008c) Contrast Media Mol Imaging 3:38–43 Tweedle MF (1989) In: Bünzli JCG, Choppin GR (eds) Lanthanide probes in life, chemical and earth sciences: theory and practice. Amsterdam, Elsevier Uggeri F, Aime S, Anelli PL et al (1995) Inorg Chem 34:633 Vasalatiy O, Gerard RD, Zhao P, Sun X, Sherry AD (2008) Bioconjug Chem 19:598–606 Ward KM, Balaban RS (2000) Magn Reson Med 44:799–802 Ward KM, Aletras AH, Balaban RS (2000) J Magn Res 143:79 Weinmann HJ, Mühler A, Radüchel B (2000) In: Young IR (ed) Biomedical magnetic resonance imaging and spectroscopy. Wiley, Chichester Weissleder R, Mahmood U (2001) Radiology 219:316–333 Winter PM, Caruthers SD, Kassner A, Harris TD, Chinen LK, Allen JS, Lacy EK, Zhang HY, Robertson JD, Wickline SA, Lanza GM (2003) Cancer Res 63:5838 Winter PM, Cai K, Chen J, Adair CR, Kiefer GE, Athey PS, Gaffney PJ, Buff CE, Robertson JD, Caruthers SD, Wickline SA, Lanza GM (2006) Magn Reson Med 56:1384–1388

13 MR Contrast Agents Woessner DE, Zhang S, Merritt ME, Sherry AD (2005) Magn Res Med 53:790 Wolber J, Rowland IJ, Leach MO, Bifone A (1999) Magn Reson Med 41:442–449 Wolber J, Cherubini A, Leach MO, Bifone A (2000) Magn Reson Med 43:491–496 Woods M, Woessner DE, Zhao P, Pasha A, Yang MY, Huang CH, Vasalitiy O, Morrow JR, Sherry AD (2006) J Am Chem Soc 128:10155–10162

193 Zandt R et al (2009) Abstracts n. 4380 and 4381, 17th ISMRM Scientific Meeting & Exhibition, Honolulu, Hawaii (USA), April 18–24 Zhang SR, Wu KC, Sherry AD (1999) Angew Chemie Int Ed 38:3192 Zhang S, Winter P, Wu K, Sherry AD (2001) J Am Chem Soc 123:1517–1518 Zhang ZD, Greenfield M, Spiller TJ, McMurry RB, Lauffer RB, Caravan P (2005) Angewandte Chemie Int Ed 44:6766

MR Spectroscopy

14

Markus Becker

Contents 14.1 Introduction............................................................ 195 14.1.1 The New Contrast..................................................... 195 14.2 Applications............................................................ 195 14.2.1 What to Expect from Spectroscopy?........................ 196 14.2.2 Brain and Liver......................................................... 196 14.2.3 Muscle and Subcutaneous Tumor............................ 197 14.3 MR Sequences and Spectrum Processing............ 199 14.3.1 PRESS...................................................................... 200 14.3.2 STEAM.................................................................... 200 14.3.3 ISIS........................................................................... 201 14.3.4 CSI............................................................................ 201 14.3.5 EPSI.......................................................................... 202 14.3.6 FID........................................................................... 202 14.4 Processing................................................................ 202 14.4.1 Time Domain Processing......................................... 202 14.4.2 Frequency Domain Processing................................. 202 14.5 MR Coils................................................................. 203 14.5.1 Transmit Coil............................................................ 203 14.5.2 Receive Coil............................................................. 203 14.6 Examination............................................................ 204 References............................................................................ 205

14.1 Introduction Spectroscopy can be a very useful tool in a variety of purposes; unfortunately, we are still waiting for the final breakthrough in clinical applications. Therefore, nuclear magnetic resonance spectroscopy (NMRS) is mainly used in preclinical research facilities. This chapter tries to explain where spectroscopy is a helpful and strong tool and to give guidance for successful work.

14.1.1 The New Contrast To get the best result out of a spectrum, NMRS should be considered as an additional contrast beside standard MR imaging examinations like T1, T2, and proton density (PD) measurements. For a successful recording and analysis of a spectrum, the imaging information is needed not only just for the placement of the voxel, but also to get the best spectroscopy result, after collecting and analyzing all anatomic information. For diagnostic purposes or for monitoring of treatments, the only need is to define the area of interest, for example, necrotic regions of a tumor.

14.2 Applications M. Becker Bruker BioSpin MRI GmbH, Rudolf-Plank-Strasse 23, Ettlingen, Germany e-mail: [email protected]

For a beginner in the field of NMRS, it is difficult to choose the right examination protocol for a specific application or research project, because there are various methods and sequences available which have

F. Kiessling and B.J. Pichler (eds.), Small Animal Imaging, DOI: 10.1007/978-3-642-12945-2_14, © Springer-Verlag Berlin Heidelberg 2011

195

196

M. Becker

different pros and cons depending mainly on the tissue type. In general, the sequences can be separated in localized, nonlocalized, and imaging comparable techniques (spectroscopic imaging).

14.2.2 Brain and Liver

Localized:

The main examination pathway for brain and liver models is shown in Fig. 14.1. This diagram leads to the preferred examination method.

• Point resolved spectroscopy (PRESS) • Stimulated echo acquisition mode (STEAM) • Image-selected in vivo spectroscopy (ISIS)

14.2.2.1 Brain

Nonlocalized: • Free induction decay (FID) Spectroscopic imaging (SI): • Chemical shift imaging (CSI-FID) • Hybrid chemical shift imaging (CSI-PRESS, CSISTEAM) • Echo planar spectroscopic imaging (EPSI) Table 14.1 can be used as a starting point to assume quality and difficulty to achieve useful results.

14.2.1 What to Expect from Spectroscopy? The examiner has to choose what kind of substance he wants to monitor. Is it a native signal or an exogenously applied chemical, like a drug or marker? What is the expected concentration and what is the chemical shift? For the different body parts, the following section will discuss few examples how to set up an examination protocol.

Examination of the brain is the easiest task for spectroscopy. Since the overall magnetic field homogeneity is good, automatic shimming routines are available enabling proper shimming (Fig. 14.2). In a brain tumor model NMRS allows to monitor the membrane metabolism of the tumor cells by detection of the choline signal. The concentration of the choline is approximately 2 mM and the chemical shift is 3.24 ppm. Absolute quantification of the choline is possible, but complicated (Dong et al. 2006). For tumor staging and longitudinal studies, it is in most cases sufficient to quantify choline by calculating the quotient I(choline)/I(creatine). The creatine concentration is almost independent of the activity of the tumor cells with one exception: High tumor grades develop a necrosis where the dead cells accumulate lipids. These necrotic areas should be identified by imaging methods prior to the spectroscopic experiments. The other possibility to identify the different active areas of the tumor is chemical shift imaging (CSI) with a metabolic map. However, CSI is very time-consuming and the spatial resolution is poor. As an example, a CSI from a mouse brain is shown in Fig. 14.3.

Table 14.1 Selection of measurement methods; quality and difficulty. Tissue/body part

1

H single voxel quality/difficulty

1

H SI quality/difficulty

X NMR quality/difficulty

Brain

PRESS good/easy

PRESS CSI good/easy

31

STEAM good/easy

STEAM CSI fair/easy

13

EPSI good/difficult

23

PRESS good/easy

PRESS CSI fair/easy

31

STEAM good/easy

STEAM CSI fair/easy

13

Muscle

Liver

PRESS good/easy

P good/easy C fair/difficult Na fair/difficult P good/easy C fair/difficult P good/fair

31

STEAM good/easy Tumor, subcutaneous

PRESS good/difficult

31

STEAM good/fair

13

P good/easy C good/easy

197

14 MR Spectroscopy SVS STEAM or PRESS longitudinal study

Tumor

grading with native substance choline,lactate,lipids

CSI PRESS

single exam Baseline spectrum necessary

SVS short TE STEAM, healthy and tumor tissue SVS long TE STEAM, healthy and tumor tissue

Brain

CSI PRESS,side comparison 31P Spectroscopy

drug monitoring

Animal Model

CSI FID,side comparison Formation and degradation of membrances can be monitored seperatly (PME/PDE).

depending on the structure of the drug 1H,31P,13C,19F all techniques possible

various diseases NAA, creatine,choline, glutamine,glutamate SVS STEAM or PRESS

31P Spectroscopy, NTP

Liver

31P FID 31P CSI FID

Energy metabolism 1H short TE STEAM glycogen drug monitoring

depending on the structure of the drug 1H,31P,13C,19F all tecniques possible

Fig. 14.1 Pathway, from animal model to preferred examination method for brain and liver

14.2.2.2 Liver Liver spectroscopy has not been performed frequently due to problems in reproducibility and quality of the spectra. A major problem is respiration, and in some areas, heart motion. Additionally, the high quantity of iron in the liver makes shimming difficult and the resulting signals are board. Respiratory gating combined with cardiac triggering is necessary to avoid motion artifacts. The aim of a liver spectroscopy is to study metabolism such as glycogen storage diseases and drug metabolism (Bollard et al. 2000). 31P, 13C, 1H, and 19F are therefore the most interesting nuclei where 31P is used to study the energy metabolism, 13 C and 19F are assessed for monitoring drug metabolism, and 1H for the development of metastasis (Fig. 14.4).

14.2.3 Muscle and Subcutaneous Tumor The main examination pathway for muscle and subcutaneous tumors is shown in Fig. 14.5, leading to the preferred examination method. 14.2.3.1 Muscle Muscle spectroscopy is important for studies focusing on mitochondrial diseases, enzymatic deficiencies, and energy metabolism. 31P spectroscopy is the most useful nucleus for muscle examinations, because it allows to measure a, b, g−NTP (nucleoside-5¢-triphosphate1), posphocreatine creatine, and free phosphate. The chemical It is not possible to separate the signals of adenosine from all other nucleobases, therefore NTP is the general term.

1

198

M. Becker Magnitude

Real

water

5

4

3

3

2

2

1

lipids

1

aromatic protons

lipids 0

0 ppm

ppm 8

6

4

2

4

0

3

2

1

NAA 1.5

1.5

tCr Cho 1.0

tCr

Glu Gln

1.0

µl

Glu

µl

0.5 0.5

Tau

NAA µl

0.0

0.0 ppm 3.0

2.5

ppm 4.0

2.0

3.8

3.6

3.4

3.2

Fig. 14.2 Rat brain spectrum (PRESS) from a healthy tissue. Up left: full spectrum. Other quadrants: spectrum magnified. Following signals can be identified: creatine and phospho-

creatine (tCr), N-acetylaspartate (NAA), glutamine (Glu), glucose (Glc), myo-inositole (mI), choline (Cho), taurine (Tau)

shift of the phosphate provides the pH of the muscle, calculated with the following formula (Barker et al. 1999; Petroff et al. 1985):

IMCL plays an important role in the development of type 2 diabetes, while EMCL level is relatively constant (Pan et al. 1997; Renema et al. 2003) (Fig. 14.6).

pH = pK + log

δ H PO − δ P 2

4

δ P − δ HPO 2 4

= 6.803 + log

5.73 − δ P . δ P − 3.22

Additionally, it is possible to calculate the manganese (Mg2+) concentration by the shift from the NTP signals (Barker et al. 1999; Bottomley and Hardy 1989). With 1H spectroscopy, triglycerides can be identified and seperated in the triglycerides in intra and extramyocellular lipids (IMCL, EMCL) (NeumannHaefelin et al. 2003). It has been suggested that

14.2.3.2 Subcutaneous Tumor With an implanted tumor model, NMRS allows to study tumor growth and treatment effects. Sub cutaneous tumors are easy to access, and are therefore, ideal for X nuclei spectroscopy. The membrane and energy metabolism can be monitored with 31P NMRS, 13C labelled drugs and even with

199

14 MR Spectroscopy

Fig. 14.3 Mouse brain CSI-PRESS from a healthy animal with metabolic (lower left) and spectral map (upper right). The metabolic map is calculated from the fitted NAA signal (upper left).

The map has a centric symmetry, because of the combination of the imperfect pulse shape and interpolation (zerofilling)

spectroscopy, it is possible to examine treatment response (Fig. 14.7). With 31P spectroscopy, it is possible to monitor the energy and membrane metabolism of the tumor simultaneously. A highly active tumor produces PE and PC to build up membranes. During successful treatment, the amount of metabolites of the membrane degradation increases (GPC, GPE) (Podo 1999). For quantification of the signals of a subcutaneous implanted tumor, an external reference can be attached easily. Nevertheless, absolute quantification is difficult (Street et al. 1997).

14.3 MR Sequences and Spectrum Processing The number of different sequences or methods is very large. The usage of the expressions “method” and “sequence” is vendor specific, but did not diverge in substance. A sequence is the chronology of radio frequency (rf) pulses, gradients, and delays and receive events during the measurement. The most important sequences are PRESS, STEAM, FID, CSI, EPSI, and ISIS. A complete list can be found at de Graaf (1998), In vivo NMR Spectroscopy.

200

M. Becker

14.3.1 PRESS

-CH2-

-C=C-CH2-

CH3-

-CH2-CO-

GIc

This single voxel sequence is based on three spatially localized pulses, one 90° excitation and two 180° refocusing pulses. PRESS is the technique with the highest S/N, but has some limitations. The minimum echotime is bigger than with STEAM, therefore the number of detectable metabolites is limited. The 180° pulse has a great demand on the power amplifier to minimize the chemical shift artifact which increases linear with field strength. In general, when going to higher field strength more amplifier power is needed. To optimize the shape of the voxel, it is necessary to use saturation regions, called outer volume suppression. This technique is available from all vendors, the measurement is easy, and the results are good in at least 70% of the examinations.

CHo -C=C-CH2ppm 4

3

2

1

14.3.2 STEAM

Fig. 14.4 1H spectrum of a mouse liver (STEAM). Mainly signals of triglycerides are visible. The multiplet of glucose (Glc) cannot be resolved

Energy metabolism Muscle

Three 90° spatially localized pulses are used to generate a stimulated echo. The advantages to PRESS are a smaller minimum echotime, less chemical shift artifacts, and that

31P Spectroscopy, 31P FID PCr,NTP 31P CSI FID 13C Spectroscopy, 13C FID gycogen,glucose 13C CSI FID

lipid metabolism, intra- extracellular SVS STEAM or PRESS

Animal Model drug monitoring

Subcutaneous tumor

depending on the structure of the drug 1H,31P,13C,19F all techniques possible

longitudinal study

SVS STEAM or PRESS

grading with native substance single exam Baseline spectrum necessary SVS short TE STEAM and long TE PRESS, choline,lactate,lipids different types of tumor tissue, prolifferent and necrotic tissues 31P Spectroscopy, FID or CSI FID Formation and degradation of membrances can be monitored sepeately (PME/PDE).

Fig. 14.5 Pathway leading to a preferred examination method for nuclear magnetic resonance spectroscopy of muscles and subcutaneous tumors

201

14 MR Spectroscopy [rel]

one localized spectra. This sequence is one of the oldest techniques to get localized spectra (Ordidge et al. 1986). For proton spectroscopy, ISIS is replaced by PRESS and STEAM, but for X nuclei it is still useful and the first choice for localization. The 180° pulse has a great demand on the power amplifier, the pulse is therefore very long, resulting in a small excitation bandwidth. Only a small part of the X-nucleus chemical shift scale can be measured localized.

CH2(EMCL)

Water 12 10 Cre 8 6 Cre

4

CH3 Tau

CH2(EMCL)

2 0

14.3.4 CSI 5

3

2

1

[ppm]

Fig. 14.6 1H spectrum of a rat thigh (PRESS). In this case only signals of the extramyocellular lipids are detected (EMCL). In 1 H spectra, it is possible to separate creatine from phosphocreatine (3.85 and 3.9 ppm)

outer volume suppression is not necessary. The main drawback is the reduced S/N which is only 50% of the PRESS sequence. For a very precise definition of a voxel, the STEAM sequence is more appropriate; however, for a high S/N, the PRESS method is superior.

14.3.3 ISIS A combination of eight experiments with three 180° slice selective pulses and one 90° pulse is leading to

CSI is a combination of spectroscopy and imaging to achieve localized spectra in a whole slice. CSI can be used in two or three dimensions. It is possible to calculate images from various signals to get metabolic information of the examined tissue. In proton CSI the sequence is a combination of a large voxel, prepared with PRESS or STEAM with phase encoding gradients to divide this large volume into smaller voxels. In each voxel a full spectrum is measured. An advantage of this technique is that one gets a large number of localized spectra is obtained (in a short time). This is very helpful for comparisons between healthy and diseased regions within one subject. The disadvantage is that the voxels are not well defined, due to the point spread function. The point spread function describes how a strong signal in one voxel is also visible in the neighbored voxels (de Graaf 1998).

PE α-NTP

PC Pi

γ-NTP

β-NTP NADPH

GPE

10

5

GPC

PCr

0

Fig. 14.7 31P NMR spectrum of an subcutaneous implanted human breast cancer (1H decoupled, FID). Signals are: O-phosphoryl ethanolamine (PE), O-phosphoryl choline (PC), phosphate (Pi), 1-glyceryl phosphoryl ethanolamine (GPE),

-5

-10

-15

δ/ppm

1-glyceryl phosphoryl choline (GPC), phosphocreatine (PCr), nucleosid-5¢-triphosphate (a-,b-,g-NTP), Nicotinacid amid dinucleotidphosphate (NADPH)

202

14.3.5 EPSI EPSI is derived from an EPI sequence where the k-space is filled with an EPI readout. The advantage of EPSI is the high S/N and the option to measure a small voxel size. The main drawback is the processing of the spectra. The vendors do not support the processing of the native EPSI data, so in most of the cases the user needs some special home-built software to view and analyze the spectra.

14.3.6 FID The FID or pulse-acquire sequence is mainly used for X-nucleus spectroscopy. For 1H spectroscopy, it is only used when the water concentration is low, because of the not sufficient water suppression in most of the applications. FID is the technique with the highest S/N, but with no localization. The localization of the signal can be done with the choice of the coil. A surface coil has a limited sensitive volume, which can be used to acquire spectra of a specific organ, e.g., liver, muscle, subcutaneous tumor.

14.4 Processing The type of processing of the spectrum is not as important as one may think. However, the most important thing is that processing is always done in the same way, so that you can rely on it and compare data from different subjects of imaging sessions. The processing steps affect the signal intensities and the resulting area calculation. The processing steps can be divided in two parts, the time and frequency domain:

14.4.1 Time Domain Processing (a) Deleting/adding of time domain points Sometimes the first complex data points of the X-nucleus FID are corrupted from insufficient hardware, resulting in broad background signals and phase distortions. Unfortunately, the first time domain point is responsible for the intensity of the whole spectrum, so by deleting points, the S/N is going down. To

M. Becker

minimize this effect linear prediction can be used to restore the missing points. (b) Zerofilling and linear prediction Zerofilling is just adding points with the intensity zero to the FID. After the Fourier transformation (FFT), the spectrum is higher resolved, similar to an interpolation in images. Because zerofilling is not the real measured FID, the number of added zero-points is limited. In practice, measured points can be doubled or quadrupled the number of measured points, e.g., 1,024 recorded complex data points can be filled up to 4,096 points without affecting the spectrum too much. However, the better way than zerofilling is linear prediction. With linear prediction, points can be added to the FID, simulating measured data. (c) Filter functions After zerofilling an intensity jump is often observed between the measured FID and the added points. In the spectrum this appears as ringing artifacts at the foot of the signals, comparable to Gibbs ringing in imaging. To smooth these artifacts, the FID is multiplied with a filter function. The basic filter function is the exponential function; Gaussian and Hanning filter influences the resulting resolution and S/N more precisely. (d) Eddy current compensation Localized spectra are often affected by eddy currents, because of the necessary gradients, especially at small voxel sizes. The water signal can be used to correct these distortions. (e) Frequency correction The B0 field of the magnet is changing during time, resulting in a mild shift of the signals during the recording of the spectrum with more than one acquisition. The resulting spectrum has therefore broader signals than necessary. By correction of this magnet shift, the summarized spectrum has signals with a smaller half width. Again, different schemes to compensate this effect are implemented from the vendors (Fig. 14.8).

14.4.2 Frequency Domain Processing (a) Phase correction After FFT one gains a complex spectrum with signals having a 90° phase shift between the real and imaginary part of the spectrum. By

203

14 MR Spectroscopy

multiplying the spectrum with a constant complex phase factor, the phase of the signal is changed to display the peak in absorption mode. The absorption mode means that all signals are on one side of the baseline. If there is more than one signal, there is often a phase difference between the signals which can be corrected by multiplying the spectrum with a linear complex phase factor. If it is not feasible to get all signals to the absorption mode, a quadratic phase correction can be applied. (b) Baseline correction The baseline of the spectrum is in most of the cases not straight because of large signals in the neighborhood of the target signal, macro molecules with broad resonances, signals from the housing of the coil, and the support material where the animal is placed on. A baseline correction is done by fitting a spline curve to the spectrum and subtract the curve from it. (c) Curve fitting To analyze the spectrum, it is necessary to calculate the signal intensities and the area under the signal. The area under the signal reflects the number of NMR detectable molecules in the measured volume. For calculation of the area, time, and frequency domain fitting routines are available. Time domain algorithms are more precise and phase- and baseline correction-independent. There are freeware fitting processing software packages available. Frequency domain fitting routines are not very reliable, but in most of the cases sufficient for a quick analysis of the spectrum, e.g., for tumor staging. The variations of the native signal intensities between

individual animals are much greater than the fitting error. For detecting and quantifying the signal of a drug, a time domain fit is recommended (Fig. 14.9).

14.5 MR Coils 14.5.1 Transmit Coil In general, it is possible to use the same coils for 1H spectroscopy as for imaging; however, it is recommended to use a coil with a homogeneous B1 field for transmit of the RF pulse. This leads to a more precise definition of the flip angles than with a surface transmit coil. The drawback is the higher power demand and the higher RF deposit in the animal so that the animal can heat up. This is especially important for X-nucleus application with 1 H decoupling. Circular polarized coils are advantageous over linear polarized because of their lower power consumption and bigger homogeneous B1 volume.

14.5.2 Receive Coil To receive the signal, it is important that the coil picks up the signal with minimum signal loss. A multichannel receive coil offers the highest possible S/N. The signal of each element is combined to a summarized and phase-corrected spectrum. X-nucleus coils are usually not designed as multichannel coils, only as linear or circular polarized coils. They are mainly combined transmit/receive coils. 6

filter function, gaussian

200

4

100 2

0

signal in absorption mode Phase correction necessary

zerofilling

0

-100

baseline correction

0.0

0.5

1.0

ms x10 ^ 3

Fig. 14.8 Time domain processing. Free induction decay with filter function (blue) and zero filling

ppm 6

5

4

3

2

Fig. 14.9 Frequency domain processing. Spectrum with phase and baseline correction (blue)

204

M. Becker

Examination Imaging T1/T2/ contrast media / diffusion

Identify interesting area, e.g.tumor, inflammation,ischemia

Single Voxel

CSI

FID

Place Voxel in interesting area

Place Slice in suspicious area

Adjustments Shim/water suppression

Adjustments Shim/water suppression

Measurement repeat with changed parameters

Measurement repeat with changed parameters

Adjustments Shim

Measurement

repeat with changed parameters

choose interesting voxel Processing single spectrum

Processing spectral map/Metabolic Image

Processing single spectrum

Fig. 14.10 Flow chart of a spectroscopic examination. After imaging and placing the voxel or slice in the interesting area, frequent measurements with different parameters can be done

Again, circular polarized2 coils are preferred over linear coils due to higher S/N by factor of 1.41.

14.6 Examination In vivo examinations start with high resolution imaging to define the region for spectroscopy as good as possible. Therefore, it might be necessary to measure images with T1, T2 contrast and contrast media.3 After defining the region-of-interest, the voxel or slice is placed in this area. The voxel definition is not very accurate because Circular polarized coils are often named as quadrature coils.

2

of the chemical shift artifact in single voxel spectroscopy and the point spread function in CSI. After placement of the voxel, the frequency and amplifier adjustment has to be performed, following by shimming and water suppression (only 1H). The adjustments are either automatically done or require user interaction, depending on the system. It is often necessary to control the quality of the adjustments, especially the shim and the water suppression (Fig. 14.10). After all these steps it is possible to measure. The final processed spectrum is usuable in most of the cases and gives the missing information about the tissue composition. Washout of contrast media in animals is much faster than in humans resulting in no interference with a following spectroscopy exam.

3

14 MR Spectroscopy

References Barker PB, Butterworth EJ, Boska MD, Nelson J, Welch KMA (1999) Magnesium and pH imaging of the human brain at 3.0 tesla. Magn Reson Med 41:400–406 Bollard ME, Garrod S, Holmes E, Lindon JC, Humpfer E, Spraul M, Nicholson JK (2000) High-resolution 1H and 1H-13C magic angle spinning NMR spectroscopy of rat liver. Magn Reson Med 44:201–207 Bottomley PA, Hardy CJ (1989) Rapid, reliable in vivo assays of human phosphate metabolites by nuclear magnetic resonance. Clin Chem 35:392–395 de Graaf R (1998) In vivo NMR spectroscopy. Wiley, New York Dong Z, Dreher W, Leibfritz D (2006) Toward quantitative short-echo-time in vivo proton MR spectroscopy without water suppression. Magn Reson. Med 55:1441–1446 Neumann-Haefelin C, Kuhlmann J, Belz U, Kalisch J, Quint M, Gerl M, Juretschke P, Herling AW (2003) Determinants of intramyocellular lipid concentrations in rat hind leg muscle. Magn Reson Med 50:242–248

205 Ordidge RJ, Connelly A, Lohmann JAB (1986) Image-selected in Vivo spectroscopy (ISIS). A new technique for spatially selective nmr spectroscopy. J Magn Reson 66:283–294 Pan DA, Lillioja S, Kriketos AD, Milner MR, Baur LA, Bogardus C, Jenkins AB, Storlien LH (1997) Skeletal muscle triglyceride levels are inversely related to insulin action. Diabetes 46:983–988 Petroff OAC, Prichard JW, Behar KL, Alger JR, den Hollander JA, Shulman RG (1985) Cerebral intracellular pH by 31P nuclear magnetic resonance spectroscopy. Neurology 35:781–788 Podo F (1999) Tumour phospholipid metabolism. NMR Biomed 12(7):413–439 Renema WKJ, Schmidt A, van Asten JJA, Oerlemans F, Ullrich K, Wieringa B, Isbrandt D, Heerschap A (2003) NMR spectroscopy of muscle and brain in guanidinoacetate methyltransferase (GAMT)-deficient mice: validation of an animal model to study creatine deficiency. Magn Reson Med 50:936–943 Street JC, Szwergold BS, Matei C, Kappler F, Koutcher JA (1997) Study of the metabolism of choline and phosphatidylcholine in tumors in vivo using phosphonium-choline. MRM 3:769–775

15

Ultrasound Stuart Foster and Catherine Theodoropoulos

Contents 15.1 Introduction............................................................. 207 15.2 Microultrasound Imaging Technology.................. 208 15.3 Oncology Applications............................................ 210 15.4 Tumor Sizing and Quantification in 2- and 3-Dimensions............................................... 211 15.5

Evaluation and Quantification of Blood Flow........................................................... 212

15.6

Targeting Injections and Biopsies......................... 214

15.7 Contrast Agents for Performing Real-Time Tumor Perfusion and Targeted Molecular Imaging of Biomarkers........................................... 214 15.8 Cardiovascular Applications................................. 214 15.9

Discussion and Conclusion..................................... 215

References............................................................................ 216

S. Foster (*) Department of Medical Biophysics, Sunnybrook Health Sciences Centre, 2075 Bayview Avenue, Toronto, ON, M4N 3M5 Canada e-mail: [email protected] C. Theodoropoulos Scientific Applications, VisualSonics Inc, 3080 Yonge Street, Suite 6100, Toronto, ON, M4N 3N1 Canada

15.1 Introduction Over the past several decades, mouse models have been used in cancer and cardiovascular research to aid in the investigation of the basic biological underpinnings of disease and as a tool for discovering new clinical agents and assays. Murine models are the most commonly used basic science and preclinical animal systems in academic and pharmaceutical research. This is due to the significant similarities and the plethora of commonly shared inherited diseases between humans and mice. Some of these diseases include atherosclerosis, cancer, heart disease, hypertension, obesity, bleeding disorders, asthma, and neurological disorders. Notwithstanding, there are other mammals which also share this similarity. However, one of the major advantages of the mouse model is the wealth of resources available on its genetics, molecular, and cellular pathways. Furthermore, mice are ideal for research environments as they are small and easy to maintain, cost-effective, have short gestation periods (~19–20 days), and finally, genetically uniform inbred strains are readily obtained. Imaging has contributed substantially to small animal research by enabling quantitative visualization of the living biology-associated development, growth, and progression of disease. As in clinical imaging, no single imaging modality suits all biological applications. In fact, each imaging modality has its own strengths and weaknesses with respect to its temporal and spatial resolution, accessibility, sensitivity, ease of use, cost, the specific regions of the body, and biological processes that can be imaged. The availability of suitable contrast agents is also important for each modality. Often preclinical imaging is performed using a multimodality format to allow researchers a more

F. Kiessling and B.J. Pichler (eds.), Small Animal Imaging, DOI: 10.1007/978-3-642-12945-2_15, © Springer-Verlag Berlin Heidelberg 2011

207

208

complete understanding of disease processes and/or the efficacy of a preclinical therapy. Here, we will focus on micro-ultrasound as an imaging modality in preclinical research. Clinically, ultrasound is used in hospital departments for real-time imaging and quantification of soft tissue disease, cardiac function, fetal development and diagnosis, emergency medicine, screening and assessment of tumors, and as a tool for guidance of invasive procedures. The traditional strengths of ultrasound are its speed of image formation, high resolution, portability, and low cost compared to other modalities. In recent years, preclinical ultrasound systems (“micro-ultrasound”) have been developed that provide the functionality available clinically, yet allow for ultrahigh spatial resolution (30–150 mm) in order to view and quantify the detailed structures, flow dynamics, and molecular biomarkers inherent in murine microimaging. In addition, these systems have been developed to minimize perturbation of the mouse, rat, or other small animal model while maximizing reproducibility of both the procedure and the data derived. Several hundred peer-reviewed papers using the single element scanned micro-ultrasound platform have now been published. In the area of cardiovascular research (see, for example, Liu et al. (2007), Lee et al. (2008), Trivedi et al. (2007), Ino et al. (1996), Zhou et al. (2003, 2004), and in the area of cancer (see, for example, Goessling et al. (2007), Kiguchi et al. (2007), Olive and Tuveson (2006), Wirtzfeld et al. (2005, 2006), Xuan et al. (2007)). In this chapter the basic technology of micro-ultrasound will be reviewed. Scanning techniques, animal handling, and applications will be described in the context of relevant biological problems.

15.2 Microultrasound Imaging Technology There is only one manufacturer of dedicated pre clinical high-resolution micro-ultrasound systems (VisualSonics, Toronto, Canada). These systems operate at frequencies between 15 and 70 MHz compared to frequencies of 3–15 MHz for clinical systems. Both mechanical and array technologies are currently available for ultrasound microimaging. A block diagram of the latest array system is illustrated in Fig. 15.1. It shows a 256 element transducer in combination with a

S. Foster and C. Theodoropoulos

64 element beamformer. A high speed bus transfers the multiplexed transducer aperture data to processing systems for calculation of B-mode and Doppler images. Figure 15.2 shows the consoles and scanheads for both the mechanical (Fig. 15.2a, b) and array (Fig. 15.2c, d) imaging systems. Each system is connected to a series of application-specific scanheads, (Fig. 15.2b, d) which transmit and receive high-frequency ultrasound signals that generate the real-time images. Operation of the scanner is illustrated in Fig. 15.3. As Fig. 15.3 shows, the scanhead is configured on an imaging platform called a rail system to allow for hands-free positioning and capture of the images. The animal is anesthetized and maintained on a heated platform to ensure the comfort and maintain body temperature of the animal during imaging. Key physiological parameters are captured from the platform including heart rate, temperature, respiration, and ECG. These signals are gathered through paw and temperature probes and integrated with the real-time micro-ultrasound images. An additional accessory to this Imaging Station is an injection mount that allows for the guidance of precise injections or biopsies in vivo through the real-time imaging of the micro-ultrasound system. Figure 15.3b shows a typical abdominal cross-section in a normal mouse. Visible structures include the right kidney, adrenal gland, liver, and a variety of vascular structures. Figure 15.4 shows a cross-section through an orthotopic hepatocellular carcinoma pre and postcontrast injection showing the increase in signal intensity caused by the contrast agent (discussed in the Chap. 16). The analytic software is designed to allow for study-based aggregation of animal data for longitudinal experiments with multiple animals, multiple time points, and multiple therapeutic interventions. Over 500 disease-specific measurements are implemented within the software to allow for quantification and analysis of the images. Two-dimensional images are gathered and viewable in real time without any reconstruction or delay, while 3D images are acquired rapidly and reconstruction into a 3D image is done in seconds by the software. Animal physiological data can be stored with the imaging data for cardiotoxicity studies, for example. All images and measurements can easily be exported to databases for further analysis or archiving. The software is networked for high-throughput studies or for use in core imaging facilities where multiple researchers use the same instrument and can quickly image their animals

209

15 Ultrasound

Main computer Image display processor

Ultrasound signal processor

Data memory

Main software application

Beamformer controller

Tx channel 2

Tx channel 64

Rx channel 1 +

Rx channel 2

Rx channel 64

Transducer connector

Receive multiplexer

Tx channel 1

Transmit multiplexer

Display

Transducer cable

Front panel & keyboard

256 element array transducer

Fig. 15.1 Block diagram of the Vevo 2100 scanner architecture. The System consists of a 256 element transducer coupled to a 64 element beamformer. A high speed data connection to the main

processing computer enables B = mode and Doppler data to be computed in real time and displayed on the system monitor

and then review their data and perform measurements over the network. High-frequency mechanical sector scanning for mice was originally developed by Foster et al. (2002). The mechanical scanhead shown in Figs. 15.2b has been optimized with sophisticated low inertia actuation and is capable of frame rates of over 100/s. The 15–55 MHz transducer executes a shallow sector scan over about a 15-mm field of view in these devices. The single element transducers used in the mechanical probes are composed of PVDF copolymer devices configured as described by Sherar and Foster (1989) and Foster (2000) and optimized to achieve high sensitivity and bandwidth. Foster et al. introduced highfrequency linear arrays for mouse imaging in 2008 (Foster et al. 2009; Foster 2008). These scanheads (Fig. 15.2d) contain no moving parts, but rather create

the ultrasound image through careful orchestration transmit and receive beamforming that is well de scribed in numerous publications (Cobbold 2007). The choice of scanhead frequency for either the mechanical or array systems is driven by the required resolution and depth of penetration. Table 15.1 illustrates the expected resolution for various ultrasound frequencies. Resolution in the axial (depth) direction is approximately equal to the wavelength of ultrasound where as in the two lateral directions resolution is about twice the wavelength. Resolution is typically between 30 and 150 mm depending on the choice of frequency. Total depth of penetration is on the order of 30 mm at 20 MHz and is reduced to <10 mm at 50 MHz. A thorough description of resolution and depth of penetration for high-frequency imaging are given in (Foster et al. 2009).

210

S. Foster and C. Theodoropoulos

a

b

c

d

Fig. 15.2 Consoles for mechanical (a) and array-based (c) high-frequency ultrasound imagers. Corresponding probes for the two systems are given in (b) and (d), respectively

15.3 Oncology Applications For cancer research applications, micro-ultrasound is an imaging modality that, because of its high spatial resolution and real-time temporal resolution, can be used to visualize, characterize, and quantify prepalpable orthotopic and subcutaneous tumors in a multitude of small animal models. Tumors can be monitored and quantified from tumorgenesis through their growth stages and through to metastases to surrounding organs, lymph nodes, and tissue. Cheung et al. (2005) describe the use of micro-ultrasound in the assessment of xenograft growth analysis. The current software allows

real-time visualization and measurement of the tumors in 2- and 3-dimensions. Blood flow, blood architecture, and assessment of feeder vessels can be quantified using Power Doppler (for vessels larger than 30 mm) and using contrast agents for tumor perfusion and microcirculatory flow in vivo. Early applications of flow analysis and measurement in tumor growth are provided by Franco et al. (2006), Shaked et al. (2006) and Cheung et al. (2007). Xuan et al. (2007) described neoangiogeneic development and tumor blood flow in prostate tumors including visualization of each of the early stages of such development can be clearly visualized and quantified in longitudinal studies.

211

15 Ultrasound

a

Fig. 15.3 Linear array scanhead positioned on rail system for imaging (a). The mouse is monitored for temperature, ECG, and other physiologically relevant parameters, while the operator

a

b

makes instrument adjustments during a scanning session. (b) A still frame from a typical abdominal cross-section of a normal mouse

b

Fig. 15.4 Images of a hepatocellular tumor pre (a) and post (b) contrast injection

15.4 Tumor Sizing and Quantification in 2- and 3-Dimensions One of the most important measurements in oncology is quantifying the change in tumor size during disease progression and changes related to tumor response to anticancer therapy (Wirtzfeld et al. 2005; Feldmann et al. 2008; Wang et al. 2007). Because micro-ultrasound is noninvasive, tumor growth and changes can be monitored repeatedly and longitudinally in the same animal, which can serve as its own control, thereby

increasing accuracy of the experiment and reducing the number of cohort animals required. Microultrasound is routinely used in serial 2D and 3D volumetric quantification of tumor sizing in vivo in a variety of murine and nonrodent cancer models (Goessling et al. 2007; DeRosier et al. 2007; Graham et al. 2005; Huizen et al. 2005; Kiguchi et al. 2005; Lyshchik et al. 2007; Wu et al. 2005). For example, Wirtzfeld et al. 2005 were the first to publish on the use of micro-ultrasound to track 3D tumor volumes in progression in a transgenic prostate cancer mouse model.

212

S. Foster and C. Theodoropoulos

Table 15.1 Resolution and other relevant parameters vs. frequency for ultrasound imaging Frequency (MHz)

Wavelength (mm)

Axial resolution (mm)

Lateral resolution (mm)

Max imaging depth (mm)

1

1.5

1.5

3

450

5

0.3

0.3

0.6

90

10

0.15

0.15

0.3

45

20

0.075

0.075

0.15

22.5

50

0.03

0.03

0.06

9

100

0.015

0.015

0.03

4.5

3D micro-ultrasound images correlated closely to serial histology (a correlation coefficient of 0.998 (p < 0.001)). Furthermore, 3D micro-ultrasound measurements accurately confirmed the size and shape of these tumor masses in vivo (Wirtzfeld et al. 2005). The technique is highly reproducible as tumor detection sensitivity and specificity were both >90% when diagnoses were based on repeated micro-ultrasound examinations performed on separate days. Further studies went on to show the utility of 3D micro-ultrasound to not only noninvasively track the growth of liver metastases (tumor diameter, volume, and growth curve), but also to evaluate potential chemotherapeutics on these parameters in a longitudinal murine metastases model (Graham et al. 2005). In this particular study the authors showed high accuracy in detecting tumor mass progression using four different cell lines; B16F1 (murine melanoma), PAP2 (murine H-ras–transformed fibroblast), HT-29 (human colon carcinoma), and MDA-MB-435/HAL (human breast carcinoma) in a liver metastases model. Furthermore, the longitudinal imaging of B16F1 liver metastases revealed distinct textural changes which had anechoic regions found to be areas of liquefactive necrosis. The tracking of tumor volume and textural changes can be a useful tool in tracking subtle responses to different antiangiogenic therapeutic treatments. More importantly, this study illustrated that longitudinal microultrasound imaging is a powerful tool in preclinical trials and drug development since liver metastases treated with Doxorubicin, a cytotoxic chemotherapeutic agent, were found to have significant decreases in tumor volume at day 12 post cell injection. The authors concluded that micro-ultrasound improved the evaluation of therapeutic efficacy on sequential stages of tumor development (Graham et al. 2005).

15.5 Evaluation and Quantification of Blood Flow A key component of studying cancer in animal models is understanding how specific drugs affect angiogenesis, tumor growth, and metastases. As such, microimaging modalities that provide insight into how a drug influences tumor vasculature and molecular expression of specific biomarkers, such as those involved in angiogenesis, have become important life science research tools. Micro-ultrasound provides such utility, allowing for the mapping and visualization of tumor vasculature using the Power Doppler which can detect blood flow ranging from 2 mm/s up to 4 m/s in tumors at multiple time points (Xuan et al. 2007; Goertz et al. 2002; Jugold et al. 2008). Power Doppler can be used to detect subtle changes in tumor perfusion and blood vessel architecture in a noninvasive longitudinal manner (unpublished observation Xuan et al. 2006 Robarts Research Insitute, London, Canada). Studies by Goertz et al. (2002) reported the first use of high-frequency micro-ultrasound 2D Power Doppler in studying the effects of an antivascular drug on blood flow in superficial human melanoma MeWo tumors. The authors reported a significant reduction in blood flow 4 h after injection of the tumor vascular targeting agent ZD6126 followed by a recovery of flow by 24 h after injection. They concluded that high-frequency Power Doppler was a highly effective noninvasive quantitative tool for longitudinally following the effects of antivascular therapy on blood flow in superficial tumors (Goertz et al. 2002). Similarly Xuan et al. (2007) recently reported the first application of high-frequency 3D Power Doppler micro-ultrasound imaging in a genetically engineered mouse prostate cancer model. They showed that 3D Power Doppler could sensitively, specifically, and reproducibly depict

15 Ultrasound

functional neoangiogenic blood flow in prostate tumors when compared to normal prostate tissue which had little or no flow. These observations were confirmed using micro-CT and by correlation with microvessel distributions measured by immunohistochemistry and enhanced vascularity visualized by confocal microscopy. Figure 15.5 shows examples of the correlation

Fig. 15.5 Comparison of 3-dimensional power Doppler (left) and 3-dimensional micro-CT (right) for prostate tumors at early (top), middle (centre), and late (bottom) stages of development. Arrows point to common features in the vascular patterns. Bar = 1 mm. (Image from Xuan et al. (2007 with permission)

213

between micro-ultrasound and micro-CT from this study. A separate study from Jugold et al. (2008) investigated the effects of blocking VEGF-mediated pathways using a VEGFR2-blocking antibody treatment. This study, using Power Doppler imaging, showed that after 6 days of treatment in subcutaneous tumor (spontaneously immortalized human skin keratinocytes) in

214

nude mice, tumor vascularity significantly decreased. These findings were additionally confirmed with immunohistochemistry staining for vascular markers, CD31, and smooth muscle actin (Jugold et al. 2008). Finally, the high-resolution imaging capabilities of micro-ultrasound allow cancer researchers to study not only the effect of a novel drug on angiogenesis and tumor growth, but also the influence of that drug on the surrounding tissues. This advantage is highly beneficial when monitoring side effects and determining the toxicity of therapeutics during drug development.

15.6 Targeting Injections and Biopsies Because of its real-time nature, a unique utility to micro-ultrasound is its ability to guide targeted injections or biopsies without having to do surgical intervention. As such, intricate procedures such as the placement of probes or invasive monitors in animals, the targeted injection of stem cells into specific tumor or anatomical targets, or even the specific placement of probes in vivo can be visualized noninvasively using micro-ultrasound. As a result of this noninvasive guidance facilitated by micro-ultrasound, surgery and stress on the animal are minimized or eliminated, accuracy of the procedures and injections is increased significantly, and results are more quantitative and reproducible. This was illustrated by Goessling et al. (2007) who demonstrated image-guided tumor biopsy in adult zebrafish with hepatic tumors. A 23G needle was guided into the tumor and a sample was taken and the sample was subsequently implanted in a separate recipient zebrafish (Goessling et al. 2007). The implanted tumor was still detected in vivo by microultrasound 6 weeks later.

15.7 Contrast Agents for Performing Real-Time Tumor Perfusion and Targeted Molecular Imaging of Biomarkers Microbubble contrast agents allow more sensitive detection of microvascular flow down to the capillary level. By adding ligands to endothelial cell surface

S. Foster and C. Theodoropoulos

markers, it is also possible to evaluate expression of important signaling, inflammation, and adhesion molecules. In angiogenic studies, for example, the targetready contrast agent (MicroMarker, VisualSonics) can be conjugated with an antibody against VEGFR2. The agent is then introduced into the animal through venous injection, and the animals are screened for relative expression of the tagged contrast agent. The animals can then be injected with an antiangiogenic drug and the effect of the therapeutic on the expression of the VEGFR2 receptor can be quantified. This whole procedure can be performed in vivo in the same animals and at multiple time points over a longitudinal time course. For micro-ultrasound, this procedure was first described by Rychak et al. (2007) and subsequently by Lyshchik et al. (2007), concluding that targeted contrast-enhanced high-frequency micro-ultrasound allows in vivo molecular imaging of VEGFR2 expression on the tumor vascular endothelium and may be used for noninvasive longitudinal evaluation of tumor angiogenesis in preclinical studies. More recently, Willmann et al. (2008a, b) showed high VEGFR2 expression in subcutaneous tumor (mouse angiosarcoma SVR cells) grown in nude mice using targeted VEGFR2 contrast agent. VEGFR2 expression was confirmed by immunohistochemistry. Furthermore, this contrast agent was mixed in in vitro cell culture to examine the binding affinity of the targeted contrast agent to angiosarcoma SVR cells. These results showed an average of 2 targeted VEGFR2 microbubbles binding to each angiosarcoma cell, suggesting very high binding affinity exists between targeted VEGFR2 microbubbles and angiosarcoma cell (Willmann et al. 2008a, b). High-frequency contrast agents are described more fully in the next chapter of this book.

15.8 Cardiovascular Applications Another major application of micro-ultrasound is in the field of cardiovascular research (see for example, references Zhou et al. 2004; Franco et al. 2006; Kulandavelu et al. 2006). The principle advantages of ultrasound in this regard are real-time frame rates to image the rapidly beating (6/s) heart rate, functional imaging of blood flow hemodynamics via pulsed Doppler and the ability to achieve phenomenal (<1 ms) temporal resolution using ECG-registered reconstruction (Cherin et al.

215

15 Ultrasound

2006). Figure 15.6 shows an example of a day 13 mouse embryo in which both ventricles are visible and the total size of the heart is little more than 1 mm in diameter. Other structures visible include neural tubes in the brain, the heart, and torso. Measurement of essential cardiac performance factors is possible using semiautomated methods in which the cardiac chambers are outlined in various phases of the cardiac cycle as illustrated in Fig. 15.7. Quantitative analysis of stroke volume, ejection fraction, fractional shortening, area change, fractional area change, cardiac output, stroke volume, LV mass, and average wall thickness is possible. Pulsed Doppler is also important for the assessment of the hemodynamics of the heart. Figure 15.8 shows the use of pulsed Doppler to measure flow velocity through the

mitral valve. Visible are both the E and A waves critical for the assessment of diastolic function. Other features that are directly measureable include the isovolumic relaxation times. ECG-registered reconstruction enables useful studies of myocardial biomechanics in which quantitative measurements of displacement and strain are needed. Speckle tracking studies (Li et al. 2007, 2008) have shown that high-resolution ultrasound image tracking methods provide for the detection of cardiac dyssynchrony in the noninfarcted regions in the murine left ventricles late after MI by identifying the temporal and spatial disparity of regional myocardial contraction. Elastography based on full radiofrequency reconstructions has demonstrated capability of accurately characterizing normal myocardial function throughout an entire cardiac cycle, at high resolution, and detecting and localizing myocardial infarction in vivo (Luo et al. 2007; Luo and Konofagou 2008).

15.9 Discussion and Conclusion

Fig. 15.6 Forty megahertz micro-ultrasound image of a day 13 mouse embryo showing the ventricles of the heart, neural, and body structures. The field of view is about 12 × 10 mm

a

Fig. 15.7 Long axis imaging of the left ventricle at end systole (a) and end diastole (b). Data from these views are combined with measurements of endo and epicardium in the short axis view to generate quantitative estimates of stroke volume, ejec-

The mechanical and array ultrasound technologies reported here extend the bandwidth of high performance ultrasound imaging from 15 to 50 MHz enabling significant improvement in performance for preclinical and potentially clinical applications. The established value of the mechanical high-frequency imaging systems is now augmented by new systems based on a 256 element linear array configuration with a 64-channel b

tion fraction, fractional shortening, fractional area change, cardiac output, stroke volume, LV mass, and average wall thickness. These measures are transferred directly to excel compatible data files for analysis

216

Fig. 15.8 Mitral valve flow waveforms such as that shown above can be obtained using the four-chamber view. The E (arrow) and A (double arrow) waveforms and isovolumic relaxation times (IVRT, IVCT) can be used for quantitative assessment of diastolic function

beamformer. Laser machining is used to define the array structure providing narrowly spaced elements with only 5–10 mm between elements. The architecture of the transducer and beamformer enables beam profiles that approach the diffraction limit and ensure high image quality. One of the major advances associated with the development of the high-frequency arraybased micro-ultrasound is the ability to perform realtime color and power Doppler as demonstrated in Fig. 15.5. In the preclinical arena, the new technology will find immediate application in a wide range of small animal biomedical research. In addition to providing accurate measures of tumor burden in the parenchymal organs, micro-ultrasound enables real-time visualization of flow in the tumor microcirculation. Other applications in cardiovascular disease, developmental biology, musculoskeletal disease, inflammation, atherosclerosis, and a wide variety of other mouse models of human disease are developing rapidly.

References Cherin E, Williams R, Needles A, Liu G, White C, Brown AS, Zhou Y-Q, Foster FS (2006) Ultrahigh frame rate ultrasound in micro-imaging and blood flow visualization in mice in vivo. Ultrasound Med Biol 32:683–691 Cheung AM, Brown AS, Hastie LA, Cucevic V, Roy M, Lacefield JC, Fenster A, Foster FS (2005) Three-dimensional ultrasound biomicroscopy for xenograft growth analysis. Ultrasound Med Biol 31:865–870 Cheung AM, Brown AS, Cucevic V, Roy M, Needles A, Yang V, Hicklin DJ, Kerbel RS, Foster FS (2007) Detecting vascular

S. Foster and C. Theodoropoulos changes in tumour xenografts using micro-ultrasound and micro-ct following treatment with VEGFR-2 blocking antibodies. Ultrasound Med Biol 33:1259–1268 Cobbold RSC (2007) Foundations of biomedical ultrasound. Oxford University Press, New York DeRosier LC, Buchsbaum DJ, Oliver PG, Huang ZQ, Sellers JC, Grizzle WE, Wang W, Zhou T, Zinn KR, Long JW, Vickers SM (2007) Combination treatment with TRA-8 anti death receptor 5 antibody and CPT-11 induces tumor regression in an orthotopic model of pancreatic cancer. Clin Cancer Res 13:5535s–5543s Feldmann G, Fendrich V, McGovern K, Bedja D, Bisht S, Alvarez H, Koorstra JB, Habbe N, Karikari C, Mullendore M, Gabrielson KL, Sharma R, Matsui W, Maitra A (2008) An orally bioavailable small-molecule inhibitor of Hedgehog signaling inhibits tumor initiation and metastasis in pancreatic cancer. Mol Cancer Ther 7:2725–2735 Foster FS (2000) Transducers and probe construction. Ultrasound Med Biol 26S:2–5 Foster FS (2008) Micro-ultrasound takes off (in the biological sciences). IEEE International Ultrasonics Symposium, pp. 120–125 Foster FS, Zhang MY, Zhou YQ, Liu G, Mehi J, Cherin E, Harasiewicz KA, Starkoski BG, Zan L, Knapik DA, Adamson SL (2002) A new ultrasound instrument for in vivo microimaging of mice. Ultrasound Med Biol 28: 1165–1172 Foster FS, Mehi J, Lukacs J, Hirson D, White C, Chaggares C, Needles A (2009) A new 15 – 50 MHz array based microultrasound scanner for preclinical imaging. Ultrasound Med Biol 35:1700–1708 Franco M, Man S, Chen L, Emmenegger U, Shaked Y, Cheung AM, Brown AS, Hicklin DJ, Foster FS, Kerbel RS (2006) Targeted anti-vascular endothelial growth factor receptor-2 therapy leads to short-term and long-term impairment of vascular function and increase in tumor hypoxia. Cancer Res 66:3639–3648 Goertz DE, Yu JL, Kerbel RS, Burns PN, Foster FS (2002) High-frequency Doppler ultrasound monitors the effects of antivascular therapy on tumor blood flow. Cancer Res 62:6371–6375 Goessling W, North TE, Zon LI (2007) Ultrasound biomicroscopy permits in vivo characterization of zebrafish liver tumors. Nat Methods 4:551–553 Graham KC, Wirtzfeld LA, MacKenzie LT, Postenka CO, Groom AC, MacDonald IC, Fenster A, Lacefield JC, Chambers AF (2005) Three-dimensional high-frequency ultrasound imaging for longitudinal evaluation of liver metastases in preclinical models. Cancer Res 65:5231–5237 Huizen IV, Wu G, Moussa M, Chin JL, Fenster A, Lacefield JC, Sakai H, Greenberg NM, Xuan JW (2005) Establishment of a serum tumor marker for preclinical trials of mouse prostate cancer models. Clin Cancer Res 11:7911–7919 Ino T, Akimoto K, Ohkubo M, Nishimoto K, Yabuta K, Takaya J, Yamaguchi H (1996) Application of percutaneous transluminal coronary angioplasty to coronary arterial stenosis in Kawasaki disease. Circulation 93:1709–1715 Jugold M, Palmowski M, Huppert J, Woenne EC, Mueller MM, Semmler W, Kiessling F (2008) Volumetric high-frequency Doppler ultrasound enables the assessment of early antiangiogenic therapy effects on tumor xenografts in nude mice. Eur Radiol 18:753–758

15 Ultrasound Kiguchi K, Ruffino L, Kawamoto T, Ajiki T, Digiovanni J (2005) Chemopreventive and therapeutic efficacy of orally active tyrosine kinase inhibitors in a transgenic mouse model of gallbladder carcinoma. Clin Cancer Res 11:5572–5580 Kiguchi K, Ruffino L, Kawamoto T, Franco E, Kurakata S, Fujiwara K, Hanai M, Rumi M, DiGiovanni J (2007) Therapeutic effect of CS-706, a specific cyclooxygenase-2 inhibitor, on gallbladder carcinoma in BK5.ErbB-2 mice. Mol Cancer Ther 6:1709–1717 Kulandavelu S, Qu D, Sunn N, Mu J, Rennie MY, Whiteley KJ, Walls JR, Bock NA, Sun JC, Covelli A, Sled JG, Adamson SL (2006) Embryonic and neonatal phenotyping of genetically engineered mice. Ilar J 47:103–117 Lee DJ, Lyshchik A, Huamani J, Hallahan DE, Fleischer AC (2008) Relationship between retention of a vascular endothelial growth factor receptor 2 (VEGFR2)-targeted ultrasonographic contrast agent and the level of VEGFR2 expression in an in vivo breast cancer model. J Ultrasound Med 27:855–866 Li Y, Garson CD, Xu Y, Beyers RJ, Epstein FH, French BA, Hossack JA (2007) Quantification and MRI validation of regional contractile dysfunction in mice post myocardial infarction using high resolution ultrasound. Ultrasound Med Biol 33:894–904 Li Y, Garson CD, Xu Y, French BA, Hossack JA (2008) High frequency ultrasound imaging detects cardiac dyssynchrony in noninfarcted regions of the murine left ventricle late after reperfused myocardial infarction. Ultrasound Med Biol 34:1063–1075 Liu J, Du J, Zhang C, Walker JW, Huang X (2007) Progressive troponin I loss impairs cardiac relaxation and causes heart failure in mice. Am J Physiol Heart Circ Physiol 293: H1273–H1281 Luo J, Konofagou EE (2008) High-frame rate, full-view myocardial elastography with automated contour tracking in murine left ventricles in vivo. IEEE Trans Ultrason Ferroelectr Freq Control 55:240–248 Luo J, Fujikura K, Homma S, Konofagou EE (2007) Myocardial elastography at both high temporal and spatial resolution for the detection of infarcts. Ultrasound Med Biol 33: 1206–1223 Lyshchik A, Fleischer AC, Huamani J, Hallahan DE, Brissova M, Gore JC (2007) Molecular imaging of vascular endothelial growth factor receptor 2 expression using targeted contrast-enhanced high-frequency ultrasonography. J Ultrasound Med 26:1575–1586 Olive KP, Tuveson DA (2006) The use of targeted mouse models for preclinical testing of novel cancer therapeutics. Clin Cancer Res 12:5277–5287 Rychak JJ, Graba J, White C, Cheung AMY, Mistry B, Lindner JR, Kerbel RS, Foster FS (2007) Micro-ultrasound molecular imaging of VEGFR-2 in a mouse model of tumor angiogenesis. Mol Imaging 6:289–296 Shaked Y, Ciarrocchi A, Franco M, Lee CR, Man S, Cheung AM, Hicklin DJ, Chaplin D, Foster FS, Benezra R, Kerbel RS (2006) Therapy-induced acute recruitment of circulating endothelial progenitor cells to tumors. Science 313:1785–1787

217 Sherar MD, Foster FST (1989) The design and fabrication of high frequency Poly(vinylidene fluoride) transducers. Ultrasonic Imaging 11:75–94 Trivedi CM, Luo Y, Yin Z, Zhang M, Zhu W, Wang T, Floss T, Goettlicher M, Noppinger PR, Wurst W, Ferrari VA, Abrams CS, Gruber PJ, Epstein JA (2007) Hdac2 regulates the cardiac hypertrophic response by modulating Gsk3 beta activity. Nat Med 13:324–331 Wang XF, Birringer M, Dong LF, Veprek P, Low P, Swettenham E, Stantic M, Yuan LH, Zobalova R, Wu K, Ledvina M, Ralph SJ, Neuzil J (2007) A peptide conjugate of vitamin E succinate targets breast cancer cells with high ErbB2 expression. Cancer Res 67:3337–3344 Willmann JK, Chen K, Wang H, Paulmurugan R, Rollins M, Cai W, Wang DS, Chen IY, Gheysens O, Rodriguez-Porcel M, Chen X, Gambhir SS (2008a) Monitoring of the biological response to murine hindlimb ischemia with 64Cu-labeled vascular endothelial growth factor-121 positron emission tomography. Circulation 117:915–922 Willmann JK, Lutz AM, Paulmurugan R, Patel MR, Chu P, Rosenberg J, Gambhir SS (2008b) Dual-targeted contrast agent for US assessment of tumor angiogenesis in vivo. Radiology 248:936–944 Wirtzfeld LA, Wu G, Bygrave M, Yamasaki Y, Sakai H, Moussa M, Izawa JI, Downey DB, Greenberg NM, Fenster A, Xuan JW, Lacefield JC (2005) A new three-dimensional ultrasound microimaging technology for preclinical studies using a transgenic prostate cancer mouse model. Cancer Res 65:6337–6345 Wirtzfeld LA, Graham KC, Groom AC, Macdonald IC, Chambers AF, Fenster A, Lacefield JC (2006) Volume measurement variability in three-dimensional high-frequency ultrasound images of murine liver metastases. Phys Med Biol 51:2367–2381 Wu G, Wang L, Yu L, Wang H, Xuan JW (2005) The use of three-dimensional ultrasound micro-imaging to monitor prostate tumor development in a transgenic prostate cancer mouse model. Tohoku J Exp Med 207:181–189 Xuan JW, Bygrave M, Jiang H, Valiyeva F, Dunmore-Buyze J, Holdsworth DW, Izawa JI, Bauman G, Moussa M, Winter SF, Greenberg NM, Chin JL, Drangova M, Fenster A, Lacefield JC (2007) Functional neoangiogenesis imaging of genetically engineered mouse prostate cancer using threedimensional power Doppler ultrasound. Cancer Res 67: 2830–2839 Zhou YQ, Foster FS, Parkes R, Adamson SL (2003) Developmental changes in left and right ventricular diastolic filling patterns in mice. Am J Physiol Heart Circ Physiol 285:H1563–H1575 Zhou YQ, Foster FS, Nieman BJ, Davidson L, Chen XJ, Henkelman RM (2004) Comprehensive transthoracic cardiac imaging in mice using ultrasound biomicroscopy with anatomical confirmation by magnetic resonance imaging. Physiol Genomics 18:232–244

Ultrasound Contrast Agents

16

Michel Schneider

Contents 16.1 Main Characteristics of Ultrasound Contrast Agents (UCA)........................................ 219 16.2 Preparation of Ultrasound Contrast Agents...... 220 16.3 Contrast-Specific Imaging Modes....................... 221 16.4 Contrast-Enhanced Destruction Replenishment Method........................................ 222 16.5 Applications of Ultrasound Contrast Agents..... 223 16.5.1 Vascular Imaging.................................................... 223 16.5.2 Tissue Perfusion...................................................... 223 16.5.3 Lesion Characterization and Follow-Up................. 223 16.5.4 Site-Specific Ultrasound Contrast Agents.............. 224 16.6 The Importance of Quantification....................... 225 16.7 Discussion and Conclusion................................... 225 16.8 Appendix: Laboratory Protocols......................... 226 16.8.1 Example 1: Mouse Kidney Perfusion..................... 226 16.8.2 Example 2: Use of P-Selectin: Binding Microbubbles in a Mouse Inflammation Model............................................... 226 16.8.3 Example 3: Mean Transit Time Measurements in a Mat B III Tumor Model........... 228 References............................................................................ 229

M. Schneider Bracco Research SA, Geneva, Switzerland e-mail: [email protected]

16.1 Main Characteristics of Ultrasound Contrast Agents (UCA) Contrast agents are extensively used in medical imaging to enhance particular body compartments or tissues. They are usually administered intravenously. The relative enhancement observed provides infor mation on gross anatomy, lesion extent, perfusion abnormalities, and other morphological aspects, i.e., key elements that provide essential aids for diagnosis. Computed tomography (CT), magnetic resonance imaging (MRI), single photon emission computed tomography (SPECT), and positron emission tomography (PET) make extensive use of contrast agents. As with these other imaging modalities, there is an added value in using contrast agents in ultrasound imaging. Gramiak and Shah were the first authors to report in 1968 (Gramiak and Shah 1968) that air bubbles could be used in vivo as strong reflectors of ultrasound. Their observations led to the first ultrasound contrast agent preparation obtained by rapidly transferring saline between two syringes in the presence of air. This socalled “agitated saline” method is still in use today, primarily in cardiology departments for the detection of cardiac shunts. However, these air-based microbubbles are too short lived (a few seconds) to be of any value outside of this very particular indication, which does not require passing the lung capillaries. It is only in the 1980s that the first bubble stabilization methods were described, in particular the work of Rasor et al. (1992), using saccharides, and Keller et al., who used albumin stabilized bubbles (1987). In the 1990s it was realized that further stabilization, and consequently, improved persistence in the bloodstream could be obtained by using poorly soluble gases rather than air.

F. Kiessling and B.J. Pichler (eds.), Small Animal Imaging, DOI: 10.1007/978-3-642-12945-2_16, © Springer-Verlag Berlin Heidelberg 2011

219

220

M. Schneider

Perfluorinated gases such as sulfur hexafluoride (SF6), perfluoropropane (C3F8), and perfluorobutane (C4F10) are today the preferred gases for ultrasound contrast agents. Bubble size is another important parameter. After intravenous administration, large bubbles (>8 mm) are trapped in the lungs and therefore do not reach the general circulation. The bubble sizes should preferably be in the 2–5 mm range, i.e., close to the size of red blood cells (6–8 mm). Bubbles in this size range are able to move unimpeded in the entire vasculature, including in the smallest capillaries. Micron-sized bubbles are outstanding ultrasound reflectors due to the fortunate coincidence that their resonance frequency lies right within the range of medical frequencies (2–20 MHz). Albumin, biocompatible polymers, or lipids are generally used to stabilize the bubbles, to prevent coalescence, and to reduce gas dissolution in blood. Microbubbles stabilized by a monomolecular layer of phospholipids are particularly effective from an acoustic point of view due to their thin (2–3 nm) and highly elastic shell (see Fig. 16.1). They are used in several commercial ultrasound contrast agents. Today, all commercially available ultrasound contrast agents contain high concentrations of micron-sized gas bubbles (108–1010 bubbles per ml) stabilized by phospholipids or heat-denatured albumin. Polymerstabilized bubbles have been described by several authors, but none of them has so far reached the marketplace. The half life of ultrasound contrast agents after intravenous administration is short, typically a few minutes. This duration is, however, in most cases sufficient for imaging purposes and has the added advantage that several consecutive injections can be performed in the same animal. Although not specifically discussed in the present chapter, ultrasound contrast agents are extremely safe and repeated phospholipid membrane

SF6 or C4F10

Fig. 16.1 Structure of a phospholipid-stabilized microbubble

administrations to test animals are not an issue. The gas present in the bubbles is exhaled via the lungs within minutes. The shell constituents are taken up in the liver where they are metabolized. Only minute amounts of gas and stabilizing material are administered in an imaging dose, typically 0.3 mL of gas and about 8 mg of phospholipids in a mouse imaging dose. These minute quantities reflect the remarkable potency of bubble-based ultrasound contrast agents.

16.2 Preparation of Ultrasound Contrast Agents As indicated above, ultrasound contrast agents are made of small size gas bubbles surrounded by a stabilizing shell in suspension in an isotonic aqueous carrier. Several methods have been described for the preparation of such stabilized microbubbles. Albumincoated bubbles have been described by Keller et al. (1987) and by Porter et al. (1993). The preparation procedure involves the sonication of human serum albumin using a probe sonicator in the presence of a gas (air or perfluorinated gas). The preparation of phospholipid-stabilized bubbles has been described by Unger et al. (1992) and by Klibanov and coworkers (Kim et al. 2000). Both authors make use of sonication, at high output power, of an aqueous suspension of phospholipids in the presence of a perfluorocarbon gas. The stable bubbles obtained can be separated from the nonincorporated phospholipids by flotation or lowspeed centrifugation. While these preparation methods may sound straightforward, standardization of the procedure is not easy. The total bubble concentration and the bubble size distribution – and consequently also the acoustic properties – vary from batch to batch. The stability of these preparations is also questionable even when stored at 5°C. Moreover, the preparation techniques described in these articles are laboratory procedures using probe sonicators and are not easily applicable to the preparation of clean, sterile, and pyrogen-free products. Ultrasound contrast agents were initially developed for blood pool imaging. Recently, some authors have attempted and succeeded to link molecular entities such as antibodies to the phospholipid-stabilizing monolayer. The general concept was to develop bubbles that would not simply circulate passively such as

221

16 Ultrasound Contrast Agents

red blood cells, but could adhere to selected sites in the vascular compartment and generate detectable signals after clearance of circulating bubbles. A general scheme for the preparation of such targeted bubbles consists of incorporating small amounts of a biotinylated phospholipid in the phospholipid monolayer. The resulting bubbles are then incubated with avidin (or streptavidin) which will bind to the biotin residues. Thanks to the multivalency of avidin/streptavidin molecules, biotinylated molecular entities can then be bound to the avidin/streptavidin-coated bubbles (see Fig. 16.2) (Klibanov 1999). In this way, a targeting moiety may be attached to the microbubble shell through a biotin-avidin-biotin bridge. This procedure is very flexible and convenient for small-animal imaging, as it allows the preparation of microbubbles coated on their surface with a biotinylated molecule selected by the investigator. It requires, however, several centrifugation steps (to eliminate the excess avidin and the unreacted biotinylated moieties) and this has a detrimental effect on the overall yield of targeted bubbles recovered at the end of the procedure. Recently, commercial preparations designed for small-animal imaging and including both nontargeted and streptavidin bubbles have been made available. The latter are particularly versatile since they can be used to prepare targeted bubbles according to the user’s needs using an appropriate ligand. The ligand can be a protein, an antibody or antibody fragment, an oligopeptide, or even a small molecule, the only requirement being that a biotin residue be attached to the molecule selected. No separation stage is usually necessary since the amount of biotinylated ligand added can be adjusted in order to be captured entirely by the streptavidin bubbles. Examples of the use of these

commercial agents are described in the Appendix on Laboratory protocols.

16.3 Contrast-Specific Imaging Modes An ultrasound beam consists of traveling pressurepulse waves with alternating compression and rarefaction phases. Because of their high compressibility, microbubbles exposed to ultrasound alternatively contract and expand during the compression and rarefaction phases, respectively, and scatter ultrasound waves in all directions, including back to the transducer. At the so-called resonance frequency, bubble-oscillations are particularly high and, consequently, also the scattered ultrasound waves and the signal detected by the probe. The resonance frequency decreases with bubble size. For bubbles with diameters in the 0.5–6 mm range, the resonance frequencies are in the 2–20 MHz range. The mechanical index (MI) is a measure of the maximum acoustic pressure delivered to the tissue. At very low acoustic pressures, bubbles expand and contract in a symmetrical way – the so-called linear oscillation mode. The frequency of ultrasound reflected waves is equal to the transmitted frequency. At slightly higher acoustic pressures, nonlinear oscillations occur. In this mode, echo signals comprise not only frequency components present in the transmitted waveforms, but also harmonics thereof. These nonlinear signal components are specific of bubbles and are not present to a significant extent in tissue echoes. Several imaging modes have been designed to capture preferentially these nonlinear components. They are usually designated with terms such as second harmonic imaging, pulse

avidin biotin

Fig. 16.2 Coupling of a biotinylated antibody to streptavidin molecules bound onto biotinylated microbubbles

222

M. Schneider

inversion, or power modulation. In the latter cases, these methods exploit the nonlinear responses of bubbles exposed to successive pulses of different amplitudes or phases. Bubble-specific signals can thus be generated using specific imaging methods called “coherent contrast imaging,” “contrast pulse sequencing,” and such. For more detail on the various contrastspecific imaging modes available, the reader is referred to a detailed description by Rafter et al. (2004). The advent of contrast-specific imaging has dramatically improved the sensitivity for the detection of microbubbles, allowing real-time imaging at low MI. At baseline, in the absence of bubbles, the signals from the tissues are quasi-completely extinguished. Only when the bubbles arrive in the tissue, the vessels start to light up until complete enhancement (see Fig. 16.3). Thanks to this very high sensitivity, only very tiny amounts of contrast microbubbles are necessary to generate strong enhancement. This is in contrast with other imaging

modalities, such as CT or MRI, which require much larger amounts of contrast agent. In addition to their nonlinear acoustic behavior, microbubbles have another important characteristic when exposed to high acoustic pressures. Such exposure results in more intense oscillation of the bubbles and leads to their bursting. This outstanding feature of microbubbles has found applications for blood flow measurements and will be discussed in the next paragraph.

16.4 Contrast-Enhanced Destruction Replenishment Method Exposure to a few frames at high MI results in the destruction of microbubbles in the scan plane. This property can be used to derive data both on blood flow and blood volume as shown initially by Wei et al.

Mouse kidney in contrast mode (12.5 MHz) before MicroMarker injection

Mouse kidney in contrast mode (12.5 MHz) 3 s after MicroMarker injection

Fig. 16.3 Contrast perfusion in a mouse kidney. Upper: baseline image; middle: arrival of the contrast agent in the cortex (3 s after MicroMarker injection); lower: complete filling of the kidney (cortex and medulla), 7 s after MicroMarker injection

Mouse kidney in contrast mode (12.5 MHz) 7 s after MicroMarker injection

223

16 Ultrasound Contrast Agents

(1998) for myocardial perfusion. This approach has then been extended to the evaluation of the same parameters in various tissues, lesions, tumors, organs such as kidneys, etc. Continuous infusion of contrast agent is usually preferred, in order to reach first a steady-state concentration of microbubbles in the tissue of interest. Contrast microbubbles are then destroyed within the imaging plane by applying two or more frames at high mechanical index (MI > 0.8) and, preferably, at a low frequency. Following the destruction, the UCA replenishment within the scan plane is monitored by contrast-specific ultrasound imaging under nondestructive conditions, i.e., at low MI. The rate of replenishment is proportional to blood velocity and the product of the blood volume by the blood velocity is a measure of blood flow. Any change in blood flow (for instance, through the action of a drug), therefore, translates into a modification of the replenishment rate. After complete refilling with bubbles, the final plateau reached at steady-state is a measure of the capillary blood volume and any change of signal reflects changes in the vascular blood volume.

16.5 Applications of Ultrasound Contrast Agents As already pointed out, tiny doses of microbubbles are sufficient to generate strong enhancement. Typical volumes for mouse or rat imaging are 20–50 mL, followed by a saline flush (50–150 mL). These volumes are well tolerated by the animals with no evidence of hemodynamic effects.

presence of contrast may help in determining precisely the size of the vessel, to evaluate the vessel wall thickness, and to detect atherosclerotic plaque, stenosis, occlusion, and blood leakages. Real-time contrastenhanced imaging of vessels is usually superior to color or power Doppler for the detection of vessel abnormalities, due to a higher spatial and temporal resolution and the absence of artifacts (e.g., blooming).

16.5.2 Tissue Perfusion The level of tissue perfusion is the most important indicator of blood-, oxygen-, and nutrient supply to parenchymal cells. Standard echography shows the overall tissue architecture, but provides no information about perfusion defects, vascular occlusions, and other circulatory anomalies. Doppler US is classically used to assess blood flow. However, Doppler ultrasound can only be used to estimate flow rate in vessels with a diameter greater than 200 mm, i.e., in relatively large vessels. Thanks to the high signal-to-noise ratio of contrast-specific imaging modes, it is now possible to detect bubbles in smaller vessels, down to the capillary level (Ferrara and Merritt 2000). The combination of UCA and imaging modes allows to assess the microcirculation and to detect microvascular flow in real time. By observing the filling pattern of a tissue or an organ, abnormal regions can be identified. Stenoses in major feeding vessels result in areas with delayed filling and occlusions translate into a complete absence of perfusion in an entire territory. Contrast ultrasound also clearly outlines areas of necrosis.

16.5.1 Vascular Imaging

16.5.3 Lesion Characterization and Follow-Up

After intravenous injection, UCAs enhance all bloodcontaining compartments. These include heart ventricles and large vessels such as the aorta. Left ventricle opacification and endocardial border delineation were actually the first clinical indications for UCAs. Contrast enhancement of cavities such as the left ventricle allows to identify precisely the endocardial border, and therefore, facilitates the detection of areas with abnormal ventricular contraction. Similarly, in large vessels, the

Different lesions show different filling profiles and this can be used for their characterization. Exploiting dynamic enhancement patterns for the characterization of focal hepatic lesions is a major clinical indication for SonoVue™, a commercial UCA developed by Bracco. Tumors often show rapid contrast enhancement first at the periphery, sometimes extending in a second stage to the central part of the lesion, often followed by a rapid washout of the contrast signal.

224

Several imaging modalities such as CT, MRI, and ultrasound are used to assess tumor size and structure. Contrast ultrasound provides additional information on tumor perfusion, vessel density, blood volume, blood flow, and blood distribution in the tumor. Cytostatic therapies do not usually lead to shrinkage of tumors in a short period of time. But changes in tumoral perfusion occur very rapidly after the onset of chemotherapy, and contrast ultrasound is an ideal tool to detect early response. Drug-induced perfusion changes can be monitored by contrast-enhanced ultrasound. In patients with gastrointestinal stromal tumors (GIST) for instance, the response to antiangiogenic therapy (imatinib) was shown to translate into detectable changes in tumor perfusion as early as one day after treatment initiation, therefore providing an early prediction of tumor response (Lassau et al. 2006). Classical morphologic criteria based on tumor size measurements are unlikely to be able to assess tumor response at such an early stage of treatment. Similar studies have been performed in small animals. Changes in blood flow and vascular volume could be measured quantitatively in tumor-bearing rats, providing useful data for treatment monitoring (Pollard et al. 2007). This rapid effect of test drugs might also become important for identifying the optimal time window and dosing of drugs. Additional important uses of contrast ultrasound in the follow-up of treatment include the detection of residual blood flow in treated lesions (indicating incomplete tumor destruction) and the monitoring of tumor recurrence.

16.5.4 Site-Specific Ultrasound Contrast Agents Due to their sizes, microbubbles are unable to extravasate and remain strictly localized in the vascular compartment. Therefore, only the vascular endothelial surface is accessible to site-specific (targeted) microbubbles. In a number of pathophysiological conditions, cell surface markers specifically associated with disease are expressed or upregulated on the vascular endothelium. Inflammation, for instance, is characterized by endothelial overexpression of selectins (such as P and E selectins), immunoglobulin-like intercellular adhesion molecule-1 (ICAM-1), and vascular cell adhesion molecule-1 (VCAM-1). This so-called

M. Schneider

dysfunctional endothelium can be interrogated by bubbles-bearing specific affinity ligands on their surface, which causes sticking of the bubbles to the receptor specific for the selected ligand. The capacity of targeted bubbles to image vascular receptors allows the detection and characterization of markers at the molecular level, and therefore, performs what is classically designated as molecular imaging. In this regard, two application areas are particularly important: inflammation and neoangiogenesis. Infla mmation is a common reaction present in many diseases. During inflammation, several receptors are expressed or upregulated on the luminal endothelium and are therefore accessible to targeted microbubbles. For instance, postischemic injury is known to result in the expression of P-selectin. Microbubbles targeted to P-selectin have been prepared using an anti-P-selectin monoclonal antibody and showed a much stronger contrast signal in a mouse kidney model compared to the one observed with similar bubbles bearing a control antibody (Lindner et al. 2001). The second important area is neoangiogenesis, i.e., the formation of new blood vessels which is a fundamental process occurring during tumor progression. The major angiogenesis-related targets studied include the vascular endothelial growth factor receptor-2 (VEGF-R2), the avb3 integrin, and endoglin (CD105). A biotinylated antimouse VEGF-R2 antibody coupled to streptavidin microbubbles showed significantly higher signal intensity in two murine models of breast cancer compared to control microbubbles. In addition, the level of signal observed correlated well with the expression of VEGF-R2 expression in the two tumor types (Lyshchik et al. 2007). Similar results were reported in two murine models of angiosarcoma and malignant glioma (Willmann et al. 2008). The same strategy has been used for the preparation of a whole range of target-specific bubbles for other neoangiogenesis receptors such as av integrins (Ellegala et al. 2003; Leong-Poi et al. 2005) and for various inflammatory receptors such as ICAM-1 (Weller et al. 2005), P-selectin (Takalkar et al. 2004; Villanueva et al. 2007; Kaufmann et al. 2007a), and VCAM-1 (Kaufmann et al. 2007b). It is beyond the scope of the present chapter to cite exhaustively the numerous receptors which have been imaged so far. For a more extensive list, the reader is referred to a review article by PA Dayton and JJ Rychak (Dayton and Rychak 2007). In addition to endothelial receptors, site-targeted microbubbles can be used for the visualization of thrombi associated with stroke. Immunobubbles

225

16 Ultrasound Contrast Agents

bearing Abciximab, a glycoprotein IIb/IIIa receptor inhibitor, were shown to improve visualization of human clots both in vitro and in vivo in a model of acute arterial thrombotic occlusion (Alonso et al. 2007). Several receptors can also be imaged in the same animals during tumor growth, for instance, VEGF-R2 and avb3 integrin (Palmowski et al. 2008) or endoglin (CD105), VEGF-R2, and the VEGF-VEGFR complex (Korpanty et al. 2007), and the level of expression of each of these markers can be monitored separately during treatment. As shown in these studies, targeted microbubbles can be used not only to detect the presence of different markers expressed on endothelial cells, but also to monitor changes in receptor levels longitudinally in response to therapy. Treatment with antiangiogenic or cytotoxic therapy resulted in a reduction in binding of microbubbles targeted to these receptors. This reduction correlated with a reduction in tumor microvessel density. Changes in receptor levels might be very early signs of treatment efficacy, well before any morphological change is detected, and might, therefore, be used as surrogate markers and good predictors of response to antiangiogenic or cytotoxic chemotherapy. Whether these changes take place before or simultaneously with drug-induced perfusion changes remains to be established on a case-by-case basis. The imaging strategy for detecting attached bubbles is different from the regular approach used in perfusion imaging. As a matter of fact, only a very small fraction of the injected site-specific bubbles actually reach and stick to endothelial receptors. The large majority transit freely throughout the microcirculation and are then cleared from the bloodstream. Sitetargeted bubbles generally become detectable only after complete elimination of circulating bubbles. In many cases, these bubbles are detectable by low MI contrast-specific imaging. In some circumstances, the number of bubbles attaching at the receptor site is too low to include bubbles of resonant size and results in insufficient acoustic signal. In this case, a destructive ultrasound imaging approach can be used. This involves the application of high mechanical-index pulses, causing bubble destruction. The destruction generates strong acoustic echoes, easily detectable by ultrasound systems. The advantage of this approach is its very high sensitivity. A major drawback, however, is that the destruction of targeted bubbles is a one-shot event, which cannot be repeated unless a new dose of site-specific bubbles is administered.

16.6 The Importance of Quantification Whether the objective is to follow therapy, to detect the level of expression of a given receptor, or to screen different molecules in an animal model, the life scientist is always faced with the need to extract quantitative information from ultrasound imaging sequences. At baseline, when contrast-specific imaging modes are used, the residual signal in an area of interest (AOI) is usually minimal. Following administration of nontargeted bubbles, the measurement of the mean signal intensity in the AOI as a function of time (time-intensity curve) provides information on the local concentration of bubbles and hence – since UCAs are pure blood-pool agents – on blood perfusion parameters in the scanning plane. Contrast ultrasound allows following the wash in and wash out phases in AOIs in real time. Following curve fitting, several parameters can be derived from the time-intensity curve such as time to peak, peak intensity, mean transit time (MTT), area under the curve, etc. In the case of targeted bubbles, the measurement of the mean intensity in an AOI after clearance of circulating bubbles will provide information on the presence of a given receptor and on its level of expression. These parameters can be measured in several test animals and averaged. Alternatively, they can be followed longitudinally in each individual animal if the objective is, for instance, to follow therapy as a function of time. In the bubble destruction/replenishment approach, the refilling kinetics of contrast in the AOI provide quantitative data related to blood flow and blood volume.

16.7 Discussion and Conclusion Imaging methods offer several advantages over other practices currently used in small-animal experimentation. As they are noninvasive, they allow for longitudinal studies in a single animal, which results into a considerable reduction in the number of animals required. In addition, the use of imaging endpoints can significantly decrease the workload involved in tissue analyses and radioactive tracer measurements. Ultrasound is a real-time imaging modality ideally suited for small-animal imaging, thanks to its very broad range of applications, its high spatial and temporal

226

resolution, and low cost. Access to the animal during ultrasound imaging is easy and the animal can be continuously monitored during a procedure. In the absence of contrast, ultrasound provides information on the density and compressibility of tissues, and is, therefore, primarily an anatomical imaging modality with excellent spatial resolution (30 mm with high frequency probes). Additional information on blood flow in major vessels can be obtained by Doppler processing. With the adjunct of a contrast agent, ultrasound imaging is able to provide a new level of functional information on blood flow, not only in large vessels but also in small ones. Quantitative software packages capable of measuring enhancement levels and extracting kinetic parameters are becoming available. They provide valuable quantitative data, necessary for evaluating and comparing different drug treatments and for assessing interanimal variability. Blood flow and vascular volume changes can be quantitatively determined using the microbubble destruction/replenishment technique. The strict intravascular localization of microbubbles provides clean perfusion information, which is not corrupted by the diffusion of a contrast agent in extravascular spaces, as observed with most CT or MR contrast agents. Contrast ultrasound allows the detection and characterization of lesions, the follow-up of their evolution, and the effects of treatment over time in the same animal. The development of site-specific contrast agents has expanded the role of ultrasound beyond blood flow or blood volume measurements into molecular imaging applications. Microbubbles targeted to specific epitopes present in the vascular lumen or on intravascular cells can be detected due to the very high sensitivity of contrast-specific imaging modes capable of detecting even individual bubbles. Molecular ultrasound imaging allows to visualize the expression (presence) of a receptor and to assess receptor density over time and/or during treatment with drugs. It can be used to detect early response to therapy or to provide surrogate endpoints for measuring drug efficacy or therapy success. Com pared to methods based on the use of radiolabeled agents, molecular ultrasound imaging shows much higher spatial resolution, and therefore, anatomical information without ionizing radiation. Molecular ultrasound imaging will be particularly useful for scientists involved in inflammation and/or angiogenesis research in animal models. Interestingly, once the benefits of molecular imaging studies are established in animal models, the same methodology can be transferred into

M. Schneider

clinical settings, for instance for following response to therapy and/or for distinguishing patients responding to a given therapy from nonresponders. As ultrasound imaging has not reached full maturity yet, a number of new and exciting developments can be expected in the coming years. In particular, three-dimensional imaging in real-time is the object of intense development. It is likely to increase even further the diagnostic value of contrast ultrasound in small-animal research, by providing 3D information on lesion size, tissue, or tumor perfusion, areas of inflammation or neoangiogenesis, etc.

16.8 Appendix: Laboratory Protocols 16.8.1 Example 1: Mouse Kidney Perfusion The example of Fig. 16.3 makes use of a commercial contrast agent (Vevo MicroMarker Contrast Agent, Visualsonics Inc., Toronto, Canada). The agent is supplied in the form of a lyophilized cake. The microbubble suspension is generated by adding 0.7 mL of saline (0.9% NaCl) through the rubber stopper, followed by gentle agitation of the vial for 10 s. The reconstituted product contains 2–3 109 bubbles per ml with a mean diameter in number (Dn) of 1.2–1.3 mm. The product is stable within a vial for 3–4 h after reconstitution. A typical dose for mouse imaging is 50 mL, followed by a 50-mL flush of saline. The agent is usually administered intravenously in a tail vein or in the jugular vein. Figure 16.3 illustrates contrast images of a mouse kidney using a contrastspecific mode on the Vevo2100 (Visualsonics, Toronto, Canada) at successive instants: before injection of the contrast agent, upon arrival of the agent in the cortex, and after filling of the entire organ.

16.8.2 Example 2: Use of P-Selectin: Binding Microbubbles in a Mouse Inflammation Model This example makes use of commercial streptavidincontaining microbubbles (MicroMarker Target-Ready

227

16 Ultrasound Contrast Agents

Contrast Agent, Visualsonics Inc., Toronto, Canada). In this contrast agent, streptavidin is already incorporated in the bubble-stabilizing lipid shell. Contrary to the biotin-labeled bubbles, these bubbles are ready to be used for conjugation with any biotinylated ligand (antibody, peptide, etc.). No centrifugation or other separation step is necessary. This agent is supplied in the form of a lyophilized cake. Saline (NaCl 0.9%, 0.5 mL) is injected through the rubber stopper for reconstitution of the bubble suspension. The syringe is then removed, leaving in place the needle for a few seconds for venting the excess pressure in the vial. The needle is then removed and the vial is gently agitated for 10 s and let to sit at room temperature for 5 min. At this point, the vial contains 4–6 109 streptavidin-containing microbubbles which are ready to be conjugated with a biotinylated molecule. For the coupling of a biotinylated antibody, the antibody solution (10–20 mg in 0.4 mL saline) is injected through the rubber stopper of the reconstituted vial. The vial is then gently agitated for 1 min and let to rest at room temperature for 10 min. The desired volume (usually around 70 mL) of antibodycoupled microbubbles is drawn up in a 1-mL syringe, preferably fitted with a 21G5/8 needle and injected

intravenously (tail vein or jugular vein) to the animal. In the case of several withdrawals, the vial should be gently agitated before sample collection. In the following example, hind limb inflammation was induced by IM injection of mouse rTNF-a in OF1 mice of 6–10 weeks of age (30 mL/mouse of a 0.017-mg/mL solution). Ultrasound imaging of the hind limb was performed 24 h later using an ultrasound scanner (Siemens, Sequoia 512) in CPS mode at 7 MHz with a MI of 0.25. Anti-P-selectin microbubbles were obtained by injecting a biotinylated anti-Pselectin antibody (20 mg, BD Biosciences) into a vial containing streptavidin microbubbles as described above. Similarly, control microbubbles were obtained by coupling a biotinylated isotype control antibody (from eBioscience) to streptavidin microbubbles. Boluses (25 mL) of targeted or control bubbles were administered in the tail vein followed by flush of saline (80 mL). Figure 16.4 shows the time-intensity curves measured in the inflamed muscle. In the case of P-selectin-targeted microbubbles, the signal increased very rapidly and stayed almost at the same level for at least 10 min, whereas control bubbles washed out very rapidly. The images on the right hand side show the enhancement in the inflamed muscle at 10 min.

120 100

Echo signal

80 60 40 20 0

0

100

200

300 400 Time (sec)

500

600

Fig. 16.4 Time-intensity curves in a TNF-a-induced mouse inflammatory model in the muscle after administration of bubbles bearing an anti-P-selectin antibody (green curve) or an isotype control antibody (red curve). Images on the right hand side

700

show the enhancement in the inflamed muscle, 10 min after bolus injection of P-selectin-binding bubbles (upper picture) and control bubbles (lower picture)

228

16.8.3 Example 3: Mean Transit Time Measurements in a Mat B III Tumor Model The example of Fig. 16.5 refers to the use of the commercial ultrasound contrast agent SonoVue™ (Bracco, Italy) for perfusion measurements in a mammary tumor model in the rat. Rat mammary adenocarcinoma was induced by orthotopic injection of 13762 Mat B III tumor cells (0.5 × 106 cells in 0.1 mL) in the mammary fat pad of anesthetized female Fischer 344 rats (150 g). Tumor imaging is performed at day 8 after implantation using a Siemens Sequoia ultrasound equipment (Siemens, Mountain View, California) with a 15L8 probe (transmit frequency 7 MHz). The following settings are used: CPS mode at an MI of 0.18; dynamic range 83 dB. Gain was set at 0 dB with flat (neutral) TGC settings.

M. Schneider

SonoVue™ (50 mL) is injected in the jugular vein. The bolus injection is followed by a saline flush (0.2 mL). The time-intensity curve was recorded as DICOM files. The sequence was then analyzed using proprietary software (SonoFit, Bracco). The SonoFit software package provides off-line analysis of video sequences or DICOM files. The main features include linearization of video data, motion compensation, and curve fitting in up to three areas of interest. A first AOI can, for instance, be drawn around the entire tumor and other ones in selected areas such as in the peripheral part of the tumor. The software provides time-intensity curves in the areas of interest. A number of parameters can be derived from the fitted curves, in particular peak intensity (expressed as echo power), MTT, time to peak, area under the curve, etc. Figure 16.5 shows fitted timecurves in two areas and their corresponding MTT measurements.

Fig. 16.5 Time-intensity curves in two areas of interest selected in a Mat B III tumor in the rat after administration of SonoVue. The mean transit times in the two areas are 20.9 s in the highly perfused area and 29.2 s in the peripheral zone

16 Ultrasound Contrast Agents

References Alonso A, Della Martina A, Stroick M et al (2007) Molecular imaging of human thrombus with novel abciximab immunobubbles and ultrasound. Stroke 38:1508–1514 Dayton PA, Rychak JJ (2007) Molecular ultrasound imaging using microbubble contrast agents. Front Biosci 12:5124–5142 Ellegala DB, Leong-Poi H, Carpenter JE (2003) Imaging tumor angiogenesis with contrast ultrasound and microbubbles targeted to avb3. Circulation 108:336–341 Ferrara KW, Merritt CRB (2000) Evaluation of tumor angiogenesis with US: imaging, doppler, and contrast agents. Acad Radiol 7:824–839 Gramiak R, Shah PM (1968) Echocardiography of the aortic root. Invest Radiol 3(5):356–366 Kaufmann BA, Lewis C, Xie A (2007a) Detection of recent myocardial ischaemia by molecular imaging of P-selectin with targeted contrast echocardiography. Eur Heart J 28: 2011–2017 Kaufmann BA, Sanders JM, Davis C (2007b) Molecular imaging in atherosclerosis with targeted ultrasound detection of vascular cell adhesion molecule-1. Circulation 116(3):276–284 Keller MW, Feinstein SB, Watson DD (1987) Successful left ventricular opacification following peripheral venous injection of sonicated contrast agent: an experimental evaluation. Am Heart J 114(3):570–575 Kim DH, Klibanov AL, Needham D (2000) The influence of tiered layers of surface-grafted poly(ethylene glycol) on receptor-ligand-mediated adhesion between phospholipid monolayer-stabilized microbubbles and coated glass beads. Langmuir 16:2808–2817 Klibanov KA (1999) Targeted delivery of gas-filled microspheres, contrast agents for ultrasound imaging. Adv Drug Deliv Rev 37:139–157 Korpanty G, Carbon JG, Grayburn PA (2007) Monitoring response to anticancer therapy by targeting microbubbles to tumor vasculature. Clin Cancer Res 13(1):323–330 Lassau N, Lamuraglia M, Chami L (2006) Gastrointestinal stromal tumors treated with imatinib: monitoring response with contrast-enhanced sonography. AJR 187:1267–1273 Leong-Poi H, Christiansen J, Heppner P (2005) Assessment of endogenous and therapeutic arteriogenesis by contrast ultrasound molecular imaging of integrin expression. Circulation 111:3248–3254

229 Lindner JR, Song J, Christiansen J (2001) Ultrasound assessment of inflammation and renal tissue injury with microbubblestargeted to P-selectin. Circulation 104:2107–2112 Lyshchik A, Fleischer AC, Huamani J (2007) Molecular imaging of vascular endothelial growth factor receptor 2 expression using targeted contrast-enhanced high-frequency ultrasonography. J Ultrasound Med 26:1575–1586 Palmowski M, Huppert J, Ladewig G (2008) Molecular profiling of angiogenesis with targeted ultrasound imaging: early assessment of antiangiogenic therapy effects. Mol Cancer Ther 7(1):101–109 Pollard RE, Broumas AR, Wisner ER (2007) Quantitative contrast enhanced ultrasound and CT assessment of tumor response to antiangiogenic therapy in rats. Ultrasound Med Biol 33(2):235–245 Porter TS, D’Sa A, Turner C et al (1993) Myocardial contrast echocardiography for the assessment of coronary blood flow reserve: validation in humans. JACC 21(2):349–355 Rafter P, Phillips P, Vannan MA (2004) Imaging technologies and techniques. Cardiol Clin 22:181–197 Rasor NS, Hilmann J, Lange L et al (1992) Ultrasonic contrast medium comprising gas bubbles and solid lipophilic surfactant-containing microparticles and use thereof. US Patent 5’141’738 Takalkar AM, Klibanov AL, Rychak JJ (2004) Binding and detachment dynamics of microbubbles targeted to P-selectin under controlled shear flow. J Control Rel 96:473–482 Unger EC, Lund PJ, Shen DK et al (1992) Nitrogen-filled liposomes as a vascular US contrast agent: preliminary evaluation. Radiology 185:453–456 Villanueva FS, Lu E, Bowry S (2007) Myocardial ischemic memory imaging with molecular echocardiography. Circulation 115(3):345–352 Wei K, Jayaweera AR, Firoozan S (1998) Quantification of myocardial blood flow with ultrasound-induced destruction of microbubbles administered as a constant venous infusion. Circulation 97(5):473–483 Weller GER, Villanueva FS, Tom EM (2005) Targeted ultrasound contrast agents: in vitro assessment of endothelial dysfunction and multi-targeting to ICAM-1 and sialyl lewisx. Biotechnol Bioeng 92(6):780–788 Willmann JK, Paulmurugan R, Chen K (2008) US imaging of tumor angiogenesis with microbubbles targeted to vascular endothelial growth factor receptor 2 in mice. Radiology 246(2):508–518

17

PET and SPECT Sibylle I. Ziegler

Contents 17.1 PET and SPECT Principles................................... 231 17.1.1 Positron Emitters vs. Single Photon Emitters........... 231 17.1.2 SPECT Imaging Devices.......................................... 233 17.1.3 PET Imaging Devices............................................... 233 17.2 Image Reconstruction, Quantification.................. 234 17.2.1 Image Reconstruction Algorithms............................ 234 17.2.2 Quantification........................................................... 234 17.3 Future Developments.............................................. 235 References............................................................................ 236

17.1 PET and SPECT Principles Both positron emission tomography (PET) and single photon emission computed tomography (SPECT) use radioactively labeled molecules that interact with a specific cellular target after injection. Images are formed by the detection of gamma rays, X-rays, or annihilation quanta (in the case of positron imaging). If single photon emitters are used, the direction of flight has to be determined by geometric collimation. In contrast, coincidence detection exploits the unique feature of positron annihilation which results in two high energy gamma rays simultaneously emitted back-to-back. In most cases, PET or SPECT instrumentation for small animals consists of scintillation crystals to convert gamma-ray energy into visible light, suitable light sensors, readout electronics, and image processing units. Depending on the application, PET and SPECT have different advantages and disadvantages, so there is not one technology superior for all cases.

17.1.1 Positron Emitters vs. Single Photon Emitters

S.I. Ziegler Nuklearmedizinische Klinik und Poliklinik der TU München, Klinikum rechts der Isar, Ismaninger Strasse 22, 81675 München, Germany e-mail: [email protected]

Depending on the process to be imaged, adequate substances and labeling radioisotopes need to be selected. While PET nuclides offer the opportunity to label molecules without changing their biological behavior (e.g., 11 C labeling), for some slow biological processes it is advantageous to use SPECT tracers labeled with longerlived radioisotopes, such as 99mTc or 125I. This allows much longer scan sequences, following slow processes, which would not be possible with the short-lived positron emitting isotopes. Furthermore, labeling chemistry may

F. Kiessling and B.J. Pichler (eds.), Small Animal Imaging, DOI: 10.1007/978-3-642-12945-2_17, © Springer-Verlag Berlin Heidelberg 2011

231

232

S.I. Ziegler

Table 17.1 Characteristics of the most commonly used radio nuclides in nuclear medicine Radionuclide Decay Photon energy Half life (keV) C

b+

511

20.3 min

N

b

+

511

10.0 min

O

b

+

511

2.07 min

F

b

+

511

110 min

Ga

b

+

511

68 min

Rb

b

+

511

1.25 min

Tc

IT

140

6.03 h

In

EC

172, 247

2.81 day

I

EC

159

13.0 h

I

EC

27–35

59.4 day

Tl

EC

68–80

3.05 day

11 13 15 18 68 82

99m 111 123 125 201

be very different for SPECT and PET isotopes. Typically, it is much simpler and less expensive for SPECT. A unique feature of SPECT imaging is the possibility to acquire data in several energy windows at the same time, offering the opportunity to coinject tracers with different radiolabels for simultaneous detection of different processes (Table 17.1). The radiopharmaceuticals used in PET are labeled with beta emitters. The positron, depending on its object containing positron-emitting radionuclide

Detector

initial energy, will travel a small distance in the surrounding material, losing energy until it annihilates with an electron resulting in the emission of two photons with 511 keV, each in the opposite direction. Imaging is based on the detection of the two annihilation photons by two opposing detectors with a short time window (ns) (Fig. 17.1). Every pair of opposing detectors defines a line-of-response along which the emission point of the two annihilation photons is located; this can be described as an electronic collimation in contrast to geometric collimation in SPECT. In case the positron and electron are not at rest during annihilation, the angle between the two annihilation quanta can deviate from 180° by up to 0.25°. Positron range and noncolinearity of annihilation quanta are physical limits of the spatial resolution in PET imaging. Obviously, these limits especially influence the quality of preclinical imaging. There are no such limits for SPECT nuclides. The drawback in SPECT is found in the necessity of geometrical collimation, which reduces detection sensitivity. Collimators are made of highly absorbing material, such as lead or tungsten, and their exact geometry needs to be defined according to the energy of the radionuclide and the spatial resolution. In clinical SPECT systems, collimators with parallel hole geometry are most widely used. For preclinical measurements, in which small structures are being measured, pinhole collimators offer excellent submillimeter

annihilation event

Detector

accepted by coincidence detection rejected by coincidence detection

Fig. 17.1 Volume (shaded area) from which a pair of simultaneously emitted annihilation photons can be detected in coincidence by a pair of detectors. Not all decays in this volume will lead to recorded events, because it is necessary that both photons

strike the detectors. Outside the shaded volume, it is impossible to detect annihilation photons in coincidence unless one or both undergo a Compton scatter in the tissue and change direction (after Cherry et al. (2003))

233

17 PET and SPECT Fig. 17.2 (a, b) Two types of collimators used to project “g-ray images” onto the detector of a gamma camera. 0, radioactive object; I, its projected image (after Cherry et al. (2003))

a

b Image

Image

Object

Object

spatial resolution (see Fig. 17.2). In general, caused by the geometric collimation, a strong dependence of spatial resolution vs. distance from the collimator is observed in SPECT. To improve sensitivity, there are systems available with many pinholes, measuring overlapping or nonoverlapping projections. Advanced image reconstruction techniques have been developed for these geometries (Schramm et al. 2003; Beekman and van der Have 2007).

17.1.2 SPECT Imaging Devices The main detector module of a SPECT imaging device is a gamma camera. A typical gamma camera consists of a collimator, a scintillator crystal, a light guide, and a number of photomultiplier tubes (PMTs). The energy information is acquired by the sum of the PMT signals and the position information of the gamma-ray interaction within the crystal is determined using a centroid algorithm (Anger 1958). A light guide is used in order to efficiently collect the scintillation light and distribute it among the PMTs. Segmented scintillator crystals and position-sensitive PMTs (PSPMTs) are also used in current systems in order to improve the spatial resolution for the preclinical application (Zeniya et al. 2006). A typical SPECT device contains one, two, or more gamma cameras which rotate around the object. Stationary systems with several detector heads are currently available for small animals (Wirrwar et al. 2005; van der Have et al. 2009). Cost-effective approaches implemented stationary detectors in combination with the possibility of rotating the animal inside the field-of-view.

17.1.3 PET Imaging Devices Block detectors are traditionally used in clinical PET, which consist of a number of individual crystal elements read out by a small number of photomultiplier tubes (typically four). The sum of the four signals from the PMTs is used to determine the energy information, while the crystal where the interaction took place is determined by the four PMT signals using Anger logic (Casey and Nutt 1986). Another option is the readout of a large number of crystals with a continuous light guide and an array of PMTs, similar to a gamma camera (Surti and Karp 2004). Both concepts have been translated into preclinical tomograph designs (Cherry et al. 1997; Huisman et al. 2007), with the commercial version of the microPET concept currently being the most widely used principle (Kemp et al. 2009). Alternative to the block detector concept is the individual crystal readout which can be achieved by using more compact photodetectors such as avalanche photodiodes (Lecomte et al. 1996; Ziegler et al. 2001). The use of small size crystals is advantageous in PET, since the crystal size will determine a lower limit in the achieved spatial resolution. However, the need of a large number of electronic channels to individually process the large number of detector signals significantly increases the cost of such a design. On the other hand, this concept was the base for developments toward integrated PET/MR imaging (e.g., (Judenhofer et al. 2007)). The thickness of the crystal is critical for both PET and SPECT imaging. A thick crystal increases the gamma detection efficiency and hence the sensitivity

234

of the imaging device. However, in SPECT large thickness enhances the spread of the scintillation light before reaching the PMT; therefore, a larger uncertainty in the localization of the photon interaction is introduced. In PET, and especially in small animal systems, making use of depth of interaction (DOI) information plays an important role in the improvement of the spatial resolution at the edges of the FOV when using long crystals. In order to minimize the crystal penetration effects and thus to determine the position along the crystal where the photon interaction took place, PET scanner geometries were suggested with two radial crystal layers of either the same or different material instead of a single scintillation radial crystal layer (Schmand et al. 1998; Seidel et al. 2003; Chung et al. 2004). Advanced readout schemes of the scintillation light spread in a continuous crystal offer continuous depth measurement. DOI measurement is not yet a standard feature of commercial systems, but will become more important as the ring diameter becomes smaller and the reconstructed field-of-view covers a large fraction of the detector diameter.

17.2 Image Reconstruction, Quantification 17.2.1 Image Reconstruction Algorithms Tomographic image reconstruction is based on measured estimation of the integral of radiotracer distribution as seen under different angles (projections). The reconstruction algorithms used for both SPECT and PET are the same, although, as previously described, the hardware and the principle of operation are different for the two modalities. These algorithms are used to produce a volume representation of the radiotracer distribution. The reconstruction algorithms can be divided into two groups: analytical and statistical. For each projection, the acquired data are organized as number of counts along each line-of-response. In SPECT the lines are defined as the radial extensions of the collimator holes across the field-of-view, while in PET a line-of-response is the line connecting a detector pair. For the total number of projections, it is convenient to reorganize the data into a 2-D matrix called sinogram, which contains the number of counts for

S.I. Ziegler

every LOR at each angular view, namely for every radial distance r and every angle q. Sinograms are then used as input to the reconstruction algorithms in order to generate the final image. Analytical reconstruction algorithms model the measurement of radiotracer distribution in a simplified way so that an exact solution may be calculated analytically. However, these algorithms ignore a number of physical effects during acquisition such as limited sampling, Poisson statistics in photon counting, attenuation, or radiotracer decay, resulting in reduced image accuracy. Filtered back projection (FBP) has long been the most widely used analytical reconstruction algorithm. Data in each LOR are homogeneously backprojected and filtered. The filter introduces negative contributions which cancel out the positive counts in areas of zero activity. The choice of filter and cut-off frequency determines resolution and noise level in the reconstructed image. On the other hand, statistical, iterative reconstruction algorithms compensate for the inaccuracies introduced to the image by including in the initial estimate of the image the above-mentioned physical processes. Thus, the model becomes more complicated and an analytical solution is impossible to compute. An iterative calculation is thus needed which starts with an estimate of tracer distribution, computes the forward-projection, and compares calculated and measured projections. The image is updated according to the differences found between calculated and measured projections until the estimate agrees with the measured data. The model used in the algorithm can account for geometrical and detector effects, yielding improved image quality especially in the case of low counting statistics. Maximum likelihood expectation maximization (MLEM) is most commonly used. Since convergence is slow, accelerated algorithms have become most important in clinical routine, such as ordered subsets expectation maximization (OSEM) (Hudson and Larkin 1994).

17.2.2 Quantification Quantification of activity concentration in the reconstructed images is most important, e.g., if changes of tracer accumulation are being measured during therapy. It is the basis for tracer kinetic modeling, which yields parameters characterizing the biological process under

235

17 PET and SPECT

investigation. Several factors affect quantification. While PET is widely accepted to be a quantitative imaging method, this is not the case for SPECT, although reconstruction and correction techniques have become available for quantitative SPECT. The main effects which determine quantification are attenuation and scatter in the object of interest and spatial resolution of the imaging device. Gamma rays traversing material undergo scattering and photoabsorption effects; the magnitude of these effects depend on the energy of the gamma rays. In tissue, annihilation quanta mainly interact via Compton scattering, while lower energy gamma rays, typical for SPECT, undergo photo absorption as well. In either case, gamma ray intensity along a certain line is attenuated and needs to be corrected. Compton scattering in the object causes gamma rays to deviate from their original line of flight and to lose some energy. Depending on the energy resolution of the detector system, they may still be detected and falsely attributed to a different line. In contrast to the clinical case, it can be questioned whether attenuation and scatter need to be corrected for when imaging small animals, since the amount of tissue through which the gamma rays pass is much smaller. It should be kept in mind that these effects are much more pronounced when imaging a low-energy nuclide such as I-125. For example, it has been shown for commercial small animal pinhole SPECT systems that for determination of activity within 5%, appropriate correction methods need to be applied (Hwang et al. 2008; Chen et al. 2009). Both, in PET and SPECT, attenuation and scatter correction algorithms rely on the availability of structural information, which can be gained from radionuclide-transmission or CT scans. The latter is routinely used in dual-modality scanners (Chow et al. 2005; Hwang and Hasegawa 2005; Vanhove et al. 2009). The distance-dependent spatial resolution of SPECT systems also needs to be taken into account if quantitative reconstruction is aimed for. Developments in SPECT reconstruction algorithms include descriptions of the resolution degradation, based on measurements, in order to minimize this effect (Liang et al. 1992). In PET imaging, spatial resolution is nearly independent of the position along a line-of-response. For noncentral lines-of-response, the so-called parallax error introduces spatially variable resolution. Oblique angles cause the volume of coincidence to be much wider, thus worsening the spatial resolution. Parallax

error is more pronounced for small detector ring diameters, as it is the case in devices for small animal imaging. Several schemes are being investigated for reducing the parallax effect in PET by exploiting DOI detection schemes. The spatial resolution is also affecting quantification significantly due to the partial volume effect. The partial volume effect is the bias introduced to the estimation of the radiotracer concentration due to the limited spatial resolution of the imaging device and is dependent on both the size and shape of the object (Hoffman et al. 1979) and on the relative radiotracer concentration with respect to the surroundings. In general, the partial volume effect is minimized if the size of the object is relatively large in comparison to the system’s spatial resolution. In order to acquire quantitative information about the radiotracer distribution, compensation for partial volume effects is a prerequisite (Chen et al. 1999). Advanced reconstruction algorithms include the detector blur in the system matrix description, thus offering reconstructed images with enhanced resolution (e.g., (Rafecas et al. 2004)). There are other less important effects which influence image quality. In the case of PET, the detection of random coincidences may increase background in the reconstructed image. Owing to the finite time resolution of the detectors, a coincidence event is the detection of two photons by two opposing detectors within a specific time window. This window is usually chosen to be twice the system’s time resolution. Since from theory, the random coincidence rate is proportional to the width of the time coincidence window, the number of random events may be reduced by improving the system time resolution, namely by choosing fast scintillators, photodetectors, and electronics. The detection and subtraction of random coincidences from measured data are performed by various methods which may be software- or hardware-based (Brasse et al. 2005).

17.3 Future Developments Obviously, spatial resolution is the primary requirement for animal imaging instruments. But sensitivity is as important since the amount of activity which can be injected may be limited by the specific activity (ratio of labeled tracer to nonlabeled substance) of the radiotracer.

236

Highest spatial resolution in preclinical SPECT in the range of a few hundred mm can be achieved with pinhole collimators, which, on the other hand, have a very low sensitivity (Meikle et al. 2005). In order to increase sensitivity, several detectors are used. The spatial resolution of commercial PET devices for imaging small animals has reached the 1.0–1.5 mm region (Kemp et al. 2009), with sub-mm imaging possible (Schafers et al. 2005). Sensitivity of small animal PET scanners is increased by a long axial extent and 3D data acquisition. Detector development will aim at cost-effective designs covering a large field-of-view. Parallel readout of many electronic channels with high count rate capability is a challenge. Both PET and SPECT reconstruction algorithms are continuously being refined, including models for improved system matrix generation, which will incorporate the specific detector configurations and characteristics.

References Anger HO (1958) Scintillation camera. Rev Sci Instr 29:27–33 Beekman F, van der Have F (2007) The pinhole: gateway to ultra-high-resolution three-dimensional radionuclide imaging. Eur J Nucl Med Mol Imaging 34(2):151–161 Brasse D, Kinahan PE et al (2005) Correction methods for random coincidences in fully 3D whole-body PET: impact on data and image quality. J Nucl Med 46(5):859–867 Casey ME, Nutt R (1986) Multicrystal two dimensional BGO detector system for positron emission tomography. IEEE Trans Nucl Sci 33:460–463 Chen CH, Muzic RF Jr et al (1999) Simultaneous recovery of size and radioactivity concentration of small spheroids with PET data. J Nucl Med 40(1):118–130 Chen CL, Wang Y et al (2009) Toward quantitative small animal pinhole SPECT: assessment of quantitation accuracy prior to image compensations. Mol Imaging Biol 11(3):195–203 Cherry SR, Shao Y et al (1997) MicroPET: a high resolution PET scanner for imaging small animals. IEEE Trans Nucl Sci 44:1161–1166 Cherry SR, Sorenson JA et al (2003) Phys Nucl Med. Saunders, Philadelphia Chow PL, Rannou FR et al (2005) Attenuation correction for small animal PET tomographs. Phys Med Biol 50(8):1837–1850 Chung YH, Choi Y et al (2004) Characterization of dual layer phoswich detector performance for small animal PET using Monte Carlo simulation. Phys Med Biol 49(13):2881–2890 Hoffman EJ, Huang SC et al (1979) Quantitation in positron emission computed tomography: 1. effect of object size. J Comput Assist Tomogr 3(3):299–308

S.I. Ziegler Hudson HM, Larkin RS (1994) Accelerated image reconstruction using ordered subsets of projection data. IEEE Trans Med Imag 13:601–609 Huisman MC, Reder S et al (2007) Performance evaluation of the Philips MOSAIC small animal PET scanner. Eur J Nucl Med Mol Imaging 34(4):532–540 Hwang AB, Hasegawa BH (2005) Attenuation correction for small animal SPECT imaging using x-ray CT data. Med Phys 32(9):2799–2804 Hwang AB, Franc BL et al (2008) Assessment of the sources of error affecting the quantitative accuracy of SPECT imaging in small animals. Phys Med Biol 53(9):2233–2252 Judenhofer MS, Catana C et al (2007) PET/MR images acquired with a compact MR-compatible PET detector in a 7-T magnet. Radiology 244(3):807–814 Kemp BJ, Hruska CB et al (2009) NEMA NU 2-2007 performance measurements of the Siemens Inveon preclinical small animal PET system. Phys Med Biol 54(8):2359–2376 Lecomte R, Cadorette J et al (1996) Initial results from the Sherbrooke avalanche photodiode positron tomograph. IEEE Trans Nucl Sci 43:1952–1957 Liang Z, Turkington T et al (1992) Simultaneous compensation for attenuation, scatter and detector response for SPECT reconstruction in three dimensions. Phys Med Biol 37:587–603 Meikle SR, Kench P et al (2005) Small animal SPECT and its place in the matrix of molecular imaging technologies. Phys Med Biol 50(22):R45–R61 Rafecas M, Mosler B et al (2004) Use of Monte-Carlo based probability matrix for 3D iterative reconstruction of MADPET-II data. IEEE Trans Nucl Sci 51(5):2597–2605 Schafers KP, Reader AJ et al (2005) Performance evaluation of the 32-module quadHIDAC small-animal PET scanner. J Nucl Med 46(6):996–1004 Schmand M, Eriksson L et al (1998) Performance results of a new DOI detector block for a high resolution PET-LSO research tomograph: HRRT. IEEE Trans Nucl Sci 45:3000–3006 Schramm NU, Ebel G et al (2003) High-resolution SPECT using multipinhole collimation. IEEE Trans Nucl Sci 50(3):315–320 Seidel J, Vaquero JJ et al (2003) Resolution uniformity and sensitivity of the NIH ATLAS small animal PET scanner: comparison to simulated LSO scanners without depth-of-interaction capability. IEEE Trans Nucl Sci 50:1347–1350 Surti S, Karp JS (2004) Imaging characteristics of a 3-dimensional GSO whole-body PET camera. J Nucl Med 45(6):1040–1049 van der Have F, Vastenhouw B et al (2009) U-SPECT-II: an ultra-high-resolution device for molecular small-animal imaging. J Nucl Med 50(4):599–605 Vanhove C, Defrise M et al (2009) Improved quantification in single-pinhole and multiple-pinhole SPECT using micro-CT information. Eur J Nucl Med Mol Imaging 36(7):1049–1063 Wirrwar AK, Nikolaus S et al (2005) TierSPECT: performance of a dedicated small-animal-SPECT camera and first in vivo measurements. Z Med Phys 15(1):14–22 Zeniya T, Watabe H et al (2006) Use of a compact pixellated gamma camera for small animal pinhole SPECT imaging. Ann Nucl Med 20(6):409–416 Ziegler SI, Pichler BJ et al (2001) A prototype high resolution animal positron tomograph with avalanche photodiode arrays and LSO crystals. Eur J Nucl Med 28(2):136–143

Radiotracer I: Standard Tracer

18

Walter Mier and Uwe Haberkorn

Contents

18.1 Introduction

18.1 Introduction............................................................. 237

Molecular imaging combines non-invasive imaging methods with tools of molecular and cellular biology. Traditional imaging approaches in nuclear medicine were solely based on the detection of anatomical and structural features. Complementary to these conventional methods, advances in biology made it possible to visualize cellular and biochemical processes. Until the event of PET imaging, the spectrum of nuclides used for tracers was largely restricted to iodine isotopes and 99 m-technetium. Today, several PET nuclides have largely extended the possibilities of tracer design. The uptake of several archetypical diagnostic radiopharmaceuticals is illustrated in Figs. 18.1–18.3. Yet most of the commonly used tracers are derivatives of natural products. At present, the main effort in the field of radiotracer development is the identification of specific probes for targeting cellular structures, e g. tumourassociated receptors. These new ligands can be used for diagnosis as well as for treatment, a concept known as theranostics. Further, non-invasive imaging paradigms studied in small animals are currently being translated into the clinics. In the future, molecular imaging will provide new ways in the diagnosis of cancer and monitoring treatment response in oncology. For instance, studies using PET-based imaging of reporter genes have already been described in gene therapy approaches.

18.2 PET Tracers............................................................ 237 18.2.1 18F-FDG, [18F]-Fluoro-2-Deoxy-2-d-Glucose......... 237 18.2.2 18F-FLT, 3¢-Deoxy-3¢-[18F]Fluorothymidine............ 239 18.2.3 18F-FMISO, [18F]Fluoromisonidazole...................... 240 18.2.4 11C-MET, L-[methyl-11C]Methionine....................... 240 18.2.5 18F-Fluoride, [18F]Fluoride....................................... 240 18.2.6 18F-FECH, [18F]Fluoroethylcholine......................... 241 18.2.7 13N-Ammonia, [13N]Ammonia................................. 241 18.2.8 68Ga-DOTATOC, [68Ga](1,4,7,10Tetraazacyclododecane-N,N¢,N¢¢,N¢¢¢Tetraacetic Acid)-Tyr3-Octreotide............................. 242 18.2.9 11C-Acetate, [11C]Acetate........................................ 242 18.3 SPECT Tracers....................................................... 242 18.3.1 99mTc-Microspheres................................................... 242 18.3.2 123I-MIBG, Meta-[123I]Iodobenzylguanidine........... 243 18.3.3 99mTc-Pertechnetate, [99mTc]TcO4−. ......................... 243 18.3.4 99mTc-MIBI, [99mTc] Tc-2-Methoxyisobutylisonitrile................................ 244 18.3.5 111In Zevalin, [111-In] In-Ibritumomab Tiuxetan...... 244 18.3.6 99mTc-MAG3 [99mTc] Tc-Mercaptoacetyl-Triglycine)................................. 245

18.2 PET Tracers

W. Mier and U. Haberkorn University Hospital Heidelberg, Nuclear Medicine, Im Neuenheimer Feld 400, 69120 Heidelberg, Germany e-mail: [email protected]

18.2.1 18F-FDG, [18F]-Fluoro-2-Deoxy2-d-Glucose The glucose analogue 18F-FDG is the most versatile PET radiopharmaceutical with major applications in

F. Kiessling and B.J. Pichler (eds.), Small Animal Imaging, DOI: 10.1007/978-3-642-12945-2_18, © Springer-Verlag Berlin Heidelberg 2011

237

238

W. Mier and U. Haberkorn

Fig. 18.1 Receptor binding, Metabolism and Mineralization: (1) Receptor-mediated endocytosis of the antibody Zevalin allows the targeting of NHL. (2) Myocard cells metabolize

11

C-acetate in the tricarboxylic acid (TCA) cycle. (3) 18F-Fluoride ions substitute hydroxide ions in hydroxyapatite of the cancellous bone matrix

Fig. 18.2 Transport: (1) The amino acid 18F-O-(2-fluoroethyl)tyrosine (FET) is taken up into cells by the L-type amino acid transporter system. (2) Iodide is taken up by the human sodium iodide symporter (NIS) on thyroid cancer tissues. (3)

F-Fluorodeoxyglucose (FDG) is taken up into cells by glucose transporters. (4) 123I-Meta-iodobenzylguanidine (MIBG) is transported by the norepinephrine transporter (NET) into adrenergic neurons

Fig. 18.3 Hypoxia, enzyme activity and morphology: (1) 18 F-Fluoromisonidazole (FMISO) is reduced and covalently bound within hypoxic cells. (2) The nucleoside 18F-3¢-deoxy-3¢-

fluorothymidine (FLT) is phosphorylated by thymidine kinase-1. (3) 99mTc-labelled microspheres (i.e. macroaggregated human serum albumin) can be trapped in the sentinel lymph node

18

239

18 Radiotracer I: Standard Tracer

oncology, neurology, and cardiology. Most malignant tumours show an upregulation of glycolysis, resulting in increased glucose consumption. FDG is taken up into cells by glucose transporters and then phosphorylated to FDG-6-phosphate by the enzyme hexokinase, the first enzyme of glycolysis. FDG-6-phosphate is not metabolized within the cells because it is a poor substrate for the further enzyme systems of glycolysis. Since the dephosphatase activity is downregulated in tumours, a dephosphorylation does not occur. As a consequence, FDG-6-phosphate is trapped in the cells because the charge prevents the diffusion through the cellular membrane (Fig. 18.4). OH O

HO HO

OH 18F

Fig. 18.4 Chemical structure of deoxy-2-D-glucose

assessment of myocardial viability and inflammatory plaque detection in cardiovascular diseases, the diagnosis of dementia and the detection of inflammatory foci in patients with fever of unknown origin.

18.2.2 18F-FLT, 3¢-Deoxy-3¢ -[18F]Fluorothymidine F-FDG uptake is not tumour-specific and false-positive findings can occur in inflammatory lesions. In contrast, nucleoside metabolism correlates with the proliferative activity of malignant tumours. 18F-FLT is transported via nucleoside transporters and phospho rylated to 3¢-fluorothymidine monophosphate by thymidine kinase 1 and reflects thymidine kinase 1 activity. The phosphorylated tracer is trapped in the tumour cells (Fig. 18.5). 18

F-FDG, [18F]-Fluoro-2-

18

O HN

18.2.1.1 Target Glucose utilization, cellular uptake predominantly by glucose transporter 1 (GLUT-1) followed by phosphorylation by hexokinase II (HK-II).

HO

O

N O

18

F

18.2.1.2 Applications The clinical impact of 18F-FDG PET has been reported for many different tumour types, such as lung tumours, colorectal carcinomas, breast carcinoma, and lymphomas. Since PET allows quantitative values, the tracer has also been used for therapy monitoring by the assessment of changes in glucose metabolism. These changes usually occur very early during treatment, thus allowing a rapid decision about the effectiveness of a specific treatment schedule. This may lead to early changes in therapy management. Furthermore, the values obtained with FDG-PET have been used as predictive markers in a couple of tumours including lymphomas and oesophageal cancer. Non-oncological applications include the

Fig. 18.5 Chemical structure of fluorothymidine

F-FLT, 3¢-deoxy-3¢-[18F]

18

18.2.2.1 Target Proliferation, nucleoside uptake and phosphorylation by thymidine kinase-1 (TK-1).

18.2.2.2 Applications The tracer has been used as a proliferation marker in a variety of tumours.

240

W. Mier and U. Haberkorn

18.2.3 18F-FMISO, [18F] Fluoromisonidazole

S

Large tumours show areas with low blood perfusion. The reduced oxygenation of these areas causes a reductive milieu. Tumour cells in hypoxic areas show reduced metabolism and sensitivity against radiation and chemotherapy. The hydrophilic tracer 18F-FMISO diffuses across cell membranes. It is subsequently reduced and covalently attached to macromolecules within hypoxic cells (Fig. 18.6).

NO2 18

F

N

11CH3

OH

H2N O

Fig. 18.7 Chemical structure of Methionine

C-MET, L-[methyl-11C]

11

18.2.4.1 Target Amino acid transport, by L-type amino acid transporters sodium-dependent transport systems as well as protein synthesis.

N

OH

Fig. 18.6 Chemical structure of [18F]Fluoromisonidazole

18.2.4.2 Applications C-Methionine has been applied in brain tumours, head and neck tumours and lung tumours, mainly for diagnosis, but also in selected cases for therapy monitoring.

11

18.2.3.1 Target Hypoxia, reductive potential in hypoxic tissues.

18.2.5 18F-Fluoride, [18F]Fluoride 18.2.3.2 Applications Used for the detection of hypoxia in tumours, myocardial infarcts, or cerebral ischemia.

18.2.4 11C-MET, L-[methyl-11C] Methionine

F-Fluoride is a sensitive indicator of skeletal pathology. It is taken up by mineralizing bone in proportion to osteoblastic activity; normal bone structures demonstrate uniform uptake of 18F-fluoride. The 18F-fluoride ion exchanges with hydroxyl groups on hydroxyapatite to form fluoroapatite. 18F-fluoride is used as an alternative to the clinical bone scanning 99mTc-diphosphonate SPECT agents (Fig. 18.8).

18

18 –

The malignant transformation goes along with the overexpression of a variety of amino acid transporters. Consequently, radiolabelled amino acids can be used as functional biomarkers to visualize amino acid transport and to some extent protein synthesis. 11C-Methionine, the most commonly used amino acid-based PET tracer, is considered to be more tumour-specific than FDG because its uptake in inflammatory cells is low (Fig. 18.7).

F

Fig. 18.8 Chemical structure of [18F]Fluoride

18.2.5.1 Target Mineralization, malignant processes stimulate the osteoblastic activity, this causes increased fluoride incorporation.

241

18 Radiotracer I: Standard Tracer

18.2.7 13N-Ammonia, [13N]Ammonia

18.2.5.2 Applications Fluoride is increasingly used as an alternative to the classical bone scan for the detection of bone metastases. The possibility to obtain quantitative values should also offer new aspects such as the therapy monitoring of benign bone diseases.

18.2.6 18F-FECH, [18F]Fluoroethylcholine FECH is a specific substrate of choline kinase, an enzyme commonly overexpressed in malignant lesions. Phosphorylation with choline kinase results in intracellular trapping of FECH. Phosphorylated choline derivatives can be incorporated into phospholipids. The major application of FECH is prostate cancer as FDG PET-CT is generally not suitable for diagnosing prostate cancer showing low glycolysis rates (Fig. 18.9).

N-ammonia allows to measure regional myocardial perfusion in normal and diseased states. The small molecule moves from the vascular space to tissue by both active transport and passive diffusion. Once in the myocardial cells, retention of 13N-ammonia involves predominantly the enzymatic conversion of 13N-ammonia and glutamic acid to 13N-labelled. Because of the requirement of ATP for the metabolism, intracellular levels of 13N-ammonia reflect cellular processes dependent on viable myocardium (Fig. 18.10). 13

Fig. 18.10 Chemical structure of [13N]Ammonia

18.2.7.1 Target Myocardial blood flow followed by metabolism by the glutamic acid-glutamine pathway. 18.2.7.2 Applications

18F

+ N

CI OH

The tracer is used for the quantitative determination of myocardial blood flow.

Fig. 18.9 Chemical structure of [18F]Fluoroethylcholine

18.2.6.1 Target Choline metabolism, phosphorylation by choline kinase and subsequent incorporation into phospholipids of cell membranes.

18.2.6.2 Applications Choline-based tracers are used for the diagnosis of primary and recurrent prostate carcinoma, the planning of biopsy, and therapy monitoring.

18.2.8 68Ga-DOTATOC, [68Ga] (1,4,7,10-TetraazacyclododecaneN,N¢,N¢¢,N¢¢¢-Tetraacetic Acid)Tyr3-Octreotide Due to its excellent performance, the peptide derivative DOTATOC is the gold standard of peptide radiopharmaceuticals. A large number of carcinoid tumours have been shown to express receptors for somatostatin. Somatostatin receptor-binding peptides containing chelators for various radiometals can be used for the diagnosis and therapy of somatostatin receptor-positive tumours. DOTATOC labelled with 68Ga allows the diagnosis by PET; 90Y labelled DOTATOC has been proven to be effective for endoradiotherapy (Fig. 18.11).

242

W. Mier and U. Haberkorn

Fig. 18.11 Chemical structure of [68Ga] (1,4,7,10-tetraazacyclododecaneN,N¢,N¢¢,N¢¢¢-tetraacetic acid)Tyr3-octreotide

OH

O OH

OH O

N N 68Ga N N

O

O NH

NH O

S

O

HO

NH

HO

O HN

O NH

O

NH

O

S HO

NH

NH

H3C

NH O

NH2

OH

18.2.8.1 Target

18.2.9.1 Target

Peptide receptor binding, the binding to the somatostatin receptor subtypes SSTR2 and SSTR5 induces receptor-mediated endocytosis of the tracer.

Oxidative metabolism, fatty acid metabolism, TCA cycle in myocard cells, and fatty acid synthetase (FAS) in tumour cells.

18.2.8.2 Applications

18.2.9.2 Applications

DOTATOC is used for the diagnosis and therapy monitoring of neuroendocrine tumours.

PET tracer for myocardial oxidative metabolism and regional myocardial blood flow as well as renal, pancreatic, and prostate tumours.

18.2.9 11C-Acetate, [11C]Acetate Fatty acids are the primary metabolic energy source for the myocardium. 11C-acetate, a readily utilized myocardial substrate, is predominantly metabolized to carbon dioxide in the tricarboxylic acid cycle (TCA). Consequently, PET imaging of 11C-acetate allows the measurement of both myocardial flow and oxidative lipid metabolism. In contrast to the predominant oxidative metabolism in myocardial cells, tumour cells involve the fatty acid synthesis pathway in 11C-acetate metabolism. Consequently, 11C-acetate can be used as a marker for the expression fatty acid synthetase, an enzyme overexpressed in prostate carcinomas and other cancers (Fig. 18.12).

Fig. 18.12 Chemical structure of [11C]acetate

18.3 SPECT Tracers 18.3.1 99mTc-Microspheres Localization of the sentinel node can be achieved using a radiolabelled colloid such as 99mTc-HSA microspheres. The “sentinel” node is the very first lymph node(s) to receive drainage from a cancer-containing area; it can be the only site of lymphatic metastasis. A negative result from the intraoperative sentinel lymph node negates the possibility of more-distant lymphatic metastatic spread. This process is part of “staging” the cancer. Once the labelled tracer has reached the nodes, the surgeon scans the area with a handheld gamma counter to detect the tracer. The sentinel node is removed and examined by a pathologist procedure. Thin intraoperative procedure is used to decide whether further dissection is required, most commonly for axillary lymph node dissection to avoid extensive removal of underarm nodes in case of breast cancer (Fig. 18.13).

243

18 Radiotracer I: Standard Tracer

18.3.2.1 Target Transport by the norepinephrine transporter followed by accumulation in neurosecretory storage granules of adrenergic tissue. Fig. 18.13 Structure of [99mTc]-microspheres

18.3.2.2 Applications 18.3.1.1 Target Filtration of particles from the lymphatic ducts that drain from a tumour into the sentinel lymph node.

Diagnosis of phaeochromocytoma or paraganglioma, measurement of myocardial sympathetic innervation.

18.3.3 99mTc-Pertechnetate, [99mTc]TcO4−

18.3.1.2 Applications Breast cancer, Melanoma, Colon cancer, Prostate cancer.

18.3.2 123I-MIBG, Meta-[123I] Iodobenzylguanidine Some tumours show a metabolic pathway, in which the hormones dopamine, adrenaline and noradrenaline are synthesized from phenylalanine. The guanidine analogue 123I-MIBG is chemically similar to these hormones. It enters a metabolic pathway of noradrenaline, a catecholamine that acts as an adrenergic transmitter and is synthesized and stored in the storage granules (synaptosomes) of the sympathomedullary cells. MIBG uptake is studied in nuclear cardiology to demonstrate myocardial sympathetic innervation to characterize acute and chronic alterations in sympathetic nerve function (Fig. 18.14).

Due to its comparable charge/radius ratio, pertechnetate is incorporated into the thyroid gland similarly as iodide. This uptake is specific for tissues expressing the sodium iodide symporter such as the thyroid, the stomach, etc. 99m Tc-pertechnetate has shown advantages over 123I- and 131 I-iodide with respect to image quality, procedure and radiation dose for examinations of thyroid uptake and scintigraphy. Consequently, studies of the thyroid gland (morphology, vascularity, and function) are performed primarily using pertechnetate. The pertechnetate ion is not incorporated into thyroglobulin, and therefore, not stored in the thyroid gland. 99mTc-pertechnetate cannot pass the blood–brain barrier; it accumulates primarily in the choroid plexus. It is therefore used to study the blood perfusion, regional accumulation, and cerebral lesions in the brain. 99mTc pertechnetate can be used for the diagnosis of papillary and follicular thyroid carcinoma (Fig. 18.15).

99m[TcO4] H2N

NH2

Fig. 18.15 Chemical structure of [99mTc]Pertechnetate

N

18.3.3.1 Target 123

Fig. 18.14 Chemical structure of meta-[123I]iodobenzylguanidine

Thyroid-specific uptake by the human sodium iodide symporter (hNIS).

244

W. Mier and U. Haberkorn

18.3.3.2 Applications Evaluation of thyroid diseases. Diagnosis of thyroid cancers and their metastases. Detection of ectopic gastric mucosa as a source of gastrointestinal bleeding.

CH3 O O C H 3C C CH3 H3C

O

H 3C

18.3.4 99mTc-MIBI, [99mTc] Tc-2Methoxyisobutylisonitrile The cationic, lipophilic complex 99mTc-MIBI enters the cell by passive diffusion across negatively charged plasma and mitochondrial membranes. It was originally developed for myocardial perfusion imaging, which is still its major application. However, it has been shown that 99mTc-MIBI can be used to evaluate oncological questions, i.e. overexpression of P-glycoprotein (Pgp), one of the primary mechanisms of multidrug resistance (MDR) in several diseases, including multiple myeloma (Fig. 18.16).

18.3.4.1 Target Passive diffusion across negatively charged membranes. The retention is influenced by transport proteins such as the Pgp pump and the multidrug resistance-associated protein pump.

18.3.4.2 Applications Myocardial perfusion, detection of parathyroid adenoma, used to estimate drug efflux.

N

CH 3

C O

C

H 3C

C

N

C

C

N

O

O

CH3

CH 3

99mTc

H3C H 3C

O

O C

C

C

N

CH3

C C

N

N H3C C CH3

O C

H 3C

C

CH3 O

CH3

O C O CH3

Fig. 18.16 Chemical structure of [99mTc] Tc-2-methoxyisobu tylisonitrile

18.3.5 111In Zevalin, [111-In] In-Ibritumomab Tiuxetan Due to its excellent performance, the protein derivative Zevalin is the gold standard of protein radiopharmaceuticals. Rituximab is a chimeric monoclonal antibody directed against the B cell-specific antigen CD20 expressed, which is primarily found on the surface of B cells. The high expression of CD20 on non-Hodgkin’s lymphomas (NHL) allows the treatment with this protein. Zevalin is a derivative of rituximab and can be radiolabelled via the chelator MX-DTPA. Zevalin labelled with 111In allows the diagnosis by SPECT; 90Y labelled Zevalin has been proven to be effective for endoradiotherapy of NHL (Fig. 18.17).

NH

NH COOH

S

HOOC HOOC

Fig. 18.17 Schematic structure of [111In] In-Zevalin

C

C

N

N 111In

N

COOH COOH

245

18 Radiotracer I: Standard Tracer

18.3.5.1 Target

18.3.6.1 Target

Antigen binding to the B cell-specific antigen CD20 allows the targeting of NHL.

Kidney perfusion due to a high glomerular filtration and low kidney retention.

18.3.5.2 Applications Non-Hodgkin’s lymphomas (NHL).

18.3.6 99mTc-MAG3 [99mTc] Tc-Mercaptoacetyl-Triglycine) Radiotracers with low binding to plasma proteins that are filtrated through the glomerular capillary membranes and not secreted or absorbed by the renal tubules can be used to study renal function. 99mTc-MAG3 has a high extraction efficiency from functional kidneys, following active excretion by the tubular and glomerular system. It is used to determine the kidney function and kidney outflow rates (Fig. 18.18).

Fig. 18.18 Schematic structure of [99mTc] Tc-mercaptoacetyltriglycine)

18.3.6.2 Applications Follow-up of kidney transplants. Kidney clearance rate (combined perfusion and functional scintigraphy). Molecular imaging combines non-invasive imaging methods with tools of molecular and cellular biology. Traditional imaging approaches in nuclear medicine were solely based on the detection of anatomical and structural features. Complementary to these conventional methods, advances in biology made it possible to visualize cellular and biochemical processes. Until the event of PET imaging, the spectrum of nuclides used for tracers was largely restricted to iodine isotopes and an 99 m-technetium. Today, several PET nuclides have largely extended the range of different tracers. Today, a broad range of SPECT and PET tracers are available to visualize differentiated uptake in diseased organs/tissues in contrast to healthy tissue. The most commonly targeted cellular differences include transport, enzyme activity, receptor expression, metabolism, and hypoxia. This article provides a comprehensive description of a selection of the most commonly used tracers with respect to their mode of action, structure, molecular target, and clinical application.

Radiotracer II: Peptide-Based Radiopharmaceuticals

19

Roland Haubner and Clemens Decristoforo

Contents

19.1 Introduction

19.1

Recently, radiolabeled receptor-binding peptides have emerged as an important class of radiopharmaceuticals for diagnosis and therapy. Major advantages of peptides compared with antibodies are that they are not immunogenic, show fast diffusion and target localization and can be modified concerning metabolic stability and pharmacokinetics. Advantages compared to small molecular weight compounds are, that they are more tolerant concerning modification necessary for appropriate labeling (e.g., introduction of chelating systems for radio metalation) and strategies for optimizing pharmacokinetics. Most prominent members of this class of tracer are radiolabeled peptides for targeting somatostatin receptors. Some of them are already clinical routine for diagnosis as well as peptide receptor radionuclide therapy of somatostatin expressing tumors. Currently a variety of other peptides including bombesin derivatives, cholecystokinin/gastrin analogs and RGD-containing peptides are evaluated. In this chapter we want to give some guidelines for the development and evaluation of radiolabeled peptides. This includes a summary of strategies for lead structure finding, approaches for optimization of binding affinity, metabolic stability and pharmacokinetic behavior, possible labeling strategies and discussion of techniques for in vitro and in vivo evaluation. Obviously this chapter cannot cover all aspects of the different topics in detail. The first part is thought to give some brief insights in different strategies how to find a new lead structure. The second part gives an overview of different approaches to optimize tracers for the use in nuclear medicine and my help to choose appropriate strategies for the development of new peptide based radiopharmaceuticals. The third part deals with labeling strategies

Introduction............................................................. 247

19.2 Tracer Development............................................... 248 19.2.1 Lead Structure Finding............................................. 248 19.2.2 Tracer Optimization.................................................. 249 19.3 Labeling Strategies................................................. 253 19.3.1 Halogens................................................................... 253 19.3.2 Metals....................................................................... 256 19.4 Tracer Evaluation................................................... 260 19.4.1 In Vitro Evaluation................................................... 260 19.4.2 In Vivo Evaluation.................................................... 263 References............................................................................ 265

R. Haubner (*) and C. Decristoforo Medical University Innsbruck, Anichstrasse 35, 6020 Innsbruck, Austria e-mail: [email protected]

F. Kiessling and B.J. Pichler (eds.), Small Animal Imaging, DOI: 10.1007/978-3-642-12945-2_19, © Springer-Verlag Berlin Heidelberg 2011

247

248

and covers the most suitable labeling techniques for the most important isotopes used for labeling of peptides and will try to give the reader some guidelines for choosing the appropriate labeling techniques. The last part includes theoretical and practical details for carrying out corresponding in vitro and in vivo evaluations of peptides based tracers.

19.2 Tracer Development 19.2.1 Lead Structure Finding Human disease is often caused and/or associated with alterations in protein expression. Peptide-protein interactions play a central role in controlling and modulating cellular function, intracellular communication, immune response, and signal-transduction pathways. With some exceptions, like calcitonin, insulin and oxytocin, naturally occurring peptides are not directly used as drugs. Especially for radiolabeling modifications are required. The different strategies for the design of peptides as drugs, drug candidates and biological tools include structural, conformational, dynamic and topographical considerations. Anyway, the development of new drugs is a challenging task. Many are discovered by chance observations, the scientific analysis of folk medicines or by noting side effects of other drugs. More systematic methods include: (a) the “classical” approach, where at least the structure of the ligand or the structure of the protein and the binding site is known, (b) approaches based on combinatorial chemistry, where synthetic libraries of peptides were produced and screened, and (c) approaches based on phage display libraries, where the properties of bacteriophages are used to produce a great diversity of binding epitopes. The following part gives a brief overview of these major strategies in lead structure finding and is not thought as a comprehensive review of the state-of-the-art. Nevertheless, it should provide some basic information for those who are not familiar with this topic.

19.2.1.1 Rational Drug Design This approach uses information about the structure of a drug receptor or one of its natural ligands to identify

R. Haubner and C. Decristoforo

or create candidate drugs (for more information see (Leach and Harren 2007)). The three-dimensional structure of a protein can be determined using methods such as X-ray crystallography or nuclear magnetic resonance spectroscopy. Based on this structural information databases containing structures of many different chemical compounds can be searched. These approaches allow selection of those compounds that are most likely to interact with the receptor, which can subsequently be synthesized and tested in corresponding biological assays. If the structure of the receptor’s binding pocket is not known, the development can start by using the structure of a known ligand or, at least, binding sequence of the ligand. Resulting from these information, by using structure activity relationship (SAR) studies, molecules with high binding affinity are synthesized. In this process searching of databases to identify compounds with similar properties can narrow down the search as much as possible to avoid large-scale screening. For example, with the knowledge, that the three amino acid sequence Arg-Gly-Asp is essential for binding of a variety of extracellular matrix proteins to the corresponding integrins comprehensive SAR studies resulted in the cyclic pentapeptide cyclo(-ArgGly-Asp-dPhe-Val-), which was the lead structure for development of a diversity of radiolabeled RGDpeptides for monitoring avb3 integrin expression.

19.2.1.2 Combinatorial Chemistry About two decades ago techniques became available to produce and screen a great variety of different compounds in an acceptable time frame. In the beginning the emphasis of combinatorial chemistry was to prepare enormous collections of compounds, but the kind of compounds resulting from this screening process did not have the properties that drugs typically have. Even though large collections of compounds were made, they were not much different from each other due to the limited variability of the synthetic routes used. Therefore, strategies changed and the emphasis now is on preparing smaller collections of compounds not millions, but dozens to hundreds of compounds which are more varied, are pure, and based on basic chemical principles that are consistent with drugs. E.g., Feher and Schmidt (2003) noted that combinatorial chemistry libraries suffer particularly from the lack

249

19 Radiotracer II: Peptide-Based Radiopharmaceuticals

of chirality, as well as structure rigidity, both of which are widely regarded as drug-like properties. The concept of the high throughput organic synthesis (HTOS) starts with preparing the core part of the molecules that constitutes the library by classical organic synthesis, and then “decorating” this core by using databases which allow selection from a wide array of different variations (Chighine et al. 2007). Some key foci of HTOS are automation and development of standard procedures that are likely to work with a wide variety of different starting cores so that there is very minimal chemistry development time required for every new class of compounds. Further improvements will lead more into automated multi-step libraries. Currently a lot of effort is focused on libraries that involve only one chemical step. The automation in chemistry is now evolving to the point where multiple steps synthesis on a single platform can be conducted.

19.2.1.3 Phage Display Libraries The principle underlying all phage display systems is to physically link phenotypes of peptides to the corresponding genotypes (for overview see (Brissette and Goldstein 2007)). The single-stranded DNA of the peptide to be expressed is packed inside bacteriophages. This allows expression of peptides as fusion proteins with the phage coat protein (e.g., pIII or pVIII) on the surface of the phage. Using this technique large phage display libraries containing up to 1010 individual members can be created from batch-cloned gene libraries. These phage libraries can be used for finding new peptide sequences that bind to a given target structure. The enrichment of phages presenting the binding peptide is achieved by affinity selection. For this “panning” process the corresponding purified target receptor is immobilized and phages binding to the target are isolated and amplified by re-infection of E. coli cells. Typically the panning and amplification step is repeated several times before the isolated phages are analyzed for the peptide they present. In addition, phage display libraries are used to identify new peptides targeting particular cells or even organs without knowledge of the target receptor identity. Therefore cultured cells or biopsy specimens were used for in vitro panning. Even in vivo phage library selection was described. In this case, the intravenous administered

phage display library is allowed to circulate for a certain period of time in the vascular system followed by perfusion and washing of the tissue to remove nonspecifically bound phage clones. Phages selectively homing to the vascular bed of the tissue of interest are recovered and amplified for subsequent rounds of selection. This approach has been used predominantly in murine models, where it mainly supplies information on changes in the expression patterns of the vasculature during tumorinduced angiogenesis or during the shift from benign to metastatic state. One limitation of this approach is, that the screening is carried out in an animal model and normally data from such an experiment can’t directly be transferred to the situation in humans. Thus, first attempts are made to use phage display systems in patients (Trepel et al. 2002). Anyway, this approach includes very strict regulatory and ethical requirements and has yet not been established as a routine procedure.

19.2.2 Tracer Optimization After the lead structure is identified strategies to optimize the labeling precursor concerning binding affinity, metabolic stability and pharmacokinetics have to be considered. In many cases binding affinity of the lead structure is already sufficient for monitoring the corresponding target structure. Nevertheless, in some cases improvement of the binding affinity is of advantage. This may be achieved by modification of the peptide sequence or by the formation of multimeric compounds. For the improvement of pharmacokinetic properties a variety of strategies are available. Additionally, especially peptide sequences derived from screening procedures may exhibit low stability towards metabolic degradation requiring modifications to stabilize the peptide. In the attempt to reach these goals a major advantage of peptides as tracers is found in the greater tolerance concerning modification without effecting binding properties compared with small molecular weight compounds.

19.2.2.1 Optimization of Binding Affinity: Multimerisation A variety of strategies can be used to optimize binding affinities of biological active lead structures. Most are focused on modifications of the first generation

250

compounds resulting from the above mentioned lead structure finding process. This can be a comprehensive iterative process involving replacements and modifications within the found amino acid sequence and is out of the scope of this chapter. Another approach introduces multimeric compounds presenting more than one receptor binding-sites in one molecule (Handl et al. 2004). This “multimerisation” approach may result in an improved target affinity and prolonged target retention mainly due to an increased apparent ligand concentration and/or, especially by lager ligand molecules, due to strong cooperative binding (see Fig. 19.1). This strategy was successfully used for the improvement of the binding affinity of RGDcontaining peptides. In almost all approaches the biological active cyclic pentapeptide with the general sequence cyclo(-Arg-Gly-Asp-dPhe/dTyr-Xxx-), where Xxx is either lysine or glutamic acid, was used and conjugated via different linker and/or spacer units. This includes lysine and glutamic acid as branching units and polyethylen glycol as spacer/pharmacokinetic modifier. The resulting derivatives present two to eight binding epitopes per molecule. It was demonstrated in in vitro binding studies using immobilized integrin receptors that the affinity increased depending on the amount of cyclic RGD-peptides presented. Subsequent in vivo studies confirmed improved

R. Haubner and C. Decristoforo

tracer accumulation and retention for some of the multimeric compounds. For example, Janssen et al. (2002) synthesized a dimeric RGD-peptide by coupling two cyclo(-RGDfK) via a glutamic acid linker. For radiolabeling DOTA or HYNIC were conjugated to the free amino function of the linker moiety. The dimeric 99mTc-HYNICE-[c(RGDfK)]2 revealed a tenfold higher affinity for the avb3 integrin as the monomeric 99mTc-HYNICc(RGDfK). Moreover, also activity retention in the tumor was improved compared with the monomeric compound. However, activity retention was also high in kidneys. A more systematic study on the influence of multimerisation on receptor affinity and tumor uptake was carried out by the groups of Kessler and Wester (Poethko et al. 2004a). They synthesized a series of monomeric, dimeric, tetrameric and octameric RGDpeptides. These compounds contain different numbers of c(RGDfE) peptids which are connected via PEG linker and lysine moieties, which are used as branching units. Labeling was based on a chemoselective oxime formation between an aminooxo function at the peptide site and a 18F-labeled aldehyde. They found an increasing binding affinity in the series monomer, dimer, tetramer and octamers in in vitro binding assay. Initial PET images confirmed these findings. The image of mice with both a receptor positive and a a

b

Fig. 19.1 Scheme showing the concept of multimerisation to improve tumor uptake and retention. This concept includes that in one compound multiple ligands binding to the same target are combined. The binding epitopes are connected via corresponding scaffolds (e.g., lysine trees) which also carry the radiolabel.

The increased binding can be due to a strong cooperative binding, which is only possible if the designed molecule is large enough to bind two receptors at the same time (a). Or binding affinity is increased due to an increased apparent ligand concentration (b)

251

19 Radiotracer II: Peptide-Based Radiopharmaceuticals

receptor negative melanoma showed an increasing activity accumulation only in the receptor positive tumor in the series monomer, dimer and tetramer. Comparable findings were reported for other peptides, e.g., for minigastrin derivatives, where dimerisation resulted in considerable improvement of binding affinity. However, this approach will mainly be helpful for radiopharmaceuticals targeting extracellular receptors, which only show limited internalizion of their ligands.

19.2.2.2 Optimization of Metabolic Stability One major drawback when using peptide-based tracer is their low metabolic stability. A variety of peptidases specialized either to cleave N- or C-terminal amide bonds (exopeptidases) or amide bonds within the peptide sequence (endopeptidasis) are found in blood and other tissue such as liver and kidneys. There are several strategies to overcome this problem. The easiest way to protect from degradation by exopeptidases is formation of cyclic peptides where the N-terminal end is conjugated with the C-terminus. This was carried out during the development of avb3-selective RGDpeptides. Radiopharmaceuticals based on this lead structure demonstrated high metabolic stability (for review see (Haubner and Decristoforo 2009)). The problem of this approach is that the biologically active sequence will be fixed in a certain spatial conformation. This can lead to improved target selectivity, as found in the case of the RGD-peptides, but may also result in a complete loss of binding affinity and, thus A1

a Fig. 19.2 Structure of some peptide bond isosters (b–g) and comparison with a dipeptide fragment with classical peptide bond (a): thioamide y(CSNH) (b), methyleneamino y(CH2NH) (c), E-alkene y(CH=CH) (d), ketomethylen y(COCH2) (e), methylenoxy y(CH2O) (f), and carba y (CH2CH2) (g) bond (A1 and A2 indicates corresponding amino acid side chains)

N H

A1

b

N H

O

H N O

A2

e

c

A2

A1 N H

A1

O

H N S

can not be used as a general strategy to improve metabolic stability. In some cases already modification of the C- or/and N-terminal amino acid via conjugation with corresponding chelating moieties required for labeling (e.g., DOTA) may improve stability towards degradation by exopeptidases. Other approaches are focused on the replacement of natural amino acids by their enantiomeric d-analogs or by unnatural amino acids. For example the sequence of octreotide includes dPhe at the N-terminus, dTrp at position 4 and threoninol (threonine were the carboxylate is reduced to an alcohol function) at the C-terminus (Heppeler et al. 2000). In other radiolabeled octreotide derivatives dPhe1 is replaced by the unnatural d-naphtylalanine. Some endopeptidases have special cleavage sites. For example trypsin is specific for Arg and Lys and chymotrypsin for Trp, Tyr and Phe, thus replacement of these amino acids, if found in the corresponding sequence, will be of particular benefit. Obviously, not all amino acids in a sequence can be replaced by either the corresponding d-amino acid or by an unnatural amino acid without influencing the binding affinity. In some cases the essential amino acids are already known, if not an “alanine scan” (for details see (Nicole et al. 2000)) may supply this information. Chemically much more challenging approaches deal with modifications of the amide bounds. A variety of peptide bond isosteres are possible (Vagner et al. 2008). They range from replacement of the nitrogen by a methylene group via reduction of the carbonyl group to the replacement/modification of both the carbonyl group and the nitrogen (see Fig. 19.2). These modifications results

N H

O

A2

N H

f

d

A2

A1

O

A1

O

H N

N H

O

A2

A1

O A2

O

N H

g

O

A2

252

in ketomethylene analogs, in reduced amid bonds, and in carba or alkene analogs, respectively. This approach has been successfully applied to stabilize neurotensin analogs to be used as tracers (Garcia-Garayoa et al. 2009). Anyway, all of these modifications have drastic effects on the properties of the peptide. Depending on the modification carried out this includes higher flexibility of the modified amide bond and changes in electron densities and thus, changes in the hydrogen bridge donor acceptor character. In many cases these modifications result in changes of the peptide structure and will influence the binding affinities of the peptides. Moreover, most of the modifications cannot be carried out on the final peptide, meaning that via complex synthesis routes the corresponding pseudo dipeptide including the peptide bond isostere have to be synthesized, which are in a second step included into the peptide sequence. The complexity of such synthetic routes strongly depends on the amino acids involved. Pseudo dipeptides including complex amino acids with additional side chain functionalities are in general more difficult to produce than pseudo dipeptides including “simple” amino acids like Gly or Phe. In any case, determination of the binding affinity of the peptide after modification is a prerequisite.

19.2.2.3 Optimization of Pharmacokinetics Depending on their amino acid sequence peptides can greatly differ in pharmacokinetic properties. In some cases, they already meet the needs of radiopharmaceuticals for molecular imaging and PRRT. In most cases, improvement of these properties (e.g., blood clearance, renal elimination, protein binding) is necessary. Here it is of advantage that normally not all amino acids of the sequence are required for high affinity binding. Usually there are parts of the peptide where modifications are tolerated. The strategies are manifold and include conjugation of carbohydrates, polyethylen glycol (PEG) moieties, or hydrophilic amino acids. The conjugation of carbohydrates (glycosylation approach) was introduced to e.g., optimize pharmacokinetics of RGD-peptides (Haubner et al. 2001a, b) and octreotide derivatives (Schottelius et al. 2005). For example sugar amino acids (SAA) were conjugated via the e-amino function of the lysine in the sequence cyclo(-Arg-Gly-Asp-dPhe-Lys-). In a murine tumor model the resulting [*I]Gluco-RGD (Haubner et al. 2001a) and [18F]Galacto-RGD (Haubner et al. 2001b)

R. Haubner and C. Decristoforo

showed an initially increased activity concentration in blood, very similar kinetics in kidneys, a clearly reduced activity concentration in liver and an increased activity uptake and retention in the tumor compared to the first generation, unmodified peptides. In another approach hydrophilic d-amino acids were introduced to improve pharmacokinetics. Therefore, peptides containing three d-serine or d-aspartic acids and a g-amino butyric acid for 18F-labeling via prosthetic groups were coupled with the corresponding cyclic RGD-peptide. d-amino acids were used to improve the metabolic stability of the compounds. The peptides showed high avb3 selectivity in vitro and receptor specific accumulation in vivo. The tumor uptake in a murine melanoma model was lower as found with the glycosylated RGD-peptides. However, due to the rapid predominantly renal elimination of [18F]dAsp3-RGD tumor/ background ratios calculated from small animal PET images were comparable with [18F]Galacto-RGD. PEGylation is known to improve many properties of peptides and proteins including plasma stability, immunogenicity and pharmacokinetics (Harris and Chess 2003). In many cases it is used to prolong median circulation times and half lives of proteins and polypeptides by shifting the elimination pathway from renal to hepatobiliary excretion. Since renal filtration is dependant on both: molecular mass and volume occupied, this effect strongly depends on the molecular weight of the PEG moiety. For example, Chen et al. (Chen et al. 2004b) attached a 2-kDa PEG moiety to the e-amino function of cyclo(-Arg-Gly-Asp-dTyr-Lys-) and compared the 125 I-labeled PEGylated derivative (125I-RGD-PEG) with the radioiodinated cyclo(-Arg-Gly-Asp-dTyrLys-) (125I-RGD). 125I-RGD-PEG showed a more rapid blood clearance, a decreased activity concentration in the kidneys and slightly increased activity retention in the tumor. However, tumor uptake for 125I-RGD was higher as found for 125I-RGD-PEG for all time points. Moreover, increased activity retention in liver and intestine was found. In another study [18F]FB-RGD, a [18F]fluorobenzoyl labeled RGD peptide, and the PEGylated analog [18F]FB-PEG-RGD (PEG, MW = 3.4 kDa) have been compared (Chen et al. 2004a). Again activity retention of the PEGylated peptide in the tumor was improved compared with the lead structure. However, initial elimination from blood was slower and activity concentration in liver and kidneys was higher as for [18F]FB-RGD.

253

19 Radiotracer II: Peptide-Based Radiopharmaceuticals

Altogether, these studies revealed very different effects of PEGylation on the pharmacokinetics and tumor uptake of RGD-peptides which seems to strongly depend on the nature of the lead structure, indicating that the effects are not completely predictable. Anyway, in general pharmacokinetic modifiers are helpful tools to adjust lipophilicity and charge towards the desired pharmacokinetic properties. In some cases, however, radiolabeling strategies themselves can serve to modulate pharmacokinetics. The best example for this is the development of radiolabeled somatistatin analogs. Initially 123I-labeled Tyr3-octreotide was used to image neuroendocrine tumors in patients. This compound, although only showing very limited modification from the lead structure octreotide showed considerably hepatobiliar excretion. By introducing DTPA as chelator for 111In labeling pharmacokinetics were switched to a predominant renal excretion pathway, which was a major breakthrough for imaging applications (Bakker et al. 1991). Also other chelator systems have shown to act as pharmacokinetic modifiers for octreotide derivatives. Especially HYNIC-derivatisation for radiolabeling with 99mTc allows using coligands, which can serve as pharmacokinetic modifiers (Decristoforo et al. 2006). Also hydrophilic chelating systems like DOTA for labeling with e.g., 90Y, 177Lu, and 111In have shown to serve this purpose.

19.3 Labeling Strategies Labeling strategies for peptides can be divided into two major groups. One includes halogens like 18Ffluorine and iodine radioisotops. The other includes a variety of different radiometals like 99mTc-technetium, 111 In-indium, 68Ga-gallium, 64Cu-cupper, 90Y-yttrium and 177Lu-lutetium (for some physical characteristics of the different nuclides see Table 19.1). Due to the different properties of halogens and metal isotopes, labeling of the first focuses on direct labeling (iodine) or labeling via prosthetic groups (18F and iodine) whereas labeling of the latter is carried out via chelating systems conjugated to the peptides. Here we want to summarize some standard labeling protocols for radioiodination, describe the most common prosthetic groups for 18F-labeling and discuss major aspects of labeling with radiometals.

19.3.1 Halogens 19.3.1.1 Radioiodine One of the easiest ways to label peptides is by direct labeling using radioiodine isotops. For this approach at least one tyrosine in the amino acid sequence is necessary. Labeling is carried out via electrophilic substitution at the activated position 3/5 of the aromatic ring of the tyrosine (Fig. 19.3). Oxidation of the commercially available radioiodide can be carried out by different strategies (overview in ref. (Coenen et al. 2006)). Standard procedures use chloramine-T or better chloramine-T on a polymeric support (Iodobeads®) or Iodogen® for oxidizing *I− to *I+. The advantage of Iodobeads® and Iodogen® is, that under standard conditions (labeling in puffer systems) these reagents are not soluble and can easily be separated from the solution containing the radiopharmaceutical after labeling. Of disadvantage, especially for chloramine-T, is that they can oxidize sensitive amino acids (e.g., methionine) in the peptide sequence, which may lead to changes in affinity. A gentler oxidation can be carried out via enzymatic systems (e.g., peroxidases) and hydrogen peroxide. For labeling of peptides which do not include tyrosine, are sensitive towards oxidation or if direct labeling has negative effects on binding affinity the Bolton-Hunter reagent, N-succinimidyl-3-(4-hydroxyphenyl) propionate, can be used. Labeling is a two step reaction. In the first step the Bolton-Hunter reagent is radioiodinated using e.g., Iodobeads®. In the second step the radioiodinated reagent is conjugated to an amine function of the peptide sequence (e.g., a-amino function of the N-terminal amino acid, or e-amino function of lysine). In many cases for the first evaluation of a new class of peptides 125I-iodine is used. The advantages are a comparable low radiation exposure due to the low energy and long half life allowing experiments over a long period. Due to the availability of a great variety of iodine isotopes, depending of the isotope used peptides can be utilized for SPECT (123I), PET (124I) and therapeutic approaches (131I). However, in any case it has to be taken into account that the iodine-carbon bond can easily be cleaved by corresponding enzymes. Thus, radioiodinated peptides may show low metabolic stability reducing the applicability of this class of tracer. In some cases proof-of-principle is carried out with iodinated derivatives followed by modifications allowing labeling with 18F or radiometals.

254

R. Haubner and C. Decristoforo

Table 19.1 Overview of the most frequently used isotopes for peptide labeling and some physical characteristics Half life [h]a

Decay mode

Max. b-energy [MeV]b

g-Energy [MeV]b

Production route

Tc

6.01

100% ITc

–

0.140

Generator

In

2.8 days

100% EC

–

0.171

Cyclotron

Nuclide SPECT 99m 111

d

0.245 I

13.2

100% EC

–

0.159

Cyclotron

I

59.4 days

100% EC

–

0.035

Reactor

1.83

97% b+

0.63

–

Cyclotron

89% b+

2.92

1.077

Generator

11% EC

0.82

17% b

1.67

1.345

Cyclotron

Cyclotron

123

125 e

PET F

18

3% EC Ga

68

Cu

64

1.13

12.7

+

44% EC

I

124

4.2 days

39% b−

0.58

23% b

1.53

0.603

77% EC

2.14

1.691

+

Therapy Y

64.1

100% b−

2.3

–

Generator

Lu

6.7 days

100% b

0.50

0.208

Reactor

0.38

0.113

90

177

−

0.18 I

131

8.0 days

100% b

−

0.61

0.636

0.33

0.364

Reactor

0.284 As indicated for some nuclides half life is given in days b If more only the most prominent decay energies are listed c IT: internal transfer d EC: electron capture e Most commonly 125I-iodine is used for biodistribution experiments but small animal imaging using 125I and SPECT is also possible a

19.3.1.2 18F-flourine Direct labeling of peptides using 18F-fluorine is in most cases not appropriate (harsh reaction conditions, formation of [18F]HF in the presence of acidic protons).Thus, most approaches use prosthetic group labeling for introducing 18F into peptides and proteins. Therefore, the nature of peptides, which include amino acids with multiple side chain functionalities like amino, thiol and alcohol

groups, is exploited. Thus, standard labeling strategies are based on 18F-fluoroalkylation, 18F-fluoroacylation and 18F-fluoroamidation (Okarvi 2001; Wester and Schottelius 2007) (see also Fig. 19.4). Most prosthetic groups for labeling of peptides focus on acylation of an amino function via an activated ester. One of the most prominent members is succinimidyl-4-[18F]fluorobenzoate ([18F]SFB) (Fig. 19.4). This prosthetic group can be produced via a three step

19 Radiotracer II: Peptide-Based Radiopharmaceuticals Fig. 19.3 (a) Radioiodine labeling of peptides via the phenol system of included tyrosine moieties. For electrophilic substitution the commercially available radioiodine has to be oxidized. (b) Structure of chloramine T, Iodobeads (polymere-bound chloramine T), and of Iodogen, which are the most commonly used oxidizing reagents for radioiodinations

255

a

b

a

b

c

Fig. 19.4 (a) Structure of succinimidyl-4-[18F]fluorobenzoate ([18F]SFB) and 4-nitrophenyl-2-[18F]fluoropropionate ([18F] NFP), which belong to the most commonly used prosthetic groups for 18F-labeling of peptides via fluoracylation. (b) Labeling of peptides via oxime formation using 4-[18F]fluorobenzaldehyde and amino-oxo-modified peptides. By using this strategy care has to be taken because of the high reactivity of the

amino-oxo-function. Any aldehyde or keto function has to be avoided. Otherwise quenching reactions can lead to the loss of the coupling group. (c) Schematic structure of a prosthetic group which is selective for thiol-containing peptides. Many of them combine an aromatic system for labeling via a linker system with the maleimido group which allows selective addition of HS-groups to the double bond

256

synthesis starting with the 18F-labeling of 4-N,N,Ntrimethylamonium ethyl benzoate followed by the hydrolysis of the ethyl ester and subsequent activation using O-(N-succinimidyl)-N,N,N¢,N¢-tetramethyluronium tetrafluoroborate (TSTU). Another aliphatic prosthetic group is 4-nitrophenyl-2-[18F]fluoropropionate (Fig. 19.4). Again synthesis includes 18F-labeling of the corresponding precursor, hydrolysis of the ester and subsequent activation. In many cases, at least, before conjugation with the peptide a HPLC-separation is carried out making these labeling strategies very complex and time consuming. Anyway, for most of the yet described 18F-labeled peptides acylation is used for conjugation of the prosthetic group. Recently, alternative strategies designed to optimize 18 F-labeling of peptides are discussed. This includes oxime and hydrazone formation, conjugation via “click chemistry” or by using thiol reactive groups. It has been demonstrated that amino-oxy functionalized peptides allows regioselective labeling using aldehydes and ketones. E.g., it has been shown, that multimeric RGD-peptides modified with amino-oxo acetic acid could be labeled successfully by using [18F]fluorobenzaldehyde (Poethko et al. 2004b). Based on similar chemistry HYNIC-modified peptides can be labeled with [18F]fluorobenzaldehyde. Click chemistry, the 1,3-dipolar cycloaddition of an azide and an alkyne at room temperature using a Cu catalyst, was first introduced in radiochemistry to produce 99mTc(CO)3-labeled compounds (Mindt et al. 2006). Meanwhile this approach is also discussed to be used for 18F-labeling. E.g., using the corresponding tosylate w-[18F]fluoroalkynes were produced by nucleophilic fluorination and conjugated via cycloaddition to peptides functionalized with 3-azidopropionic acid. Other routes use propagylglycine for introducing an alkyne function into the peptide and [18F]fluoroethylazide, again produced via the tosylate, as prosthetic group (Thonon et al. 2009). It has been found that glycosylation can improve pharmacokinetics of peptide-based tracer (for details see above). An elegant approach would combine introduction of the label and the sugar moiety in one step. Most elegant would be to use [18F]FDG, which is available in almost all PET centers, as the prosthetic group. Because fluorodeoxyglucose exists in small amounts in the “aldehyde form,” the chemoselective labeling strategy via oxime formation with amino-oxy functionalized peptides can be used. Most recently, by

R. Haubner and C. Decristoforo

using this approach, Wester and coworkers produced an 18F-labeled glycosylated RGD-peptide (Hultsch et al. 2009). Of importance for the production of larger activities is, that n.c.a. [18F]FDG is used, which means that after [18F]FDG production a separation of the radiolabeled sugar from excess glucose via HPLC is necessary. Other approaches modify [18F]FDG before conjugation with the corresponding peptide. This can be carried out by producing thiosulfonate derivatives of the fluorodeoxyglucose which can form disulfide bridges with thiol-containing peptides, by synthesizing a thiol-reactive prosthetic group via oxime formation of the corresponding maleimide derivative and [18F]FDG (Prante et al. 2007), or by using an “[18F] FDG-azide” which is conjugated to a propagylglycincontaining peptide via click chemistry (Maschauer and Prante 2009). However, all these strategies have to be evaluated more detailed before it will become clear which will be the most useful labeling procedure.

19.3.2 Metals Today a number of radiometals are available for radiolabeling including labeling of peptides. Many radiometal labeled peptides have proven to have a high stability in biological systems and residualizing properties in cells, once taken up. In contrast to radiohalogens, metals have to be introduced via formation of a strong complex with the peptide structure. While direct labeling approaches of peptides have been described, it is today recognized that a specific metal chelator has to be introduced to achieve suitable targeting properties of the radiolabeled peptide in vivo in the great majority of cases. This is achieved by the so called post labeling bifunctional chelate approach. Thereby a metal chelating moiety is introduced into the peptide by chemical synthesis, either using peptide chemistry approaches using specific, chemically protected chelating moieties or alternatively by attaching the chelator using e.g., activated ester functions after synthesis of the peptide. It is well recognized that an introduction of the chelating moiety will have considerably influence on specific interaction between the peptide sequence and the receptor, dependent on the individual targeting sequence. Information on the site of chelator attachment without impairing receptor binding ability can be

19 Radiotracer II: Peptide-Based Radiopharmaceuticals

derived from basic research on the peptides themselves and their respective receptors. Such derivatised peptides can be radiolabeled by incubating the radiometal with an excess of peptidechelate conjugate directly, typically achieving almost quantitative labeling yields. In contrast to radiohalogenation, the radiolabeled peptide is usually not separated from excess of the labeling precursor (here peptide-chelate). As peptides bind to so called low capacity targets, usually receptors, with high affinity, the total peptide amount has to be limited to avoid saturation of the target. Therefore, radiolabeling reactions have to be performed at very high specific activities, meaning with very low amounts of peptide conjugate and radiometal. This limits the choice of both the radiometal and the chelator. Additionally, the chelator has to ensure a high in vivo stability, especially avoiding the release of the radiometal and should not deteriorate pharmacokinetic behavior or ideally even act as pharmacokinetic modifier (see Sect. 19.2.2.3). Besides high specific activity the radiometal should have a high chemical purity, mainly being free from other metallic impurities and be easily available at reasonable costs. The most frequently used radiometals applied in peptide labeling for imaging purposes are, on the one hand, 99mTc and 111In for SPECT, on the other hand, 68 Ga and 64Cu for PET. 99mTc and 68Ga show advantages in terms of availability, as they are generator products providing high flexibility and low cost in use. Whereas 99m Tc requires the use of dedicated chelators and will be described separately in the following chapter, the other radiometals have some common properties in their coordination chemistry allowing the use of similar labeling strategies.

19.3.2.1 Technetium Different chelating systems have been employed in an attempt to develop 99mTc-labeled peptides for targeting peptide receptors. Early attempts for radiolabeling peptides using bifunctional chelators were based on propylenediamine-dioxime (PnAO) or so called N3S-ligands such as mercaptoacetyltriglycine (MAG3). The relatively high lipophilicity of these chelating systems resulted in relatively low tumor and a high liver uptake with predominant excretion via the hepatobiliary system

257

(Decristoforo and Mather 1999a, b). Only introduction of pharmacokinetic modifiers could improve properties to achieve suitable imaging properties, such as 99mTc-Depreotide based on a N3S monothiolbisamide-monoamine ligand and multiple lysine residues to increase hydrophilicity (Vallabhajosula et al. 1996). A very frequently used labeling moiety is 2-hydra zinonicotinic acid (HYNIC). It can be attached via the carboxylic function to a free amine group of the peptide either by using carboxylic group activation via succinimidyl-ester (succinimidyl-6-hydrazinopyrdine3-carboxylic acid) or by using e.g., the hydrazine-BOC protected building block (6-BOC-hydrazinopyridine3-carboxylic acid) during peptide synthesis (Fig. 19.5). In both cases the required HYNIC-derivative is easily accessible via chemical synthesis or is commercially available. In the radiolabeling process with HYNIC derivatized peptides a coligand is required that has shown to be of major importance for in vivo properties (Decristoforo and Mather 1999a, b). Coligands that allow radiolabeling at every high specific activities are usually aminocarboxylates such as tricine and ethylendiamin-N,N`diacetic acid (EDDA) or even ternary complexes with tricine and a monodentate ligand such a phoshines or N-heterocycles, imidazoles or pyridines. Ternary complexes and EDDA have shown a higher stability especially resulting in a reduced plasma protein binding. Radiolabeling is achieved by reacting the HYNIC-peptide with 99mTc-pertechnetate in the presence of stannous chloride, the choice of co-ligand may also be a matter of peptide stability as some coligands require heating. Tetraamine (N4) based chelators are also suitable to prepare 99mTc-labeled peptides. Such a tetraamine chelator can be labeled at high specific activities under relatively mild alkaline conditions at room temperature and is hydrophilic, due to the formation of a positively charged Tc-dioxo unit. This 99mTc-labeling approach has shown to result in very suitable in vivo properties when attached e.g., to octreotide derivatives (Maina et al. 2002). The N4 chelator has to be introduced as protected building block during peptide synthesis, which is not commercially available. A very versatile 99mTc-labeling approach is based on the so called carbonyl-core (Fig. 19.5). Thereby 99m Tc-pertechnetate is first reacted to the Tc-tricarbonyltriaqua-ion ([99mTc(CO)3(H2O)3]+), where positively

258 Fig. 19.5 (a) Structure of HYNIC-Boc, which can be directly used for conjugation of the carboxylate with amino functions in the peptide sequence. For complexation of 99mTc the Boc-protection group has to be removed and coligands have to be added. Here EDDA was used as coligand. (b) Shows an example of a peptide labeled using the Tc(CO)3-core. As tridentate ligand a thiol modified cysteine is used. (c) The “chlick to chelate” approach uses the cupper(I) catalyzed 3+1 cyclo addition reaction to produce the triazole-based tridentate chelating system and subsequently the 99m Tc-complex in an one pot reaction

R. Haubner and C. Decristoforo

a b

c

charged 99mTc in oxidation state (I) is bound very stable to three CO and weakly to three water molecules. This can be achieved by a commercially available kit (Isolink Covidien, NL) in a straight forward procedure. [99mTc(CO)3(H2O)3]+ is then reacted with a chelating system replacing the weakly bound water molecules by more stable bound chelator. It has been shown that some amino acids form very stable complexes with the 99m Tc-(CO)3 core with very high specific activities (Egli et al. 1999). Especially using histidine, but also cysteine provides a tridentate ligand system. However, if these amino acids are incorporated into a peptide chain, one coordination site is lost and then these amino acids only act as bidentate chelators, resulting in a much lower in vivo stability. While complexes with tridentate coordinated ligand systems generally showed a fast and complete clearance from all organs and tissues, those complexes with only bidentate coordinated ligands showed a significantly higher retention of activity in the liver, the kidneys, and blood pool (Schibli et al. 2000).The likely explanation for this behavior is that the co-ordination site which is unoccupied by these ligands become filled by donor atoms from circulating plasma proteins resulting in a high degree of protein binding and a radically altered pattern of biodistribution. Therefore, specific building blocks for introduction into peptide chains for 99mTc-tricarbonyl labeling have

been described, based on inverse histidine derivatives or by introducing chelators using click chemistry approaches resulting in triazole derivatives providing corresponding coordination sites for the 99mTc-tricarbonyl core (Mindt et al. 2006) (see also Fig. 19.5). To date, however, none of these chelator systems for 99m Tc-tricarbonyl cores are commercially available and have to be synthesized individually. Whether the tricarbonyl approach offers advantages in terms of biological properties as e.g., compared to HYNICderivatization remains to be shown.

19.3.2.2 Other Radiometals Labeling of other metals (e.g., 64Cu, 68Ga, 111In) can be carried out by a class of very similar chelating systems. Basically they can be distinguished between open chain analogs and macrocyclic ligands. The main open chain ligands are based on diethylenetriamine-tetraacetic acid (DTPA) (Fig. 19.6), an octadentate ligand for many metals. DTPAanhydride can directly be coupled to free amino groups in the peptide chain. This, however, is not site specific and results in loss of one carboxylic group required for high stability coordination of many radiometals. To overcome this limitation DTPA-derivatives have been described, e.g., based

19 Radiotracer II: Peptide-Based Radiopharmaceuticals

on S-2(4-aminobenzyl)-diethylenetriamine pentaacetic acid. They are commercially available and can be used to introduce DTPA as a peptide building block or via active esters. It has to be considered, that especially introduction of aromatic linking moieties will increase lipophilicity and may hamper pharmacokinetic properties. A major advantage of DTPA derivatization of peptides is that it allows radiolabeling at room temperatures, thereby avoiding destruction of chemically sensitive moieties in the peptide. However, especially stability in vivo remains a challenge, especially in case of 68Ga and 64Cu derivatives. In case of 68/67Ga also desferoxamine (DFO) as linear chelator have been used for radiolabeling through the included three hydroxamate groups. So far this approach was mainly used for proteins such as antibodies. However, DFO-Octreotide has been attempted to be used for tumor imaging labeled with 67Ga (SmithJones et al. 1994), but was achieving only poor clinical results. The most frequently used chelating system for radiometal labeling of peptides is the cyclic chelator 1,4,7,10-tetraazacyclododecane-N,N¢,N²,N²¢-tetraacetic acid (DOTA) (Fig. 19.6). It contains 4 amino groups and

259

4 carboxylic groups for coordination of metals, the carboxylic group can be utilized for attachment to amino groups of peptides. This can be achieved either by specifically protected building blocks during the peptide synthesis or again via activated esters after the peptide has been prepared. Examples of commercially available building blocks are shown in Fig. 19.6. Radiolabeling of DOTA derivatized peptides is achieved under acidic conditions and requires elevated temperatures (Breeman et al. 2003). Temperature and pH vary dependent on the radionuclide being used. The ideal pH is a compromise between the tendency of the radiometal to form colloids at elevated pH and the reduced speed of complex formation at lower pH. Whereas for 111 In-labeling optimal pH is in the range between 4.5 and 5 usually achieved with acetate or ascorbate buffers, lower pH values are required for 68Ga-labeling whereby HEPES or acetate buffers are frequently used to provide high radiolabelling yields (Velikyan et al. 2004). The prototype of a DOTA-derivatized peptide is DOTA-Tyr3Octreotide (DOTATOC) and related derivatives that have been labeled with 111In and 68Ga for diagnostics but also with 90Y and 177Lu for therapeutic applications. However, coordination chemistry of DOTA may vary dependent on the radiometal. Whereas 111In result in a octacoordinated

a

b

c

Fig. 19.6 (a) Structure of DTPA as an example of a linear chelating system. It can be used especially for 111In-labeling. (b) Macrocyclic chelating systems: DOTA is used for 111In, 90Y, 177 Lu and 68Ga (left). NODAGA is a nine-membered ring which fits very well for 68Ga and shows for this isotope a drastically increased complex binding affinity compared with DOTA

(middle). TE2A is a bridged chelating system especially designed for complexing 64Cu (right). (c) Special building blocks to introduce DOTA as chelator into peptides: left: Tris-BOC protected for conjugation in peptide synthesis with final deprotection, right: activated N-succidimidyl-ester that can be reacted with free amine groups of unprotected peptides

260

complex, 68Ga is hexacoordinated resulting in a free carboxylic group. This may result in different biological behavior of the radiolabeled peptides. In the case of DOTATOC 68Ga labeling resulted in higher tumor uptake and more rapid clearance (Froidevaux et al. 2002). In contrast a DOTA-RGD derivative showed higher blood levels and impaired tumor/background ratios when labeled with 68Ga compared to the corresponding 111 In-labeled compound (Decristoforo et al. 2008). Especially Ga3+ is a relatively small cation and therefore cyclic chelators with smaller ring size may be more suitable for derivatisation of peptides for radiolabelling. 1,4,7- triazacyclononane-1,4,7-triacetic acid (NOTA) is a more suitable ligand for radiolabelling with 68/67Ga. In contrast to DOTA, all three carboxylic functions are required resulting in stable hexacoordinated Gallium-complexes. Therefore, specific derivatives of NOTA are required providing a coupling site with the peptide chain, an example hereof is NODAGA (1,4,7-triazacyclononane-1-glutaric acid-4,7-diacetic acid) (Eisenwiener et al. 2002) (Fig. 19.6), overall aliphatic substitutions should be preferred over aromatic linker due to comparably reduced lipophilicity values. An increasing number of NOTA building blocks are becoming commercially available allowing straight forward derivatisation of peptides. A major advantage of NOTA derivatives is the possibility of radiolabelling at room temperature, therefore being suitable for sensitive peptide structures. DOTA has been used as chelator for 64Cu-labeling of RGD peptides (Chen et al. 2004c), Tumor/blood and tumor/muscle ratios of approximately 7 and 8, respectively, allowed acquisition of clear tumor/background contrast images 1 h p.i. using a small animal scanner. However, highest activity concentration was found in liver, intestine and bladder indicating that further optimization of the tracer is needed. Therefore alternative chelating systems have been described including 1,4,8,11-tetraazacycotetradecane-1,4,8,11-tetraacetic acid (TETA) and 4-(1,4,8,11-tetraazacyclotetradec-1-yl) methyl)benzoic acid (CPTA). Recently, cross-bridged tetraamine chelators such as 4,11-bis(carboxymethyl)-1,4,8,11-tetraazabicyclo[6.6.2]hexadecane (TE2A) (Fig. 19.6) have been proposed that show improved in vivo stability of radiolabeled peptides compared with their DOTA analogs (Garrison et al. 2007; Hausner et al. 2009). Further studies are required to judge the superiority of one of these derivatives in terms of targeting properties of 64Cu labeled peptides.

R. Haubner and C. Decristoforo

Overall if peptides are to be labeled with different radiometals for SPECT and PET applications DOTA may be the chelator of initial choice allowing radiolabeling with 111In, 68/67Ga and 64Cu, and additionally with 177 Lu and 90Y if therapeutic applications are to be considered.

19.4 Tracer Evaluation For the evaluation of the radiolabeled peptides adequate in vitro and in vivo studies have to be carried out. A comprehensive evaluation has to include studies determining affinity, selectivity, stability and pharmacokinetic properties of the tracer. The most relevant assays to generate these data are discussed below. Besides the described studies noninvasive imaging data resulting from studies using dedicated small animal scanner belong to the main criteria to determine the imaging properties of a new tracer. However, discussion of corresponding scanner, imaging techniques and protocols are not included and described in detail in other chapters.

19.4.1 In Vitro Evaluation 19.4.1.1 Target (Receptor) Binding Assays Binding affinity can be determined using a variety of radioligand binding assays, whereby description of all details is out of scope of this chapter. Theoretical and practical descriptions can be found in standard textbooks. To determine the receptor binding, first a source for the receptors has to be found. This can either be: • A cell line naturally expressing the receptor • A cell line (stably) transfected with the receptor, such as CHO (Chinese hamster ovarian cancer cells) • Fresh tissue or frozen tissue homogenates from animals • Tissue sections from human or animal tissue (normal organs, tumors) • Isolated receptors, ideally commercially available Dependent on the source of the receptor different experimental techniques have to be applied. However receptor binding assays always will include:

261

19 Radiotracer II: Peptide-Based Radiopharmaceuticals

• A radioligand: This can either be the radiolabeled peptide itself or a known radiolabeled ligand for the respective receptor (e.g., commercially availably as 125 I or 14C-labeled compound). A major requirement is that the concentration of the radioligand has to be very low to avoid saturation of the receptor system. A rule of thumb is that the radioligand should be used in a concentration below the (expected or known) Kd value of the radioligand to the receptor. Therefore, radioligands of high specific activity are required still to provide sufficient count statistics in the assay. • A suitable buffer system: This depends on the receptor system studied and specific additives such as metal ions, protease inhibitors or additives to reduce non specific binding (e.g., albumin) may have to be added. • A competitor of the radioligand: This is a non radioactive compound known to bind to the receptor and may serve in the simpliest case to determine non specific binding, in this case, typically 1–10 µM of the competitor is used. In more complicated assays different dilutions of the competitor are used. • A washing buffer: Typically this is refrigerated before use to remove non specific binding but reduces the probability of loss of specific binding. • A separation mechanism between bound and free radioligand: This may be in the simpliest system pipetting of a supernatant, e.g., when tissue sections, coated receptors plates or sticking cells are used. It may be achieved by centrifugation or more often by filtration, typically using glass fiber filters. Some different assays can be designed that also require different values of complexity: 1. Simple binding assays, where a radioligand is incubated with receptors with and without competitor. This assay will provide answer whether a compound specifically binds to a receptor only in qualitative terms. 2. Competition assays: Here a radioligand is incubated with increasing concentration of a non radioactive ligand. The results can be fitted in dedicated computer programs (such as ORIGIN or SIGMA-Plot) and are expressed as IC50 values in a concentration unit (typically nM). The IC50 value is a parameter for the binding affinity of the competitor (not the radioligand!!). 3. Saturation assays: Here increasing concentrations of the radioligand are used and results again fitted

in a computer program. This assay results in Kd values, giving the binding affinity of the radioligand. 4. Internalization assays: Here the radioligand is incubated with intact cells and after certain time points cells are washed from unbound radioligand, followed by a washing step with an acidic buffer (typically acetate or glycine buffers) that removes cell surface receptor bound activity and at the end cells are collected and counted. In this assays information is gained on active cell uptake via receptor internalization and results are expressed as internalized activity related to the total activity added or related to surface bound activity. Non specific cell uptake can be determined either by co-incubation with excess of cold receptor ligand or at 4°C inhibiting energy dependent internalization. 5. Washout assays: This assay may be performed if information is required whether an internalizing radioligand is released after intracellular metabolism or via active excretion mechanisms. To achieve this first a radioligand is incubated with the respective cells, then extracellular radioactivity is washed of. The cells are incubated again and samples of the supernatant are collected over time. The result may be expressed as % washout of total activity in the cell. It must be stressed that for assays giving information on internalization and externalization intact cells have to be used, whereas for determination of binding affinities membranes or isolated receptor assays may be preferred as equilibrium required for exact determination of e.g., a Kd value cannot be reached when the system is “open,” e.g., by allowing internalization of ligands. In any case the non specific binding has to be determined, usually by co-incubation with an excess of a ligand blocking the target, without this no information on target binding whatsoever can be gained. Special assays may also require additional techniques, such as when tissue sections are used. In this case determination of radioligand binding must be performed by autoradiography either electronically or by means of classical film autoradiography.

19.4.1.2 Partition Coefficient One parameter, that is easy to determine and allows some prediction of the pharmacokinetics in vivo, is the partition coefficient (log P). It is a major

262

parameter for the hydrophilicity and describes the distribution of the radiolabeled peptide between an octanol and a buffer phase. Especially for the partition of peptides a defined pH-value is of great importance (the protonation stage has influence on the hydrophilicity). Mostly phosphate buffered saline pH 7.4 (PBS) is used as buffer system. The more hydrophilic the radiolabeled peptide is the higher is the activity found in the buffer phase. It is difficult to define an optimal range for this parameter as it strongly depends on the particular nature and application of the radiolabeled peptide. In most cases radiolabeled peptides are developed to target receptors which are upregulated in tumor tissue. Thus, peptides that are rapidly accumulated in the tumor and eliminated from the body are searched for. It has been shown, that peptides with very low logP values (typically below −3) are preferable in this case (see also tracer optimization). Anyway, fast elimination from blood is accompanied by short time for tumor uptake. Thus, a balance between elimination and accumulation has to be found, which is best reflected in the tumor-to-background ratio (see later). For the determination of the log P value it is suggested to add aprox. 500,000 cpm of the radiolabeled peptide in <50 µL aqueous solution to an Eppendorf cap containing 500 µL octanol and 500 µL PBS. After rigorous shaking for 10 min samples of 100 mL from each phase are collected and counted. As peptides usually are very hydrophilic, most of the activity is found in the aqueous phase and great care has to be taken during sample drawing to avoid cross contamination of the organic layer.

19.4.1.3 Protein Binding A second important predictor of pharmacokinetics is the amount of radioligand bound to plasma proteins. High values of plasma protein binding will lead to prolonged blood circulation and reduced target to nontarget ratios. Plasma protein binding increases with increased lipophilicy, which is a known phenomenon for drugs. In the case of radiometal labeled peptides protein binding may also reflect reduced complex stability resulting in increased binding of the radiometal to proteins. The protein bound fraction is determined by incubation of the radiolabeled peptide in low concentration

R. Haubner and C. Decristoforo

(<1 µM) at 37°C in human or, if not available, rat plasma or serum for varying time points. Care has to be taken, that the radioligand solution does not contain organic solvents or may contain buffers influencing the pH. The protein bound fraction is determined by separation of the radiolabeled peptide from plasma proteins. This may be achieved by size exclusion, precipitation or ultra-filtration. The latter two may overestimate the protein bound fraction due to unspecific effects. A straight forward approach is to use miniaturized Sephadex based size exclusion columns (Microspin). These can be loaded with the sample and centrifuged under controlled conditions thereby only releasing the protein bound fraction. Eluate and column are counted in a gamma counter and the protein bound fraction is calculated as % protein bound of total activity. The assay should always be performed using a negative control (i.e., incubation in buffer). Increase of protein binding over time may be an indicator of a limited complex stability (release of radiolabel over time). It should be stated that the method only is applicable if the molecular weight of the peptide is comparably low (i.e., <5,000). Low protein binding is usually of advantage, absolute values are difficult to give, however e.g., for RGD peptides low protein binding was considered to be clearly below 10%.

19.4.1.4 Stability Assays Stability assays should be performed for different reasons. First they can give an indication of the stability of the radiolabel, especially when radiometal-based labels are used. On the other hand, they can provide information on the stability of the peptide itself. In general these assays are performed by incubation of the radiolabeled peptide in a certain medium, followed by analysis of the incubation solution at different time points, usually by radio-HPLC analysis, in certain cases TLC or SPE may be applied, however will only provide limited information (especially no information of the stability of the peptide is included). Assays to determine the stability of the radiolabel, specifically when radiometals are used should include incubation in different media. First stability in solution and in buffers at neutral pH (e.g., PBS) should be included. Additionally, so called challenging assays should be performed. In this case an excess of a ligand

263

19 Radiotracer II: Peptide-Based Radiopharmaceuticals

with good chelating properties for the radiometal should be co-incubated. This can be DTPA in the case of 111In, 67/68Ga or 64Cu. In the case of 99mTc the challenging agent should be adapted to the radiolabelling approach used. In the case of 99mTc(V) ligands (e.g., N3S and N4 ligands) cysteine can be used as naturally occurring competing ligands. In case of 99mTctricarbonyl compounds histidine may be a better choice. Such assays can be designed in different ways. One possibility is co-incubation with a great excess of the competing ligand (e.g., 1,000-fold molar excess of ligand over radiolabeled peptide) and determination of the stability at different time points (up to 24 h). Another possibility is to use one time point (e.g., 4 h) and different molar ratios of competing ligand and peptide. The amount of released radionuclide is usually determined by means of radio-HPLC, but may be simplified by using e.g., TLC methods. However, it should be validated whether the separation of radiolabeled peptide and released radiometal can be determined in this way. Stability of the peptide itself is usually determined by incubation of the radiolabeled peptide in serum at 37°C, followed by radio-HPLC analysis at different time points, expressing the result as half lives of the radiolabeled peptide. To allow this analysis, first serum proteins have to be separated by means of precipitation (e.g., by pipetting the same volume ethanol or acetonitrile to the serum sample, followed by centrifugation). An alternative approach uses fixation of the peptide on a C-18 cartridge. In both cases the protein bound fraction should be determined in parallel by one of the methods described above to avoid artifacts due to protein bound material. Additionally to incubation in serum, other incubation media may be chosen such as tissue homogenates (e.g., from rat), as it is known that among other organs kidneys and liver are responsible for peptide degradation. It also may be of interest to determine stability in tumor cells expressing the respective target in similar way. With such assays information on the metabolic stability can be gained. However, interpretation should be made very carefully, as stability can be very variable dependent on assay conditions (volume, age of serum, concentration). It therefore should be considered as an intra-laboratory standard rather than a comparison with published data or results from other laboratories.

19.4.2 In Vivo Evaluation 19.4.2.1 Normal Distribution/Metabolic Stability First in vivo studies with a new tracer can be carried out using normal mice. This study will supply information about the overall biodistribution of the radiolabeled peptide. E.g., it will show if tracer elimination is predominantly hepatibiliary or renal, if blood-pool clearance is fast or slow, and if blood–brain-barrier is penetrated. For this study the tracer is injected into the tail vein of the mouse. At different time points (depending on the isotope used) the animals are sacrificed, dissected and tissue of interest is weighed and activity is measured using a gamma-counter. Activity concentration in the different organs will be expressed as percent injected dose per organ (%ID) or/and per gram tissue (%ID/g). Further important data in tracer evaluation results from studies of the metabolic stability. A first indicator of tracer performance results already from determination of this parameter in serum/plasma. However, studying of the metabolic stability in whole body may supply a deeper insight in tracer properties. A standard protocol may include investigation of the metabolic stability of the tracer in blood, tumor, kidney and liver. Therefore the tracer is injected i.v. and blood and organs are sampled at corresponding time points. Subsequently organs are homogenized (e.g., organs were frozen with liquid nitrogen and homogenized using a Mikro Dismembranator), suspended in buffer and organ homogenates and blood are centrifuged. Workup of the supernatant follows the same protocols as described above (in vitro stability assays). 19.4.2.2 Murine Tumor Models Most approaches using radiolabeled peptides are focused on targeting receptors involved in tumor biology. Most important in the development of these tracers is to choose the appropriate tumor model. One very important factor is, that it is proven that the target molecule is expressed on the used tumor cell line as well as, and even more important, on the tumor grown in the corresponding animal. The latter can be carried out by dissecting the tumor and immunohistochemical staining of the tumor tissue sections using corresponding

264

antibodies. If this is not possible at least Western blotting, determining that the mRNA is produced should be carried out. In cases where a set of subtypes belongs to a receptor family (e.g., somatostatin receptors, integrins) and the development is focused on a subset or even on one special subtype out of the receptor family it has to be guaranteed that only the target receptor is expressed. This can be achieved by using genetically modified cell lines where the target receptor is inserted into the host cell genome or by selecting cell lines which have already the desired expression pattern e.g., by FACS. Standard procedures to demonstrate selective receptor mediated tracer accumulation are using blocking experiments (see below). However, these include in general that high peptide amounts have to be injected which may induce some down stream signals or/and influence the pharmacokinetics of the tracer. Thus, preferable assays will include negative control cell lines which are selected or modified not to express the target receptor. Such approaches will allow to grow receptor positive as well as receptor negative tumors within the same animal eliminating effects may be occurring due to injection of “cold” peptide or due to the fact that experiments are carried out in different sets of animals. In most cases for the first in vivo evaluation the animal models of choice are based on mice or in some cases rats, which are injected subcutaneously with the corresponding tumor cell line. One advantage is that adequate animal housing and handling is relatively easy. Moreover, a variety of immunodefficient nude mice strains, required for the use of human cell lines, are available. In some cases transgenic mice may be included in the evaluation. Transgenic mice are very useful systems for studying mammalian gene function and regulation. Moreover, transgenic mice can be used to model human diseases that involve the over- or missexpression of a particular protein. E.g., the RIP-TAG model is a transgenic mouse model of carcinogenesis and angiogenesis were the oncogene SV40 T antigen is expressed under the control of the rat insulin promotor. The advantage of such a model is that the tumor arises from normal cells in its natural tissue environment and progresses through multiple stages as human cancer do. In the RIP-TAG model within 12–14 weeks insulin producing beta-cells develop to islet cell hyperplasia, adenomas and finally carcinomas. Anyway,

R. Haubner and C. Decristoforo

production and handling of transgenic mice is not trivial and is not a prerequisite in the evaluation of new radiolabeled peptide, but can add additional information to get a deeper insight into the molecular mechanism of tracer accumulation.

19.4.2.3 Tumor Accumulation/Blocking Experiments Information on the distribution of the tracer in normal tissue and the metabolic stability in vivo are important factors. But finally adequate and specific accumulation of the tracer at the target structure is the most important prerequisite of a radiopharmaceutical. Thus, a major step in radiotracer development is the evaluation of the radiolabeled peptide using corresponding diseases models. As already mentioned most radiolabeled peptides yet are developed to target receptors which are overexpressed in tumor tissue. Thus, tumor xenografts grown in mice (or nude mice; depending on the origin of the tumor cells) are used in most cases. These were generated by either injection of tumor cells subcutaneously into the mouse or by directly transplanting tumor tissue form one animal to the next (which will not be discussed in detail here). For biodistribution studies injection site is not important. Preferable the tumor cells (approximately 5 Mio, but it strongly depends on the in vivo growth rate of the cell line) are injected subcutaneously at the musculus femoris. Mice are ready for biodistribution studies when the tumors reached a weight of approximately 100–200 mg. There is a great variation in the growth period, which strongly depends on the tumor cell type used. Many cell lines can be injected in buffer systems (e.g., PBS). For some cell lines injection in Matrigel, is recommended. Therefore, it is mixed with immortalized human cancer cells and the mixture is injected subcutaneously in immunodeficient mice. Using this procedure a human tumor usually forms in 2–4 weeks. Major components of Matrigel are structural proteins such as laminin and collagen. Additionally, it contains growth factors that promote differentiation and proliferation of many cell types as well as numerous proteins in small amounts. Anyway, its exact composition is unknown. After the tumor reached desired size distribution of the tracer will be studied following the same protocols as described above. An important parameter to draw conclusions regarding tracer performance is the

19 Radiotracer II: Peptide-Based Radiopharmaceuticals

tumor-to-background ratio, which can be extracted from this experiment. Not only the absolute amount of tracer accumulation in tumor tissue is important but also it is even more important to show that accumulation is tumor or better receptor specific. As mentioned this can be confirmed by either using negative control tumors or, if not available, by carrying out blocking or competition experiments. Therefore, prior (approximately 10 min) or together with the radiolabeled peptide a great excess of unlabeled peptide, which is known to bind with high affinity to the target structure, is injected. This has to lead to a significant reduction in tumor uptake. Sometimes not only tracer accumulation in the tumor but also in other organs decreases. This can be due to expression of the target structure on organs other than the tumor. Anyway, this has to be considered carefully, because the high amount of peptide can influence elimination, which also may result in changes of the distribution pattern.

References Bakker WH, Albert R, Bruns C et al (1991) [111In-DTPAdPhe1]-octreotide, a potential radiopharmaceutical for imaging of somatostatin receptor-positive tumors: synthesis, radiolabeling and in vitro validation. Life Sci 49: 1583–1591 Breeman WA, De Jong M, Visser TJ et al (2003) Optimising conditions for radiolabelling of DOTA-peptides with 90Y, 111 In and 177Lu at high specific activities. Eur J Nucl Med Mol Imaging 30:917–920 Brissette R, Goldstein NI (2007) The use of phage display peptide libraries for basic and translational research. Methods Mol Biol 383:203–213 Chen X, Park R, Hou Y et al (2004a) Micropet imaging of brain tumor angiogenesis with 18F-labeled pegylated rgd peptide. Eur J Nucl Med Mol Imaging 31:1081–1089 Chen X, Park R, Shahinian AH et al (2004b) Pharmacokinetics and tumor retention of 125I-labeled rgd peptide are improved by PEGylation. Nucl Med Biol 31:11–19 Chen X, Park R, Tohme M et al (2004c) Micropet and autoradiographic imaging of breast cancer alpha v-integrin expression using 18F- and 64Cu-labeled RGD peptide. Bioconjug Chem 15:41–49 Chighine A, Sechi G, Bradley M (2007) Tools for efficient highthroughput synthesis. Drug Discov Today 12:459–464 Coenen HH, Mertens J, Maziere B (2006) Radioiodination reactions for pharmaceuticals – compendium for effective synthesis strategies. Springer, Dordrecht Decristoforo C, Mather SJ (1999a) 99m-Technetium-labelled peptide-HYNIC conjugates: effects of lipophilicity and stability on biodistribution. Nucl Med Biol 26:389–396

265 Decristoforo C, Mather SJ (1999b) Technetium-99m somatostatin analogues: effect of labelling methods and peptide sequence. Eur J Nucl Med 26:869–876 Decristoforo C, Faintuch-Linkowski B, Rey A et al (2006) [99mTc]HYNIC-RGD for imaging integrin alpha(v)beta3 expression. Nucl Med Biol 33:945–952 Decristoforo C, Hernandez Gonzalez I, Carlsen J et al (2008) 68 Ga- and 111In-labelled DOTA-RGD peptides for imaging of alpha(v)beta3 integrin expression. Eur J Nucl Med Mol Imaging 35:1507–1515 Egli A, Alberto R, Tannahill L et al (1999) Organometallic 99m Tc-aquaion labels peptide to an unprecedented high specific activity. J Nucl Med 40:1913–1917 Eisenwiener KP, Prata MI, Buschmann I et al (2002) NODAGATOC, a new chelator-coupled somatostatin analogue labeled with [67/68Ga] and [111In] for SPECT, PET, and targeted therapeutic applications of somatostatin receptor (hsst2) expressing tumors. Bioconjug Chem 13:530–541 Feher M, Schmidt JM (2003) Property distributions: Differences between drugs, natural products, and molecules from combinatorial chemistry. J Chem Inf Comput Sci 43:218–227 Froidevaux S, Eberle AN, Christe M et al (2002) Neuroendocrine tumor targeting: study of novel gallium-labeled somatostatin radiopeptides in a rat pancreatic tumor model. Int J Cancer 98:930–937 Garcia-Garayoa E, Blauenstein P, Blanc A et al (2009) A stable neurotensin-based radiopharmaceutical for targeted imaging and therapy of neurotensin receptor-positive tumours. Eur J Nucl Med Mol Imaging 36:37–47 Garrison JC, Rold TL, Sieckman GL et al (2007) In vivo evaluation and small-animal pet/ct of a prostate cancer mouse model using 64Cu bombesin analogs: Side-by-side comparison of the CB-TE2A and DOTA chelation systems. J Nucl Med 48:1327–1337 Handl HL, Vagner J, Han H et al (2004) Hitting multiple targets with multimeric ligands. Expert Opin Ther Targets 8:565–586 Harris JM, Chess RB (2003) Effect of PEGylation on pharmaceuticals. Nat Rev Drug Discov 2:214–221 Haubner R, Decristoforo C (2009) Radiolabelled RGD peptides and peptidomimetics for tumour targeting. Front Biosci 14:872–886 Haubner R, Wester HJ, Burkhart F et al (2001a) Glycosylated RGD-containing peptides: tracer for tumor targeting and angiogenesis imaging with improved biokinetics. J Nucl Med 42:326–336 Haubner R, Wester HJ, Weber WA et al (2001b) Noninvasive imaging of alpha(v)beta3 integrin expression using 18Flabeled RGD-containing glycopeptide and positron emission tomography. Cancer Res 61:1781–1785 Hausner SH, Kukis DL, Gagnon MK et al (2009) Evaluation of [64Cu]Cu -DOTA and [64Cu]Cu-CB-TE2A chelates for targeted positron emission tomography with an alpha(v) beta(6)-specific peptide. Mol Imaging 8:111–121 Heppeler A, Froidevaux S, Eberle AN et al (2000) Receptor targeting for tumor localisation and therapy with radiopeptides. Curr Med Chem 7:971–994 Hultsch C, Schottelius M, Auernheimer J et al (2009) 18 F-fluoroglucosylation of peptides, exemplified on cyclo(RGDfK). Eur J Nucl Med Mol Imaging 36(9): 1469–1474

266 Janssen ML, Oyen WJ, Dijkgraaf I et al (2002) Tumor targeting with radiolabeled alpha(v)beta3 integrin binding peptides in a nude mouse model. Cancer Res 62:6146–6151 Leach AR, Harren J (2007) Structure-based drug discovery. Springer, Berlin Maina T, Nock B, Nikolopoulou A et al (2002) [99mTc]Demotate, a new 99mTc-based [Tyr3]octreotate analogue for the detection of somatostatin receptor-positive tumours: synthesis and preclinical results. Eur J Nucl Med Mol Imaging 29:742–753 Maschauer S, Prante O (2009) A series of 2-o-trifluoromethylsulfonyl-d-mannopyranosides as precursors for concomitant 18 F-labeling and glycosylation by click chemistry. Carbohydr Res 344:753–761 Mindt TL, Struthers H, Brans L et al (2006) “Click to chelate”: synthesis and installation of metal chelates into biomolecules in a single step. J Am Chem Soc 128:15096–15097 Nicole P, Lins L, Rouyer-Fessard C et al (2000) Identification of key residues for interaction of vasoactive intestinal peptide with human Vpac1 and Vpac2 receptors and development of a highly selective vpac1 receptor agonist. Alanine scanning and molecular modeling of the peptide. J Biol Chem 275:24003–24012 Okarvi SM (2001) Recent progress in fluorine-18 labelled peptide radiopharmaceuticals. Eur J Nucl Med 28:929–938 Poethko T, Schottelius M, Thumshirn G et al (2004a) Chemo selective pre-conjugate radiohalogenation of unprotected mono- and multimeric peptides via oxime formation. Radiochimica Acta 92:317–327 Poethko T, Schottelius M, Thumshirn G et al (2004b) Two-step methodology for high-yield routine radiohalogenation of peptides: 18F-labeled RGD and octreotide analogs. J Nucl Med 45:892–902 Prante O, Einsiedel J, Haubner R et al (2007) 3, 4, 6-tri-oacetyl-2-deoxy-2-[18F]fluoroglucopyranosyl phenylthiosulfonate: a thiol-reactive agent for the chemoselective 18F-glycosylation of peptides. Bioconjug Chem 18: 254–262

R. Haubner and C. Decristoforo Schibli R, La Bella R, Alberto R et al (2000) Influence of the denticity of ligand systems on the in vitro and in vivo behavior of 99mTc(I)-tricarbonyl complexes: a hint for the future functionalization of biomolecules. Bioconjug Chem 11:345–351 Schottelius M, Rau F, Reubi JC et al (2005) Modulation of pharmacokinetics of radioiodinated sugar-conjugated somatostatin analogues by variation of peptide net charge and carbohydration chemistry. Bioconjug Chem 16:429–437 Smith-Jones PM, Stolz B, Bruns C et al (1994) Gallium-67/gallium-68-[DFO]-octreotide–a potential radiopharmaceutical for pet imaging of somatostatin receptor-positive tumors: synthesis and radiolabeling in vitro and preliminary in vivo studies. J Nucl Med 35:317–325 Thonon D, Kech C, Paris J et al (2009) New strategy for the preparation of clickable peptides and labeling with 1-(azidomethyl)-4-[18F]-fluorobenzene for PET. Bioconjug Chem 20:817–823 Trepel M, Arap W, Pasqualini R (2002) In vivo phage display and vascular heterogeneity: implications for targeted medicine. Curr Opin Chem Biol 6:399–404 Vagner J, Qu H, Hruby VJ (2008) Peptidomimetics, a synthetic tool of drug discovery. Curr Opin Chem Biol 12:292–296 Vallabhajosula S, Moyer BR, Lister-James J et al (1996) Preclinical evaluation of technetium-99m-labeled somatostatin receptor-binding peptides. J Nucl Med 37:1016– 1022 Velikyan I, Beyer GJ, Langstrom B (2004) Microwave-supported preparation of 68Ga bioconjugates with high specific radioactivity. Bioconjug Chem 15:554–560 Wester HJ, Schottelius M (2007) Fluorine-18 labeling of peptides and proteins. In: Schubiger AP, Lehmann L, Friebe M (eds) PET chemistry – the driving force in molecular imaging. Springer, Berlin

Optical Imaging

20

Ralf B. Schulz and Vasilis Ntziachristos

Contents 20.1 Introduction to Optical Imaging........................... 267 20.2 Light and Tissue...................................................... 268 20.3 Sources and Detectors............................................ 269 20.3.1 Choosing the Right Light Source.............................. 269 20.3.2 Photomultiplier Tubes and Multichannel Plates....... 270 20.3.3 Charge Coupled Device............................................ 271 20.3.4 The Right Choice?.................................................... 275 20.4 Optical and Fluorescence Tomography................ 275 20.5 Photoacoustic Tomography.................................... 277 References............................................................................ 278

R.B. Schulz (*) and V. Ntziachristos Institute for Biological and Medical Imaging, Helmholtz Center Munich, Technical University Munich, Ingolstädter Landstrasse 1, 85764 Neuherberg, Germany e-mail: [email protected]

20.1 Introduction to Optical Imaging The term “optical imaging” is not well defined; it refers to techniques ranging from simple photography to tomographic methods requiring most complex mathematical reconstruction algorithms. Common to all of these, however, is the use of a set of light sources to probe the object of interest, and usually a set of lightsensing devices to capture the resulting photon distribution from which the desired information is extracted by some kind of postprocessing. Recently, also hybrid techniques employing ultrasonic instead of optical detection have been reported and will be discussed further below. Optical imaging methods can be categorized according to the type of source-detector setting, and also according to the underlying contrast mechanism employed for imaging (Table 20.1). There are a lot of scatterers and absorbers present in the body to provide intrinsic contrast. Scattering contrast alone is used in clinical methods such as OCT (optical coherence tomography), which are outside the scope of this book. Intrinsic tissue absorption contrast is most often used to visualize the oxygenation statuses of tissue, as deoxy and oxyhemoglobin have significantly different absorption spectra; otherwise, intrinsic tissue contrast has been neglected as it did not show significant differences between malign and benign lesions in areas such as breast cancer. To overcome this contrast limitation, targeted fluorescent probes or activatable markers have been developed (Rudin and Weissleder 2003). They show either the attachment of fluorescence at target molecules, such as receptors, or the accumulation of activated probe near regions of increased enzymatic activity. With these powerful mechanisms, it became possible

F. Kiessling and B.J. Pichler (eds.), Small Animal Imaging, DOI: 10.1007/978-3-642-12945-2_20, © Springer-Verlag Berlin Heidelberg 2011

267

268

R.B. Schulz and V. Ntziachristos

Table 20.1 Overview over available contrast mechanisms and corresponding (tomographic) imaging techniques Contrast mechanism

Imaging method

Scattering

Optical coherence tomography (OCT)

commonly found obstacles and problems that might have to be faced when setting up experiments or imaging systems.

Diffuse optical tomography (DOT) Absorption Emission

Lifetime

Optoacoustic tomography (OAT) Multispectral optoacoustic tomography (MSOT) Fluorescence-mediated tomography (FMT) Fluorescence reflectance imaging (FRI) Bioluminescence imaging (BLI) Fluorescence lifetime microscopy (FLIM)

to distinguish better the molecular differences of tissues and diseases. Methods for imaging of such probes are different in intrinsic contrast-based methods as they require (a) the excitation of the probe at a certain wavelength and (b) the specific detection of the emission signal at a significantly different wavelength. From a system point of view, creating informative images under high scattering conditions has been the major challenge. Efforts initially concentrated on physically eliminating signal from scattered photons to enable imaging in thick tissues (Helmchen and Denk 2005; Yuste 2005; Low et al. 2006). While these methods – such as OCT and two-photon microscopy – are able to resolve tissue structures or the presence of fluorophores with a spatial resolution in the order of a few hundred nanometers, the maximum penetration depth in tissue is strongly limited. For deeper tissue penetration in the order of millimeters up to centimeters, the information from highly scattered photons needs to be included in the imaging process. Early macroscopic optical imaging methods used flat illumination for probe excitation and acquired planar images of the fluorescence. However, while these planar methods are nowadays often used for preclinical studies involving optical imaging, there is a strong need for quantitative tomographic techniques as it is crucial to quantify probe concentration and spatially resolve probe localization. Furthermore, planar images are highly surface-weighted, and detected intensities originating from fluorescent probes deep inside might be misleading. In this chapter, we will introduce how biological tissue interacts with light, how light can be guided into tissue, and how it is detected. In the final section, also the necessary postprocessing steps for more modern tomographic advances will be discussed. The main objective is to introduce the reader to the most

20.2 Light and Tissue The term light is referring to electromagnetic radiation at wavelengths in the range between a few hundred and one thousand (or slightly more) nanometers. This includes the visible spectrum from 400 to 700 nm, but also ultraviolet radiation and, of course, the near and far infrared region. Of special interest for macroscopic whole-body biological imaging are wavelengths above 650 nm, where the attenuation of light due to intrinsic absorbers is at a minimum and the water resonance is still low. This region, albeit not having strictly fixed limits, is termed water window. For a comprehensive list of tissue optical properties and corresponding references, please refer to the reference book by Vo-Dinh (2003). When light travels through biological tissues, it is not only absorbed, but also heavily scattered due to the many different layers with different optical densities of which tissue or even single cell layers are composed, and which refract and diffract incident light. The mean scattering free path of photons in tissue, i.e., the average distance they travel before being scattered, is only in the order of 0.1–0.01 mm. This explains why microscopic techniques requiring exact focusing – such as confocal or two-photon microscopy – fail after a few hundred microns. However, this figure is a bit misleading as most scattering events occur in a strictly forward direction. Tissue scattering is largely anisotropic with an average scatter angle of 25°, i.e., a scattered photon on average deviates from its original direction of flight by only 25°. To compare the effects of scattering better, instead of using the scatter mean free path as scattering coefficient ms, physicists often prefer the reduced scattering coefficient, m ¢s = (1 -g ) m s , which has been corrected for anisotropy. It is equivalent to totally isotropic scattering. Factor g is the expected cosine of the scattering angle; for a more graphical explanation of anisotropy (Fig. 20.1). Standard methods for scatter reduction, a standard tool in nuclear imaging and X-ray tomography, will fail due to the extreme number of scattering events detected photons have undergone. Scatter reduction tries to fit scatter-affected measurements to a scatter-free model.

269

20 Optical Imaging 140

40

150

30 g=0

160

20 g=0.5

170 Photon

180 190

g=0.9

10 0 350

200

340

210 220

330 320 Scattering Angle

Fig. 20.1 Anisotropic scattering. Typical values for tissue are an anisotropy factor of g >0.9, corresponding to an average scatter angle of ~25°

Optical tomographic methods in contrast have to be based on a model that takes into account heavy scattering. This is described further below. Scattering is decreasing with longer wavelengths, but otherwise, it remains relatively constant over the visible range. Absorption, on the other hand, is much more specific and contains many resonance peaks from different chromophores. When choosing an appropriate fluorochrome or the optimal wavelength for a specific imaging purpose, there is also another, counter-intuitive effect one might have to take into account: the choice of a wavelength in a strongly absorbing region will result in the preferential detection of photons that have undergone fewer scattering events, as scattering increases the length of the propagation path, and the higher absorption will suppress the detection photons propagating for too long. Thus, scattering can be significantly reduced – however, signal intensities are decreased as well. This has been extensively studied for different fluorescent proteins (Deliolanis et al. 2008), showing that far-red-shifted proteins do not necessarily enhance usability.

20.3 Sources and Detectors 20.3.1 Choosing the Right Light Source The light source is responsible for delivering sufficient numbers of photons at the desired wavelengths into the tissue or at least some volume of interest, harming neither the cells (burns, phototoxicity) nor the targeted dye

(photobleaching). For skin, generally the power limit ranges between ~20 mJ/cm2 for ultrashort pulses (~1 ns) and 100 mW/cm2 for long exposures. The exact numbers strongly depend on skin color, wavelength, and exact exposure times used and are, of course, much lower for photosensitive tissues such as cornea and retina, and also generally lower for UV radiation due to the photochemical and photophysical damage caused in cells. Generally, for planar reflectance imaging, one will never have to get anywhere near that limit while for photoacoustic imaging – still in its infancy – these limits might have to be exceeded in order to obtain a useful signal. For any application not requiring sharp wavelength spectra, excellent light collimation, high power, or short pulsing, light emitting diodes (LEDs) or incandescent lamps will suffice. Using appropriate bandpass filters, tunable filters, or interference gratings, they can reproduce nearly any desired wavelength within the broad output spectrum of the lamp. Incandescent lamps therefore are standard parts of fluorescence microscopes or spectrometer setups. Precise light collimation is only possible if the output spectrum is sharply peaked. If a sharply peaked spectrum or precise collimation is necessary, a laser source needs to be chosen. Laser light is strongly coherent, which, as a side effect, will automatically cause speckle noise if the beam is widened to illuminate a large area. Speckle noise is the occurrence of interference of strong coherent light when it is reflected from a surface. While light quickly loses coherence and polarization when undergoing many random scattering events, and thus, speckle noise is not a problem when using transillumination for thick tissue regions, reflectance-based imaging methods will be very susceptible to it, as most of the illumination light is directly reflected at superficial tissue regions. For planar imaging, laser sources are therefore far from optimal; for tomographic point source excitation schemes, however, they are perfect. Laser diodes – the cheapest member of the laser family – are nowadays available at many different wavelengths for a reasonable price. Other laser types can usually be tuned to emit at certain distinct wavelengths. If high powers are required at arbitrary wavelengths, but pulsing is not an issue, a supercontinuum source should be considered. These are based on nonlinear effects in photonic crystals that lead to the conversion of a short laser pulse into a continuum of wavelengths. The output power is in the order of 1–5 mW/nm, which allows trading sharp

270

R.B. Schulz and V. Ntziachristos

Table 20.2 Overview of available light sources for biomedical imaging Type of source

Advantages

Disadvantages

Laser (Solid state, gas, diode)

Discrete wavelengths, high photon density can be pulsed, limited wavelength tuning

Expensive, speckle noise, intensity fluctuations, not all wavelengths available

Widely tunable laser (OPO, supercontinuum)

Tunable output wavelength, high photon density, can be pulsed

(Very) expensive, unstable, often complex to operate

Light emitting diode

Cheap, simple to operate, stable output, can be modulated

Wide spectrum needs to be filtered, cannot be pulsed, not all wavelengths available

Broadband source

Simple to operate, standard tool for fluorescence microscopy

Wide spectrum, needs to be filtered, cannot be pulsed or modulated

wavelength selection for increased power and the cost of a supercontinuum source is comparable to about a dozen smaller laser diodes at fixed wavelengths. For photoacoustic imaging, where a single pulse needs to deliver 10–20 mJ within a nanosecond, however, high-power optical parametric oscillators (OPOs) are the only available and very expensive choice if wavelengths need to be selectable (Table 20.2).

20.3.2 Photomultiplier Tubes and Multichannel Plates For decades, photon multiplier tubes (PMT, Fig. 20.2) have been used as standard detectors for low photon currents. The momentum of photons incident on a charged photocathode frees electrons from that plate. Electrons can be accelerated in an electric field and then guided onto another charged cathode, freeing even more electrons. After several of these stages, a measurable voltage change is observed between the first cathode and the final anode. This voltage curve needs to be digitized and evaluated in real time. Generally, the shape of

the curve and the integral of it contain information on the total energy delivered by the incident photons. While there are new designs of multichannel photomultipliers, the number of measurement channels is generally limited due to size constraints of the tube. Thus, PMTs are employed when ultimate sensitivity is required at relatively low spatial resolution. They play a major role in nuclear imaging methods (PET, SPECT) and are used in single photon counting applications as well as in some early optical tomographic systems. Depending on the size of the tube, the number of amplification stages, and the gain, time resolution is around 200 ps at a gain of several millions. Aside from limited spatial resolution, there is another disadvantage of PMTs. As the voltage curve needs to be digitized and characterized for each single photon event individually, only a single photon may arrive within a time frame similar to the time resolution of the PMT; otherwise, the signal will be corrupted: first of all, when the first photon arrives, the electrodes will become partially depleted of electrons, and the amplification gain is reduced. Secondly, the peaks created by incident photons might not be distinguishable from each other, especially with the simple

dynodes photon

photon electron

R

Fig. 20.2 Schematic diagram of a photon multiplier tube (PMT)

R

R

R

secondary electrons

photo cathode accelerating voltage

Ra R

resulting voltage

271

20 Optical Imaging Fig. 20.3 Layout of a multichannel plate, with a single channel shown on the right

Primary electron

Secondary electrons

Final electron current

methods that have to be employed to analyze the signal in real time. Both effects will lead to a strong nonlinearity in photon fluence measurements as soon as the mean time between photon arrivals is nearing the time resolution limit of the PMT. Thus, PMTs can only be employed for extremely low photon currents. A modern form of PMT are microchannel plates (MCP, Fig. 20.3), which contain in a compact package several thousands of channels offering response times as low as 20 ps. They consist of a single semiconducting slab of ~2 mm thickness with metal coating on the opposing surfaces to produce an accelerating voltage potential for electrons. The semiconducting slab contains several thousands of fine channels (typically 6–10 mm in diameter separated by walls of 10–15 mm thickness), which act as individual electron multipliers. The primary electron is created from the metal coating and accelerated in the channel. As the channel walls are tilted slightly, it is guaranteed that the first electron will hit a channel wall where secondary electrons are created. A final two-dimensional (2D) current distribution can be measured at the bottom side of the MCP, opposite to the incident light. This measurement is often performed using charge coupled device (CCD). Due to the limited number of secondary electrons and the small length of each channel, the amplification gain is typically much lower than in PMTs but still in the order of 1,000×.

20.3.3 Charge Coupled Device Today, the most often used detector type for imaging is the so-called CCD, which directly measures a time-integrated 2D photon distribution incident on its

surface. CCDs are employed in a variety of imaging machinery, from digital cameras, scanners, to spectrometers. The importance of this device is underlined by the fact that its inventors were awarded with the Nobel Prize in physics 2009.

20.3.3.1 Working Principle CCDs were derived from analog shift registers. They consist of a number of elements similar to a capacitor that act as a potential well for electrons. These wells are aligned on a regular grid. A maximum number of electrons can be stored in each well without moving between wells. When an electrical field is applied, electrons can move out of their well in the direction of the field; as all electrons move at the same time, charges from different wells will not mix. In this way, rows or columns can be shifted horizontally or vertically, without changing the charge distribution. Electrons are injected into the wells either by shifting or through the photoelectric effect. This process is, for scientific-grade CCDs, nearly linear – i.e., for each arriving photon, the same number of photoelectrons will be injected – although some saturation effects will occur when wells are nearly full. The negative field generated by the electrons will either prevent further accumulation or lead to electrons flowing over to their neighbors (blooming), which will render parts of the image useless. Readout occurs by shifting the charges row by row and column by column into a single ADC chain (ADC: analog-to-digital converter, Fig. 20.4). This single chain is also the biggest bottleneck of CCDs: the readout of a 1,000×1,000 pixel image on a 100 kHz

272

R.B. Schulz and V. Ntziachristos

a

b

c

A/D A/D

A/D

Fig. 20.4 Different CCD types and their respective readout scheme. Light-sensitive pixel areas are shown in gray. (a) Classical CCD. Readout is performed by shifting the image by one column or row into a light-protected shift register whose contents are subsequently fed into the digitizer. (b) Frame transfer. The whole image is shifted to an area of the same size and

then read out as in a classical CCD. (c) Interline CCD. Lightsensitive and insensitive pixels are neighboring each other. Shifting the image by one column moves the whole frame into a light-protected area from which it can be read out using an additional shift register shown as an additional row

digitizer would take 10 s; at 1 MHz, this is still 1 s. During readout or charge-shifting, the photosensitive area of a standard CCD needs to be protected from light, as the additionally injected charges during readout would create smear in the resulting image. To minimize the time between readouts, special versions of CCDs have been developed, namely frame transfer and interline CCDs. In frame transfer (Fig. 20.4b), the photoactive region covers only one half of the actual chip. A complete image can be moved to an inactive region at once and then read out from there while the next image is formed on the photoactive side. This technology is quite expensive – as the chip area doubles – and it still requires a mechanical shutter as the whole image has to be shifted over a large area, and this takes enough time to create smear in light-intense applications. On interline CCDs (Fig. 20.4c), light-sensitive and light-protected pixels are neighboring each other in one direction. To shift the contents of a whole image from the light-sensitive to the light-protected side, a single pixel shift is required. To read electrons from the light-protected area, an additional shift register is required. With interline CCDs, a mechanical shutter is no longer required. In this cost-effective technology, the total size of the CCD is not increased as active pixel sizes are reduced. Reducing the size of the photoactive region of course reduces sensitivity by the same factor. This effect can be partially corrected by attaching light-collecting microscopic lenses on top of the CCD to guide photons to the photoactive pixels.

20.3.3.2 Increasing Sensitivity and Time Resolution As far as sensitivity is concerned, modern CCDs offer – for dedicated wavelength regions – up to 95% of quantum efficiency, i.e., a single photon will in 95% of the cases add electrons to the pixel well, so there is no significant improvement to be expected. This efficiency is in particular reached either using special coating on the front side of the chip to enhance UV sensitivity or, for the red region of the spectrum, using back illumination. In this case, light is incident not on the front side of the CCD but on its back side, the actual silicon waver, which has been thinned to near translucency. For these sensors, there is no interference between the low energy photoelectrons created by red light and the electric circuits that are located on the CCD’s front side. For far red and near infrared imaging at low light levels, the use of back-illuminated CCDs is preferable due to their increased sensitivity. However, when it comes to wavelengths >700 nm, an adverse effect might occur, called etaloning (illustrated in Fig. 20.5), which is often neglected when choosing an appropriate imaging system. The thinned waver has a thickness of 10–20 mm. Both sides of the waver will slightly reflect some of the incident light, which will lead to interference effects within the waver as soon as the photon wavelength approaches the thickness of the waver. As the thickness will vary slightly, interference fringes will be visible on the detector. This is independent of whether the incident light was coherent or not.

273

20 Optical Imaging Fig. 20.5 Etaloning shows interference artifacts on the silicone substrate of a back-illuminated CCD at wavelengths >700 nm. The patterns depend on the chip, not on the object imaged or the light source used. This is showcased here in three examples: (a, b) Transillumination of a scattering and absorbing liquid with a laser in central image position at different light levels and signal-to-noise ratios. (c) Transillumination of a mouse with a collimated laser source, showing the same pattern in well-illuminated regions

a

b

c

For application cases where etaloning is a problem, companies offer specially treated CCDs with improved antireflectance coating and thicker wavers (40–50 mm) made of a slightly different material. Those CCDs are sometimes referred to as deep depletion CCDs and offer improved sensitivity in the near infrared, however slightly less sensitivity in the visible. Once a photon actually injected some charges into a given pixel well, this charge still has to be amplified and digitized for the photon to be detected. The number of electrons injected per photon depends on the CCD and photon wavelength. Depending on the chosen preamplifier gain, which can often be changed by software, about two to eight electrons then correspond to one ADU (analog-digital-unit) of the A/D converter. However, this value does not correspond to the actual CCD sensitivity. Within the amplifier and ADC, a number of electrons might be injected or lost before digitization, which causes so-called read noise. Read

noise is depending on the actual electronics and is in the order of 10+ electrons. This value stays constant during pixel binning. Another source of noise is dark noise, which denotes randomly injected electrons, mostly due to thermal energy. Dark noise decreases with temperature, which is why deeply cooled sensors are used in low light applications. However, dark noise is a problem for long exposure times only (in the order of seconds) and is often overrated by distributors. In fact, with a CCD cooled down to 0° or slightly less, typically exposure times of up to 10 s are achievable without decreasing dynamic range. Both noise sources decrease sensitivity and dynamics. A further gain increase can be reached by amplifying either the injection process or readout. Amplifying charge injection can be achieved by adding an additional photocathode at a distance to the actual CCD sensor and applying an accelerating voltage. Photons will then create free photoelectrons, which will be accelerated by the

274

R.B. Schulz and V. Ntziachristos

applied field and hit the CCD sensor with high kinetic energy, creating secondary electrons within the sensor. This type of CCD is called electron-bombarded CCD (EB-CCD). A gain of 100–1,000× is achieved in this way – depending on accelerating voltage – which levitates single photon events to a level above read noise. Amplification before readout is performed in an electron-multiplying CCD (EM-CCD). Here, the serial register used for readout and digitization is much longer than a single line or row of the CCD. Instead of shifting the electrons with a small electric field as on the CCD, electrons are accelerated between wells using a stronger field, creating a number of secondary electrons on impact. This process is repeated several times. The total gain achieved is in the order of EB-CCDs, but this process inherently increases noise due to the nonconstant creation of secondary electrons, which are subsequently amplified. EM-CCDs are not used in conditions where signal is extremely low, but rather in conditions where readout speed is important: due to the strong amplification, faster digitizers (at rates of ~10 MHz) that produce higher read noise can be used. Another tricky question is how to acquire highresolution time-dependent data with CCDs. CCDs integrate over time, and any readout occurs through a single ADC chain, which is the major bottleneck of detection. Even with high-speed systems, the readout time is still in the order of 100 ms for highly sensitive megapixel cameras, creating a maximum frame rate of 10 frames/s. There are some extremely fast systems that push frame rates to 1,000 frames/s, but at the cost

of sensitivity and noise. Thus, very fast signals – on timescales of microsecond and nanoseconds – such as fluorescence lifetime cannot usually be recorded using a CCD but require the use of PMTs. If the process is repeatable, for example, by using multiple excitation pulses – as is the case for lifetime if the distribution of dye and its chemical environment is invariant of time – one can use a time-gated MCP coupled to a CCD sensor. In time-gated MCPs, the amplification voltage is only applied in a very short time-interval in which amplification can then occur. At all other times, electrons will be blocked from detection. These devices can then act as an electronic shutter, which opens repeatedly for a very short time, down to 50–100 ps. Total CCD exposure times are then chosen in the order of seconds, while the gating signal is derived from some kind of trigger, such as the modulation voltage of a pulsed laser diode. With these intensified CCDs (ICCD), the propagation of photons through diffusive media can be imaged at a time resolution of 100 ps with surprisingly good signal-to-noise ratio (Table 20.3).

20.3.3.3 Alternative Detectors CCD sensors essentially use an array of time-integrating capacitors to store the 2D photon distribution. Pixels on a CCD do not contain any electronic circuits; thus, they are sometimes referred to as passive pixels. Using CMOS (complementary metal oxide semiconductor)

Table 20.3 Typical light-sensing devices and their most prominent properties Detector

Advantages

Disadvantages

PMT

Time-resolved (20 ps)

Each PMT channel needs its own digitizer

Very high sensitivity

Photon density needs to be low (signal can only be evaluated if a single photon arrives)

Single photon applications

Low spatial resolution

Integrating detector

Not time-resolved (slow digitization)

CCD

High sensitivity High spatial resolution High and low light applications, adaptable sensitivity Only one digitizer channel necessary Standard tool ICCD

Time-resolved (100 ps)

High noise, medium sensitivity

Gating of repeatable events

Expensive

20 Optical Imaging

technology – a standard technique to create highly integrated circuits – an array of miniature photodiodes plus according amplifiers for each diode can be implemented. As each pixel now contains its own amplifying circuitry with in principle adjustable gain, they are termed active pixels. CMOS devices do not yet reach the sensitivity of stateof-the-art CCDs, but they have inherent advantages: • Pixels can be read out individually without destroying the content of neighboring pixels. • Blooming does not occur. • The gain can, in principle, be regulated on a perpixel basis, increasing the dynamic range without increasing the read noise (as is the case with multiple exposures using a CCD).

20.3.4 The Right Choice? Each of the presented light sources and photon sensing devices has its own peculiarities; the right choice for a given application is often hard to determine. For the biological or medical end user, choices might seem limited, as usually a turn-key system with integrated sourcedetector arrangement has to be selected. However, knowledge on the technical properties and potential problems – hardly described by any manufacturer – will help in determining which system offers the best compromise. As a general remark, optical imaging need not be expensive, and cheaper solutions are often best.

20.4 Optical and Fluorescence Tomography Exciting new developments over the past years have emerged from interdisciplinary research involving mathematics, physics, and biology. Those developments comprise physical models of light propagation in tissue that allow some kind of inversion process to yield the threedimensional distribution of either tissue optical properties (Boas et al. 2001), nowadays mostly employed for brain imaging (Boas et al. 2004), or fluorescence following a noninvasive in vivo measurement (Weissleder and Ntziachristos 2003; Ntziachristos et al. 2005).

275

The according imaging techniques – referred to as DOT (diffuse optical tomography) for the nonfluorescent case, and FMT (fluorescence-mediated molecular tomography), sometimes fDOT (fluorescence DOT) in the case of fluorescence imaging – allow for quantitative volumetric reconstructions instead of providing only qualitative planar images of fluorescence. Their aim is to noninvasively estimate a spatial map of either tissue optical properties or fluorescent probe concentration by illuminating the object under investigation with light at the excitation wavelength and subsequently measuring the generated fluorescence light distribution on the outer surface. By changing the position of the (usually collimated) light source, a number of measurements become available, allowing for tomographic reconstruction. The first FMT systems included a number of continuous-wave sources and time-integrating detectors coupled to the object under investigation via lightguiding fibers (Fig. 20.6a). The geometrical arrangement of sources and detectors was fixed such that either the imaged object had to be compressible enough to conform to that geometry or resulting gaps between object and source/detector fibers had to be filled with an optically matching liquid, usually a mixture of intralipid, water, and ink to match scattering and absorption coefficients of the object. Due to the fixed number of sources and detectors, resolution was limited; due to the use of a matching fluid, imaging procedures proved to be complex for small animals, which had to be embedded in the fluid and still be able to breathe. However, the use of a fixed geometry minimizes the computational effort as it greatly simplifies solving the forward model. A second system generation restricted the angular coverage to a mere ~120° by placing the animal in a chamber having a slab geometry instead of a cylindrical geometry. One side of the chamber contained a fixed array of source fibers; the other side was made of glass and allowed a CCD camera to be used as a more flexible detection system, providing literally hundreds of detector measurements by using individual groups of binned pixels as one detector in a postprocessing step (Fig. 20.6b). In phantom experiments, submillimeter resolution was reported for this kind of system. An exciting new approach in optical tomographic techniques (Hielscher 2005) was the development of noncontact imaging systems (Fig. 20.6c) in which a collimated free laser beam instead of light-guiding

276

R.B. Schulz and V. Ntziachristos

Fig. 20.6 Three generations of optical tomography setups. (a) First generation setup with source and detector fibers directly coupled to the object of interest, possibly embedded in a scattering fluid. (b) Fiber-based sources and free-space, high-density detectors, i.e., a CCD camera. The object is still embedded in a scattering fluid. (c) Free-space, 360° setup where either the object or the source-detector arrangement is rotated around the object. The object does not have to be embedded

a Filters Light Source

Detectors

b Filters CCD

Light Source

c Filters Light Source

fibers delivers light to the specimen, and light is detected via a lens-coupled CCD camera (Ntziachristos et al. 2005). By obtaining images from the outer surface of the animal, its shape can first be deduced and subsequently used to set up an adequate photon propagation model for that specific geometry. Latest generations of noncontact systems allow full 360° of view around the animal, and a large number of sourcedetector pairs (10,000–100,000) to yield high-resolution images. In general, optical tomographic methods are based on the following partially independent steps: • Measurement of optical projection data –– Using one of the system types described above –– Possibly minimizing the influence of noise by adapting system parameters such as source power, source positions, detector sensitivity, etc • Creation of a mathematical model (forward model) describing the expected measurements –– Under given optical properties and/or fluorescence biodistribution and

CCD

–– Adapted for the geometry of the measurement system and the specimen under investigation • Determining the most probable distribution of optical properties and/or fluorescence biodistribution (inverse problem, reconstruction) –– Using the model generated previously and –– Iteratively changing the input parameters (optical parameters or fluorescence) to minimize the difference between prediction and measurement The last step, the actual reconstruction, is significantly simplified if the derived model is linear. A linear model can be described by a matrix equation, y = Ax, with A being the model matrix, x being the unknown parameter (such as fluorescence distribution), and y being the measurement. Such linear systems arise in the case of fluorescence tomography, as fluorescence – under common conditions – is linear, meaning that the addition of some fluorescence in some location to a given distribution of fluorophores will add a specific signal to the measurement without disturbing the

277

20 Optical Imaging

signal created by the previous distribution. If linearity can be assumed, then the process of reconstruction simply comprises the calculation x = A -1 y, a standard problem in linear algebra with accordingly many possible solution techniques outside the scope of this chapter. For nonlinear problems, techniques (such as sequential quadratic programming) are available as well, however much more time-consuming. The model itself is derived from physical descriptions of light propagation in turbid media usually treating photons as ballistic particles and ignoring effects such as interference and diffraction caused by their wave nature. Depending on specimen size and thus the actual importance of scattering, different solutions can be derived, allowing, for example, highresolution imaging in very small organisms such as Drosophila pupae (Vinegoni et al. 2008), or medium resolution in thick specimen. For detailed descriptions of forward models and solving the inverse problems, refer to the reviews by Arridge (1999; Arridge and Schotland 2009). Generally, while tomography is mathematically complex and computationally expensive, requiring between several minutes and many hours for reconstruction, depending on the number of measurements and the desired resolution, especially FMT has the following unique advantages not found in conventional planar techniques: • (Semi)quantitative results that can be used to compare disease progression or therapeutic efficacy over time in the same subject. • Probably results quantitative enough to be used for comparisons between subjects. • Correct localization of fluorescent lesions within the specimen that can be overlaid with images obtained from other modalities (such as MRI and X-ray CT) to correlate with animal morphology, yielding results comparable to PET–CT. It has to be noted, though, that FMT is still an area of intense active research, and not all promises mentioned above have been realized to full capacity as of yet. One of the most challenging problems is the fact that currently, most tomographic algorithms consider the optical properties of the animal to be known and largely constant that is obviously a severe approximation. However, it has been shown that measuring the ratio between acquired fluorescence emission and

transmitted excitation source light can correct for most of the inhomogeneities of tissue. Another active field of interest is the combination of morphological imaging with optical imaging, such as MRI/optical and X-ray-CT/optical systems for better colocalization. Another intrinsic challenge in optical tomography is the fact that the reconstruction process in itself belongs to a class of mathematical problems termed ill-posed and ill-conditioned. A problem is ill-posed if the number of available measurements is much lower than the number of unknowns, i.e., the number of voxels required in the reconstruction. For free-space systems offering an arbitrary number of excitation source positions and acquiring full CCD images of the diffuse projections for each source, this is no longer the case, but it is commonly true for fiber-based systems. Ill-conditioning, much more severe, means that a large number of different possible solutions (in this case, biodistribution of fluorophores) will produce nearly the same predictions in the forward model – so close together, in fact, that measurement noise is larger than the difference between these predictions. In this case, a favorable solution has to be chosen arbitrarily, based, for example, on prior knowledge of typical fluorescence biodistribution shape properties. Of course, this arbitrary choice of a solution is subject of discussion and renders optical tomography to date less reliable than MRI, X-ray CT, or nuclear imaging modalities. These obstacles that ultimately reduce achievable resolution and reliability can potentially be overcome with techniques based on the photoacoustic effect, as described in the next section.

20.5 Photoacoustic Tomography Photoacoustic tomography (PAT) (Wang 2003; Ntzia christos et al. 2005) is an emerging technique capable of recovering the optical absorption coefficients in tissues by detecting ultrasonic waves originating from a photoacoustic interaction of a short laser pulse with tissue. The incident light pulse will be absorbed by tissue cells, instantaneously creating a small temperature elevation, and thus a thermal expansion. The expansion in turn gives rise to an acoustic pressure wave that will propagate through the sample and subsequently be detected on its outside as an acoustic signal. While the

278

effect can be observed with any radiation that will increase sample temperature, such as microwaves, the use of visible light is especially useful in the context of bioimaging due to its importance for fluorescence excitation and its relatively good tissue penetration at least in some wavelength ranges. Since its first reported use for biological purposes, different systems have been developed for preclinical use, and even clinical applications for optical mammography have been proposed (Ku et al. 2005; Gu et al. 2006). Compared to other optical wholebody techniques, it offers an excellent resolution of a few hundred microns, which can even be improved by coupling the technique to a microscope (Zhang et al. 2006). By using several different wavelengths, it is possible to distinguish the very distinct peaks or spectra of different chromophores and thus use the technique for fluorophore detection or blood oxygenation measurements (Razansky et al. 2009). Note, however, that activation or deactivation of a probe can be detected only if the absorption profiles of the respective dye change significantly as fluorescent emission cannot be detected using PAT techniques. In cases where quenched dyes are activated by simple enzymatic cleavage, the absorption spectra of quenched and activated probe might correspond exactly. PAT offers the following advantages over classical optical tomographic techniques: • The time-resolved ultrasonic signals can be reconstructed using fast backprojection algorithms quite similar to X-ray CT algorithms. • PAT offers a direct, potentially quantitative, measurement of tissue absorption, as only absorption – not scattering – is responsible for the thermal elevation. • The achievable resolution, even within thick tissue samples, is, in principle, in the order of 100 mm or less. Of course, all of these potential benefits need to be carefully considered. The resolution might be much better than in FMT or DOT, but this requires the excitation pulse to deliver sufficient energy into the tissue. Also, the delivered pulse still needs to be a very short pulse still creating an instantaneous temperature elevation; in scattering tissues, however, the pulse will disperse and decay exponentially with penetration depth. Secondly, even the analytical backprojection algorithms demand prior knowledge on the energy distribution of the

R.B. Schulz and V. Ntziachristos

incident pulse at least if quantitative results need to be obtained. This can only be performed by solving the aforementioned forward problem for DOT or FMT, which, in turn, would require knowledge on the attenuation coefficients that should be obtained from the backprojection; thus, the proper solution can only be obtained in an iterative fashion. Thirdly, and most importantly, regions with strongly differing sound propagation properties or that pose a barrier to ultrasound propagation (such as lungs or bowels that might contain a lot of air) need to be properly modeled mathematically in order to obtain useful images – if possible at all.

References Arridge SR (1999) Optical tomography in medical imaging. Inverse Probl 15(2):R41–R93 Arridge SR, Schotland JC (2009) Optical tomography: forward and inverse problems. Inverse Probl (in press with 25:123010) Boas DA, Brooks DH et al (2001) Imaging the body with diffuse optical tomography. IEEE Signal Process Mag 18(6): 57–75 Boas DA, Dale AM et al (2004) Diffuse optical imaging of brain activation: approaches to optimizing image sensitivity, resolution, and accuracy. Neuroimage 23:S275–S288 Deliolanis NC, Kasmieh R et al (2008) Performance of the redshifted fluorescent proteins in deep-tissue molecular imaging applications. J Biomed Opt 14(4):044008 Gu HM, Yang SH et al (2006) Photoacoustic tomography and applications in the medical clinic diagnosis. Prog Biochem Biophys 33(5):431–437 Helmchen F, Denk W (2005) Deep tissue two-photon microscopy. Nat Methods 2(12):932–940 Hielscher AH (2005) Optical tomographic imaging of small animals. Curr Opin Biotechnol 16(1):79–88 Ku G, Fornage BD et al (2005) Thermoacoustic and photoacoustic tomography of thick biological tissues toward breast imaging. Technol Cancer Res Treat 4(5):559–565 Low AF, Tearney GJ et al (2006) Technology insight: optical coherence tomography – current status and future development. Nat Clin Practice Cardiovasc Med 3(3): 154–162 Ntziachristos V, Ripoll J et al (2005) Looking and listening to light: the evolution of whole-body photonic imaging. Nat Biotechnol 23(3):313–320 Razansky D, Distel M et al (2009) Multispectral opto-acoustic tomography of deep-seated fluorescent proteins in vivo. Nat Photon 3(7):412–417 Rudin M, Weissleder R (2003) Molecular imaging in drug discovery and development. Nat Rev Drug Discov 2(2):123–131 Vinegoni C, Pitsouli C et al (2008) In vivo imaging of Drosophila melanogaster pupae with mesoscopic fluorescence tomography. Nat Methods 5:45–47 Vo-Dinh T (2003) Biomedical photonics handbook. CRC, Boca Raton

20 Optical Imaging Wang LHV (2003) Ultrasound-mediated biophotonic imaging: a review of acousto-optical tomography and photo-acoustic tomography. Dis Markers 19(2–3):123–138 Weissleder R, Ntziachristos V (2003) Shedding light onto live molecular targets. Nat Med 9(1):123–128

279 Yuste R (2005) Fluorescence microscopy today. Nat Methods 2(12):902–904 Zhang HF, Maslov K et al (2006) Functional photoacoustic microscopy for high-resolution and noninvasive in vivo imaging. Nat Biotechnol 24(7):848–851

21

Optical Probes Kai Licha

Contents 21.1 Introduction: Optical Probes for Animal Imaging...................................................... 281 21.2 State-of-the-Art Technologies................................ 282 21.2.1 Basics on Fluorescent Dyes...................................... 282 21.2.2 Nontargeted and Vascular Dyes as Contrast Agents........................................................ 283 21.2.3 Reactive Dyes and Labeled Targeting Molecules.................................................................. 284 21.2.4 Activatable Probes.................................................... 285 21.2.5 Fluorescent Particles as Imaging Agents.................. 286 21.3 The Choice of Parameters...................................... 286 21.3.1 Optical Properties..................................................... 286 21.3.2 Pathophysiological Paradigm.................................... 287 21.3.3 Type of Labeling Chemistry..................................... 288 21.4 Conclusion and Outlook......................................... 290 21.5 Laboratory Protocols.............................................. 290 21.5.1 Labeling of Antibody with Dye NHS-Ester............. 290 21.5.2 Labeling of Antibody with Dye Maleimide.............. 291 References............................................................................ 291

K. Licha mivenion GmbH, Robert-Koch-Platz 4, 10115 Berlin, Germany e-mail: [email protected]

21.1 Introduction: Optical Probes for Animal Imaging The application of optical imaging technologies for drug discovery research and the development of novel preclinical animal models has expanded with tremendous vigor in the past few years. The fact that fluorescent dyes can be detected at low concentrations and nonionizing and harmless radiation is applied with rather low technical effort makes optical techniques attractive for routine use in the animal imaging laboratory. Novel imaging probes and contrast agents have been designed in a broad variety addressing the various requirements given by the disease problem at the preclinical animal imaging stage. Moreover, the industry has identified this field as a market from the side both the imaging equipment and the fluorescent probes applied as readily injectable contrast agents or reactive labels for bioconjugation chemistry. Imaging devices with powerful and versatile capabilities are now commercially available, addressing different needs of the customer, such as multicolor imaging, time-resolved imaging, planar or tomographic geometry, signal quantification, spectral unmixing for background correction, or zooming within macroscopic scale. A closer look into these topics is provided elsewhere in this book. The design of fluorescent imaging probes for in vivo imaging applications has likewise emerged and is reflected by an increasing number of commercialized probes and labels. The underlying chemistries have been optimized for their capability to monitor disease-specific anatomic, physiological, and molecular parameters by way of directly applicable imaging probes or by specific reactive labels, which can be used by the customer to fluorescently label his own drug candidate for further studies or derivatize important biological ligands to investigate the basics of a given targeting approach.

F. Kiessling and B.J. Pichler (eds.), Small Animal Imaging, DOI: 10.1007/978-3-642-12945-2_21, © Springer-Verlag Berlin Heidelberg 2011

281

282

K. Licha

The main objective of this contribution is to provide an overview on the different choices available when setting up an imaging protocol using optical probes. Hereby, the review sets a focus on practical aspects of the application of optical probes for animal imaging and drug discovery research. The state-of-the-art of imaging probes is first discussed covering nontargeted and vascular agents, approaches to generate targeted agents via reactive fluorophores, the field of enzymatically activatable probes, and imaging agents based on particles. To facilitate the researcher’s choice, a selection of relevant parameters is discussed in the subsequent chapter, discussing appropriate examples out of the literature. At this point is mentioned that examples include published work and refer to selected commercial compounds and companies as well. The intention of this review is neither to provide a comprehensive data set on these probes nor to accentuate particular commercial suppliers. Figure 21.1 summarizes the general parameters that contribute to the design of optical imaging probes.

21.2 State-of-the-Art Technologies 21.2.1 Basics on Fluorescent Dyes An efficient fluorescence imaging probe should employ a combination of different properties, which are (1) a

high fluorescence quantum yield in the desired wavelength spectrum, (2) sufficient biological stability and photostability to permit unimpaired image acquisition, and (3) reasonable solubility in (most) aqueous environment to enable bioconjugation and labeling reactions without substantial triggering of aggregation and/ or precipitation. Among the many possible organic fluorophore classes, the polymethine dyes (e.g., cyanines, hemicyanines, benzopyrillium dyes), xanthene dyes (rhodamines and fluoresceins), and also some oxazoles and thiazoles (e.g., Methylene Blue, Nile Blue) are basically the ones that are widely established and have been contributing to the majority of scientific results (Licha et al. 2008). Figure 21.2 illustrates typical absorption and fluorescence spectra of these types of fluorophores, thereby showing the wavelength range within imaging is possible. While xanthene dyes do not reach fluorescence emission far beyond 700 nm, they exhibit extremely high fluorescence quantum yields often close to 100%. In turn, cyanine dyes are capable to cover the entire spectrum from VIS to NIR (>900 nm) with high extinction coefficients (up to 250,000 L/ mol cm), but rather moderate fluorescence quantum yields (up to 30%). Besides organic fluorophores, inorganic semiconductor nanoparticles have emerged as alternative fluorescent reporters (see Chap. 21.2.5). As further detailed in the subsequent chapters, the most relevant applications for these dyes as imaging tools are the direct use as injectable contrast agents in

Optical properties, quantum yield, FRET and/or quenching

Targeting or drug molecules

Properties of fluorophore

Fluorescent conjugates for animal imaging

Conjugation chemistry, dye-to-protein ratio, targeting properties Fluorophore: • Compound class (cyanine, fluorescein …) • Solubility, hydrophilicity • Chemical stability • Self-made or commercial

Optical properties: • Wavelength range (UV, VIS, NIR) • Fluorescenc quantum yield • Quenching or FRET effects • Photostability

Fig. 21.1 Basic idea and general parameters for the generation of optical probes

Conjugation: • Type of targeting molecule • Selection of reactive groups and type of conjugation chemistry • Dye-to-protein ratio • Functionality of targeting molecule

283

21 Optical Probes

absorption − / fluorescence....

carbocyanin dyes +

N

n=1

n

xanthen dyes

O

O

OH

+

N

N

N COOH

n=2

COOH

n=3 fluoresceins

500

O

600 700 wavelength(nm)

800

400

rhodamines

500 600 wavelength(nm)

700

Fig. 21.2 Typical absorption and fluorescence spectra of cyanine dye and xanthenes dyes shown as idealized illustration

a nontargeted or passively distributing format, the use as reactive fluorophore labels to create targeted conjugates of biological or synthetic origin, and the use as quenched components for activatable imaging probes to monitor different kinds of enzyme activity.

21.2.2 Nontargeted and Vascular Dyes as Contrast Agents A variety of fluorescent dyes for passive targeting and vascular circulation have been applied as fluorescent imaging probes. These probes, most of them carbocyanine dyes, represent the basic chemicals for the design of labels and probes. Using these dyes as imaging agents by themselves, they act by their physicochemical and structural properties since they do not employ structures designed for a specific molecular interaction. However, the suitability for animal studies, e.g., the characterization of experimental tumor models, has been demonstrated in various publications. The near-infrared dye Indocyanine Green (ICG, absorption ~780–800 nm) was approved in the 1960s as a diagnostic drug, was rediscovered in the 1990s as an imaging agent, and is today, together with fluorescein, frequently applied for fluorescence angiography in ophthalmology (Richards et al. 1998). Similarly, novel cyanine dye derivatives have been created as passively targeted contrast agents of simple synthetic availability. Examples are the dye SIDAG, which leads to enhanced uptake in a variety of tumor models and

permits differentiation of the degree of angiogenesis (Wall et al. 2008). Furthermore, the compound was shown to enable fluorescence imaging of inflammation in a model of rheumatoid arthritis (Fischer et al. 2006). It therefore provides a powerful tool for the validation of preclinical models, imaging equipment and therapeutic studies (TryX750; mivenion). With the IRDyes (LI-COR Biosciences) or Cy5.5 (GE Healthcare Life Sciences), other examples of nontargeted or control dyes with published results for animal imaging and human applications, such as lymphatic mapping or studies on T-cell migration, accompany ICG (Foster et al. 2008; Rasmussen et al. 2009). Another class that has recently attracted attention is the benzopyrriliumbased polymethine dyes (Dyomics) (Pauli et al. 2009). In Fig. 21.3, a panel of selected chromophores is depicted. Beyond ICG, which by itself is not directly applicable to synthesize reactive derivatives for biolabeling due to the absence of chemical functionalities, the compounds discussed here represent basic structures to create in few further steps reactive labels and subsequently any desired bioconjugate. Cyanine dye conjugates with substantially increased molecular weight have been designed to enable the visualization of vascular structures and to study effects on the vascularity. For this purpose, the dyes were covalently conjugated to macromolecular carriers, such as albumin, dextrane, or other biopolymers (Klohs et al. 2009) or non-covalently entrapped into dendritic architectures used as passive drug delivery systems (Quadir et al. 2008). These conjugates have been useful for imaging tumors, but also to visualize and characterize

284

K. Licha

a

b HO

N+

HO

N

O3S

SO3−Na+

c

SO3−Na+

Na+−O3S

d

O N+

−

o3s

OH

SO3−Na+

−

H N

HO

N+

N

O3S

OH

HO

OH

SO3−

−

O3S SO3−

−O S 3

e

O OH

N

O

N

SO3−Na+

N

O OH

OH

SO3−Na+

−

SO3−

H N

HO

OH

N+

N

O

O

O+

OH

Fig. 21.3 Selected near-infrared fluorescent dyes, which have been used as nontargeted probes for in vivo fluorescence imaging: (a) Indocyanine Green, (b) SIDAG (Wall et al. 2008), (c)

IRDye® 800CW (LI-COR, Foster et al. 2008), (d) Cy5.5® carboxylic acid (GE Healthcare) and (e) DY-750 (Dyomics)

tissue architecture in vivo on a macro to microscopic imaging level. Commercialized are, e.g., streptavidin, ovalbumin, protein A or dextran conjugates with a variety of Alex Fluor® labels, marketed under the brand SAIVI™ (Invitrogen) as imaging probes and secondary detection materials. Other available probes are the Angio Sense® probes (VisEn Medical) based on large, long-circulating macromolecules at 250 kD (type of macromolecular carrier not published).

and cellular level to the in vivo application in higher organisms. Generally, optical imaging probes in the format of bioconjugates range from protein conjugates, antibody conjugates, structures based on oligonucleotides to receptor-avid peptide conjugates. Furthermore, a large panel of small molecule substrates and ligands, such as folate, cyanocobalamine, glucose, bisphosphonates, as well as rationally designed receptor ligands has been conjugated with fluorophores for in vivo studies. This work has ultimately led to a large variety of targeted imaging agents published in the past few years. A selection of different reactive dye labels and resulting conjugates is depicted in Fig. 21.4. Being not the intention of this chapter to discuss these versatile approaches in detail, the reader is referred to respective reviewing publications (Achilefu 2004; Tung 2004; Licha and Olbrich 2005). A few conjugates have been made commercially available addressing fundamental pathways in angiogenesis and tumor growth, for instance EGF conjugates (LI-COR). Oligonucleotides and small peptides, as opposed to, e.g., targeting molecules of biotechnological origin, are accessible via solid phase synthesis with the fluorophore conjugation capable of being implemented into the synthetic protocol. Most of the dyes used have shown to be stable enough with respect to the chemical

21.2.3 Reactive Dyes and Labeled Targeting Molecules This chapter addresses the increasing demand for fluorescently labeled drug candidates based on molecules of synthetic (small organic molecules, peptides, oligonucleotides) or biotechnological origin, such as engineered proteins, antibodies and novel protein formats. In order to get an insight into the targeting capabilities of these drugs in vivo, optical imaging is establishing as a fast and easy-to-apply technology in the animal imaging laboratory. As further outlined in Chap. 21.3. fluorescence labeling offers the opportunity to study biological and physicochemical properties of a desired candidate along the entire path from synthesis, in vitro

285

21 Optical Probes

a

b

SO3−Na+ −

Na+ −O3S

SO3 Na

+

N O

O

O

SO3−

O N

O

N+

N

SO3−Na+ O+

−

O3S

O O

N

HN

N

O

c

N SO3−

O

−

O3S

N+

H N

H N

O O

O

N H

NH2 NH

N

H N

HO2C

N

O

H O

N

d

O

SO3− N+

−

O3S

H N

DPhe-Cys-Phe-DTrp-Lys-Thr-Cys-Thr-OH

N

SO3−Na+

Fig. 21.4 Selected dyes in a reactive format for biolabeling and published bioconjugates: (a) IRDye® 800CW Maleimide (LI-COR) (b) DY-781 NHS-ester (Dyomics), (c) Cy5.5-folate

conjugate (Moon et al. 2003) and (d) Indotricarbocyanineoctreotate (Becker et al. 2001)

conditions required for peptide cleavage from the resin and other interventions (e.g., forming disulfide bridges via oxidation). Examples for the synthesis of lowmolecular weight peptide conjugates comprising different cyanine or rhodamine dyes have been published (Becker et al. 2001; Mier et al. 2002). For the 3¢ and 5¢ labeling of oligonucleotides, conjugation chemistry can be applied similarly. A broad set of labels and quenchers covering the entire VIS to NIR is for instance offered with the ATTO dyes (ATTO-TEC).

unique opportunity of modulating optical signals through intramolecular fluorescence quenching effects (Weissleder et al. 1999). These probes are “activatable” polymeric conjugates based on a poly-l-lysine/ polyethylene glycol graft polymer, which are “switched on” when single fluorochrome fragments are cleaved from the fluorescence-quenched polymeric structure by tumor-specific proteolytic enzymes, thus leading to signal amplification of up to 200-fold. In various preclinical disease models, the agents have demonstrated the imaging enzymatic activity and report the in vivo efficacy of enzyme-inhibiting drugs (Bremer et al. 2001). The underlying imaging probes are based on cleavable graft polymers of high molecular weight and are commercialized by VisEn Medical Inc. Under different trade names (ProSense®, MMPSense®), the compounds are cleaved and activated by enzymes such as cathepsins and matrix metalloproteinases at spectral ranges of 680

21.2.4 Activatable Probes Weissleder and coworkers have introduced an optical imaging approach, which is of fundamental difference to the approaches described above and utilizes the

286

K. Licha

and 750 nm, respectively. More recently, rationally designed conjugates of defined chemical structures, activatable photosensitizers for Photodynamic Therapy (Stefflova et al. 2007), as well as constructs sensitive to cleavage by other enzymes, such as thrombin and caspases (Barnett et al. 2009), have expanded this versatile technology.

Nanoparticles are particularly interesting for multi modality imaging purposes due to their modular composition. Magnetic cores for MRI imaging can be functionalized with additional fluorophores to allow fluorescent tracking, and vice versa, fluorescent cores have been combined with paramagnetic complexes (Azar and Intes 2008).

21.2.5 Fluorescent Particles as Imaging Agents

21.3 The Choice of Parameters

Fluorescent imaging agents based on particles have shown great utility, especially for applications on cellular staining (in vitro) and vascular imaging (in vivo). In principle, there are two different groups of nanoparticle probes. On the one hand, nanosized materials based on organic and/or inorganic chemistry have been equipped with organic fluorophores, by either incorporating the dye molecules into the nanoshells or attaching it to surface functionalities (Azar and Intes 2008). In particular, semiconductor nanoparticles (quantum dots) have recently emerged as powerful imaging tools with brilliant optical properties (Resch-Genger et al. 2008). These semiconductor nanocrystal materials consist of atoms such as Cd, Se, Te, S, and Zn. The fluorescence emission range depends on the diameter of the particles, which can be synthesized in a very controlled fashion with respect to size and surface modifications (Fig. 21.5). Surfaces employing stabilizing polymers as well as targeting molecules, such as antibodies and peptides, have expanded their applications for in vivo animal imaging (Bentolila et al. 2009).

Fig. 21.5 Typical absorption and fluorescence spectra of quantum dots shown as idealized illustration; principal design of targeted quantum dot conjugates

21.3.1 Optical Properties Generally, the choice of the spectral range of absorption and fluorescence of the dye determines whether it is detectable on tissue surfaces (UV-VIS dyes; 400– 700 nm) or from deeper-located tissue areas (NIR dyes; >700 nm). A prerequisite for sensitive detection of a contrast-enhancing dye is a high extinction coefficient at the desired absorption wavelength. If fluorescence is recorded within the UV/VIS spectral region, both autofluorescence and the administered contrast agent will contribute to the observed signal, while in the NIR spectral region, tissue autofluorescence is negligible due to the absence of endogenous NIR fluorophores. In the latter case, the detected signal nearly exclusively reveals the distribution of the optical imaging probe. The broad availability of fluorophore labels allows the selection of probes exactly matching the technical parameters of a given analytical instrument. Under practical circumstances is the application of one particular fluorophore for a research program starting at the in vitro

287

21 Optical Probes

level with cell microscopy, FACS, and other assays, then reaching in vivo imaging studies in animals, often not possible. Analytical instruments are usually not equipped with excitation and detection components permitting detection with sufficient sensitivity over the entire spectral range. Therefore, a work-around strategy often involves the choice of an appropriate series of fluorophores that differ only in minor extent in their chemical structures. This strategy minimizes unexpected effects on the resulting targeting conjugate when changing the fluorophore backbone. Here, cyanine dyes with their carbo, dicarbo, and tricarbocyanine homology are well suited. Examples of dyes offered as series are the CyDye™ series (GE Healthcare Life Sciences), which have been one of the first comprehensive product lines for fluorescence labeling (Cy3, Cy5, Cy5.5, Cy7). The Alexa Fluor® Dye series (Invitrogen/Molecular Probes) have expanded the available spectral emission range from approximately 450 to 800 nm including derivatives (cyanine and xanthene) of improved photostability and fluorescence quantum yields. Alexa Fluor 488, Alexa Fluor 546, Alexa Fluor 555, and Alexa Fluor 750 are examples for products (consistently named by their excitation maximum). Another group is based on a different chromophore type, the borondipyrromethens, known as BODIPY dyes (Invitrogen) (Ulrich et al. 2008). For a representative cyanine dye series, the typical color appearance of solutions (approximately 0.1 mM) in water is depicted in Fig. 21.6, which might facilitate the identification of

c

a

b

Fig. 21.6 Photograph of cyanine dye solutions: (a) indocarbocyanine with abs/fluor 550/580 nm, (b) indodicarbocyanine 650/680 nm, and (c) indotricarbocyanine 750/780 nm. Concentration of solutions approximately 0.1 mM in water

the correct probe in case of some disorientation caused by too many flasks at the same time on the bench. Generally, reactive dyes are offered in different version of chemoselectivity (see Chap. 21.3.3) so that the most reasonable labeling chemistry can be applied. The obtained amount of conjugate has to match the extent of the planned animal imaging study. Typically, fluorescent conjugates have been applied at doses in the range of 0.1–1 mmol/kg (i.v.), which corresponds to approximately 0.1–1 mg/kg amount of fluorophore (calculation based on the molecular weight of the fluorophore). Under technical aspects, the dose should be chosen such that a reasonable enhancement of the overall signal intensity in the animal directly after administration is detected (e.g., at least tenfold increase compared to background signal before injection).

21.3.2 Pathophysiological Paradigm Viewing the applications of optical imaging probes from the angle of their pathophysiological function, it appears as common sense to group the probes into nontargeted or vascular agents, targeted conjugates, and activatable probes. As detailed in the preceding chapter on state-ofthe-art technologies, a host of compounds have been designed and studied for each of these three fundamental strategies. It depends on the rationale of the respective scientific problem to be solved by the application of in vivo optical imaging. Vascular or nontargeted fluorophores, as well as circulating fluorescent particles might be used first when a general imaging protocol is to be established, for the characterization of novel disease models (growth characteristics, vascularity, tumor sizes, formation of metastases etc.). Other interesting aspects refer to the dose finding, detection limits, and image acquisition parameters. Targeted conjugates permit the imaging of molecular information, such as specific cellular expression profile. Fluorescently labeled drug candidates can be monitored with respect to their in vivo uptake into the target tissue and their capability to bind the molecular target. The underlying rationale is often to study the pharmacokinetic behavior of the probe, which has to fulfill certain properties with respect to their adsorption, distribution, metabolization, and excretion (ADME) for the living organism. Optical imaging can provide a fast and easy access to semiquantitative ADME data; however, a true quantification of drug

288

K. Licha

amounts and the fate of metabolites in the organs is technically difficult. Here, radioactive techniques are a method of choice. Activatable probes provide insight into the third level of detecting protein function, such as the activity of proteolytic enzymes. Here, the injected imaging probe is silent until activated by the corresponding enzyme. Accordingly, a concentration ratio between diseased tissue and normal areas is not necessarily a requirement as long as the circulating fraction of the probe remains undetectable. This technology has shown to be applicable for the imaging of enzyme inhibitors in drug screening protocols (Fig. 21.7). Probably, the combined exploitation of these mechanisms would allow that the deepest insight into the physiology and molecular biology of disease formation and progression be gained. At this point, it is worth to emphasize that a particular opportunity of fluorescence detection is the principal capability of multicolor imaging given that the imaging instrumentation permits multiple excitation and/or detection wavelengths. To give an example, the tumor vascularity would be detectable by a long-circulating vascular probe (e.g., 680 nm excitation), while the analysis of a certain receptor expression could be accomplished by a target-specific conjugate working at 750 or 800 nm.

21.3.3 Type of Labeling Chemistry

Fig. 21.7 Ilustration of the different mechanisms of contrast generation by fluorescent imaging probes. In principle, the target-to-background ratios of signals and resulting imaging con-

trasts increase from nonspecific optical probes to targeted conjugates and activatable probes. Modified with permission from Sonja Vollmer

For the labeling of proteins, antibodies, or other macromolecules, the type of labeling reaction has to be considered. Generally, different residues in the macromolecule are available for covalent attachment to reactive fluorophores. The most relevant reaction sites are amino groups (lysins), thiol groups (cystein), hydroxy groups (threonine, serine, and glycans), and aldehydes (generated from oxidative cleavage of glycosilated structures). The principle pathways of covalent labeling of amino and thiol groups (provided by the amino acids, lysine and cysteine, respectively) are illustrated in Fig. 21.8. Labeling of lysins is most established by using N-hydroxysuccinimidyl esters of dyes, easily accessible from carboxylic acid groups. Another approach is the utilization of isothiocyanate groups (most prominent is fluorescein-isothiocyanate, FITC). Labeling of cysteins occurs most conveniently with maleimides or with pyrridinium disulfides. A combined strategy involves the thiolation reagent iminothiolane (Traut’s reagent), which enables the conversion of an amine group into a linker with free thiol group for subsequent maleimide labeling (Hermanson 1996).

289

21 Optical Probes Fig. 21.8 Illustration of synthetic pathways for selective labeling of antibodies or other proteins; the fluorophore is simplified as bulb. Depicted are reactions of (a) maleimidedye with cysteine, (b) pyridinium disulfide dye with cysteine, (c) NHS-ester dye with lysine, (d) isothiocyanate dye with lysine, and (d) modification of lysine with Traut’s reagent to give thiol, followed by reaction with maleimide dye

a

b

c

d

e

Labeling usually occurs with high chemoselectivity, allowing that cysteins or lysins be targeted independently. In native proteins, labeling occurs statistically yielding an average dye-to-protein ratio with randomly distributed fluorophores due to the randomly located cysteins or lysins in the protein. In this respect, reactions at sensitive binding sites or amino acids involved

in the molecular function can hamper the affinity of the biomolecule. Therefore, the labeling approach should consider the protein sequence in order to select the appropriate labeling reaction. A complementary approach can be followed by employing structures of glycosilation by reducing sugar moieties in a way that aldehyde functionalities are

290

generated. These aldehyde groups react readily with amino groups (forming imines) or hydrazine groups (forming hydrazones), both of which are acid-cleavable but can be chemically stabilized by reduction with sodium cyanoborohydride (Hermanson 1996). Thiols are usually accessible by reductive breaking of disulfide bonds in the protein. However, since these bonds are crucial for protein folding and stability, denaturation and loss of target binding capability can occur. A structurally more defined approach can be followed by using engineered antibodies with additional cystein residues, which are not involved into protein structure and are free to be used for covalent labeling (Lee et al. 2008). As outlined in Chap. 21.2, a host of fluorophores are available for the different bioconjugation chemistries. In a typical protein-labeling protocol, a buffered solution (pH 7.5–8) of the protein (1–10 mg/mL) is incubated with a molar excess of reactive dye (typically 10–50 moles of dye per mole biomolecule) for a few minutes to several hours depending on dye reactivity and desired labeling ratio. Purification and separation from unbound dye can be achieved by separation methods such as dialysis, ultrafiltration, or size exclusion chromatography. In order to determine the quality of the resulting conjugate, suitable analytical methods are needed subsequently.

21.4 Conclusion and Outlook Optical imaging has an established role in preclinical research and animal imaging. The application of imaging techniques for the detection for fluorescence signals in living animals has emerged to a powerful routine level and utilizes a versatile set of commercially available imaging equipment. The detection of fluorophores as exogenously applied contrast agents can be accomplished with convenient handling and without a substantial effort regarding technical installation, safety issues, and training. A broad variety of fluorescent probes have been created and validated for their ability to facilitate the set-up of imaging protocols, to support the characterization of disease models, and to accompany the drug discovery process from drug design, in vitro characterization to the in vivo proof-of-concept. Both the selection and application of commercially available products

K. Licha

optimized for a specific function, as well as the implementation of self-induced synthetic chemistry adapting to the requirements in a given drug discovery project will play a central role in the optical imaging laboratory.

21.5 Laboratory Protocols 21.5.1 Labeling of Antibody with Dye NHS-Ester Generally, antibodies should be used in a buffer free of amines. For instance, phosphate, acetate, or borate buffers are recommended, whereas TRIS and HEPES buffers are less suited. Incubation is performed with large excess of dye, followed by removal of excess nonreacted dye by using gel chromatography, dialysis, or ultracentrifugation. The latter two methods bring the dye conjugate into contact with material surfaces for which precipitation and loss of reaction yield are sometimes observed especially when using less hydrophilic dyes with tendency to aggregate. The following steps are recommended. • Dissolve antibody or protein to be labeled in reaction buffer at concentration of 1–10 mg/mL depending on solubility (buffers, e.g., 100 mM sodium phosphate pH 7.4–8.0). • Dissolve the reactive NHS-ester dye in a small amount of DMSO or DMF (concentration approximately 10 mM or 1 mg/100 mL). In case of high water solubility, the dye can be dissolved in buffer but should then be used immediately due to starting hydrolysis of the NHS-ester group. • Add the stock solution of dye to the antibody/protein solution to achieve the desired molar excess. For instance, 15-fold molar excess is generated by adding 10 mL of dye solution (concentration 10 mM) to 1 mL of antibody solution (concentration 1 mg/ mL, IgG molecular weight 150 kD). • Incubate for 3–24 h under gentle shaking at room temperature and under protection from light. • Remove excess dye by gel filtration, e.g., by using desalting columns. Best results are obtained when not loading the columns at the maximal possible volume, such as using a NAP10 column for not more than 0.5 mL.

21 Optical Probes

21.5.2 Labeling of Antibody with Dye Maleimide This example describes the labeling of a Cystein after reductive generation of free thiol. A mild reducing agent is, e.g., TCEP (triscarboxyethylphopshine), which reaches surface disulfides while leaving deeper bridges in the protein backbone intact. Usually, maleimide dyes can be used with less molar excess compared to NHS esters. The following steps are recommended. • Dissolve antibody or protein to be labeled in reaction buffer at concentration of 1–10 mg/mL (buffers e.g., 100 mM sodium phosphate pH 6.5–7.0). • Dissolve TCEP in reaction buffer (concentration approximately 10 mM). • Add the TCEP solution to the antibody/protein solution to achieve a molar excess of 5–10 molecules TCEP per disulfide to be reduced. • Incubate for 2 h under gentle shaking at room temperature. • Add the stock solution of dye to the freshly reduced antibody/protein solution to achieve the desired molar excess (5–10 fold). • Incubate for another 2–4 h under gentle shaking at room temperature. • Remove excess dye by gel filtration, e.g., by using desalting columns.

References Achilefu S (2004) Lighting up tumors with receptor-specific optical molecular probes. Technol Cancer Res Treat 3:393–409 Azar S, Intes X (2008) Translational multimodality optical imaging. Artech House, Boston Barnett EM, Zhang X, Maxwell D, Chang Q, Piwnica-Worms D (2009) Single-cell imaging of retinal ganglion cell apoptosis with a cell-penetrating, activatable peptide probe in an in vivo glaucoma model. Proc Natl Acad Sci USA 106:9391–9396 Becker A, Hessenius C, Licha K, Ebert B, Sukowski U, Semmler W, Wiedenmann B, Grötzinger C (2001) Receptor-targeted optical imaging of tumors with near-infrared fluorescent ligands. Nat Biotechnol 19:327–331 Bentolila LA, Ebenstein Y, Weiss S (2009) Quantum dots for in vivo small-animal imaging. J Nucl Med 50:493–496 Bremer C, Tung CH, Weissleder R (2001) In vivo molecular target assessment of matrix metalloproteinase inhibition. Nat Med 7:743–748 Fischer T, Gemeinhardt I, Wagner S, Stieglitz DV, Schnorr J, Hermann KG, Ebert B, Petzelt D, Macdonald R, Licha K, Schirner M, Krenn V, Kamradt T, Taupitz M (2006) Assessment

291 of unspecific near-infrared dyes in laser-induced fluorescence imaging of experimental arthritis. Acad Radiol 13:4–13 Foster AE, Kwon S, Ke S, Lu A, Eldin K, Sevick-Muraca E, Rooney CM (2008) In vivo fluorescent optical imaging of cytotoxic T lymphocyte migration using IRDye800CW near-infrared dye. Appl Opt 47:5944–5952 Hermanson GT (1996) Bioconjugate techniques. Academic, San Diego Klohs J, Steinbrink J, Bourayou R, Mueller S, Cordell R, Licha K, Schirner M, Dirnagl U, Lindauer U, Wunder A (2009) Near-infrared fluorescence imaging with fluorescently labeled albumin: a novel method for non-invasive optical imaging of blood-brain barrier impairment after focal cerebral ischemia in mice. J Neurosci Methods 180(1):126–32 Lee SB, Hassan M, Fisher R, Chertov O, Chernomordik V, KramerMarek G, Gandjbakhche A, Capala J (2008) Affibody molecules for in vivo characterization of HER2-positive tumors by near-infrared imaging. Clin Cancer Res 14:3840–3849 Licha K, Schirner M, Henry G (2008) Optical agents. In: Schwaiger M (ed) Handbook of experimental pharmacology, vol 185, part 2. Springer, Heidelberg, pp 203–222 Licha K, Olbrich C (2005) Optical imaging in drug discovery and diagnostic applications. Adv Drug Delivery Rev 57:1087–1108 Mier W, Beijer B, Graham K, Hull WE (2002) Fluorescent somatostatin receptor probes for the intraoperative detection of tumor tissue with long-wavelength visible light. Bioorg Med Chem 10:2543–2552 Moon WK, Lin Y, O’Loughlin T, Tang Y, Kim DE, Weissleder R, Tung CH (2003) Enhanced tumor detection using a folate receptor-targeted near-infrared fluorochrome conjugate. Bioconjug Chem 14:539–545 Pauli J, Vag T, Haag R, Spieles M, Wenzel M, Kaiser WA, Resch-Genger U, Hilger I (2009) An in vitro characterization study of new near infrared dyes for molecular imaging. Eur J Med Chem 44:3496–3503 Quadir QA, Radowski MR, Kratz F, Licha K, Hauff P, Haag R (2008) Dendritic multishell architectures for drug and dye transport. J Control Release 132:289–294 Rasmussen JC, Tan IC, Marshall MV, Fife CE, Sevick-Muraca EM (2009) Lymphatic imaging in humans with near-infrared fluorescence. Curr Opin Biotechnol 20:74–82 Resch-Genger U, Grabolle M, Cavaliere-Jaricot S, Nitschke R, Nann T (2008) Quantum dots versus organic dyes as fluorescent labels. Nat Methods 9:763–775 Richards G, Soubrane G, Yanuzzi L (1998) Fluorescein and ICG angiography. Thieme, Germany Stefflova K, Chen J, Zheng G (2007) Using molecular beacons for cancer imaging and treatment. Front Biosci 12:4709–4721 Tung CH (2004) Fluorescent peptide probes for in vivo diagnostic imaging. Biopolymers 76:391–403 Ulrich G, Ziessel R, Harriman A (2008) The chemistry of fluorescent bodipy dyes: versatility unsurpassed. Angew Chem Int Ed 47:1184–1201 Wall A, Persigehl T, Hauff P, Licha K, Schirner M, Müller S, von Wallbrunn A, Matuszewski L, Heindel W, Bremer C (2008) Differentiation of angiogenic burden in human cancer xenografts using a perfusion-type optical contrast agent (SIDAG). Breast Cancer Res 10:R23 Weissleder R, Tung CH, Mahmood U, Bogdanov A Jr (1999) In vivo imaging of tumors with protease-activated near-infrared fluorescent probes. Nat Biotechnol 17:375–378

Multi-modal Imaging and Image Fusion

22

Vesna Sossi

Contents 22.1 Introduction........................................................ 293 22.2 Nuclear Imaging (PET/SPECT) and CT......... 295 22.2.1 Beyond Anatomy and Function: Impact on Data Quantification.............................................. 296 22.3 Nuclear Imaging/Optical................................... 299 22.4 PET/MRI............................................................ 299 22.4.1 Technical Challenges........................................... 300 22.4.2 Quantification Aspects......................................... 305 22.4.3 PET/MRI: Beyond Function and Anatomy......... 305 22.5 Nuclear Medicine/Ultrasound........................... 309 22.6 Interaction Between Multi-Modality Imaging on Imaging Probe Design................... 309 22.7 Image Fusion...................................................... 310 22.7.1 Image Registration............................................... 312 22.7.2 Image Fusion........................................................ 312 22.8 Conclusion.......................................................... 312 References............................................................................ 312

V. Sossi Department of Physics and Astronomy, University of British Columbia, Vancouver, BC, Canada e-mail: [email protected]

22.1 Introduction In vivo imaging in small animals is playing an increasingly important role in the understanding of basic physiological processes, in the study of disease and evaluation of therapies; in addition to providing unique information in its own right, it also provides an es sential bridge between invasive techniques, which often require postmortem analysis, and noninvasive approaches that are applicable to studies in human subjects. A large number of in vivo imaging techniques have become available in the last decades including optical bioluminescence/fluorescence, single photon emission computed tomography (SPECT), positron emission tomography (PET), computed tomography (CT) and ultrasound (US), magnetic resonance imaging (MRI) in its various flavors, such as magnetic resonance spectroscopy (MRS), diffusion tensor imaging (DTI), functional MRI (fMRI), and others. These imaging approaches differ in both, characteristics of the imaging probes and detection techniques. An imaging probe can be thought of as a biologically relevant signal particularly suitable for detection by a specific imaging technique. The probe can be either endogenously present (for example, tissue electron density investigated with CT) or administered exogenously. Probes belonging to the latter category may serve to either enhance a signal suitable for an imaging modality (for example, CT contrast agents) or to generate a signal (for example, radioactive probes in PET/ SPECT). In general, all imaging techniques use either electromagnetic radiation of different wavelengths and intensity and different emission/detection combinations (for example, emission and detection of radio waves in MRI, X-rays in CT, and detection only of annihilation photons originating from positron decay

F. Kiessling and B.J. Pichler (eds.), Small Animal Imaging, DOI: 10.1007/978-3-642-12945-2_22, © Springer-Verlag Berlin Heidelberg 2011

293

294

in PET) or emission/detection of high frequency sound waves in US. Although in general, the instrumentation is specialized at detecting a particular characteristic of the probe, the biological relevance of the probe may be related to some of its other characteristics: for example, in PET/SPECT, the number of useful positron or gamma emitters is rather limited; however, the number of biologically relevant molecules that can be tagged with each positron emitter is large. Since the queried biological function is related to the specific molecule as opposed to the specific radionuclide, many different probes can be created for the same imaging modality. Considering the large number of probes that can be used with different imaging modalities, it becomes apparent that there is a wealth of imaging approaches available for in vivo imaging. At first glance, such wealth could be considered a great asset. Unfortunately, part of the reason for its existence is the fact that each imaging modality only operates within a relatively narrow set of parameters. Spatial and temporal resolution, detection sensitivity, tissue penetration, and signal-to-noise ratio vary greatly between imaging modalities rendering them better or worse suited for different applications. The suitability depends mostly on the match or mismatch between instrumentation performance, probe characteristics, and the properties of the tissue function under investigation. For example, an important requirement is the match between the temporal resolution of the imaging instrumentation and the timescale of the process under investigation. For instance, fMRI is better suited than 15 O-water PET to explore brain function-related activation due its much higher temporal resolution. In addition to considering the technical limitations of each modality, it is also important to remember that different imaging modalities are naturally best suited at providing fundamentally different information. For example, PET can easily provide information on tissue metabolism, while CT is able to provide exquisite anatomical details. Given their often intrinsically complementary nature, the knowledge derived from two imaging modalities can enhance the reliability of the interpretation of the images obtained from a single one. As each modality is technically maturing in its own right, the existing and potential synergy between them becomes more apparent, and the development of methods to seamlessly combine information from different modalities is gaining an increasing amount of interest.

V. Sossi

Although many combinations might prove useful, the combination of functional and structural information is often the most crucial. Nowhere is this more evident than in cancer research: the attempt to design increasingly more selective tumor-related markers means that there is less and less anatomical information available to correctly locate the tumor with respect to anatomical landmarks. A combination of a modality best suited to provide structural information (MRI, CT, and US) and one capable of providing functional information (for example, PET, SPECT, fMRI, and MRS) is thus most desirable. The traditional solution to combining anatomical and functional information consisted of performing sequential scans with different imaging devices and co-registering the resulting images. This approach, denoted as the software approach, was soon found to be suboptimal; while co-registration was relatively easily feasible when imaging rigid body objects, such as the brain, the task became daunting when trying to co-register whole body images. Different subject supports led to different apparent body shapes, and involuntary visceral movement further increased the degree of nonlinearity in the correspondence between the two images, thus rendering accurate co-registration difficult (Townsend 2008). Recognizing the limited success of software approaches, the next step was to build systems that would be capable of integration of dual imaging modalities. The first such system was built by Hasegawa and his colleagues at the University of California, San Francisco in the early 1990s (Hasegawa et al. 2002a). This system provided a truly simultaneous X-ray/ SPECT imaging but was not pursued further due to technical difficulties. It did however mark a milestone in the field of multi-modality imaging and paved the road to the development of systems where two imaging modalities are performed sequentially, sharing a common bed, and finally to truly hybrid techniques that are capable of acquiring data simultaneously. A guiding principle in the evolution of multi-modality imaging is the requirement that the performance of each individual imaging component should not be significantly compromised when operating in a multimodality mode. The rest of this chapter will be devoted to the description of the development of multi-modality imaging as relevant to small animals and latest advances in hybrid imaging. The term multi-modality

22 Multi-modal Imaging and Image Fusion

will be used when addressing systems where imaging is performed sequentially while sharing a common support mechanism, while the term hybrid imaging will be reserved for different modality imaging devices capable of truly simultaneous data acquisition. While several combinations will be described, those receiving most attention are the nuclear imaging/CT combinations and especially the currently emerging hybrid PET/MR. The last section will discuss image coregistration and fusion, which is still required both, in multi-modality imaging and to a lesser degree in hybrid imaging.

22.2 Nuclear Imaging (PET/SPECT) and CT As mentioned earlier, the concept of simultaneous SPECT/CT was developed first in human studies by Hasegawa and his colleagues in the early 1990 for human whole body scanning (Hasegawa et al. 2002a, b). The design was based on a linear array of high purity germanium (HPGe) detectors that were used to simultaneously detect the 100–200 keV emission gamma rays originating from an injected SPECT tracer and the transmitted X-rays originating from a low-power 120 kVp X-ray tube. The limitation of the system resided in the fact that the detector operated in pulse mode, thus seriously limiting the possible X-ray flux (Hasegawa et al. 1991). The system was used primarily for cardiovascular and oncology studies, but it was never developed commercially due to high cost of the HPGe detectors. In spite of the limitations, these studies clearly demonstrated the benefits and potential of a combined system leading to further development of the concept of combined imaging. To overcome the inability of a hybrid system to measure higher X-ray fluxes, the hybrid design was abandoned in favor of a multi-modality approach, where the SPECT and X-ray imaging devices could be separately optimized and placed adjacent to each other while sharing a common bed. This approach was subsequently adopted by commercial designs both in human and small animal imaging. Almost simultaneously to the SPECT/CT development, Townsend and his colleagues introduced the idea of a combined PET/CT. The first PET/CT prototype used a rotating PET scanner (ECAT ART; CPS

295

Innovations) and a single slice spiral CT scanner (SOMATOM AR.SP, Siemens Medical Solutions) placed adjacent to each other (Beyer et al. 2000). Approximately 300 patient studies were sufficient to demonstrate the clinical benefit of this multi-modality imaging approach (Kluetz et al. 2000). The recognition of the synergy provided by a combination of an anatomically and a functionally oriented modality prompted similar developments in preclinical imaging. Several research and commercial SPECT/CT are now available, based on either pinhole or parallel-hole collimation, low-power X-ray tubes (40–80 kVp), and digital X-ray detectors (Franc et al. 2008); and an example of a fused SPECT/CT image is shown in Fig. 22.1. Dedicated commercial PET/CT preclinical systems, capable of approximately 1.5 mm isotropic resolution for PET and a 50–100 mm resolution for CT, are also offered by several manufacturers (Fig. 22.2) (Pichler et al. 2008a). The comparison between relative advantages and disadvantages of PET/CT compared to SPECT/CT follows the same arguments encountered when comparing PET and SPECT; they are optimized for different radionuclides, and thus tracers, and preclinical SPECT is generally characterized by higher resolution (~0.3 mm compared to ~1.5 mm) while PET is capable of much higher sensitivity (~5% compared to <1%). Multi-modality PET/SPECT/CT systems are now also available. The development of a hybrid preclinical either PET/CT or SPECT/CT system is still hampered by the same problems faced in the first human prototype, i.e., the requirement of detectors and electronics that would be capable to simultaneously measure high X-ray fluxes and count gamma rays. Recent advances in single photon counting X-ray CT, together with developments of highly pixelated detectors designs with parallel processing electronics, might eventually lead to a hybrid design (Brunner et al. 2009; Berard et al. 2007). Attempts to develop hybrid PET/SPECT/ CT systems have also been explored (Saoudi and Lecomte 1999), but no full systems are currently available. While it is reasonable to assert that the development of nuclear medicine/CT combination revolutionized clinical practice for many diagnostics applications, the same degree of success was not achieved in small animal imaging. Of particular concern is the fact that in small animal imaging, a relatively high radiation dose

296 Fig. 22.1 Example of a SPECT (111Indium labeled antibody)/CT image: (a) CT only; (b) SPECT only; (c) fused SPECT/CT indicating the improved anatomical localization of the SPECT tracer uptake

V. Sossi

a

b

c

must be administered during a CT scan to obtain sufficient tissue contrast. For example, for a whole body mouse CT scan with a spatial resolution of 50m, the dose per scan is approximately 0.6 Gy, which is approximately 5% of the LD 50 (lethal dose for 50%) value for mice (Weissleder 2002). There are studies that show that even X-ray doses below 1 Gy can influence the expression of cytokines such as IL-10 or IL-12 in animal models (Liu et al. 2003). This is of particular relevance when longitudinal studies are performed on the same animal that would expose it to multiple CT scanning sessions.

22.2.1 Beyond Anatomy and Function: Impact on Data Quantification 22.2.1.1 Attenuation Correction One of the main advantages of nuclear medicine-imaging techniques, especially PET, is the ability to provide quantitative assessments of radioactivity concentrations present in the imaged subjects. Several corrections must be applied to the data before accurate quantification is achieved; an important one is the

297

22 Multi-modal Imaging and Image Fusion

correction for gamma interaction in tissue, which leads to a reduced number of measured counts or signal attenuation. Most correction methods rely on the measurement of the tissue linear attenuation coefficients, which are generally calculated by performing a transmission scan with an external source. In the early days, the energy of the radiation produced by the external

Fig. 22.2 Example of a commercial PET/CT system with CT docking on the PET unit. Image courtesy of Siemens Medical Solutions USA, Inc

a

Fig. 22.3 Linear attenuation coefficient (or m-map) image with; (a) 68Ge transmission source (rat); (b)a 57Co transmission source (rat); (c) a CT (whole body mouse). All images were obtained with standard protocol settings, and the difference in the statistical quality of the images is clearly visible. Note that the mouse was scanned in a different orientation

b

source used to closely match the energy of the radiation emitted by the tracer under investigation. In PET, this approach produced fairly noisy transmission images for small animals: the attenuation length of 511 keV photons in tissue (~10 cm) is larger than the size of most animals, and therefore, only a relatively low number of photons interact with the tissue. This problem was minimized by the introduction of a 57Co transmission source (gamma energy 122.1 keV 85.6% and 136.5 keV (10.7%), attenuation length in tissue ~6.4 cm), and the improvement observed in the linear attenuation coefficient image quality is clearly visible in Fig. 22.3. Tissue segmentation algorithms are employed to define different tissue types, and corresponding energy scaling algorithms are used to correct the linear attenuation coefficient to its value corresponding to the energy of the detected radiation (Townsend 2008). Given the good quality of 57Co-based transmission data, the use of CT for the same purpose is less crucial compared to human scanning. However, the use of CT-derived attenuation map has been proven advantageous in SPECT (Franc et al. 2008) and when compared to using a 511 keV gamma producing transmission source (typically 68Ge) (Chow et al. 2005).

c

298

V. Sossi

22.2.1.2 Partial Volume Effect Another factor limiting quantification is the partial volume effect (PVE). PVE is due to two phenomena that alter image intensity values compared to the true values of the corresponding observables. The first effect is due to the image blurring introduced by the finite size of the detector; any object that is smaller than approximately twice the size of the imaging resolution will appear blurred. The second reason is the fact that an image voxel might include different tissue types and the resulting signal is not single tissue specific. This is generally called the tissue fraction effect. It is important to realize that PVE can lead to an under- or over-estimate of a signal as seen in Fig. 22.4. PVE-introduced inaccuracies can be especially severe when using PET to image small structures in mice, where organs of interest might measure 1 mm or less. Several partial volume correction algorithms have

been developed exhibiting various degrees of complexity (Soret et al. 2007). The simplest correction method is based on the use of premeasured recovery coefficients (RC); typically spherical phantoms of different size are imaged and the RC correction factors are determined by comparing the measured signal to the independently measured radioactivity concentration. Although simple, this method requires a priori knowledge of the object size and suffers from the limitation that the shape is always assumed to be well described by a sphere. More refined methods, which are able to correctly account for the size of the structure and the tissue fraction effect, require a very well co-registered structural image (Soret et al. 2007). Since the PVE correction factors can be as large as 10, it is very important to ensure both a correct image co-registration and tissue definition; both are aided by the availability of a CT image acquired in a multimodality environment.

a

b 1

1

0.5

0.5

0 5

0 5

5

0

5

0

0

0

–5 –5

–5 –5

c 1 0.8 0.6 0.4 0.2 0 –5

Fig. 22.4 Illustration of the partial volume effect. In this case, the diameter of the object was taken to be equal to twice the resolution (FWHM) of the system; (a) original object, (b) imaged

0

5

object, and (c) one dimensional profile through the imaged object. The arrows indicate the regions where the radioactivity concentration is underestimated and overestimated (spill-over)

22 Multi-modal Imaging and Image Fusion

299

22.3 Nuclear Imaging/Optical

most studies are still performed sequentially on different imaging devices. There are however ongoing research efforts to develop an optical/PET system, which are expected to increase as more multi-modality probes are being developed (Alexandrakis et al. 2006).

Nuclear imaging and optical multi-modality imaging fall in a slightly different category compared to the combinations described earlier, since none of them excel at providing structural information. The nature of their complementary resides in a different area: while both techniques have very high detection sensitivity, they differ greatly in spatial resolution and penetration depth (Culver et al. 2008; Ntziachristos et al. 2005). Optical imaging methods are characterized by very high spatial resolution and low penetration depth, while nuclear medicine techniques have relatively low spatial resolution and much higher penetration depth. Another complementary aspect is the fact that optical techniques generally do not yield quantitative measurements, unlike nuclear medicine-imaging methods. The combination of the two modalities thus permits to investigate tissue characteristics and disease at different spatial and temporal scales. For example, nuclear imaging could be used to localize disease in tissue, and optical techniques can be used to validate biopsied tissue. Likewise, optical signals can be detected in tissue for several weeks after administration of the imaging agent, unlike nuclear medicine techniques, where the duration of the signal is limited by the radiotracer half life. In order to take advantage of this type of multimodality imaging, it is important to develop multimodality imaging probes. This is a very vibrant area of research; an interesting approach is to use quantum dots and encapsulate them in functionalized phospholipid micelles and covalently label them with 18F-fluorine (Duconge et al. 2008); the kinetics and whole body distribution can be then evaluated by PET, while their cellular behavior can be explored with optical imaging techniques. Semiconductor nanocrystals, or “quantum dots” (Murray et al. 2000), are potentially excellent fluorescent probes for in vivo imaging due to their intrinsic optical properties, such as high quantum yields, excellent resistance to photobleaching, broad excitation spectra, and narrow emission bands tunable from the UV to the NIR regions (Michalet et al. 2005). The ability of imaging with such a dual modality probe presents two advantages: a biologically relevant question can be addressed at a multi-scale level, and the characterization of a new probe can likewise be done at several levels. Multi-modality nuclear medicine/optical imaging system development is currently still in its infancy, and

22.4 PET/MRI The idea of developing hybrid PET and MRI systems started in the late 1990s. It first originated as a tool to improve the spatial resolution of the PET scanners. The positron range, a limiting factor in PET resolution, decreases as the positrons are exposed to a magnetic field (Wirrwar et al. 1997). One of the first efforts was based on coupling scintillation crystals to optical fibers (Shao et al. 1997a, b) and then inserting them into a 1.5 T MR system. The use of optical fibers was dictated by the need for placing the photomultiplier tubes (PMTs) outside the magnetic field, which adversely affects their performance, since they rely on electric fields for signal amplification. The prototype system was able to acquire images, but presented two drawbacks: low PET detection sensitivity (0.03%) due to the low number of crystals and light loss in the optical fibers (Catana et al. 2006; Marsden et al. 2002). Almost simultaneously another group was pursuing a different approach (Pichler et al. 1998a); they used a novel Avalanche photodiode (APD)-based PET detector and inserted it into a 9.4 T animal MRI. This approach is to date considered to be the most versatile. Further progress in the development of hybrid PET/ MRI systems was stimulated by the great success of combined functional/anatomical information provided by PET/CT together with the recognitions that for small animal imaging, there were additional reasons to prefer MRI over CT: the soft tissue contrast achievable with MRI is much superior (Fig. 22.5), and there is no additional radiation exposure (particularly relevant for longitudinal studies). MRI is capable of additional functional imaging, for instance perfusion, blood tissue oxygenation level (BOLD) effect, spectroscopy, diffusion, and others. A combined PET/ MRI thus offers a very diverse and comprehensive imaging environment where several parameters can be imaged at once with, in principle, perfect image

300 Fig. 22.5 Whole body CT (a) and MRI (b) image of a mouse

V. Sossi

a

co-registration. Several technical challenges, however, needed to be addressed before a reliable system could be developed.

22.4.1 Technical Challenges 22.4.1.1 Mutual Interference Between PET and MRI When developing a hybrid imaging device, it is important to preserve the full functionality of each and therefore minimize possible mutual interference. There are several reasons whereby PET and MRI can negatively influence each other: the high main and gradient magnetic fields and radiofrequency needed in MRI can alter PET signal detection; the electronics and materials used in PET instrumentation can degrade the MRI performance with radiofrequency noise or magnetic inhomogeneities present in the detector housing or other material. The eddy currents induced from the gradient system in the conducting structures of the PET imaging elements can further contribute to mutual interference by reducing the uniformity of the magnetic fields, which is critical for accurate MRI.

b

These collective requirements have an impact on both the geometry of the design of the hybrid system and the selection of the PET components. At present, there are three main design approaches (Fig. 22.6): (i) a PET insert placed into the main magnet. This design works best with the RF generating coils placed in between the object and the insert itself to optimize the MRI signal (Judenhofer et al. 2007); (ii) a hybrid approach, which uses a PET insert in the magnetic field and short optical fibers (~12 cm) linking the PET insert to the analog PET electronics placed outside the magnetic field (Catana et al. 2006); and (iii) PET detectors placed into two halves of an MRI magnet, thus allowing the use of nearly standard PET equipment (Lucas et al. 2006). In addition, a very recent design is proposing a combination of a standard PET scanner placed adjacent to a field cycle MRI system (thus not leading to a truly hybrid system) (A., B.G.A.S.T.J.H.W.B.C.B 2009). The first approach is in principle the most versatile and capable of highest performance; the hybrid approach does not easily allow for an arbitrarily large PET field-of-view (FOV); in order to minimize light attenuation in the fibers, their length must be kept relatively short. In the latter hybrid design, the maximum strength of the magnetic field is limited to approximately 1 T. A PET insert and

301

22 Multi-modal Imaging and Image Fusion Fig. 22.6 Three current approaches to a hybrid PET/ MR system. (a) PET system in the magnet; (b) hybrid design, where the length of the optical fibers is limited to ~10 cm; (c) split magnet design

b

a

Light fiber Scintillator

APD

APD

Scintillator

Electronics

Electronics

Magnet

Magnet

PMT

c Light fiber

Scintillator

SplitMagnet

related electronic placed entirely in the FOV in principle does not suffer from any of these shortcomings, but it is also the most prone to signal interference between the two imaging modalities: new instrumentation development and careful component selection (Fig. 22.7) are required to minimize image distortion.

22.4.1.2 PET Crystal Selection With the PET insert placed into the magnetic field, it is important to ensure that the scintillating crystal exhibits

magnetic susceptibilities close to those of human tissue to minimize MR artifacts. Thallium doped sodium iodide [NaI(Tl)], thallium doped caesium iodide [CsI(Tl)], bismuth germinate (BGO), and cerium-doped lutetium oxyorthosilicate [LSO(Ce)] satisfy these requirements, unlike two other scintillators also commonly used in PET imaging, germanate oxyorthosilicate (GSO) and lutetium germanate oxyorthosilicate (LGSO) (Yamamoto et al. 2003). Of these, LSO(Ce) is currently considered the scintillator with optimal detection characteristics for PET, thus rendering it also the preferred choice for a hybrid PET/MRI system.

302 Fig. 22.7 Integration of the multi-slice PET scanner into a 7-T MRI apparatus. (a) Drawing of PET–MRI combination showing the PET insert placed inside the MRI scanner, matching the centers of both fields of view. (b) Photograph of the MRI-compatible PET insert consisting of ten radially arranged detector modules. (c) Single PET detector module showing the LSO scintillator block, APD array, and preamplifier built into an MRI-compatible copper shielding. Reprinted with permission from (Judenhofer et al. 2008)

V. Sossi

a

22.4.1.3 PET Signal Amplification and Electronics The main challenge faced when placing the entire PET detection system in the MR magnetic fields is the development of a reliable amplification of the signal generated by the detected radiation in the crystals. In traditional PET, PMTs are most commonly used due to their high gain, good energy linearity, and stability. However, due to their intrinsic mode of operation, they are not suitable for use in a magnetic field (Fig. 22.8). An alternative detector for PET, which is capable of detecting the light originating from the scintillator material, is the APD (Lecomte et al. 1990, 2001; Pichler et al. 1998b), a semiconductor-based light detecting device. The APDs are thin, compact, and resistant to the effects of the magnetic field, since they have a high internal electric field and the charge carriers only travel for a short distance. In contrast to PMTbased signal amplification, the amplification process in APD is affected by an excess statistical noise factor that increases with the gain (Webb et al. 1974). A multiplication gain of up to 103 can be obtained, but due to the excessive noise factor, they tend to be operated in a multiplication range of approximately 50–150. They have been proven to operate in high magnetic fields (above 9 T) without compromising their per

b

c

formance (Pichler et al. 1998a). Since their size is small, a large number of them can fit in the bore of the magnet, and thus, they do not become a limiting factor for the extent of the scanner axial FOV. These features not only promoted the use of the APDs in the development of a preclinical hybrid imaging device, but also enabled the development of a first APD-based PET/ MR prototype for human brain imaging (Schlemmer et al. 2008). An interesting further potential improvement for gamma detection are APDs with the development of Geiger-mode diodes (G-APDs, or silicon photomultipliers (siPMs)) (Otte et al. 2005). While maintaining most of the positive characteristics of APDs, including insensitivity to the presence of a magnetic field, they are also capable of much higher signal gains, reaching multiplicative factors up to 106. This is achieved by subdividing the active surface of an siPM into very small cells (of the order of mm), which are operated in Geiger mode. Individually, they do not provide a linear response as a function of signal energy; overall linearity is maintained by ensuring that the number of cells in each siPM is much larger than the number of ex pected photons (Pichler et al. 2008b). Just as important as the signal amplification instrumentation is a careful design of electronic circuit boards; all their components such as capacitors, amplifiers, and

303

22 Multi-modal Imaging and Image Fusion APD PET/MR

Conventional PET-Detectors

B=0

Bπ0

B=0

Bπ0

Fig. 22.8 Crystal position map distortion observable when the PMTs are placed in the magnetic field; no such distortions are present when APDs are placed in the magnetic field. Reprinted with permission from (Pichler et al. 2008c)

resistors must be nonmagnetic, and large solid areas of conductive material must be avoided.

22.4.1.4 PET Detector Shielding and System Stability The presence of any conducting structures in the PET insert can cause RF coil detuning, with consequent distortion of the MR image. The use of externally shielded coils can reduce the risk of inducing currents caused by the RF radiation in the conductive structure of the PET insert. Since such currents are larger in closed loop conductors, any close metallic loops must be avoided in the PET/MR device. The switching of the MR gradients can also cause eddy currents in the PET instrumentation, which can linger during MR readout, thus leading to image distortion. Proper shielding of the PET insert is thus of utmost importance: it should be thin and have nonconductive materials interleaved with slit conductive layers. Another potential source of mutual interference is temperature variation, which can arise either because of suboptimal temperature stabilization protocols or

by eddy currents induced in the PET insert. A temperature drift will alter the PET signal, since the radiation detection sensitivity of most scintillator materials and certainly of the APDs varies as a function of temperature. Temperature variation will also introduce distortions in the MR signal and ultimately image. It is thus essential to appropriately stabilize and monitor the temperature on the system.

22.4.1.5 PET/MR Imaging Space Co-Registration Since the hybrid systems acquire data with both imaging modalities simultaneously with the subject in the same physical location; a perfect co-registration between the PET and MR-derived image is theoretically possible. In order to achieve it, it is however still necessary to co-register the PET and MR image space. This is generally achieved by scanning the same phantom with both modalities and co-registering the respective images (Fig. 22.9) (Judenhofer et al. 2008). This method may be sufficiently accurate when comparing PET to anatomical MR images. Considering however that the presence of the object itself in the MR FOV may introduce

304

V. Sossi

Fig. 22.9 Phantom used to co-register the PET and the MRI image space in the PET/ MR system described in (Judenhofer et al. 2008). Reprinted with permission from (Judenhofer et al. 2008) MR

PET/MR

PET

object-dependent image space distortions during more complex MRI protocols, it would be additionally desirable to develop MR/PET compatible markers that could be imaged simultaneously with the animal and would serve as fixed landmarks when co-registering the two image spaces. An example is small tubing filled with a radioactive solution and MRI contrast agent to make it visible in both modalities (Zhang et al. 2008). The problem might be additionally compounded by object-dependent nonlocal image distortions, occurring in fast MRI sequences, which cannot be compensated by

shimming (Fig. 22.10). This problem is particularly severe when imaging small size objects and has been plaguing the MRI world for a long time (Li et al. 2008). Many different algorithms have been proposed to minimize it; it remains to be seen if the additional information provided by a simultaneous PET image can provide further information to help solve the problem. For the time being, it is important to remember that even simultaneous data acquisition, while always resulting in the best possible co-registration between PET and MR-derived images, does not always guarantee perfect image overlap.

305

22 Multi-modal Imaging and Image Fusion

a

utilizing CT and MRI data to extract relevant features in the MRI data are all being explored for human brain and whole body PET/MRI (Hofmann et al. 2009), not much is done however for preclinical PET/MR imaging. The effect of attenuation and gamma ray scatter are often considered relatively minor problems in small animal imaging; nevertheless correction for them is necessary, if absolute data quantification is desired. 22.4.2.2 Image Analysis and Partial Volume Correction

b

Fig. 22.10 Example of a nonlinear image distortion: (a) anatomical image (TSE sequence) and (b) DTI image (EPI sequence). The arrows indicate the areas where distortions are most apparent

22.4.2 Quantification Aspects 22.4.2.1 Attenuation Correction There is one potential drawback in using MRI instead of CT: MRI-derived images are less suitable to providing attenuation correction factors, since they provide information on proton density and not on photon interaction in the tissue. For example, bone, with a relatively high tissue density, has a higher attenuation factor in CT images compared to tissue, as it does in PET images. On the other hand, it has relatively low proton density, thus providing low MR signal, similar to that provided by air. The use of external sources to measure attenuation does not seem viable in PET/MR systems due to space and mechanical constraints; alternative approaches to attenuation correction must thus be developed. Methods based on image segmentation, aid of atlas-based information, training-algorithms

The excellent soft tissue contrast provided by MRI can be used to great advantage for several different applications. The theoretically perfect co-registration between the MRI and PET images allows for a much better localization of the structures on interest in the PET images, and such knowledge can be used for many different applications. For example, in typical neuroreceptor studies, it is generally very difficult to correctly identify the cerebellar region, which is often used as reference region, since generally devoid of any tracer specific binding; small variations in the regionof-interest (ROI) placement can lead to significant variations in the estimates of binding potentials (Sossi et al. 2007). The ability to place an ROI on the MRI image and confidently copy it onto the PET image not only greatly facilitates image analysis from a practical point of view, but is also expected to increase its accuracy and thus reproducibility. This is of particular importance when the magnitude of a biologically relevant effect is small, of the order of a few percent. Another important area where the combination of the theoretically perfect co-registration between the PET and MRI images and the excellent soft tissue contrast provided by MRI can improve PET quantification is in the development and application of partial volume correction methods: in particular, MRI can offer a much more accurate estimate of the tissue fraction compared to CT and thus enhance the accuracy of the PVE correction.

22.4.3 PET/MRI: Beyond Function and Anatomy MRI imaging can be used to query much more than just anatomical structure as shown in Fig. 22.11. Using

306

a

d

V. Sossi

b

c

e

Fig. 22.11 Examples of different types of MRI images: (a) ASL image of a mouse brain slice; (b) anatomical image of a rat brain; (c) DTI image overlaid on the anatomical image of a rat

brain (note the distortions); and (d, e) MRS image in a mouse brain and corresponding spectra

different sequences, MRI can measure perfusion and blood flow (arterial spin labeling – ASL), (BOLD signal), metabolic processes and products (MRS and chemical shift imaging (CSI)), diffusion, and diffusion directionality (DI and DTI). The current hybrid imaging prototypes have demonstrated that the performance of the MRI system is not degraded by the presence of the PET insert (Judenhofer et al. 2007, 2008), and consequently, the full range of MR applications can be used in conjunction with PET (Wehrl et al. 2009). Small animal imaging, where ethical restrictions are less severe than in human studies, provides an ideal environment where to explore the full potential of this hybrid imaging technique. While many applications of this technique still need to be defined, it is relatively easy to speculate about some very interesting scientific questions that can be addressed with such an instrument. In this context, it is important to remember that in most cases, neither modality provides direct information on the biological process of interest, but does so only

indirectly: PET only detects radioactivity, and complex modeling with several underlying assumptions is required to translate the temporal and spatial distribution of radioactivity into parameters bearing biological meaning, such as standard uptake values, uptake rate constants, or binding potentials. Further assumptions are required to relate these parameters to tissue function in health or disease. Likewise, MRI only relies on different magnetic properties of tissue and uses sophisticated radiofrequency sequence designs to tease out these differences: the relation between such signals and underlying function and ultimately pathology is often found empirically and requires educated assumptions for interpretation. An exciting application for the hybrid system would thus be to obtain seemingly similar information from both modalities and compare their relation to the underlying process or pathology. In addition to providing a deeper understanding of the question under investigation, this approach could also be used to validate the assumptions leading from measurement to underlying biology. Along the same

307

22 Multi-modal Imaging and Image Fusion

line of thought, PET and MR biomarkers could be compared in terms of their diagnostic utility to select the most cost-effective and less-invasive approach to be used clinically. Another category of applications may be represented by those studies, where the two imaging modalities are used to provide truly multi-functional and multiparameter information.

22.4.3.1 Examples of Related Process Dual Modality Investigation A very good example of this type of investigations is blood perfusion. Typically in PET, blood perfusion is measured with 15O-labeled water, while ASL is often employed in MRI. Recent studies in humans have shown similar changes in ASL signal and PETmeasured signal following visual stimulation and hypercapnea (Chen et al. 2008). However, no such studies have been completed in small animals, nor absolute measurements of blood perfusion have been compared between the two techniques. Another tissue parameter that can be measured with PET and MRI is hypoxia. Detection of hypoxia has therapeutic implications in tumors, stroke, and myocardial ischemia. Several PET tracers are used to measure hypoxia (for example, 18F-EF5, 18F-MISO, 64Cu-ATSM) (Krohn et al. 2008). They provide information about MRI

hypoxia on a cellular level. In MRI, hypoxia measurements can be done with BOLD by determining differences between oxyhaemoglobin (O2Hb) and deoxyhaemoglobin (dHb). A correlation between the oxygen partial pressure (PO2) and BOLD has been reported (Al-Hallaq et al. 1998); however, a given BOLD signal response may reflect different changes in PO2 (Baudelet and Gallez 2002). A simultaneous monitoring of hypoxia with PET and MRI might thus provide new insights in tissue metabolism. Finally, there is an interest in developing truly dual modality probes, that is, markers that are observable simultaneously by PET and MRI. Here, PET could be used to generally localize the probe and quantify its uptake, while MRI could provide exact localization in an anatomical context. An example of such probe is the use of a 124I serum albumin (SA)-MnMEIO for imaging the sentinel lymph node (LN) (Choi et al. 2008). Accurate imaging of the LNs is critical in cancer staging, since the presence of cancer cells in the LNs can be often taken as an indicator of metastatic disease. In this dual probe, radioactive iodide ions are combined into the innate tyrosine residue of SA and are then coated onto the surface of MnMEIO magnetic nanoparticle. The resulting images are shown in Fig. 22.12. In this particular case, the PET and MRI images were obtained sequentially; it is however easy to imagine how such studies would additionally benefit from simultaneous imaging.

PET

PET/MRI

b

c

d

e

f

high

PET signal

a

low

Fig. 22.12 (a–f) PET/MR images of sentinel lymph nodes (LN) in a rat at 1 h postinjection of 124I-SA-MnMEIO into the right forepaw (I = nanoprobe injection site); coronal (a) MR and (b) PET images in which a brachial LN (white circle) is detected; (c) PET/MR fusion image. Four small pipette tips containing

Na124I solution are used as a fiducial marker (white arrowheads) for the concordant alignment in PET/MR images; (d–f) in the transverse images, axillary (red circle) and brachial LNs (white circle) are detected. Adapted with permission from (Choi et al. 2008)

308

V. Sossi

22.4.3.2 Examples of Multi-Function Multi-Parameter Imaging Changes in blood flow induced either pharmacologically or as a consequence of injury can alter flowdependent kinetic constants measured with PET in a manner that is not relevant to the specific process of interest (Schmidt and Turkheimer 2002). In order to control for this potential confounding factor, often a PET blood flow measurement immediately before the PET measurement of interest is performed. Simultaneous ASL/PET would be of methodological interest in those cases, since it would eliminate the need of a second PET measurement and provide truly simultaneous information. Another interesting PET/MR application is the characterization of tumor proliferation and its morphological composition. Figure 22.13 shows an example of a simultaneous PET/MR study, where PET 3¢-[18F]fluro3¢deoxythymidine ([18F]FLT) was used to determine

tumor proliferation, and contrast-enhanced static and dynamic MRI were used to determine tumor morphology and viability (Judenhofer et al. 2008). This particular study was performed in a living BALB/c mouse bearing a colon carcinoma of 1.1 cm2. The areas of slower contrast agent enrichment, reflecting necrosis and inflammation (later confirmed by histological assessment), were found to correspond with the areas of lower PET tracer uptake. The information provided by the two techniques was thus found to be highly complementary. Such complementary is anticipated to be of great advantage when investigating complex characteristics of tumor models, evaluation of possible therapies, and also when investigating the suitability of possible new biomarkers. While some of the described applications can also be performed with multi-modality imaging, they would still benefit from simultaneous data acquisition. Co-registration between images is much more easily achieved, and possible artifacts introduced by d

a

b

f

c

e

g

Fig. 22.13 (a) [18F]FLT PET image of a BALB/c mouse with a CT26 colon cancer. High tracer uptake reflects cell proliferation; (b, c) pre and postcontrast T1-weighted MRI acquired simultaneously with PET. Tissue inhomogeneity can be observed; (c, d) curves with higher slopes correspond to areas showing tumor proliferation in the PET images, while flatter curves correspond

to areas with low PET tracer uptake; (e) histology confirms areas with high cell proliferation (asterisk) observed with PET; (f, g) arrows in the histology images reveal necrotic and inflamed tissue, consistent with low contrast in T1-weighted MRI and very low PET tracer uptake. Reprinted with permission from (Judenhofer et al. 2008)

309

22 Multi-modal Imaging and Image Fusion

mismatches due to breathing and visceral motion can be minimized. Furthermore, the length of time during which the animal is subject to anesthesia is minimized, thus reducing the stress to the animals. There is thus an enormous wealth of possible applications for this hybrid device in preclinical imaging, and more will emerge as the modality matures. Some of these applications will undoubtedly extend into human PET/MR research, where an additional prominent role of fMRI can be easily envisioned. At this stage, it is also possible to foresee a role of PET/MR in whole body clinical studies, mostly due to the fact that MR can provide excellent anatomical localization with no additional radiation, a serious drawback in PET/CT. Long-term stability of the system hardware, MRI-based attenuation correction, and better software tools to integrate information from the two modalities are currently areas of research and need to be achieved before PET/ MR can widely reach the clinics.

22.5 Nuclear Medicine/Ultrasound There are currently no preclinical multi-modality imaging devices combining SPECT or PET and US. US is however widely available, safe, and inexpensive and can provide high spatial resolution without ionizing radiation. These characteristics, together with emerging new contrast-material-enhanced US techniques, have recently led to an increasing amount of interest to combine the two techniques, mostly, at this stage as a validation tool for new US probes. Contrast-material-enhanced US uses targeted micro bubbles (MBs), which are gas-filled echogenic US contrast agents that can be targeted to specific molecular markers by means of attachment of appropriate ligands to the surface of the MBs (Lindner 2004). When administered intravenously, they distribute throughout the body and attach at tissue sited expressing the targeted molecular marker. Because of their relatively large size (several microns), they stay mostly in the vascular compartment and are most suitable to image events within the vascular compartment, such as inflammation and angiogenesis. By adding a radioactive tag to such MB and image it with PET or SPECT, it becomes possible to investigate their whole body distribution and kinetics. This is of great relevance, since in order to obtain a reliable US signal, it is important to differentiate the

specific molecular marker from a background signal. The knowledge of the whole body distribution can thus help in the evaluation of the suitability and specificity of an MB as a specific molecular marker. An example is provided by a recent study where such a dual modality probe was developed. US was explored as a tool to monitor tumor angiogenesis by using MB targeted to integrin anb3 integrin, endoglin, and vascular endothelial growth factor (VEGF) receptor 2 (VEGFR2), which are involved in regulation of angiogenesis. These MBs were subsequently radiolabeled with 18F to investigate their whole body distribution and kinetics. The study clearly demonstrated the feasibility and usefulness of this type of imaging (Willmann et al. 2008).

22.6 Interaction Between Multi-Modality Imaging on Imaging Probe Design In order to fully appreciate the impact of multi-modality and hybrid instrumentation development, it is necessary to place it into the context of the revolutionary progress occurring in the development of multi-modality imaging probes. Both aspects of imaging, signal creation (probes), and detection (imaging instrumentation) must proceed together to further the field of in vivo imaging, especially in the preclinical world. Advances in nanotechnology and other sciences are providing multi-modality molecular-targeted imaging probes, such as multiple targeting motifs, including peptides, proteins, and small molecules that provide enhanced binding and selectivity. Novel metals or biopolymers are being developed, which improve physiochemical properties and delivery efficiency. There are efforts to develop new probes for virtually any combination of imaging modalities. In addition to those mentioned in the previous sections, there is increasing interest to developed optical/MRI probes consisting of magnetofluorescent nanoparticles used, for example, in the preclinical investigation of myocardial disease (Lee and Chen 2009). The combination of various radioisotopes, fluorophores, novel metals, small molecules, and various nanoparticles has significantly widened the range of the observable parameters in the spatial and temporal domains, thus creating an optimal environment for the synergy of multi-modality imaging.

310

22.7 Image Fusion The display of images acquired from multi-modality imaging is very important to extract visual and quantitative information from the acquired data. Two steps are required to do this accurately: image registration and visually informative image display. Often, the term “fusion” is used for both steps; here, it will be limited to the process of overlaying co-registered images so as to maximize their synergistic information. The amount of potential displacement between images decreases dramatically with simultaneous data acquisition, but even in this optimal situation where the subject is in the same physical position, its location in the image space of each imaging modality might be different, as discussed above. A brief overview of image registration concepts and methods will be presented; they have been primarily developed for human imaging, but they are equally well applicable in preclinical imaging. More details can be found in (Hawkes et al. 2003).

22.7.1 Image Registration The fundamental concept in image registration is to determine spatial correspondences. Depending on the type of analysis and information needed, image co-registration can establish correspondences between (i) images of the same subject – intra-subject co-registration – and (ii) images of different subjects – inter-subject co-registration. Co-registration methods can be further used to establish correspondences between images acquired with (i) the same imaging modality (single modality co-registration) or (ii) with different imaging modalities (multi-modality co-registration). The approach and the algorithms used in each case might be different. In the context of multi-modality imaging, intra-subject, multi-modality co-registration is most often needed. There are several steps where choices must be made in image registration: (i) type of registration, broadly divided into two categories: rigid body, with or without scaling, affine, or nonrigid body where more degrees of freedom are allowed; (ii) definition of a similarity measure and method used to optimize it, and (iii) transformation of the source image (image to be aligned) to the target image (image used as reference).

V. Sossi

22.7.1.1 Type of Registration Rigid body transformations can be used if it is reasonable to assume that the shape of the imaged structure does not change between images as is often the case when performing brain images on the same subject. Exceptions might be nonlinear image distortions introduced by some MRI sequences. In general, a rigid body transformation is fully defined by six parameters, rotations around the three Cartesian axes, and translations along them. A more general rigid body transformation is the affine transformation, which includes shear, and increases the number of free parameters to twelve. Any affine transformation is characterized by the fact that parallel lines are always transformed into parallel lines. When performing whole body imaging, the spatial relationship between organs most often changes due to breathing and cardiac and visceral motion. Nonrigid body transformation with a variable number of degrees of freedom is required to co-register such images. Basis functions, in particular B-splines or shape descriptors are often used to describe the transformation (Rueckert et al. 1999). A particularly difficult problem emerges when not only the spatial relation between any two points in the image changes, but also when the shapes of different structures of interest change; it must be then decided whether a global or a local nonrigid body transformation is to be used. Unfortunately, each nonrigid body transformation produces slightly different results, and accurate validation methods are still under development.

22.7.1.2 Similarity Measures The choice of similarity measures is one of the defining factors in establishing an algorithm. The intrinsic nature of the imaging modalities producing the images to be co-registered (for example, MRI tends to produce images with very high definition, while images obtained in PET/SPECT tend to be statistically noisy with blurred edges), potential accuracy of the outcomes, and practicality of implementation are important considerations when choosing a similarity measure. They can be broadly divided into three categories: landmark measures, surface and edge measures, and voxel intensity measures.

311

22 Multi-modal Imaging and Image Fusion

Landmark measures. Landmark measures use external markers that are visible in the two images to be co-registered (Maurer et al. 1997; West et al. 1997). The landmarks must be placed in a fixed relationship with the structures that are being imaged; this requirement is often difficult to satisfy; for example, skin might move independently from internal organs, and any external fixture must be secured in such a way as to minimize the probability of it moving independently. Furthermore, limitations in the size of the imaging device FOV might limit the choice of possible useful markers. Nevertheless the use of external landmarks has the great advantage of providing absolute reference points in the two images. In the case of multimodality imaging, the landmarks must have properties that make them visible in both (or more) modalities. Surface and edge markers. This technique attempts to minimize the average distance between corresponding surfaces or edges, usually preidentified by either automated or user-defined segmentation (Pelizzari et al. 1989; Pietrzyk et al. 1994). While originally designed for brain studies, this method was later adapted to other applications. This method however works best when high resolution images are available and is not very well suited for registration of images with very poor intrinsic landmarks, such as, for example, functional images with highly localized contrast. Voxel intensity measures. Since these methods operate directly on voxel intensities, different approaches must be used for alignment of intra- or inter-modality images. A number of measures are thus possible; (i) minimization of intensity difference, which minimizes the sum of the absolute or square intensity difference between two images (Eberl et al. 1996). This technique assumes that images differ only by noise and is only suitable for intra-modality co-registration; (ii) correlation of voxel intensities assumes that the images are strongly correlated, again best suited for intra-modality images; and (iii) minimum variance; this approach comes in two main versions. In the first version, a voxel-by-voxel ratio of the two images is performed, and the variance of the ratio is minimized (Woods et al. 1992). This is particularly suitable for intra-modality registration. In the second approach, initially developed for PET to MRI co-registration, it is assumed that voxels with similar MRI signal intensity will correspond to the same tissue type and thus have the same PET signal (Woods et al. 1993). Minimization of the

variance of the voxel values classified to belong to the same tissue is used to define the optimum transformation. Voxel intensity histograms. This class of similarity measures relies on the co-occurrence of specific intensity pairs regardless of their location. In this approach, a joint intensity histogram is formed from a pair of images. A correct realignment between images will reduce the dispersion in the joint histogram. The amount of dispersion can be analyzed either by variance reduction or by using information theory. Maximization of the mutual information approach, which seems one of the most successful similarity measures for inter-modality registration, belongs to the latter category and is based on the statement that when two images are correctly realigned (minimum distortion in the joint histogram), the image intensity in one image provides maximum information about the image intensity in the other image (Maes et al. 1997; Hill et al. 2001). The determination of the optimal transformation generally proceeds using optimization algorithms, such as simplex, conjugate gradient, and Levenberg Marquardt algorithms (Press et al. 1993). A potential pitfall is the encounter of a local optimum, and multiresolution methods (Hawkes et al. 1997) can be used to minimize this problem. In these approaches, an initial estimate of the transformation is achieved by using a coarse image representation, which is then refined in subsequent iterations.

22.7.1.3 Transformation Methods When registering two images, it might occur that only part of the image is of interest. For example, in small animal brain imaging, it is only the brain that is of interest and not the entire head. In order to achieve the best possible registration between two brain images, it is desirable to define a mask around the region of interest. Using this approach, simpler realignment methods might be appropriate: for example, rigid body transformations can be used for the brain, while nonrigid body transformation might have been necessary without the use of the mask if the animal mouth might have been in a different position. Another step that can cause significant resolution degradation as a consequence of realignment is the image interpolation step necessary when the transformation is applied. The simplest interpolation methods,

312

V. Sossi

the nearest neighbor, determine the voxel nearest to the point of interest, and it assigns that value to that point. Depending on the size of the voxels relative to the size of the structure of interest, errors up to 5–10% can be introduced into the quantitative accuracy of the images. It is thus always important to investigate the effect of interpolation: several registration packages now allow a choice of interpolation. There is always a trade-off between accuracy of interpolation and time required to perform the transformation. In this context, the use of B-splines has been found to preserve the quantitative accuracy of the transformed images fairly well, while still being relatively fast (Hutton and Braun 2003). Finally, it is also important to consider that some imaging modalities introduce nonlinear and nonlocal geometric distortions into the images, which persist even in hybrid imaging; determination and correction for such distortions are an active area of research, particularly relevant in the context of multi-modality imaging.

22.7.2 Image Fusion Image fusion relates to displaying the information contained in two co-registered images in such a way as to visually convey the maximum amount of information. Most often, the pixel intensities of the fused image are derived from individual pixel intensity in each image using a transparency factor b for each color component separately. For example, if one has a 24-bit display, with 8 bits for each color (Red Green Blue), then pixel intensity for the color blue (and similar for the other two colors) in the fused image follows the relation Bb = b B 1 + (1− b ) B 2 ,

where the indices 1 and 2 refer to the two images. The user can adjust the transparency factor b, from a value of 1, in which case Bb = B1 to a value of 0, leading to Bb = B2. In order to maximize visual contrast, often a gray scale is used for one modality (typically the anatomical) and a color scale for the other (typically the functional).

22.8 Conclusion Multi-modality is the next logical step in the field of in vivo imaging. Technical advances in each modality together with an increasing understanding of their limits

and strengths are permitting to design synergistic imaging combinations that are much more powerful than any modality alone. These developments together with fascinating progress in the area of multi-modality imaging probes are providing a new context for preclinical imaging, where function and disease can be investigated at a multi-scale level, thus providing new knowledge and strengthening the bridge between basic science and clinical application.

References Alexandrakis G, Rannou FR, Chatziioannou AF (2006) Effect of optical property estimation accuracy on tomographic bioluminescence imaging: simulation of a combined opticalPET (OPET) system. Phys Med Biol 51(8):2045–2053 Al-Hallaq HA et al (1998) Correlation of magnetic resonance and oxygen microelectrode measurements of carbogeninduced changes in tumor oxygenation. Int J Radiat Oncol Biol Phys 41(1):151–159 Baudelet C, Gallez B (2002) How does blood oxygen leveldependent (BOLD) contrast correlate with oxygen partial pressure (pO2) inside tumors? Magn Reson Med 48(6):980–986 Berard P et al (2007) Investigation of the LabPET (TM) detector and electronics for photon-counting CT imaging. Nucl Instrum Methods Phys Res A 571(1–2):114–117 Beyer T et al (2000) A combined PET/CT scanner for clinical oncology. J Nucl Med 41(8):1369–1379 Brunner FC et al (2009) Imaging performance of the hybrid pixel detectors XPAD3-S. Phys Med Biol 54(6):1773–1789 Catana C et al (2006) Simultaneous PET-MRI- from detector modules to imaging system. Society for Molecular Imaging, Fourth Annual Meeting in Cologne, Germany, 7–10 September 2005, p 4 Catana C et al (2006) Simultaneous acquisition of multislice PET and MR images: initial results with a MR-compatible PET scanner. J Nucl Med 47(12):1968–1976 Chen JJ et al (2008) Cerebral blood flow measurement using fMRI and PET: a cross-validation study. Int J Biomed Imaging 2008:516359 Choi JS et al (2008) A hybrid nanoparticle probe for dual-modality positron emission tomography and magnetic resonance imaging. Angew Chem Int Ed Engl 47(33):6259–6262 Chow PL, Rannou FR, Chatziioannou AF (2005) Attenuation correction for small animal PET tomographs. Phys Med Biol 50(8):1837–1850 Culver J, Akers W, Achilefu S (2008) Multimodality molecular imaging with combined optical and SPECT/PET modalities. J Nucl Med 49(2):169–172 Duconge F et al (2008) Fluorine-18-labeled phospholipid quantum dot micelles for in vivo multimodal imaging from whole body to cellular scales. Bioconjug Chem 19(9):1921–1926 Eberl S et al (1996) Automated interstudy image registration technique for SPECT and PET. J Nucl Med 37(1):137–145 Franc BL et al (2008) Small-animal SPECT and SPECT/CT: important tools for preclinical investigation. J Nucl Med 49(10):1651–1663

22 Multi-modal Imaging and Image Fusion Gilbert KM et al (2009) Evaluation of a positron emission tomography (PET)-compatible field-cycled MRI (FCMRI) scanner. Magn Reson Med. 62(4):1017–1025 Hasegawa BH et al (1991) A prototype high-purity germanium detector system with fast photon-counting circuitry for medical imaging. Med Phys 18(5):900–909 Hasegawa BH et al (2002a) Dual-modality imaging of function and physiology. Acad Radiol 9(11):1305–1321 Hasegawa BH et al (2002b) Dual-modality imaging of cancer with SPECT/CT. Technol Cancer Res Treat 1(6):449–458 Hawkes DJ, Studholme C, Hill DL (1997) Accuracy, precision, and robustness of fully automated 3D neuro-image registration by multi-resolution optimisation of mutual information (MOMI). Radiology 205:111–111 Hawkes DJ, Hill D, Hallpike L, Bailey DL (2003) Coregistration of structural an functional images. In: Bailey D, Valk P, Townsed DW, Maisey M (eds) Positron emission tomography: basic science and clinical practice. New York, Springer, pp 181–198 Hill DLG et al (2001) Medical image registration. Phys Med Biol 46(3):R1–R45 Hofmann M et al (2009) Towards quantitative PET/MRI: a review of MR-based attenuation correction techniques. Eur J Nucl Med Mol Imaging 36(suppl 1):S93–S104 Hutton BF, Braun M (2003) Software for image registration: algorithms, accuracy, efficacy. Semin Nucl Med 33(3): 180–192 Judenhofer MS et al (2007) PET/MR images acquired with a compact MR-compatible PET detector in a 7-T magnet. Radiology 244(3):807–814 Judenhofer MS et al (2008) Simultaneous PET-MRI: a new approach for functional and morphological imaging. Nat Med 14(4):459–465 Kluetz PG et al (2000) Combined PET/CT Imaging in Oncology. Impact on patient management. Clin Positron Imaging 3(6):223–230 Krohn KA, Link JM, Mason RP (2008) Molecular imaging of hypoxia. J Nucl Med 49(2):129–148 Lecomte R et al (1990) High resolution positron emission tomography with a prototype camera based on solid state scintillation detectors. IEEE Trans Nucl Sci 37(2):805–811 Lecomte R et al (2001) Performance analysis of phoswich/ APD detectors and low-noise CMOS preamplifiers for high-resolution PET systems. IEEE Trans Nucl Sci 48(3): 650–655 Lee S, Chen X (2009) Dual-modality probes for in vivo molecular imaging. Mol Imaging 8(2):87–100 Li Y et al (2008) Geometric distortion correction for echo planar images using nonrigid registration with spatially varying scale. Magn Reson Imaging 26(10):1388–1397 Lindner JR (2004) Microbubbles in medical imaging: current applications and future directions. Nat Rev Drug Discov 3(6):527–532 Liu XD, Ma SM, Liu SZ (2003) Effects of 0.075 Gy x-ray irradiation on the expression of IL-10 and IL-12 in mice. Phys Med Biol 48(13):2041–2049 Lucas AJ et al (2006) Development of a combined micro-PETMR system. In: IEEE Nuclear Science Symposium Conference Record, doi.10.1109/NSSMIC.2006.354384:2345–2348 Maes F et al (1997) Multimodality image registration by maximization of mutual information. IEEE Trans Med Imaging 16(2):187–198

313 Marsden PK et al (2002) Simultaneous PET and NMR. Br J Radiol 75:53–59 Maurer CR et al (1997) Registration of head volume images using implantable fiducial markers. IEEE Trans Med Imaging 16(4):447–462 Michalet X et al (2005) Quantum dots for live cells, in vivo imaging, and diagnostics. Science 307(5709):538–544 Murray CB, Kagan CR, Bawendi MG (2000) Synthesis and characterization of monodisperse nanocrystals and closepacked nanocrystal assemblies. Annu Rev Mater Sci 30:545–610 Ntziachristos V et al (2005) Looking and listening to light: the evolution of whole-body photonic imaging. Nat Biotechnol 23(3):313–320 Otte AN et al (2005) A test of silicon photomultipliers as readout for PET. Nucl Instrum Methods Phys Res 545: 705–715 Pelizzari CA et al (1989) Accurate 3-dimensional registration of CT, PET, and or MR images of the brain. J Comput Assist Tomogr 13(1):20–26 Pichler B et al (1998) Performance test of a LSO-APD PET module in a 9.4 tesla magnet. In: IEEE Nuclear Science Symposium and Medical Imaging Conference, Piscataway, NJ, pp 1237–1239 Pichler B et al (1998b) Studies with a prototype high resolution PET scanner based on LSO-APD modules. IEEE Trans Nucl Sci 45(3):1298–1302 Pichler BJ, Judenhofer MS, Pfannenberg C (2008a) Multimodal imaging approaches: PET/CT and PET/MRI. Handb Exp Pharmacol 185(1):109–132 Pichler BJ, Wehrl HF, Judenhofer MS (2008b) Latest advances in molecular imaging instrumentation. J Nucl Med 49(suppl 2) : 5–23 Pichler BJ et al (2008c) Positron emission tomography/magnetic resonance imaging: the next generation of multimodality imaging? Semin Nucl Med 38(3):199–208 Pietrzyk U et al (1994) An interactive technique for 3-dimensional image Registration – validation for PET, SPECT, MRI and CT brain studies. J Nucl Med 35(12):2011–2018 Press WH, F.B, Teukolsky SA, Vetterling WT, Flannery BP et al (1993) Numerical Recipes in C, 2 edn. Cambridge University Press, Cambridge Rueckert D et al (1999) Nonrigid registration using free-form deformations: Application to breast MR images. IEEE Trans Med Imaging 18(8):712–721 Saoudi A, Lecomte R (1999) A novel APD-based detector module for multi-modality PET/SPECT/CT scanners. IEEE Trans Nucl Sci 46(3):479–484 Schlemmer HP et al (2008) An integrated MR/PET system: prospective applications. Abdom Imaging. doi:10.1007/s00261008-9450-2 Abdom Imaging. 2009 Nov;34(6):668–74 Schmidt KC, Turkheimer FE (2002) Kinetic modeling in positron emission tomography. Q J Nucl Med 46(1):70–85 Shao Y et al (1997a) Simultaneous PET and MR imaging. Phys Med Biol 42(10):1965–1970 Shao Y et al (1997b) Development of a PET detector system compatible with MRI/NMRsystems. IEEE Trans Nucl Sci 44(3):1167–1171 Soret M, Bacharach SL, Buvat I (2007) Partial-volume effect in PET tumor imaging. J Nucl Med 48(6):932–945 Sossi V et al (2007) In vivo measurement of density and affinity of the monoamine vesicular transporter in a unilateral

314 6-hydroxydopamine rat model of PD. J Cereb Blood Flow Metab 27(7):1407–1415 Townsend DW (2008) Dual-modality imaging: combining anatomy and function. J Nucl Med 49(6):938–955 Webb PP, McIntyre RJ, Conradi J (1974) Properties of Avalanche photodiodes. RCA Rev 35:234–278 Wehrl HF et al (2009) Pre-clinical PET/MR technological advances and new perspectives in biomedical research. Eur J Nucl Med Mol Imaging 36(1):56–68 Weissleder R (2002) Scaling down imaging: molecular mapping of cancer in mice. Nat Rev Cancer 2(1):11–18 West J et al (1997) Comparison and evaluation of retrospective intermodality brain image registration techniques. J Comput Assist Tomogr 21(4):554–566 Willmann JK et al (2008) Targeted microbubbles for imaging tumor angiogenesis: assessment of whole-body biodistribution with dynamic micro-PET in mice. Radiology 249(1):212–219

V. Sossi Wirrwar A et al (1997) 4.5 Tesla magnetic field reduces range of high-energy positrons - Potential implications for positron emission tomography. IEEE Trans Nucl Sci 44(2): 184–189 Woods RP, Cherry SR, Mazziotta JC (1992) Rapid automated algorithm for aligning and reslicing PET images. J Comput Assist Tomogr 16(4):620–633 Woods RP, Mazziotta JC, Cherry SR (1993) MRI-PET registration with automated algorithm. J Comput Assist Tomogr 17(4):536–546 Yamamoto S, Kuroda K, Senda M (2003) Scintillator selection for MR-compatible gamma detectors. IEEE Trans Nucl Sci 50(5):1683–1685 Zhang M et al (2008) Accuracy and reproducibility of tumor positioning during prolonged and multi-modality animal imaging studies. Phys Med Biol 53(20):5867–5882

Part Ex Vivo Validation Methods

IV

Ex Vivo and In Vitro Cross Calibration Methods

23

Albertine Dubois, Julien Dauguet, and Thierry Delzescaux

Contents

23.1 Introduction

23.1 Introduction........................................................... 317

Biomedical imaging has been increasingly utilized in medical research to study brain and its associated dysfunctions. Biomedical imaging provides valuable in vivo information particularly in diagnosing diseases and guiding therapies. In the past years, the development of dedicated small animal imaging systems has enabled anatomical (MRI, Morris, 1999; CT, Paulus et al. 1999), or even functional (SPECT, Weber et al. 1994; PET, Cherry et al. 1997; fMRI, Mandeville et al. 1998), three-dimensional (3D) imaging of in vivo rodent brain. It has also opened up the possibility of performing repeated studies in the same animal: longitudinal follow-up of brain atrophy, cerebral glucose metabolism, blood flow and oxygen metabolism; studies of protein synthesis; tracing of neural pathways; and the investigation of neurotransmission processes. However, despite huge technological advances, these systems remain subject to technical limitations, including poor sensitivity and low spatial resolution (~0.5–3 mm), which do not match the level of anatomical or functional details revealed by postmortem imaging (e.g., histology, autoradiography).

23.2 Image Processing and Histological Preparation............................................................ 319 23.2.1 Definitions and Terminology.................................. 319 23.2.2 Types of Brain Deformation Caused by Histological Preparation......................................... 324 23.2.3 Propagation-Based Registration............................. 327 23.2.4 The Photographic Modality.................................... 330 23.3 Correction of the Secondary Deformations........ 331 23.3.1 Manual Registration................................................ 333 23.3.2 Fiducial Marker or Artificial Landmark-Based Methods...................................... 333 23.3.3 Feature-Based Methods.......................................... 335 23.3.4 Gray Level-Based Methods.................................... 335 23.3.5 Interslice Intensity Normalization.......................... 336 23.4 Correction of the Primary Deformations........... 338 23.4.1 3D Registration....................................................... 338 23.4.2 In Vivo/Postmortem Registration in Primates........ 339 23.4.3 In Vivo/Postmortem Registration in Rodents......... 340 23.5 Conclusion............................................................. 342 References............................................................................ 343

A. Dubois (*) CEA-DSV-I2BM-SHFJ, Inserm U1023, 4, place du Général Leclerc, 91401 Orsay Cedex, France e-mail: [email protected] J. Dauguet and T. Delzescaux CEA-DSV-I2BM-MIRCen Neurodegenerative diseases laboratory, 18, route du Panorama, 92265 Fontenay-aux-Roses Cedex, France

F. Kiessling and B.J. Pichler (eds.), Small Animal Imaging, DOI: 10.1007/978-3-642-12945-2_23, © Springer-Verlag Berlin Heidelberg 2011

317

318

Histology and autoradiography are low-cost, widely available techniques for anatomical and functional neuroimaging in small animals. Histological analysis is still the gold standard for the accurate description of neuroanatomy and for tissue characterization of the brain. High-resolution autoradiography (~100–200 mm spatial resolution) is usually required to validate the in vivo functional results obtained with small animal MRI and PET imaging systems (Houston et al. 2001; Thanos et al. 2002; Toyama et al. 2004; Schmidt and Smith 2005) and remains the reference technique most widely used in small animal research for functional brain imaging. Correlation of high-resolution postmortem histological and/or autoradiographic data with lower resolution in vivo images, the ultimate geometric reference for analysis, is hence of high interest for the biomedical community. It enables to assess the relationship between measured signal, microanatomy, and function of the brain. In most studies, postmortem correlates of parameters obtained with in vivo imaging systems are still assessed visually by selecting similar 2D slices in both modalities (e.g., the longitudinal investigation of the morphological changes in a rat model of stroke, Chen et al. 2007 or the location of senile plaques resulting from amyloid-b deposits in a mouse model of Alzheimer’s disease, Wengenack et al. 2008). In some cases, a 3D volumetric relationship between functional or histopathologic features and in vivo imaging may be required for meaningful comparisons, especially when these correlates are performed at a local scale. This in vivo/postmortem 3D volumetric comparison involves the registration of two 3D volumes, from which at least one has been obtained from a set of two-dimensional (2D) brain sections. Indeed, the major disadvantage of histology and autoradiography is that they require brain tissue sectioning, entailing the production of up to several hundreds of serial sections and the inherent loss of the 3D spatial consistency. The mapping of postmortem data onto in vivo images remains challenging, due to the many differences between the brain in its original geometry and the resulting series of 2D brain sections. Generally speaking, there are two types of brain deformations caused by histological preparation: (1) global 3D deformations caused by the death of the animal and the extraction of the brain from the skull, and (2) individual slice-specific 2D deformations,

A. Dubois et al.

due to the serial sectioning of the brain into 2D slices and the immersion of slices into various solutions and specific staining dyes. The correction of the global deformations requires to estimate a 3D transformation between the brain in its postmortem geometry and the actual in vivo brain and hence leads to the 3D in vivo/postmortem matching. To estimate 3D transformations for the registration of two brains, several techniques have been proposed in the literature, either performing rigid transformations using mutual information as similarity criterion, or nonlinear transformations. The correction of the individual slice-specific 2D deformations requires the reconstruction of the 3D volume by alignment of the series of slices and hence leads to the 3D reconstruction of postmortem data. Many works have been done on the alignment of series of 2D slices, using global affine transformations or nonlinear transformations. The correction of these two different types of deformations that will ultimately link the analysis of in vivo and postmortem data can benefit from the use of an intermediate modality. This intermediate modality consists of photographs or videos of the brain face taken during the sectioning process, which are usually called blockface photographs. In this chapter, we aim to describe the methodological problems associated with the fusion of 2D histological and/or autoradiographic sections with a 3D volume (MRI, PET) and to summarize methods and strategies proposed in literature. In the first section, entitled Image processing and histological preparation, we briefly define terminology and concepts related to image processing and to describe the deformations caused by histological preparation. This description is essential to well understand the difficulties and the challenges of in vivo/postmortem matching. In the second section, entitled Correction of the secondary deformations, we describe various methods reported in literature allowing to correct for the deformations associated with the sectioning process and hence leading to the 3D reconstruction of postmortem data. In the third and last section, entitled Correction of the primary deformations, we describe methods reported in literature allowing to correct for the deformations which distinguish a postmortem brain from an in vivo brain and hence leading to the 3D in vivo/postmortem matching.

23 Ex Vivo and In Vitro Cross Calibration Methods

23.2 Image Processing and Histological Preparation 23.2.1 Definitions and Terminology To begin with, it seems necessary to briefly clarify some image processing terminology which will be used later in the chapter. This part does not claim to give a complete definition of these expressions, but rather to present the rudiments on which we are going to build our speech.

23.2.1.1 Image The term “medical image” covers a wide variety of types of images, with very different underlying physical principles and very different applications. The sort of images used in healthcare and medical research vary from microscopic images of histological sections to video images used for remote consultation, and from images of the eye taken with a fundus camera to whole body radioisotope images. From the point of view of computer science, an image of dimension n is a discrete function from the intervals of integers [0, dim1] × [0, dim2] × ... × [0, dimn] to a set I. In general, n is 2 or 3, dimi, 1
Fig. 23.1 2D and 3D sampling: from pixels to voxels. (Figure adapted from Bruno Nazarian’s lesson: “Imagerie Médicale 3D Visualizations, segmentations et reconstructions,” http://bnazarian.

319

the intensities is an interval of integers or floats limited by the image format. For U8 format, I is a subset of 8-bits coded unsigned integers (e.g., [0,255]). I can be a set of scalars, but also of vectors as in the case of color images (each point in the image is defined by three scalars, associated with the three channels for red, green, and blue components and interpreted as coordinates in some color space). Medical images are hence discrete and stored as an array of numbers, each corresponding to a different position in the patient. In 2D the elements in the array are called picture elements (pixels). Pixels are not necessarily square, but all pixels have the same dimensions. In three dimensions (3D), the pixel becomes the volume element (voxel). The third dimension can be spatial as above, (e.g., MRI or CT) or temporal (e.g., fluoroscopy). We can have a fourth dimension, most commonly a 3D image volume acquired over time. An image has dimensions in pixels or voxels (e.g., 256 × 256 × 128 for a typical MR image volume). Each pixel or voxel has spatial dimensions (e.g., 0.8 × 0.9 × 2 mm3). Image data are normally stored as a list of raw values, with a separate header (at the beginning of the file, or in a separate file) to describe the image. The third dimension is often treated differently from the first two. A 3D image is commonly treated as a sequence of 2D images. The process of forming a discrete image from a continuous object is called sampling (Fig. 23.1). The

free.fr/MyUploads/ImagerieMedicale3D.pdf. Bruno Nazarian, CHU, ’La Timone’ Hospital, Marseille, France)

320

A. Dubois et al.

gap between the centers of adjacent samples in a given direction is the sampling interval. The sampling interval is often anisotropic. Sampling interval in the through-slice direction in tomographic images is often larger than in the slice plane. The sampling can be at regular intervals, or irregular intervals: normally (but not always) regular in a 2D image, or in the slice plane of a multislice image, sometimes irregular in the through-slice direction. Most medical images are made up of one or more slices (Fig. 23.2). The orientations of the slices are defined anatomically: • Horizontal – plane normal to a vector from head to toe • Coronal – plane normal to a vector from front to back • Sagittal – plane normal to a vector from left to right

a

• Oblique – a slice that is not (at least approximately) one of the above A medical image might be made of a single slice, a series of parallel slices with uniform spacing, a series of slices with varying spacing, and/or orientation, or be volumes. In some types of medical image, data are acquired from an entire volume in one go (note this is different from acquiring multiple slices, one after the other). Hence, in this chapter, a slice will indifferently refer to a physical slice of a postmortem brain or to the image resulting from the digitization of this slice. More generally, this term may indicate a 2D image extracted from a 3D volume such as a MRI or a CT scan. By volume, we will also often mean a spatially consistent image of dimension 3.

Coronal slice FRONT

RIGHT

LEFT Horizontal slice

BACK

b

Sagital slice

Fig. 23.2 (Up) Representation of the possible orientations of the slices on a model of 3D human brain (gray). (Bottom) MR image of a human brain in the three planes of section. (Images credit: Florent Sureau and Claire Leroy, CEA-Service Hospitalier Frederic Joliot, Orsay, France)

321

23 Ex Vivo and In Vitro Cross Calibration Methods

23.2.1.2 Image Registration The meaning of the word registration is determining a transformation T that can relate the position of features in one image (the test image) with the position of the corresponding feature in another image (the reference image). The purpose of image registration (also called image matching) is therefore to spatially transform the test image to align with the reference one so that voxels (or pixels in 2D) representing the same underlying image features may be superimposed. Registration algorithms that make use of geometrical features in the images such as points, lines, and surfaces determine the transformation T by identifying features such as sets of image points xA and xB that correspond to the same physical entity visible in both images, and calculating T for these features. When these algorithms are iterative, they iteratively determine T, and then infer the final transformation when the algorithm has converged. Registration algorithms that are based on image intensity values work differently. They iteratively determine the image transformation that optimizes a voxel similarity measure. A standard framework for such an image registration consists in minimizing a cost function or maximizing a “similarity measure” that expresses the pixel or voxel similarity between the images to be registered. When the reference and test images are acquired of a single subject, we refer to it as intrasubject registration. If the registration is accomplished using two images of different subjects, this is referred to as intersubject registration. In the case of 2D slices, all the slices result generally from the same subject, but the registration of a slice with a nearby slice is similar to an intrasubject registration. If one image is acquired from a single patient and the other image is somehow constructed from an image information database obtained using imaging of many subjects, we name it atlas registration. Intrasubject registration is by far the most common of the three, used in almost any type of diagnostic and interventional procedure.

23.2.1.3 Types of Registration We have not, so far, discussed the nature of the transformation T. For most current applications of medical image registration, the discrete domains of both images

are three-dimensional, so T transforms from 3D space to 3D space. Where both images being registered are two-dimensional (e.g., two radiographs before and after contrast are injected or two intrinsically 2D images like histological slices), the appropriate transformation might be from 2D space to 2D space.1 Compared to 3D/3D registration, 2D/2D registration is less complex by an order of magnitude both where the number of parameters and the volume of the data are concerned, so obtaining a registration is in many cases easier and faster than in the 3D/3D case. In some circumstances, such as the registration of a radiograph to a CT scan, it is useful to consider transformations from 3D space to 2D space (or vice versa). Single-modality registration methods tend to register images of the same modality acquired by the same scanner/sensor type, while multimodality registration methods tend to register images acquired by different scanner/sensor types. Multimodality registration methods are often used in medical imaging as images of a subject are frequently obtained from different scanners. Monomodal tasks are well suited for growth monitoring, intervention verification, rest-stress comparisons, and many other applications. The applications of multimodal registration are abundant and diverse, predominantly diagnostic in nature. A coarse division would be into anatomical-anatomical registration, where images showing different aspects of tissue morphology are combined, and functional-anatomical, where tissue metabolism and its spatial location relative to anatomical structures are related. Examples include registration of brain PET/MR images in order to relate an area of dysfunction to anatomy or whole body PET/CT images for improving tumor localization (Fig. 23.3). The use of both modalities combined is beneficial, as the one is better suited for delineation of anatomical or tumor tissue (and has in general better tissue contrast), while the second is needed for accurate computation of the metabolic activity. Image-based registration can be divided into extrinsic, i.e., based on foreign objects introduced into the imaged space, and intrinsic methods, i.e., based on the image information as generated by the patient. Extrinsic methods rely on artificial objects attached to the patient, objects which are designed to be well lthough the images being registered may be 2D, the structures A being imaged are often free to translate, rotate, and deform in three dimensions.

1

322

A. Dubois et al.

Fig. 23.3 (Up) Registration of MRI and PET sagittal brain images. (Left) T1-weighted MRI alone, (middle) [18F]-FDG PET alone, and (Right) fusion image corresponding to the superimposition of both registered images. (Bottom) Registration of CT and PET coronal whole body images. (Left) CT image alone,

(middle) [18F]-FDG PET alone, and (right) fusion image corresponding to the superimposition of both registered images. [18F]-FDG PET images appear in false color rainbow in fusion images. (Images credit: Claire Leroy and Maria-Joao Ribeiro, CEA-Service Hospitalier Frederic Joliot, Orsay, France)

visible and accurately detectable in all of the pertinent modalities. Intrinsic methods rely on patient-generated image content only. Registration can be based on a limited set of identified salient points (landmarks), on the alignment of segmented binary structures (segmentation based), most commonly object surfaces, or directly onto measures computed from the image gray values (voxel property based). The operation that consists of registering every pair of consecutive 2D slice images in the series from a same brain is called alignment (Fig. 23.4). The series of 2D aligned sections are then stacked in the direction

orthogonal to the cutting plane, hence leading to a 3D reconstructed volume. Alignment enables 2D slice images to recover their geometrically coherent 3D alignment.

23.2.1.4 Nature of the Transformation If the image coordinates transformation maps parallel lines onto parallel lines it is called affine. If it maps lines onto lines, it is called projective. Finally, if it maps lines onto curves, it is called nonlinear or elastic.

23 Ex Vivo and In Vitro Cross Calibration Methods

323

Fig. 23.4 Principle of alignment: from 2D individual slices to a 3D reconstructed volume. (Images credit: Dubois Albertine and Thierry Delzescaux, CEA-MIRCen, Fontenay-aux-Roses, France)

A transformation is called global if it applies to the entire image, and local if each of the subsections of the image has its own transformations defined. Image affine transformations describe the set of transformations P which are written in the form P(x)=Ax+T for any element x of the n-dimensioned space of spatial coordinates of an image (n is 2 in 2D and 3 in 3D) and where A is a real-number matrix of size n × n and T any real vector of size n. If we note R the matrix of a rotation (i.e., verifying the conditions R T R = Id n and det(R) = 1, where RT is the transpose of matrix R, Id n denotes the identity matrix of order n, and det the determinant function), the most commonly encountered affine transformations are (Fig. 23.5): • • • • •

A translation ( A = Id n and T is any) A rigid body transformation (A=R and T is any) A similitude (A=s.R where s is a scalar) A homothety (A=D where D is diagonal and T=0) A general affine transformation (A and T are any)

A rigid transformation includes global translation and rotation thereby preserving distances, planarity of surfaces, and angles. Most medical image registration algorithms additionally assume that the transformation is rigid body, i.e., there are six degrees of freedom (or unknowns) in the transformation: three translations and three rotations. It is typically used to compensate for global patient repositioning. If the objects are the same size and shape, then you should be able to register any two images of the objects. An affine transformation includes anistropic global scaling and shearing, in addition to rigid transformation. It preserves planarity of surfaces and parallelism, but not the angle. It compensates for additional global size changes and shears. It is typically used to account for

Fig. 23.5 Schemes illustrating each of the above-mentioned affine transformations

the scale and size differences in intersubject brain registration. Affine registration algorithms increase the number of degrees of freedom by allowing for anisotropic scaling (giving nine degrees of freedom) and skews (giving 12 degrees of freedom). For any affine transformation, A can be written as the product of a scale matrix (a diagonal matrix in which the entries outside the main diagonal are all zero), of a cutting matrix (a lower triangular matrix

324

where the entries above the main diagonal are zero), and of a rotation matrix. A vector application L : x ® L ( x ) = Ax is hence associated to any affine transformation P : x ® P( x ) = Ax + T . This linear application verifies any points M and N of the   for space: P( M ) P( N ) = L( MN ) . A transformation that includes scaling and skews as well as the rigid body parameters is referred to as affine and has the important characteristics that it can be described in matrix form and that all parallel lines are preserved. A rigid body transformation can usefully be considered as a special case of affine, in which the scaling values are all unity and the skews all zero. The use of an affine transformation rather than a rigid body transformation does not greatly increase the applicability of image registration, as, except for brain, there are not many organs that only stretch or shear. Tissues usually deform in more complicated ways. By abuse of language, we will mean by a nonlinear transformation a nonaffine transformation (Fig. 23.6). A nonlinear transformation can have any number of degrees of freedom, but more than the affine transformation. A nonlinear transformation is said parametric if it can be written with a finite number of parameters. By contrast, the transformations that are said free are non-parametric. Among the nonlinear transformations, there is a distinction between the algorithms called elastic, which simultaneously minimize a term of data discrepancy and a term of energy regularization (initially inspired by the theory of the continuum mechanics), and the fluid algorithms which alternately minimize these two terms, entailing less smooth but more precise deformations. The Free Form Deformation transformation (Rueckert et al. 1999) belongs to the family of elastic nonlinear transformations: it is a parametric transformation contrary to what its name (Free) suggests.

Fig. 23.6 Scheme of registration between the test and reference images on the basis of the computation of nonlinear transformations

A. Dubois et al.

Most applications represent nonlinear transformations in terms of a local vector displacement (disparity) field or as polynomial transformations in terms of the old coordinates. A nonlinear transformation includes local transforms in addition to the global affine transform. It does not preserve distance, planarity, parallelism, or the angle. It is typically used to compensate for longitudinal tissue changes and deformation. For most organs in the body, many more degrees of freedom are indeed necessary to describe the tissue deformation with adequate accuracy. Even in the brain, development of children, lesion growth, or resection can make an affine transformation inadequate. Moreover, while an affine transformation is widely used to provide an approximate alignment of different subjects, additional degrees of freedom are needed for more accurate registration. Figure 23.7 summarizes the main transformations that can be applied to a brain volume for registration purpose.

23.2.2 Types of Brain Deformation Caused by Histological Preparation Generally speaking, there are two types of brain de formations caused by histological preparation and fixation: global 3D deformation and slice-specific 2D deformations.

23.2.2.1 Fixation by Perfusion and Extraction Many neuroscience protocols require to perfuse and fixate the brain via the circulatory system before its extraction from the skull. The rationale behind this process is that fixation is necessary to stabilize the proteins or peptides in the brain in order to allow for subsequent antibody binding. The classic animal transcardiac perfusion procedure with physiological saline prewash followed by formaldehyde fixative results in about 20% shrinkage of brain volume and gross anatomical distortion. After euthanasia of the animal, the brain is extracted from the skull and other deformations may then occur.

23 Ex Vivo and In Vitro Cross Calibration Methods Fig. 23.7 Illustration of the main transformations to be applied to a human brain for registration purpose

325

326

A. Dubois et al.

Fig. 23.8 Fusion image corresponding to the superimposition of a histological brain slice of a primate brain onto MRI data from the same animal. A global shrinkage of the postmortem brain compared to the MRI data is clearly visible (white arrows). Results are presented in coronal (left) and horizontal views (right). (Images credit: Julien Dauguet, CEA-MIRCen, Fontenay-aux-Roses, France)

The deformations caused by the fixation by perfusion and the extraction of the brain from the skull are denominated by global 3D deformations. They include: • A global shrinkage due to the loss of cerebrospinal fluid, the loss of blood irrigation, and a global dehydration (Fig. 23.8). • Various deformations due to gravity or other mechanical effects (manipulation of the brain) and possible specific treatments (such as PFA/sucrose cryopreservation and paraffin embedding). Following nomenclature proposed in Dauguet 2005, we defined these global 3D deformations as “primary deformations” because they are first to appear and they occur before sectioning the brain. In order to correct these deformations, it is therefore necessary to estimate a 3D transformation between the brain in its postmortem geometry and the actual in vivo brain.

23.2.2.2 Sectioning Process After the brain has been snap-frozen in isopentane at −40°C, it is embedded in an embedding matrix and cut on a cryostat or a microtome into thin sections (some tens of microns thickness) so as to obtain a series of 2D slices. In the great majority of cases, the sectioning process is performed in coronal incidence. The serial sectioning of the brain into 2D sections entails not only an inherent loss of the 3D spatial consistency of the brain, but also individual slice-specific 2D deformations (Fig. 23.9). The latter includes:

• Specific deformations due to sectioning (shears, folds, and tears). • Shrinkage due to temperature changes (see Gardella et al. 2003, for a quantitative study of de formations). • Contraction of tissues due to immersion of slices into various solutions and specific staining dyes. • Changes in the original geometry of each slice due to the mounting on glass slides (torn or missing parts, flattening) and local displacements of smaller parts like occipital cortex or cerebellum. Following nomenclature name proposed in Dauguet 2005, we defined these individual 2D deformations as “secondary deformations” because they appear during the cutting step, after the primary deformations. The correction of these deformations requires the reconstruction of the 3D volume by alignment of the series of slices. The scheme in (Fig. 23.10) summarizes the main steps of the preparation of cerebral histological slices and the associated deformations which have to be corrected to create an image comparable to the original brain. 23.2.2.3 Deformations and 3D Reconstruction It is always necessary to take into account and to correct for the secondary deformations to allow the reconstruction of a 3D consistent postmortem volume from a series of 2D slices (see Jones et al. 1994 for a study of the distortions in the rat). The correction of the primary deformations is, however, essential if the final objective of the study is to

23 Ex Vivo and In Vitro Cross Calibration Methods

327

Fig. 23.9 Illustration of the various slice-specific deformations and artifacts associated with the sectioning process and mounting on glass slides. (Images credit: CEA-MIRCen, Fontenayaux-Roses, France)

match a series of stained histological and/or autoradiographic 2D slices from a rodent brain with in vivo 3D data (anatomical MRI and/or PET scan) of the same animal. The correction of the primary deformations solely enables to lead to the 3D in vivo/postmortem matching. As we will see later in this chapter, the literature demonstrated that the correction of these two different types of deformations that will ultimately link the analysis of postmortem and in vivo data can benefit from the use of an intermediate modality. This intermediate modality consists of photographs or videos of the brain face taken during the sectioning process, which are usually called “blockface” photographs.

23.2.3 Propagation-Based Registration As mentioned above, the 3D reconstruction of a postmortem volume generally involves the alignment, i.e., the sequential registration of each slice to its adjacent slice in the series using image registration techniques based on an affine or a nonlinear transformation. Registration techniques based on a rigid body transformation are very often used. This type of registration is standard, robust, and well adapted for brain slices obtained with a cryostat because a rigid body transformation is mostly sufficient to superimpose one slice on the next one. Although the data can in some cases exhibit important deformation artifacts as a result of

328

A. Dubois et al.

Fig. 23.10 The main steps for the preparation of cerebral histological slices and the associated deformations which have to be corrected to create an image comparable to the original brain. Note that the deformations of the biological material have to be corrected in the images in the reversed order compared to the one along with which they occurred during the histological preparation. (Figure adapted from Dauguet et al., 2007)

sectioning process and tissue shrinkage, nonrigid (affine) or nonlinear transformations between the adjacent slices can distort the brain structures. As noticed by Lee et al. (2005b), it appears better to preserve the shape of each slice without compensating for the deformation artifacts than to take the risk of distorting the overall and regional image information during the registration process. Most rigid registration methods are based on a propagation scheme, consisting in serially propagating the transformations estimated between consecutive slices

relative to a reference slice in the series (Fig. 23.11). Each pair of consecutive slices in the series Si and Si +1 is first rigidly registered, yielding a 2D transformation Ti ¬i +1 (or equivalently Ti + 1¬i ). A reference slice of index ref1 is then chosen (because it carries few artifacts such as folds or tears and is located in the middle of the volume to limit error propagation), and the 2D transformations Ti ¬ref are computed by composition of the Ti ¬i +1 (if i < ref1 ) or Ti +1 ¬i (if i > ref1 ) transformations. The resampled slices, Si oTi¬ref , can then be stacked to build a geometrically coherent 3D volume. 1

1

329

23 Ex Vivo and In Vitro Cross Calibration Methods Fig. 23.11 (Left) Scheme of propagation-based registration for 3D reconstruction. (Right) Surface renderings of the histological volume, before and after registration. Quality of the registration and 3D reconstruction process can be assessed by visual inspection of internal structures (e.g., hippocampus) viewed in an incidence orthogonal to the cutting plane. We can thus observe that before registration, the hippocampal region is not accurately defined. (Images credit: Thierry Delzescaux, CEA-MIRCen, Fontenayaux-Roses, France)

At this stage, visual inspection is necessary to detect some eventual misalignments due to a misregistration between two consecutive slices. These misalignments separate consistent subvolumes in the 3D reconstruction. Indeed, consider a misregistration for the couple of slices of indices e and e+1. The 2D transformations with respect to the reference slice of index refi are computed by Ti ¬ref = Ti ¬i + 1ooTe-1¬ eoTe ¬ e+ 1ooTref -1¬ref for i < e < ref1 . The sections from index i to index e are then correctly coaligned together (they form a consistent subvolume), but are not correctly coaligned with section of index e+1. By visually inspecting the reconstructed volume in the two incidences orthogonal to the cutting plane, these misregistrations are thus easily identifiable. Typically, this only concerns a small number of slices per series, less than 2% of the total. However, this percentage highly relies on the quality of postmortem data sets: the higher the quality of the sections, the more accurate will be the registration. To correct them, it is possible to change the default registration parameters and coregister again the misaligned couple of consecutive sections, until the transformation is satisfactory. This propagation-based approach for 3D reconstruction of postmortem volumes has been criticized because 1

1

1

it can lead to different types of misregistrations. According to Nikou et al. (2003), if an error occurs in the registration of a section with the previous section, this error will be propagated through the entire volume. Thus, if the number of sections to be registered is large, an overall offset of the volume, due to error accumulation, may occur and a modification of its global shape may be observed. This problem is clearly most marked when distant sections are involved in registration. When one slice is missing or damaged due to sectioning (a common occurrence when dealing with 3D reconstruction of postmortem data), it can be simply removed from the series before the registration. When the volumes are reconstructed from about 100 20 mmthick consecutive slices covering a total of 2,000 mm, one 20 mm-thick slice therefore only represents 0.33% of the total volume. Hence, removing one 20 mm-thick cryosection from a volume of 200 consecutive ones has a negligible impact on the global shape of the final volume. When several sections (two or more) are missing or damaged due to sectioning, using the previous or the next sections for additional registration is a more appropriate solution than straightforwardly removing them. This way, we make sure that the global shape of the volume is preserved.

330

A. Dubois et al.

Propagation-based registration techniques have been intensively validated and are acknowledged to be the most comprehensive registration algorithms, in the absence of a 3D geometrical reference. Indeed, another difficulty in the reconstruction of a 3D volume from 2D sections (either histological or autoradiographic) is the so-called “banana problem”: a 3D curved object cannot be reconstructed from cross-sections in the absence of additional information (Streicher et al. 1997; Malandain et al. 2004; Fig. 23.12). In other words, in the absence of any 3D geometric reference, a rigid pairwise registration of 2D sections tends to make lose the curvature of the brain and hence do not provide a volume with the real geometry of the object under study, but a rectilinear one (i.e., the brain after extraction from the skull but before sectioning and staining). Although the 3D curvature of the postmortem brain is less pronounced in rodents than in biggest animals like nonhuman primates, this artifact can be very annoying when we seek to spatially match postmortem data with in vivo images, displaying the brain in its original geometry. However, this artifact cannot be really and directly corrected given that the curvature of the brain, i.e., its 3D spatial consistency, is inherently lost during the sectioning process, inducing the discretization into 2D slices. In the literature, there are nevertheless several solutions to this problem. The acquisition protocol can be designed to include additional information in the form of fiducial markers. To do so, needles can be stuck in the material before slicing, which provides correspondences to drive the geometry-based reconstruction (FordHolevinski et al. 1991; Goldszal et al. 1995, 1996; Humm et al. 1995). However, reconstruction can be awkward, particularly if needles are not orthogonal to the cutting plane or if needle holes collapse during the Fig. 23.12 The banana problem: the 3D reconstruction of a 3D curved object is not easy. (a) Take a 3D curved object (e.g., a banana); (b) cut it into slices; (c) digitize the slices; (d) mix the digitized slices; (e) the 3D reconstruction results in a cylindrical banana. (f) Using a shape prior (e.g., MRI) may help to reconstruct the curved banana. (Figure extracted from Malandain et al., 2004)

a

b

histological treatment. Another way is to acquire photographs or videos of the unstained surface of the brain during the slicing process, including a reference system fixed on the cryostat or the microtome (Toga et al. 1994; Schormann et al. 1995; Kim et al. 1997; Mega et al. 1997; Schormann and Zilles 1998; Ourselin et al. 2001; Bardinet et al. 2002; Meyer et al. 2006; Dauguet et al. 2007; Dubois et al. 2010). Alignment of these photographs using the reference system then provide a photographic 3D volume with the real geometry of the object under study, for example, the brain after extraction from the skull but before slicing and staining. Thus, further reconstruction of the 2D histological sections into a geometrically consistent 3D volume can be made by independent registrations of the 2D sections with the corresponding directly aligned photographs. When such a priori additional information about the real geometry of the object under study is available, the banana problem does not occur. Nevertheless, for practical reasons, such information is not always available. In that case, the only information about the geometry of the brain (its real anatomy) before slicing is given by the anatomical MRI volume (Malandain et al. 2004). However, in this case, the registration method is necessarily more complex seeing that the transformation being computed must simultaneously correct for both secondary and primary deformations.

23.2.4 The Photographic Modality 23.2.4.1 Acquisition of “Blockface” Photographs Images from the surface of the brain, usually called “blockface” photographs, are recorded with a digital or a video camera before the sectioning of each

c

d

e

f

331

23 Ex Vivo and In Vitro Cross Calibration Methods

section. These photographs are taken at the end of the cryostat wheel crank course or of the microtome sledge course, ensuring that the brain is in the same position for all sections (Figs. 23.13a, b). An optical system such as an optic fiber-ring light can be fixed close to the objective of the camera to ensure uniform, reproducible illumination of the sample. A laptop computer is used to control the camera, making it possible to take images remotely, and the images are stored directly on the hard disk of this computer. The calibration factor between the number of pixels in the images and the corresponding distance in millimeters is determined by obtaining an image of a piece of graph paper in the focal plane before sectioning. Then, the series of blockface photographs are directly stacked in the Z direction so as to create a 3D consistent blockface volume (Figs. 23.13c, d). No registration is required, due to the natural spatial consistency of the images. Small discrepancies due to accidental displacements of the imaging device, which shifted between the positions of two consecutive images, remain a rare occurrence. If necessary, we compensated for this shift locally, by applying a 2D rigid registration estimated between the blockface photographs taken before and after the displacement.

23.2.4.2 Interest of “Blockface” Photographs Blockface imaging does not properly constitute a biological modality because no specific histological staining is possible. Nevertheless, blockface imaging is of great interest because it allows to dissociate the correction of the two main types of deformations of the histological slices. The photographic volume presents “primary deformations” without displaying slice-specific deformations inherent to the sectioning process. The local secondary deformations are very limited because each blockface photograph is realized while the slice still stands by the block brain. It is hence free from possible deformation artifacts associated with the sectioning process (displacements, torn or missing parts, folds). The photographic volume thus constitutes a consistent 3D geometric reference for the correction of the secondary deformations. If each histological slice is rigidly registered to its corresponding blockface photograph, the 3D curvature of the postmortem

brain is preserved and the “banana problem” is thus avoided. It also represents the ideal modality to estimate the primary deformations with regard to the geometry of the in vivo brain. Indeed, the photographic volume preserved the original curvature of the brain (or a curvature close to the original one because this modality is a postmortem modality). Consequently, the blockface volume can be used to estimate the transformation with the in vivo data providing the greatest anatomical detail and the best resolution and contrast between specific organ tissues (e.g., MRI for brain or CT for whole body). Indeed, among postmortem modalities, it suffered fewer artifacts than the histological volume. The correction of the curvature and the possibility of correcting for both the deformations are interesting for the 3D reconstruction of any type of brain, but the introduction of the photographic modality is particularly useful for the studies for which the correlation with the in vivo imaging is envisaged. High-quality, well-defined blockface photographs are essential for 3D brain reconstruction from postmortem images. The tissues are unstained for blockface photographs and can be acquired only remotely, with a digital camera in reflection mode. The resolution and contrast between specific brain tissues are, therefore, more limited than for stained histological sections. Hence, the blockface photographs remain an intermediary modality used mainly for methodology purposes. Previous works in nonhuman primates overcame some of these limitations by using a special dye injected in vivo. This dye enhanced the contrast on blockface photographs of postmortem brains (Annese et al. 2006). However, the use of such a dye was incompatible with further quantitative [14C]-2-DG autoradiography. Indeed, it required prior fixation of the brain tissue by perfusion with 4% paraformaldehyde solution and cryopreservation by successive immersion in graded solutions of sucrose, which would wash out the 2DG trapped in the brain.

23.3 Correction of the Secondary Deformations Reconstruction of a 3D volume from a stack of 2D images (histological sections or autoradiographs) has been largely investigated in the literature. As

332 Fig. 23.13 Acquisition system of blockface photographs from rodent (a; cryostat) and primate (b; microtome) brains. Each blockface photograph is recorded with a digital camera (right corner). Corresponding blockface volumes achieved by stacking the series of photographs in the Z direction. Volumes are represented in coronal and horizontal views as indicated. (Images credit: Thierry Delzescaux and Julien Dauguet, CEA-MIRCen, Fontenay-aux-Roses, France)

A. Dubois et al.

a

b

c

d

333

23 Ex Vivo and In Vitro Cross Calibration Methods

previously mentioned, it consists of registering every pair of consecutive images in the stack so as to recover their geometrically coherent 3D alignment. It however differs from the image matching problem, as classically understood, in several ways as follows: • Slices to be registered are not images of the same object, but rather of similar objects (i.e., two consecutive sections) that can exhibit differences in size, shape, and/or intensity. • Noncoherent 2D deformations may occur from one slice to the next. Over the past two decades, many authors have reported how affine transformations can be applied to automatically bring consecutive sections into alignment (Toga and Arnicar-Sulze 1987; Hibbard and Hawkins 1988; Rydmark et al. 1992; Andreasen et al. 1992; Zhao et al. 1993; Goldszal et al. 1995; Schormann et al. 1995; Rangarajan et al. 1997; Hess et al. 1998; Cohen et al. 1998; Ourselin et al. 2001; Bardinet et al. 2002; Nikou et al. 2003; Malandain et al. 2004; Dauguet et al. 2007; Dubois et al. 2007). The simplicity of affine transformations not only reduces the complexity of computation, but also allows convenient manual adjustment through software interfaces (Chen et al. 2003; Karen et al. 2003; Andrey and Maurin, 2005). However, as shown in several studies (Deverell et al. 1989; Jones et al. 1994; Boyle et al. 1997), the section distortions induced by the preparation process are local in nature. With the advance of image registration techniques, reconstruction methods were proposed to correct localized section distortions by using local 2D nonlinear deformations (Kim et al. 1997; Timsari et al. 1999; Ali and Cohen 1998) as well as elastic 3D surface deformations (Thompson and Toga 1996; Mega et al. 1997; Schormann and Zilles 1998; Gefen et al. 2003; Pitiot et al. 2006; Chakravarty et al. 2006; Ju et al. 2006). These registration methods include: • • • •

Manual registration. Fiducial marker or artificial landmark-based methods. Principal axes alignment. Feature-based methods, using contours, crest lines, or characteristic points extracted from the images. • Gray level-based registration techniques using the intensities of the whole image, through similarity or correlation functions.

23.3.1 Manual Registration The most common method, manual registration (Laan et al. 1989; Rydmark et al. 1992; Deverell et al. 1993; Carlbon et al. 1994; Andrey and Maurin 2005), despite its simplicity, suffers from several drawbacks: it is not reproducible, it is user-dependent, and it is very timeconsuming. Those inherent difficulties make it not suitable to process large data sets. A benefit, nevertheless, of this technique is the very large freedom left to the user who can focus on a specific part of the images to be registered, and thus to exercise an active control over the registration (Fig. 23.14).

23.3.2 Fiducial Marker or Artificial Landmark-Based Methods Alternatively, fiducial markers (obtained by sticking needles in the material before slicing) (Fig. 23.15a) can be tracked over the whole stack to recover the original geometry (Toga and Arnicar 1985; Toga and Arnicar-Sulze 1987; Hibbard and Hawkins 1988; Ford-Holevinski et al. 1991; Goldszal et al. 1995; Humm et al. 1995; Goldszal et al. 1996; Hess et al. 1998). However, since it is usually done by least squares minimization, a bias may appear if needles are not orthogonal to the cutting planes. Moreover, tracking can sometimes be difficult, especially when needle holes collapse. The resulting tracks may be unreliable if the cutting planes are not perfectly orthogonal to the needle. Furthermore, the needles may destroy part of the tissues of interest in the sections and then impede any postmortem diagnosis. In addition, registration methods based on the use of intrinsic landmarks are not satisfactory in the case of autoradiography due to the time lapse between animal sacrificing and brain freezing. This causes the images to be unsharp (broadening of the point spread function), and therefore, it becomes difficult to find valid intrinsic markers. Considering these limitations, Streicher et al. (1997) and Simonetti et al. (2006) proposed an alternative to the use of needles by developing a method of registration based on external markers (Fig. 23.15b). Brain embedding provides the opportunity to add fiducial markers outside the organ, and thus to perform extrinsic registration without inducing tissue damage.

334

A. Dubois et al.

Fig. 23.14 Slice image registration. The images to be registered can be directly selected in the slice list on the right. Using the mouse, each slice can be transparently overlaid and moved (translation and rotation) on top of the previous one. Here, the

display of the segmented items has been switched off and the slice bounding frames are shown in color. Additionally, the user can select the most convenient magnification and transparency mode. (Image extracted from Andrey and Maurin, 2005)

Classical embedding methods, whether applied with or without external markers, are performed at room, or even at higher temperatures (Williams and Doyle 1996; Streicher et al. 1997; Sorensen et al. 2000; Bjarkam et al. 2001). They cannot be applied to frozen tissues, in particular to autoradiographic sections that may not defrost during processing. To the best of our knowledge, only Simonetti et al. (2006) proposes a strategy whereby four external fiducial markers are positioned outside the rat brain with the use of a low temperature brain embedding procedure, thus avoiding tissue degradation due to sample heating. The fiducial markers can be rapidly added to any frozen tissue block with no impact on the subsequent histological operations. Tissues are wrapped around four tubes of

Plexiglas, which are removed after hardening of the embedding medium. After the sections have been mounted on glass slides and digitized, the fiducial markers are clearly visible at the corners of the sections eventually obtained. Their initial positions are readily identified following segmentation of the digitized sections. These positions may be used to calculate the rotation/translation/scale/factor matrices that optimally register the entire set of sections. This technique gives very good results, but requires very important and uncommon equipments in laboratory. Other images registration methods have also been investigated (van den Elsen et al. 1993; Maintz and Viergever 1998; Hill et al. 2001). Those methods can be divided in two classes: the geometrical approaches

335

23 Ex Vivo and In Vitro Cross Calibration Methods

a

line 2

AC ST

anterior reference 2 in the striatum

TH

anterior reference 1 in the striatum

line 1 posterior reference in the thalamus

anterior

posterior

Fig. 23.15 (a) Autoradiograph of a horizontal section through a Gerbil brain: line 1 connects the anterior reference point 1 in the striatum (ST) with the posterior reference point in the thalamus (TH); line 2 connects the anterior reference point 1 with the

a nterior reference point 2 in the striatum (ST); AC, auditory cortex. Image extracted from Hess et al. (1998). (b) An embedded rat brain mounted inside the cryotome. The four markers appear in black (Indian ink). (Image extracted from Simonetti et al., 2006)

which require the segmentation of some features (contours, crest lines, characteristic points extracted from the images, or even objects of interest) and the iconic approaches based solely on image intensity.

may be used as initialization (Hibbard and Hawkins 1988; Hess et al. 1998). More precise features, for example, contours (Hibbard and Hawkins, 1988; Zhao et al. 1993; Cohen et al. 1998), edges (Kim et al. 1995), or points (Rangarajan et al. 1997) can also be used.

23.3.3 Feature-Based Methods Concerning the geometrical methods, registration can first be achieved with global descriptors, for example, centers of mass and principal axes. This proved to yield only limited precision (Schormann and Zilles 1997), but

23.3.4 Gray Level-Based Methods Because of the difficulty to design a fully automatic and reliable segmentation method, or to manually extract features of interest, iconic (intensity based)

336

methods have also been investigated. They are based on the minimization (or maximization) of a similarity measure of the images’ intensities. These criteria can be calculated on the whole image (global approach) or on the succession of subsections of the whole image (local approach). Among others, one finds cross-correlation (Hibbard and Hawkins 1988; Ourselin et al. 2001) or mutual information (Kim et al. 1997). The correlation coefficient is usually used when an affine relationship between the gray levels of both images is assumed (typically with single-modality registration methods). The mutual information is more adapted when there is no affine relationship between intensities of both images (typically with multimodality registration methods). In both cases, the absolute value of these two criteria is an increasing function of the quality of the registration. The most famous works in the field are those published by Hibbard and Hawkins (1988). After completing a prealignment based on moments of inertia, they align two successive slices by using the Fourier analysis. This approach is very fast, but unfortunately not very robust. As regards Andreasen et al. (1992), they minimized the sum of squared intensity differences between images to be registered, weighted by a term of contrast between the intensities. The works of Ourselin et al. reported the choice of the correlation coefficient as a local similarity criterion (Ourselin et al. 2001; Ourselin 2002). Their method, namely block matching, is very similar to the ICP2 algorithm (Besl and McKay 1992) which consists in extracting feature points in the two images to be registered and in iterating the following steps until convergence: (1) to pair each feature point of the test image with the closest feature point in the reference image; (2) to compute the transformation that will best superimpose the paired points; and (3) to apply this transformation to the feature points of the test image. Indeed, after applying the transformation, the pairings may have changed, thus a better transformation may be found by iterating these three steps. In the block matching algorithm, authors do not extract feature points, but consider subimages (i.e., blocks) in the test image that will be paired to the most similar subimage in the reference image. The computed transformation is the one that will best superimpose the centers of the paired subimages. This transformation is then applied to each subimage of the Iterative Closest Point

2

A. Dubois et al.

test image. The block matching algorithm can therefore be considered as a hybrid ICP-like approach between geometrical and iconic methods. Blocks can be considered as geometrical features while coregistration of blocks is achieved by minimization (or maximization) of a similarity measure. Many other authors have found block matching algorithm highly suitable for the purpose of 3D volume reconstruction from 2D brain sections obtained with a cryostat (Malandain et al. 2004; Annese et al. 2006; Dauguet et al. 2007; Dubois et al. 2007) (Fig. 23.16). This one turns out to be robust and to be able to take into account dissimilarities between slices due to the deformations undergone by tissues during the stages of histological preparation (tearings, stretchings, destructions, withdrawals, etc.). By using a local approach and a robust estimation scheme, this technique is, indeed, capable of estimating the good transformations even between sections, the information of which is partial, in a more robust way than global estimation registration methods. Moreover, its ability to register slightly distorted or damaged sections made it possible to automate it without supplementary data handling such as removal or replacement of those sections. The assumption of an affine relationship between intensity distributions is locally verified between the gray levels of the blockface photographs and the gray levels of either the histological or the autoradiographic sections because the regions (blocks) are usually sufficiently small. At last, Nikou et al. have developed a registration method based on the minimization of a global energy function and on robust statistics (similarity measures between a slice and its neighborhood in the 3D volume) (Nikou et al. 2003). Because such an energy function is isotropic, this method does not favor any direction during the registration process. Thus, the final alignment is not biased.

23.3.5 Interslice Intensity Normalization So far, we only mentioned the 3D reconstruction of a geometrically coherent 3D volume from 2D section images. In the previously cited literature, the problem of intensities alignment, that is, the correction of global intensity inhomogeneities from one image to the next, has rarely been addressed. Clearly, 2D sections may exhibit intensity inhomogeneities whose causes are multiple: autoradiographic sections of varying thickness

23 Ex Vivo and In Vitro Cross Calibration Methods

337

Fig. 23.16 3D reconstructions of histological rat (left; Nissl stain) and primate (right; Acetyl Cholinesterase stain) brains. All results are presented in coronal, horizontal, and sagittal views as indicated. (Images credit: Anne-Sophie Hérard and Julien Dauguet, CEAMIRCen, Fontenayaux-Roses, France)

present different amount of radioisotope and hence display different level of radioactivity fixation, sensitivity may vary from autoradiographic film to another, concentration of the various baths and time of incubation may have an influence on the intensity of the staining, etc. The stage of digital acquisition, using a CCD camera or a flatbed scanner, can also increase these differences. Because intersection differences in intensity can be annoying for the visual evaluation of the 3D reconstructions, it appears useful to compensate for these changes. Methods of intensity compensation were proposed in the literature, mainly to correct the drift in intensity noticed with certain acquisitions or certain antennas in MRI. For example, the works of Mangin

(2000) looking for a corrective field of the way by minimization of the entropy. These methods, well adapted to the slow drift of MRI scanners, are not really convenient for the histological data. More relevant methods dedicated to the correction of interslice intensity differences within histological data were proposed by Malandain and Bardinet (2003) and Dauguet et al. (2004). They both operate an iconic histogram registration for each pair of consecutive sections, yielding an affine intensity transformation fi ¬i + 1 (or equivalently fi + 1¬i ) between two consecutive sections. A reference section, of index ref 2 , is chosen and the intensity transformations f ref ¬i are computed by composition of the fi ¬i + 1 or fi + 1¬i intensity transformations. 2

338

23.4 Correction of the Primary Deformations In the previous section, the cited literature was limited to the problem of the reconstruction of a 3D volume from 2D slices. This required the correction of slicespecific 2D deformations associated to the sectioning process. However, it does not allow for an easy comparison with more classical 3D modalities (e.g., MRI, CT, or PET). The fusion of such histological or autoradiographic data with other 3D data has not been often addressed. In this case, the primary deformations, namely all the 3D deformations which differentiate the brain extracted postmortem from the brain in its in vivo original geometry, have to be compensated. Because MRI is doubtless the in vivo modality providing the most reliable and accurate morphometric and structural information, an anatomical T1-weighted MR image of the brain is considered as the modality of geometric reference. To perform in vivo/postmortem matching, the 3D reconstructed histological data, whose secondary deformations have been previously corrected, have to be registered with the modality of geometric reference. This time it is a 3D registration.

23.4.1 3D Registration Various registration techniques have been proposed in the literature to estimate the 3D transformations between two 3D in vivo brain images. They are mainly characterized by the space of transformations chosen as well as by the similarity measure used that both depend on the type of registration problem to be resolved (see the works of Roche (2001) on registration and similarity measures).

23.4.1.1 Intrasubject Registration Considering the registration of images arising from the same subject, the proper space of transformations is generally the space of rigid transformations. In the case of mono-subject single-modality registration, considered as the most favorable case, “simple” similarity

A. Dubois et al.

measures such as the correlation coefficient or even the sum of the squared intensity differences (SSD), which are minimized during registration, are often the most effective. The SSD measure is widely used for serial MRI registration, for example, by Hajnal et al. (1995a, b). It is also used in a slightly different framework by the registration algorithm within Friston’s statistical parametric mapping (SPM) software (Friston et al. 1995; Ashburner and Friston 1999). With mono-subject multimodality registration, such as the registration of MRI and TEP images from the same subject, the situation is quite different. There is, in general, no simple relationship between the intensities in the two images. More general similarity measures, such as the mutual information or the correlation ratio, are often used (Maes et al. 1997; Viola and Wells 1997; Maes et al. 2003).

23.4.1.2 Intersubject Registration In neuroscience research it is a common practice to register images from different individuals into a standard space, in order to study variability between subjects or to compare normal subjects with volunteers. This type of registration is an intersubject singlemodality registration, frequently called intersubject normalization. The alignment is most commonly carried out with an affine transformation (e.g., Collins et al. 1994), but more degrees of freedom are sometimes used. Elastic methods based on the modeling of deformation field as a linear combination of threedimensional discrete cosine transform (DCT) basis functions and using mutual information as similarity measure (see Friston et al. 1995; Ashburner and Friston 1997) are classically used. Many other efficient nonlinear registration techniques, adapted to the intersubject single-modality registration, exist (Christensen et al. 1996; Davatzikos 1997; Collins and Evans 1998; Thirion 1998; Woods et al. 1998; Christensen 1999; Cachier 2002; Clatz et al. 2005). Multisubject multimodality registration is rarely required, because it is the most unfavorable case and because it often makes little sense. A more complete overview of medical image registration can be found in Maintz and Viergever (1998) and Hill et al. (2001).

23 Ex Vivo and In Vitro Cross Calibration Methods

23.4.1.3 In Vivo/Postmortem Registration: A Special Case Of Mono-subject Registration Registration between postmortem data and in vivo brain images is part of the family of mono-subject multimodality registration. In theory, a proper transformation would be a rigid transformation as between MRI and TEP images from the same subject. However, as postmortem data have undergone many deformations, assumed limited to the primary deformations since the secondary deformations were the object of a preliminary correction, the strategy is necessarily different. To the best of our knowledge, histology has previously been coregistered with MRI (Schormann et al. 1995; Schormann and Zilles 1998; Jacobs et al. 1999; Ourselin et al. 2001; Bardinet et al. 2002; Zarow et al. 2004; Meyer et al. 2006; Malandain et al. 2004; Dauguet et al. 2007) and with PET data (Mega et al. 1997; Humm et al. 2003). In most cases, photographs were acquired during the cutting process to serve as an intermediate modality. A 3D linear transformation was used to map the photographic volume onto the MRI volume, while 2D highly nonlinear transformations compensated residual in-plane misalignments between the histological sections and the photographs. The transformations estimated between the blockface volume and the MRI data are then applied to the reconstructed histological volume.

23.4.2 In Vivo/Postmortem Registration in Primates For the particular 3D registration of postmortem and in vivo primate brains, different strategies have been proposed in the literature. Ourselin et al. (2001) and Bardinet et al. (2002) have proposed an affine 3D transformation for the registration of the basal ganglia. When the region of interest is very spatially limited, a simple 3D affine registration focused on the structure of interest (by windowing for example) can locally be sufficient. Indeed, some brain regions as the basal ganglia display relatively slight deformations between the two images. For global matching of one hemisphere, Malandain et al. (2004) proposed a method based on an alternated

339

correction of primary and secondary deformations using 3D and 2D affine registrations. They address the fusion of 2D autoradiographic slices with a 3D anatomical MR image, when no intermediate modality, such as photographs, is available. The autoradiographic slices are aligned at first, without the help of the MR image, by affine registration (correction of the secondary deformations), leading to a first autoradiographic volume, coherent with respect to both geometry and intensity. The reconstructed autoradiographic volume is then coregistered with the MRI volume using a 3D rigid transformation, providing an initial solution to the problem. This one is registered in 3D in an affine way with the MRI (correction of the primary deformations) taken as re ference imaging in terms of geometry. A fusion loop alternates between the reconstruction update of the autoradiographic volume with the help of the MR image and the 3D registration of the autoradiographic volume to the MR image. This registration allows to obtain by resampling a MRI corresponding to the histological slices. This way, authors ensure that the reconstructed 3D volume can be superimposed onto the MR image that represents the reference anatomy. This correspondence is used to realize a new adaptation by 2D affine registration of every autoradiographic slice on the corresponding slice in MRI. The process is iterated: the correction of the primary and secondary deformations is thus made gradually and alternately. The final transformation is a series of compositions of 2D/3D affine transformations alternately and can be likened to a class of particular 3D nonlinear transformation. This method is adapted to the case where the photographic modality was not acquired. Schormann et al. (1995) used an optimized affine registration of labeled regions to correct for the deformations between histological slices and the MRI section. Thompson and Toga (1996) developed a 3D warping method based on surface that was used in 2D for histology registration in Mega et al. (1997). The methods of purely nonlinear deformations specific to this problem of registering histological data with a geometrical reference as MRI remain relatively rare in the literature. Indeed, given the large number of degrees of freedom of these methods and the artifacts associated to the histological data, robustness and precision are difficult to reconcile. Thompson and Toga (1996) developed a method of deformation based on the mapping of surfaces, which seems to be adapted to the correction of the primary deformations.

340

Other methods of nonlinear registration were proposed and seemed to give interesting results with histological data. A 3D viscous fluid transformation model has been proposed by Christensen et al. (1997) and applied to the deformation of blockface photographs, whereas Schormann and Zilles (1998) performed a 3D transformation based on the elastic medium theory with a full multigrid strategy. Zarow et al. (2004) have developed an agar-embedding method that minimizes the physical distortions usually associated with slicing and handling of the postmortem brain. This method can also be used to embed and slice previously cut material. Incorporation of external fiducials in the agar surrounding each brain slice enables 3D reconstruction of the postmortem brain. The reconstructed brain can then be coregistered to the MRI using the specially developed coregistration method which uses polynomial warping and sliceto-volume transformation (Kim et al. 2000). In order to reduce the number of distortions within and between slices and to facilitate MRI coregistration, each cerebral hemisphere is processed separately. Authors describe postmortem MRI and brain tissue registration obtained using a technique using the Pearson crosscorrelation coefficient first and mutual information last as two separately and sequentially applied objective functions and an n th order polynomial as a 2D (only) geometric warping interpolant. More recently, other authors were interested in estimating an elastic transformation of the blockface volume to the MRI using a Free Form Deformation (FFD) model and hence to correct the primary deformations (Dauguet et al. 2007; Fig. 23.17). The deformation is modeled by cubic B-splines driven by a grid of control points and mutual information is used as similarity criterion. This method was originally dedicated to the correction of breast deformations (Rueckert et al. 1999) and it was also used to successfully correct the deformations for thorax respiration motions between positron emission tomography (PET) and computer tomography (CT) scan (Delzescaux et al. 2003; Mattes et al. 2003). The method is generic and offers a good control on the flexibility of the deformation controlled by the grid chosen. A coarse-to-fine approach can assure a progressive correction and robustness can be obtained because of the quite limited final degrees of freedom, which avoided wrong warping in the case of histological distortions. A trade-off concerning the number of control points used was found to allow an

A. Dubois et al.

accurate registration of the different regions and to avoid spurious deformations. The 3D rigid and elastic FFD-based transformations estimated between the blockface volume and the MRI data are then applied to the reconstructed histological volume. The elastic transformations like FFD are in general more sensitive to artifacts, such as tearing or missing parts because of the high number of degrees of freedom compared to linear transformations.

23.4.3 In Vivo/Postmortem Registration in Rodents For the particular 3D registration of postmortem and in vivo rodent brains, different strategies have also been proposed in the literature. Jacobs et al. (1999) present a method for coregistration and warping of MRI to histological sections for comparison purposes. This methodology consists of a modified head and hat surface-based registration algorithm using Pelizarri’s automatic surface fitting method for rigid structures (Pelizzari et al. 1989), followed by a new automated warping approach using nonlinear thin plate splines to compensate for distortions between the data sets. Elastic methods using thin plate splines with control points have been proposed to be used for rodent brains (Kim et al. 1997, which do not, however, present the same deformations as that of primate brains. Kim et al. (1997) demonstrated an automatic method for warping autoradiographic slices back to their geometry in the tissue block and then one final manually initialized 3D warping from the geometry of the tissue block to its geometry in the in vivo brain MR image using mutual information as the objective function and thin plate splines as the geometric warping interpolants. Humm et al. (2003) have developed and then validated a stereotactic fiduciary marker system for tumor xenografts in rodents which could be used to coregister MRI, PET, tissue histology, autoradiography, and measurements from physiologic probes. A Teflon fiduciary template has been designed which allows the precise insertion of small hollow Teflon rods into a tumor. These rods can be visualized by MRI and PET, as well as by histology and autoradiography on tissue sections. The fiduciary marker system provides a coordinate system used to

23 Ex Vivo and In Vitro Cross Calibration Methods

341

Fig. 23.17 Final matching of the series of stained histological slices (postmortem data) with the MR image (in vivo data) of a baboon brain. The histological volume appears in false color

orange in the fusion images with the MRI data. All results are presented in coronal, horizontal, and sagittal views as indicated. (Figure extracted from Dauguet et al., 2007)

correlate information from multiple image types, on a voxel-by-voxel basis, with submillimeter accuracy – even among imaging modalities with widely disparate spatial resolution and in the absence of identifiable anatomic landmarks. Meyer et al. (2006) present a method for registering histology and in vivo imaging that requires minimal microtoming and is automatic following the user’s initialization (Fig. 23.18). They describe an automated method that is relatively free from user bias in locating homologous points in that it is driven by maximization of mutual information, can operate on multimodal data, and can handle mixed combinations of 2D and 3D

datasets. They registered a single hematoxylin-andeosin-stained histological slide of a coronal section of a rat brain harboring a 9L gliosarcoma with an in vivo 7T MR image volume of the same brain. Because the spatial resolution of the in vivo MRI is limited, they add the step of obtaining a high spatial resolution, ex vivo MRI in situ for intermediate registration. The approach taken was to maximize mutual information in order to optimize the registration between all pairings of image data whether the sources are MRI, tissue block photograph, or stained sample photograph. The warping interpolant used was thin plate splines with the appropriate basis function for either 2D or 3D

342

A. Dubois et al.

Fig. 23.18 Final summary figure showing an hematoxylinand-eosin (HE)-stained section of a rat brain restored to blockface geometry (a) and registered to a slice of the in vivo MRI animal volume (b). The outlined regions represent identical vox-

els in each image. (c) Checkerboard verification of registration of in vivo MRI volume onto HE slice also restored to blockface geometry. (Figure extracted from Meyer et al., 2006)

applications. All registrations were implemented by user initialization of the approximate pose between the two data sets, followed by automatic optimization based on maximizing mutual information.

h igh-resolution histological information (Yelnik et al. 2007). It enables the validation and testing of imaging techniques processed on in vivo data (fMRI, DTI, and PET) by the confrontation with the gold standard embodied by the information derived from the actual tissues. It is also now possible to investigate new methods of analysis of 3D biologic data in groups of rodents, based on voxel-based statistical approaches and allowing rapid exploratory statistical studies of the whole brain without any anatomical a priori, complementarily with manual segmentation if needed (Schweinhardt et al. 2003; Nguyen et al. 2004; Lee et al. 2005a; Holschneider et al. 2006; Dubois et al. 2008; Holschneider and Maarek, 2008; Dubois et al. 2010). Finally, in vivo/postmortem correlation opens perspectives for the development and evaluation of new dedicated small animal imaging systems or the improvement of the former ones. For human data obtained by PET or fMRI, the planes are aligned at the time of acquisition. For postmortem data, the acquisition of all sections and the reconstruction of 3D volumetric images for each animal are first required. This requirement constitutes a major drawback, which restricts in vivo/postmortem matching to only specific studies. However, automated 3D reconstruction software and computerized procedures based on some of these registration methods have become available to the small animal postmortem imaging community (Diaspro et al. 1990; Lohmann et al. 1998; Thevenaz et al. 1998; Dubois et al. 2007; NIH-Image; MacPhase; VoxelView; 3D-BrainStation and SURFdriver). The automation of these tools and their integration into a graphical

23.5 Conclusion In conclusion, this chapter summarizes various methodological solutions proposed in literature in order to merge postmortem data sets with in vivo imaging. This allows to bridge the gap between fundamental biological research relying on postmortem microscopic material analysis and preclinical applications relying on in vivo. Since 2D brain slices (autoradiographs or histological stained sections) can be considered as imaging the ground truth, to merge them with in vivo 3D imaging modalities makes it possible to assess the reliability of in vivo measurements and to establish relationships between histological features and regional brain function. Pathological studies could also highly benefit from in vivo/postmortem matching. The longitudinal follow-up of in vivo imaging biomarkers in correlation with the “gold standard” of postmortem analysis could be very helpful to clarify disease mechanisms and to screen candidate treatments, guiding their selection for testing in expensive, time-consuming clinical trials. In vivo/postmortem correlation offers new opportunities for biomedical studies involving histology. Accurate 3D atlases representing the whole brain in the in vivo geometry can be created based on the

23 Ex Vivo and In Vitro Cross Calibration Methods

interface facilitate their use and make possible the simple, rapid, operator-independent prior processing of large data sets. These approaches have recently been extended to other organs, such as lung in human patients with lung cancer (Stroom et al. 2007; Dahele et al. 2008; van Baardwijk et al. 2008), or even to whole body mouse images (Dogdas et al. 2007). This opens new perspectives in another domain of application, oncology: (1) Stroom et al. (2007) evaluate the feasibility of pathology-correlated imaging for accurate target definition of lung tumors, taking into account lung deformations after surgery; (2) van Baardwijk et al. (2008) evaluated the feasibility to correlate intratumor heterogeneity as visualized on [18F]-FDG PET with histology and the possibility to investigate the histological background of this phenomenon; and (3) Dogdas et al. 2007 have constructed a 3D whole body mouse atlas from coregistered CT, PET, and cryosection data, whose potential applications include the automated labeling of anatomical structures and computer studies phantom to simulate small animal PET, SPECT, CT, or MRI imaging systems. Acknowledgments Authors would like to warmly thank Vincent Frouin for initiating dedicated small animal image processing in CEA/DSV/I2BM and Anne-Sophie Hérard, their favorite biologist, for believing in this from the beginning and always providing so beautiful postmortem data!

References Ali WS, Cohen FS (1998) Registering coronal histological 2-D sections of a rat brain with coronal sections of a 3-D brain atlas using geometric curve invariants and b-spline representation. IEEE Trans Med Imaging 17(6):957–966 Andreasen A, Drewes A, Assentoft J, Larsen N, Nielsen H, Holm I, Geneser F (1992) Computer-assisted three-dimensional reconstruction of the hippocampal region based on serial sections. J Neurosci Methods 37(2):151–160 Andrey P, Maurin Y (2005) Free-D: an integrated environment for three-dimensional reconstruction from serial sections. J Neurosci Methods 145(1–2):233–244 Annese J, Sforza DM, Dubach M, Bowden D, Toga AW (2006) Postmortem high-resolution 3-dimensional imaging of the primate brain: blockface imaging of perfusion stained tissue. Neuroimage 30(1):61–69 Ashburner J, Friston K (1997) Spatial transformation of images. In: Frackowiak R, Friston K, Frith C, Dolan R, Mazziotta J (eds) Human brain function. Academic Press, San Diego, pp 43–58

343 Ashburner J, Friston KJ (1999) Nonlinear spatial normalization using basis functions. Hum Brain Mapp 7(4):254–266 Bardinet E, Ourselin S, Dormont D, Malandain G, Tandé D, Parain K, Ayache N, Yelnik J (2002) Co-registration of histological, optical and MR data of the human brain. In: Dohi T, Kikinis R (eds) Medical image computing and computerassisted intervention (MICCAI’02). Vol. 2488 of LNCS. Springer, Tokyo, pp 548–555 Besl P, McKay N (1992) A method for registration of 3D shapes. IEEE Trans Pattern Anal Mach Intell 14(2):239–256 Bjarkam CR, Pedersen M, Sørensen JC (2001) New strategies for embedding, orientation and sectioning of small brain specimens enable direct correlation to mr-images, brain atlases, or use of unbiased stereology. J Neurosci Methods 108(2):153–159 Boyle JJ, Jenkins J, McKay IC, McPhaden AR, Lindop GB (1997) An assessment of the distortion of arteries due to sectioning in endomyocardial biopsies. J Pathol 181(2):243–246 Cachier P (2002) Recalage non rigide d’images medicales volumiques – contribution aux approches iconiques et geometriques. Ph.D. Thesis, Ecole Centrale des Arts et Manufactures Carlbon I, Terzopulos D, Harris M (1994) Computer-assisted registration, segmentation and 3D reconstruction from images of neuronal tissue sections. IEEE Trans Med Imaging 13(2):351–362 Chakravarty MM, Bertrand G, Hodge CP, Sadikot AF, Collins DL (2006) The creation of a brain atlas for image guided neurosurgery using serial histological data. Neuroimage 30(2):359–376 Chen X, Ai Z, Rasmussen M, Bajcsy P, Auvil L, Welge M, Leach L, Vangveeravong S, Maniotis AJ, Folberg R (2003) Three-dimensional reconstruction of extravascular matrix patterns and blood vessels in human uveal melanoma tissue: techniques and preliminary findings. Invest Ophthalmol Vis Sci 44(7):2834–2840 Chen F, Suzuki Y, Nagai N, Jin L, Yu J, Wang H, Marchal G, Ni Y (2007) Rodent stroke induced by photochemical occlusion of proximal middle cerebral artery: evolution monitored with mr imaging and histopathology. Eur J Radiol 63(1): 68–75 Cherry S, Shao Y, Silverman R, Chatziioannou A, Meadors K, Siegel S, Farquhar T, Young J, Jones W, Newport D, Moyers C, Andreaco M, Paulus M, Binkley D, Nutt R, Phelps M (1997) MicroPET: a high resolution PET scanner for imaging small animals. IEEE Trans Nucl Sci 44:1161–1166 Christensen G (1999) Consistent linear-elastic transformations for image matching. In: Information processing in medical imaging. Vol. 1613 of LCNS. Springer, Berlin, pp. 224–237 Christensen GE, Rabbitt RD, Miller MI (1996) Deformable templates using large deformation kinematics. IEEE Trans Image Process 5(10):1435–1447 Christensen G, Joshi S, Miller M (1997) Volumetric transformation of brain anatomy. IEEE Trans Med Imaging 16(6):864–877 Clatz O, Delingette H, Talos I-F, Golby AJ, Kikinis R, Jolesz FA, Ayache N, Warfield SK (2005) Robust non-rigid registration to capture brain shift from intra-operative mri. IEEE Trans Med Imaging 24(11):1417–1427 Cohen F, Yang Z, Huang Z, Nissanov J (1998) Automatic matching of homologous histological sections. IEEE Trans Biomed Eng 45(5):642–649

344 Collins D, Evans A (1998) In: Toga A (ed) Animal: automatic non-linear image matching and anatomical labeling. Academic Press, San Diego Collins DL, Neelin P, Peters TM, Evans AC (1994) Automatic 3d intersubject registration of MR volumetric data in standardized talairach space. J Comput Assist Tomogr 18(2): 192–205 Dahele M, Hwang D, Peressotti C, Sun L, Kusano M, Okhai S, Darling G, Yaffe M, Caldwell C, Mah K, Hornby J, Ehrlich L, Raphael S, Tsao M, Behzadi A, Weigensberg C, Ung YC (2008) Developing a methodology for three-dimensional correlation of PET-CT images and whole-mount histopathology in non-small-cell lung cancer. Curr Oncol 15(5):62–69 Dauguet J, Mangin J-F, Delzescaux T, Frouin V (2004) Robust inter-slice intensity normalization using histogram scalespace analysis. In: Barillot C, Haynor DR, Hellier P (eds) Medical image computing and computer-assisted intervention (MICCAI 2004). Vol. 3216 of LNCS. Springer, SaintMalot, pp. 242–249 Dauguet J (2005) L’imagerie post mortem tridimensionnelle cérébrale: constitution et apport pour l’analyse conjointe de données histologiques anatomo-fonctionnelles et la mise en correspondance avez l’imagerie in vivo. Ph.D. Thesis, Ecole Centrale de Paris Dauguet J, Delzescaux T, Condé F, Mangin J-F, Ayache N, Hantraye P, Frouin V (2007) Three-dimensional reconstruction of stained histological slices and 3D non-linear registration with in-vivo MRI for whole baboon brain. J Neurosci Methods 164(1):191–204 Davatzikos C (1997) Spatial transformation and registration of brain images using elastically deformable models. Comput Vis Image Underst 66(2):207–222 Delzescaux T, Foehrenbach H, Frouin V (2003) A performance study for whole-body helicoidal CT/PET-FDG a posteriori registration using rigid and non-rigid ffd-based methods. In: Proceedings of Society of Nuclear Medicine (SNM), Nouvelle-Orléans Deverell MH, Bailey N, Whimster WF (1989) Tissue distortion in three-dimensional reconstruction of wax or plastic embedded microscopic structures. Pathol Res Pract 185(5):598–601 Deverell MH, Salisbury JR, Cookson MJ, Holman JG, Dykes E, Whimster WF (1993) Three-dimensional reconstruction: methods of improving image registration and interpretation. Anal Cell Pathol 5(5):253–263 Diaspro A, Adami M, Sartore M, Nicolini C (1990) IMAGO: a complete system for acquisition, processing, two/threedimensional and temporal display of microscopic bio-images. Comput Methods Programs Biomed 31(3–4):225–236 Dogdas B, Stout D, Chatziioannou AF, Leahy RM (2007) Digimouse: a 3D whole body mouse atlas from CT and cryosection data. Phys Med Biol 52(3):577–587 Dubois A, Dauguet J, Hérard A-S, Besret L, Duchesnay E, Frouin V, Hantraye P, Bonvento G, Delzescaux T (2007) Automated three-dimensional analysis of histological and autoradiographic rat brain sections: application to an activation study. J Cereb Blood Flow Metab 27(10):1742–1755 Dubois A, Hérard A-S, Flandin G, Duchesnay E, Besret L, Frouin V, Hantraye P, Bonvento G, Delzescaux T (2008) Quantitative validation of voxel-wise statistical analyses of autoradiographic rat brain volumes: application to unilateral visual stimulation. Neuroimage 40(2):482–494

A. Dubois et al. Dubois A, Hérard AS, Delatour B, Hantraye P, Bonvento G, Dhenain M, Delzescaux T (2010) Detection by voxel-wise statistical analysis of significant changes in regional cerebral glucose uptake in an APP/PS1 transgenic mouse model of Alzheimer’s disease. Neuroimage 51(2):586–598 Ford-Holevinski TS, Castle MR, Herman JP, Watson SJ (1991) Microcomputer-based three-dimensional reconstruction of in situ hybridization autoradiographs. J Chem Neuroanat 4(5):373–385 Friston K, Ashburner J, Frith C, Poline J, Heather JD, Frackowiak R (1995) Spatial registration and normalization of images. Hum Brain Mapp 2:165–189 Gardella D, Hatton WJ, Rind HB, Rosen GD, von Bartheld CS (2003) Differential tissue shrinkage and compression in the z-axis: implications for optical disector counting in vibratome-, plastic- and cryosections. J Neurosci Methods 124(1):45–59 Gefen S, Tretiak O, Nissanov J (2003) Elastic 3-D alignment of rat brain histological images. IEEE Trans Med Imaging 22(11):1480–1489 Goldszal A, Tretiak O, Hand P, Bhasin S, McEachron D (1995) Three-dimensional reconstruction of activated columns from 2-[14C]deoxy-D-glucose data. Neuroimage 2(1):9–20 Goldszal AF, Tretiak OJ, Liu DD, Hand PJ (1996) Multimodality multidimensional image analysis of cortical and subcortical plasticity in the rat brain. Ann Biomed Eng 24(3):430–439 Hajnal JV, Saeed N, Oatridge A, Williams EJ, Young IR, Bydder GM (1995a) Detection of subtle brain changes using subvoxel registration and subtraction of serial mr images. J Comput Assist Tomogr 19(5):677–691 Hajnal JV, Saeed N, Soar EJ, Oatridge A, Young IR, Bydder GM (1995b) A registration and interpolation procedure for subvoxel matching of serially acquired mr images. J Comput Assist Tomogr 19(2):289–296 Hess A, Lohmann K, Gundelfinger E, Scheich H (1998) A new method for reliable and efficient reconstruction of 3-dimensional images from autoradiographs of brain sections. J Neurosci Methods 84(1–2):77–86 Hibbard L, Hawkins R (1988) Objective image alignment for three-dimensional reconstruction of digital autoradiograms. J Neurosci Methods 26(1):55–74 Hill DL, Batchelor PG, Holden M, Hawkes DJ (2001) Medical image registration. Phys Med Biol 46(3):R1–R45 Holschneider DP, Maarek J-MI (2008) Brain maps on the go: functional imaging during motor challenge in animals. Methods 45(4):255–261 Holschneider DP, Yang J, Sadler TR, Nguyen PT, Givrad TK, Maarek J-MI (2006) Mapping cerebral blood flow changes during auditory-cued conditioned fear in the nontethered, nonrestrained rat. Neuroimage 29(4):1344–1358 Houston GC, Papadakis NG, Carpenter TA, Hall LD, Mukherjee B, James MF, Huang CL (2001) Mapping of brain activation in response to pharmacological agents using fMRI in the rat. Magn Reson Imaging 19(7):905–919 Humm JL, Macklis RM, Lu XQ, Yang Y, Bump K, Beresford B, Chin LM (1995) The spatial accuracy of cellular dose estimates obtained from 3D reconstructed serial tissue autoradiographs. Phys Med Biol 40(1):163–180 Humm J, Ballon D, Hu Y, Ruan S, Chui C, Tulipano P, Erdi A, Koutcher J, Zakian K, Urano M, Zanzonico P, Mattis C, Dyke J, Chen Y, Harrington P, O’Donoghue J, Ling C (2003) A stereotactic method for the three-dimensional registration of

23 Ex Vivo and In Vitro Cross Calibration Methods multi-modality biologic images in animals: NMR, PET, histology, and autoradiography. Med Phys 30(9):2303–2314 Jacobs MA, Windham JP, Soltanian-Zadeh H, Peck DJ, Knight RA (1999) Registration and warping of magnetic resonance images to histological sections. Med Phys 26(8):1568–1578 Jones A, Milthorpe B, Howlett C (1994) Measurement of microtomy-induced section distortion and its correction for 3-dimensional histological reconstructions. Cytometry 15:95–105 Ju T, Warren J, Carson J, Bello M, Kakadiaris I, Chiu W, Thaller C, Eichele G (2006) 3D volume reconstruction of a mouse brain from histological sections using warp filtering. J Neurosci Methods 156(1–2):84–100 Karen P, Jirkovská M, Tomori Z, Demjénová E, Janácek J, Kubínová L (2003) Three-dimensional computer reconstruction of large tissue volumes based on composing series of high-resolution confocal images by glueMRC and linkMRC software. Microsc Res Tech 62(5):415–422 Kim B, Frey K, Mukhopadhyay S, Ross B, Meyer C (1995) Co-registration of MRI and autoradiography of rat brain in three-dimensions following automatic reconstruction of 2D data set. In: Ayache N (ed) Computer vision, virtual reality and robotics in medicine. Vol. 905 of LCNS. Nice, France. Springer, Berlin, pp. 262–266 Kim B, Boes J, Frey K, Meyer C (1997) Mutual information for automated unwarping of rat brain autoradiographs. NeuroImage 5:31–40 Kim T, Singh M, Sungkarat W, Zarow C, Chui H (2000) Automatic registration of postmortem brain slices to MRI reference volume. IEEE Trans Nucl Sci 47:1607–1613 Laan A, Lamers W, Huijsmans D, Kortschot AT, Smith J, Strackee J, Los J (1989) Deformation-corrected computeraided three-dimensional reconstruction of immunohistochemically stained organs: application to the rat heart during early organogenesis. Anat Rec 224:443–457 Lee E-F, Jacobs RE, Dinov I, Leow A, Toga AW (2005a) Standard atlas space for C57BL/6J neonatal mouse brain. Anat Embryol (Berl) 210(4):245–263 Lee J, Ahn S-H, Lee D, Oh S, Kim C, Jeong J, Park K, Chung J-K, Lee M (2005b) Voxel-based statistical analysis of cerebral glucose metabolism in the rat cortical deafness model by 3D reconstruction of brain from autoradiographic images. Eur J Nucl Med Mol Imaging 32(6):696–701 Lohmann K, Gundelfinger E, Scheich H, Grimm R, Tischmeyer W, Richter K, Hess A (1998) Brainview: a computer program for reconstruction and interactive visualization of 3D data sets. J Neurosci Methods 84:143–154 Maes F, Collignon A, Vandermeulen D, Marchal G, Suetens P (1997) Multimodality image registration by maximization of mutual information. IEEE Trans Med Imaging 16(2):187–198 Maes F, Vandermeulen D, Suetens P (2003) Medical image registration using mutual information. Proc IEEE 91(10): 1699–1722 Maintz J, Viergever M (1998) A survey of medical image registration. Med Image Anal 2(1):1–36 Malandain G, Bardinet E (2003) Intensity compensation within series of images. In: Ellis RE, Peters TM (eds) Medical image computing and computer-assisted intervention (MICCAI 2003). Vol. 2879 of LNCS. Springer, Montreal, pp 41–49 Malandain G, Bardinet E, Nelissen K, Vanduffel W (2004) Fusion of autoradiographs with an MR volume using 2-D and 3-D linear transformations. NeuroImage 23(1):111–127

345 Mandeville JB, Marota JJ, Kosofsky BE, Keltner JR, Weissleder R, Rosen BR, Weisskoff RM (1998) Dynamic functional imaging of relative cerebral blood volume during rat forepaw stimulation. Magn Reson Med 39(4):615–624 Mangin J-F (2000) Entropy minimization for automatic correction of intensity nonuniformity. In: IEEE Work. MMBIA. IEEE Press, Hilton Head Island, South Carolina, pp. 162–169 Mattes D, Haynor D, Vesselle H, Lewellen T, Eubank W (2003) PET-CT image registration in the chest using free-form deformations. IEEE Trans Med Imaging 22:120–128 Mega MS, Chen SS, Thompson PM, Woods RP, Karaca TJ, Tiwari A, Vinters HV, Small GW, Toga AW (1997) Mapping histology to metabolism: coregistration of stained wholebrain sections to premortem PET in alzheimer’s disease. Neuroimage 5(2):147–153 Meyer CR, Moffat BA, Kuszpit KK, Bland PL, Mckeever PE, Johnson TD, Chenevert TL, Rehemtulla A, Ross BD (2006) A methodology for registration of a histological slide and in vivo MRI volume based on optimizing mutual information. Mol Imaging 5(1):16–23 Morris PG (1999) Magnetic resonance imaging and magnetic resonance spectroscopy assessment of brain function in experimental animals and man. J Psychopharmacol 13(4):330–336 Nguyen P, Holschneider D, Maarek J, Yang J, Mandelkern M (2004) Statistical parametric mapping applied to an autoradiographic study of cerebral activation during treadmill walking in rats. NeuroImage 23(1):252–259 Nikou C, Heitz F, Nehlig A, Namer I, Armspach J (2003) A robust statistics-based global energy function for the alignment of serially acquired autoradiographic sections. J Neurosci Methods 124(1):93–102 Ourselin, S., 2002. Recalage d’images médicales par appariement de régions – application à la construction d’atlas histologiques 3d. Ph.D. Thesis, Université de Nice Sophia Antipolis Ourselin S, Roche A, Subsol G, Pennec X, Ayache N (2001) Reconstructing a 3D structure from serial histological sections. Image Vis Comput 19(1–2):25–31 Paulus M, Sari-Sarraf H, Gleason S, Bobrek M, Hicks J, Johnson D, Behel J, Thompson LH, Allen W (1999) A new X-ray computed tomography system for laboratory mouse imaging. IEEE Trans Nucl Sci 46:558–564 Pelizzari CA, Chen GT, Spelbring DR, Weichselbaum RR, Chen CT (1989) Accurate three-dimensional registration of CT, PET, and/or MR images of the brain. J Comput Assist Tomogr 13(1):20–26 Pitiot A, Bardinet E, Thompson PM, Malandain G (2006) Piecewise affine registration of biological images for volume reconstruction. Med Image Anal 10(3):465–483 Rangarajan A, Chui H, Mjolsness E, Pappu S, Davachi L, Goldman-Rakic P, Duncan J (1997) A robust point-matching algorithm for autoradiograph alignment. Med Image Anal 1(4):379–398 Roche A (2001) Recalage d’images médicales par inférence statistique. Université de Nice Sophia-Antipolis, Thèse de sciences Rueckert D, Sonoda LI, Hayes C, Hill DL, Leach MO, Hawkes DJ (1999) Nonrigid registration using free-form deformations: application to breast MR images. IEEE Trans Med Imaging 18(8):712–721

346 Rydmark M, Jansson T, Berthold CH, Gustavsson T (1992) Computer-assisted realignment of light micrograph images from consecutive section series of cat cerebral cortex. J Microsc 165(pt 1):29–47 Schmidt K, Smith C (2005) Resolution, sensitivity and precision with autoradiography and small animal positron emission tomography: implications for functional brain imaging in animal research. Nucl Med Biol 32(7):719–725 Schormann T, Zilles K (1997) Limitations of the principal-axes theory. IEEE Trans Med Imaging 16(6):942–947 Schormann T, Zilles K (1998) Three-dimensional linear and nonlinear transformations: an integration of light microscopical and MRI data. Hum Brain Mapp 6(5–6): 339–347 Schormann T, Dabringhaus A, Zilles K (1995) Statistics of deformations in histology and application to improved alignment with MRI. IEEE Trans Med Imaging 14:25–35 Schweinhardt P, Fransson P, Olson L, Spenger C, Andersson JLR (2003) A template for spatial normalisation of MR images of the rat brain. J Neurosci Methods 129(2): 105–113 Simonetti AW, Elezi VA, Farion R, Malandain G, Segebarth C, Rémy C, Barbier EL (2006) A low temperature embedding and section registration strategy for 3D image reconstruction of the rat brain from autoradiographic sections. J Neurosci Methods 158(2):242–250 Sorensen JC, Bjarkam CR, Danielsen EH, Simonsen CZ, Geneser FA (2000) Oriented sectioning of irregular tissue blocks in relation to computerized scanning modalities: results from the domestic pig brain. J Neurosci Methods 104(1):93–98 Streicher J, Weninger WJ, Müller GB (1997) External markerbased automatic congruencing: a new method of 3D reconstruction from serial sections. Anat Rec 248(4): 583–602 Stroom J, Blaauwgeers H, van Baardwijk A, Boersma L, Lebesque J, Theuws J, van Suylen R-J, Klomp H, Liesker K, van Pel R, Siedschlag C, Gilhuijs K (2007) Feasibility of pathology-correlated lung imaging for accurate target definition of lung tumors. Int J Radiat Oncol Biol Phys 69(1):267–275 Thanos PK, Taintor NB, Alexoff D, Vaska P, Logan J, Grandy DK, Fang Y, Lee J-H, Fowler JS, Volkow ND, Rubinstein M (2002) In vivo comparative imaging of dopamine D2 knockout and wild-type mice with (11)C-raclopride and microPET. J Nucl Med 43(11):1570–1577 Thevenaz P, Ruttimann UE, Unser M (1998) A pyramid approach to subpixel registration based on intensity. IEEE Trans Image Process 7(1):27–41 Thirion J-P (1998) Image matching as a diffusion process: an analogy with maxwell’s demons. Med Image Anal 2(3): 243–260 Thompson P, Toga A (1996) A surface-based technique for warping three-dimensional images of the brain. IEEE Trans Med Imaging 4:1–16 Timsari B, Tocco G, Bouteiller J, Baudry M, Leahy R (1999) Accurate registration of autoradiographic images of rat brain

A. Dubois et al. using 3D atlas. In: International conference on imaging science, systems, and technology Toga A, Arnicar T (1985) Image analysis of brain physiology. Comp Graph Appl 5(12):20–25 Toga AW, Arnicar-Sulze TL (1987) Digital image reconstruction for the study of brain structure and function. J Neurosci Methods 20(1):7–21 Toga AW, Ambach K, Quinn B, Hutchin M, Burton JS (1994) Postmortem anatomy from cryosectioned whole human brain. J Neurosci Methods 54(2):239–252 Toyama H, Ichise M, Liow J-S, Modell KJ, Vines DC, Esaki T, Cook M, Seidel J, Sokoloff L, Green MV, Innis RB (2004) Absolute quantification of regional cerebral glucose utilization in mice by 18F-FDG small animal PET scanning and 2-14C-DG autoradiography. J Nucl Med 45(8): 1398–1405 van Baardwijk A, Bosmans G, van Suylen RJ, van Kroonenburgh M, Hochstenbag M, Geskes G, Lambin P, Ruysscher DD (2008) Correlation of intra-tumour heterogeneity on 18F-FDG PET with pathologic features in non-small cell lung cancer: a feasibility study. Radiother Oncol 87(1):55–58 van den Elsen PA, Pol E-JD, Viergever MA (1993) Medical image matching-a review with classification. IEEE Eng Med Biol 12(1):26–39 Viola P, Wells WM (1997) Alignment by maximization of mutual information. Int J Comput Vis 24:137–154 Weber DA, Ivanovic M, Franceschi D, Strand SE, Erlandsson K, Franceschi M, Atkins HL, Coderre JA, Susskind H, Button T (1994) Pinhole SPECT: an approach to in vivo high resolution SPECT imaging in small laboratory animals. J Nucl Med 35(2):342–348 Wengenack TM, Jack CR, Garwood M, Poduslo JF (2008) Mr microimaging of amyloid plaques in Alzheimer’s disease transgenic mice. Eur J Nucl Med Mol Imaging 35(suppl 1): S82–S88 Williams BS, Doyle MD (1996) An internet atlas of mouse development. Comput Med Imaging Graph 20(6):433–447 Woods RP, Grafton ST, Watson JD, Sicotte NL, Mazziotta JC (1998) Automated image registration: II. Intersubject validation of linear and nonlinear models. J Comput Assist Tomogr 22(1):153–165 Yelnik J, Bardinet E, Dormont D, Malandain G, Ourselin S, Tandé D, Karachi C, Ayache N, Cornu P, Agid Y (2007) A three-dimensional, histological and deformable atlas of the human basal ganglia. I. Atlas construction based on immunohistochemical and MRI data. Neuroimage 34(2): 618–638 Zarow C, Kim T-S, Singh M, Chui HC (2004) A standardized method for brain-cutting suitable for both stereology and MRIbrain co-registration. J Neurosci Methods 139(2):209–215 Zhao W, Young T, Ginsberg M (1993) Registration and threedimensional reconstruction of autoradiographic images by the disparity analysis. IEEE Trans Med Imaging 12(4): 782–791

In Vitro Methods for In Vivo Quantitation of PET and SPECT Imaging Probes: Autoradiography and Gamma Counting

24

David Stout

Contents 24.1 Introduction............................................................. 347 24.2 Autoradiography..................................................... 348 24.2.1 Tissue Distribution of Radiolabeled Compounds..... 348 24.2.2 QWBA Methodology................................................ 349 24.2.3 Special Considerations for the Design of a QWBA Study..................................................... 352 24.2.4 Combination of QWBA with Other Imaging Approaches................................................. 352 24.3 Gamma Counters.................................................... 353 24.3.1 Principles of Detection............................................. 353 24.3.2 Geometric Efficiency................................................ 354 24.3.3 Factors Altering Measurement Values...................... 355 24.3.4 QC Measurements.................................................... 355 24.4 PET and SPECT Quantitation Issues................... 355 24.4.1 Factors Related to Image Data.................................. 355 24.4.2 Calibration Constant................................................. 356 24.4.3 Blood Sampling........................................................ 356 24.5 PET and Autoradiography Combined.................. 357 24.5.1 Working Fast: Using PET Isotopes with Autoradiography....................................................... 357 24.5.2 Quantitation Comparisons........................................ 358 24.6 Summary................................................................. 359 References............................................................................ 359

D. Stout Preclinical Imaging Center, UCLA Crump Institute for Molecular Imaging, 570 Westwood Plaza, CNSI room 2151, Los Angeles, CA 90095, USA e-mail: [email protected]

24.1 Introduction The development of noninvasive imaging methods such as positron emission tomography (PET) and single photon emission computed tomography (SPECT) has opened up tremendous opportunities to investigate biological function in vivo. These methods applied to small animals enable the whole body to be imaged over time, so the biodistribution of labeled probes can be followed within the same animal for hours or days (Phelps 2004; Rudin and Weissleder 2003). This is in sharp contrast to previous biodistribution methodology, where many animals had to be injected and groups sacrificed at various time points to see where the probe localized. The ability to observe changes within the same animal over time drastically reduces the number of animals and complexity of the experiment, while at the same time providing better temporal sampling and the ability to track individual metabolic function. One drawback of in vivo small animal PET and SPECT systems is the limited spatial resolution. When various factors such as blurring due to respiratory and cardiac motion, positron range, and image reconstruction methods are considered, most in vivo imaging systems have a resolution in the 1–2 mm range. Higher resolution is possible, but may be prohibitively expensive in terms of costs, time, radiation dose, and image noise (Cherry and Gambhir 2001). While this resolution is excellent for many purposes, it is not sufficient to discriminate exactly where the signal originated. Precise localization is particularly important in the development of new imaging probes, where the uptake, clearance, and metabolism must be understood in order to know exactly what information the images are providing. A good example of this is in Fig. 24.1, where the PET signal shows uptake in the bowel (Radu et al.

F. Kiessling and B.J. Pichler (eds.), Small Animal Imaging, DOI: 10.1007/978-3-642-12945-2_24, © Springer-Verlag Berlin Heidelberg 2011

347

348

S. David

Fig. 24.1 Cryosection photograph, autoradiography, and PET image of an F-18 imaging probe used in a mouse

2008). Only with the use of higher resolution autoradiography, could we see that the uptake was in the intestinal wall, rather than the bowel contents. This is important, since it means the probe was delivered by the bloodstream, not through hepatobiliary excretion through the liver and bile duct. The radiolabeling of compounds to assess tissue distribution is a common practice in research and drug development. The fate of radiolabeled probes can be determined by directly counting the radioactivity in dissected tissues using a gamma or liquid scintillation counting for gamma or beta-related isotopes, respectively. This methodology is known as “tissue dissection” or “cut and count.” For SPECT and PET probe biodistributions, the two most com mon methods used are gamma counting of tissues and autoradiography. Gamma counting is relatively simple, with individual samples placed in test tubes to measure radioactivity; however, this method typically only provides organ level information. Autor adiography is the process of visualizing radioactivity distributions in either tissue samples or over the whole body. Systems used for autoradiography can image over a range of resolutions, from 200 to 5 mm, thus activity distributions can be seen down to the cellular level.

24.2 Autoradiography 24.2.1 Tissue Distribution of Radiolabeled Compounds Autoradiography is used to determine the distribution of radiolabeled molecules from whole-body sections or specific tissues of interest by exposing the sample to an energy capturing matrix such as photographic film, phosphor-imagers, or emulsions. There are two types of autoradiography: whole-body and micro or tissue sectioning. Whole-body autoradiography is increasingly popular in the pharmaceutical industry as it allows an accurate evaluation of the tissue distribution in the intact animal. Whole-body sections also can be matched with in vivo imaging data. In addition to the qualitative assessment, it offers the capability of quantitating the concentrations of radioactivity in tissues by the use of standard curves, accounting for the commonly used terminology, quantitative whole-body autoradiography (QWBA). Micro-autoradiography (MAR) detects radioactivity at a microscopic level providing information about the cellular, subcellular, and topographical tissue deposition of the radiolabeled compound (Stumpf 2003) (Fig. 24.2).

24 In Vitro Methods for In Vivo Quantitation of PET and SPECT Imaging Probes: Autoradiography and Gamma Counting 349 Fig. 24.2 Digital images of sagittal whole-body sections of a male rat at different sectioning levels

The advantages of whole-body autoradiography relative to traditional cut-and-count strategies are several-fold. First, whole-body autoradiography allows for a comprehensive and quantitative anatomical survey of compound tissue distribution within the intact animal. This helps avoid the need to preselect tissues of interest and aids in the identification of distribution to unexpected tissues. Second, the methodology enables localization of compound in organ substructure, which can be important for better understanding compound mechanism of action and toxicity. Third, by virtue of this methodology, the animals are quickly frozen at the completion of prespecified time points, which immediately arrests potential compound redistribution. When performing cut-and-count, the time from time-point completion to tissue dissection is often variable. Lastly, this methodology allows for the in situ evaluation of placental transfer of a compound, melanin binding, and penetration of the blood/brain barrier. The key disadvantages of QWBA are that the methodology is laborintensive, expensive, and time-consuming.

24.2.2 QWBA Methodology Tissue distribution and fate of radiolabeled molecu les by QWBA were introduced by Ullberg (1954). Immediately after euthanasia, animals are pinned onto a stationary board and immersed in an alcohol bath at −70°C until completely frozen. Unpinned frozen animal carcasses are then embedded in 3% carboxymethylcellulose (CMC) and placed into a freezing chamber at −70°C. Whole-body cryosections are obtained by using a cryostat microtome at −20°C. The section thickness depends on the strength of the energy emitted by

the radioisotope in use, (i.e., 20 mm (3H, 125I) and 40 mm (14C)) (Fig. 24.3)). There are several sectioning levels that can be collected, each level representing the various major tissues (organs and fluids) (see Fig. 24.2, as an example). Sectioning levels to be collected are selected according to the tissues of interest. A digital picture of the level is taken before sectioning to be used as anatomical reference during imaging analysis (see Fig. 24.4). The collection of multiple sectioning levels not only permits for a wide tissue sampling, but also captures the possible heterogeneity of the distribution within the same tissue. A minimum of three sections per level allow for a statistically meaningful value at the time of quantitation. Although sagittal sections are the standard practice, other planes can be adopted to better represent the tissues of interest. As an example, Fig. 24.4 depicts coronal sections of a mouse bearing two tumors, in the left and right flanks, respectively. Unlike PET or SPECT, where the same animals can be observed over time, autoradiography measurements at different time points must be done using different animals for each sampling time. Whole-body sections are collected by placing a layer of clear tape on the animal, then cutting the tissue free, leaving the tissue mounted on the tape. The tissue sections are lyophilized at −20°C for 24–48 h before exposing them on film or phosphor-imaging plates (PIs) (Fig. 24.5). Suitable imaging plates for each specific isotope are commercially available and have replaced the X-ray films used in the past for autoradiography. The PIs are composed of phosphors designed to absorb radiation (Miyahara 1989; Hamaoka 1990). To avoid contamination of the plates, tissue samples can be covered with thin sheets of Mylar film so the tissue is not in direct contact with the plate.

350

S. David

Fig. 24.3 Diagram showing the steps required in a QWBA study, starting with injection of a radiolabeled probe, freezing and embedding, sectioning, exposure of whole-body sections, and image analysis for quantification

Fig. 24.4 Representative whole-body images and autoradiograms of mice bearing two tumors (arrowheads) at different time points after an IV bolus injection I-125-labeled SPECT

imaging agent. Uptake by tumors increased over time, showing specific localization of the probe

However, this step does not apply to samples containing 3H as the low energy radioactivity emission of this isotope requires direct contact of the tissue to the plate to avoid quenching of the signal.

For quantitation purposes there are two types of standards: a separate quantitation block and internal standards within the animal block. Polymer calibration standards are commercially available; however,

24 In Vitro Methods for In Vivo Quantitation of PET and SPECT Imaging Probes: Autoradiography and Gamma Counting 351

Fig. 24.5 Phosphor imaging plate and cassette (left), Fuji BAS 5000 plate reader (right)

standards prepared with biological matrixes are more reliable. Whole blood is the preferred matrix, as it is similar to most tissues in the body (Solon and Kraus 2002). Thus, several known concentrations of radioactivity are spiked into whole blood according to the concentration range expected in the whole-body samples. Blood standards are placed into drilled holes in the frozen CMC block and allowed to freeze at −20°C prior to sectioning. Quantitation standards are placed in an independent block and the section thickness should match that of the whole-body sections. The internal standard consists of few concentrations that are placed along the animal block (where the animal is embedded) and they are used as a parameter to determine the section thickness variations in any given section. Exposure time of the whole-body sections to the PIs is determined by the radioisotope used and the concentration of radioactivity administrated to the animals. In general, recommended exposures times are: 2–24 h for 125I, 124I, 64Cu, 18F, 11C, 3–8 days for 14C, and 15–30 days for 3H. Longer exposure times can improve detection of low radioactivity signals; however, caution should be applied as flare intensity, nonlinear response of high radioactivity signal, and increased signal-to-noise ratio may occur. Exposure conditions such as artificial light, environmental radiation, and temperature of the room where the tissue samples are exposed can affect the drug-related radioactivity detection. Because light and environmental radiation are absorbed by the PIs,

lead- and copper-lined cabinets are recommended to decrease background signals. Typically, exposure of sections to PIs at room temperature is sufficient. With short-lived PET isotopes, there is not sufficient time to dehydrate overnight so the exposure process must be conducted below freezing to prevent the tissues from thawing and loss of signal localization. This requires that the cassettes and the PIs be precooled to the exposure temperature to avoid water condensation (Maas et al. 2000) and to prevent the tissue sections from thawing. After exposure, PIs are scanned by a phosphorimager and the digital autoradiogram image is obtained. These digital images are then analyzed and can be quantified using various software programs available from several sources. A diagram of the steps involved in a typical QWBA study is shown in Fig. 24.3. The ability to quantify the radioactivity concentration in tissues has made QWBA an important tool in drug discovery and development. The quantitation is done by densitometry, pairing the optic density values of the standard to the corresponding radioactivity concentration values. A resulting standard curve converts the tissue density values to units of radioactivity concentration as described by Potchoiba et al. (1995). The quantitation method described above has been demonstrated to be as accurate and precise as the values obtained from the gamma or liquid scintillation counters (Potchoiba et al. 1995, 1998; Steinke et al. 2000; Busch et al. 2000). Results of tissue distribution by QWBA of a 14C-labeled molecule are demonstrated in Fig. 24.6.

352

S. David

Fig. 24.6 Representative whole-body rat autoradiograms at different time points (0.083, 1, 6, 12, and 120 h, respectively) after an IV bolus injection of C-14-labeled compound. Quantitation

of tissue radioactivity concentrations, converted to mEq/g of tissues, was obtained

24.2.3 Special Considerations for the Design of a QWBA Study

Another common artifact is caused by the conditions and length of the tissue exposure, as detailed above, which can result in poor signal-to-noise ratio or flare induced by very high radioactivity uptake in a given tissue. Flare will not only impact the ability to accurately quantitate radioactive signal in the tissue of origin, but also in surrounding tissues. In addition, tissues not properly frozen or tissues affected by surgical manipulation (e.g., bile, venous catheter; Zimmer 2007) can result in a false radioactivity signal. It is important to consider that incomplete erasing of residual radioactivity on plates will result in the so-called “phantom images” where new whole-body sections appear in the midst of old exposures. Plates are reusable if properly protected and blanked between uses. Given their high cost ($800–1,200), it is worth the effort to keep tissue and radiation from contaminating the surface. There are many challenges with QWBA experiments, so many that an entire society and conference has been created around this topic. For more information, refer to http://www.autoradiography.net/index.html

The radioactive concentration of the dose will depend on the isotope. Suggested doses are: 14C = 100–300 mCi/ kg, 3H = 1,000–2,500 mCi/kg, 125I = 100–800 mCi/kg. Either too high or too low radioactivity doses can impact imaging quality, quantitation, and overall outcome of the study (Maas et al. 2000; Potchoiba and Nocerini 2004). To accomplish the desired dose of the drug or test article, the total concentration can be adjusted by mixing radiolabeled with nonradiolabeled material. Caution must be exercised when the drug formulation contains any oil-based component as the oil may impair complete tissue freezing. Other considerations include the fact that when working with tumor-bearing models, the localization of the tumor tissue in the autoradiograms can be challenging (i.e., when working with orthotopic tumors). In this case, a complementary histological staining of adjacent whole-body sections may be a useful tool to discriminate normal from diseased tissue. Similarly, tumor detection can become more difficult in studies where drug administration affects tumor size. Lastly, artifacts can influence the outcome of QWBA data: one of the most common causes of misreading tissuerelated radioactivity is due to variations of the sectioning thickness within the whole-body section. Thus, internal standards placed in the animal block can help to identify thickness variations among the section.

24.2.4 Combination of QWBA with Other Imaging Approaches A multimodal approach to assessing distribution is often useful, as QWBA can complement in vivo imaging strategies such as PET or SPECT/CT. QWBA adds to

24 In Vitro Methods for In Vivo Quantitation of PET and SPECT Imaging Probes: Autoradiography and Gamma Counting 353

Fig. 24.7 Coronal views of whole body autoradiography (left), SPECT-CT (center), and SPECT (right) images of the same two tumor-bearing mouse at 96 h postdose of a 125I-labeled compound. Tumors are shown by the arrowheads

PET and SPECT/CT imaging by enabling an enhanced resolution of the distribution of radioactivity to tissues. An example of QWBA complementing SPECT/CT imaging is depicted in Fig. 24.7 (Pastuskovas et al. 2008).

24.3 Gamma Counters 24.3.1 Principles of Detection Gamma counters are sensitive instruments designed to measure very small quantities of radioactivity, usually less than 0.5 mCi. Unlike dose calibrators, which measure a wide range of activities and are proportional

gas detectors that measure a total signal, gamma counters are made of solid scintillator material, typically sodium iodide (NaI). Each gamma ray interacts within the scintillation crystal and creates many light photons, which are collected by a photomultiplier tube (PMT). The PMT has a photocathode that converts the light photons into electrons, then accelerates them using an electric field through a series of anodes to amplify the charge sufficiently for electronic processing. The electrical pulse is then processed using both hardware and software to determine a count rate. Software settings can be used to set energy thresholds and windowing, to identify certain isotopes and to eliminate spurious background noise (Knoll 1989). Each isotope has a characteristic emission of gamma rays, alpha particle or beta particles, either positrons or

354

electrons. The gamma counter is not able to detect alpha particles, as those cannot penetrate even the plastic sample tube. Beta particles are typically stopped within a few millimeters, so again the tube and housing for the crystal will stop those emissions. Gamma counters, as the name implies, are only suitable for detecting gamma emissions. The rays must be strong enough to penetrate into the scintillator and deposit energy, roughly a minimum of 25 keV. The ability of the detector to stop the gamma ray and deposit its energy is called the stopping power, which is determined by both the energy of the gamma rays and the density of the detector material. Over a certain energy threshold required to penetrate completely through the detector, as gamma energy increases, there will be less activity detected as more rays pass completely through the scintillator or only deposit some of their energy. This can be compensated by knowing the isotope, thus knowing the energy emissions and accounting for scintillator penetration using a calibration factor. For SPECT imaging, typical energies range from ~30 to 200 keV, with the most common isotope Tc-99 m at 140 keV. With PET isotopes, there are always 511 keV gammas from the annihilation process, which may also sum to 1,022 keV. Some PET isotopes may emit multiple types of radiation, such as both positrons and gammas. These “dirty” isotopes often deliver more dose to animals; however, they can be quite useful for their biological applications and longer half lives (Cu-64: 12 h, I-124: 4 days) compared to the more frequently used F-18 (2 h) and C-11 (20 min) isotopes. A reasonable trade-off for both cost and stopping power for NaI detectors is 2 in. (5 cm), which is suitable for both SPECT and PET isotopes. For a 5-cm NaI detector measuring PET isotopes at 511 keV, the detection efficiency is around 55%. To convert the counts measured to the true activity, this efficiency must be taken into account, converting counts per min to disintegrations per minute or Becquerel.

24.3.2 Geometric Efficiency Two types of detector configurations are possible: a pass through hole in the center of a cylinder and a cylinder with a hole going only part way through, commonly

S. David

Nal Scintillator PMT

Well Geometry

Pass-Through

Fig. 24.8 Two geometries used in gamma counters: well and pass-through. Some of the solid angle emitted from the blue source is not covered, shown in light blue. Light from the scintillation crystal is collected, converted to an electrical signal, and amplified by the photomultiplier tube (PMT)

called a well detector (Fig. 24.8). The pass-through design has an advantage for automated sample handling, since samples can be pushed or raised into a counting position, then lowered and moved aside to make room for the next sample. Well configurations have an advantage of more detector material, thus higher efficiency for stopping more activity. Unfortunately, well configurations often have problems with sample handling, since the sample must be held from the top, requiring very specific tubes and positioning. If anything drops out of the holder, it can be tricky to try and fish out things from the well. Sample positioning is a crucial consideration for any gamma counting work. Ideally, the sample should be placed in the center of the detector for a passthrough configuration or at the bottom of a well detector, where the maximum geometric efficiency is located for signal measurements. The solid angle of detector coverage directly affects the signal measured (Katz and Blantz 1972). As the sample moves toward the opening, the signal strength decreases and counts are lost, making the measurements inaccurate. For this reason, it is important to make sure all sample materials containing radiation are placed on the bottom of the test tube used to hold samples. Sample volume can also affect measurements through attenuation and displacement toward the top of the detector, where the geometric efficiency drops off. Attenuation is more important for lower energy isotopes such as Thallium, as they will be more subject to this effect compared to either PET or higher energy SPECT isotopes.

24 In Vitro Methods for In Vivo Quantitation of PET and SPECT Imaging Probes: Autoradiography and Gamma Counting 355

24.3.3 Factors Altering Measurement Values Apart from the geometric positioning of samples, there are several other factors that can alter the signals measured from the gamma counters. There are specific settings and counting protocols used that may specify certain energy windows, calibrations files used to correct measurements for specific isotopes and the duration of how long the samples are counted. The longer the samples are counted, the lower the noise will be in the measurements (Knoll 1989). Often there is a tradeoff in how long to count each sample vs. the half-life of the isotope and the potential for very low activity in some of the samples. The clock in the counter computer should also be configured to match those of other related imaging systems, so that the timing is precisely known and decay corrections are accurate. The ability to detect small or weak sources relies mainly upon the background radiation level where the counter is located. Ideally, the gamma counter is located well away from other sources of radioactivity, or at a minimum that other sources nearby be adequately shielded. A common rule of thumb is that samples are only significant if they are three times background, so the background level plays an important role in the data analysis. Commercial gamma counters are well shielded using a substantial amount of lead to reduce background. This shielding is good, but it can also make servicing and moving the system difficult. The maximum amount of activity that can be counted is vendor-specific, but is generally in the range of 0.5 mCi. High activity levels can saturate the electronics, causing system dead time and loss of counts. Systems typically have a software encoded limit of ~20% dead time and give some sort of error message rather than providing an inaccurate reading. One way to work with a wide range of activities is to have samples automatically counted numerous times until the activity decays to a working range. This enables measurements of both weak samples at early times and strong samples at a later time. Another factor is how samples are prepared and located in the counter. Blood or HPLC samples may be fractionated into individual samples or various tissues might be collected and placed in tubes for measurements. If a high activity sample is located immediately adjacent to a low activity sample, there may be shine

contamination from the hot sample to the cold one. A simple solution could be separating the samples by several spaces in the holding rack, to enable the shielding of the detector to function better.

24.3.4 QC Measurements Gamma counters should be checked periodically to make sure the response to a known standard is consistent with the expected measurement, corrected for decay. Various calibration standards for a range of energies are commercially available in vials suitable for either well or pass-through geometries. Most often Cs-137 and Co-57 are used for PET and SPECT energy ranges. For short-lived isotopes, the measurements can be compared to those obtained with a dose calibrator using either dilutions or waiting for decay to bring the sample activity down from the minimum accurate dose calibrator range (~10 mCi) to the reasonable range of the gamma counter (<0.5 mCi). Keep in mind that dose calibrators have a stated accuracy in the 10–15% range and that there are both instrument settings and geometric factors that also affect these measurements. Dilutions add another source of potential error; however, this is perhaps the only viable option for short-to-medium-lived isotopes (hours to days).

24.4 PET and SPECT Quantitation Issues 24.4.1 Factors Related to Image Data An in-depth review of PET and SPECT instrumentation and image reconstruction is beyond the scope of this chapter; nonetheless, there are some settings and choices that must be considered when comparing image data with gamma counter or autoradiography data. Foremost is that the image data must be as accurate as possible to account for any differences in how the data are measured. This means all the necessary correction factors must be applied, factors such as detector normalization, dead time correction, random count correction, isotope decay and decay scheme, attenuation and scatter corrections, partial volume

356

effects related to system resolution, and possibly probe metabolism. The amount of dose injected needs to be accurately known, accounting for any residual in the syringe or any activity stuck in the injections site that is not bioavailable. The choice of image reconstruction method can also change image values, depending on the filtering used, iterations, and the use of system performance parameters. Although the use of filtered back projection (FBP) is common, there is increasing use of iterative methods such as ordered subset maximum likelihood (OSEM) or for some systems maximum a posteriori (MAP), which makes use of the system response to activity to create higher resolution images (Chatziioannou et al. 2000). If an object is small relative to the resolution, roughly 3× the resolution or less in size, then there will be partial volume effects that blur the uptake into adjacent voxels, resulting in both a larger apparent size and a lower uptake measurement (Hoffman et al. 1979). For objects with mostly spherical shapes, such as tumors or the left ventricle, a recovery coefficient can be used to correct the activity level to account for the resolution loss.

24.4.2 Calibration Constant To compare image values and gamma counter results, the best method is to measure a known volume and amount of radioactivity in both the scanner and gamma counter. Knowing the true activity and measured counts, the gamma counter efficiency can be determined, which should be very stable over time. The corresponding measurement in the imaging system allows the system response to activity to be assessed and a calibration constant determined, which can be used to convert the counts per voxel per second to activity amounts. These two measurements can be done in one process, using a small vial approximating the size of a mouse (a liquid scintillation vial works well). By weighing the vial empty and full of a known amount of radiation, the concentration is known and can be compared to the image data. Aliquots of the vial can be carefully weighed and measured in the gamma counter to provide the counter efficiency measurement. Measurement of the efficiency and calibration constant is an excellent quality

S. David

Blood

Gamma Counter

Counts/sec

Gamma Ctr Efficiency

uCi

Tissue

microPET

Counts/voxel

Calibration Constant

uCi

Tissue

Autorad

Counts/pixel

Calibration Constant

uCi

Injected Dose

uCi

Fig. 24.9 To compare information from different sources, measurements need to be calibrated to the same unit of measure. Calibration constants and gamma counter efficiency are needed to compare PET, autoradiography, and gamma counter data

control measure to ensure that each system is functioning as expected, primarily because the process follows exactly that used to image animals and samples. The same process is used with autoradiography using known radioactivity standards or measured standards embedded within the sectioning block. Figure 24.9 shows how each type of measurement system has its own efficiency for data collection. Using the appropriate correction factor, the measurements can be converted into the same unit of measure, enabling comparisons between divergent data sets. There are many corrections and steps involved in this process, so careful measurements and reproducible protocols are essential. If the temperature and environmental conditions remain constant, these efficiency values should be very stable over time. When using these values with data, a running average is best to use to avoid the noise coming from any one measurement. In general, these systems should have a variability of less than 3–5%, thus any errors greater than this probably can be attributed to human measurements and might be best repeated.

24.4.3 Blood Sampling One place where gamma counters and SPECT or PET imaging often overlap is the measurement of blood or plasma radioactivity samples taken during the imaging process. The blood time activity profile is needed for kinetic modeling of metabolic rate constants. The ability to measure in vivo metabolism is the primary

24 In Vitro Methods for In Vivo Quantitation of PET and SPECT Imaging Probes: Autoradiography and Gamma Counting 357

purpose and most valuable application of SPECT and PET systems. Blood sampling is unfortunately not easy in small animals, especially mice (Laforest et al. 2005). While a simple poke of the tail may be good for obtaining a small amount of blood for a glucose measurement, a well-sampled blood time activity curve requires 15–20 samples, with accurate known timing and sufficient volume to enable handling, counting, and perhaps plasma separation and metabolite analysis. The most challenging part is sampling fast enough over the first few seconds to adequately sample the initial rise and fall of activity during the first pass extraction by the tissues (Raylman et al. 1993). Sometimes it is possible to use the imagederived input function using the left ventricle (Shoghi and Welch 2007); however, this approach may not work well for FDG in mice, where there is substantial myocardial muscle uptake over time, causing spillover of heart muscle signal into the late low activity blood pool values. Recently, several people have validated using both the LV and liver signals to get an accurate input function using FDG (Ferl et al. 2007). Others have worked on microfluidic blood sampling systems that can take relatively fast and tiny samples (Wu et al. 2007). If blood samples are to be used in a gamma counter, several factors must be considered. First is that an accurate sampling time must be recorded. Second is that the sample needs to be large enough to accurately weigh in order to get the volume. Most analytical balances are only useful to a few microliters, so it may be necessary to have 25 mL or more for the sample size. Third is that the samples must contain only blood and not any saline or other solution that might be flushed into the catheter between sampling to keep the line open and free from blood clots. A good clean sample requires first flushing with about three times the volume of the tubing and any valves before taking the sample. This flush can either be discarded or reinjected after the sample has been taken. For small animals, it is therefore very important to minimize the sampling system. Often a short piece of PE-10 tubing and a TB syringe with a fixed needle are used, since the dead volume associated with a Leur fitting and 3-way stopcock can easily add up to volumes over 500 mL. As mentioned previously, to get a good measurement of the sample, it must be placed on the bottom of the test tube used in the gamma counter in order that there be reproducible and optimal geometric efficiency.

24.5 PET and Autoradiography Combined 24.5.1 Working Fast: Using PET Isotopes with Autoradiography Conducting a SPECT or PET imaging session followed immediately by autoradiography is truly a logistical challenge. Each imaging method requires careful and timely attention to many details, so the combination of the two makes for an intensive experiment (Dogdas et al. 2007). Practice and experience are essential for success. The in vivo imaging is conducted first, then the animal must be quickly euthanized, frozen, cut, and the sections exposed to the imaging plates. Usually autoradiography experiments have the luxury of time, often 3 days or so to prepare everything for cutting. Timing only matters for the injection and subsequent sacrifice, since after freezing the tissues are stable and the half-lives of C-14 or tritium are measured in years. The most common isotope used with PET is F-18, with a 2-h half-life, so the entire process of autoradiography must be sped up accordingly. Figure 24.10 shows the linear autoradiography process and the corresponding times often required for each step. On the right are the times required for working quickly with F-18 in only a few hours. One group in Sweden has even sped this up for working with C-11, though the number of sections that can be imaged is very limited (Sihver et al. 1999). Something worth consideration when conducting both in vivo imaging and QWBA is that animals are typically anesthetized during the imaging process, often for an hour or more. The physiology and possibly the probe biodistribution may be different compared with animals injected and left conscious for the uptake period. One way to test this is to inject the probe into two groups: unconscious animals and conscious animals that are anesthetized only right before the imaging process. Comparing the resulting images will show the difference between anesthetized and awake animals. Another option is to inject two groups of animals, one left awake for QWBA work and another group anesthetized for in vivo imaging, though this will mean the in vivo image data will not necessarily match the QWBA data coming from different animals. Different anesthetics also have different effects in vivo, so the anesthetic choice

358

S. David

Fig. 24.10 Typical autoradiography processing times and steps required vs. the fast times required for working with short-lived isotopes. The minimal process here was optimized for F-18, with a 2-h half live

Typical Times Overnight in freezer Overnight in freezer 2-3 hours

Prep Animal Injection, uptake, sacrifice

Freeze Animal Positioning, cracking

Imbed in Block Ensure proper orientation

Freeze Block to Plate Reinforce front and back sides

Slice

Minimal Times <1 min using liquid freezing 15 min using liquid freezing ~1 hour

3 hours

Identify slices to take Thickness to cut?

1-2 hours

10 min each

Assemble Cassettes

2 min each

Keep slices frozen

3 days 1 week

Expose Plates

>4 hours Overnight

How long?

30 min each

Read Out Plates

30 min each

Resolution, Exposure?

may be necessary to consider during the experimental design phase. For long-lived isotopes, sections can be hung on a rack inside the cryostat and a defrost cycle run overnight to dehydrate the samples. Once dry, the tape can be moved to phosphor-imaging plates at room temperature for exposure. With short-lived isotopes, this is not an option, so the sections must be placed immediately on the PI plates. To keep the sections from melting and losing all the valuable positioning information, the PI plates and cassettes must be kept frozen. Exposure must be done in a −20 or −80 freezer, typically overnight for F-18. Condensation should be kept at a minimum to prevent water from collecting on any surfaces, which can degrade the image quality. Most of the signal on the PI plates will come from the positron, which only has a short range. The plates are relatively insensitive to the 511 keV gamma rays, so it is best to keep the sections as close as possible to the PI plates with only minimal material between them, usually just a thin layer of plastic wrap (Fig. 24.11).

24.5.2 Quantitation Comparisons Using the calibration constant and QWBA standards, the amount of radioactivity in tissues can be determined. The challenge then becomes trying to match up

Fig. 24.11 Sectioning of animal using F-18. Note that the phosphor plate and cassette are placed within the freezer compartment to keep the sections completely frozen after slicing

the locations from the two methods. A minor technical hurdle is that images are in different formats and will have different resolutions, slice thicknesses, and orientations. More problematic is that the freezing and slicing process deforms the tissues, so that they are no longer in the same orientation as in the in vivo imaging experiment. The mismatch in resolution may make image warping and coregistration difficult. Often QWBA experiments only look at specific sections within the animal to get a representative sampling of the activity. One option is to survey the entire animal; however, this is very time-consuming and may not be feasible given the short half life of the isotope. A

24 In Vitro Methods for In Vivo Quantitation of PET and SPECT Imaging Probes: Autoradiography and Gamma Counting 359

reasonable trade-off might be to take several back-toback slices to acquire data from a representative thick section of the tissue, which may give sufficient information for coregistration or cover the entire area of interest. Fiducial markers have also been used to help with this difficult process. Often tissue sections are stained for histology and pathology work. For whole-body QWBA, a problem is posed in that the sections are acquired directly on to tape as a support surface. If the tape does not stick to the section, it is then not useful as the tissues curl and fall off or get wrinkled. If samples stick properly, it is then hard to do the staining process. A solution has been devised where a degradable polymer is used as the tape, which is then dissolved by exposure to UV light. Using this process, it is possible to both image and stain the same section, rather than resorting to adjacent sections each going to one method (CryoJane tape transfer system, Instrumedics http://www.instrumedics.com/cryojane.htm). Fortunately, exact image coregistration is not often necessary. Looking back at Fig. 24.1, simply being fairly close in orientation and location is sufficient to derive useful information about the content of both in vivo and cryosection images. QWBA plus PET or SPECT is a challenging process, but fortunately one that might only need to be used sparingly to validate probe biodistributions, then perhaps no longer necessary at all or only as an end point to follow the same animal over time. The challenges of radiolabeling probes may prevent easy use of in vivo methods, plus the cost and complexity may not be warranted. QWBA remains a reasonable alternative, though one requiring more animals and a known optimal timing sequence for when to look at the biodistribution.

24.6 Summary By measuring the efficiencies of each radioactivity sensing device, the resulting data can be converted into a common unit of reference, usually either Becquerel or Curies. This conversion into the same unit makes it possible to compare values across different types of measurement systems, making use of the value of each system, whether it be high spatial resolution or in vivo measurements over hours or days. To be accurate, the various corrections and calibration factors must be

carefully measured and applied. Fortunately, regular assessments of the calibration factors are an excellent quality control measure and are relatively simple to obtain. Combining autoradiography with in vivo imaging is challenging but certainly possible, enabling more exact knowledge of the precise location of the imaging probe used in vivo.

References Busch U, Heinzel G, Nehmiz G (2000) Precision of measurement of tissue concentration by radioluminography. Regul Toxicol Pharmacol 31:S45–S50 Chatziioannou A, Qi J, Moore A, Annala A, Nguyen K, Leahy R, Cherry S (2000) Comparison of 3-D maximum a posteriori and filtered backprojection algorithms for high-resolution animal imaging with microPET. IEEE Trans Med Imaging 19(5):507–512 Cherry SR, Gambhir SS (2001) Use of positron emission tomography in animal research. ILAR J 42(3):219–232 Dogdas B, Stout D, Chatziioannou A, Leahy R (2007) Digimouse: a 3D whole body mouse atlas from and cryosection data. Phys Med Biol 52(3):577–587 Ferl GZ, Zhang X, Wu H, Huang SC (2007) Estimation of the 18F-FDG input function in mice by use of dynamic smallanimal PET and minimal blood sample data. J Nucl Med 48(12):2037–2045 Hamaoka T (1990) Autoradiography of a new era replacing traditional X-ray film. Cell Technol 9:456–562 Hoffman EJ, Huang SC, Phelps ME (1979) Quantitation in positron emission computed tomography: 1. Effects of object size. J Comp Asst Tomogr 3(3):299–308 Katz MA, Blantz RC (1972) Geometric error in tissue gammacounting: methods for minimization. J Appl Physiol 32(4): 533–534 Knoll G (1989) Radiation detection and measurement. Wiley, New York Laforest R, Sharp TL, Engelbach JA, Fettig NM, Herrero P, Kim J, Lewis JS, Rowland DJ, Tai YC, Welch MJ (2005) Measurement of input functions in rodents: challenges and solutions. Nucl Med Biol 32(7):679–685 Maas J, Binder R, Steinke W (2000) Quantitative whole-body autoradiography: recommendations for the standardization of the method. Regul Toxicol Pharmacol 31:S15–S21 Miyahara J (1989) Visualizing things never seen before. The imaging plate: a new radiation image sensor. Chem Today 223:29–36 Pastuskovas CV, Williams S, McFarland L, Khawli L (2008) A multimodal imaging approach for the characterization of antibody distribution in a preclinical model of mice bearing high and low Her2 expressing tumors. Abstract 1415, The Society of Nuclear Medine 55th Annual Meeting Phelps ME (2004) PET: molecular imaging and its biological applications. Spinger, Berlin Potchoiba MJ, Nocerini M (1998) Utility of whole-body autoradiography in drug discovery for the quantification of tritiumlabeled drug candidates. Drug Metab Dispos 26(3)

360 Potchoiba MJ, Tensfeldt TG, Nocerini MR, Silber BMA (1995) Novel quantitative method determining the biodistribution of radiolabeled xenobiotics using whole-body cryosectioning and autoradioluminography. JPET 272:953–962 Potchoiba MJ, West M, Nocerini MR (1998) Quantitative comparison of autoradioluminography and radiometric tissue distribution studies using carbon-14 labeled xenobiotics. Drug Metab Dispos 26:272–277 Potchoiba MJ, Nocerini MR (2004) Utility of whole-body autoradioluminography in drug discovery for the quantification of tritium-labeled drug candidates. Drug Metab Dispos 32(10):1190–1198 Radu CG, Shu CJ, Nair-Gill E, Shelly SM, Barrio JR, Satyamurthy N, Phelps ME, Witte ON (2008) Molecular imaging of lymphoid organs and immune activation using positron emission tomography with a new 18F-labeled 2¢-deoxycytidine analog. Nat Med 14(7):783–788 Raylman R, Caraher J, Hutchins G (1993) Sampling requirements for dynamic cardiac PET studies using image-derived input functions. J Nucl Med 34:440–447 Rudin M, Weissleder R (2003) Molecular imaging in drug discover and development. Nat Rev 2(123):123–131 Shoghi KI, Welch MJ (2007) Hybrid Image- and blood-sampling (HIBS) input function for quantification of microPET data. Nucl Med Biol 34(8):989–994 Sihver S, Sihver W, Bergström M, Höglund AU, Sjöberg P, Långström B, Watanabe Y (1999) Quantitative autoradiog-

S. David raphy with short-lived positron emission tomography tracers: a study on muscarinic acetylcholine receptors with n-[11C]methyl-4-piperidylbenzilate. Pharmacology 290(2): 917–22 Solon EG, Kraus L (2002) Quantitative whole-body autoradiography in the pharmaceutical industry. Survey results on study design, methods, and regulatory compliance. J Pharmacol Toxicol Methods 46:73–81 Steinke W, Archimbaud Y, Becka M, Binder R, Busch U, Dupont P, Maas J (2000) Quantitative distribution studies in animals: cross-validation of radioluminography versus liquid-scintillation measurements. Regul Toxicol Pharmacol 31:S33–S43 Stumpf WE (2003) Drug localization in tissues and cells. Receptor microscopic autoradiography. International Institute of Drug Distribution, Cytopharmacology and Cytotoxicology (IDDC) Press Ullberg S (1954) Studies on the distribution and fate of 35S-labeled benzylpenicillin in the body. Acta Radiologica Supplementum 118:1–110 Wu HM, Sui G, Lee CC, Prins ML, Ladno W, Lin HD, Yu AS, Phelps ME, Huang SC (2007) In vivo quantitation of glucose metabolism in mice using small-animal PET and a microfluidic device. J Nucl Med 48(5):837–845 Zimmer M (2007) Catheter rats: influence of surgery quality on distribution pattern. Abstract, The Society for Whole-Body Autoradiography, 2007 Meeting

Part Data Postprocessing

V

Qualitative and Quantitative Data Analysis

25

Felix Gremse and Volkmar Schulz

Contents 25.1 Common Image Properties.................................... 363 25.1.1 Image Structure......................................................... 363 25.1.2 Image quality............................................................ 364 25.1.3 Partial Volume Effect................................................ 364 25.2 Modality Characteristics........................................ 365 25.2.1 Magnet Resonance Imaging..................................... 366 25.2.2 Computed Tomography............................................ 366 25.2.3 Fluorescence-Mediated Tomography (FMT)............ 366 25.2.4 Fluorescence Reflectance Imaging (FRI)................. 366 25.2.5 Ultrasound................................................................ 368 25.2.6 Positron Emission Tomography................................ 368 25.2.7 Single Photon Emission Computed Tomography ............................................................. 369 25.2.8 Multimodal Imaging Devices................................... 370 25.3 Image transformations........................................... 370 25.3.1 Image Resizing......................................................... 370 25.3.2 Affine Transformations............................................. 371 25.3.3 Image Registration.................................................... 371 25.4 ROI Definition and Analysis.................................. 372 25.4.1 Geometric ROIs and Segments................................. 372 25.4.2 Manual Definition of Geometric ROIs..................... 373 25.4.3 ROI Transfer............................................................. 373 25.4.4 ROI Evaluation......................................................... 373 25.5 Segmentation........................................................... 373 25.5.1 ISO Surfaces............................................................. 373 25.5.2 Automated Segmentation.......................................... 374

F. Gremse (*) Experimental Molecular Imaging, University of Aachen (RWTH), Pauwelsstraße 20, 52074 Aachen, Germany e-mail: [email protected] V. Schulz Department Molecular Imaging Systems, Philips Research Europe – Aachen, RWTH/UKA Aachen, 52074 Aachen, Germany

25.6 Statistical Tests........................................................ 374 25.6.1 Principle of Statistical Tests...................................... 374 25.6.2 Parametric and Nonparametric Tests........................ 376 25.6.3 Power Analysis......................................................... 376 25.6.4 Tests Against Zero Mean or Median (Paired t-Test and Wilcoxon Signed Rank Test)..................................................... 376 25.6.5 Tests for Group Difference (t-Test, U-Test, and Kolmogorov-Smirnov Test).................. 376 25.6.6 Tests Against Normality........................................... 377 25.6.7 Multiple Comparisons.............................................. 377 25.6.8 Tests for Dependence Between Measurements (Linear Correlation, Rank Correlation)..................................................... 377 References............................................................................ 378

25.1 Common Image Properties 25.1.1 Image Structure Nearly, all preclinical and clinical image data sets are stored in a rasterized format, i.e., intensities are sampled using a regular grid. In this chapter, the terms “image data set” and “data set” are used to cover rasterized data sets of all dimensions. In the following, the most common image data sets of different dimensionalities are listed. • 1D: a sequence of values, e.g., over time or along a line through a data set of a higher dimensionality. • 2D: images, but also slices through higher dimensions. • 2D over time or “cineloop”: sequence of 2D images over time. Each 2D image is called a frame. • 3D: volumetric data set, i.e., a stack of 2D slices. • 4D: 3D data sets over time. Each 3D data set is called a phase.

F. Kiessling and B.J. Pichler (eds.), Small Animal Imaging, DOI: 10.1007/978-3-642-12945-2_25, © Springer-Verlag Berlin Heidelberg 2011

363

364

A voxel is the generalization of a pixel for higher dimensions than two. The following is a list of the properties that should be available for the image data set for quantitative analysis. • The dimensionality, i.e., 2D, 3D, 2D over time, 4D (3D over time). • The number of voxels in each dimension (x, y, z, and t). • The data type, e.g., unsigned char, short, int, float, float2. Vectors like float2 are used for data sets with multiple channels. The advantage of float over short or int is that float has a higher precision and does not require a scaling parameter. • The unit of the elements, e.g., in arbitrary units (a.u.), (pmol/mm³). • The voxel size for each dimension in a metric unit, e.g., in mm for spatial dimensions or seconds for temporal dimensions. This is required to compute metric values, e.g., length, area or volume, of a region of the image data set. • The voxel intensities. Other application-specific properties like the type of capturing device, date of acquisition, patient name, capturing protocol, room temperature, etc. are called meta data, which are stored in the DICOM (digital imaging and communications) file format for example.

25.1.2 Image quality The image quality of a medical imaging technique significantly depends on the physical property being measured. For example, as computed tomography (CT) is measuring the absorption of gamma photons (40– 140 keV), highly absorbing structures like cortical bone appear bright while low absorbing ones like air appear darker. In contrast, magnet resonance imaging (MRI) is measuring the net response of exited spins, which leads to low image intensities for both bone and air. These imaging modality-dependent contrasts have to be taken into account and can significantly influence study results, e.g., tumor volume measurements. Thus, a basic understanding of the physical principles of the involved modalities is essential, especially when modalities are being compared. This is even more important when anatomic images, e.g., from MRI, X-ray, CT, or ultrasound (US), are compared with functional or metabolic images, e.g., from positron emission tomography (PET), single photon emission computed tomography (SPECT), functional MRI (fMRI), or Magnetic Particle

F. Gremse and V. Schulz

Imaging (MPI). Some of the imaging modalities, like FMT, PET, MPI, and CT, claim to be quantitative in the way they produce the voxel intensities, and they display the tracer concentration in the final image. For those images, threshold-based techniques to extract image features are very easy to apply. However, especially when contrast agents for MRI and CT or radioactively labeled tracers are used (as in PET and SPECT), the preparation and execution of the experiment significantly influence the image contrast in a quantitative way. In addition, every modality that acquires data is affected by noise that significantly influences the image features. To quantify the usefulness of an image, two features are often used: the signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR). The SNR is defined as mean signal intensity m divided by the standard deviation s of areas where only noise is to be expected, e.g., outside the object. SNR a , c =

m (a ) s(c )

The CNR is defined as the contrast difference of two regions normalized by the noise level. CNR ( a ,b|c ) =

m (a) - m (b) = SNR a , c SNR b , c s (c )

The areas a, b, and c are defined by the user, e.g., rectangular regions in 2D slices. a and b should contain the targeted image feature, while c should be an area that contains only image noise. Obviously, the larger the sizes of the areas a, b, and c are, the more stable the calculation of m and s is. In general, SNR and CNR should be maximized under the given acquisition conditions, e.g., acquisition time (for MRI, US, PET, SPECT) or radiation dose (for CT, SPECT, and PET). Both SNR and CNR increase proportional with the voxel volume and the square root of the acquisition time, which finally limits the useful resolution of an image. Normally, the vendors deliver tools with the imaging device that supports the extraction of these values.

25.1.3 Partial Volume Effect Image data sets have a limited resolution, which is defined by the voxel size. Small details or sharp edges cannot accurately be represented with a finite resolution. Real objects being imaged often have sharp

365

25 Qualitative and Quantitative Data Analysis

contrast borders, e.g., between cortical bone and tissue and therefore would ideally require an infinite resolution. Therefore, voxels at such borders end up with a mixed value of multiple tissues, which is called partial volume effect (PVE). This can be seen in MR imaging for example when a large slice thickness is used as shown in Fig. 25.1. It is important to distinguish between the digital resolution of an image data set, i.e., the voxel size, and the effective resolution that can be achieved by an imaging device. The effective resolution is limited by technical constraints such as sensor size, aperture size, motion blur, size of a “point” light source, or the mathematical reconstruction algorithm and is typically coarser than the digital resolution. Seen from the other side, the digital resolution should be higher than the effective resolution to not lose any valuable information. The common blurring effect caused by the physical constraints can be thought of as a smoothing operation that is applied to the true image during the process of acquisition. The result of scanning a (theoretically infinitely) small object is called the “point spread function” (PSF) of a device. For many devices, it can be approximated by a Gaussian filter whose full width at half maximum (FWHM) is a good indication of the effective resolution. This limit of resolution through physical constraints is often also called PVE; however, “intensity diffusion” would be a better term (Skretting 2007). It has a larger relative effect on small objects, which will

Fig. 25.1 Partial volume effect (PVE) and “intensity diffusion”: Several T1-weighted slices of a murine kidney are shown on the left four images. Due to the high slice thickness of 1 mm compared to the pixel size of 0.2 mm, strong PVEs are visible. The sharp transition between kidney and surrounding fat is not captured at some round regions. The right image shows a CT slice

appear larger or may be smoothed away completely into the background. In particular, in combination with thresholding, this will lead to an under-or overestimation of the size of small objects. If the PSF of a device is known, this can be used to improve the effective resolution. Unfortunately, this technique is limited by the image noise, and furthermore, some modalities have a position-dependent PSF, e.g., leading to higher blurring at the borders of the field of view (FOV). The quantitative evaluation of a ROI (region of interest) also has an aspect to the PVE. ROIs that have a geometrical representation, e.g., a sphere, have a partial overlap with voxels on the boundaries as shown in Fig. 25.2. Care should be taken to how a program evaluates the sum of intensity over such an ROI. Those partially overlapping voxels should only be considered as much as they are included in the ROI. Otherwise, a systematic bias will be created, particularly for small ROIs.

25.2 Modality Characteristics Not all imaging modalities provide signal intensities that can be used directly for quantitative measurements. In the following, the most common imaging modalities are described with respect to quantitative analysis.

of a mouse finger bone reconstructed with a voxel size of 15 mm. The transition between bone and tissue is sharp in reality which is not reflected in the image due to the effect of intensity diffusion, which goes beyond the PVE. The effective resolution is higher than 100 mm

366

F. Gremse and V. Schulz

cortical bone, soft-tissue, and air. 4D data sets can be acquired for kinetic measurements. The typical unit of the voxel intensities is Hounsfield Units (HU), which is normalized using the following formula. HU tissue =

Fig. 25.2 The PVE can bias measurements such as sum, mean, or volume of an ROI if the program does not handle this properly. Shown is an image that was enlarged so that the quadratic pixels are visible. A circular ROI partially overlaps the pixels at the border. These pixels should only contribute proportionally to their overlap to the sum, mean, and volume

25.2.1 Magnet Resonance Imaging (MRI) T1- or T2-weighted MRI images provide signal intensities that not only depend on the T1 and T2 relaxation times but also on the proton density and the position in the coil. The intensity is typically higher in the center of the coil and falls off on the outside. Therefore, the signal intensities do not represent the actual T1 or T2 relaxation times. To estimate them, a relaxometry protocol is required, which makes several images (e.g., 3 or 7) with varying TE and TR and then estimates the T1 and T2 times. These calibrated T1 and T2 values can then be used for quantitative assessment of specific contrast enhancement. It is also helpful for segmentation because the intensity now depends only on the tissue and not on the position in the coil anymore. Dynamic contrastenhanced MRI imaging (DCE-MRI) uses a fast sequence of 2D or 3D images during and after a bolus injection of a contrast agent. From the time course of the intensities, quantitative values such as fractional blood volume in [%] or perfusion in [min−1] can be derived.

25.2.2 Computed Tomography (CT) X-ray-based CT imaging provides information about the X-ray absorption coefficients of tissue. A 3D data set is reconstructed that provides good contrast between

m tissue - m water 1,000 HU . m water

The advantage is that the HU are independent of the X-ray energy used, because in the formula, the coefficient m water is specific to the energy. Water, therefore, always ends up with HU of 0 while air has a value of −1,000 HU. Unlike with MRI signal intensities, the HU do not depend on the position of the voxel, which allows easy threshold-based classification of tissue into cortical bone, soft-tissue (water), fat, and air. Water, cerebrospinal fluid (CSF) CSF, muscle, and most tissue of the inner organs differ very little in HU contrast, making them hardly distinguishable without contrast agents. Most soft-tissue types have around the same HU as water, while fat is around 100 HU lower (Fig. 25.3).

25.2.3 Fluorescence Reflectance Imaging (FRI) FRI cameras acquire 2D images of fluorescent reemission of photons through a narrow-banded light filter. Example images are shown in Fig. 25.4. FRI is very useful to quickly assess or monitor the presence of fluorescence; however, it is hardly quantitatively useful. Typically, multiple channels, e.g., 3, can be used simultaneously to acquire a multichannel image. The appearance of a fluorescent region depends not only on the amount of fluorescence but also on the depth which makes it hard to quantify the amount based on the pixel intensities. Many devices store images that were normalized relative to the maximal brightness which means that the absolute intensity value is lost. Furthermore, the pixel size cannot be associated uniquely to a metric unit, e.g., [mm], because of the perspective transformation inherent in the image capturing process.

25.2.4 Fluorescence-Mediated Tomography (FMT) Fluorescence Mediated Tomographs (FMT) FMT capture 2D images of the absorption and reemission of fluorescent contrast agents, usually in combination with a

25 Qualitative and Quantitative Data Analysis

367

Fig. 25.3 CT slices of a mouse heart using a blood pool contrast agent. The reconstruction kernel is varied from top to bottom (sharp to smooth), while the amount of projections used for reconstruction is increased from left to right from one to four full rotations. The choice of the kernel allows a trade-off between

noise and blurring, which is shown in each column. The best choice depends on the task, e.g., classification of voxels into bone, muscle, blood, and air. Threshold-based segmentation is very sensitive to the amount of noise present in an image

spacially varying light source for excitation. Wavelengths in the visible light spectrum are used, preferably near the infrared spectrum because it penetrates through greater tissue depths than shorter wavelengths such as green light. From these 2D images, a 3D data set is reconstructed, which provides the concentration of fluorescence in pmol mm−3 at each voxel. FMT is sensitive to

very low concentrations, but the effective resolution is typically around 1 mm voxel size. This technology along with many specific contrast agents is commercially available for mice and used in many research facilities. The device can also be used as an FRI camera. Drawbacks of the technology are that the current reconstruction algorithms make the assumption of uniform tissue that fills

368

F. Gremse and V. Schulz

Fig. 25.4 Fluorescence reflectance imaging (FRI) can be used to quickly assess the presence of fluorescence. Both images show subcutaneously implanted tumors, which were genetically altered to express green and red fluorescent proteins. The left image shows a single red channel, whereas the right image shows both channels. Colocalization of red and green areas is displayed in yellow. In the right image are two small tumors,

one deep below the surface, resulting in a large red blob because the green light was absorbed through the large depth. The little yellow blob is a tumor shallow under the skin. The positions of the tumors were verified with T2-weighted MR images. FRI images are hardly quantitatively useful because the intensity depends on the size, depth, and wavelength

out the whole mouse bed, which in practice leads to inaccuracies in the image data set. Prototype FMT+CT scanners have been developed, which use the anatomical information for an improved FMT reconstruction (Schulz et al. 2010).

By exploiting the Doppler effect, blood flow velocity (mm s -1 ) in the direction towards the scan head can be measured. The contrast mode, which makes use of nonlinear resonance of microbubbles, provides signal intensities that depend on the concentration of microbubbles. Fig. 25.5 shows several example images.

25.2.5 Ultrasound (US)

25.2.6 Positron Emission Tomography (PET) Preclinical US scanners use scan heads with higher frequencies than clinical scanner because higher resolutions can be achieved and less depth penetration is required. Anatomical scans, called B-Mode, provide signal intensities in arbitrary units, which depend on the sound-reflectance properties of the tissue. Cineloops of high frequencies such as 300 fps are possible, which is particularly useful in combination with cardiac gating. By mounting the scan head on a servo, 3D scans can be acquired. These anatomical scans allow measuring anatomical parameters such as tumor size, vessel diameter, or even the size of atherosclerotic plaques.

Compared to anatomic imaging modalities like MRI, CT, X-ray, or US, the PET image represents a quantitative measure of the radioactive tracer distribution c(t,n) per voxel n at a given time t. The unit of this voxel’s intensity is typically [kBq/mL]. The image quality of PET depends critically on a number of parameters like ligand, radioisotope, applied dose, time between injection and scan, scan duration, image acquisition parameters, reconstruction parameters, object weight, and size. In one way or another, these parameters should

Fig. 25.5 Shown left is an ultrasound image of the aortic arch with a plaque in the left common carotid artery. The intensity of the pixels along the dotted line is concatenated over time in the

right image. This allows estimating the plaque thickness at several time points and averaging them for a more accurate measurement

369

25 Qualitative and Quantitative Data Analysis

always be mentioned for comparison or assessment of the images. Generally, the PET image suffers from missing anatomical information, especially when tracers with very specific binding are used. It is therefore common to fuse PET images with anatomic modalities like CT or MRI, using software or hardware registration. If the structures being investigated are rather rigid, like for the head, software registration might be sufficient. Furthermore, a quantitative PET image requires the use of a proper scatter and attenuation correction map, which is mostly derived from the preceding transmission Ge-68 or a dedicated CT scan. Other anatomical images could also be used to generate these attenuation maps, e.g., MRI-based attenuation correction. The correction is of less importance for smaller objects, like mice. The absolute value of the tracer concentration is a function of input tracer dose d and weight w of the object under investigation. Therefore, the semiquantitative measure standardized uptake value (SUV), sometime also referred to as differential uptake ratio (DUR), is used, which is defined as follow: SUV(n) =

w * c(n) . d

There are other definitions that are using the body surface area (BSA) or lean body mass (LBM). In order to archive comparable results within a study, it is very essential to take the images always at the same time point (Boellaard et al. 2004). Fortunately, it is not required to use blood sampling, which is one of the reasons that SUV is very popular. If blood samples have been measured, a simple but more quantitative alternative to SUV is Fractional Uptake Rate (FUR). FUR and SUV are proportional, related by plasma clearance rate and a dimensionless initial distribution volume (Thie 1995). Image noise, poor resolution, and ROI definition affect the SUV and may hamper their use, especially in multicenter trials (Boellaard et al. 2004). The effective image resolution, not to be mixed with the voxel size, is commonly specified in terms of fullwidth-half-max (FWHM) of a point or a line source surrounded by water using filtered back projection (FBP) reconstruction. Generally, the resolution of an image is affected by the scanner design (scintillator crystal dimensions, gantry dimension, etc), by the object size, by the object position, by the count statistics per voxel, and on the positron range of the used radioisotope. For any commercially available PET systems, this image resolution is not constant in the FOV. In a detector ring arrangement (which is the standard geometry in PET

systems), the resolution is highest in the center region, while it degrades as a function of distance to the center. In other words, the studies’ organ should always be close to the central axis of the PET scanner. Very important and thus relevant for image parameter extraction is the positron range of the tracer being used. For example, a scanner with a spatial resolution of 1.7 mm using F-18 or Cu-64 will show a resolution of 4.3 mm with Rb-82 or I-120 (Palmer et al. 2005). Iterative reconstruction techniques allow in principle increasing the image resolution, which unfortunately could end up in an increased background noise and thus on a low CNR. Fig. 25.6 shows typical PET- CT images.

25.2.7 Single Photon Emission Computed Tomography (SPECT) In contrast to PET, SPECT measures individual gamma photons in the energy range of 40–450 keV (sometimes even 600 keV), which are emitted from radiopharmaceuticals. Similar to PET, SPECT is preferably used to visualize functional and metabolic information and also morphological or organ specific information. In

Fig. 25.6 PET and PET/CT images of a mouse after injection of the bone tracer 18F-fluoride. Curtesy of P. Marsden, Division of Imaging Sciences, King’s College London

370

difference to PET, a SPECT camera is able to detect more radiopharmaceuticals when they emit at different energy levels. The SPECT image is stored using the [kBq/mL]. As for PET, the relevant parameters of the investigation, like ligand and radioisotopes, dose, time between injection and scan, object weight, object size, scan duration and image acquisition parameters, reconstruction parameters, and collimator, determine the image quality and should therefore always be mentioned. In order to archive quantitative SPECT data, similar normalizations on input dose and body weight as for PET should be aimed for but are not common today. However, there are efforts to propose a similar measure like SUV for SPECT. Quantitative SPECT imaging requires attenuation correction, scatter correction and a compensation of spatially dependent sensitivity (called normalization). Thus, SPECT is very often combined with CT in a hybrid system. For imaging of small animals, pinhole collimation offers distinct advantages over conventional parallel-hole collimation. Nevertheless, it is common to use a proper collimator for the different radioisotopes emitting at different energies. Thus, the resolution of a SPECT system is determined by the radioisotope with corresponding collimator and via the distance of the object and the detector. Image segmentation, geometric parameter extraction, and temporal investigations are more robust if 3D acquisition and reconstruction are used. Compared to PET, the acquisition time of SPECT image could easily cover 45–60 min. Thus, the proper animal handling significantly influences the quantitative nature of the image. Fig. 25.7 shows typical SPECT-CT images.

F. Gremse and V. Schulz

Fig. 25.7 Mouse scan. Left: SPECT only; right fused SPECT and CT. Mouse: 35 g, 30 × 90 mm, 1mCi Tc99m-MDP, 30 min. scan, 1.5 h p.i., 36 pinholes, Ø = 1.0 mm, scan range 110 mm. Curtesy of M. de Jong, ErasmusMC Rotterdam, Netherlands

25.3 Image transformations In many situations, an image transformation such as upsizing, downsizing, rotation, or elastic transformation is necessary for processing, registration, or visualization. Several problems can arise with respect to a quantitative evaluation, in particular with respect to the unit of voxel intensities and volume computations.

25.2.8 Multimodal Imaging Devices More and more facilities use multimodal imaging devices or a combination of devices, e.g., CT+FMT, PET+CT, PET+MRI, MRI+FMT, or SPECT +CT. The advantage is that the molecular data of the PET, SPECT, or FMT scanners can be better associated to anatomical regions, which can be better localized with CT and MRI. Multimodal imaging devices can use more accurate hardware registration because the setup and position of the animal are known. Software-based registration commonly requires manual interaction for registration, which may be hard to be used routinely. Moving the animal between devices without changing the position is also a challenge and requires animal fixation on dedicated beds.

25.3.1 Image Resizing Handling of large data sets, which are common for preclinical MR or CT scanners, may be slow or even prohibited by the available memory of the computer or operating system. This may require a downsizing step before the analysis. Integer factors such as 2, 3, or 4 are preferable since they are handled better and faster by many programs. If a 3D image with 1,000 × 1,000 × 1,000 voxels is downsized to 500 × 500 × 500 voxels, then the memory footprint is reduced to 1/23, while the voxel size increases by a factor of 2. Downsized images often have a better SNR; however, this depends on the type of

371

25 Qualitative and Quantitative Data Analysis

noise and the effective resolution. Proper algorithms should be used for resizing to minimize the loss of quality. Upsizing is generally possible with marginal or zero reduction in quality.

25.3.2 Affine Transformations During analysis, image data sets might have to get resized, rotated, translated, flipped, or more generally transformed by an affine transform. Registration of scans from two different devices can also be achieved through an affine transform. Any combination of rotation, translation, isotropic scale, anisotropic scale, and shear transformations can be combined into one affine transformation, which can be expressed with 12 parameters in 3D. An anisotropic scaling transformation allows different scaling factors for each dimension. Any combination of rotations, translations, and isotropic scaling results in an RST (rotate, scale, translate) transform, which is a special affine transform (i.e., the set of RST transformations is a subset of the set of affine transformations). RST transforms require seven parameters and preserve angles. This can be further restricted to a rigid body transformation, which consists only of rotation and translation and requires six parameters in 3D. Aligning two partial scans from one device might require only a translation, which is also a special affine transformation. These four groups, affine transformation, RST, rigid transformation, and translation, are closed under composition, which means that any sequential composition of two or more transformations results in a transformation of the same group. An affine transformation can be inverted, if it is not degenerate.

Fig. 25.8 Marker-based alignment of CT and FMT images. A phantom was used with a cylindrical cavity (yellow) that was filled with a fluorescent agent. Three markers were selected in each image by pointing and clicking. The average registration

Affine transformation and subsequent resampling of an image data set may result in a loss of quality, particularly if rescaling to smaller sizes is involved. This effect adds up when many operations are done consecutively, e.g., through interactive rotation and translation. Therefore, most programs keep the image data sets in their original spaces and apply the transformation only for visualization and evaluation of ROIs. Geometrically defined ROIs can be transformed from one coordinate system to another without loss of accuracy using an affine transformation.

25.3.3 Image Registration Rigid body transformations are commonly used for registration of multimodal images because the scaling factor is known and determined by the voxel size. Affine transformations or RST transformations should be avoided because they would estimate a different scaling factor than the one that is given by the voxel sizes and they are not angle preserving. The parameters of the transformation are estimated using markers or some minimization criterion based on the pixel intensities. Markers are selected by the user through pointing and clicking. For a rigid body transformation, three markers are required in each of the two data sets. The parameters are then estimated by minimizing the squared distances. To get an accurate transformation, markers that are as distant as possible from each other and are not positioned along a line should be chosen. Selecting more than three markers further reduces the alignment error. Figure 25.8 shows the software registration of FMT and CT image data sets of a phantom. The average registration error is

error is approximately one FMT voxel (1 mm). It can be optimized by setting up more markers and positioning them as far as possible from each other. Registration was performed with the free software amide

372

region-dependent and can be estimated using phantom studies or approximated from the markers (Fitzpatrick et al. 1998). It is called target registration error (TRE) as opposed to the fiducial registration error (FRE), which is the minimization criterion when computing the transformation parameters. Many registration transformations other than the rigid body transformation have been proposed (Hutton et al. 2006). The main problem with other transformations, e.g., ones that include shears or local stretches, is that the sum of intensity of a region is changed through the transformation. This can be compensated by adapting the intensity values but is rarely considered. Unless you are sure that this is done properly, you should only use the rigid body transformation when you plan to extract quantitative results from registered image data.

25.4 ROI Definition and Analysis ROI stands for region of interest, meaning a sub region of an image data set of which measurements such as diameter, volume, sum, mean, or max of intensity are interesting.

Fig. 25.9 Several simple ROIs are shown. (a) The upper torso of a mouse was scanned in an FMT after injection of a catepsinactivated fluorescent contrast agent. (b) A box-shaped ROI was manually edited to encapsulate a bright region. (c, d) Elliptic

F. Gremse and V. Schulz

25.4.1 Geometric ROIs and Segments Two different types of ROIs are commonly used, geometric ROIs and segments. Geometric ROIs are defined by geometric primitives such as ellipses or polygons. A 3D sphere, for example, is defined as the set of points within a certain distance from a center point. This sphere has four parameters, i.e., three for the center point and one for the radius. Some simple geometric ROIs are shown in Fig. 25.9. Complex ROIs can be represented by lists of vertices and connections between them as shown in Fig. 25.10. A segment is a subset of the voxels of the image data set. Segments are defined along pixel boundaries, making them dependent on a given resolution. They typically are digitally represented as a binary mask or an integer mask if multiple segments are required. A classification of voxels into segments is called a segmentation. If every voxel is assigned to exactly one segment, it is called a partition. Both geometric ROIs and segments have advantages over the other, and most algorithms require either one of them, which makes a conversion necessary. This conversion is not trivial however and might require high amounts of

and cylindrical ROIs. (e) Slice of registered FMT and CT data sets. An elliptic ROI is set around an axillary lymph node. (f) Line measurements in a CT image of mouse paw bones

373

25 Qualitative and Quantitative Data Analysis Fig. 25.10 Left: A subcutaneous tumor on the hind leg of a mouse appears bright in a T2-weighted image. The ROI was drawn with a few mouse clicks. Right: A murine heart ventricle imaged with a gated 3D ultrasound scan was extracted by manually drawing contours in several slices and then letting the program interpolate for the other slices

memory and CPU time for complex segments, which might crash or stall programs that do not properly handle this.

25.4.2 Manual Definition of Geometric ROIs Many programs allow manual definition of simple geometric ROIs like rectangles, boxes, ellipses, spheres, or cylinders as shown in Fig. 25.9. Even though a line is not a region, it is also called a ROI. These simple ROIs are typically entered by the user through clicking and dragging until covering the desired region. More complex 2D forms are usually entered by clicking at some points at the border of the ROI. 3D ROIs can be entered in some programs by drawing 2D shapes in a few reference slices, which are then interpolated for the rest of the slices.

25.4.4 ROI Evaluation Once a ROI is defined, interesting values such as the volume or the sum of intensity can be extracted. For image sequences, the time course of the value is often relevant. Many programs allow exporting this data as a text file for further spreadsheet computations. When generating quantitative values such as the total tracer uptake or the sum of fluorescence over an ROI, the unit should be carefully considered. The unit of a 3D molecular FMT is [ pmol mm-3] and not -1 [ pmol voxel ] because then a resizing operation would require changing the voxel intensities. The unit of the sum over an ROI then is [pmol], whereas the unit of the mean is [ pmol (mm) - (-3)] . Similar considerations apply to PET and SPECT images. Some devices, e.g., US, store the logarithm of the intensities in order to reduce the quantization error. Those values have to be converted before computation of measurements.

25.5 Segmentation 25.4.3 ROI Transfer ROIs that are defined in the coordinate system of one device can be warped to the coordinate system of another device if the affine transformation between them is known, e.g., through markers. Geometric ROIs can be transformed without loss of accuracy by transforming their parameters, e.g., an ellipse remains an ellipse after affine transformation; only the center point and shape are changed.

Sophisticated segmentation methods usually operate on segments, i.e., sets of voxels, rather than geometric ROIs because the former can be manipulated more efficiently.

25.5.1 ISO Surfaces Geometric shapes like ellipses, boxes, or polygons strongly depend on the user’s input and therefore easily lead to a subjective bias in the evaluation. Often, the

374

F. Gremse and V. Schulz

simple shapes don’t fit the object well enough either. ROI definition through ISO surfaces is a method that generates more complex ROIs with less user interaction and less subjective bias. The user has to supply a threshold, and the software computes the region with intensity above this threshold. The points on the surface of this region all have the same intensity as the threshold, therefore the name ISO surface or ISO contour. The region is often split into several disconnected parts, which the user would have to disambiguate if he is interested in only one of them. If this is done by selecting one seed point before computation of the ISO surface, this is called region growing. Volume renderers can be used to visualize ISO surfaces by setting voxels with intensity above the threshold to opaque and below the threshold to transparent. Surface renderers require a conversion of the ISO surface into a geometric representation before rendering. Examples for both are shown in Fig. 25.11. The threshold of an ISO surface still has to be provided by the user, and this should be done consistently throughout an experiment. Therefore, a common approach is to set the threshold to a fraction of a local maximum, e.g., 50% or 30%, which is commonly used for estimating the tumor size in clinical PET images. Unfortunately, the intensity diffusion during image acquisition leads to a systematic overestimation of the size and underestimation of the maximum intensity value of small objects. Figure 25.12 shows a simulation of this effect, assuming a Gaussian PSF and Gaussian-shaped objects. Noise in the image leads to a higher variance and therefore to higher local maxima, which causes higher thresholds if they are selected relative to a local maximum. Therefore, noise leads to a systematic underestimation of the volume of small objects.

25.5.2 Automated Segmentation

Fig. 25.11 An ISO surface consists of the points that have an intensity value equal to a threshold provided by the user. The ROI is defined as the set of points with higher intensity than the threshold, i.e., within the ISO surface. The left two images are

volume renderings of CT data sets. The right two images are surface renderings of a data set acquired with an optical tomograph. Thresholds were set relative to a local maximum

In a typical experiment, many data sets of equal type are generated for which the same analysis needs to be performed. If a human operator would have to manually analyze and export the interesting features, this would be limited to simple operations. Also, the analysis is not reproducible, error prone, and could be severely biased. Instead, the effort could be justified to write a program to perform batch analysis with a structured export of intermediate and final data. Once such a program is available, it can be reused as many times as needed and is free of subjective bias, which increases the credibility of the results. Figure 25.13 shows a fully automated segmentation of a mouse captured with a SPECT/CT. An overview over the technology used for these results is given in (Baatz et al. 2009).

25.6 Statistical Tests In the following, an overview over the most common statistical tests in the context of small animal imaging is given. Details can be found in, e.g., Sachs (1984).

25.6.1 Principle of Statistical Tests The purpose of a statistical test is to turn measurements of an experiment into a clear statement and estimate a probability falsely accepting the statement. The

375

25 Qualitative and Quantitative Data Analysis

Ratio of estimated vs.true size

.... 8 1D 2D 3D

4

2

1

0

1

2

3

4

5

6

7

8

9

10

Ratio of object width to point spread function width

Fig. 25.12 Small objects are systematically overestimated when their size is estimated through threshold-based region growing. This is simulated by assuming Gaussian-shaped objects and a Gaussian point spread function (PSF) with full width at half max (FWHM) = 1. The upper three graphs show three blue “objects” with FWHM of 1, 2, and 3. They are all smoothed with the PSF with FWHM of 1 resulting in the green “objects.” When applying a threshold of 50% relative to the max to estimate the size, the purple bars would result, which systematically overestimate the

“true” size shown by the red bars. The effect is stronger for smaller objects. The lower graph shows how the overestimation depends on the object size. The curves were computed using two facts: When convoluting two Gaussians, another Gaussian results with a variance equal to the sum of the two variances. The Gaussian FWHM is proportional to the standard deviation parameter of the Gaussian, i.e., FWHM = 2 2in(2)s » 2.35482s

Background Mouse Skeleton Oral Cavity Trachea Right Lobe Left Lobe Brain Bone Inflammation Arthritis Bladder Gall Bladder Intestine Right kidney Left kidney

Fig. 25.13 Fully automated segmentation of a mouse scanned with a SPECT/CT device using the Tc99m-GCA tracer. The algorithm was developed by (Baatz et al. 2009) and used on 11

mice to compute volume and tracer activities of regions such as skeleton, brain, kidney, lungs, and bladder

statement is that a so-called null hypothesis can be rejected. The term null hypothesis means that there is no effect, e.g., no difference between two groups or no correlation between two sets of measurements. The most common null hypothesis is that there is no difference between the means of two sample sets, e.g., measurements of control and therapy groups. The words rejecting, accepting, proving, and disproving are all

relative to an error rate a in this context. To estimate this error rate is the responsibility of the test. If a is below a significance level of, e.g., 5%, the test result is called significant; the null hypothesis is then said to be rejected with p < 5%. Typical significance levels are 10%, 5%, and 1% of which 5% is the most common. A significant result is typically indicated by a star in a chart.

376

It is important to understand that a high a, i.e., an insignificant result, does not mean that the null hypothesis can be accepted; it just means that the test failed to reject the null hypothesis. However, by performing a power analysis, a relaxed version of the null hypothesis can be proven. Each test implies an alternative hypothesis, which is the opposite of the null hypothesis.

25.6.2 Parametric and Nonparametric Tests Statistical tests can be based on assumptions about the distributions of the samples, e.g., the t-test assumes that the samples are distributed according to a normal density. Tests that assume parameterized densities such as the normal density with its parameters mean and variance are called parametric tests. Other tests are called nonparametric. For most parametric tests exists a nonparametric analog test, e.g., the U-test for the t-test. If the assumptions of a parametric test are not met, this may lead to an unrealistical a. Nonparametric tests often convert measurements into ranks by sorting them by the magnitude. Each measurement is then associated to its rank in the sorted sequence. By this conversion, the information about the magnitude of the samples is lost; tests that apply such a rank conversion are therefore less powerful than their parametric counterparts whose assumptions are met, i.e., require more samples to generate a significant result. On the other hand, these tests become more robust to outliers, making them more trustworthy if they succeed at rejecting the null hypothesis.

25.6.3 Power Analysis If a test has an insignificant result, i.e., failed to prove an effect, this may be because there really is little effect or because there are not enough samples. Without knowing which of these two reasons caused the result, the test does not provide any useful information yet. Therefore, insignificant results are published less often, which is called the publication bias. A power analysis however quantifies which of these two reasons is more likely and can therefore extract useful information from this test. It computes the minimal effect size D that should lead to a significant result with a low error

F. Gremse and V. Schulz

rate b. Then, the true effect is less than D with error rate b, i.e., a relaxed null hypothesis is “proven.” Power analysis can also be performed before an experiment to estimate how many samples are needed for a significant result given an assumed effect size D.

25.6.4 Tests Against Zero Mean or Median (Paired t-Test and Wilcoxon Signed Rank Test) These tests are called paired tests because they can be used on paired data, i.e., when two measurements are available for each subject and their difference is the quantity of interest. The powerful and often-used paired t-test was published by William Gosset in 1908 under the synonym “Student” because using statistics was considered a trade secret by his employer. The t-test assumes that the samples of a group are normally distributed. The null hypothesis is that the mean of the density is zero, therefore the name null hypothesis. The alternative hypothesis is therefore that the mean is different from zero. The test statistic t of the test is the ratio of estimated mean and the estimated standard error of the mean multiplied by N , i.e., N m / S . It is distributed according to the t-distribution, which is parameterized by the number of observations N. Values far from zero are unlikely under the null hypothesis and lead to a rejection. The exact rejection thresholds for a significance level of, e.g., 5% can be looked up in tables in many books (Sachs 1984) or computed by statistics software. The t-test is a parametric test because it assumes a normal density. The Wilcoxon Signed Rank test is less powerful than the t-test but more reliable if the assumption of a normal density is not met. It is the nonparametric analog to the t-test. The samples are sorted and converted to ranks on which the test statistic is computed. The null hypothesis is that the median of the samples is zero.

25.6.5 Tests for Group Difference (t-Test, U-Test, and KolmogorovSmirnov Test) These tests can be used to prove the difference between sample distributions of two groups. The tests are called unpaired because they require independent samples of each group; the number of samples for each group

377

25 Qualitative and Quantitative Data Analysis

does not have to be the same however. The most powerful test in this category is Student’s t-test for unpaired samples. Its null hypothesis is that both groups have the same mean. Assumptions are that the samples of both groups are normally distributed with the same variance. In praxis, this assumption can rarely be proven, and it has become common praxis to use the t-test anyways and leave the final judgment to the reader. The t-test is known (Sachs 1984) to be robust to deviations from these assumptions, except when the group with less samples has a lower variance. One Way Analysis of Variance (Anova) is an extension of the t-test for more than two groups. Anova for two groups is equivalent to the t-test and should therefore be called t-test because it is more common to more people. The U-test is the nonparametric analog to the t-test and operates on ranks; in fact, it simply converts the measurements into ranks before applying the t-test. The null hypothesis is that both groups have the same distribution. The alternative hypothesis is that the distribution is different, which implies that both samples have different means and medians. The U-test does not assume normal densities but requires that both groups’ densities have the same shape, i.e., are identical except for a shift. The U-test requires more samples than the t-test for a significant result if the assumptions of the t-test are met, but when it succeeds, it is more reliable. An extension of the U-test to three or more groups is the Kruskal-Wallis test. The U-test can be used in combination with a test against normality, i.e., if there is evidence that the normal assumptions are violated, the U-test should be used instead of the t-test. Many other names are common for the U-test, such as WilcoxonMann-Whitney test, Mann-Whitney- Wilcoxon test, Mann-Whitney-U test, Mann-Whitney test, or Wilcoxon rank sum test. The Kolmogorov-Smirnov test is even more general than the t-test and the U-test. The null hypothesis is that both groups have different distributions. In contrast to the other tests, this test does assume that both groups have the same shape of distribution; the only assumption is that the distributions are continuous. Therefore, this test is also called nonparametric. While this test requires more samples than the other tests, it is also much more reliable if it succeeds.

25.6.6 Tests Against Normality The t-test assumes normal densities. This assumption is often justified, especially if the measurement

is computed by averaging several random values. The central limit theorem states that when averaging N samples of any distribution, the resulting distribution becomes more and more similar to a normal distribution with increasing N. If nothing is known, the normal distribution is the best bet. The Kolmogorov-Smirnov test can be used as a test against normality, instead of comparing two distributions, a set of samples is compared with a fixed normal distribution. In this case, the Lillifors correction can be applied to improve the power. However, this test still requires many samples for a significant result. It is important to understand that this test will never tell that something is normally distributed; it can only provide evidence against it.

25.6.7 Multiple Comparisons Special care is advised, if many pairs of sample sets are compared, because for any comparison, a false positive has the chance to slip through. A simple and conservative method to adjust the alpha values is the Bonferroni correction, which uses the significance level divided by the number of comparisons as the threshold for rejection. If different tests are applied, the outcome of all tests should be conveyed to the reader. Trying all tests and reporting the lowest p-value leads to overconfident results.

25.6.8 Tests for Dependence Between Measurements (Linear Correlation, Rank Correlation) These tests analyze bivariate measurements, i.e., experiments where paired values are measured and a dependency between the paired values needs to be proven. Measurements are statistically dependent if your chances at guessing one can be improved knowing the other, which means that there has to be some kind of connection between them. A typical example is to compare noninvasive measurements with results from histology from the same subject to show that one is related to the other or can even be replaced if the dependency is strong enough. Linear correlation is a special case of dependence. In order to visualize the correlation of samples, the

378

F. Gremse and V. Schulz y = 0.813x + 0.0988 R2 = 0.63 P = 0.033 0,2

y = 1.0145x + 0.1195 R2 = 0.75850 P = 0.011 0,2

0,2

0,15

0,15

0,15 0,1 y

0,1

y

0,1

y

y = 1.133x + 0.0799 R2 = 0.944 P = 0.000248 0,25

0,05 0,05

0,15

0,1

0,05

0

0,05

0 02 0

0,05

0,15

0,1 0,05

0,1

0.05 x

0

0,05

01

0,05

x

0,1

0,2

0,1

0 0,05

0

0,1

02

0,1 0,15

x

Fig. 25.14 Each image shows artificially generated correlated samples. The strength of correlation is visible by how close the points are to the line fit. For highly correlating samples (right), a stronger significance can be achieved with fewer samples

bivariate measurements can be plotted as shown in Fig. 25.14. If the point cloud is close to a diagonal line, the samples are correlated linearly. The strength and direction of correlation are captured by the correlation coefficient r, which is also called Pearson’s r. r is a number between −1 and 1, where the sign and magnitude indicate the direction and magnitude of correlation respectively. r2 is called the coefficient of determination; if r2 is 1, then there is a direct functional dependency between the measurements. Some examples are shown in Fig. 25.14. The null hypothesis is that there is no correlation, i.e., r = 0. The significance a is a function of r² and the number of measurements and can be looked up in a table. If measurements correlate stronger, less samples are required for a significant a. The values of r2 and a are the same if x and y axes are swapped. Linear regression fits a linear curve (y = ax + b) to the data cloud. a and b are computed by minimizing the squared distances in y direction between the sample points and the line. Different parameters will be computed if x and y are swapped, i.e., the results are not symmetric in x and y. This linear curve can be used to extrapolate values. Spearman’s Rank correlation first converts the measurements into ranks by sorting them and then performs linear correlation on the ranks. The resulting coefficient is called Spearman’s r or rs. This test is less powerful than linear correlation if the samples are binormally distributed. On the other hand, it is more robust to outliers and deviations from the assumptions.

References Baatz M, Zimmermann J, Blackmore CG (2009) Automated analysis and detailed quantification of biomedical images using definiens cognition network technology. Comb Chem High Throughput Screen 12:908–916 Boellaard R, Krak NC, Hoekstra OS, Lammertsma AA (2004) Effects of noise, image resolution and ROI definition on the accuracy of standard uptake values: a simulation study. J Nucl Med 45:1519–1527 Eckelman WC, Tatum JL, Kurdziel KA, Croft BY (2000) Quantitative analysis of tumor biochemistry using PET and SPECT. Nucl Med Biol 27:633–635 Fitzpatrick JM, West J, Maurer CR (1998) Predicting error in rigid-body, point-based registration. IEEE Trans Med Imaging 17:694–702 Hutton BF, Braun M, Slomka P (2006) Image registration techniques in nuclear medicine imaging. In: Quantitative analysis in nuclear medicine imaging. Springer, New York Palmer MR, Zhu X, Parker JA (2005) Modeling and simulation of positron range effects for high resolution PET imaging. IEEE Trans Nucl Sci 52(5):1391–1395 Sachs L (1984) Applied statistics: a handbook of techniques. Springer, Berlin Schulz R, Ale A, Sarantopoulos A, Freyer M, Soehngen E, Zientkowska M, Ntziachristos V (2010) Hybrid system for simultaneous fluorescence and X-ray computed tomography. IEEE Trans Med Imaging 29:465–473 Skretting A (2007) Intensity diffusion’ is a better description than ‘partial volume effect. Eur J Nucl Med Mol Imaging 34(10):1658–1669 Thie JA (1995) Clarification of a fractional uptake concept. J Nucl Med 36:711–712

Guidelines for Nuclear Image Analysis

26

Martin S. Judenhofer, Stefan Wiehr, Damaris Kukuk, Kristina Fischer, and Bernd J. Pichler

Contents 26.1 Quantitative Image Analysis.................................. 379 26.2 Quantification Errors............................................. 380 26.2.1 Photon Attenuation................................................... 382 26.2.2 Partial Volume Effect................................................ 383 26.3 Reporting of Nuclear Image Quantification......... 384 26.4 PET Images from Miscellaneous Tracer.............. 386 References............................................................................ 386

M.S. Judenhofer (*), S. Wiehr, D. Kukuk, K. Fischer, and B.J. Pichler Department of Radiology, Laboratory for Preclinical Imaging and Imaging Technology of the Werner Siemens-Foundation, University of Tübingen, Röntgenweg 13, 72076 Tübingen, Germany e-mail: [email protected]

26.1 Quantitative Image Analysis The information of a nuclear emission image depends mainly on the used radiolabeled biomarker, which interacts – in an ideal case – specifically with the in vivo target. Although several sections of this chapter are applicable for single photon emission computed tomography (SPECT), however, the major focus is on positron emission tomography (PET) imaging since accurate quantification is more complicated in SPECT, which is why SPECT is often considered as a semiquantitative imaging modality. Figure 26.1 shows examples of PET images of the same rat, scanned on 2 consecutive days with different tracers, the glucose analog [18F]FDG and the proliferation marker [18F]FLT. The rat bears a xenograft tumor on the right flank. Figure 26.1 clearly shows the strength of PET imaging, which delivers information based on the molecular target of the tracer. However, it depicts also a major weakness of PET, which is the limited availability of anatomical references. This drawback makes the quantitative analysis sometimes very difficult as the tumor size, appearing in the image, depends not only on the physical lesion size and metabolic rate but also on the image contrast, noise, and chosen upper and lower image intensity threshold for display. Figure 26.2 shows an example of one rodent displayed with an image analysis software at two different upper thresholds. According to Fig. 26.2, it is obvious that the intensity threshold for image analysis needs to be chosen very carefully. Too low thresholds might lead to blasting of the image, which makes the tumor larger than it is. Drawing the region of interest (ROI) for quantification based on such an image would result in an underestimation of the average tumor activity due to the

F. Kiessling and B.J. Pichler (eds.), Small Animal Imaging, DOI: 10.1007/978-3-642-12945-2_26, © Springer-Verlag Berlin Heidelberg 2011

379

380

M.S. Judenhofer et al.

• Normalize all images of one study to each other and use an ROI size based on visual, color-based boarders. The normalization factor needs to account for the individual-injected radioactivity doses and isotope half-life. However, the uptake time and physiological parameters need to be maintained constant for all animals, especially if the tracer has a fast or time-variant pharmacokinetic. • Placing automatically ROIs by a region growth algorithm using predefined isocontours (e.g., 50% of the maximum value within a seeding point) (Fig. 26.4). There is no ideal approach or gold standard for PET or SPECT image analysis; the choice of the best option depends on the study design and target area. For example, in brain studies where the striatum or cerebellum is the region being investigated, an atlas-based ROI positioning in combination with the use of a fixed ROI size might be ideal since the brain has a fairly constant spatial distribution of regions across a group of subjects. However, an anatomical atlas would not be a choice for tumor studies. Here, the size and location of the tumor are random or at least have high variations among a group of subjects. Besides the variations arising from the biological effects of the model being investigated, there is also a major source of user-determined factors, which influence the measured activity in nuclear imaging. It is therefore definitely helpful to attend the following advice if the study design allows:

Fig. 26.1 The strength of PET lies in the ability to acquire the in vivo biodistribution of different tracers in one single animal. This example shows two different tracers, [18F]FDG and [18F]FLT, commonly used in oncology. The rat, bearing a subcutaneous tumor, was imaged for 20 min on 2 consecutive days. Each tracer shows different uptake patterns

large volume. However, the other way, displaying the image too dimly by choosing a very wide display range (low lower and high upper threshold) would lead to smaller visual display of the tumor and subsequently to a too small ROI, which would then not represent the real tracer uptake value in the entire tumor. Therefore, different approaches are used to avoid an image quantification, which is biased by the chosen threshold: • Using an anatomical reference like a CT or an MR image of the same subject (Fig. 26.3). • Using an anatomical atlas for defining the ROIs.

• Normalize all images of one study to the injected dose (keep injected dose as constant as possible among a group of animals). • Keep the scan time of all animals within a study constant. • Keep the tracer uptake time of all animals within a study constant. • If more than one PET or SPECT scanner is used for a study, the scanners need to be cross-calibrated. • For quantitative studies, the dose calibrator, animal scanner, and gamma counter need to be crosscalibrated.

26.2 Quantification Errors There are several error sources that limit the image quantification accuracy. Although other chapters deal more specifically with these errors, here are some more

26 Guidelines for Nuclear Image Analysis

381

Fig. 26.2 The appearance of PET images and the size of organs or lesions are determined by upper (max) and lower (min) image intensity thresholds. The same PET image is displayed at different lower and upper intensity levels showing clearly the change of tumor size, which would influence the quantification when ROIs are based on these different displayed images

Fig. 26.3 Example of the importance of using an MR image as anatomical reference for defining a PET ROI. Since in this example, the PET tracer uptake in the tumor is very low, the tumor cannot be visualized by PET. However, if the MR image,

which clearly shows the subcutaneous tumor, is used to delineate an ROI around the tumor volume, the PET data can be quantified

382

M.S. Judenhofer et al.

Region level

Volume [mL]

Mean Activity [kBq/mL]

25%

890

44,5

50%

499

57,7

80%

214

68,2

Fig. 26.4 Isocontours are often used as nonuser biased method for generating ROIs. This [18F]FDG-PET image of a tumor-bearing rat shows the different sized volumes, which are generated for different chosen 3D isocontour levels (percentage of the maximum pixel intensity). The table shows that this can lead to significant differences in volume and estimated mean activity

practical related recipes how to account for these errors in small animal imaging studies.

26.2.1 Photon Attenuation A major problem in nuclear imaging, when accurate quantification is required, is the attenuation of the gamma quanta in tissue or scanner materials. Photon

attenuation is driven by the gamma energy (PET annihilation photons have 511 keV) as well as the tissue density and volume of the subject being measured. Neglecting the effect of attenuation will always lead to an underestimation of the total amount of measured activity. For PET, on average, the underestimation in a mouse-sized object can range between 10 and 20% whereas in a rat-sized object, tracer quantification is underestimated by 20–35%. With SPECT, the attenuation is also variant for the different isotopes respectively their gamma energies. SPECT uses lower energy isotope, such as 125I, which will be much more affected by attenuation than an isotope with higher gamma energy (e.g., 99mTc). For a proper correction of the attenuation effects, information about the size and density of the object needs to be acquired. In PET, this is, in general, achieved by measuring the transmission (attenuation) properties for gamma rays at 511 keV for the specific object (Ostertag et al. 1989). There are several ways how this can be achieved. The most straightforward approach is simply using an external radioactive source (preferably with the same gamma energy as the used isotope), which is rotated around the object and then measures the transmitted photons. By this, the attenuation properties of the object can be estimated for all necessary projectional angles resulting in a so-called m-map. For PET, the use of a 511 keV transmission source delivers images with a very limited contrast, which is why alternatively some small animal PET systems provide a lower energy transmission source (e.g., 57Co). In this case, the measured values need to be rescaled to match the energy of the 511 keV photons. As more and more PET systems are combined with a CT scanner, the CT images can be used for PET attenuation correction (Townsend 2001). This has the benefit of reducing the examination time significantly since a CT acquisition is performed faster than a regular transmission scan on a PET system. As an alternative to measured transmission information, attenuation correction can also be applied by using a calculated correction based on segmented images with a known tissue-specific attenuation factor. For mice PET imaging, attenuation correction is often neglected, and a quantification error of about 10–20% might be accepted to save transmission scan time and therefore shorten the time under anesthesia of the animal. In addition, a nonproper performed transmission measurement can add significant noise to the emission image. Thus, mouse attenuation correction is only recommended when the study design

383

26 Guidelines for Nuclear Image Analysis

requires a proper quantification or, when the scanner periphery, such as the bed, adds significant attenuation. However, especially for larger rats, attenuation correction is recommended, specifically when interobject variations, e.g., different tumor positions, apply.

26.2.2 Partial Volume Effect Besides photon attenuation, a second source for false estimations of absolute activity values is the partial volume effect, which is also described in more detail in the next chapter of this book. In brief, since the PET or SPECT system has a finite spatial resolution on the order of 1–1.5 mm, it cannot properly display structures that are smaller than this resolution. Therefore, the whole image and especially sharp edges will always be displayed blurred. The effect for the case of a 2D system can be seen in Fig. 26.5, which shows that all shapes suffer the same amount

of blurring at the edges; however, the smaller the structures are, the more severe this effect is propagated when the values are quantified. This is demonstrated by the profiles through the spheres in Fig. 26.5. In the case of a 3D system, the amount of underestimation will much more severe. As this effect is intrinsic to the system and cannot be avoided, it is difficult to correct for. Currently, there are only a few experimentally used algorithms that provide a correction of the partial volume effect in the image data (Srinivas et al. 2009). However, standard systems do not usually provide any sort of correction. It is therefore the responsibility of the investigator to consider this effect if results are compared to other methods like in vitro gamma counting of tracer uptake in isolated organs. Ex vivo gamma counting is anyway advisable to be done whenever possible – in terminal studies – to cross-validate the in vivo measured results with the gold standard of tracer distribution measurement. As in vivo PET is a convolution of several parameters, superimposed by

Real object 1x resolution 3x resolution

0.5 x resolution

2x resolution

Object after acquisition

Fig. 26.5 Simplified schematic, demonstrating the partial volume effect for the case of a 2D sytem. Objects with the same signal intensity but different sizes suffer from partial volume effect when imaged by a system with finite resolution. The analysis of the 1D profile on the bottom shows severe underestimation esspecially for the smaller objects

Profile Underestimation: 4%

9%

22%

52%

384

errors such as attenuation or partial volume effects, the in vivo and ex vivo results can differ significantly. Figure 26.6 shows an example of in vivo PET results and ex vivo biodistribution measurements by gamma counting from the same mouse. The animal was bearing a 250 mL tumor and was injected with a 64 Cu-labeled tumor-specific antibody. The PET scan was performed 2 days after injection, and the animal was sacrificed immediately after the scan to avoid any further tracer kinetic. Interestingly, the in vitro and in vivo quantification differs by a factor of approximately 2.5. Based on phantom scans, one can expect that the underestimation from not performing attenuation scans is approximately 20% and that approximately 50% underestimation is a result of the partial volume effect. This was quite surprising as the tumor had a size of ~5 mm in diameter, being more than 3 times larger than the PET scanner’s resolution of 1.4 mm. Further measurements confirmed that partial volume effects in small animal scanners are more predominate than in larger clinical scanners. The underestimation can reach the values of ~65% for 3 mm and 30% for 8 mm diameter lesions. Thus, the common rule of thumb for clinical scanners that the recovery of a scanned object in PET is close to 100% of the real activity concentration if the lesion size is 2.5 times the scanner’s resolution does not necessarily apply for the geometry and intrinsic physical parameters of small animal PET scanners. It is therefore advised that for accurate PET quantification, the impaired recovery for a specific lesion size, caused by partial volume effect, should be known.

M.S. Judenhofer et al.

26.3 Reporting of Nuclear Image Quantification There are different ways to report the quantitative tracer uptake in specific organs of interest. A PET scanner is usually calibrated to deliver values of activity per volume (e.g., kBq/mL) from the image data. Based on this, tracer uptake of specific organs can be easily reported as percent-injected dose per cubic-centimeters (%ID/mL) if the values are put in relation to the total injected amount of activity. The %ID/mL can be calculated according to (26.1).

a% ID =

aAbs AInjected

(26.1)

aAbs measured activity within a defined ROI in the PET image (kBq/mL) AInjected injected activity (kBq) a%ID measured value in PET (%ID/mL) The beauty of reporting the tracer uptake of a specific organ in %ID/mL is the direct link of this value to the quantification measured by ex vivo biodistribution with a gamma counter, which is usually reported in %ID/g. Assuming a density of 1, meaning that 1 mL tissue is equally to 1 g, the two values, %ID/mL and %ID/g, can be directly compared. A prerequisite to compare biodistribution from in vivo PET and ex vivo gamma counting results is that the tracer kinetic is known or, alternatively, that the animal is immediately sacrificed while or directly after the PET scan to freeze the tracers pharmacokinetic.

40%

30%

Fig. 26.6 In vivo PET example of the partial volume effect. Due to the small tumor volume of approximately 250 mL, the quantification based on the PET image reveals a much lower activity (%ID/mL) than what was found by examining the same tumor by ex vivo gammacounting (%ID/g), which is considered as the gold standard

20%

10%

0%

PET (%ID/mL)

gamma counter (%ID/g)

385

26 Guidelines for Nuclear Image Analysis 60 sec

5 min

15 min

30 min

40 min

55 min

2000

activity [kBq/mL]

1500

1000

500 striatum cerebellum 0 0

60 sec

500

1000

5 min

1500 2000 time [s]

15 min

30 min

2500

40 min

3000

3500

55 min

750

activity [kBq/mL]

Fig. 26.7 The strength of PET lies in its ability to track the in vivo biodistribution of a tracer over time. This example shows the application of dynamic PET acquisitions performed over one hour with two common tracers used in brain research. The top row shows an acquisition with the 11C-labeled dopamine transporter ligand [11C]Methylphenidate. The bottom shows a similar acquisition when using the 11 C-D2 receptor ligand [11C] Raclopride. The small image icons show visually the tracer uptake in the mouse brain at specific time points. While images at early time points after tracer injection are dominated by perfusion, later time points show specific tracer uptake in the striatum. The cerebellum is the reference tissue region where no specific tracer binding is found. The harderian glands behind the eyes light up in all images; most of the PET tracers accumulate in these glands. In both acquisitions, quantitative time activity curves were derived based on ROIs (white circles in the images) in the striatum and cerebellum of the animal

500

250 striatum cerebellum

0 0

500

Another way of quantitative reporting in PET, which is mainly used in the clinical field, is the so-called standard uptake value (SUV), which is a simplification of the more complex Patlak analysis (Weber et al. 2000) used mainly for FDG. The SUV is defined as specified in (26.2). Even though the SUV does not provide a “cut-off” value which would allow for a sharp separation of malign and benign tumors it can be very well used to compare values among patients with the same type of tumor and

1000

1500 2000 time [s]

2500

3000

3500

especially for follow-up studies within the same patient to determine a response to a certain therapy. SUV aAbs AInjected m

SUV =

a Abs ×m AInjected

(26.2)

standard uptake value (1/1) measured value in PET image (kBq/mL) injected activity (kBq) objects body weight (g)

386

Note: The assumption is made that the density of the tissue is 1, which is why the SUV is dimensionless. Since the SUV is a so frequently used parameter in clinical diagnosis, it also becomes more and more applied in the preclinical field. However, in most cases, the clinical SUV values are not comparable with SUVs reported in preclinical studies. For example, a prostate xenograft tumor of 200 mL, grown in a nude mouse, can provide a SUV of only 0.5, which is a value measured in most mouse tumors. In sharp contrast, FDGPET SUV values measured in tumors of humans are typically on the order of 2–5 and higher.

26.4 PET Images from Miscellaneous Tracer Figure 26.7 shows an example of PET being used in brain research experiments. The in vivo biodistribution of different commonly used PET brain receptor tracers in mice has been investigated by performing dynamic acquisitions over one hour. A time series of PET images can be reconstructed, and based on ROIs, a socalled time activity curve (TAC) can be generated, which shows quantitatively the regional tracer uptake over time. These TAC data can later be used as input data to perform kinetic modeling. For small animal PET studies, the amount of required injected activity depends on the size of the animal, the tracer, the isotope, and the sensitivity of the animal PET scanner. For scanner sensitivities of <5%,

M.S. Judenhofer et al.

in mice, injected dose of 5–13 MBq for a 18F tracer is a good lead, while the dose is escalated to approximately 11–18 MBq in larger rats. For these amounts of tracer, static scan times of 10 min are sufficient. Using 11C compounds and data analysis by kinetic modeling requiring dynamic data analysis and time framing over one hour, injected activities of 18–22 MBq for mice are advisable. However in receptor binding studies, where tracer mass may effect binding parameter estimates lower injected activities may be required to support the tracer principle. In rats, amounts of 35 MBq should be considered. However, for longer living isotopes like 64Cu (12.7 h half-life), amounts of 5–13 MBq are sufficient providing enough activity to perform scans over 24 h. When using 64Cu, one has to keep in mind that only ~20% of the decay is by positrons, requiring more injected activity or longer scan times than isotopes with a higher positron decay branch.

References Ostertag H et al (1989) Measured attenuation correction methods. Eur J Nucl Med 15(11):722–726 Srinivas SM et al (2009) A recovery coefficient method for partial volume correction of PET images. Ann Nucl Med 23(4):341–348 Townsend DW (2001) A combined PET/CT scanner: the choices. J Nucl Med 42(3):533–534 Weber WA, Schwaiger M, Avril N (2000) Quantitative assessment of tumor metabolism using FDG-PET imaging. Nucl Med Biol 27(7):683–687

27

Kinetic Modelling Jo¨rg van den Hoff

27.1 Introduction

Contents 27.1

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 387

27.2

Mathematical Basics of Tracer Kinetic Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 388 27.2.1 Linear Systems and Convolution . . . . . . . . . . . . . . . . 389 27.3 Compartment Models . . . . . . . . . . . . . . . . . . . . . . . . . . 27.3.1 General Properties of Compartment Models . . . . . 27.3.2 Extraction Fraction and the Renkin Crone Formula . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27.3.3 Perfusion Measurements . . . . . . . . . . . . . . . . . . . . . . . . . 27.3.4 Fractional Blood Volume . . . . . . . . . . . . . . . . . . . . . . . . 27.3.5 Distribution Volume . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27.3.6 Reference Tissue Methods . . . . . . . . . . . . . . . . . . . . . . .

389 390 394 395 395 396 396

27.4

Specific Problems of Quantitative Small Animal Imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27.4.1 Spatial Resolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27.4.2 Time Scales . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27.4.3 Arterial Input Function . . . . . . . . . . . . . . . . . . . . . . . . . .

397 397 397 397

Practical Considerations . . . . . . . . . . . . . . . . . . . . . . . Computing the Convolution Integral . . . . . . . . . . . . . Input Function Delay and Dispersion . . . . . . . . . . . . Interpolation of the Measured Data . . . . . . . . . . . . . . Influence of Metabolites . . . . . . . . . . . . . . . . . . . . . . . . . Model Fitting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . How to Decide Which Model to Use . . . . . . . . . . . .

398 398 398 399 399 399 400

27.5 27.5.1 27.5.2 27.5.3 27.5.4 27.5.5 27.5.6 27.6

Appendix: A Short Outlook on General Compartmental Theory . . . . . . . . . . . . . . . . . . . . . . . . 400

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

402

J. van den Hoff PET Center, Forschungszentrum Dresden Rossendorf (FZD), Bautzner Landstrasse 400, 01328 Dresden, Germany e mail: j.van den [email protected]

A mathematical description of the time-dependent tissue uptake and tissue clearance after injection of contrast agents or radioactively labelled tracers with suitable models (kinetic modelling) allows a detailed analysis of transport processes and metabolism in vivo. Such an analysis can provide at once quantitative information for several interesting parameters such as local blood volume, blood flow, distribution volumes, metabolic rates, binding potentials and so forth. While the mathematical techniques are with some reservations in principle suited for data analysis in other tomographic modalities as well (notably CT and MRI), the broadest field of application is found in PET. We will focus especially on small animal PET in the following. Small animal PET is a quantitative tomographic imaging technique enabling non-invasive and repeated determination of regional accumulation and distribution of suitable radioactively labelled tracers in vivo. The full potential of this quantitative capability is only utilized if the well-established techniques of tracer kinetic modelling are used to derive from the measurements of regional tracer concentrations quantitative values of relevant physiological and biochemical transport constants. This chapter provides an introduction to the basic techniques while drawing attention to some specific problems of translating techniques established in human investigations to small animals. For reasons of limited space, this chapter can not cover all aspects of kinetic modelling. A comprehensive discussion of concepts in classical tracer kinetic experiments can be found in (Lassen and Perl 1979). Some recent references, which partly cover complementary aspects of tracer kinetic modelling in PET, are

F. Kiessling and B.J. Pichler (eds.), Small Animal Imaging, DOI: 10.1007/978 3 642 12945 2 27, # Springer Verlag Berlin Heidelberg 2011

387

388

(Carson 2003; Carson et al. 2003; Mu¨ller-Schauenburg and Reimold 2008; van den Hoff 2005; Willemsen and van den Hoff 2002). A useful reference regarding the physics and mathematics of physiological transport processes in general is (Weiss 1996). What are the benefits of kinetic modelling? This is a justified and an important question since as we will see kinetic modelling requires substantial methodological effort both during data acquisition and during data evaluation. The answer is that only in this way we are able to really measure relevant parameters. Valuable information can be derived from the image data solely by visual inspection, determination of regional contrast, or computation of regional values for the accumulated percentage of injected dose. But all these approaches are essentially qualitative and purely phenomenological. They are frequently not suited to gain a deeper understanding of the investigated processes. Moreover, tracer experiments are investigating inherently time-dependent processes, namely the transport and metabolism of certain substrates. Only by suitable evaluation of the whole time course is it possible to actually get information concerning the nature of these processes. What are the limits and pitfalls of kinetic modelling? One serious danger lies in the attempt to derive “too much” information from the data by using too complex models. Non-invasive PET imaging provides only an integral signal, namely the timedependent total tracer concentration in the target region. This signal is in general the sum over several compartments. Only for very simple models is it possible to derive from this superposition all the contributing curves (and thus the corresponding rate constants connecting the compartments). Another danger is a wrong interpretation of the model parameters that is related to the fact that the parameters always express relative rates (of concentration changes) with respect to total volume and not with respect to the individual pools.

27.2 Mathematical Basics of Tracer Kinetic Models Kinetic modelling is concerned with the mathematical description of physiological and metabolical transport processes by analysis of mass balance equations.

J. van den Hoff

Contrary to classical pharmacokinetics, which is essentially dealing with whole body kinetics such as blood clearance of a substance, tracer kinetic modelling of tomographic data is investigating regional processes. No attempt is made to account for the body as a whole. A direct consequence is that the concentration in the arterial blood ca(t) is treated as a known quantity, i.e. it has to be determined either via arterial blood sampling or in the tomographic image data. The fundamental mass balance is given by dms ðtÞ ¼ F ca ðtÞ F cv ðtÞ dt

(27.1)

or equivalently, if ms(t 0) = 0, in integral form Z ms ðtÞ ¼ F 0

t

Z ca ðuÞdu F

t

cv ðuÞdu;

(27.2)

0

where ms(t) is the tracer amount (mass, activity, . . .) residing in the system at a time t and is measured, e.g., in becquerel (Bq), F is blood flow (units: mL/min), and ca(t) and cv(t) are the tracer concentrations in arterial and venous blood (units, e.g., Bq/mL), respectively. Equation (27.2) expresses the simple fact of mass conservation: the amount ms(t) of tracer in the system at time t is the difference between cumulative inflow and outflow until this time.1 Since the tomographic techniques yield concentration proportional data (e.g. tracer amount per image voxel), it is necessary to divide Eq. (27.1) on both sides by the system volume, which leads to dcs ðtÞ ¼ f ca ðtÞ f cv ðtÞ; dt

(27.3)

where f = F/V (units: mL/min/mL) is called specific blood flow or perfusion and represents the blood flow per unit system volume.2 For local quantification, Eq. (27.3) is not directly usable since the local venous concentration cv(t) can generally not be measured. On

1

If the tracer is finally completely washed out from the system, i.e. limt!1 ms ðtÞ 0, mass conservation implies that the areas under R 1 and outflow curves (AUC) are identical: R 1 the inflow 0 ca ðtÞdt 0 cv ðtÞdt:

2

It’s rather common to use both terms (perfusion and blood flow) synonymously to denote specific blood flow per unit volume. We will adopt this custom in this chapter.

27

Kinetic Modelling

389

the other hand, ca(t) is the same in all arteries (apart from a possible bolus delay and dispersion that can be accounted for during further data evaluation). Thus, it can very well be measured, e.g. via blood sampling from an accessible larger artery or, using the image data, in the cavum of the left ventricle. Quite obviously, cv(t) is determined by how the tracer is transported back and forth between the capillary blood space and tissue. Thus, cv(t) is not an independent quantity and should be eliminated from the equations. It is exactly at this point that specific model assumptions have to be used to proceed. A central, still rather general assumption of tracer kinetic modelling is that we are dealing with a so-called linear system. The following section discusses the implications of this assumption.

27.2.1 Linear Systems and Convolution A linear system “translates” a given input function ca(t) into the corresponding system response function cs, which might be expressed symbolically as ca ! cs , in such a way that a scaling of the input leads to an identical scaling of the response and, more generally, ð1Þ ð2Þ the response to the sum of two inputs ca ; ca ; equals ð1Þ ð2Þ the sum of the responses cs ; cs ; to the separate ð1Þ ð2Þ ð1Þ ð2Þ inputs: ca þ ca ! cs þ cs . This “translation” from input to response is mediated by the system’s (unit) impulse response function (IRF), which we denote as h: h

ca ! cs : Mathematically, linear systems are described by linear differential equations whose formal solution can be expressed with the help of a so-called convolution integral. In their most general form, the model equations are partial differential equations for the time- and space-dependent tracer concentration c(x, y, z, t) at each point in the system (distributed modelling), but it turns out that explicit treatment of the spatial concentration variations is not necessary when analyzing tomographic image data. Instead, it suffices to subdivide the system into a number of compartments, which eliminates the spatial coordinates from the mathematical description. We are thus

left with solely time-dependent functions. The resulting compartment models are described by ordinary differential equations for the time-dependent concentrations c1(t), c2(t), . . ., cN(t) in the different compartments constituting the model. Since, furthermore, all concentrations are equal to zero before t = 0, the solution is given by this specific convolution integral: Z cs ðtÞ ¼ ca ðtÞ hðtÞ ¼

t

ca ðuÞ hðt uÞdu: (27.4)

0

A linear system is completely characterized by its IRF. Once you know the IRF, you can compute the system response for any input ca(t). This special property of linear systems separates the specific system properties that control the functional shape of h(t) and the influence of the specific input function actually used in an experiment. Equation (27.4) is a direct consequence of the linearity of the system. Simply speaking, for any given observation time t, the convolution integral is adding up time-shifted impulse responses h(tu) to successive short pulses of intensity ca(u) and duration du occurring at all times between u = 0 and the observation time u = t. An illustration of this process is given in Fig. 27.1.

27.3 Compartment Models Compartment modelling is definitely the most important kinetic modelling technique used for quantification of PET investigations. In this approach, it is assumed that we can decompose the system into a finite (preferably small) number of components the compartments that correspond to distinguishable states of the tracer. Compartments might represent different spatial components where the tracer resides (e.g. extra- and intracellular space) or different chemical modifications of the tracer. In any case, the central assumption is that a compartment does not exhibit any internal structure but is completely characterized by the (generally time-dependent) tracer concentration in this compartment. In other words, there are no internal concentration differences: a compartment is by definition a space with homogeneous tracer concentration (“well-stirred pool”).

390

1 pulse

Input

Response

1 pulse

Time

Time 5 pulses

Input

Response

5 pulses

Time

Time 20 pulses

Input

Response

20 pulses

Time

Time 100 pulses

Response

100 pulses Input

Fig. 27.1 Response of a linear system (in this case with mono exponential IRF) to different numbers of discrete input pulses with negligible duration. Pulse positions are indicated by solid circles (left column). The system response at a given time point is obtained by summing up the contributions from the impulse responses to all pulses prior to that time point. With increasing number of pulses, the input approaches an infusion with finite duration while the system response asymptotically approaches the shape given by the convolution integral Eq. (27.4)

J. van den Hoff

Time

All concentration differences are attributed to transitions from one compartment to another. Even complex systems can be described by compartment models if one is willing to accept a large number of compartments.3 The real value and usefulness of compartment models depend, however, on sufficient model simplicity. This is especially true for non-invasive imaging, where one cannot access the individual compartments directly but has to deduce their properties via their contribution to the total tracer concentration in the system.

Time

Compartments are represented by boxes. Tracer exchange is represented by arrows labelled with the

a

b

ca

ca

K1

K1

c1

c1

k2

c

d

ca

ca K1

k3 c1

The structure of compartment models can be illustrated as shown in Fig. 27.2.

3

The so called distributed models (Bassingthwaighte et al. 1992; Bassingthwaighte and Goresky 1984; Bassingthwaighte and Holloway 1976), which permit description of tracer trans port in the presence of spatial concentration gradients, can be seen as limiting cases of compartment models where the number of compartments tends to infinity.

k5 c2

k4 k2

c2

k2

K1

27.3.1 General Properties of Compartment Models

k3

c3

k3 c1

k6

c2 k4

k2

Fig. 27.2 Examples of compartment models. (a) Reversible one compartment model. This model is very frequently used and includes the case of freely diffusible (flow) tracers. (b) Irreversible two compartment model. This is the well known “FDG model”. (c) Reversible two compartment model. Config urations (b) and (c) are used broadly for description of tracers, which are first entering a free pool (c1) and are then metabolized, either irreversibly or reversibly, which is represented by the second compartment (c2). (d) Three compartment model. This model is suitable, e.g. for receptor ligands with competitive specific (c2) and non specific (c3) binding to different sites. Typical response curves for these models are shown in Figs. 27.3 and 27.4.

Kinetic Modelling K1, k2, k3

IRF (=h1)

0.4

0.4

K1, k2

IRF h1 h2

0.1 0.0

0.0 0

10

20 30 40 time (a.u.)

50

0

60

10

K1, k2, k3, k4

20 30 40 time (a.u.)

50

60

0.4

K1, k2, k3, k4, k5

IRF h1 h2

IRF h1 h2 h3

d

0.1

0.2

0.3

c

0.0 0

10

20 30 40 time (a.u.)

50

60

0

10

K1, k2

20

30 40 time (a.u.)

a

50

60

0

10

10

20

30 40 time (a.u.)

60

20

30 40 time (a.u.)

b

50

60

K1, k2, k3, k4, k5

c

50

input response c1 c2 c3

0.0 0.2 0.4 0.6 0.8 1.0

concentration (a.u.) 0.0 0.2 0.4 0.6 0.8 1.0

K1, k2, k3, k4 input response c1 c2

0

50

input response c1 c2

0.0 0.2 0.4 0.6 0.8 1.0

concentration (a.u.) 0.0 0.2 0.4 0.6 0.8 1.0

10

20 30 40 time (a.u.)

K1, k2, k3

input response (=c1)

0

Fig. 27.4 Response curves ct = ca h for the models from Fig. 27.2 to an idealized bolus input with recirculation using the impulse responses from Fig. 27.3

b

0.2

0.3

concentration (a.u.) 0.1 0.2 0.3

a

0.0

Fig. 27.3 Total impulse PN response (IRF), h n 1 hn , of the models from Fig. 27.2. Non zero transport constants in each case are indicated on top of the plots (a.u.: arbitrary units). The transport constants were set to the following values (in the appropriate units): K1 = 0.4, k2 = 0.2, k3 = 0.05, k4 = 0.075, k5 = 0.05, k6 = 0. Note how the impulse responses hn in the different compartments change from (a) to (d) and how the different compartments contribute to the tissue response. Further note the rather similar but not identical shape of the tissue responses in (a, c) and (b, d), respectively

391

concentration (a.u.) 0.1 0.2 0.3 0.4

27

60

0

10

20

30 40 time (a.u.)

d

50

60

392

J. van den Hoff

transport constants K1, k2, . . ..4 The input into the system is given by the time-dependent arterial tracer concentration ca(t). In the examples from Fig. 27.2, tracer is lost from the system via a single outlet characterized by the rate constant k2. Each model is mathematically defined by a certain differential equation. For instance, the reversible two-compartment model of Fig. 27.2c5 has the model equation dc1 ¼ K1 ca ðk2 þ k3 Þc1 þ k4 c2 ; dt dc2 ¼ k3 c1 k4 c2 : dt

ð27:5Þ

Remember that all concentrations are specified with respect to the same volume, and Eq. (27.5) thus describes the fundamental mass balance for the system: the rate of concentration change in a compartment equals the difference of inflow into and outflow from the compartment which in turn are proportional to the respective “source” concentrations. Inspection of Eq. (27.5) shows that the transport constants have dimensions of inverse time. They are typically specified in units of min 1. From dimensional considerations alone, it appears that K1 is measured in the same units as the rate constants k2, k3, . . .. But contrary to all other model parameters, K1 relates ca, the tracer concentration in blood, to the rate of concentration change in a tissue compartment. For this reason, the units are commonly specified as ½K1 ¼

5

hðtÞ ¼ ¼ ¼

hm ðtÞ ¼

N X N X

ln t

m¼1 n¼1

N X

N X

n¼1

m¼1

N X

amn e

an e

!

amn ln t

:

e

ln t

(27.6)

n¼1

mLðtissueÞ

Actually, constancy of these parameters (i.e. time invariance of the investigated process during the measurement) is assumed and not to be taken for granted. Variability of the transport constants during the measurement does complicate the situation considerably. In the older literature, the blood pool was frequently counted (erroneously) as one of the compartments, so that this model was called a three compartment model. In order to be completely unambiguous, it might be wise to use the notation two tissue compartment model (Innis et al. 2007), but omitting the explicit reference to “tissue” should not lead to misunder standings anymore.

N X m¼1

mLðbloodÞ min

and formal cancellation of the volume units is not performed. From this, one might construct the following interpretation of a given numerical value of K1: it

4

represents the volume of arterial blood whose tracer content is completely extracted and transferred within 1 minute to 1 mL of tissue. Note that in general, tracer extraction during the capillary transit is actually incomplete, so that the blood volume flowing through the tissue per unit time will be larger than K1 (cf. Sect. 3.2). A non-invasive experiment does not allow separate determination of individual compartment concentrations. Rather we only can measure the average tracer concentration in tissue. For example, in the case of a two-compartment model, this is given by the sum ct = c1 + c2 (remember that all concentrations refer to a common volume). The IRF of the system, h(t), is correspondingly given by the sum of the IRFs in the individual compartments: h = h1 + h2. The IRF in each compartment of an N-compartment system is a sum over N decreasing exponentials: P hm ðtÞ ¼ Nn¼1 amn e ln t ðln 0Þ: The decay constants ln are identical in all compartments, but the amplitudes amn differ between compartments (and some amplitudes might be negative or zero). The IRF of the whole system is given by summing over all compartments:

As can be seen, the IRF of an N-compartment model has 2N free parameters (N amplitudes, N decay constants). This is the maximum number of parameters derivable from an experiment that only accesses the total concentration in the system.6 In the case of Eq. (27.5), the IRF is explicitly given by

6

Theoretically, the number of transport constants in an N compartment model can get larger than 2N (in the most general case, there are N2 rate constants and N blood to tissue uptake constants). In this case, determination of all model parameters from a non invasive experiment would be princi pally impossible. In practice, this usually poses no problem since the relevant models have more specific configurations, such as those given in Fig. 27.2, where the number of rate constants never exceeds 2N.

27

Kinetic Modelling

393

K1 h ðk4 l2 Þe l2 t þ ðl1 k4 Þe l 1 l2 i K1 h k3 e l2 t k3 e l1 t ; h2 ¼ l 1 l2 h1 ¼

h 1 þ h2 h ¼ a 2 e

l2 t

K1 ¼ l 1 l2

þ a1 e l 1 t h ðk3 þ k4 l2 Þe

þ ðl1 k3 k4 Þe

l1 t

i

l1 t

i

;

l1=2 ¼

q

h ¼ K1 e

l2 t

;

ð27:7Þ

ðk2 þ k3 þ k4 Þ2 4k2 k4 2

:

(27.8) As can be seen, the analytical relation between the parameters an, ln of Eq. (27.6) and the transport constants (K1, k2, . . .) already is rather complicated for this simple model. For multi-compartment models, these relations are usually no longer obtainable in closed form.7 The Appendix gives some hints on the numerical methods that can be used to derive the IRF in these cases. It is straightforward to deduce the IRF for the simpler model in Fig. 27.2b by letting k4 = 0 in Eqs. (27.7) and (27.8). Hence, l1 = k2 + k3, l2 = 0, and we obtain h1 ¼ K1 e ðk2 þk3 Þt ; K1 h h2 ¼ k3 k3 e k2 þ k3 K1 h k3 þ k2 e h¼ k2 þ k3

ðk2 þk3 Þt ðk2 þk3 Þt

i i

; :

ð27:9Þ

For models with irreversible binding such as this one (the well-known “FDG-model” developed for description of the tracer [18F]2-fluoro-2-deoxy-D-glucose, which is routinely used for investigating glucose metabolism), the IRF approaches a non-zero value at sufficiently late times, which corresponds to the rate of binding in the irreversible pool. For example, for the above model, this binding rate is given by

7

K1 k3 : k2 þ k3

Additionally letting k3 = 0 in Eq. (27.9) yields the solution for the one-compartment model in Fig. 27.2a:

where ðk2 þ k3 þ k4 Þ

Km ¼

Note, however, that the basic functional form of the IRF always remains the sum of exponentials according to Eq. (27.6).

k2 t

:

(27.10)

The typical shape of the IRF for different models is shown in Fig. 27.3. The corresponding responses to a finite input are presented in Fig. 27.4. One further remark concerning the interpretation of the impulse response Eq. (27.6): in this equation, we started by summing over the contributions hm(t) from the N individual compartments and ended up with a sum over the N exponentials an e ln t . This has the important implication that one always can envisage an artificial compartment model consisting of N non-interacting (“decoupled”) one-compartment systems independently exchanging tracer with the blood pool with uptake constants an and clearance rate constants ln that always exhibits exactly the same total tissue response curve as the original compartment model. This is exemplified in Fig. 27.5. The indistinguishability between different models containing the same number of compartments has the following consequences:

By measuring only the total tissue response (rather than the concentration in the individual compartments), one cannot determine the actual compartmental structure of the system under investigation but only the number N of kinetically discriminable compartments. In order to be able to derive specific model parameters, the model structure must be decided a priori.

a

b

ca K1

ca a1

ca a2

k3 c1

c2 k4

k2

⇔

e1 λ1

e2 λ2

Fig. 27.5 A reversible two compartment model (a) and the equivalent decoupled model (b). Each compartment en of (b) corresponds to one of the exponentials in the tissue response of (a). The parameters of both models are related according to Eqs. (27.7) and (27.8). Both models are equivalent in the sense that for all times c1(t) + c2(t) e1(t) + e2(t) so that they behave completely identical with respect to the total tissue response

394

J. van den Hoff Uptake K1

0.2

0.4

E

0.6 0

1

2 3 f (ml / min / ml)

In general, it would be an error to interpret individual exponentials in the impulse response as representation of individual compartments. This would only be true if the actual structure of the system were that of a decoupled model.

The qualitative behaviour of the system under consideration can frequently be more easily understood in terms of the equivalent decoupled model.

K1 !

PS=f

and

K1 ðf Þ ¼ f 1 e

PS=f

0

1

2 3 f (ml / min / ml)

;

4

5

f for f PS (flow limited transport) PS for f PS (diffusion limited transport),

which are a direct consequence of the Renkin Crone formula.8 Figure 27.6 shows the flow dependence of E and K according to Eq. (27.12) for PS = 1 mL/min/mL. In practice, deviations from Eq. (27.12) are frequently observed under high-flow conditions and the modified formula PS0 =f

(27.13)

is empirically more appropriate. This equation might be derived from Eq. (27.12) by assuming a flow dependence of PS according to

1 PSðf Þ ¼ PS0 þ ln T1

(27.11)

As indicated, the extraction fraction is generally flow-dependent (as is K1). Under certain assumptions, the flow dependence is given by the following relation (Renkin Crone formula, Crone 1963; Renkin 1959): Eðf Þ ¼ 1 e

5

K1 = PS

Eðf Þ ¼ 1 T1 e

The unidirectional tracer uptake rate K1 can never exceed the tissue perfusion f. The ratio of these parameters is called the (unidirectional) extraction fraction: K1 ðf Þ : f

4

(

27.3.2 Extraction Fraction and the Renkin–Crone Formula

Eðf Þ ¼

K1 = f

0.8

K1(ml / min / ml) 0.0 0.2 0.4 0.6 0.8 1.0 1.2

1.0

Extraction E

0.0

Fig. 27.6 Unidirectional extraction fraction and uptake rate according to the Renkin Crone formula for PS = 1 mL/min/mL. In this example, the uptake rate K1 essentially equals perfusion f below f 0.3 mL/min/mL and becomes nearly constant beyond f 3 mL/min/mL

f:

(27.14)

PS0 is the low-flow limit of the PS-product, and T1 represents the non-extracted (“transmitted”) fraction at very high flows. One way in which such a flow dependence of PS could be understood is successive capillary recruitment (increasing S) with increasing flow (but not in the brain, where capillary recruitment is not operational).

(27.12) where PS is called the permeability surface area product, which is specific of the given tracer and diffusion boundary. The two limiting cases of very high and very low permeability are of special interest:

8

The low permeability limit follows from first order Taylor expansion x 1 ) ex 1 x in the Renkin Crone for mula.

27

Kinetic Modelling

395

27.3.3 Perfusion Measurements Perfusion measurements in PET require the use of highly extracted tracers (cf. Sect. 3.2). Ideally, a perfusion tracer is freely diffusible between blood and tissue. In this case, K1 = f and in the one-compartment case one gets the Kety Schmidt model dct f ¼ fca ct : dt Vd

(27.15)

This model results directly from the primary mass balance Eq. (27.3) using a single but rather restrictive assumption, namely that the tracer entering the system equilibrates so rapidly between blood and tissue space that the venous outflow cv (t) is proportional to ct (t) at all times: cv ðtÞ ¼

1 ct ðtÞ: Vd

This very simple model usually works well in PET, since the time resolution of the measurements is relatively low in comparison to the capillary tracer transit time ( 1 2 s), so that violations of the model assumptions do not manifest themselves in the data.

27.3.4 Fractional Blood Volume Up to now, we have silently assumed that we are really able to measure directly the tracer concentration in tissue when performing a PET measurement. Actually, we are measuring the concentration in a macroscopic region of interest (ROI)9 and are thus averaging over the enclosed tissue space as well as the microvasculature. In order to account for this, we introduce the fractional blood volume (fbv) as a further parameter. The measured PET signal is then given by cs ðtÞ ¼ ð1 fbvÞ ct ðtÞ þ fbv ca ðtÞ;

9

(27.16)

The smallest possible ROI being a single image voxel.

i.e. the volume-weighted average of tissue and blood concentration. Since one finds fbv 0.03 0.05 in most tissues, the scaling of ct(t) with (1fbv) is generally only a small correction and sometimes omitted in the literature. This omission amounts to a redefinition of K1 by referring it (and ct(t)) not to the unit of tissue volume but rather to the unit of total volume (including blood space). As explained, the difference is usually (although not always) small, but conceptually noteworthy. From the mathematical side, fbv simply represents an additional model parameter, which has to be determined together with the transport constants of the compartment model. The inclusion of fbv in the operational equations is primarily important to avoid serious bias in the estimates of the other parameters. On the other hand, fbv might be interesting in itself since it provides direct information regarding the degree of regional vascularization, e.g. in tumours. The identifiability of fbv in the measured data depends on sufficiently rapid variation of ca(t) in some time window, usually the early phase of bolus transit. A further remark seems necessary: Eq. (27.16) assumes that the tracer concentration in the whole enclosed microvasculature follows the arterial tracer concentration. This assumption is only sensible for sufficiently low extraction of the tracer. For highly extracted (“perfusion”) tracers, on the other hand, one usually assumes rapid equilibration between capillary and tissue space, so that the distinction between both spaces vanishes completely for perfectly diffusible tracers (cf. Sect. 3.3). This is the reason why the fractional blood volume usually is not included in this case (although one might argue that there still is a very small fractional blood volume contribution from the arterial part of the microvasculature). Formally, Eq. (27.16) amounts to assuming an instantaneous impulse response of the capillary space to the arterial input. If this assumption is dropped, one is led to more elaborate models, one example being the two-compartment system Fig. 27.5a, where in this case, the first compartment represents the capillary and the rate constants can be expressed in terms of perfusion f, PS-product PS and the fractional blood and tissue spaces. These type of models become relevant, e.g., in the context of dynamic contrast agentenhanced CT (DCE-CT).

396

J. van den Hoff

27.3.5 Distribution Volume For reversible (untrapped) tracers, the regional distribution volume10 Vd can be defined as11 Rt Vd ¼ lim R 0t t!1

0

ct ðuÞdu ca ðuÞdu

:

(27.17)

If ca(t) approaches zero at late times, this is simply the ratio of the (finite) total areas under both time activity curves. This definition implies that Vd is also identical to the equilibrium (i.e. time-independent) tissue-to-blood concentration ratio ceq t ceq a

(27.18)

if (and only if) ca(t) and consequently ct(t) approaches a finite constant value at late times.12 Vd is related to the impulse response of the system according to Z

1

Vd ¼

hðtÞdt;

In terms of the equivalent decoupled model (cf. Fig. 27.5), this is directly understandable considering the equilibrium at late times, when all compartments en have equilibrated with the blood. The tissue blood concentration ratio is then equal to the ratio of the forward/backward transport rates, hence equal to an/ln. By summing over the independent components, Eq. (27.20) follows. Expressed in terms of the rate constants of the original model, this leads, e.g. to K1 k3 k5 Vd ¼ 1þ þ k2 k4 k6 for the reversible three-compartment model of Fig. 27.2d, which includes the simpler models of Fig. 27.2a, c by setting either k3 = k5 = 0 or only k5 = 0. Apart from deriving Vd from a determination of all transport constants, there are other strategies such as the Logan approach, which make use of the asymptotic behaviour of the system at sufficiently late times (Logan et al. 1990). Bolus-plus-infusion techniques (see, e.g. Carson et al. 1997) aiming at speeding up the approach to equilibrium Eq. (27.18) are another way to facilitate Vd determination.

(27.19)

0

which follows from the definition Eq. (27.17) together with Eq. (27.4). For our compartment models, h(t) is given by Eq. (27.6) and computation of the above integral yields Vd ¼

N X an n¼1

10

11

12

ln

:

(27.20)

The terms distribution volume and tissue blood partition coefficient, although conceptually different, denote essen tially equivalent quantities. We choose the former. If tracer is irreversibly trapped in the system, the ratio of the above integrals tends towards infinity with increasing time, so that Vd formally would become infinite. The tissue to blood ratio can also become constant for suffi ciently slowly varying ca(t), but this ratio can deviate sub stantially from Vd and is thus not suitable for estimation of this parameter.

27.3.6 Reference Tissue Methods Reference tissue methods have gained considerable importance in receptor imaging. In other settings, the usefulness is questionable. The basic idea is to use a reference region devoid of receptors as an indirect input function for the target region. This is achieved by assuming one-compartment kinetics in the reference region and an identical K1/k2 ratio in, both, target and reference region (note that these are non-trivial assumptions). ca(t) is then expressed in terms of the tissue response and rate constants of the reference region, and the resulting expression is substituted in the model equation of the target region. For details, we refer the reader to the literature (e.g. (Lammertsma and Hume 1996; Mu¨ller-Schauenburg and Reimold 2008) and references given therein). Here, we only want to draw attention to the fact that reference tissue approaches must be carefully validated in each individual case before reliable results can be derived from measurements.

27.4 Specific Problems of Quantitative Small Animal Imaging

0.0

0.2

While the principal techniques of tracer kinetic modelling are equally valid for human PET and small animal PET, there are a number of important practical differences.

397 1.0

Kinetic Modelling

signal intensity 0.4 0.6 0.8

27

0

27.4.1 Spatial Resolution Probably, the most obvious difference is the fact that the limited spatial resolution of the PET scanner is potentially more problematic in small animal investigations: the reconstructed spatial resolution of small animal scanners is about a factor of three better than in clinical PET scanners (about 1.5 2 vs. 4.5 6 mm) while the body size between a mouse and a human differs by a factor of about 20. Therefore, in small animal PET, the relative spatial resolution is smaller by a factor of about 7. Consequently, in small animal PET, the size of many target structures is not much larger than the spatial resolution, and the limited recovery of true signal intensity frequently plays a much larger role than in human PET. Figure 27.7 illustrates how serious this effect becomes when the 3D object size approaches the resolution limit. It is very important to realize that incomplete signal recovery directly translates into a corresponding reduction of the derived uptake parameters (perfusion f or, more generally, uptake K1). It is thus at least very difficult, if not unfeasible, to reliably quantify tracer uptake in structures whose size approaches the tomographic resolution: accurate recovery correction becomes mandatory, which in turn necessitates precise knowledge of the object size and shape. In real data, the situation becomes even more complicated by the spill-over of signal intensity from the surrounding background into the target region. The different model parameters are affected differently by recovery effects. Actually, only the uptake parameter K1 is directly effected. All other rate constants are not altered as long as no sizable spill-over from the vicinity into the target region occurs. The latter assumption is of course only valid if the actual target to background contrast is sufficiently high.

0.5

1 1.5 2 2.5 3 sphere diameter / FWHM

3.5

4

Fig. 27.7 Cross sections through the centres of differently sized homogeneous spheres of unit intensity (solid grey) together with the corresponding measured line profiles (solid black curve). Sphere diameter is specified in units of the full width at half maximum of the scanner’s point spread function (FWHM). The dashed curve shows the variation of the maxi mum measured signal for each sphere (as fraction of true inten sity) with varying size, the so called peak recovery. The dotted curve shows the variation of the signal averaged over the sphere volume, the so called mean recovery. A sphere with a diameter of twice the FWHM, for instance, has a peak recovery of 86% and a mean recovery of 52%

27.4.2 Time Scales Naturally, circulation time is much smaller in small animals than in humans. Moreover, many metabolic processes are occurring on shorter time scales, too. These differences have to be taken into account, when designing optimal scan and reframing protocols.

27.4.3 Arterial Input Function Accurate knowledge of the input function is mandatory for many quantification approaches. The practical limitations (e.g. limited available blood volume) usually prevent direct determination by arterial blood sampling in small animal experiments. Continuous or repeated arterial blood sampling in small animals requires advanced techniques that minimize the sampled blood volume or allow direct re-injection of the blood after the measurement. Such techniques are currently far from being routinely available. The apparently obvious solution to this problem is determination of the input function from the cavum of the left heart ventricle, since contrary to human

398

J. van den Hoff

investigations, the heart is nearly always included in the field of view. But this solution is hampered by the already mentioned problems of limited recovery (of the blood signal) and spill-over (from the myocard and other tissues in the vicinity). Using this approach requires serious effort to avoid massive bias in the derived input functions. The critical importance of a correct input function should be re-emphasized: each overall scaling error of ca(t) directly translates into a corresponding (reciprocal) scaling error of K1 and f. Erroneous curve shape (e.g. by spill-over from neighbouring regions into an image-derived input function) affects all model parameters. Erroneous input functions do not necessarily manifest themselves in bad fits of the used model to the data: scaling errors, e.g. do not affect the goodness of fit at all. Two plausibility checks can be performed, however: if the fitted fractional blood volume is much larger than expected, this is a strong indication of a scaling error (e.g. due to limited recovery) in ca(t). If there are strong deviations of the data from the fit of a model known to be correct for the given tracer, this might be an indication of shape errors in the input function.

27.5 Practical Considerations 27.5.1 Computing the Convolution Integral The special convolution integral occurring in compartment modelling is computed by summing over the contributions from the different exponentials in Eq. (27.6). We end up with a sum over integrals of the form an ca ðtÞ e

ln t

Z

t

¼ an

ca ðuÞe

ln ðt uÞ

Z

t

ca ðuÞe

ln ðt uÞ

du ¼ e

0

ln t

Z

t

ca ðuÞeln u du:

0

The actual problem, then, is to accurately compute Z

t

In ðtÞ ¼

ca ðuÞeln u du;

(27.21)

0

which can be done incrementally for all given time points tm, thus saving computation time. This comes at a cost, however: the naive approach would be direct numerical integration of Eq. (27.21) on the grid tm of the measured time points using the trapezoidal rule. This approach is prone to substantial error in the resulting integral since due to the exponential the integrand increases, for realistic input functions, very rapidly with increasing lnu, and the implied piecewise linear approximation continuously overestimates the actual In(t). A solution to this problem is to choose a suitable phenomenological model for the input function and use this in integration of Eq. (27.21). The easiest approach is to approximate ca(t) itself by a piecewise linear function, i.e. to describe it by a polygon connecting the measured data points. One can then proceed by piecewise analytical integration of Eq. (27.21) using integration by parts. This is generally much more accurate than the naive approach.

27.5.2 Input Function Delay and Dispersion

du:

0

Only for very special choices of ca(t), this integral can be computed analytically.13 Otherwise, one must resort to numerical methods. In this case, computation of the tissue response for all interesting time points 13

directly from the above formula would imply repeated integrations between u = 0 and u = t since the integrand itself depends on the upper integration boundary t. This is disadvantageous since it increases computation time substantially. The obvious solution is to rewrite the integral as

Notably, if ca(t) is a sum of exponentials or piecewise linear (this includes ca(t) = const. as a special case).

Frequently, the input function has to be determined remote from the investigated tissue region or derived by arterial blood sampling. In small animal imaging, especially the latter approach will introduce a sizable time delay and shape modification (dispersion) of the input function relative to the true input function as it arrives in the target tissue. These two parameters (delay and dispersion) have to be corrected or incorporated in the operational

27

Kinetic Modelling

equations (cf., e.g. Meyer 1989; van den Hoff et al. 1993) in order to avoid potentially significant errors in the derivation of the relevant model parameters. This is at least true for perfusion measurements with freely diffusible tracers, since in this case, the correct modelling of the early phase of the tracer kinetic is most important.

399

In the second case (metabolites do enter target tissue), one usually has to give up on the tracer, since synchronous determination of all model parameters (including those describing the metabolite kinetics) is in general out of the question.

27.5.5 Model Fitting 27.5.3 Interpolation of the Measured Data It is frequently necessary to interpolate the measured time activity curves to a common time grid. Quite generally it can be said that linear interpolation is sufficient, if the temporal sampling of the curves is adequate. More elaborate techniques (e.g. spline interpolation) might even be dangerous by deviating locally too much from the data.

27.5.4 Influence of Metabolites If local or systemic metabolism leads to accumulation of sizable amounts of radioactive metabolites in the blood, there are two possibilities: 1. The metabolites do not enter the target tissue at all. 2. The metabolites do enter the target tissue in parallel to the original tracer. In the first case, the effect of the metabolites is restricted to modifying the time activity curve in the blood. If the time dependency of metabolite accumulation in the blood is known (either individually or at least in a representative control group), the metabolites can be subtracted from the whole blood time activity curve in order to derive the true input function for the used compartment model. Note, however, that in the fractional blood volume term in Eq. (27.16), one has nevertheless to use the uncorrected blood curve.14 14

A related problem, which can lead to substantial errors if ignored, should be mentioned: sizable plasma protein bind ing of the tracer (e.g. neuroreceptor ligands), which necessi tates to distinguish between plasma concentration of the tracer and the free fraction in plasma that actually defines the input function.

The actual determination of the model parameters is usually achieved by adjusting the parameters in such a way that the sum of squared differences between data and fit is minimized (Least Squares fitting). In the case of compartment modelling, the model functions are non-linearly dependent on the rate constants k2, k3, . . . , which necessitates non-linear Least Squares fits. It is strongly advisable to use dedicated fitting routines instead of general purpose function minimizers, since the special properties of Least Squares problems can be used to improve stability and runtime of the fit. The central problem with a non-linear fit (apart from sizable runtime, which prevents use for generation of parametric maps) is that the algorithms are iterative procedures with dependence on reasonable start values for all parameters. Moreover, convergence to the best parameter estimate (in the least squares sense) is not guaranteed. For these reasons, trying to avoid non-linear fits is worthwhile. One possibility of achieving this is to concentrate on the asymptotic behaviour of the model equations. The so-called Logan-Plot (Logan et al. 1990) for determination of distribution volumes and the Patlak-Plot (Patlak et al. 1983) for determination of irreversible binding rates are the most important examples. These approaches have in common that they do not provide complete information for all model parameters but rather only for certain ratios of the parameters (which frequently can be estimated more reliably than the individual parameters). Another possibility consists in directly using the integral versions of the model equations, which are linear in all parameters (cf., e.g. Blomqvist 1984). This approach also allows inclusion of corrections for input function delay and dispersion (van den Hoff et al. 1993). These linearized fits provide estimates for all model parameters. The procedure is fast and thus wellsuited for generation of parametric maps. One notable disadvantage is the fact that it is in general not possible

400

to impose constraints on the fit (e.g. by fixing some of the parameters to known non-zero values or demanding positive parameter estimates). Moreover, the parameter estimates are not optimal in the least squares sense, although in practice, this does seldom lead to sizable effects.

27.5.6 How to Decide Which Model to Use Assume you have performed an investigation with a new tracer and managed to measure a good arterial input function as well as tissue response curves. Now the question arises as to which kinetic model to use for further data evaluation. There are two ways to approach this task. 1. Set up the model from first principles based on your knowledge of the transport and metabolization of the tracer. Use this model for data evaluation. 2. Choose a phenomenological model of minimum complexity that is able to describe the data adequately. There is of course no clear-cut difference between both approaches. More often than not one has no complete a priori knowledge of the tracer’s fate. Moreover, even if possible, a detailed description of the tracer’s fate will lead to model configurations far too complex to use directly in data evaluation. It will therefore be necessary to simplify the model sufficiently. What is sufficient in this context can be decided by the second approach: clarify how complex the model has to be in order to completely describe the data. It does not make sense to use a complex model whose parameters can no longer be determined with adequate accuracy from the data. It is therefore not a good idea to start with a too-complex model. Rather start with a simple configuration. A pragmatic approach is the following:15

J. van den Hoff

2. Try to fit this model to the data. Inspect the residues (differences between data and fit). If you get systematic deviations of the fit from the data, analyze these. For example, a sharp peak in the tissue response during the bolus passage of the input is a sure sign of sizable fractional blood volume. Continuously rising tissue uptake in the late phase, whereas the fit tries to “level out”, is rather a sure sign of additional compartments. According to your analysis, successively increase model complexity and add further parameters. Repeat the fit. 3. Inspect the parameter estimation errors (it is mandatory that the used algorithm does provide these!). If the estimation error approaches or even exceeds the parameter estimate itself, this is indication that the parameter is not identifiable in the data and should therefore be eliminated from the model. A difficult case is shown in Fig. 27.8 were simulated noisy data are fitted with different models. Identification of the correct model (Fig. 27.8d) is possible but requires great care. The one-compartment model (Fig. 27.8b) cannot be ruled out with certainty.

27.6 Appendix: A Short Outlook on General Compartmental Theory We give here a short overview of general compartmental theory, which is intended to provide some background for the topics covered in Sect. 3.1. It moreover shows the way for efficient simulations of multi-compartment models, which otherwise rapidly becomes tedious and numerically inaccurate. If the reader does not feel comfortable with the mathematics, this Appendix can be skipped without problem. We use lower case boldface for vectors and upper case boldface for matrices. In matrix notation, the general N-compartment model is then defined by: dcðtÞ ¼ bðtÞ þ KcðtÞ dt

1. Start with a very simple model. Usually, this will be the reversible one-compartment model.

(27.22)

or abbreviated: 15

There are also statistical criteria, notably the F test and the Akaike information criterion that can help to decide whether an increase in model complexity is justified by the concomi tant improvement in the goodness of fit.

c_ ¼ b þ Kc; where c is a (column) N-vector of concentrations in the different compartments, b is the corresponding vector

a

k1 = 0.091 +/− 0.005 k2 = 0.053 +/− 0.004 0

10

20

30 40 time (min)

k1 = 0.09 +/− 0.001 k2 = 0.054 +/− 0.003 k3 = 0.006 +/− 0.001 fbv = 0.111 +/− 0.005 10

20

30 40 time (min)

of inputs into the different compartments, and K is the N N transfer matrix defining the structure of the model. All concentrations are assumed to be zero until t = 0. With this notation, one gets, for example, for the reversible two-compartment model (Fig. 27.2c) c_ 1 c_ 2

¼

K1 ca 0

þ

ðk2 þ k3 Þ k4 k3 k4

c1 ; c2

which is just another way of writing Eq. (27.5). Very generally, one could consider differing time-depenðnÞ dent input functions ca ðtÞ with different uptake rates qn from arterial blood into the N compartments (superscript T denotes matrix transposition): ðNÞ T b ¼ ðq1 cð1Þ a ; . . . ; qN ca Þ

(27.23)

This kind of situation could for instance arise, if metabolites of the tracer accumulate in the blood and do then enter the system as well. But to keep the situation a bit simpler, we assume instead a common input function for all compartments: b ¼ q ca ¼ ðq1 ; . . . ; qN ÞT ca :

(27.24)

The IRF in the N different compartments can be expressed via a so-called matrix exponential. With the input vector of Eq. (27.24), the IRF is given by the vector

50

k1 = 0.085 +/− 0.001 k2 = 0.044 +/− 0.001 fbv = 0.118 +/− 0.006 0

60

c

0

50

b

10

concentration (Bq/ml) 0 1000 2500

Fig. 27.8 Comparison of the performance of different models in fitting a simulated noisy tissue response of a reversible two compartment model including a fractional blood volume parameter. All model parameters were set to a value of 0.1 in the simulation. The data were simulated using a bolus input with a shape similar to that used in Fig. 27.4. In this example, identification of the correct underlying model (d) is difficult, which is signalled, e.g. by the substantial estimation errors of k2 and k3 in (d)

concentration (Bq/ml) 0 1000 2500

401 concentration (Bq/ml) 0 1000 2500

Kinetic Modelling

concentration (Bq/ml) 0 1000 2500

27

60

20

30 40 time (min)

50

60

d k1 = 0.107 +/− 0.004 k2 = 0.135 +/− 0.026 k3 = 0.138 +/− 0.039 k4 = 0.09 +/− 0.007 fbv = 0.092 +/− 0.005 0

10

20

30 40 time (min)

h ¼ ðh1 ; . . . ; hN ÞT ¼ eKt q:

50

60

(27.25)

The response to a finite input is obtained by (component-wise) convolution of the IRF with the input function: c ¼ h ca .16 Due to the appearance of the matrix K in the exponent, the precise meaning of the matrix exponential might not be immediately clear. In fact, it is defined as the N N matrix obtained by the formal Taylor expansion eKt ¼

1 X ðKtÞn n¼0

n!

;

where (Kt)n is the n-fold matrix product of the matrix (Kt) with itself. Accurate computation of the matrix exponential is far from trivial in the general case, but reliable algorithms are known (Moler and Loan 2003). Freely available implementations can be found, e.g., in the numerical open source systems Octave (Eaton 2008) and R (R-Development-Core-Team 2008). For more details concerning the matrix exponential

16

For the general case of input according to Eq. (27.23) one gets: Z t c eKt b eKðtuÞ bðuÞdu; 0

where integration of the vector function is performed component wise.

402

J. van den Hoff

formalism in the context of compartment modelling, cf. Sect. 5.2 in (Bates and Watts 2007). Given a reliable matrix exponential implementation, the formalism allows an easy, efficient and accurate computation of the IRF for arbitrary compartment models via Eq. (27.25) irrespective of their complexity. This is a very valuable tool, e.g. for simulation of complex systems in order to gain insight into consequences of model simplifications. Concerning numerical accuracy (and speed), the matrix exponential approach is especially superior to numerical integration of the model equations. For reference, we finally look a bit closer at the case where K is diagonizable. The matrix exponential can then be expressed via the eigenvector matrix V, its inverse V 1, and the eigenvalues (ln) of K. Equation (27.25) then becomes h ¼ VEV 1 q with Emn ¼

e 0

ln t

m ¼ n; (27.26) m 6¼ n:

This implies that the IRF in each compartment is a weighted sum over the exponentials e ln t . Explicit evaluation of Eq. (27.26) yields: hm ¼

N X

ln t

amn e

with amn ¼ Vmn

n¼1

N X

Vnk1 qk ;

k¼1

(27.27) which simplifies to amn ¼ Vmn Vn11 K1

(27.28)

if there is input only into the first compartment with a rate q1 = K1. The total system response is finally given by h¼

N X m¼1

hm ¼

N X

an e

ln t

with

n¼1

an ¼

N X

amn :

m¼1

Equation (27.26) also shows the relation to the equivalent “decoupled” model introduced earlier. The latter is defined in such a way that its transfer matrix is already diagonal (namely K = diag (l1, . . . , lN)) and has the input vector a = (a1, . . . , aN)T. The corresponding impulse response is g ¼ Ea;

(27.29)

which leaves the total system response unaltered: g¼

N X n¼1

gn ¼

N X

an e

ln t

h:

n¼1

References Bassingthwaighte J, Chan I, Wang C (1992) Computationally efficient algorithms for convection permeation diffusion models for blood tissue exchange. Ann Biomed Eng 20:687 725 Bassingthwaighte JB, Goresky CA (1984) Modeling in the anal ysis of solute and water exchange in the microvasculature. In: Renkin EM, Michel CC (eds) Handbook of physiology. Sect. 2, The cardiovascular system, vol IV, The microcircu lation. American Physiological Society, Bethesda, pp. 549 626 Bassingthwaighte JB, Jr Holloway GA (1976) Estimation of blood flow with radioactive tracers. Semin Nucl Med 6:141 161 Bates D, Watts D (2007) Nonlinear regression analysis and its applications (Wiley Series in Probability and Statistics). Wiley, New York. ISBN 0470139005 Blomqvist G (1984) On the construction of functional maps in positron emission tomography. J Cereb Blood Flow Metab 4:629 632 Carson R (2003) Tracer kinetic modeling in PET. In: Valk PE, Bailey DL, Townsend DW et al (eds) Positron emission tomography: Basic science and clinical practice. Springer, London, pp. 147 79. ISBN 1852334851 Carson R, Breier A, de Bartolomeis A, Saunders R, Su T, Schmall B, Der M, Pickar D, Eckelman W (1997) Quantifi cation of amphetamine induced changes in [11 C] raclopride binding with continuous infusion. J Cereb Blood Flow Metab 17:437 447 Carson R, Cunningham V, Gunn R, van den Hoff J, Knudsen G, Lammertsma A, Leenders K, Maguire R, Mu¨ller Schauen burg W (2003) PET pharmacokinetic course manual. Uni versity of Groningen, Groningen Crone C (1963) The permeability of capillaries in various organs as determined by the use of the indicator diffusion method. Acta Physiol Scand 58:292 305 Eaton J (2008) GNU Octave, http://www.octave.org Innis R, Cunningham V, Delforge J, Fujita M, Gjedde A, Gunn R, Holden J, Houle S, Huang S, Ichise M et al (2007) Consensus nomenclature for in vivo imaging of reversibly binding radioligands. J Cereb Blood Flow Metab 27 (9):1533 1539 Lammertsma A, Hume S (1996) Simplified reference tissue model for PET receptor studies. Neuroimage 4(3):153 158 Lassen N, Perl W (1979) Tracer kinetic methods in medical physiology. Raven, New York. ISBN 0 89004 114 8 Logan J, Fowler J, Volkow N, Wolf A, Dewey S, Schlyer D, MacGregor R, Hitzemann R, Bendriem B, Gatley S et al (1990) Graphical analysis of reversible radioligand binding

27

Kinetic Modelling

from time activity measurements applied to [n 11c methyl] ( ) cocaine PET studies in human subjects. J Cereb Blood Flow Metab 10:740 747 Meyer E (1989) Simultaneous correction for tracer arrival delay and dispersion in CBF measurements by the H2150 autora diographic method and dynamic PET. J Nucl Med 30 (6):1069 1078 Moler C, Loan CV (2003) Nineteen dubious ways to compute the exponential of a matrix, twenty five years later. SIAM Review 45(1):3 49 Mu¨ller Schauenburg W, Reimold M (2008) PET pharmacoki netic modelling. In: Lemoigne Y, Caner A (eds) Molecular imaging: computer reconstruction and practice. Springer, Dordrecht, pp. 189 210. ISBN 978 1 4020 8751 6 (Print) 978 1 4020 8752 3 (Online) Patlak C, Blasberg R, Fenstermacher J (1983) Graphical evalu ation of blood to brain transfer constants from multiple time uptake data. J Cereb Blood Flow Metab 3:1 7

403 R Development Core Team (2008) R: a language and environ ment for statistical computing, R Foundation for Statistical Computing, Vienna. ISBN 3 900051 07 0. http://www. R project.org Renkin E (1959) Transport of potassium 42 from blood to tissue in isolated mammalian skeletal muscles. Am J Physiol 197:1205 1210 van den Hoff J (2005) Principles of quantitative positron emis sion tomography. Amino Acids 29(4):341 353 van den Hoff J, Burchert W, Mu¨ller Schauenburg W, Meyer G, Hundeshagen H (1993) Accurate local blood flow measure ments with dynamic PET: fast determination of input func tion delay and dispersion by multilinear minimization. J Nucl Med 34:1770 1777 Weiss T (1996) Cellular biophysics, vol 1: Transport. MIT, Cambridge. ISBN 0 262 23183 2 Willemsen A, van den Hoff J (2002) Fundamentals of quantita tive PET data analysis. Curr Pharm Des 8:1513 1526

Data Documentation Systems

28

Sönke H. Bartling and Wolfhard Semmler

Contents 28.1 Introduction............................................................. 405 28.2 Necessary Data Management in Small Animal Imaging...................................................... 406 28.3 Solutions.................................................................. 407 28.3.1 Difference from Clinical Radiology Data Management..................................................... 407 28.3.2 Necessary Storage Space.......................................... 408 28.3.3 DICOM for Small Animal Imaging Data................. 409 28.3.4 DICOM Data Storage/PACS..................................... 409 28.3.5 Proprietary and Other Non-DICOM Image Formats.......................................................... 410 28.3.6 Third Party Software Solutions in Small Animal Imaging.............................................. 410 28.3.7 Research Data Handling Systems............................. 410 28.4 Ideal Small Animal Imaging Data Handling Environment........................................... 411 28.5 Software and How-Tos........................................... 411 28.5.1 DICOM Sending and Receiving............................... 411 28.5.2 Understanding Image Files....................................... 412 28.5.3 Manipulating the DICOM Header............................ 412 28.5.4 Pointers to Software and Other Resources............... 412

S.H. Bartling (*) and W. Semmler Department of Medical Physics in Radiology, German Cancer Research Center, Im Neuenheimer Feld 280, 69120 Heidelberg, Germany e-mail: [email protected]

28.1 Introduction With the growing awareness of the advantages of small animal imaging in preclinical research, its use is constantly increasing and small animal scanner systems are constantly improving. Besides imaging with high spatial resolution, dynamic contrast enhanced scans, diffusion weighted imaging, etc. are more and more often part of preclinical routine imaging protocols. This leads to an increasing amount of data that are produced in small animal imaging and have to be handled. Additionally, data have to be stored safely for later retrieval and reinvestigation. Beside the growing amount of data, its handling is aggravated by the lack of suitable standards for small animal imaging data. DICOM data can be generated with many modalities and imaging tools. However, too often these DICOM data are incompatible with standard postprocessing equipment, which particularly holds true for 3D or 4D data. In these cases DICOM can just be used to store and retrieve data and to look at single slices, while 3D or 4D postprocessing is not possible in a straightforward way. In the worst case, manufactures are using their own imaging file formats or third-party formats that can only be read and postprocessed by the scanner system itself. If one wants to access these proprietary data formats, handling the data on a low level becomes necessary. If standards are proprietary and/or not released by the manufactures, accessing data can become a tedious and scientifically unsatisfying hacking experience. The glance at established tools from clinical radiology is often fruitful; however, this approach hits limitations when it comes to advanced postprocessing such as analyzing 3D or 4D data. Some radiology equipment is just not able to postprocess and handle high

F. Kiessling and B.J. Pichler (eds.), Small Animal Imaging, DOI: 10.1007/978-3-642-12945-2_28, © Springer-Verlag Berlin Heidelberg 2011

405

406

resolutions that come along with small voxel sizes that are acquired (or reconstructed) in small animal imaging. Postprocessing may then result in erroneous data. Postprocessing of multimodality data and/or timecourse data using clinical workstations requires compliance with tight image format standards (e.g., within DICOM) that are often not satisfied by small animal imaging scanners and are not compatible with other manufacturers. Again some hacking and data manipulating can make image data compatible with clinical equipment – however, this needs substantial knowledge and/or a long process of trial and error programming – which is often just not possible for a busy researcher. Data management in small animal imaging not only consists of handling imaging-related data: With a growing amount of studies, an effective workflow and resource management (e.g., scheduling of technicians, scientists, and modalities) also become important issues. Furthermore, secondary data that are generated during the analysis process – ranging from simple tumor sizes toward complex time-course data – have to be stored and analyzed. Secondary data should be distinguished from metadata – metadata are data that are not directly related to imaging data, but are also of interest for a research study, e.g., body weight, heart rate, sex, weight, body temperature, behavior, information about the scan procedure itself such as intolerance to narcotics, problems during contrast media injection, and so on. Currently, data handling in small animal imaging is solved on a low-level base, often data are handled on a per-file base and with standard office tools. The same holds true for workflow management. In most institutions data handling in small animal imaging consists of solutions that are a patchwork of work around and labor-intensive makeshift solutions that work for basic scientific experiments; however, as small animal imaging becomes more routinely employed, data handling in small animal imaging should become streamlined and effective. In this chapter, we will give valuable information for data handling for both basic researchers as well as effective routine small animal imagers. Therefore, we will first identify necessary data documentation and handling tasks in small animal imaging. Then we will introduce and discuss solutions that are currently found in small animal data management followed by a description of an ideal small animal data storage/

S.H. Bartling and W. Semmler

workflow environment. Through a glance at clinical radiology data management solutions, the reader will understand how small animal data management varies from clinical radiology management. The focus is on the practical side of the data management in small animal imaging. In the end, practical hints and tips will be given that include a list of mostly freely available and useful software solutions.

28.2 Necessary Data Management in Small Animal Imaging Data are generated during several steps in a small animal imaging study. A scheme of a small animal imaging study together with the related data flow is outlined in Fig 28.1. Often small animal studies involve follow-up studies. During the planning of a small animal imaging study, the data structure becomes defined – the researcher might be more or less aware of this step. Maybe the most visible step is the definition of the primary (e.g., lung parenchyma density in CT or tumor size) and secondary/tertiary target criteria (e.g., lung fibrosis or growing/shrinking of tumors). The amount of scans and the scan protocols are also defined. Then, large amount of data are produced. The raw imaging data (in terms of not scientifically postprocessed imaging data) have to be stored. Often metainformation occurs that should be saved for later retrieval. For the postprocessing, the image data have to be retrieved together with the metainformation. During postprocessing secondary data (e.g., tumor size, perfusion parameters) are generated that might be further stored and analyzed at the end of the study. The data size of this secondary data might also be large – a good example is the generation of perfusion parameter maps. For the final analysis, all primary data and secondary data together with relevant metainformation are re trieved and taken into account. Often complex statistical analysis is performed during this step. The workflow data management encases the whole imaging process. Resource management and time scheduling as well as postprocessing time management have to be performed during all steps of the small animal imaging study.

407

28 Data Documentation Systems Workflow Study design Animal scanning Data postprocessing Follow-up studies

Data analysis Final study analysis

Datahandling

Data flow

Calender

Secondary image data (ROIs, segmented volumes, perfusion maps, ...)

File system

Pen and paper

a

Standard office tools

Manual work

Tools Calender Spread sheet

Spread sheet

Clinical radiological tools

All data

Network

CD-Rom, Other media

Tools

Raw image data (CT, PET, MRI, ...)

Parameters (measurements, volumes, perfusion parameters, ...)

Manual work

Network

Study design related information (scheduling, scan parameters, statistical analysis, ...)

Clinical PACS

Highly integrated software solution

Standard office tools Compiled databases

Compiled databases

Pen and paper

b

c

Fig. 28.1 Data handling. To the left, key steps of a typical small animal imaging study are symbolized and the data flow is exemplified. To the right, three more or less integrated data handling environments are symbolized. An example for a typical patch-

work solution (a), followed by a solution that cointegrates a (clinical) PACS (b) and a highly integrated solution that offers one solution that can handle all aspects of small animal imaging data handling (c) are provided

28.3 Solutions

28.3.1 Difference from Clinical Radiology Data Management

Currently, solutions for small animal imaging data management are patchworks of several nonoptimal interacting components. Often laborious and tedious user tasks are involved in the process of exchanging data between steps of the data flow process. Standard office tools as well as handling data on a low-level base (meaning, e.g., on a per file base) and/ or data in spreadsheets and databases play an important role. Often CDs/DVDs as well as external hard drives are employed for archiving (Fig. 28.1a). Sometimes a solution as pictured in Fig. 28.1b is employed. Here, a soft-/hardware solution exists that can handle or store image data that are produced by the modalities. Frequently, the storage device is a DICOM server or a PACS system (see Sect. 28.3.3). All other tasks, such as postprocessing, target data storage, and workflow data management, are performed using independent systems – often by standard office tools or by using laborious user interactions. The ideal solution is a highly integrated software solution. Here, all processes from study planning to final study analysis are integrated into one software solution (Fig. 28.1c). Even resource scheduling as well as dataanalysis software might be integrated in one platform.

In clinical radiology PACS, RIS (radiology information system) together with HIS (hospital information system) perform the tasks of data management (Fig. 28.2). A clinical PACS archives image data (primary and sometimes secondary). In most cases a unique patient identification number – stored in the DICOM header – allows clear assignment of an examination to a patient. Usually, reading workstations are incorporated into the PACS solution, often proprietary and fast protocols are used to transfer patient data to the reading workstations – a real-time transfer is considered state-of-the-art. However, depending on the local network implementation and software, standard DICOM transfer may also result in fast data transfer. Workflow management, report management, and scheduling are performed in a RIS. In a HIS, radiological as well as other patient related data (metadata) are stored. In clinical radiology the value of a highly integrated software solution(s) with seaming less integration of components is considered state-of-the-art. If a radiology infrastructure is insufficient, the potential consequences affect almost all aspects of clinical radiology. Potential consequences range from patient safety issues

408

S.H. Bartling and W. Semmler

Workflow

Clinical indication

Data flow

Data handling

Scheduling, clinical question/indication, patient specific information (allergies, etc.)

RIS/HIS

Patient scan Raw image data (CT, PET, MRI, ...) Data post-processing

Secondary image data (ROIs, segmented volumes, perfusion maps, ...)

Reading

Clinical PACS

Image data, reports, earlier exams, earlier report Follow-up/Additional modalities

Fig. 28.2 Data handling in clinical radiology. A state-of-the-art highly integrated solution provides all aspects of data management. In contrast to small animal imaging, only a few data sets of one patient have to be present at once, the long-term storage

have to be highly reliable, and data access should be available at all time. The high value of a working, highly-integrated, and reliable system is widely accepted in clinical radiology

(wrong examinations, no follow-up comparison) over economic losses (waiting time, delayed communication) to legal issues. In small animal imaging mostly the economic aspects play a central role, while legal aspects and animal safety being only of secondary concern. A working and efficient highly integrated software solution in small animal imaging decreases the amount of human labor and makes data analysis faster and more precise – therefore decreasing the running costs of small animal imaging. The situation of data management in small animal imaging at the moment can be compared with clinical radiology in the prePACS era of films, laborious task, and unsecure storage. The value of a working software solution is currently beyond doubt in clinical radiology. Small animal data management systems do not have to be certified by legal institutions (yet). Current clinical radiology systems comply with most relevant aspects of the DICOM standard, while they may still hit limitations when it comes to complex data such as coregistered data sets (e.g., PET-CT) and 4D time-course data. In clinical

radiology the current and maybe one scan before that have to be available to read an examination and to write a report; the rest of the patient data may be kept untouched in the long-term archive. However, in small animal imaging, all scans of the animal have to be available at one time point for an analysis of the study. After final analysis the animal data can be omitted or saved in a less secure archive, while in clinical radiology long-term archiving has to be fail-safe (by legal requirements). Depending on the local legislation, radiology data have to be stored up to 30 years. For small animal imaging data, no such requirements exist, leaving it up to good scientific praxis to store primary research data.

28.3.2 Necessary Storage Space Image data need large amount of digital storage space. The actual amount of necessary storage space for a software solution depends largely on the amount of

409

28 Data Documentation Systems

studies performed. The amount of storage room necessary for a study can also vary largely; while single CT data are relatively small, high resolution X-ray projections do require much more storage space. To provide some idea of necessary storage space, some estimates of data sizes are provided: A usual CT or MRI slice needs 514 KB (with a 512 × 512 matrix, 2 bytes depth and standard DICOM header), a whole animal scan 40–200 MB, while a projective X-ray may be in the order of 8–20 MB. Time-course data, multisequence MRI protocols, angiographies, etc. may require a multiple of the named storage space, whereof single ultrasound images, region of interest maps, and optical images require only a fraction of the storage space. Data sizes are subject to increase, and with continuously declining storage prices, the economic relevance of storage capacity varies significantly over time. A small animal imaging data management solution should use a scalable data storage space, if long-term archiving is required. To give a house number, a storage space of 10 TB is certainly a good start for a small animal imaging facility. Data compression may virtually increase storage capacity, but comes along with an increased need of calculation power and potential data incompatibilities. If lossy compression methods are used, data compression methods have to be selected carefully with respect to specific image information.

tags can be assigned for study name, animal number, and so forth. A local standard can be very helpful in retrieving data at a later point. The complexity of the DICOM standard leaves manufactures with a certain amount of freedom in handling their data. Some scanners produce single large DICOM files, while other setups produce multiple files – both basically representing the same data and both in full accordance with the DICOM standard. This openness is prone to some confusion: • The slice position within a 3D data set is stored in the DICOM header and read by software in several different ways. Some software tools are relying on the actual slice position expressed as a number in a series of slices and some other rely on the actual position of the slice in a 3D coordinate system. • Voxel size within a data set is represented in several different ways. Some software packages use slice positions to calculate voxel sizes, while others use actually given voxel sizes. Due to the different representation, wrong calculation of voxel sizes may occur. This problem aggravates if largely overlapping thick slices or slices with large gaps in between are acquired (such as in MRI scanning) and exact volumes have to be calculated.

28.3.4 DICOM Data Storage/PACS 28.3.3 DICOM for Small Animal Imaging Data DICOM is a standard format that was developed for the storage and sending of medical imaging data. It is a very open standard, and from a range of imaging modalities, still images, movies, 3D/4D, and even ECGs and sounds can be encapsulated within DICOM. A DICOM file is some (image) data together with a DICOM header. The header contains – depending on the modality – several tags that contain information about the DICOM image data. This ranges from length and position of the header to actual scan parameters, names of the “patient,” and so on. DICOM was developed for clinical image data – therefore DICOM tags such as patient names and hospital name should not confuse the small animal imager. If the DICOM environment is solely by itself and not integrated to a clinical system/network, these DICOM

If the measured data of the imaging modalities comply at least with the basics of the DICOM standard, standard DICOM sending and retrieving procedures can be used for networking data from the modalities to the storage facility. Most modalities employ DICOM sending standard, here – assuming parameters are set correct – DICOM data can then conveniently be sent through a network. A DICOM network is based on standard computer/ office networks and is another layer of the standard IP/ TCP network. For storage, import, and export of DICOM data, a wide range of solutions exists. They vary in terms of cost, abilities, interfaces, installation effort, etc. (some are listed in the software list below). A PACS can be seen as a very large and flexible DICOM server. Setting up a DICOM server and/or a PACS requires some advanced knowledge of computer networks and DICOM protocols which is beyond the scope of this book.

410

The same holds true for archiving tools. Archiving of DICOM data can be performed by using several ways: CDs/DVDs, external hard drives, up to professional archive solutions that are integrated into a PACS system. If a clinical PACS system is used for small animal imaging, besides legal and cost issues, often certain rules have to be followed for the naming of a study (e.g., unique Patient ID). The unique labeling of studies cannot be overemphasized since a good labeling can ease a later retrieval tremendously. The local system administrator should also be aware of the fact that large research data are stored in the clinical PACS since local implementations of the PACS might be jeopardized by sending nonclinical data.

28.3.5 Proprietary and Other Non-DICOM Image Formats If the imaging modalities create non-DICOM data and/ or other potentially proprietary file formats, the data management cannot rely on DICOM facilities. Data management has to be done either by the manufactures software or on a per file base. A wide range of standard imaging formats exists, often some libraries exist, that can be used to import such file formats. In other cases the standard is public and the image data can be accessed by using the raw file data. Most often, an image header with DICOM metainformation of the study as well as scan parameters is found (Here, DICOM metainformation may contain study metainformation as defined above, but is not limited to them; it may also include information that is associated to the DICOM structure itself, etc.). Using available programs and libraries, DICOM data can be generated from those non-DICOM files – however, great care has to be taken here to generate right and useful DICOM headers.

28.3.6 Third Party Software Solutions in Small Animal Imaging The use of third party solutions to analyze and further postprocess image data is wishful if not all analysis tools are available in the software packages that accompany the imaging modality.

S.H. Bartling and W. Semmler

If the modality and the third party software comply with DICOM standard, usually DICOM networking tools are implemented and communication becomes straightforward. The only potential pitfalls here are the lack and/or misinterpretation of DICOM header information as mentioned above. If no DICOM networking capabilities on the modality exist, but the image format is DICOM, usually CDs or DVDs can be burned that can then be imported into a DICOM server or PACS that can then be used to send the DICOM data to the third-party analysis solution. If no DICOM data are produced, data handling has to be performed on a per file base. Potentially DICOM data could be generated– however, sometimes it is easier to import the image data using a raw-image data import function of third-party software. Then the image header has to be omitted and the format of the raw-image file data has to be provided, which requires some knowledge of the image data format or some trial and error hacking. If the data should be handled as 3D or 4D, data basic geometric and time information have to be provided as well. If this information is not incorporated in an image header or difficult to find in the header, it can be necessary to type it in manually.

28.3.7 Research Data Handling Systems Nowadays, with the increasing amount of all kind of data that are produced in biomedical research, the problem of handling large amounts of data is not limited to small animal imaging. Therefore, general research data management systems have been developed. Such systems can generally handle a wide variety of data, store it, and archive it. Often, an exchange of data over the internet and communication between researchers and research groups is possible. Often, small animal imaging data can also be included in those solutions. Depending on the flexibility of the system and the import/export functionality, this can be done on a per-file base, per DICOM, or other file formats. Often workflow functionality is integrated into such systems. The advantage of these systems is that nonimage data can easily be coincorporated in the image data, making it possible to store related metainformation and results for a later analysis.

411

28 Data Documentation Systems

28.4 Ideal Small Animal Imaging Data Handling Environment An ideal data handling environment for small animal imaging would facilitate and accelerate several aspects of the workflow of small animal imaging. Small animal imaging in service would benefit more from it than small animal imaging research by itself, since the latter is more focused on the development of imaging methods, while the former relies usually on a large number of studies performed. To the best of our knowledge, no highly integrated small animal imaging environment exists so far. Therefore, a potential solution is outlined in the following: An ideal data handling environment would incorporate study planning, workflow management, PACS or DICOM-server functionality, import-/export functionality, nonimage data storage, and basic postprocessing functions with well integrated and communicating components (Fig 28.1c). Functional components could be the following: Study design: The study design functionality provides tools to design a study, define data points, and set the scanning schedule. The planning tool communicates with the workflow management and the other parts of the environment to set up schedules for scanning and prepare the necessary data fields of the nonimage data storage. Workflow management: Workflow management would consist of a scheduling functionality as well as worklists for necessary postprocessing steps. Potentially automatic export and import functions might be incorporated into export data to third-party postprocessing tools. Scheduling functionality: The scheduling functionality will reserve necessary scan time slots on scanners as well as on required personal and equipment. The scheduling works with the workflow management tool to set up scheduling times that are within the range of the designed study setup. The workflow management system will handle workflow tasks in a smart way: For example, it will place an imaging study on the work list for the next postprocessing/image analysis step after successful scanning, providing automatically an interface to store metainformation that may be generated during postprocessing. PACS or DICOM-server functionality: Basic PACS functionality that should incorporate data import and export, safe-handling, and storage should be available.

Image data that are not DICOM should be changed into DICOM data through versatile DICOM import functions. Import/Export functionality: It is unlikely that a small animal imaging environment will be able to combine all functionality that could become necessary at some point. Therefore, concise export and import functionality have to be implemented for all kind of data ranging from image data to results. Ideally, this would not only allow a DICOM export and sending function, but would also support other common image and data formats. For example, target data and metainformation should be exportable in standard formats, e.g., spreadsheet formats. The data can then be used for further analysis using advanced statistical programs.

28.5 Software and How-Tos In this section concrete hints and pointers to solutions for the daily work of small animal imagers are provided. Detailed procedural descriptions, parameter lists, and hard links are largely avoided since they can be retrieved from internet search engines with ease.

28.5.1 DICOM Sending and Receiving DICOM and the DICOM networking standard can be used to easily transfer an examination or a whole set of examinations in between a workstation, PACS, and DICOM server or modality (“DICOM nodes”). Usually, the researcher is confronted with setting up a DICOM connection between a scanner, workstation, his own computer, or other “DICOM nodes.” Additionally, some parameters have to be configured on both ends. These are IP, port, and application entity title (AET). The sending device needs to know the IP and Port of the receiving one, whereof the AET has to be the same on both machines and can be freely chosen. The IP of the receiving computer can be found in the basic system preferences. The Port of the DICOM server of the receiving machine as well as the AET can be found within the configuration of the receiving DICOM server program. The sending program needs to know the IP, Port, and AET of the receiving program.

412

Further information and How-Tos can be found using standard web search engines with the search words “DICOM,” “network,” and “How-to.” A good reference page with useful pointers to other webpages is: www.sph.sc.edu/comd/rorden/dicom.html.

28.5.2 Understanding Image Files In the optimal case, the researcher is not confronted with the structure of the underlying image files. However, in situations of data incompatibilities or if some nonstandard image manipulation has to be performed, some understanding of the image file structure is necessary. Two different image file formats can be distinguished: Image files with header and image files without header. In the latter case, usually the metainformation of the images such as scan time, scan parameters, resolution, etc. are given with another file that is usually stored in the same directory. In the other case, an image header is placed before the actual image information. The header contains – among others – information about the scan, the resolution, voxel sizes, and so on. The header is then followed by the actual image data. The actual image data are given in various formats. Omitting the header byte length allows access to the image data – however, some image formats allow the header to be of various lengths. Here, access and understanding of the header information are necessary. Image data size is usually calculated by the x-resolution times y-resolution times byte depth of one voxel (however, this holds true only for uncompressed image data). A voxel can – depending on the amount of colors or gray shades that can be distinguished – be represented by various amount of bytes (word, integer, float, signed, unsigned, etc.). Furthermore, the representation of the voxel in the file can be distinguished in “littleendian” and “big-endian.” Be aware of the fact that it matters for importing raw-image data whether it is read x–y or y–x – so trying to change the resolution in x and y direction during import might help! Making matter short, image importation needs a lot of information that have to be found out either by some documentation or by back-engineering/hacking of the file format. The latter can be a difficult and tedious task involving try and error; however, success can be very rewarding. More help and how-tos can be found in the

S.H. Bartling and W. Semmler

internet using search words such as “image,” “file format,” “hacking,” “how-to,” and so on. Sometimes it helps to search for the suffix of the file name and/or the specific modality since people posted their experience with the image format to internet.

28.5.3 Manipulating the DICOM Header As mentioned, various information are stored in DICOM headers. Changing of DICOM header tags might be necessary to sort image data, making it better accessible or even correct for wrong geometry and/or time information that might lead to erroneous postprocessing results. Some software tools exist that allow this.

28.5.4 Pointers to Software and Other Resources DCMTK-DICOM-Toolkit: A set of command line parameters for handling and manipulating DICOM files – it includes sending and receiving of DICOM files. For example, “storescu-aet SAI1 172.21.52.146 104 *” allows to send all the files in the current, local directory to a DICOM server that runs on 172.21.51.146, at the Port 104 with the AET SAI1. DICOM standard: A comprehensive overview of DICOM standard and software can be found at: http:// www.sph.sc.edu/comd/rorden/dicom.html DICOMworks: DICOMworks allows the visualization, manipulation, and sending of DICOM data – it is freely available for PC and provides a convenient graphical user interface. Efilm: Efilm is a DICOM server, local database, and image viewer. It works for PC and is designed for a clinical environment. It provides some basic postprocessing tools. ImageJ: Is a freely available image postprocessing, import, and conversation platform with a multitude of plug-ins – a script feature allows automation of tedious and complex tasks. Matlab: Matlab is a developing platform that includes, among others, comprehensive image handling and DICOM sending functionality.

28 Data Documentation Systems

Osirix: Osirix is a DICOM server, local database, and image viewer. It is freely available, but only for the Apple system. It is open-source and plug-ins can be programmed that can be used to analyze software. PACS systems: www.rtstudents.com/pacs/index. htm

413

SimpleDICOM: A simple DICOM receiver and sender. It can be used to receive DICOM images through a network as sent by a workstation or PACS system and to store them locally for further postprocessing. NEMA/DICOM: More links and Information regarding DICOM standard http://medical.nema.org/

Part Special Applications

VI

Imaging in Developmental Biology

29

Katrien Vandoorne1, Stav Sapoznik1, Tal Raz, Inbal Biton, and Michal Neeman

Contents 29.1 Animal Models for Developmental Imaging........ 417 29.1.1 Imaging Development in Fish and Frogs.................. 417 29.1.2 Imaging Avian Development.................................... 418 29.1.3 Imaging Development in Mice................................. 419 29.2 Imaging Modalities................................................. 420 29.2.1 Ultrasound Biomicroscopy....................................... 421 29.2.2 Magnetic Resonance Imaging.................................. 423 29.3 Parametric Mapping of Development................... 426 29.3.1 Anatomical Imaging of Development....................... 426 29.3.2 Functional Parametric Mapping............................... 429 29.3.3 Molecular and Cellular Mapping of Development............................................................. 430 29.3.4 Specific Applications................................................ 431 29.4 Animal Preparation................................................ 434 29.4.1 Anesthesia Procedures.............................................. 434 29.4.2 Animal Preparation and Physiological Monitoring................................................................ 434 29.4.3 Fetal Identification and Genotyping......................... 435 29.5 Summary................................................................. 435 References............................................................................ 435

29.1 Animal Models for Developmental Imaging Biological imaging studies of fetal development are frequently conducted on small laboratory animals, which offer the advantages of rapid reproductive cycle and multiparity. The first section of this chapter will screen the most widely used animal models. Animal models aiming to study human physiology or disease by noninvasive imaging should exhibit along with genetic, anatomical, and physiological similarities to humans, also the ability to provide information using available imaging modalities. Many developmental studies utilized the rapid reproduction, easy access, and optical clarity of developing avian and fish embryos for high-resolution fluorescence microscopy, while studies of mammals were frequently limited to ex vivo imaging. However, over the last years, new imaging tools allow in vivo monitoring of development also in the mouse, which is the most common mammalian model for the study of development, genetics, immune response, pathology, neurology, and cellular mechanisms of action.

29.1.1 Imaging Development in Fish and Frogs

K. Vandoorne, S. Sapoznik, T. Raz, I. Biton, and M. Neeman (*) Department of Biological Regulation, Weizmann Institute of Science, Rehovot 76100, Israel e-mail: [email protected]

An important animal model for optical imaging of developmental processes is the transparent zebrafish embryo (Ingham 2009). The adult zebrafish is small (3 cm in length) and can therefore be easily maintained in large numbers. The generation time is short, as KV and SS contributed equally to this work

1

F. Kiessling and B.J. Pichler (eds.), Small Animal Imaging, DOI: 10.1007/978-3-642-12945-2_29, © Springer-Verlag Berlin Heidelberg 2011

417

418

sexual maturity is achieved within 3 months after conception. Spawning is external and easily controlled, resulting in about 200 eggs per event. Hatching occurs within 2–3 days, and the entire developmental process is rapid, as the major organ systems largely develop by 6 days postfertilization. By regulation of the environmental temperature, it is possible to regulate the rate of physiological processes, including the embryonic development. Contrary to mammals, the entire embryogenesis of the zebrafish takes place outside the maternal body, facilitating optical imaging and enabling direct manipulations. Transparency of the embryos persists naturally for 32 h postfertilization but can be extended experimentally for several days. Importantly, many fundamental morphogenetic processes are evolutionary conserved between zebrafish and mammals (Ingham 2009). Genetical engineering of the zebrafish embryos to express fluorescent protein reporters or injection of vital dyes or microscopic particles can be used for molecular imaging. In vivo optical imaging is feasible by anesthetizing the embryos and mounting them in agarose to reduce movements. Various physiological systems of the zebrafish embryo were studied by optical imaging. Taking advantage of the superficial ventral position of the heart, cardiovascular development and its response to either pharmacological or genetic manipulations were investigated. For example, the anatomical patterning of the primary vascular network was imaged using an EGFP encoding embryos under the regulation of the fli1 promoter (Vogt et al. 2009). Macrophages mechanisms of action and the development of the nerve system in zebrafish embryos were studied by using differential interference contrast microscopy and harmonic generation microscopy, respectively. Recently, zebrafish were utilized for delineating the formation of the lymphatic vasculature (Yaniv et al. 2006) (Fig. 29.1). In addition to optical imaging, zebrafish development was also followed by ultrasound, magnetic resonance imaging (MRI), and computerized tomography (CT). Similar to zebrafish embryos, the developing frog embryos are transparent and can be easily imaged using light microscopy; therefore, they are extensively used in development studies. Notably, the xenopus oocytes were also used for imaging development by MRI, using the reporter probe EgadMe that reveals the activity of the reporter gene beta-galactosidase (Louie et al. 2000).

K. Vandoorne et al.

29.1.2 Imaging Avian Development Another widely used animal model for embryonic developmental studies is the avian egg, and particularly the chicken and quail embryos. Chick embryos are relatively transparent and flat and, like the zebrafish embryos, develop outside the maternal reproductive system (Ezin and Fraser 2008). The developmental process is relatively rapid, as blastulation, gastrulation, and early neurulation are completed within 36 h. The quail embryonic development is slightly faster. Optical imaging takes advantage of the embryo’s transparency and the various possibilities for cell labeling. Fluorescent labeling of avian embryos is often induced by microinjection of either lipophilic dyes or hydrophilic polysaccharides into living cells. Furthermore, microinjection, electroporation, or viral transfection of fluorescent proteins is commonly utilized. Interestingly, quail cells transplanted into chick embryos can be easily distinguished from the host cells due to unique nuclear characteristic. In this way, cell lineage studies can be carried out, avoiding the need for staining (Kulesa 2004; Tirosh-Finkel et al. 2006; Rinon et al. 2007). Several experimental procedures were developed to allow avian embryogenesis imaging. Many studies are based on in vitro culturing of the embryos, where the embryo is collected with as much of its vitelline membrane and laid with its ventral side up inside an incubator, which controls the temperature, humidity, and CO2 concentration. Nutrition of the embryo is provided by culturing it on a pool of thin egg white, bacto agar, or culture medium (Ezin and Fraser 2008). This in vitro method allows experimental manipulations of the embryo, as well as both dorsal and ventral imaging. Alternatively, embryos can be imaged in vivo, thro ugh a window opened in the eggshell. This method allows a long imaging duration, while disruption of the embryonic tissue is minimal. Newly developed sagittal slice culture procedures enable imaging of deep structures previously optically unexposed, such as the developing peripheral nervous system (Kulesa 2004). Furthermore, the chick embryo model was used to investigate developmental processes such as cell lineage and migration during axis formation, and cardiovascular, craniofacial, limb, skeletal, and vascular patterning (Kulesa 2004; Tirosh-Finkel et al. 2006; Rinon et al. 2007).

419

29 Imaging in Developmental Biology

a

Angiography Lymphangiography

Subcutaneous Injection

DA LV PCV

b

e

c

f

d

g

Fig. 29.1 Functional characterization of zebrafish lymphatic vessels. (a) Methods used to image blood or lymphatic vessels. Red, approximate region of the trunk imaged in (b–e) and (g). Blue, approximate region of the trunk imaged in (f). LV, lymphatic vessel (thoracic duct). (b) Angiography of a 14-d.p.f. Tg(fli1:EGFP)yl zebrafish (green) injected with fluorescent microspheres (red), labeling dorsal aorta (DA; large arrow) and PVC (asterisk) but not lymphatic thoracic duct (small arrow). (c) Lymphangiography of 3-week postfertilization Tg(fli1:EGFP) yl zebrafish (green) injected with fluorescent microspheres (red), labeling thoracic duct (small arrow) but not dorsal aorta (large arrow). (d) Time-averaged confocal image of a 7-d.p.f. Tg(fli1:EGFP)yl (green) and TG(gata1:dsRed) (red) zebrafish,

showing red fluorescence in the dorsal aorta (large arrow) and cardinal vein (CV; asterisk) but not lymphatic thoracic duct (small arrow). (e–g) Confocal imaging of an 18-d.p.f. Tg(fli1:EGFP)yl zebrafish (green) injected subcutaneously with 2 Md rhodaminedextran (red). (e) Subcutaneously injected rhodamine-dextran drains into the thoracic duct (small arrow) but does not label the adjacent dorsal aorta (large arrow). (f) Numerous rhodaminedextran–labeled vessels (red) are visible between the blood vessels (green). (g) Higher-magnification image of blind-ended rhodamine-dextran–labeled vessels. Scale bars in (b–d, g), 50 m; in (e), 20 m; in (f), 100 m. (Reprinted with permission from Macmillan (Yaniv et al. 2006))

Avian development tracking by MRI is feasible and provides the advantage that the egg shell can remain unperturbed (Hogers et al. 2009). Thus, wild type quail embryos imaged by MRI showed the entire process of heart development. Cardiac malformations including ventricular septal defects and aortic arch interruptions could be visualized in venous clipped embryos (Hogers et al. 2009).

29.1.3 Imaging Development in Mice The genetic, anatomical, and physiological similarity to human led to the adoption of the laboratory mouse as the most widely used model for the study of reproduction and development. Sexual maturity of mice is achieved early in postnatal development, at 5–8 weeks of age. The estrus cycle lasts for 4–5 days, and once

420

the female is pregnant, the whole gestation period lasts for only 19–21 days. The average litter size is usually 6–8 pups, but can reach much higher values depending on the mouse strain. The cardiovascular and respiratory systems of mice are very similar to human (Schneider and Bhattacharya 2004). Both have a four-chambered heart, with a septated outflow tract, resulting in physical separation of the pulmonary and systemic circulations. The respiratory systems in both cases consist of lungs, alveoli, and diaphragm, being the major respiratory muscle. Similarities between the two organisms are found also in other physiological processes, including reproduction and skeletogenesis. Importantly, a high degree of phylogenetic similarity between human and murine genomes facilitates identification of genetic homologues between the two species. This kind of homology provides the justification for wide efforts for identifying the role of human genes by understanding their function in the mouse model. An example for such a research strategy resulted in the identification of numerous genes involved in congenital heart defects (Schneider and Bhattacharya 2004). Moreover, manipulations of known human genes can be carried out in the mouse, enabling a better understanding of the gene function, spatiotemporal expression, and mutagenesis. Methods for genetic manipulation in mice include overexpression of genes for studying gain-of-functionrelated phenomena, whereas deletion or replacement of a specific gene by another sequence (knockout or knock-in respectively), is used for studying loss-offunction-related phenomena. Mutagenesis is often used for identification of a specific locus within a gene, bearing a unique feature, which is important for the gene function. Supplementary techniques, often used in embr yology to study development processes, may also be combined with imaging techniques. An example is tetraploid complementation, which is used to create an embryo of a specific genetic background bearing extraembryonic tissues of a different genetic background (Tanaka et al. 2009). This method is used for analysis and prevention of in utero lethality in transgenic mice due to defects in placental function. The fact that the entire embryonic development occurs in utero entails major challenges for in vivo imaging. Thus, most studies resorted to invasive methods, such as ex vivo culture and imaging of preimplantation embryos,

K. Vandoorne et al.

static postimplantation embryos culture on a microscope stage, collection and fixation of embryos for ex vivo analysis, and culturing of embryonic slices (Passamaneck et al. 2006). Novel imaging methodologies, predominantly MRI and ultrasound, now offer the possibility for longitudinal noninvasive in vivo imaging of the mouse embryonic development.

29.2 Imaging Modalities To date, most imaging studies of the developing mouse and rat embryos and neonates were performed ex vivo. The main ex vivo techniques for mouse imaging are optical projection tomography (OPT), microcomputed tomography (mCT), and micromagnetic resonance imaging (mMRI). OPT is well-suited for whole-mouse embryo during the early stages of development and can produce 3-dimensional (3D) reconstruction of embryo tissue. OPT has a spatial resolution of 10 mm and allows the ability to map gene expression patterns through the visualization of immunofluorescent staining, along with other vital stains commonly used in histology (Dickinson 2006). mCT can distinguish air, fatty, nonfatty, and bone tissue. Furthermore, by injecting radio-opaque polymers into the vascular space, blood vessels can be imaged with a resolution of 10 mm. With the use of contrast media, CT can be used to visualize the entire vascular network of embryos and their placentas ex vivo (Kulandavelu et al. 2006). A number of imaging modalities are currently available for in vivo analysis of rodents including ultrasound biomicroscopy (UBM), MRI, fluorescence and bioluminescence imaging (BMI), CT, positron emission tomography (PET), and single photon emission tomography (SPECT). In the selection of imaging modality for developmental biology, it is important to take into consideration physiological side effects on the pregnant mice and the embryos, particularly in longitudinal in vivo studies. In this respect, MRI and UBM appear to be the tools of choice for most studies (Table 29.1). The use of ionizing radiation (X radiation), such as in CT, may cause embryonic abnormalities (Bang et al. 2002), thus limiting the use of CT for longitudinal monitoring of fetal development. Specifically, exposure at embryonic day (E) E11.5 to radiation dose of 0.5–4 gray

421

29 Imaging in Developmental Biology Table 29.1 UBM and MRI for in vivo imaging of prenatal development Ultrasound biomicroscopy (UBM) (Phoon 2006)

Magnetic resonance imaging (MRI) (Turnbull and Mori 2007)

Transabdominal: limited number of embryos visible

Implantation visible with contrast-enhanced (CE) MRI (Plaks et al. 2006)

Applications Early-stage embryo

Semi-invasive: transuterine all embryos visible Late- stage embryo

Live embryo imaging: both transabdominal and transuterine imaging

Manganese-enhanced (ME)MRI for imaging central nervous system (CNS) (Deans et al. 2008) Cardiac imaging (Nieman et al. 2009)

Late embryonic brain imaging not possible (mineralization of the skull)

Molecular imaging of reporter gene expression (Cohen et al. 2007)

Neonates

Cardiovascular system and other soft tissues

Diffusion tensor (DTI) and MEMRI for CNS

Advantages

Dynamic, real-time

Three-dimensional (3D)

Functional data: Doppler M-mode

Nondestructive nature, excellent tissue contrast

Quantitative information on structure and function

Contrast resolution by administration contrast agents Mn2+, macromolecular Gd, cell tracking

Temporal resolution/image acquisition time Portable, versatile Disadvantages

Images limited by gas, bone, depth

Temporal resolution/image acquisition time

Operator-dependent image acquisition

Physiological motion

No information on biochemical, metabolic processes Limited depth of penetration and poor tissue contrast

Limited SNR Low contrast resolution with difficulty in administration of the contrast agent High cost of the equipment

Image resolution

30–40 mm axial and 70–90 mm lateral resolution

Spatial

Depth of penetration close to 5–15 mm

Temporal

Real-time, but each embryo needs to be scanned separately

(Gy) led to developmental defects, such as dilatation of the cerebral ventricles, cleft palate, growth retardation, and reduced head size with increased dose. Failed pregnancies, including embryonic mortality and resorptions, were significantly more frequent when animals were exposed to X radiation levels above 4 Gy. Similarly, repeated use of radioactive tracer material and detection of gamma rays, another type of ionizing radiation, in SPECT/PET can be potentially harmful for embryos. However, a recent study demonstrated feasibility for in vivo CT imaging of nearterm pregnant rats showing exquisite bone contrast and the use of contrast media for visualization of the

Spatial resolution 100–200 mm

Several minutes per scan, but whole uterus imaged

maternal circulation in the placenta (Winkelmann and Wise 2009) (Fig. 29.2).

29.2.1 Ultrasound Biomicroscopy Ultrasound imaging is a safe method commonly used to monitor human pregnancy. In the last decade, UBM has emerged as a key in vivo imaging tool for noninvasive in utero imaging of live rodent embryos, allowing ultrasonic visualization of living tissue. UBM transducers use higher frequencies (40–100 MHz) than diagnostic clinical transducers (typically 2–15 MHz).

422 Fig. 29.2 Micro-CT imaging of rat fetuses. (a) In vivo imaging of a pregnant rat (E21). A single coronal slice is shown on the left with multiple fetal skeletons visible (white arrows). On the right, a volume rendering of the same data set indicating individual fetuses (red arrows). (b) In vivo contrast-enhanced imaging of a pregnant rat (E21). The coronal slice on the left shows multiple contrast-enhanced placentas (white arrows). Individual fetal sacs can be visualized as well as the urinary bladder (red arrow). The volume rendering on the right details the placenta structure and vascular supply (red arrows). (c) Volume renderings of ex vivo imaging of a single rat fetus show detailed anatomical features. Scale bars = 1 cm. (Reprinted from Winkelmann and Wise (2009) with permission from Elsevier)

K. Vandoorne et al.

a

b

c

To date, there is no significant evidence for deleterious effects of UBM on the developing mouse embryo. However, slight growth retardation was documented for mice exposed to high-frequency ultrasound between E8.5–10.5 (Brown et al. 2004). Akin to clinical imaging modalities, UBM-Doppler imaging modalities include cross-sectional or twodimensional (B-mode) imaging, one-dimensional M-mode imaging, and spectral pulsed-wave Doppler.

The latter permits the detection of flow velocities within a specified region of interest, or sample volume. Two important determinants of the capabilities for embryonic imaging are the size of the sample volume and the frequency of the Doppler incident beam; the smaller the sample volume, the more specific the flow detection of regions within the developing embryo. Due to limitations inherent in mechanical transducers, and the trade-off between scan speed and

29 Imaging in Developmental Biology

minimal detectable blood flow velocities for the frequencies and spatial resolutions needed, only recently, color flow imaging with UBM became commercially available. In very small rodents, imaging developmental processes mandates high transducer frequencies to achieve adequate spatial resolution. UBM scanners cover high spatial resolution of 30–40 mm axial and 70–90 mm lateral depending on the frequency. Lower frequencies used for clinical transducers appear ideal for deeper tissue penetrations, whereas high-frequency UBM imaging penetrates only 5–15 mm deep; this range, however, is quite sufficient for subsurface, in utero imaging of embryonic mouse (Fig. 29. 3). Transabdominal visualization of early-stage embryos allows imaging up to four embryos per litter, but without localizing them within the litter. To overcome these technical limitations, a semi-invasive approach was established in which the uterine sac is exteriorized through an incision in the abdominal wall into warmed physiologic saline in a petri dish. Apart from the perturbating nature of this procedure, this technique has yielded excellent physiological imaging conditions, and it allows better resolution of embryonic structures. Furthermore, the technique allows longitudinal followup of early-stage embryos and the possibility of mic roinjection. Temporal resolution is important for moving objects, such as the rapidly beating embryonic heart (Kulandavelu et al. 2006) (Fig. 29. 4). The temporal resolution used to be better on clinical ultrasound systems than on UBM systems, but nowadays, the frame rates of UBM systems (60–100 Hz) using mechanical sector scanning approach those of clinical systems. Current transducer technology yields a relatively narrow zone of focus (depth of field), but the development of annular array transducers should increase the distance over which an image remains in focus (Phoon 2006).

29.2.2 Magnetic Resonance Imaging MRI provides isotropic 3D noninvasive imaging, acquisition of dynamic, functional information, and the possibility for noninvasive longitudinal studies of development. Ex vivo mMRI offers high spatial resolution (25–50 mm), albeit at long acquisition times (6–24 h). Spatial resolution for in vivo imaging is

423

limited by physiological motion, signal-to-noise ratio (SNR), and scanning time and is typically on the order of 100–200 mm (Deans et al. 2008; Turnbull and Mori 2007; Nieman et al. 2009). Respiratory, cardiac gating, self-gating, or image coregistration are required for in vivo studies to reduce motion artifacts (Fig. 29.5) (Turnbull and Mori 2007; Nieman et al. 2009). The available resolution may be sufficient for analyzing the developing brain and heart of mid- to late-stage mouse embryos (E13.5–18.5 days). Effective analysis of organogenesis in smaller early-stage mouse embryos (E9.5–12.5 days) requires better spatial resolution of at least 50 mm (isotropic), and thus, advances in hardware for small-animal imaging would be beneficial (Turnbull and Mori 2007). Contrast resolution is as important as spatial resolution for detection of anatomical features. Contrast resolution presents a special challenge in the developing mouse, where many cells and tissues are undifferentiated or immature, resulting in minimal differences in the endogenous MR relaxation properties. Sequence design in combination with administration of targeted or physiologically relevant contrast agents can improve MRI contrast (Turnbull and Mori 2007). The effective maternal-fetal barrier provided by the placenta excludes the transfer of most contrast media from the fetal circulation. Early-stage pregnancy can be detected by contrast-enhanced (CE) MRI, which highlights the maternal vasculature in the implantation sites (Fig. 29.6), while at later phases of pregnancy, CE MRI highlights the maternal circulation of the developing placenta (Salomon et al. 2005; Plaks et al. 2006; Salomon et al. 2006; Taillieu et al. 2006; Plaks et al. 2008). The central nervous system (CNS) of in vivo latestage embryos was imaged in a study using manganese-enhanced MRI (MEMRI), which passes the placenta barrier and can depict neuronal activity in the developing brain (Fig. 29.7) (Deans et al. 2008). The premyelinated neonatal mouse brain provides little relaxation-based (T1, T2) contrast, which is generally dominated in adult mice by regional myelin concentration. Neuronal-specific contrast can be achieved using MEMRI showing neuronal function or diffusion tensor imaging (DTI), which is sensitive to the structural formation of neuronal bundles (Turnbull and Mori 2007). DTI was applied in order to elucidate brain development in neonatal mice (Zhang et al. 2003; Mori et al. 2006).

424

Fig. 29.3 Functional and structural imaging of mice embryos. B-mode (a–g, i), pulse-wave Doppler mode (h, k), and M-mode (j). (a) Implantation at E5.5 can be visualized at the level of the uterine cavity as a bright hyper echogenic spot (arrows); (b) Living embryo at E9.5; (c) Long-axis view of embryo at E11.5 allows quantifying its length; (d) E14.5 forepaw; (e) E18.5 eye; (f) E10.5 brain with third ventricle, and telencephalic vesicle (t, future aqueduct); (g) E14.5 placenta (p), with umbilical vein (uv) and umbilical artery (ua); (h) Doppler velocity waveforms of

K. Vandoorne et al.

E14.5 umbilical vein (uv) and umbilical artery (ua); (I) E16.5 embryonic heart with left ventricle (LV), right ventricle (RV), left atrium (LA), and right atrium (RA); (j) M-mode of E16.5 embryonic heart with diastolic internal diameter (LVID(d)) and systolic diameter, (LVID(s)) of the LV; (k) Doppler waveforms showing LV inflow and outflow velocities of E16.5 embryonic heart. The inflow includes the passive E wave and the active A wave, whereas the ejection time (ET) contains the outflow into the aorta (E Embryonic day) (Courtesy of Vandoorne et al., unpublished)

29 Imaging in Developmental Biology

Fig. 29.4 Doppler blood velocity waveforms obtained at various sites in the embryonic circulation. Scale bars in cm/sec. This figure originally appeared in an article by Shathiyah Kulandavelu et al. in ILAR Journal 47(2). It is reprinted with permission from

425

the ILAR Journal, Institute for Laboratory Animal Research, The National Academies, Washington DC (www.nationalacademies.org/ilar). (Kulandavelu et al. 2006)

426

K. Vandoorne et al.

Fig. 29.5 In utero cardiac MRI of E17 mouse embryo. Embryonic displacement during the cardiac scan was followed by image registration of serially acquired 3D images. (a) Five examples of distance displacements within the initial imaging slab (b; curve 5 shows displacement outside the imaging slab). Uncorrected for embryo displacement results in blurred images (c), where longand short-axis images from the same 3D data set are shown at both diastole and systole. (d, f) Correction of image displacement

with registration improves image quality. (e, g) Segmentations of the chambers emphasize the changes in 3D shape over the cardiac cycle. Image parameters: TE/TR=6.0/40 ms, flip angle=16°, matrix size = 192 × 78 × 48, NR = 64, TA = 2 h 40 min, isotropic resolution = 130 m, reconstructed with six 3D image frames per cardiac cycle. In (b), B brain; S spine; H heart. (Reprinted with permission of John Wiley & Sons, Inc. from Nieman et al. 2009)

29.3 Parametric Mapping of Development

relatively low temporal resolution. Ultrasound provides better temporal resolution, while CT provides the best resolution for mineralized bones.

29.3.1 Anatomical Imaging of Development The most widely used application of noninvasive imaging in development is for detection of anatomical parameters. MRI, CT, and ultrasound provide three-dimensional information that can be used for parametric analysis of fetal anatomy in mice. The three methodologies are complementary due to the differences in contrast, spatial, and temporal resolution. MRI provides high contrast for soft tissues and high spatial resolution but

29.3.1.1 Following Anatomy by Ultrasound Biomicroscopy Embryonic size and cross-sectional area can be determined as for clinical monitoring of pregnancy (Fig. 29.3). Similarly, UBM can be used to obtain realtime images containing quantitative measurements of dimensions and morphological information of different fetal organs of the mouse, such as the eye, brain, heart, and placenta (Phoon 2006). Acquiring in vivo

29 Imaging in Developmental Biology

427

Fig. 29.6 Quantitative MRI assessment of mouse fetal implantation. Maximal intensity projections acquired 7.5 min after i.v. injection of biotin-BSA-GdDTPA of E4.5 (a) and E5.5 (b) implantation sites (arrows; individual implantation sites). (c, d) Ex vivo visualization of implantation sites highlighted by i.v. administration of Evans blue for the mice shown in (a) and (b),

respectively. (e) Correlation between the number of implantation sites detected by MRI in vivo and the number detected ex vivo (N = 11 total mice; N = 5 spontaneously nonpregnant; overlapping points at 0,0). ov ovaries; K kidney; B bladder. (Reprinted with permission of John Wiley & Sons, Inc. from Plaks et al. 2006)

cardiac morphology is difficult because of the small size of individual structures (valves or myocardial wall) and the rapid beating (200–300 beats/min). However, major events in cardiac morphogenesis can be visualized, such as cardiac looping (E9.5); outflow tract septation (E11.5, E12.5); and ventricular chamber septation (E12.5, E13.5), including the appearance of the atrioventricular valves (E13.5) (Fig. 29.4). Nevertheless, it remains difficult to determine the internal anatomy of the embryonic heart at early gestational stages before E15.5, since the nucleated red blood cells in the mouse embryo are echo-dense (Le Floc’h et al. 2004), thus reducing the ability to discriminate boundaries (Phoon 2006). The developing neural tube (brain, spine) in the early- and mid-gestational embryo can be visualized, although UBM is not well-suited for imaging late

embryonic brain development, because of the limited penetration of high-frequency ultrasound through the skull (Deans et al. 2008). Developmental processes such as implantation, growth, and placentation have been imaged in a qualitative manner. The inner cell mass of the recently implanted embryo (E5.5) and the three cavities of the embryo demonstrated that the amniotic cavity, the coelomic cavity, and the ectoplacental cavity (E7.5) could be visualized. The placenta could also be imaged from very early gestation. Embryonic structures such as lungs, liver, limbs, and eye can also be imaged, illuminating the process of embryonic patterning and development (Fig. 29.3) (Phoon 2006). UBM allowed detection and longitudinal monitoring of fetal development of the retina, lens and cornea in the mouse eye (Foster et al. 2003). The rate of

428

K. Vandoorne et al.

Fig. 29.7 Manganese-enhanced MRI of the mouse embryonic CNS. Developmental stage-dependent changes were detected by Mn-enhancement (Mn dose 80 mg/kg). (a–c) Sagittal images of E12.5, E14.5, and E17.5 embryos showed increasing forebrain enhancement with age (scale bar; 0.5 mm). The spinal cord (SC) and ventral brain enhanced at all time points (d–f), with progressive enhancement from posterior to anterior and ventral to dorsal

with increasing age (scale bar is 1 mm). At E17.5, the boundaries of brain subregions including the cortex (aqua), thalamus (magenta), BG (green), and pyriform cortex (red) could be appreciated (g), enabling semiautomated segmentation and 3D analysis of these structures as well as the olfactory bulbs (orange) and SE (blue) (h). (Reprinted with permission of John Wiley & Sons, Inc. from Deans et al. 2008)

growth of the major axis was linear throughout development starting at E8.5 when the eye was close to the system spatial resolution limit. At E11–12, blood vessels were still not detectable by ultrasound.

depict specific anatomical features beyond measurements of fetal and organ size. Postnatal development was studied by DTI and MEMRI for maximizing contrast in CNS in mice from the earliest postnatal stages. DTI has not been established for in utero studies due to motion artifacts. Recently, in utero MEMRI was used in studies of the embryonic CNS in combination with respiratory gating to decrease motion artifacts and allowed MEMRI-facilitated CNS segmentation and 3D analysis in mid- and late-stage embryos. Moreover, quantitative MEMRI analysis of Nkx2.1

29.3.1.2 Following Anatomy by Magnetic Resonance Imaging MRI studies of fetal development exploited the exquisite endogenous soft tissue contrast mechanisms to

429

29 Imaging in Developmental Biology

knockout mice demonstrated volumetric changes in septum and basal ganglia as well as alterations in hypothalamic structures (Fig. 29.7) (Turnbull and Mori 2007; Deans et al. 2008). Recently, a method for in utero cardiac imaging has been described using self-gating data and image coregistration to permit reconstruction of images in the presence of both periodic physiological motion and nonperiodic rigid-body motion, including gastrointestinal peristalsis, uterine contractions, shifts of the head or torso, and embryonic movement in utero (Fig. 29.5). Self-gating is achieved by acquisition of navigator echoes representative of tissue motion along with image data. For 3D imaging of the embryonic heart, detection and correction of bulk embryo motion generated images of acceptable quality. The limit of displacement was approximately 2 mm. Reconstruction of images was performed at six different phases of the heart cycle. Longaxis and short-axis slices at systole and diastole were reconstructed, along with 3D renderings of the heart chambers (Fig. 29.5), (Nieman et al. 2009).

29.3.1.3 Following Anatomy by Computerized Tomography One of the ways to assess the fetus well-being in utero consists of measurements of specific skeletal parameters such as limb length, crown-to-rump length, and cranial size. These measurements are often conducted using ex vivo micro-CT, providing images of radioopaque structures with a voxel resolution of up to 10 mm (Oest et al. 2008). Micro-CT replaced the more conventional histological method, previously used to assess skeletal development and growth. As compared with histology, micro-CT enables fast 3D imaging and the reconstruction of the entire bone. Using micro-CT, parameters such as material density, bone volume fraction, and trabecular architecture (including trabecular thickness, number, separation, connectivity, and orientation) can be retrieved and analyzed. The endogenous contrast of bones can be used for anatomical in vivo imaging of fetal skeletal development in pregnant rats (Fig. 29.3) (Winkelmann and Wise 2009). This methodology is particularly valuable during the last stages of pregnancy when bone mineralization is a critical determinant of fetal size. The inherently low soft tissue contrast and the lack of transfer of contrast media across the placenta limit the use of CT for soft tissue fetal analysis. When considering longitudinal monitoring by CT, the

cumulative effects of the administered radiation dose should be considered.

29.3.2 Functional Parametric Mapping While anatomical information can be derived from noninvasive imaging, it can also be acquired ex vivo. However, one of the key strengths of in vivo imaging is the access to functional physiological information that is lost when analyzing fixed specimens.

29.3.2.1 Functional Imaging of Development by UBM Fetal blood velocity can be detected through the Doppler frequency shift in the signal reflected from the moving red blood cells and is routinely used to assess developmental cardiovascular physiology. UBM appears wellsuited to study embryonic and postnatal developmental cardiovascular physiology (Figs. 29.3 and 29.4). It can be used on embryos as soon as the heart starts to beat on E8.5. From this moment onwards, it is possible to use Doppler to measure heart rate and to detect heart failure. Blood flow velocity in the vitelline circulation to the yolk sac is also first detected at this stage. By E9.5, the umbilical circulation has formed, and Doppler arterial waveforms can be obtained from the umbilical artery and from the inflow or outflow tracts of the heart. By E13.5, when embryos are in a suitable orientation to obtain a four-chamber view of both ventricles and both atria, the pulsed Doppler sample volume can be placed within the left or right ventricular chamber to record the mitral or tricupsid inflow, or within the ascending aorta to record the outflow blood velocity waveforms. Doppler waveforms can also be obtained at other sites such as the ascending and descending aorta, the pulmonary artery, the umbilical and vitelline arteries, the umbilical vein, ductus venosus, and inferior vena cava (Kulandavelu et al. 2006; Phoon 2006). To obtain information on the motion and dimensions of the cardiac chambers over time, the M-mode cursor can be placed over the heart. By E13.5, when embryos are in a suitable orientation to obtain a longaxis view of the left and right ventricles, it is possible to obtain M-mode recordings of the motion of the heart walls, in addition to the intraventricular waveforms. The UBM (both Doppler and M-mode) gives similar information in newborn mice. Because the operator

430

has more freedom to position and change the transducer or body angle to obtain the best imaging plane, imaging of postnatal animals might be easier. UBM can characterize the pathophysiology of embryonic cardiovascular system and has been applied to define in utero heart and vascular malformations.

K. Vandoorne et al.

ability to manipulate and deliver materials to embryos while MRI is used for monitoring reporter gene expression during cell differentiation and for tracking cell migration.

29.3.3.1 Molecular and Cellular Imaging of Development Using UBM 29.3.2.2 Functional Imaging of Development by MRI Functional imaging by MRI was used to identify fetal implantation sites and to analyze early vascular changes during implantation (Plaks et al. 2006). This approach was used for the study of implantation failure. CE MRI using high molecular Gadolinium allowed high-resolution detection and quantitative assessment of mouse embryo implantation sites as early as E4.5, and subsequent vascular expansion at E5.5. Vessel permeability was elevated in E4.5 implantation sites, relative to nonimplanted uterus, followed by increased blood volume in implantation sites between E4.5 and E5.5 (Fig. 29.6) associated with ablation of dendritic cells from the implantation site (Plaks et al. 2008). Blood oxygenation provides another important endogenous contrast mechanism that can be detected by MRI. While oxygenated hemoglobin is diamagnetic, deoxyhemoglobin is paramagnetic, resulting in reduced signal intensity in R2*-weighted images. This blood oxygenation level dependent (BOLD) contrast MRI was applied for monitoring changes in blood oxygenation in rat embryos (Girsh et al. 2007). Cloprostenol, a prostaglandin F2a analog, reduced fetal blood oxygenation, resulting in increased deoxyhemoglobin concentration and decreased signal in R2*-weighted MRI images. The changes in BOLD contrast correlated with placental expression of vascular endothelial growth factor (VEGF) (Girsh et al. 2007).

29.3.3 Molecular and Cellular Mapping of Development Over the last years, a significant progress in imaging technologies led to the possibility to map, beyond anatomical and functional information, also molecular and cellular events in live animals. A number of these technologies were also utilized for imaging fetal development. In particular, UBM provides the

UBM can facilitate the manipulation of the mouse embryo especially at the early stages of pregnancy, to help gain insight into specific developmental processes. Cells, viruses, or other agents can be injected under UBM guidance, into precise locations in the developing embryo at various developmental stages. Following microinjections into specific sites within the embryonic mouse brain, for example, investigators have studied progenitor engraftment, ectopic gene expression with gain-of-function, cell lineages, and tumor induction. Nevertheless, these intra-embryonic injections are perturbing even for the normal embryo and carry a substantial mortality rate of approximately 50% (Phoon 2006).

29.3.3.2 Molecular and Cellular Imaging of Development Using MRI In vivo cell tracking and imaging of gene expression can supply critical new insights into the dynamics of developmental processes. In vivo imaging of gene expression and cell tracking after magnetic labeling with gadolinium have been carried out in studies of frog embryo development using single-cell microinjection and micro-MRI (Louie et al. 2000). Such cell tracking has not been done yet in vivo for fetal development of mice. However, ex vivo MRI of fixed mouse embryos was used recently to detect micron-sized iron oxide particles, internalized in cells of mouse blastocyst-stage embryos, more than a week later in E11.5 embryos (Shapiro et al. 2004; Turnbull and Mori 2007). MRI contrast can be generated by overexpression of ferritin, the iron storage protein, which was put forward as a reporter for the detection of gene expression by MRI. Tetracycline-regulated overexpression of ferritin resulted in specific alterations of the transverse relaxation rate (R2) of water. Transgenedependent changes in R2 were detectable by MRI in fetal developmental induction of transgene expression in utero. Specific endothelial cell differentiation

431

29 Imaging in Developmental Biology

An important role for imaging lies in analysis of the consequence of intervention on the developing embryo.

In particular, longitudinal multiparametric imaging is critical for those interventions resulting in variable response such that the dynamic changes cannot be inferred from comparison of histological specimens taken from different embryos. Multiple studies evaluated the use of imaging as part of the phenotyping efforts on genetically modified mice (Dickinson 2006; Kulandavelu et al. 2006; Phoon 2006; Turnbull and Mori 2007). A specific type of intervention providing important insight is the selective ablation of cells. Such manipulation can be achieved by genetic introduction of suicide genes or by selective introduction of specific sensitivity to toxins. Using this approach, dynamic CE MRI revealed that ablation of dendritic cells resulted in failure of fetal implantation, due to impaired decidual angiogenesis (Plaks et al. 2008).

a

c

was visualized using mice in which expression of ferritin was regulated by the promoter of vascular endothelial cadherin (VE-cadherin). Thus, the tethfer MRI reporter mice provide a new transgenic mouse platform for in vivo molecular imaging of reporter gene expression by MRI during embryonic life (Fig. 29.8) (Cohen et al. 2007).

29.3.4 Specific Applications 29.3.4.1 Detection of Interventions

d 16

b

14 20

15

f

g

Fig. 29.8 In utero MRI detection of VE-cadherin mediated expression in E13.5 fetal liver and heart. (a) MRI of a pregnant homozygous tet-hfer transgenic dam mated with a heterozygous VE-cadherin-tTA male (8 embryos at E13.5). (b) Embryos were genotyped as dTG (VE-cadherin-hfer; green) or sTG (tet-hfer; white). (c) R2 relaxation rate was significantly elevated in the liver and heart, but not brain, of dTG (green) embryos. (d) R2 difference values (DR2 = (R2 in liver + heart) − R2 in brain) differentiated between the sTG (gray) and dTG (green) embryos.

12 10 8 6

10

e

∆R2 (s−1)

R2 liver + heart (s−1)

25

4

6 8 10 R2 brain (s−1)

4 VE-cadherin tTA

h

(e, f) Sagittal MRI and whole embryo (line, coronal slice, as in (g, h)). (g) No EGFP was detected in sTG (left) and high EGFP in the liver, with lower expression in the brain in dTG (right) embryos. (h) Coronal images (top) and R2 maps (bottom) of sTG (left) and dTG embryo (right) show elevated R2 in liver, heart, and posterior region of the dTG embryo. P placenta, H heart, L liver, B brain. (Reprinted by permission from Macmillan (Cohen et al. 2007))

432

29.3.4.2 Detection of Fluorescent and Bioluminescent Reporter Genes

K. Vandoorne et al.

Optical mapping was applied to study the patterning and organization of the conduction system and working myocardium in the developing avian and mammalian hearts. As it is impossible to apply this method in utero, the embryonic heart must be excised and bathed in a solution of voltage sensitive dyes. A light source then excites the cells and causes the emission of a fluorescent signal, proportional to the change in cell membrane voltage. Disadvantages of this method are the inability to retrieve absolute transmembrane potentials due to the absence of standard calibration, as well as its invasiveness, which prevents follow-up studies of the same embryos throughout pregnancy.

Another study examined the role of invading fetal precursor cells in maternal angiogenesis using BMI (Nguyen Huu et al. 2007). During pregnancy, fetal endothelial precursor cells invade the maternal circulation. Using VEGFR-2-luciferase males mated with wildtype females, it was possible to detect bioluminescence of endothelial cell originating from the embryo in sites of inflammation and in newly formed blood vessels. High-resolution fluorescence microscopy provides additional possibilities for imaging-based analysis of development. Recent studies described the generation of a novel transgenic mice line, in which the endothelial cell membranes were labeled with the red fluorescent protein mCherry, and the nuclei of the same embryonic endothelial cells were labeled with the yellow fluorescent protein EYFP (Larina et al. 2009). The combination of the two proteins enabled imaging of the entire vasculature as well as identification of specific endothelial cells within the blood vessels. Using confocal microscopy of E7.5–E12.5 cultured embryos, the authors were able to image ex vivo the vascular plexus in the yolk sac, the blood vessels in the embryo trunk, brain and eye, as well as the endocardium of the heart. Rapid image acquisition enabled imaging of the beating heart in cultured E8.5 embryos.

Both voltage sensitive dyes and MRI methods have been used to follow embryonic brain development (Mori et al. 2006). MRI allows a 3D characterization of the brain, consisting of different contrast mechanisms, usually based on the water relaxation parameters such as T1, T2, and T2*. Adjusting the acquisition parameters enables detection of different anatomical structures within the brain including the ventricles, as well as large white and gray matter structures. However, various white matter tracts cannot be detected using these conventional MR contrast mechanisms, and their analysis requires the application of DTI (Zhang et al. 2003).

29.3.4.3 Detection of Fluorescent Probes

29.3.4.5 Imaging Skeletogenesis

Detailed electrophysiological analysis of the heart development can be conducted ex vivo using optical probes for the cell membrane potential, namely voltage sensitive dyes (Efimov et al. 2004; Rothenberg et al. 2004). These dyes bind the plasma membrane of cardiac cells and exhibit changes in fluorescence as a result of transmembrane voltage alternations. Unlike traditional microelectrode recordings, optical signals derived from voltage sensitive dyes enable simultaneous detection of electrical activation derived from multiple sites within the embryonic heart.

Imaging of transgenic mice is important for understanding skeletogenesis and the genes involved in this process, as well as quantification of structure-function relationships in the bone. Vascularization of the bone growth plate is important for its adequate development during embryogenesis, as well as during bone repair and regeneration following injury in the adult. Imaging of this process reveals a precise spatial regulation of vascularity, based on vascular regression and invasion to the hypertrophic zone of the growth plate. Using contrast agents, 3D vascular networks within the bone can be imaged by mCT either in vivo or postmortem. Micro-CT further allows

29.3.4.4 Imaging Brain Development

29 Imaging in Developmental Biology

433

Fig. 29.9 In vivo CT imaging of late-stage pregnancy in mice (E18). 3D-rendered images of a pregnant mouse showing mineralized bone in late-stage fetal mice (white arrows) without contrast enhancement (a, c, d) and 15 min after i.v. injection of radio-opaque contrast (b). (a) Volume-rendered image showing the rostral part of a pregnant mouse visualizing the mineralized ribcage of the mother cranially and caudally a few fetuses (white

arrows). (b) Abdomen of a pregnant mouse 15 min after injection where the contrast is washed out of the blood into the kidneys (yellow arrows) and both urethras appearing radio-opaque. (c, d) Volume renderings of the caudal part a pregnant mouse picturing the fetuses (white arrows), the bladder (red arrow), and mineralized femora and pelvis of the mother (Courtesy of Vandoorne et al. unpublished)

imaging skeletogenesis of E17–E19 mouse embryos (Guldberg et al. 2004; Oest et al. 2008). The overall increase in fetal size as well as the increase in bone volume and the changes in quantity of mineralized tissue throughout the development was detected. Measurements of specific bone lengths were found to be well correlated

with measurements taken using the standard clear-stained technique, suggesting mCT as an appropriate technique for skeletogenesis analysis. Although most studies were done ex vivo, fetal bone mineralization can also be detected in utero in both rats and mice (Figs. 29.2 and 29.9) (Winkelmann and Wise 2009).

434

29.4 Animal Preparation Pregnancy and fetal development entail specific sensitivity to many of the procedures that are routinely included in in vivo imaging studies.

29.4.1 Anesthesia Procedures Anesthetic agents are frequently used to immobilize mice to reduce motion artifacts during image acquisition or other procedures. For imaging procedures under 30–45 min, injectable anesthetics are successfully used. Most commonly, a combination of Ketamine with a tranquilizer such as Xylazine is administered i.p., although i.m. or i.v. administrations are also possible. Ketamine increases cardiac output, heart rate, and blood pressure in normal patients. Xylazine on the other hand causes an initial increase in total peripheral resistance of the cardiovascular system, with increased blood pressure followed by a longer period of hypotension. Cardiac output may be reduced, but respiration is unaffected by Xylazine. Because this adrenergica2 agonist may increase uterine tone, Xylazine should be used with extreme caution, during pregnancy. Xylazine depresses thermoregulatory mechanisms, which might be a cause of considerable mortality. Diazepam is also commonly mixed with Ketamine but has been implicated as a teratogen in humans and thus less appropriate in pregnant mice. Barbiturate anesthetics, such as pentobarbital, have a widely variable response to anesthetic dose and a narrow therapeutic index, which lead to considerable anesthetic mortality with i.p. injections in mice. It is much safer to inject i.v. (Lukasik and Gillies 2003). For newborns, injectable anesthetics proved unsafe, whereas hypothermia is used as an effective and safe longterm method for reducing motion (Danneman and Mandrell 1997). For procedures lasting longer than 30–45 min, inhalant anesthetics are preferred to injectable drugs. Inhaled anesthesia results in reduced risk for intrauterine injury that can occur with intraperitoneal administration (Kulandavelu et al. 2006). In newborns, inhalation anesthetics are reported to be safer than injectable anesthetics (Kulandavelu et al. 2006). The most commonly used

K. Vandoorne et al.

inhalant anesthetic today is Isoflurane. All of the inhalant anesthetics cause myocardial and respiratory depression to some extent but undergo very little hepatic metabolism. Elimination is via the lung. Consequently, recovery is usually rapid after discontinuing inhalant administration and takes several minutes regardless of the length of anesthesia as long as the animals are supported while anesthetized (e.g. temperature). Add itionally, Isoflurane provides a stable depth of anesthesia during the entire length of the anesthetic episode, in contrast with injectable anesthesia where depth is continuously decreasing as drug metabolism takes place (Lukasik and Gillies 2003). Pregnant rodents are typically induced with 5% isoflurane in oxygen and maintained with 1–2% Isoflurane in oxygen. Isoflurane is relatively safe when used at appropriate concentrations. Single or repeated exposure to anesthesia may have long-term effects on embryonic and newborn mice, but these effects have not been studied in detail. It has been reported that chronic exposure to light Isoflurane anesthesia (4 h/day daily from em bryonic days 6–15) reduced embryonic growth and increased the incidence of cleft palate; however, the effect of repeated exposure to deeper levels of anesthesia is largely unexplored. Furthermore, Isoflurane was not found to reduce pregnancy rates and fetal survival significantly, so that it is unlikely that Isoflurane induces severe cardiac malformations. However, there may be an increased incidence of anomalies and mild growth retardation, and exposed preimplantation embryos exhibit arrested development (Kulandavelu et al. 2006).

29.4.2 Animal Preparation and Physiological Monitoring Sedated and anesthetized patients will not close their eyelids completely, leading to drying of the corneas, corneal ulcer, and blindness. Therefore, it is advised to protect the animal’s eyes with a sterile eye lubricating ointment prior to the imaging session. All animals are susceptible to hypothermia while under anesthesia, and this can be a source of considerable mortality. Maternal body temperature should be monitored using a rectal thermometer and maintained between 36 and 38°C by wrapping the animal or its extremities in an insulator, like a plastic wrap, or

435

29 Imaging in Developmental Biology

using an external heat source such as heating pad and/ or infrared lamp. The latter should be used with caution due to increased risk of thermal injury. Body temperature in newborns can be monitored using a rectal probe when older than 14 days of age, since insertion of probes in younger mice is unsafe. Ventilation and oxygenation are critically important for maintaining normal maternal and embryonic cardiovascular function, and thus, monitoring the depth of the anesthesia by electrocardiography (ECG) and respiration is vital. Even in newborns, ECG monitoring have been reported (Lukasik and Gillies 2003; Kulandavelu et al. 2006). Motion such as spontaneous uterine contractions, respiratory motion of the mother, or motion of later stage embryos can be minimized pharmacologically or taken into account through gating or data-processing (Turnbull and Mori 2007; Deans et al. 2008).

29.4.3 Fetal Identification and Genotyping A requirement of longitudinal studies is the ability to identify individual animals and their contributions to the study data. This is particularly difficult for longitudinal studies at early stages of pregnancy, due to the extensive motion of the uterus at this stage. Newborns can be labeled on a short-term basis using permanent marker. The label must be refreshed daily. Tattooing on the bottom of the paw provides a longer lasting label. Tattooing lasts for about a month and at this stage, mice are weaned, and ear clipping can be used for subsequent identification. Ideally, embryos studied in utero should be “tattooed,” so they can be identified after birth. Interlitter variance implies that it is important to compare embryos within and between pregnancies. As a rule, it was suggested that 1–4 embryos should be evaluated from at least 3 L (Kulandavelu et al. 2006). Some of the problems of identification and restraint can be resolved through the use of cultured embryos (Phoon 2006). Isolated mouse embryos remain viable and show normal growth in culture for 1–2 days despite the disruption of the placental circulation.

29.5 Summary Noninvasive in utero fetal imaging of small laboratory animals offers a growing portfolio of parameters that can aid in developmental biology. In particular, imaging is valuable for physiological, functional, and molecular mapping of fetal development of genetically modified mice. Exciting insight is obtained by imaging perturbations of development in transgenic and knockout mice and mice carrying specific reporter genes allowing following migration and differentiation of cells during organ formation. It is clear, however, that the current limitations for contrast, temporal, and spatial resolution restrict the range of questions that can be addressed. Thus, improved technologies would be welcome for meeting the challenges of functional mapping of the genome using large-scale efforts for genetic modifications in mice. Acknowledgments The authors acknowledge the support by the Israel Science Foundation ISF 93/07 and ISF Converging Technologies Equipment Award, and by the European Commission 7th framework ERC advanced project IMAGO 232640. MN is incumbent of the Helen and Morris Mauerberger Chair.

References Bang DW, Lee JH, Oh H, Kim SR, Kim TH, Lee YS, Lee CS, Kim SH (2002) Dose-incidence relationships on the prenatal effects of gamma-radiation in mice. J Vet Sci 3:7–11 Brown AS, Reid AD, Leamen L, Cucevic V, Foster FS (2004) Biological effects of high-frequency ultrasound exposure during mouse organogenesis. Ultrasound Med Biol 30: 1223–1232 Cohen B, Ziv K, Plaks V, Israely T, Kalchenko V, Harmelin A, Benjamin LE, Neeman M (2007) MRI detection of transcriptional regulation of gene expression in transgenic mice. Nat Med 13:498–503 Danneman PJ, Mandrell TD (1997) Evaluation of five agents/ methods for anesthesia of neonatal rats. Lab Anim Sci 47:386–395 Deans AE, Wadghiri YZ, Berrios-Otero CA, Turnbull DH (2008) Mn enhancement and respiratory gating for in utero MRI of the embryonic mouse central nervous system. Magn Reson Med 59:1320–1328 Dickinson ME (2006) Multimodal imaging of mouse development: tools for the postgenomic era. Dev Dyn 235: 2386–2400 Efimov IR, Nikolski VP, Salama G (2004) Optical imaging of the heart. Circ Res 95:21–33

436 Ezin M, Fraser S (2008) Time-lapse imaging of the early avian embryo. Methods Cell Biol 87:211–236 Foster FS, Zhang M, Duckett AS, Cucevic V, Pavlin CJ (2003) In vivo imaging of embryonic development in the mouse eye by ultrasound biomicroscopy. Invest Ophthalmol Vis Sci 44:2361–2366 Girsh E, Plaks V, Gilad AA, Nevo N, Schechtman E, Neeman M, Dekel N (2007) Cloprostenol, a prostaglandin F(2alpha) analog, induces hypoxia in rat placenta: BOLD contrast MRI. NMR Biomed 20:28–39 Guldberg RE, Lin AS, Coleman R, Robertson G, Duvall C (2004) Microcomputed tomography imaging of skeletal development and growth. Birth Defects Res C Embryo Today 72:250–259 Hogers B, van der Weerd L, Olofsen H, van der Graaf LM, Deruiter MC, Groot AC, Poelmann RE (2009) Non-invasive tracking of avian development in vivo by MRI. NMR Biomed 22:365–373 Ingham PW (2009) The power of the zebrafish for disease analysis. Hum Mol Genet 18:R107–R112 Kulandavelu S, Qu D, Sunn N, Mu J, Rennie MY, Whiteley KJ, Walls JR, Bock NA, Sun JC, Covelli A, Sled JG, Adamson SL (2006) Embryonic and neonatal phenotyping of genetically engineered mice. ILAR J 47:103–117 Kulesa PM (2004) Developmental imaging: insights into the avian embryo. Birth Defects Res C Embryo Today 72: 260–266 Larina IV, Shen W, Kelly OG, Hadjantonakis AK, Baron MH, Dickinson ME (2009) A membrane associated mCherry fluorescent reporter line for studying vascular remodeling and cardiac function during murine embryonic development. Anat Rec (Hoboken) 292:333–341 Le Floc’h J, Cherin E, Zhang MY, Akirav C, Adamson SL, Vray D, Foster FS (2004) Developmental changes in integrated ultrasound backscatter from embryonic blood in vivo in mice at high US frequency. Ultrasound Med Biol 30: 1307–1319 Louie AY, Huber MM, Ahrens ET, Rothbacher U, Moats R, Jacobs RE, Fraser SE, Meade TJ (2000) In vivo visualization of gene expression using magnetic resonance imaging. Nat Biotechnol 18:321–325 Lukasik VM, Gillies RJ (2003) Animal anaesthesia for in vivo magnetic resonance. NMR Biomed 16:459–467 Mori S, Zhang J, Bulte JW (2006) Magnetic resonance microscopy of mouse brain development. Methods Mol Med 124: 129–147 Nguyen Huu S, Oster M, Uzan S, Chareyre F, Aractingi S, Khosrotehrani K (2007) Maternal neoangiogenesis during pregnancy partly derives from fetal endothelial progenitor cells. Proc Natl Acad Sci USA 104:1871–1876 Nieman BJ, Szulc KU, Turnbull DH (2009) Three-dimensional, in vivo MRI with self-gating and image coregistration in the mouse. Magn Reson Med 61:1148–1157 Oest ME, Jones JC, Hatfield C, Prater MR (2008) Micro-CT evaluation of murine fetal skeletal development yields greater morphometric precision over traditional clear-staining methods. Birth Defects Res 83:582–589 Passamaneck YJ, Di Gregorio A, Papaioannou VE, Hadjantonakis AK (2006) Live imaging of fluorescent proteins in chordate embryos: from ascidians to mice. Microsc Res Tech 69: 160–167 Phoon CK (2006) Imaging tools for the developmental biologist: ultrasound biomicroscopy of mouse embryonic development. Pediatric Res 60:14–21

K. Vandoorne et al. Plaks V, Kalchenko V, Dekel N, Neeman M (2006) MRI analysis of angiogenesis during mouse embryo implantation. Magn Reson Med 55:1013–1022 Plaks V, Birnberg T, Berkutzki T, Sela S, BenYashar A, Kalchenko V, Mor G, Keshet E, Dekel N, Neeman M, Jung S (2008) Uterine DCs are crucial for decidua formation during embryo implantation in mice. J Clin Invest 118: 3954–3965 Rinon A, Lazar S, Marshall H, Buchmann-Moller S, Neufeld A, Elhanany-Tamir H, Taketo MM, Sommer L, Krumlauf R, Tzahor E (2007) Cranial neural crest cells regulate head muscle patterning and differentiation during vertebrate embryogenesis. Development 134:3065–3075 Rothenberg F, Efimov IR, Watanabe M (2004) Functional imaging of the embryonic pacemaking and cardiac conduction system over the past 150 years: technologies to overcome the challenges. Anat Rec A Discov Mol Cell Evol Biol 280:980–989 Salomon LJ, Siauve N, Balvay D, Cuenod CA, Vayssettes C, Luciani A, Frija G, Ville Y, Clement O (2005) Placental perfusion MR imaging with contrast agents in a mouse model. Radiology 235:73–80 Salomon LJ, Siauve N, Taillieu F, Balvay D, Vayssettes C, Frija G, Ville Y, Cuenod CA, Clement O (2006) In vivo dynamic MRI measurement of the noradrenaline-induced reduction in placental blood flow in mice. Placenta 27: 1007–1013 Schneider JE, Bhattacharya S (2004) Making the mouse embryo transparent: identifying developmental malformations using magnetic resonance imaging. Birth Defects Res C Embryo Today 72:241–249 Shapiro EM, Skrtic S, Sharer K, Hill JM, Dunbar CE, Koretsky AP (2004) MRI detection of single particles for cellular imaging. Proc Natl Acad Sci USA 101:10901–10906 Taillieu F, Salomon LJ, Siauve N, Clement O, Faye N, Balvay D, Vayssettes C, Frija G, Ville Y, Cuenod CA (2006) Placental perfusion and permeability: simultaneous assessment with dual-echo contrast-enhanced MR imaging in mice. Radiology 241:737–745 Tanaka M, Hadjantonakis AK, Vintersten K, Nagy A (2009) Aggregation chimeras: combining ES cells, diploid, and tetraploid embryos. Methods Mol Biol 530:287–309 Tirosh-Finkel L, Elhanany H, Rinon A, Tzahor E (2006) Mesoderm progenitor cells of common origin contribute to the head musculature and the cardiac outflow tract. Development 133:1943–1953 Turnbull DH, Mori S (2007) MRI in mouse developmental biology. NMR Biomed 20:265–274 Vogt A, Cholewinski A, Shen X, Nelson SG, Lazo JS, Tsang M, Hukriede NA (2009) Automated image-based phenotypic analysis in zebrafish embryos. Dev Dyn 238:656–663 Winkelmann CT, Wise LD (2009) High-throughput micro-computed tomography imaging as a method to evaluate rat and rabbit fetal skeletal abnormalities for developmental toxicity studies. J Pharmacol Toxicol Methods 59:156–165 Yaniv K, Isogai S, Castranova D, Dye L, Hitomi J, Weinstein BM (2006) Live imaging of lymphatic development in the zebrafish. Nature Med 12:711–716 Zhang J, Richards LJ, Yarowsky P, Huang H, van Zijl PC, Mori S (2003) Three-dimensional anatomical characterization of the developing mouse brain by diffusion tensor microimaging. NeuroImage 20:1639–1648

Imaging in Gynecology Research

30

Matthias W. Laschke and Michael D. Menger

Contents 30.1 Introduction............................................................. 437 30.2 Physiology of the Ovary and Ovarian Tissue Transplantation........................................... 438 30.3 Placental Pathology................................................ 441 30.4 Gynecologic Oncology............................................ 442 30.5 Endometriosis.......................................................... 442 30.6 Outlook.................................................................... 445 30.7 Lab Protocol............................................................ 445 References............................................................................ 446

M.W. Laschke (*) and M.D. Menger Institute for Clinical & Experimental Surgery, University of Saarland, 66421 Homburg, Germany e-mail: [email protected]

30.1 Introduction Rodents are widely used as model animals in gynecology research for several reasons. Their breeding and maintenance are uncomplicated and cost-efficient. They exhibit a short and regular estrous cycle, including the stages of proestrus, estrus, metestrus, and diestrus, which can reliably be determined by cytological analysis of vaginal lavage samples. Accordingly, the selection of animals in the same cycle stage allows for the performance of standardized studies, excluding discrepancies between individual animals due to different sex hormone levels. Alternatively, rodents can easily be ovariectomized and substituted with defined doses of estrogen and progesterone to induce constant hormonal conditions. Moreover, they exhibit a short life cycle, which reduces the time of puberty, pregnancy, and delivery, and thus, facilitates research focusing on reproductive function and dysfunction. Mainly in mice, there is the possibility of genetic manipulations for the investigation of gene functions and the establishment of novel gynecological disease models. The use of immunocompromised nude or SCID mice further bears the advantage of studying pathophysiological processes in xenografts, such as endometriotic lesions or tumor tissue of human origin. In the past, gynecology research in rodents was mainly based on histological, immunohistochemical, or gene expression analyses of tissue samples, which were isolated after surgery or necropsy. Accordingly, this invasive approach provided data based on single observation time points, not allowing sequential dynamic studies performed in the same animals. However, such studies are especially important for the examination of the female reproductive organs, which are continuously

F. Kiessling and B.J. Pichler (eds.), Small Animal Imaging, DOI: 10.1007/978-3-642-12945-2_30, © Springer-Verlag Berlin Heidelberg 2011

437

438

changing during the estrous cycle. Moreover, to assess the increasing amounts of novel, targeted therapies for the treatment of severe gynecological diseases, such as gynecologic cancer or endometriosis, it is desirable to measure changes in tissue proliferation and regression over time in clinically relevant rodent models. Therefore, sophisticated small animal imaging technologies have been introduced in the field of gynecology research during the last years, which allow for the noninvasive and repetitive analysis of physiological and pathological processes within the female reproductive organs. These include the technique of intravital fluorescence microscopy, bioluminescence, ultrasound biomicroscopy, computed tomography (CT), and magnetic re sonance imaging (MRI). All of these technologies contribute to limit the statistical variability of preclinical studies, to reduce the number of animals required for each study, and to maximize the amount of data obtained from each animal according to the philosophy of the 3Rs model (refinement, reduction, and replacement) of Russel and Burch (Flecknell 2002). This chapter highlights some of the most interesting and relevant small animal models and imaging technologies, applied in central gynecology research fields, including physiology of the ovary and ovarian tissue transplantation, placental pathology, gynecologic oncology, as well as endometriosis.

30.2 Physiology of the Ovary and Ovarian Tissue Transplantation The ovary is beside the uterus one of the dynamic organs in the adult, which exhibits periodic growth and regression. In fact, the ovarian cycle is characterized by the growth of individual ovarian follicles, which continually leave the large pool of resting, primordial follicles. During their development, they produce increasing amounts of estrogens and finally approach ovulation. Subsequently, the remaining cells of ruptured follicles form the corpus luteum, which is critical for successful maintenance of pregnancy in mammals, because it is the primary source of progesterone. All of these steps are tightly regulated by the hypothalamic-pituitary axis via the release of the gonadotropins follicle stimulating hormone (FSH) and luteinizing hormone (LH). Accordingly, folliculogenesis

M.W. Laschke and M.D. Menger

and corpus luteum formation are essential for the regulation of physiological reproductive function and hormonal state. Recently, the technique of high-resolution ultrasound imaging has been introduced for the noninvasive imaging of follicles and corpora lutea in rodents. By means of ultra-high frequencies of ~40 MHz, this technique allows the generation of images with a resolution of ~40 mm, which is near to conventional microscopy. Therefore, it is also called ultrasound biomicroscopy. Using this imaging tool, it is possible to easily localize the ovaries in both rats and mice, to assess their volume and to accurately distinguish between nonproliferating and preovulatory follicles as well as corpora haemorrhagica and corpora lutea due to their different echogenic patterns (Pallares and Gonzalez-Bulnes 2008) (Fig. 30.1). Importantly, the ultrasound waves do not induce tissue damage and genetic alterations, such as X-rays, and thus can be applied without the risk of negative effects on the reproductive process. A major prerequisite for cyclic folliculogenesis and corpus luteum formation is angiogenesis, which is defined as the formation of new capillaries from preexisting blood vessels. During the ovarian cycle, vascular changes are tightly regulated by the expression of numerous pro- and anti-angiogenic growth factors in which angiogenesis is turned on for brief periods and then completely inhibited. Correspondingly, the ovary represents a unique system for the study of physiological angiogenesis and blood vessel regression in the adult, where angiogenesis is normally restricted to pathological processes, such as tumor growth, wound healing, or inflammation. Moreover, some types of infertility may directly be caused by disturbed follicular angiogenesis. For instance, the polycystic ovarian syndrome is associated with excessive development of new blood vessels while the ovarian hyperstimulation syndrome is characterized by an increase in capillary permeability. Thus, the analysis of ovarian angiogenesis and its regulation is also of major importance for the establishment of new therapies in gynecology, which control inappropriate follicle development due to enhanced or decreased angiogenesis. The rapid changes in blood vessel development and regression within the ovary are accompanied by equally rapid changes in rates of blood perfusion. Measurement of ovarian blood flow by means of highresolution color-coded Doppler ultrasound imaging

439

30 Imaging in Gynecology Research

a

b

c

Fig. 30.1 High-resolution ultrasound imaging of a rat ovary (a, borders marked by arrows) by means of a Vevo 770 highresolution imaging system (VisualSonics) and a 704 Scanhead with a center frequency of 40 MHz and a focal depth of 6 mm. (b, c) Higher magnification shows typical stages of folliculogen-

esis, such as premature ovarian follicles (b, arrows) and mature, preovulatory follicles (c, border marked by arrows), which are characterized by a large anechoic antral cavity and an echogenic cumulus oocyte complex (c, arrowhead). Scale bars: A = 1.2 mm; B, C = 350 mm

may therefore be useful for studying angiogenesis and vascular remodeling associated with the ovarian cycle in rodents. Alternatively, pulsed arterial spin labeling MRI has been suggested for the analysis of ovarian tissue perfusion in small animal studies (Tempel-Brami and Neeman 2002). In this type of measurement, the MRI signal is tagged in the upstream of the ovary by pulsed slice selective saturation. By this, perfusion within the ovary can be calculated by monitoring the saturation transfer due to perfusion without the need for exogenous administration of contrast agents. Interestingly, pulsed arterial spin labeling MRI of the rat ovary showed that preovulatory follicles exhibit poorly perfused regions, indicating that physiological hypoxia may play a crucial role in the ovulation process (Tempel-Brami and Neeman 2002). However, the MRI technology may not only be used to assess perfusion rates. In fact, dynamic contrast-enhanced MRI additionally enables the noninvasive analysis of vascular density, vessel permeability, and interstitial convection in ovarian tissue (Israely et al. 2004). For the detailed in vivo analysis of the morphology and perfusion of individual microvessels within ovarian tissue, follicles and corpora lutea can be

transplanted into the dorsal skinfold chamber (Vollmar et al. 2001) (Fig. 30.2). This well-established rodent model allows for the repetitive analysis of microcirculatory processes in mice, hamsters, and rats by means of intravital multi-fluorescence microscopy over a time period of 3–4 weeks (Menger et al. 2002). For the implantation of the chamber, the back of the anesthetized animal is shaven and depilated before two symmetrical titanium frames are implanted on the extended dorsal skinfold, so that they sandwich the double layer of skin. One layer of skin is completely removed in a circular area of ~15 mm in diameter, and the remaining layers consisting of striated skin muscle, subcutaneous tissue, and skin are covered with a removable coverslip incorporated into the observation window of one of the titanium frames. The coverslip of the observation window can be temporarily removed to transplant ovarian follicles or corpora lutea onto the striated muscle within the chamber. After application of the fluorochrome fluorescein-isothiocyanate (FITC)-labeled dextran (MW 150,000) for contrast enhancement by staining of blood plasma, intravital fluorescence microscopy can be performed to visualize in vivo angiogenic

440

M.W. Laschke and M.D. Menger

a

b

c

d

e

f

Fig. 30.2 Dorsal skinfold chamber model for the noninvasive and repetitive in vivo imaging of physiological angiogenesis in ovarian follicles by means of intravital fluorescence microscopy. (a) Syrian golden hamster equipped with a dorsal skinfold chamber (chamber weight ~4 g). (b) Within the observation window of the chamber (arrow), all segments of the microcirculation including arterioles, capillaries, and venules of the striated skin muscle and subcutaneous tissue can be directly visualized. (c, d) For the analysis of ovarian follicles, the ovary of a donor hamster (c) is carefully microdissected using 27-gauge needles under a stereomicroscope. This handpicking procedure guarantees single connective tissue-free follicles (d), which subsequently can be transplanted onto the striated muscle within the

g

dorsal skinfold chamber. (e–g) Intravital fluorescence microscopic images of an ovarian follicle directly after transplantation (e) as well as at day 2 (f) and day 4 (g) after transplantation into the hamster dorsal skinfold chamber. Note the initial lack of nutritive capillaries within the freshly transplanted graft (e, asterisk). On day 2 after transplantation, parts of the follicular graft display newly formed microvessels (f, arrowheads), while some areas still lack vascularization (f, arrows). On day 4, a dense glomerulum-like network of newly formed microvessels covers the entire graft (g). Blue light epi-illumination with contrast enhancement by 5% FITC-labeled dextran 150,000 i.v. Scale bars: A = 2.2 cm; B = 9 mm; C, D = 1.1 mm; E–G = 200 mm

441

30 Imaging in Gynecology Research

sprouting and network formation within the ovarian tissue and to quantify in vivo morphological (vascularized area, functional capillary density) and microhemodynamic (vessel diameters, red blood cell velocity, volumetric blood flow) parameters of follicular and luteal microcirculation. Using additional fluorochromes, this microscopic in vivo approach also enables for the assessment of cellular and molecular aspects, such as leukocyte-endothelial cell interaction, platelet adhesion, or apoptotic cell death (Menger et al. 2002). By means of the dorsal skinfold chamber model of Syrian golden hamsters, it could be demonstrated that the theca interna is essential for the vascularization of follicles (Vollmar et al. 2001). Moreover, bilateral ovariectomy of the animals accelerates angiogenesis and improves vascular blood perfusion, which may be best interpreted that lack of steroid negative feedback and high gonadotropin secretion contribute to vascular growth and nutritive perfusion of ovarian follicles (Laschke et al. 2002). However, ovarian tissue transplantation not only is an experimental approach for the noninvasive in vivo analysis of physiological angiogenesis but also represents a clinical treatment strategy for preservation of ovarian function in women who suffer from premature ovarian failure due to aggressive chemotherapy or radiation therapy. For this purpose, small ovarian tissue fragments are isolated from patients and are cryopreserved until their transplantation back into the patients after successful cancer therapy. A major prerequisite for the restoration of the grafts’ hormonal and fertility function after cryopreservation is the development of a new vasculature. Therefore, the aforementioned imaging technologies have also been used to study angiogenesis and blood perfusion within cryopreserved ovarian grafts and to establish approaches that may accelerate and improve the angiogenic process. In this context, intravital fluorescence microscopic studies revealed that cryopreservation does not affect angiogenesis and revascularization of freely grafted ovarian follicles (Laschke et al. 2003). In addition, dynamic contrastenhanced MRI analyses of the vascular permeability of rat ovarian xenografts demonstrated that intramuscular implantation provides a better maintenance of functional blood vessels relative to subcutaneous implantation (Israely et al. 2003). This was associated with improved follicular preservation.

30.3 Placental Pathology During pregnancy, the placenta ensures nearly all exchanges of respiratory gases, nutrients, and metabolites between the fetal and maternal organism. Thereby, the rate of transplacental exchange is critically dependent on the maternal (uteroplacental) and the fetal (umbilicoplacental) circulation. Abnormalities in placental blood perfusion are associated with serious pregnancy disorders, such as maternal preeclampsia or fetal intrauterine growth restriction. To establish preventive strategies and effective therapies for these diseases, sophisticated small animal models are needed, which allow for the study of the mechanisms underlying placental pathology. For this purpose, mice are appropriate species, because their placenta is quite similar to the human placenta regarding cellular and circulatory architecture as well as gene expression. In addition, mutant mouse lines have already been identified, which exhibit different pathological placenta phenotypes (Rossant and Cross 2001). To adequately analyze the pathogenesis of human pregnancy disorders in these mice, noninvasive imaging methods have been introduced in gynecology research, including ultrasound biomicroscopy and dynamic contrast-enhanced MRI. High-resolution Doppler ultrasound imaging is an easy and reliable tool to quantify blood flow velocities in the uterine artery as well as in the uteroplacental and umbilicoplacental circulations throughout gestation in mice (Mu and Adamson 2006). Importantly, blood flow velocity waveforms in mice are similar in shape and show similar changes during gestation when compared to those of human pregnancy (Mu and Adamson 2006). This indicates that abnormal uterine hemodynamics, which are typically associated with severe preeclamptic and fetal intrauterine growth restriction pregnancies in humans, may be phenocopied in corresponding mutant mouse models. For this purpose, high-resolution Doppler ultrasound imaging represents an attractive technology, because it parallels the routine ultrasound imaging used in clinical practice. Due to its high resolution, it further allows for the visualization of the calcification of the placenta (Akirav et al. 2005). Finally, placental size of pregnant mice, which is also an early predictor for pregnancy outcome in humans, can be repetitively quantified in vivo (Mu et al. 2008).

442

Dynamic contrast-enhanced MRI offers the possibility to quantify the rate at which a contrast agent crosses from the intravascular compartment into other compartments. Accordingly, this technology ideally suits for the noninvasive determination of placental perfusion and permeability in mice (Taillieu et al. 2006). For this purpose, gadolinium chelates should be used as contrast agent, because it can easily cross the placental barrier due to its low molecular weight and small particle size. Moreover, high gadolinium chelate doses of 0.5 mmol/kg body weight are recommended in mice to increase fetal uptake and thereby facilitate permeability measurements. By means of dynamic contrast-enhanced MRI, it could be demonstrated that noradrenaline, which has been implicated in the pathogenesis of preeclampsia, reduces placental perfusion in mice (Salomon et al. 2006). These results indicate that dynamic contrast-enhanced MRI is an interesting technology for functional evaluation of the placenta in small animal models, which may help to gain new insights into the pathology of severe pregnancy disorders. In fact, this technology may even be a future diagnostic tool in clinical practice, although the routine use of contrast agents in humans during pregnancy is to date not recommended to prevent potential deleterious effects to the fetus.

30.4 Gynecologic Oncology Breast cancer and cancer of the female reproductive organs belong to the most common types of cancer in women. Approximately 82,000 women per year are diagnosed with gynecologic cancer only in the United States. Small animal models are important experimental tools to investigate mechanisms of tumor development and to evaluate the efficacy of novel therapeutic strategies. For this purpose, noninvasive approaches to analyze angiogenesis, metastasis, and tumor growth have also been introduced in the field of gynecologic oncology during the last years. For a long time, the standard technique for the study of different gynecologic tumor types was the subcutaneous implantation of human tumor cells or tissue fragments into immunodeficient mice. By this, the growth of the xenografts could be determined in vivo by external caliper measurement where the tumor volume is calculated by means of the modified ellipsoid

M.W. Laschke and M.D. Menger

formula 0.5 × length × width2. However, this method is often affected by errors, because of variability in tumor shape or skin thickness. Moreover, it is associated with a large interobserver variability. To overcome these problems, high-resolution ultrasound imaging, microCT, or MRI may alternatively be used for growth measurements of subcutaneous tumors in rodents. Nonetheless, one should be aware that such subcutaneous tumor models often do not reflect the real clinical situation, because they are usually based on the analysis of larger tumors. For instance, the majority of newly diagnosed breast cancers in women are very early tumor stages, such as preinvasive ductal carcinoma in situ (DCIS), which can now be detected in clinical practice due to improved mammography techniques. However, in the past, most imaging studies of mouse mammary cancer focused on large tumors that were extremely advanced and thus did not realistically reflect early breast cancer. In contrast, spontaneous developing rodent tumors progress through all disease stages and thus mimic their human counterpart much better. Interesting examples of such tumor models in gynecology research are the uterine leiomyoma model of the Eker rat (Walker et al. 2003) or the transgenic mouse models of mammary and ovarian cancer (Hensley et al. 2007; Jansen et al. 2008). Preclinical studies of spontaneous gynecologic tumors in rodents present unique challenges, including variable latency of tumor formation and detection of very small tumors, which are located deep within the female reproductive organs. Accordingly, numerous noninvasive imaging techniques have been proposed for the in vivo analysis of gynecologic tumors, such as ultrasound biomicroscopy, bioluminescence, MRI, microCT, and recently also flat-panel detector volume CT. Each technique bears many specific advantages and disadvantages relating to the analysis of tumors, whose detailed description would go beyond the scope of this section and can be found in other chapters of this book.

30.5 Endometriosis Endometriosis is a frequent gynecological disorder, which affects approximately 10–15% of all women in reproductive age. The disease is diagnosed by clinical evidence of proliferating endometrium-like tissue outside the uterine cavity. Classical symptoms of

443

30 Imaging in Gynecology Research

endometriosis are dysmenorrhoea, dyspareunia, dysuria, severe abdominal or pelvic pain, as well as infertility. Despite the high prevalence of endometriosis, many questions about its etiology and pathogenesis are still unanswered. However, the disease is most likely caused by retrograde menstruation, i.e., reflux of endometrial fragments in the fallopian tubes during menstruation and their implantation and development to endometriotic lesions in the peritoneal cavity. Because the growth of these lesions is estrogen-dependent, current medical therapies aim on the suppression of endogenous estrogen production, which may be associated with substantial side effects such as osteoporosis or episodes of depression. Alternatively, endometriotic lesions are surgically removed, however, with a high recurrence rate of up to ~50–70%. Accordingly, current endometriosis research focuses on the development of more effective treatment strategies. For this purpose, genetically welldefined rodent models, which allow for standardized preclinical studies, are an important approach. Because the development of spontaneous endometriosis is dependent on menstruation and thus restricted to humans and subhuman primates, endometriotic lesions in rodents are usually induced iatrogenically by grafting-isolated endometrium from the uterus to ectopic sites (Grümmer 2006). For this

a

purpose, the dorsal skinfold chamber model of Syrian golden hamsters can be used, which allows for the repetitive in vivo analysis of implantation, growth, and angiogenesis of the ectopic endometrial tissue using the combination of epi-illumination multi-fluorescence microscopy and computer-assisted off-line analysis techniques (see also Sect. 30.2) (Fig. 30.3). Because this experimental approach enables for the visualization of individual microvessels with a high resolution, it is in particular of interest for studies focusing on the analysis of blood vessel development and regression in endometriotic lesions as well as on the establishment of novel anti-angiogenic treatment strategies for en dometriosis therapy (Laschke and Menger 2007). However, a major disadvantage of the chamber model is the fact that it is restricted in use to a time period of only 3–4 weeks. In addition, the environment within the chamber may not exactly reflect the physiological profile of cytokines and growth factors of the peritoneal fluid and the intra-abdominal immune system. To fulfill these criteria, endometrium has to be transplanted directly into the peritoneal cavity. For the noninvasive monitoring of intraperitoneal endometriotic lesions in mice, the technique of bioluminescence has recently been introduced (Becker et al. 2006). For this purpose, luciferase-expressing transgenic mice are

b

*

Fig. 30.3 Typical microangioarchitecture and histomorphology of endometriotic lesions induced in the dorsal skinfold chamber model. (a) Intravital fluorescence microscopy of an endometrial graft (borders marked by arrows) at day 10 after transplantation into the dorsal skinfold chamber of a Syrian golden hamster. The graft exhibits a dense network of newly formed blood vessels with a glomerulum-like angioarchitecture. Blue light epi-illumination with contrast enhancement by 5% FITC-labeled dextran

*

*

150,000 i.v. (b) Hematoxylin-eosin stained cross section of an endometrial graft at day 14 after transplantation onto the striated muscle (arrows) of the dorsal skinfold chamber of a Syrian golden hamster, displaying typical histomorphological signs of an endometriotic lesion such as cyst-like dilated endometrial glands (asterisks) with an intact glandular epithelium surrounded by a richly vascularized endometrial stroma. Scale bars: A = 270 mm; B = 80 mm

444

M.W. Laschke and M.D. Menger

a

c

b

d

f

Fig. 30.4 (a, b) Surgical induction of intraperitoneal endometriosis in mice. For this purpose, a uterine tissue sample (a, arrow) is removed from a longitudinally opened uterine horn of a donor animal by means of a biopsy punch. The tissue sample is then sutured to the peritoneal wall of a recipient animal (b, arrow). (c–e) High-resolution ultrasound imaging of a developing endometriotic lesion (borders marked by white broken line) directly (c) as well as 14 and 28 days (e) after fixation of a uterine tissue sample to the abdominal wall. Note that the formation of an anechoic cyst (d, e) within the lesion can be visualized with this

e

g

noninvasive technique. Moreover, adhesion formation (e, arrows) around the lesion can easily be detected by injecting 1 mL saline solution (c–e, asterisks) into the peritoneal cavity, which separates the endometriotic lesion at the abdominal wall from the surrounding intra-abdominal organs. (f, g) Three-dimensional reconstruction of an endometriotic lesion (g, blue polygon) with two cysts (g, yellow polygon) based on manual image segmentation of two-dimensional ultrasound images of the lesion (f, borders marked by white broken line). Scale bars: A = 2.9 mm; B = 1.6 mm; C–E = 800 mm; F = 1 mm; G = 900 mm

445

30 Imaging in Gynecology Research

generated by inserting the human ubiquitin C promoter coupled to the firefly luciferase reporter. The transgenic endometrial tissue from these mice is subsequently transplanted into the peritoneal cavity of wild-type mice. By this, systemic injection of luciferin evokes a specific light signal within the endometrial grafts, which can be detected and quantified through the abdominal wall by a bioluminescence imaging system. Importantly, transabdominal luminescence correlates well with the grafts’ implantation, reflecting onset of angiogenesis and tissue growth. Accordingly, it could be shown by this imaging technology that systemic treatment with the angiogenesis inhibitors endostatin peptide mP-1 and caplostatin significantly delays and suppresses angiogenesis within endometriotic lesions (Becker et al. 2006). These results indicate that bioluminescence may be a useful tool to monitor the efficacy of anti-angiogenic drugs in the treatment of endometriosis, especially in combination with the first genetic mouse model of spontaneous endometriosis (Dinulescu et al. 2005). However, a major disadvantage of this approach is the fact that it is dependent on specific transgenic mice. Moreover, bioluminescence lacks a high spatial resolution. Alternatively, intraperitoneal endometriotic lesions may be analyzed by high-resolution ultrasound imaging (Fig. 30.4). Using high-frequency scanheads, this technology allows for the noninvasive analysis of endometriotic lesions in mice with good soft tissue contrast and without the application of i.v. contrast agents. To provide optimal conditions for ultrasound imaging, endometrial tissue samples should be sutured to the abdominal wall, which makes it easy to detect them and to differentiate them from adjacent intra-abdominal organs during repetitive analyses. Importantly, this technique provides complete new information about endometriotic lesions. The localization and extension of anechoic cysts can nicely be visualized in three-dimensional image reconstructions. Furthermore, it is possible to detect noninvasively adhesion formation around the developing lesions. Tissue perfusion can be monitored by color-coded Doppler ultrasound imaging. In addition, microbubbles targeted to specific neovascular endothelial receptors such as av-integrins may be used to assess angiogenesis in endometriotic lesions. Accordingly, high-resolution ultrasound imaging is a reliable and easy method to gain new insights in the pathogenesis of endometriosis. Moreover, it is an excellent tool in preclinical research to test the efficacy of novel medical therapies for the treatment of this frequent gynecological disease.

30.6 Outlook Rapid progress in the development of high-resolution imaging technologies for small animals allows now for the noninvasive and repetitive in vivo analysis of physiological and pathological processes within the female reproductive organs. This option will not only contribute to the reduction of animals required for future preclinical studies, but also significantly increase our knowledge about the physiology of reproduction and the pathogenesis of numerous gynecological diseases. In contrast to the ex vivo analysis of isolated tissue samples by means of genetic, biochemic, or histological techniques, noninvasive imaging methods bear the major advantage that they enable for the assessment of functional parameters in highly dynamic tissues, as they can typically be found within the ovary, the uterus, or the placenta. Important examples are the measurement of placental perfusion and permeability by dynamic contrast-enhanced MRI or the determination of functional capillary density in developing endometriotic lesions by intravital fluorescence microscopy. Accordingly, small animal imaging technologies have already a major impact on gynecology research and will essentially contribute in the future to the establishment of novel diagnostic and therapeutic strategies in clinical gynecology.

30.7 Lab Protocol The following protocol describes an easy model for the surgical induction of standardized intraperitoneal endometriotic lesions in mice and their noninvasive analysis by means of high-resolution ultrasound imaging: 1. To exclude discrepancies of sex hormone levels between individual mice, female animals of the same stage of the estrous cycle are selected. For this purpose, 15 mL of 0.9% saline are carefully pipetted into the vagina and subsequently transferred on a glass slide for examination under a phase contrast microscope. The four different cycle stages can easily be distinguished by their typical cytology: Proestrus (nucleated epithelia and leukocytes), estrus (cornified epithelia), metestrus (cornified epithelia and leukocytes), and diestrus (leukocytes).

446

2. Donor mice are anesthetized by i.p. injection of ketamine (75 mg/kg body weight) and xylazine 2% (15 mg/kg body weight). 3. Both uterine horns are removed and transferred in a Petri dish containing 37°C warm Dulbecco’s modified Eagle medium (DMEM; 10% fetal calf serum, 100 U/mL penicillin, 0.1 mg/mL streptomycin). 4. The uterine horns are opened longitudinally with micro scissors under a stereo microscope and tissue samples of comparable size are removed using a 2-mm dermal biopsy punch. 5. For induction of intraperitoneal endometriotic lesions, the recipient animals are anesthetized by i.p. injection of ketamine/xylazine 2%, and one tissue sample per quadrant is fixed with a 7–0 Prolene suture to the peritoneal wall through a midline incision. 6. The laparotomy is closed again with a 6–0 Prolene muscle and skin suture. During the following weeks to months, the developing endometriotic lesions can be repetitively analyzed by high-resolution ultrasound imaging. 7. For ultrasound image acquisition, mice are anesthetized with 2% isoflurane in oxygen and fixed in supine position on a heated stage. 8. After chemical depilation of the abdomen to prevent air trapped in the fur from interfering with ultrasound coupling into the animal, ultrasound coupling gel is applied to the skin. 9. Subsequent ultrasound imaging of developing endometriotic lesions, their cysts, and adhesion formation can, for example, be performed with a Vevo 770 high-resolution imaging system and a 704 scanhead (VisualSonics, Toronto, ON) with a center frequency of 40 MHz and a focal depth of 6 mm (Fig. 30.4).

References Akirav C, Lu Y, Mu J, Qu DW, Zhou YQ, Stevin J, Holmyard D, Foster FS, Adamson SL (2005) ultrasonic detection and developmental changes in calcification of the placenta during normal pregnancy in mice. Placenta 26(2–3):129–137 Becker CM, Wright RD, Satchi-Fainaro R, Funakoshi T, Folkman J, Kung AL, D’Amato RJ (2006) A novel noninvasive model of endometriosis for monitoring the efficacy of antiangiogenic therapy. Am J Pathol 168(6):2074–2084 Dinulescu DM, Ince TA, Quade BJ, Shafer SA, Crowley D, Jacks T (2005) Role of K-ras and Pten in the development of

M.W. Laschke and M.D. Menger mouse models of endometriosis and endometrioid ovarian cancer. Nat Med 11(1):63–70 Flecknell P (2002) Replacement, reduction and refinement. ALTEX 19(2):73–78 Grümmer R (2006) Animal models in endometriosis research. Hum Reprod Update 12(5):641–649 Hensley H, Quinn BA, Wolf RL, Litwin SL, Mabuchi S, Williams SJ, Williams C, Hamilton TC, Connolly DC (2007) Magnetic resonance imaging for detection and determination of tumor volume in a genetically engineered mouse model of ovarian cancer. Cancer Biol Ther 6(11):1717–1725 Israely T, Dafni H, Granot D, Nevo N, Tsafriri A, Neeman M (2003) Vascular remodeling and angiogenesis in ectopic ovarian transplants: a crucial role of pericytes and vascular smooth muscle cells in maintenance of ovarian grafts. Biol Reprod 68(6):2055–2064 Israely T, Dafni H, Nevo N, Tsafriri A, Neeman M (2004) Angiogenesis in ectopic ovarian xenotransplantation: multiparameter characterization of the neovasculature by dynamic contrast-enhanced MRI. Magn Reson Med 52(4): 741–750 Jansen SA, Conzen SD, Fan X, Krausz T, Zamora M, Foxley S, River J, Newstead GM, Karczmar GS (2008) Detection of in situ mammary cancer in a transgenic mouse model: in vitro and in vivo MRI studies demonstrate histopathologic correlation. Phys Med Biol 53(19):5481–5493 Laschke MW, Menger MD (2007) In vitro and in vivo approaches to study angiogenesis in the pathophysiology and therapy of endometriosis. Hum Reprod Update 13(4):331–342 Laschke MW, Menger MD, Vollmar B (2002) Ovariectomy improves neovascularization and microcirculation of freely transplanted ovarian follicles. J Endocrinol 172(3): 535–544 Laschke MW, Menger MD, Vollmar B (2003) Cryopreservation does not affect neovascularization of freely transplanted ovarian follicles. Fertil Steril 79(6):1458–1460 Menger MD, Laschke MW, Vollmar B (2002) Viewing the microcirculation through the window: some twenty years experience with the hamster dorsal skinfold chamber. Eur Surg Res 34(1–2):83–91 Mu J, Adamson SL (2006) Developmental changes in hemodynamics of uterine artery, utero- and umbilicoplacental, and vitelline circulations in mouse throughout gestation. Am J Physiol Heart Circ Physiol 291(3):H1421–H1428 Mu J, Slevin JC, Qu D, McCormick S, Adamson SL (2008) In vivo quantification of embryonic and placental growth during gestation in mice using micro-ultrasound. Reprod Biol Endocrinol 6:34 Pallares P, Gonzalez-Bulnes A (2008) The feasibility of ultrasound biomicroscopy for non-invasive and sequential assessment of ovarian features in rodents. Reprod Biol 8(3):279–284 Rossant J, Cross JC (2001) Placental development: lessons from mouse mutants. Nat Rev Genet 2(7):538–548 Salomon LJ, Sianve N, Taillieu F, Balvay D, Vayssettes C, Frija G, Ville Y, Cuénod CA, Clément O (2006) In vivo dynamic MRI measurement of the noradn-naline-induced reduction in placental blood flow in mice. Placenta 27(9–10): 1007–1013 Taillieu F, Salomon LJ, Siauve N, Clément O, Faye N, Balvay D, Vayssettes C, Frija G, Ville Y, Cuenod CA (2006) Placental

30 Imaging in Gynecology Research perfusion and permeability: simultaneous assessment with dual-echo contrast-enhanced MR imaging in mice. Radiology 241(3):737–745 Tempel-Brami C, Neeman M (2002) Non-invasive analysis of rat ovarian angiogenesis by MRI. Mol Cell Endocrinol 187(1–2):19–22 Vollmar B, Laschke MW, Rohan R, Koenig J, Menger MD (2001) In vivo imaging of physiological angiogenesis from

447 immature to preovulatory ovarian follicles. Am J Pathol 159(5):1661–1670 Walker CL, Hunter D, Everitt JI (2003) Uterine leiomyoma in the Eker rat: a unique model for important diseases of women. Genes Chromosomes Cancer 38(4):349–356

Imaging in Cardiovascular Research

31

Michael Schäfers, Klaus Tiemann, Michael Kuhlmann, Lars Stegger, Klaus Schäfers, and Sven Hermann

Contents 31.1

Clinical Challenges in Cardiovascular Medicine: Rationale for Basic Cardiovascular Research Involving Small Animal Imaging............................. 449

31.2 Mouse Models of Human Cardiovascular Diseases: Basis for Small Animal Imaging in Cardiovascular Research....................................... 451 31.2.1 Myocardial, Peripheral, and Cerebral Ischemia....... 451 31.2.2 Nonischemic Heart Failure....................................... 452 31.2.3 Atherosclerosis......................................................... 452 31.3 Applications of High-Resolution Small Animal Imaging in Cardiovascular Research: Examples.................................................................. 453 31.3.1 Computed Tomography............................................ 453 31.3.2 Ultrasound................................................................ 453 31.3.3 Magnetic Resonance Imaging.................................. 458 31.3.4 Scintigraphic Imaging: SPECT and PET.................. 461 31.3.5 Emerging Cardiovascular Applications of Molecular and Cellular Targeting by Small Animal PET and SPECT................................ 463 31.3.6 Optical Imaging........................................................ 467 References............................................................................ 468

M. Schäfers (*) and L. Stegger Department of Nuclear Medicine, University Hospital Münster, UKM, Albert-Schweitzer-Strasse 33, 48149 Münster, Germany e-mail: [email protected] K. Tiemann Dept. of Cardiology and Angiology University Hospital, Münster, UKM, Albert-Schweitzer-Str. 33, 48149 Münster, Germany K. Schäfers, M. Kuhlmann and S. Hermann European Institute for Molecular Imaging EIMI University of Münster Mendelstr. 11, 48149-Münster, Germany

31.1 Clinical Challenges in Cardiovascular Medicine: Rationale for Basic Cardiovascular Research Involving Small Animal Imaging Cardiovascular events most frequently stem from vascular diseases, namely atherosclerosis, with the clinical sequelae myocardial infarction or stroke. These clinical scenarios result from complex inflammatory changes in the vascular wall (atherosclerotic plaques). If these plaques are mechanically unstable (“vulnerable”), a rupture or erosion of the plaque surface may eventually lead to thrombosis with partial or complete occlusion of the respective vessel. Organs such as the heart or the brain, which are heavily dependent on the blood supply by arteries, thus suffer from acute ischemia. Despite optimized therapy regimes (acute PCI, thrombolytic therapies, anti-ischemic medical strategies etc.), a significant loss of cells due to apoptosis and necrosis is observed in many cases with clinical consequences ranging from dysfunction to disability to death. The incidence of chronic heart failure, a widespread cardiac disease resulting from acute or recurrent myocardial infarctions, long-term hypertension, valve diseases, or cardiomyopathies is dramatically increasing worldwide. The risk for heart failure approximately doubles with each decade of life. Although the progressive decline in cardiac contractile function is responsible for the majority of deaths due to heart failure, other deadly events might also occur from acute severe tachyarrhythmias. Despite highly potent medical therapies, the prognosis of heart failure patients is as poor as for patients with advanced tumor diseases. Prognosis in heart failure is influenced by vascular and myocardial factors.

F. Kiessling and B.J. Pichler (eds.), Small Animal Imaging, DOI: 10.1007/978-3-642-12945-2_31, © Springer-Verlag Berlin Heidelberg 2011

449

450

The armamentarium of established diagnostic clinical tools (ECG, angiography, perfusion SPECT, other stress tests etc.) together with recently emerging improvements in imaging technologies such as threedimensional echocardiography, multislice/dual source computed tomography (MSCT), or magnetic resonance imaging (MRI) allows for a detailed characterization of the cardiovascular system in men. However, these imaging modalities only provide insight into the morphological and/or functional consequences of cardiovascular diseases (coronary artery stenoses, cardiac dilation and hypertrophy, arrhythmias, perfusion deficits, impaired contractile function etc.) and do not assess the underlying cellular and molecular pathophysiology of the respective diseases. On the other hand, significant new insight into the molecular basis of cardiovascular diseases has been gained from genetic screening and the development and application of transgenic animal models. Atherosclerosis associated with the life-threatening risk of plaque ruptures is a good example of a cardiovascular disease where individual risk stratification is currently impossible. The initial atherosclerotic lesion can be triggered by various noxae, hypercholesterolemia, arterial hypertension, smoking, and such, primarily damaging the endothelium (endothelial dysfunction). In the course of the disease, the endothelium aims at neutralization of the initial noxae. If this is unsuccessful, a chronic inflammation of the vessel wall results. The inflammation is characterized and driven by infiltration of cells into the vessel wall, resulting in the long-term complex lesion (plaque) composed of activated macrophages/foam cells, T lymphocytes, mast cells etc. around a necrotic lipid-rich core. Plaques can be occlusive and be detected by clinical signs such as pectoral angina and shortness of breath. The much more dangerous type of an atherosclerotic lesion is the unstable “vulnerable” plaque, which can suddenly rupture, expelling the thrombogenic plaque content into the blood stream triggering a thrombotic vessel occlusion and the clinical sequelae myocardial infarction or stroke. A clinically available strategy towards risk stratification of individuals is based on analyzing known risk factors for atherosclerosis such as age, sex, smoking, positive family history, diabetes, and such, which were identified by large epidemiological studies. These findings have been implemented into risk calculators, which can excellently predict risk for a group of patients. However, even these approaches finally fail to

M. Schäfers et al.

identify individual risk. This means that current medicine will have to treat more patients than needed. For example, in a typical high-risk group of patients, which is identified by risk factor analysis, the 10-year risk for cardiovascular events can be as high as 20–40%. This implies that only every third to fifth patient will undergo the event. Since we cannot identify these, we will preventively treat all patients. Among other strategies such as blood biomarker analysis, vascular imaging for direct assessment/visualization of the specific processes involved in atherosclerosis in a single patient is expected to contribute to the individual risk assessment in the future. Current imaging technologies, which are predominantly based on morphological and functional imaging approaches of vessels and the myocardium, seem to fail in this respect. Conventional angiography can only detect coronary artery stenoses and is blinded to the vessel wall whereas high-resolution MSCT can add information on calcium deposition and even noncalcified soft plaques in the vessel wall. However, although these diagnostic tools can identify coronary artery disease with high sensitivity, their specificity with respect to the prediction of individual outcome and future cardiovascular events remains limited. Again, the weakness of the morphological imaging approaches might be explained by the fact that these see only secondary disease processes but not the molecular mechanisms that are decisive for the progression and outcome of the disease. In the case of atherosclerosis, it is now well understood that the molecular composition of an atherosclerotic plaque is determining its fate rather than its size. About 50% of cardiovascular events result from low-degree stenosis – diameter narrowing of the vessel of less than 50%. Molecular targets are therefore in the focus of research; the development of both target-specific molecular imaging and therapies promise substantial clinical innovations and improvements. Another example for a clinical challenge unmet by imaging of the basic pathophysiology is the field of life-threatening tachyarrhythmias, which can occur in a broad spectrum of diseases ranging from idiopathic diseases including the inherited cardiomyopathies to coronary artery disease. Although these tachyarrhythmias can be described and characterized by Holter recordings, electrophysiological studies etc., prediction of the individual risk for future arrhythmic events remains a challenge currently unmet by any diagnostic technology. In the case of tachyarrhythmias, it is well

451

31 Imaging in Cardiovascular Research

known that amongst others, electrical changes on the level of the affected myocardium and disturbances in nerve function can trigger events. In the latter respect, sympathetic nerves play an important role. Again, their function can be assessed by molecular imaging technologies quantifying catecholamine recycling, b-adrenoceptor density, and such. First studies have already demonstrated that these molecular imaging approaches can identify patients at risk for future cardiovascular events. The combination of molecular imaging with emerging knowledge about the genetic background of various different types of tachyarrhythmias may pave the way for a better risk estimation for future arrhythmic events. These facts justify an intense research effort towards both new diagnostic and therapeutic approaches with a special focus on the prevention of cardiovascular events. In this field of basic and translational cardiovascular research, high-resolution imaging in small animals recently became a substantial tool; an overview is given here. Small animal imaging can be and is frequently used in different aspects of basic and translational cardiovascular research: • Phenotyping. Imaging to characterize genetic or surgical mouse models mimicking human cardiovascular diseases. Phenotyping of new and existing animal models is substantially supported by functional and molecular small animal imaging, especially since small animal imaging characteristically can be performed noninvasively and serially in individual animal. Furthermore, whole-body imaging is a particular value to discover pathologies developing in organs “outside” the focus of the study. • Monitoring. Imaging to monitor morphological, functional, and molecular changes induced by interventional studies (pharmaceutical treatment, surgical intervention, gene therapy, stem cell therapy). Again, serial noninvasive imaging studies do allow that intraindividual changes be assessed in a few animals. • Development. Since the majority of clinical cardiovascular imaging devices are now available in a “miniaturized” fashion for small animal imaging with similar imaging characteristics, it seems straightforward to use small animal imaging in validation studies to characterize new imaging approaches in animal models first and then translate into clinical

imaging by using the analogous clinical imaging modality. This is especially useful when testing new imaging probes for molecular imaging (contrast agents, radiopharmaceutical, optical dyes etc.).

31.2 Mouse Models of Human Cardiovascular Diseases: Basis for Small Animal Imaging in Cardiovascular Research Appropriate animal models, mimicking the pathology of human cardiovascular diseases, are essential for progress in the field of cardiovascular research and for the transfer of preclinical results into the clinical setting. In principle, a variety of animal species can be used in cardiovascular research. Although larger animals such as pigs (a classical model for myocardial infarction) are used in cardiovascular research, in the context of this book, only small animal models in rats and mice are being discussed. Animal models of cardiovascular diseases can rely on both genetic manipulations and surgical interventions and do nowadays cover the wide spectrum of human cardiovascular diseases. In this context, examples for relevant human cardiovascular diseases, myocardial infarction, and atherosclerosis, already being studied by small animal imaging rather than a complete overview, are given.

31.2.1 Myocardial, Peripheral, and Cerebral Ischemia Induction of acute myocardial infarction – permanent and transient occlusion. Acute myocardial infarction can be induced in mice by a surgical intervention, in which the left ventricular blood supply will be interrupted by a permanent or transient ligation of the left anterior descending coronary artery (LAD) (Kuhlmann et al. 2006). Depending on the length of ischemia and the length of reperfusion in the case of nonpermanent ligation, the degree of ischemic damage and scar formation can be varied. However, reproducibility with respect to the amount of ischemic tissue, scarring, and localization is limited by a quite variable coronary

452

anatomy in mice. These models can either be performed in wildtype or transgenic mouse models. Induction of cerebral ischemia – stroke model. The classical animal model resembling human stroke is the so-called middle cerebral artery occlusion (MCAo) model. In this model, a thin thread is inserted into the A. carotis interna and guided into the A. cerebri media, which in turn is occluded by the thread. The resulting ischemia in the media territory causes focal lesions resembling the clinical situation of stroke. The procedure is typically controlled by measuring the regional brain perfusion by a specialized laser Doppler device (Klohs et al. 2009). Peripheral Ischemia – hind limb model. In the hind limb model, permanent unilateral ligation of the A. femoralis induces chronic peripheral ischemia – a pathophysiological situation often associated with diabetes and smoking in patients. This model drives neoangiogenesis to circumvent the femoral artery ligation and is therefore often employed when imaging of angiogenesis is in the focus (Dobrucki et al. 2009).

31.2.2 Nonischemic Heart Failure Thoracal aortic constriction (TAC). Nonischemic heart failure can be induced in mice by a surgical intervention, the so-called “thoracic aortic constriction” (TAC) model. Here, the aortic arch is partially constricted by using a surgically introduced thread inducing cardiac pressure overload, finally resulting in pathological hypertrophy and subsequently dilative cardiomyopathy (Tsujita et al. 2005). Genetically triggered heart failure. Besides the surgical approaches to induce heart failure, different genetically determined models have been developed. A good example is the survivin knock-out mouse. Survivin belongs to the class of inhibitors of apoptosis proteins (IAPs); its knock-out is resulting in a spontaneous progressive heart failure (Levkau et al. 2008).

31.2.3 Atherosclerosis Transgenic mouse models to produce spontaneous atherosclerosis. When fed a normal and speciesappropriate diet, mice or rats do typically not develop any human-like atherosclerosis in their life time.

M. Schäfers et al.

Therefore, metabolic changes incurred by gene knockin or knock-out with additional diets, e.g., cholesterolrich food, have to be employed to produce atherosclerosis. The typical examples are the Apolipoprotein E – knockout (ApoE-/-) and the lipoprotein receptor knock-out mouse, both of which both spontaneously present with hypercholesterolemia when compared with their littermates, which in the long run (weeks) results in atherosclerosis at predilection sites such as the aortic root, the aortic arch, and the brachiocephalic artery. When feeding high cholesterol diets, the progression of the atherosclerosis is accelerated, and the inflammation within the vascular lesions is enhanced (Zhang et al. 1992). Surgical local induction of atherosclerotic lesions. Based on the above mentioned genetically determined atherosclerosis models, local lesions can be triggered by surgical interventions aiming at endothelial damage. Unilateral ligation of the common carotid artery in ApoE knock-out mice causes a vascular remodeling resulting in highly inflammatory plaque-like vascular lesions within 4 weeks (Ivan et al. 2002). Another model for inducing endothelial damage, thrombosis, and vascular remodeling is the wire-induced endothelial disruption, where the endothelium of the carotid artery is surgically injured by a wire (Manka et al. 2000). In contrast to the carotid ligation model, the surgical procedure is more sophisticated, but has the advantage of affecting the common carotid artery on its whole run. Since these models are based on relatively harsh, nonpathophysiological interventions, another model of carotid atherosclerosis induction was recently introduced, the shear stress induced arteriosclerotic plaques development (“Cast” model). This model employs the surgical placement of a small plastic collar (cast), consisting of two longitudinal halves of a cylinder, forming together a cone-shaped lumen, around one common carotid artery producing a carotid stenosis. Upstream of the site of constriction-lowered shear stress induces plaque development in the Western diet-fed ApoE−/− mouse (Cheng et al. 2006). Models for spontaneous myocardial infarction. The above described models of atherosclerosis typically do not exhibit signs of unstable plaques and plaque rupture and do not result in the clinical sequelae myocardial infarction or stroke. Therefore, a great interest exists with respect to new models. A good example towards more human-like models is the model of diet-induced occlusive coronary atherosclerosis in SrB1−/−/ApoE-R61h/h mice. In this mouse, two genes

453

31 Imaging in Cardiovascular Research

involved in lipid metabolism are modified: the HDL scavenger receptor class B, type 1 (SrB1), is knocked out, and in addition, mice of that strain express a mutant form of murine ApoE, Thr61 ® Arg61 (Arg61ApoE) (Zhang et al. 2005). When fed a high fat/ high cholesterol (HFC) diet atherosclerosis rapidly progresses, including massive plaque deposition, plaque rupture, occlusive coronary events, and myocardial infarctions, causing death within approximately 4 weeks after onset of the HFC diet.

31.3 Applications of High-Resolution Small Animal Imaging in Cardiovascular Research: Examples 31.3.1 Computed Tomography Over the past three decades, microCT imaging has rapidly advanced with higher quality resolution, the introduction of the cone beam reconstruction algorithm, and an increased availability of dedicated scanners for noninvasive small animal imaging research. To achieve the same resolution in relation to object size between human and small animal CT systems, small animal CTs have to be designed with spatial resolutions of 100 mm or better. This is especially true when it comes to imaging of the heart and vessels in mice in vivo, where respiratory and cardiac movement additionally degrades image quality. Currently, CT imaging of the heart and vessels of a living mouse therefore remains a huge challenge. The typical design of a microCT involves a tube and a detector array, which are rotated around the object slowly and noncontinuously (minutes), which results in a low or nonexistent temporal resolution. With a heart rate of ~400 bpm under anesthesia in mice, the movement of the heart and the vessels cannot be assessed by microCT, leading to blurring artifacts. It has been recently shown that using special prospective triggering algorithms or employing fast flat panel technology CT imaging of the heart cycle is feasible (Dinkel et al. 2008). However, these scans do characteristically increase radiation dose and need a stable contrast enhancement resulting by an intravenously injected iodinated contrast agent to achieve good image contrast between vessels, the myocardium and surrounding tissues. While the radiation dose is

primarily an important issue when studying individual animals serially over time, the use of contrast agents remains a challenge. Contrast agents with a long blood circulation time, which are unfortunately currently rather expensive, can be used or clinically available contrast agents. The latter have to be combined with an “intelligent” infusion protocol to balance a stable contrast with a volume limited as much as possible. With these prerequisites, small animal CT in cardiovascular applications so far has mainly been used in comparative studies, where a PET, SPECT, or fluorescence signal is overlaid on the CT to project a molecular signal onto morphology. Vascular anatomy. As detailed above, the imaging of small-caliber arteries in mice and rats remains a significant challenge. Therefore, publications describing microCT applications in the aorta or coronary arteries in mice are rare to nonexistent. The principle feasibility of imaging small arteries was recently nicely demonstrated by Kiessling et al. (2004) and Schambach et al. (2009), which imaged tumor supplying vessels cerebrovascular structures at resolutions of 50 mm and better in mice. (see Fig. 31.1). Imaging of the heart. As is true for vascular CT applications in mice, small animal CT in mouse or rat hearts has only rarely reported. However, studies have shown the principle feasibility of imaging heart anatomy and infarct size by delayed contrast enhancement employing long-circulating contrast agents (Dinkel et al. 2008; Nahrendorf et al. 2007). (see Fig. 31.2).

31.3.2 Ultrasound Ventricular morphology and function. Due to substantial technical improvements, ultrasound has become a key imaging technology for phenotypization of murine models in cardiovascular research. Echocardiography is ideally suited for cardiac phenotypization and assessment of left ventricular function since it is a real-time method, which requires minimal instrumentation of the mouse, allows fast imaging protocols, and can thereby be applied even in models with severely impaired physical conditions. Most recent transducer technology ranges from single-element transducers, which offer a unique spatial resolution, to linear array probes, which allow Duplex or Triplex imaging to 4D matrix transducers. This wide range of available systems.

454

M. Schäfers et al.

Fig. 31.1 High-resolution contrast-enhanced CT (Ultravist 300®) of a 25-g mouse where a small plastic cast is placed around the left carotid artery (arrow). (a) shows horizontal and

sagittal sections, (b) shows a segmentation and rendering of the left common carotid artery (red) and the surrounding cast (blue)

A major hallmark in phenotypization is systolic left ventricular function. Higher transducer frequencies as well as higher frame rates allow proper assessment of chamber dimensions, wall thickness, and global as well as regional left ventricular function (Collins et al. 2003; Gao et al. 2000). In adult mice, the symmetric shape of the ventricle allows proper assessment of LV-wall thickness, LV mass, and global regional function assessed by M-mode using the Teichholz formula. However, in asymmetrically shaped ventricles (e.g., new born or juvenile mice) or particularly in infarct models, M-Mode technology is not adequate for the assessment of left ventricular mass and function (Ghanem et al. 2006). 2D and most recently 3D

echocardiographies have been proposed as the ideal technology for the assessment of mass and function in transgenic models presenting with asymmetrically shaped ventricles or most importantly cardiac infarction models (Gao et al. 2000; Dawson et al. 2004; Ghanem et al. 2008). High-resolution ultrasound was found to be suitable to characterize the scar border zone in infarct models and to precisely assess regional ventricular function (Ghanem et al. 2008). Specialized imaging acquisition algorithms allow frame rates of up to 1,000 Hz and offer thereby a unique inside into regional ventricular function. Stress echocardiography has been found to be suitable to identify viable myocardium and to assess contractile reserve. In general,

Fig. 31.2 Contrast-enhanced CT of a 23-g mouse (Ultravist 300®). Sectional images (a transaxial, b horizontal, c sagittal) show a good contrast of the blood in the vessels and the ventricles of the heart to the surrounding tissues and the myocardium

31 Imaging in Cardiovascular Research

2D echocardiography applying mathematical models such as the area length method or the truncated ellipsoid model is sufficient for assessment of global left ventricular function and left ventricular mass (Collins et al. 2001; Tiemann et al. 2003). In more complex models, reconstructive or real-time 3D echocardiography is more accurate and superior to 1D or 2D echocardiographies (Ghanem et al. 2006; Dawson et al. 2004). LV-chamber dimensions, wall thickness, and function can be assessed either in conscious mice or by using volatile anesthesia to avoid cardiodepressive effects as seen by, e.g., ketamine/xylazin. 2D echocardiography can be used to analyze the shape and the ventricular geometry, allows identifying missing or extra structures, and is suited to identify structures such as thrombus formation or pericardial effusion. In addition, diastolic function can be assessed using Doppler technology (PW-Doppler of mitral inflow), Tissue-Doppler imaging (TDI), or volumetime curves derived by 4D echocardiography. These techniques have been applied in models of hypertrophy, heart failure, and others. Doppler ultrasound,

455

particularly color Doppler, allows demonstrating shunt flow in congenital heart disease. Close correlation between echocardiography findings and histology has been recently demonstrated.10 In adult mice, morphology and function of heart-valves, particularly mitral valve and aortic valve, can be assessed by 2D echocardiography. In addition, regurgitated or stenotic flow can be quantified and characterized by PW-Doppler technology. However, morphological and visual assessment of valvular geometry and function is limited due to spatial and temporal resolution. Thus, the current gold standard for evaluation is Doppler technology. New high-resolution imaging probes allow the morphological and functional assessment of left ventricular structures even in utero (Gui et al. 1996; Yu et al. 2008). Due to constant movements, quantification is, however, limited. (see Fig. 31.3). Vascular anatomy. There are several mouse models of arteriosclerosis available today, which mimic lesion phenotypes comparable to human disease. Transgene murine models as well as operative manipulations (wire injury, balloon injury, or constrictive models)

Fig. 31.3 High-resolution murine echocardiography (30 MHz). Morphology and function of the murine heart can be nicely visualized and quantitatively assessed. Frame rates of up to 1,000 Hz provide an excellent temporal resolution

456

require high-resolution imaging technology for followup. Recent technical advances now offer real-time ultrasound imaging of vascular structures with a special resolution, which allows the assessment of single plaques as well as total plaque burden (Gan et al. 2007). Larger structures such as the aorta and larger blood vessels can be easily identified by this technology. Branches from the aortic arch as well as intravascular thrombi can be assessed. Vessel morphology and wall thickness, pulse wave propagation, and blood flow within the arteries can be analyzed by 2D imaging, color Doppler, and PW-Doppler technology. Identification and quantification of regional and total plaque burden are feasible by ultrasound biomicroscopy (Gan et al. 2007). This technology allows followup investigations in atherosclerotic models and thereby characterization and quantification of plaque location and growth. Intracerebral blood vessels can be identified by PW-Doppler technology, color Doppler technology, or even contrast ultrasound. Atherosclerotic plaques can be visualized and measured en face. Reconstructive 3D technology allows quantification of plaque volume/plaque burden. Close correlation between histomorphometry and ultrasonic assessment has been demonstrated (Gan et al. 2007). Reproducibility (intra and interobservability) was very good using this new technology. Blood vessels down to the level of coronary arteries can be assessed in real time, and blood flow within these vessels can be obtained. Assessment of intima media thickness and

M. Schäfers et al.

elasticity of the blood vessel in atherosclerotic plaques are feasible by ultrasound biomicroscopy with the advent of high-resolution linear area transducers; the limitation in even smaller blood vessels needs to be elucidated. In vivo assessment of plaque formation allows further insight in plaque development, particularly since, e.g., intravital, thrombus formation can be visualized, which is somehow difficult to assess by histological techniques. (see Fig. 31.4). Perfusion. Contrast specific imaging modalities opened new arena for murine echocardiography. Tissue perfusion at the level of coronary microcirculation has become possible and allows visual assessment of regional perfusion, perfusion defects, quantification of risk area and infarct size (Scherrer-Crosbie et al. 2000). It has been demonstrated that perfusion imaging is superior to wall motion analysis in the assessment of infarct size and risk area. Contrast specific imaging probes allow sufficient spatial and temporal resolution for perfusion imaging and provide real-time segmentation of tissue and regional blood volume. Commercially available as well as custom-made microbubbles have been used for this purpose (Scherrer-Crosbie et al. 2007). Besides qualitative assessment of tissue perfusion, infarct size, and risk area quantitative approaches have been proposed. This technology relies on the ultrasound-mediated destruction of contrast microbubbles and the assessment of replenishment rate at the level of microcirculation. Mathematical modeling of contrast replenishment allows the assessment of regional blood

Fig. 31.4 Left: aortic arch in a healthy mouse. Right: cross-sectional view of the aortic arch demonstrating plaque formation (arrows) in a mouse model of atherosclerosis (ApoE−/−)

31 Imaging in Cardiovascular Research

457

Fig. 31.5 Contrast-enhanced ultrasound (CEUS). Left: longaxis view of a murine heart in harmonic imaging mode. Note that almost no signals from the myocardium are visible in this mode. Center: following i.v. injection of 0.05 mL of Optison® opacification of the atria and chambers allow delineation of

endocardial borders. Since the myocardium appears still black, an optimal signal-to-noise ratio is achieved. Right: Intermittent imaging with refreshing the image once per cardiac cycle allows visualization of myocardial microcirculation by CEUS

volume as well as replenishment rate and therefore allows the assessment of even absolute myocardial blood flow (Thibault et al. 2005; Vogel et al. 2005). (see Fig. 31.5). Molecular targeting. Molecular targeting, particularly of vascular epitopes, has become feasible due to the introduction of contrast microbubbles and contrast specific imaging modalities. Ultrasound contrast agents (microbubbles) remain strictly intravascular after intravenous or intra-arterial injection. Different shell properties and the gas content characterize their resonance behavior and therefore their contrast specific detection by ultrasound as well as their stability within the sound field. Inert gases such as perfluorocarbons make them ideally suited for ultrasound imaging. Particularly their destructive behavior within the sound field provides a unique technology to visualize even single microbubbles (loss of correlation imaging). Molecular targeting relies on either intrinsic properties of the microbubble shell or the attachment of binding ligands. Phosphatidylserinecoated microbubbles, for example, stimulate the uptake of these particles by macrophages and have been therefore used for in vivo imaging of intraflammatory cells. This type of imaging (“hotspot imaging”) has been used to characterize inflammation in ischemia reperfusion models (Kunichika et al. 2003). Positively or negatively charged micro bubbles can be used in addition to facilitate unspecific binding. In contrast, specific binding ligands have been added to microbubbles for target-specific binding. In cardiovascular models, particularly selectines and integrines have been used today to characterize early stages of atherosclerosis and to image various stages of the adhesion cascade (Lindner 2009). Intracellular adhesion molecules (ICAM-1, VCAM) and selectines such as P-selectine have been targeted

(Kaufmann 2009). When combined with myocardial perfusion imaging, p-selectine has been found to serve as an “ischemic memory.” P-selectine-specific binding of imaging probes demonstrates up regulation of P-selectine in a post-ischemic area. Multimodality imaging (contrast perfusion imaging, 2D wall motion assessment, and molecular imaging) allows further insights and provides valuable information in addition to the molecular signal. Besides inflammation and early stages of atherosclerosis, avb3 has been addressed as a target for angiogenesis in infarct models. Again, in combination with perfusion and functional imaging, this molecular information can contribute to the understanding of vascular and ventricular remodeling after myocardial infarction. Although this technology was suitable in the areas of microcirculation, a general problem of contrast microbubbles was the inability to bind in high flow scenarios such as the aorta or larger arteries. The addition of Sialyl LewisX has been found to overcome this limitation. Mimicking rolling and sticking of leucocytes by Sialyl LewisX, the attachment of targeted microbubbles even in high flow settings such as the aorta could be demonstrated. Kaufmann (2009) used p-selectine as an early marker of atherosclerosis with Sialyl LewisXlabeled microbubbles and could demonstrate specific binding to p-selectine even in the aortic arch. Smaller particles (contrast nanoparticles) have been used to image thrombus formation in blood vessels. However, there are technical limitations to image contrast particles, which are smaller than 500 nm in size due to the resonance behavior of these particles. Newer imaging strategies, which might even address intracellular targets, rely on the release of gas bubbles from these nanoparticles, which might result in specific detection of even intracellular contrast agents.

458

Multimodality imaging taking advantage of contrast ultrasound for molecular targeting is a very interesting approach since imaging of strictly intravascular targets can be compared with technologies such as PET or optical imaging aiming for intracellular compartments. This complementary approach might offer new insights in the development of cardiovascular disease. The advent of smaller particles, new imaging technologies including the active release of gas bubbles, and the development of new ligands have to be evaluated for future research projects.

31.3.3 Magnetic Resonance Imaging As for the other imaging modalities discussed here for small animal cardiovascular imaging, the heart and vessels of a mouse/rat-sized animal pose significant challenges for the MRI technology with respect to spatial and temporal resolution. Since, per se, signal-tonoise ratios are dependent on the strength of the magnetic field of an MRI scanner, typically scanners with field strengths above 3 T have been used for small animal cardiovascular MRI studies (Tsui and Kraitchman 2009). Because of the small object size and fast heart rates in small animals, MRI of the heart and vessels typically must be performed employing both ECG and respiratory gating, which in itself is not straightforward in an MRI scanner on the basis of space restriction, influence of the MRI field on the ECG, and other effects. This dual-gating approach provides unblurred images from either one stage of the cardiac cycle or a dynamic series of stages from the cardiac cycle. MRI is therefore often stated as the reference for assessment of left and right ventricular volumes, myocardial wall thickness, and ejection fraction. However, since – in contrast to clinical scanners – small animal MRI scanners are not able to acquire a three-dimensional volume in a short time but rather do a slow sliceby-slice filling of the k-space, significant inaccuracies (shifts etc.) can be detected between slices of the volume due to unequal timing of the data acquisition in different slices. Furthermore, image acquisition time for multiple short- and long-axis views can result in imaging times ranging from 30 min to several hours per animal also restricting an efficient “high-throughput” phenotyping of mouse models. New methods to image multiple mice simultaneously such as employing

M. Schäfers et al.

individual receiver coil arrays and parallel imaging techniques have been reported (Ramirez and Bankson 2007; Ramirez et al. 2007). However, these are currently not routinely available. To shorten acquisition time, image acquisition protocols without cardiac gating, where images are retrospectively reconstructed on the basis of a separately acquired navigator or the image data itself, have been proposed (Heijman et al. 2007). Herewith, imaging times can be reduced from several hours to a few minutes per imaging slice. Vascular anatomy. Starting already over 10 years ago, multiple studies have shown that the assessment of atherosclerosis progression in aortas of genetically engineered mice is feasible (Manka et al. 2000, Fayad et al. 1998). In the context of high-resolution imaging of small plaques in animal models of atherosclerosis, higher field strengths have been found advantageous for enhanced spatial resolution of the components of atherosclerotic plaques. With the use of paramagnetic blood pool contrast agents (e.g., Gd-DTPA), MRI is well suited to visualize the lumen of arterial and venous vessels throughout the whole body in humans and animals. Conventional high-resolution MRI is furthermore capable of detecting lipid-rich athero sclerotic plaques in both human atherosclerosis and animal models of atherosclerosis. However, the spontaneous tissue contrast of the atherosclerotic vascular wall against neighboring tissues is often limited. Therefore, new contrast agents have been tested. A good example is the use of paramagnetic liposomes. While in conventional multispectral MRI, neointimal lesions in an ApoE−/− carotid cast model were not distinguishable from the surrounding neck tissue; Mulder et al. (2006) found that injecting paramagnetic liposomes with a mean size of 90 nm resulted in a pronounced signal enhancement of >100% immediately after injection, which was sustained largely until 24 h postinjection. In contrast, the vessel wall of all controls (left carotid artery and animals injected with Gd-DTPA) did not show significant contrast enhancement at those time points. Further alternative contrast agents to delineate atherosclerotic plaques are currently under investigation in humans and will most likely also be used in small animals soon (Ibrahim et al. 2009). (see Fig. 31.6). Ventricular anatomy and function. As a first attempt to characterize the murine heart, various ECG-gated MRI sequences have been developed for quantifying left ventricular (LV) mass in mice. These methods have

31 Imaging in Cardiovascular Research

Fig. 31.6 MR angiography of the neck regions in a mouse showing constriction of the right carotid artery by a tapered cast. Image is a courtesy of G. Strijkers, TU Eindhoven, The Netherlands

been shown to be highly accurate (Berr et al. 2005). In addition to LV mass, state-of-the-art sequences can also quantify wall thickness, wall thickening, LV volume for each cardiac phase, and subsequently, end-diastolic volume, end-systolic volume, stroke volume, cardiac output, and EF. Multiple studies have employed this technology in wild type and animal models of ischemic or nonischemic heart failure (Levkau et al. 2008; Franco et al. 1999). Recently, it was demonstrated that measuring morphological and functional cardiac parameters in mice is also feasible with the use of a clinical 3 T-MRI

459

with a dedicated small animal coil (Bunck et al. 2009). Although the wall is significantly thinner and the geometry is more complex, the right ventricle can also be assessed by cine MRI. Indeed, it has been shown that cine MRI at 7 T with in-plane spatial resolution of approximately 0.1 × 0.1 mm2 can be used to quantify right ventricle function (Wiesmann et al. 2002). Similarly, high-resolution cine MRI at 7 T has also been successfully applied in the neonatal and juvenile mouse heart (Wiesmann et al. 2000). Apart from the LV morphology and contractile function, left ventricular perfusion can be regionally assessed by first-pass imaging of contrast agents such as Gd-DTPA or quantitatively by arterial spin labeling (ASL). ASL is currently the only quantitative MRI technique that has been developed for imaging myocardial perfusion in mice and was well validated by comparison to microspheres (Streif et al. 2005). However, high-end MRI techniques such as ASL have not seen many preclinical applications so far. The same is true for the imaging of myocardial viability in ischemia/infarction mouse models by MRI. Conventional methods to quantify infarct size after myocardial infarction in mice are not ideal, requiring either tissue destruction for histology or relying on nondirect measurements such as wall motion. Bohl et al. (2009) therefore successfully implemented a fast, high-resolution method to directly measure infarct size in vivo using three-dimensional late gadolinium enhancement small animal MRI resembling the clinically established late enhancement protocol. (see Fig. 31.7). Molecular and cellular targeting. Experience with cardiovascular molecular MRI in the myocardium and

Fig. 31.7 Horizontal long axis (A) and two short axis images (B=end-diastole, C=end-systole) of a murine heart. Images are courtesy of G. Strijkers, TU Eindhoven, The Netherlands

460

vessels is significantly less extensive than that with nuclear imaging agents, although, in principle, the characteristics of MRI make it well suited for myocardial imaging. As reported above, MRI is able to assess a range of anatomical, physiological, and functional parameters in the myocardium. Although the spatial resolution of MRI is excellent, its limitation when it comes to molecular imaging applications lies in its low sensitivity for the detection of conventional imaging agents (Sosnovik et al. 2007). To increase the sensitivity for cardiovascular molecular imaging, a variety of new gadolinium containing contrast agents consisting of either small molecules or large supramolecular structures such as liposomes were developed (Mulder et al. 2007a). Furthermore, iron-oxide-based magnetic nanoparticles have been developed and are now available for functionalization by antibodies or other target-affine ligands (McAteer et al. 2010). Their small size (20–50 nm), long blood circulation, and detectability at nanomolar sensitivity with T2-weighted pulse sequences make them particularly well suited for targeted imaging in the cardiovascular system. An early attempt to targeted molecular cardiovascular MRI was the use of antimyosin antibodies labeled with iron-oxide-based magnetic nanoparticles to image myocardial necrosis following acute myocardial infarction in rats (Weissleder et al. 1992). In the same scenario of ischemia/infarction as well as in heart failure, apoptosis plays an important role in cardiovascular pathophysiology. Small animal MRI in vivo using AnxCLIO-Cy5.5, an annexinlabeled magneto-fluorescent nanoparticle, has been

Fig. 31.8 Mouse carotid artery with atherosclerotic plaque at baseline (A) and 24h (B) after injection with collagen-targeted CNA35micelles. Images are courtesy of G. Strijkers, TU Eindhoven, The Netherlands

M. Schäfers et al.

successfully employed to study apoptosis in mice (Sosnovik et al. 2005). Besides molecular targeting by functionalized contrast agents, nonfunctionalized contrast agents have also been used to label and track cells. E.g., carbohydratecoated magnetic nanoparticles are efficiently phagocytosed by macrophages and other cells. Using these myocardial macrophages, infiltration in infarct healing was demonstrated recently (Nahrendorf et al. 2007). Using similar contrast agent constructs, stem cells injected into the myocardium can be followed and tracked by MRI (Ly et al. 2008). Recently, Flögel et al. (2008) showed that fluorocarbons that are phagocytosed by macrophages can be used in combination with 19FMRI to image macrophages with high contrast due to the lack of significant background noise in 19F-MRI imaging. Besides myocardial applications of molecular MRI, a variety of applications are already pursued or can be foreseen for molecular MRI of atherosclerotic plaques, where inflammatory cells and molecular targets are of great interest. Similar to their use in myocardial infarction, iron-oxide-based nanoparticles can be used to detect macrophages invading atherosclerotic plaques in mouse models (Tang et al. 2009). A broad spectrum of liposome or nanoparticle-based molecular MRI approaches to characterize atherosclerotic plaques ranging from the targeting of proteases over integrins for assessment of angiogenesis to cell apoptosis has been described in the literature and was recently excellently reviewed by Mulder et al. (2007b) (see Fig. 31.8).

31 Imaging in Cardiovascular Research

31.3.4 Scintigraphic Imaging: SPECT and PET With the general development and excitement of molecular imaging, scintigraphic imaging, one of the traditional functional and molecular imaging modalities, has recently again received novel attention. Together with fluorescence optical imaging, scintigraphic approaches provide the highest molecular sensitivity of all imaging modalities in vivo. With this extraordinarily high sensitivity, radionuclide imaging can be based on the injection of only trace amounts of target-affine ligands with radioisotopes (radiopharmaceuticals), which do not influence the organism under investigation. Scintigraphic imaging allows the functional and molecular study of various functional aspects of the heart, including ejection fraction, re gional wall motion, perfusion, viability, oxygen consumption, glucose and fatty acid metabolism, and such. Also, coronary artery functions and related diseases, such as ischemia, infarction, and atherosclerosis, can be investigated. In contrast to any other imaging modality, a broad spectrum of commercially available radiolabelled tracers or cyclotron-produced radiopharmaceuticals is clinically available for functional and molecular imaging and used in routine algorithms. Scintigraphic methods are currently the only clinical molecular imaging tools. Preclinical molecular imaging can straightforward make use of these existing radionuclide techniques and apply these using dedicated small animal devices to animal models of cardiovascular diseases and has therefore enjoyed much progress. Furthermore, new radioactive biomarkers and radiopharmaceuticals have been developed for cardiovascular imaging. These prerequisites have fuelled the application of preclinical scintigraphic imaging in the past years. Besides “pure” preclinical applications to test new radiopharmaceuticals or to characterize mouse models, successful preclinical molecular scintigraphic imaging can be directly translated to clinical studies, where the same trace amount principle is realized. Scintigraphic methods therefore gained particular importance for drug development and translational molecular imaging from in vitro to patients (Riemann et al. 2008). To accustom the clinical scintigraphic technology to the preclinical scenario with small object sizes and – especially true for the murine heart – fast and significant motion, dedicated SPECT and PET devices have

461

been designed in the past years. In SPECT, an adequate spatial resolution for cardiovascular studies can only be achieved by using pinhole techniques, which nowadays allow for spatial resolution well beyond 1 mm FWHM (van der Have et al. 2009; Pissarek et al. 2009). The same is true for PET: while the widely installed stateof-the-art crystal-based small animal PET systems provide spatial resolution close to 1 mm FWHM only in the center of the field-of-view, a so-called multiwire chamber-based system (quadHIDAC, Oxford Positron Ltd., UK) offers a submillimetre (0.7 mm FWHM) homogeneous resolution in a large field of view (16 × 28 cm) (Schafers et al. 2005; Rowland and Cherry 2008). The resolution of all existing SPECT and PET devices in comparison to a wall thickness of ~1.0 mm in the left murine heart, of ~0.5 mm in the right murine heart and of ~0.1 mm in the murine aorta, leads to significant partial volume artifacts when imaging the murine heart and vessels. These have to be corrected for quantitative measurements and for improving detectability of small lesions. A significant improvement of cardiac small animal scintigraphic imaging was realized by the introduction of gated SPECT and PET acquisitions. Besides improvement in spatial resolution in the single gates of the acquisition, functional parameters can be derived from scans with myocardium-affine tracers such as 18 F-FDG. When using contour-finding algorithms adapted from clinical gated SPECT and PET to small animal imaging, end-diastolic and -systolic volumes and ejection fraction can be reliably and reproducibly derived with high-resolution scanners even in the mouse heart (Schafers et al. 2006; Stegger et al. 2009; Vanhove et al. 2005; Constantinesco et al. 2005). The drawback of gated imaging in scintigraphic imaging is the loss in statistics in a single gate, which is even worse when gating for both cardiac and respiratory cycles. Currently, for human and preclinical use, methods are being developed to detect and compensate myocardial motion to come up with an optimal nonblurred and maximum-statistics image (Dawood et al. 2008). These algorithms seem to be mandatory for the future imaging of novel radiopharmaceuticals especially in coronary artery plaques, which are small and heavily moving. Myocardial Perfusion and Metabolism. Translated from clinical algorithms to diagnose myocardial parameters, the field of small animal SPECT and PET was initially established using clinically available tracers.

462

M. Schäfers et al.

Still, a majority of cardiovascular applications are based on the use of various perfusion tracers and 18 F-fluorodesoxyglucose, the latter is regularly available in PET centers through in-house production involving a cyclotron or through distribution networks worldwide. Several studies have used pinhole SPECT in combination with 99mTc-labeled clinically available perfusion tracers to obtain high-quality myocardial perfusion images in rats and mice (Constantinesco et al. 2005; Hirai et al. 2000). This technique can achieve excellent quantitative accuracy for in vivo measurement of even small myocardial infarctions in rats (Acton et al. 2006). In the ninetieth of the twentieth century with the improved resolution of the developed small animal PET systems, imaging of the heart of a rat and later of mice became feasible. 13NH3 ammonia was mostly used in a transfer of clinical PET protocols to study regional and absolute myocardial perfusion in rats and mice by small animal PET, which is especially interesting when studying the effect of pharmaceutical interventions on myocardial perfusion (Inubushi et al. 2004; Croteau et al. 2004) (see Fig. 31.9). Myocardial FDG uptake is heavily determined by the concentration of glucose in the plasma at the time of injection, insulin levels, depth and type of anesthesia

(isoflurane increases uptake), temperature, and such. Re producible and standardized preparation of the animals (starved vs. fed, baseline vs. insulin/glucose load, awake vs. anesthetized etc.) is therefore crucial to compare between individuals and to optimize the myocardial signal according to the scientific question. (see Fig. 31.10). Multiple studies have shown the principle feasibility of using FDG-PET to measure infarct size and to metabolically characterize myocardial infarctions in rats (Higuchi et al. 2007) or mice (Stegger et al. 2006). Results for the quantification of myocardial infarct size were typically validated against histology and/or MRI/echocardiography (see Fig. 31.11). Apart from the analysis of relative glucose utilization in the heart of small animals, PET was also employed to study absolute cardiac substrate metabolism. To assess absolute myocardial glucose uptake, either compartmental modeling or quantification in correlation to the injected radioactive dose (% injected dose) can be used. Shogi et al. have shown that Zucker diabetic fatty (ZDF) rats utilize significantly less glucose in the myocardium than their lean littermates (Shoghi et al. 2008). Levkau et al. (2008) demonstrated that in a knock-out mouse model resulting in progressive heart failure myocardial, 18 F-FDG uptake was increasing with the degree of heart

Fig. 31.9 Perfusion tracers such as 13N-Ammonia for PET in man can also be used in mice to study myocardial ischemia. The figure shows representative 0.4-mm thick sections of the reori-

ented murine left ventricle (SA short axis; VLA vertical long axis; HLA horizontal long axis) with a physiological perfusion

463

31 Imaging in Cardiovascular Research

Fig. 31.10 Myocardial uptake of 18F-FDG is heavily dependent on animal preparation. This figure displays typical horizontal whole-body slices of the same mouse with a good heart uptake following i.v.injection under isoflurane anesthesia (a) or i.p.

coinjection with an insuline/glucose cocktail in the awake animal (b) and virtually no heart uptake in the case of i.p. injection in the awake mouse (c). In all cases, the animal was allowed food ad libitum before the start of the protocols

failure. FDG-PET was also used to show a switch in cardiac substrate utilization towards glucose in myoglobin−/− mice, which was correlated with an overexpression of GLUT4 (Flogel et al. 2005). Further tracers for assessing the myocardial metabolism were and are studied by small animal PET and SPECT. For instance, in a dual tracer PET study in rats, Herrero et al. (2006) compared 11C-acetate and H215Owater to quantitatively assess myocardial perfusion; 11 C-acetate was also used by the same group to measure free fatty acid utilization in ZDF rats (Welch et al. 2006). Free fatty acid metabolism in the myocardium was studied in the course of ischemia/reperfusion in the rat heart by 123I-BMIPP (Higuchi et al. 2005). Because of the anatomic information, soft-tissue differentiation, and additional functional information offered by coregistered MR images, dual-modality preclinical small-animal SPECT/MRI and PET/MRI systems are under active research and development

(Wehrl et al. 2009). These developments should aid the molecular imaging of both the myocardium and the vessels (see Fig. 31.12).

31.3.5 Emerging Cardiovascular Applications of Molecular and Cellular Targeting by Small Animal PET and SPECT Myocardial Innervation. The autonomous nervous system plays an important role in regulating cardiac function and is implicated in many cardiac diseases ranging from coronary heart disease, congestive heart failure, and cardiomyopathies to life-threatening arrhythmias. Although less established in clinical practice than perfusion and metabolic imaging, examination of the

464

M. Schäfers et al.

Fig. 31.11 Polarmaps (left) and single slices (SA short axis; VLA vertical long axis; HLA horizontal long axis) of high-resolution 18F-FDG-PET in a control mouse (upper row) and a mouse

following myocardial infarction (bottom row) nicely depicting the loss of myocardial viability in the apex of the left ventricle

Fig. 31.12 Pilot-integrated MRI/18F-FDG-PET scan of a murine heart in vivo. The scan was acquired on an experimental APDPET insert in a 7T preclinical MRI scanner developed at the

University of Tübingen (Prof. Bernd Pichler). ECG-gated images depict a myocardial infarction with an apical aneurysm (ED enddiastole; ES ensystole)

31 Imaging in Cardiovascular Research

cardiac autonomous nervous system is potentially equally valuable. A variety of tracers targeting the pre and postsynaptic sympathetic innervation of the heart are available and further developed. Since for imaging of receptors such as the beta-adrenoceptors on the myocardium, maximum sensitivity is needed to avoid interference between the tracer dose and the receptor, imaging of myocardial innervation remains an exclusive field of SPECT and PET. 11C-HED-PET and 123 I-MIBG-SPECT can be employed to study sympathetic innervation in rats or mice (Tipre et al. 2008, Goethals et al. 2009). PET and SPECT studies investigating postsyanaptic beta-adrenoceptor density in small animals have not been reported so far. Apoptosis. Apoptosis, or programmed cell death, occurs in many cardiovascular diseases such as myocardial infarction, atherosclerosis, cardiomyopathies, and postheart transplantation. As for other imaging methodologies, radiolabelled annexin V targeting phosphatidylserine occurring in the course of apoptosis on the outer cell membrane was used to image apoptosis in animal models and patients. Annexin V has been labeled with 99mTc for SPECT and 18F, 64Cu, or 124 I for PET studies (Kietselaer et al. 2004; Keen et al. 2005; Cauchon et al. 2007, Yagle et al. 2005). Kietselaer et al. (2004) have already successfully applied 99mTcAnnexin V in humans. However, phosphatidylserine is not an exclusive in vivo target for apoptosis. It is also elevated in other forms of cell stress not necessarily leading to apoptosis. Activation of caspases inside the cells signals the initiation of the pivotal cascade towards apoptosis. Caspases, therefore, are likely to be better targets for molecular imaging of apoptosis. Targeting an intracellular enzyme with a radiotracer, e.g., a radiolabelled caspase inhibitor, is conceptually more challenging than targeting a cell surface receptor such as phosphatidylserine. First approaches have recently been reviewed (Faust et al. 2009). Angiogenesis. In cardiovascular medicine, there is tremendous interest not only in therapeutic approaches that strive for preventing angiogenesis, e.g., in the context of restenoses after angioplasty with or without stent implantation, but also in approaches that promote angiogenesis either by administration of angiogenic factors, such as the vascular endothelial growth factor (VEGF), or by gene therapeutic approaches in patients with ischemic heart disease. The noninvasive molecular imaging of angiogenesis is a formidable means to improve therapy

465

monitoring. VEGF is a key mediator of angiogenesis. Lu et al. (2003) have shown in an in vivo animal experiment in rabbits that uptake of VEGF121 labeled with the SPECT radionuclide 111In is considerably higher in an ischemic hind leg compared to the contralateral nonischemic hind leg. Another important mediator is the avb3integrin receptor, a heterodimeric transmembrane glycoprotein with a and b subunits that mediates not only adhesion, but also proliferation and differentiation of endothelial cells. Since the expression of avb3-integrin is upregulated with angiogenesis, it should be a suitable target for molecular imaging. Peptides containing the amino acid sequence arginine-glycineaspartic acid (RGD) can interact with the avb3-integrin receptor. Consequently, a large number of radiotracers containing the RGD sequence have been developed and evaluated. The iodinated compound *I-gluco-RGD, where * can stand for any of 123I, 124I, 125I, and 131I, and the fluorinated PET radiotracer 18F-galacto-RGD have been applied in several preclinical and clinical imaging studies (Dijkgraaf et al. 2009). Most recently, 18F-galacto-RGD was employed for imaging of vascular inflammation in atherosclerotic mice (Laitinen et al. 2009). Vascular inflammation in atherosclerosis. One rapidly growing area of cardiovascular molecular imaging is the development of methods to characterize atherosclerotic plaques, which are the pathophysiological basis for most of the cardiovascular events. As detailed above, the vulnerability and instability of atherosclerotic plaques are dependent rather on the cellular and molecular composition of these plaques than on their size. Inflammatory targets play an utmost role and are therefore in the focus of new radioligand developments to image plaque vulnerability. Here, besides imaging of apoptosis and angiogenesis, both processes are involved in the progression of plaques, the molecular imaging of proteases such as metallo-proteinases (MMPs) and other inflammatory pathways is of great interest. A variety of MMP inhibitors typically based on small nonpeptidic compounds such as hydroxamates or barbiturates were labeled in the past years by different radioligands for SPECT and PET (Wagner et al. 2006). These have been successfully applied to different animal models of atherosclerosis and are now awaiting their first clinical application (Schafers et al. 2004; Fujimoto et al. 2008). Although small animal SPECT and PET offer unique molecular sensitivity, these techniques are often difficult to interpret on their own because of the lack of

466

M. Schäfers et al.

correlation with anatomic structures or biologic landmarks. This is especially true for the imaging of atherosclerosis in animal models. Because CT images provide excellent anatomic information, multimodality SPECT/CT and PET/CT or even SPECT/PET/CT have become available for clinical and small animal molecular imaging systems. However, even with these novel technologies, perfect coregistration that is needed for high-resolution imaging of the murine vessel wall is not straightforward, since motion detection and correction are currently not implemented. Furthermore, the resulting complex multimodal datasets need special workup to be visualized and interpreted. Recently, dedicated software to analyze mouse aorta PET/CT data was published (Ropinski et al. 2009). (see Fig. 31.13). Gene Reporter imaging. Therapeutic strategies based on the introduction of genes into the genome of living organisms have been developed for several cardiac diseases, e.g., atherosclerosis, myocardial infarction, heart failure, and arrhythmias. To monitor efficiency, safety etc., there is a considerable interest in molecular imaging techniques that offer the unique possibility to closely monitor gene distribution and

gene expression in vivo noninvasively as a direct way to assess the efficacy of gene transfer. SPECT and PET can be used in combination with a reporter gene, which is coexpressed with the gene of interest. Several reporter gene/probe systems have been developed for molecular radiotracer imaging, which utilize different kinds of interaction, such as enzyme-based reporter gene imaging using the herpes simplex virus type 1 thymidine kinase reporter gene (HSV1-tk) or its mutant derivative HSV1-sr39tk to induce thymidine kinase, which phosphorylates nucleoside reporter probes such as 5-iodo-1(2-deoxy-2-fluoro-b-D-arabinofuranosyl) uracil (FIAU) and 9-(4-fluoro-3-hydroxymethylbutyl) guanine (FHBG). This reporter gene/probe system has seen application in small animal PET imaging and even in humans (Wu et al. 2002). A variety of further approaches are summarized elsewhere (Min and Gambhir 2008). Cell Tracking. With the development and evaluation of novel algorithms for regenerative cardiovascular medicine involving stem cells and other cells in various cardiovascular diseases, molecular imaging and tracking of the biodistribution and fate of the applied cells have become important research areas. Here,

Fig. 31.13 Postprocessing of high-resolution multimodal imaging data is needed for integrated analysis and potential future diagnostics. This figure shows an example of multiple linked

views that are employed to allow a comparative visualization of aortic arches of mice scanned by PET and CT. For more detailed information, visit www.voreen.org

31 Imaging in Cardiovascular Research

challenges for imaging are the dynamic localization and quantification of the number of such cells in vivo, over long periods. For tracking cells with radioactive isotopes, two principle strategies exist: One is the direct labeling of cells, e.g., by incubation with 111 In-oxine or 18F-FDG in vitro or the use of gene reporters as detailed above (Zhang et al. 2008). For instance, by taking advantage of radionuclides with relatively long half-lives, small animal SPECT has been used to track the migration of radiolabeled stem cells to myocardial infarction (Brenner et al. 2004).

31.3.6 Optical Imaging In the last decade, the existing molecular imaging armamentarium is being strengthened by the development of fluorescent imaging approaches. With the heart and most vessels being located deep in tissues, especially in men, scattering of light plays an important role and is limiting the applications of optical imaging in car diovascular research. Furthermore, absorption of light by hemoglobin in the well-perfused organs, heart and

Fig. 31.14 White light (a) and fluorescence reflectance imaging (b) of an atherosclerotic plaque (P) developing after left carotid ligation in an ApoE−/− mouse using Cy5.5-labeled RGD (reproduced from Waldeck et al. 2008). In this case, the skin was opened to get direct imaging access to the carotid arteries

467

vessels is another thread for the technology. The development of near-infrared sensors and imaging probes has broadened this technique’s scope. Further developments aim at novel catheter-based approaches for imaging of vascular signals. While having a theoretically similar sensitivity, the main advantages of fluorescence imaging over radioisotope-based imaging are the stability of the imaging agents and the ability to activate/ switch-on imaging agents in vivo. Furthermore, fluorescence can also be measured by high-resolution microscopy of the same tissue sample allowing an accurate correlation of in vivo molecular imaging signals and histology parameters. As is true for radioisotope-based imaging, fluorescence imaging has the potential to provide a wealth of cellular and molecular information. However, although this technique has seen a tremendous variety of applications in tumors and other applications, cardiovascular applications in vivo in small animals such as mice and rats comparable to those involving the other imaging modalities reviewed here are still rare. An excellent overview over the current status and the potential was recently published (Sinusas et al. 2008; Nahrendorf et al. 2009). (see Fig. 31.14).

468

References Acton PD, Thomas D, Zhou R (2006) Quantitative imaging of myocardial infarct in rats with high resolution pinhole SPECT. Int J Cardiovasc Imaging 22(3–4):429–434 Berr SS, Roy RJ, French BA, Yang Z, Gilson W, Kramer CM, Epstein FH (2005) Black blood gradient echo cine magnetic resonance imaging of the mouse heart. Magn Reson Med 53(5):1074–1079 Bohl S, Lygate CA, Barnes H, Medway D, Stork LA, SchulzMenger J, Neubauer S, Schneider JE (2009) Advanced methods for quantification of infarct size in mice using threedimensional high-field late gadolinium enhancement MRI. Am J Physiol Heart Circ Physiol 296(4):H1200–H1208 Brenner W, Aicher A, Eckey T, Massoudi S, Zuhayra M, Koehl U, Heeschen C, Kampen WU, Zeiher AM, Dimmeler S, Henze E (2004) 111In-labeled CD34+ hematopoietic progenitor cells in a rat myocardial infarction model. J Nucl Med 45(3): 512–518 Bunck AC, Engelen MA, Schnackenburg B, Furkert J, Bremer C, Heindel W, Stypmann J, Maintz D (2009) Feasibility of functional cardiac MR imaging in mice using a clinical 3 tesla whole body scanner. Invest Radiol 44:749–756 Cauchon N, Langlois R, Rousseau JA, Tessier G, Cadorette J, Lecomte R, Hunting DJ, Pavan RA, Zeisler SK, van Lier JE (2007) PET imaging of apoptosis with (64)Cu-labeled streptavidin following pretargeting of phosphatidylserine with biotinylated annexin-V. Eur J Nucl Med Mol Imaging 34(2):247–258 Cheng C, Tempel D, van Haperen R, van der Baan A, Grosveld F, Daemen MJ, Krams R, de Crom R (2006) Atherosclerotic lesion size and vulnerability are determined by patterns of fluid shear stress. Circulation 113(23):2744–2753 Collins KA, Korcarz CE, Shroff SG, Bednarz JE, Fentzke RC, Lin H, Leiden JM, Lang RM (2001) Accuracy of echocardiographic estimates of left ventricular mass in mice. Am J Physiol Heart Circ Physiol 280(5):H1954–H1962 Collins KA, Korcarz CE, Lang RM (2003) Use of echocardiography for the phenotypic assessment of genetically altered mice. Physiol Genomics 13(3):227–239 Constantinesco A, Choquet P, Monassier L, Israel-Jost V, Mertz L (2005) Assessment of left ventricular perfusion, volumes, and motion in mice using pinhole gated SPECT. J Nucl Med 46(6):1005–1011 Croteau E, Benard F, Bentourkia M, Rousseau J, Paquette M, Lecomte R (2004) Quantitative myocardial perfusion and coronary reserve in rats with 13N-ammonia and small animal PET: impact of anesthesia and pharmacologic stress agents. J Nucl Med 45(11):1924–1930 Dawood M, Buther F, Jiang X, Schafers KP (2008) Respiratory motion correction in 3-D PET data with advanced optical flow algorithms. IEEE Trans Med Imaging 27(8):1164–1175 Dawson D, Lygate CA, Saunders J, Schneider JE, Ye X, Hulbert K, Noble JA, Neubauer S (2004) Quantitative 3-dimensional echocardiography for accurate and rapid cardiac phenotype characterization in mice. Circulation 110(12):1632–1637 Dijkgraaf I, Beer AJ, Wester HJ (2009) Application of RGDcontaining peptides as imaging probes for alphavbeta3 expression. Front Biosci 14:887–899

M. Schäfers et al. Dinkel J, Bartling SH, Kuntz J, Grasruck M, Kopp-Schneider A, Iwasaki M, Dimmeler S, Gupta R, Semmler W, Kauczor HU, Kiessling F (2008) Intrinsic gating for small-animal computed tomography: a robust ECG-less paradigm for deriving cardiac phase information and functional imaging. Circ Cardiovasc Imaging 1(3):235–243 Dobrucki LW, Dione DP, Kalinowski L, Dione D, Mendizabal M, Yu J, Papademetris X, Sessa WC, Sinusas AJ (2009) Serial noninvasive targeted imaging of peripheral angiogenesis: validation and application of a semiautomated quantitative approach. J Nucl Med 50(8):1356–1363 Faust A, Hermann S, Wagner S, Haufe G, Schober O, Schafers M, Kopka K (2009) Molecular imaging of apoptosis in vivo with scintigraphic and optical biomarkers – a status report. Anticancer Agents Med Chem 9:968–985 Fayad ZA, Fallon JT, Shinnar M, Wehrli S, Dansky HM, Poon M, Badimon JJ, Charlton SA, Fisher EA, Breslow JL, Fuster V (1998) Noninvasive In vivo high-resolution magnetic resonance imaging of atherosclerotic lesions in genetically engineered mice. Circulation 98(15):1541–1547 Flogel U, Laussmann T, Godecke A, Abanador N, Schafers M, Fingas CD, Metzger S, Levkau B, Jacoby C, Schrader J (2005) Lack of myoglobin causes a switch in cardiac substrate selection. Circ Res 96(8):e68–e75 Flogel U, Ding Z, Hardung H, Jander S, Reichmann G, Jacoby C, Schubert R, Schrader J (2008) In vivo monitoring of inflammation after cardiac and cerebral ischemia by fluorine magnetic resonance imaging. Circulation 118(2):140–148 Franco F, Thomas GD, Giroir B, Bryant D, Bullock MC, Chwialkowski MC, Victor RG, Peshock RM (1999) Magnetic resonance imaging and invasive evaluation of development of heart failure in transgenic mice with myocardial expression of tumor necrosis factor-alpha. Circulation 99(3):448–454 Fujimoto S, Hartung D, Ohshima S, Edwards DS, Zhou J, Yalamanchili P, Azure M, Fujimoto A, Isobe S, Matsumoto Y, Boersma H, Wong N, Yamazaki J, Narula N, Petrov A, Narula J (2008) Molecular imaging of matrix metalloproteinase in atherosclerotic lesions: resolution with dietary modification and statin therapy. J Am Coll Cardiol 52(23):1847–1857 Gan LM, Gronros J, Hagg U, Wikstrom J, Theodoropoulos C, Friberg P, Fritsche-Danielson R (2007) Non-invasive realtime imaging of atherosclerosis in mice using ultrasound biomicroscopy. Atherosclerosis 190(2):313–320 Gao XM, Dart AM, Dewar E, Jennings G, Du XJ (2000) Serial echocardiographic assessment of left ventricular dimensions and function after myocardial infarction in mice. Cardiovasc Res 45(2):330–338 Ghanem A, Roll W, Hashemi T, Dewald O, Djoufack PC, Fink KB, Schrickel J, Lewalter T, Luderitz B, Tiemann K (2006) Echocardiographic assessment of left ventricular mass in neonatal and adult mice: accuracy of different echocardiographic methods. Echocardiography 23(10):900–907 Ghanem A, Troatz C, Elhafi N, Dewald O, Heeschen C, Nickenig G, Stypmann J, Tiemann K (2008) Quantitation of myocardial borderzone using reconstructive 3-D echocardiography after chronic infarction in rats: incremental value of low-dose dobutamine. Ultrasound Med Biol 34(4): 559–566 Goethals LR, Weytjens CD, De Geeter F, Droogmans S, Caveliers V, Keyaerts M, Vanhove C, Van Camp G, Bossuyt

31 Imaging in Cardiovascular Research A, Lahoutte T (2009) Regional quantitative analysis of small animal myocardial sympathetic innervation and initial application in streptozotocin induced diabetes. Contrast Media Mol Imaging 4(4):174–182 Gui YH, Linask KK, Khowsathit P, Huhta JC (1996) Doppler echocardiography of normal and abnormal embryonic mouse heart. Pediatr Res 40(4):633–642 Heijman E, de Graaf W, Niessen P, Nauerth A, van Eys G, de Graaf L, Nicolay K, Strijkers GJ (2007) Comparison between prospective and retrospective triggering for mouse cardiac MRI. NMR Biomed 20(4):439–447 Herrero P, Kim J, Sharp TL, Engelbach JA, Lewis JS, Gropler RJ, Welch MJ (2006) Assessment of myocardial blood flow using 15O-water and 1–11C-acetate in rats with small-animal PET. J Nucl Med 47(3):477–485 Higuchi T, Taki J, Nakajima K, Kinuya S, Namura M, Tonami N (2005) Time course of discordant BMIPP and thallium uptake after ischemia and reperfusion in a rat model. J Nucl Med 46(1):172–175 Higuchi T, Nekolla SG, Jankaukas A, Weber AW, Huisman MC, Reder S, Ziegler SI, Schwaiger M, Bengel FM (2007) Characterization of normal and infarcted rat myocardium using a combination of small-animal PET and clinical MRI. J Nucl Med 48(2):288–294 Hirai T, Nohara R, Hosokawa R, Tanaka M, Inada H, Fujibayashi Y, Fujita M, Konishi J, Sasayama S (2000) Evaluation of myocardial infarct size in rat heart by pinhole SPECT. J Nucl Cardiol 7(2):107–111 Ibrahim T, Makowski MR, Jankauskas A, Maintz D, Karch M, Schachoff S, Manning WJ, Schomig A, Schwaiger M, Botnar RM (2009) Serial contrast-enhanced cardiac magnetic resonance imaging demonstrates regression of hyperenhancement within the coronary artery wall in patients after acute myocardial infarction. JACC Cardiovasc imaging 2(5):580–588 Inubushi M, Jordan MC, Roos KP, Ross RS, Chatziioannou AF, Stout DB, Dahlbom M, Schelbert HR (2004) Nitrogen-13 ammonia cardiac positron emission tomography in mice: effects of clonidine-induced changes in cardiac work on myocardial perfusion. Eur J Nucl Med Mol Imaging 31(1):110–116 Ivan E, Khatri JJ, Johnson C, Magid R, Godin D, Nandi S, Lessner S, Galis ZS (2002) Expansive arterial remodeling is associated with increased neointimal macrophage foam cell content: the murine model of macrophage-rich carotid artery lesions. Circulation 105(22):2686–2691 Kaufmann BA (2009) Ultrasound molecular imaging of atherosclerosis. Cardiovasc Res 83(4):617–625 Keen HG, Dekker BA, Disley L, Hastings D, Lyons S, Reader AJ, Ottewell P, Watson A, Zweit J (2005) Imaging apoptosis in vivo using 124I-annexin V and PET. Nucl Med Biol 32(4):395–402 Kiessling F, Greschus S, Lichy MP, Bock M, Fink C, Vosseler S, Moll J, Mueller MM, Fusenig NE, Traupe H, Semmler W (2004) Volumetric computed tomography (VCT): a new technology for noninvasive, high-resolution monitoring of tumor angiogenesis. Nat Med 10(10):1133–1138 Kietselaer BL, Reutelingsperger CP, Heidendal GA, Daemen MJ, Mess WH, Hofstra L, Narula J (2004) Noninvasive detection of plaque instability with use of radiolabeled annexin A5 in patients with carotid-artery atherosclerosis. N Engl J Med 350(14):1472–1473

469 Klohs J, Steinbrink J, Bourayou R, Mueller S, Cordell R, Licha K, Schirner M, Dirnagl U, Lindauer U, Wunder A (2009) Near-infrared fluorescence imaging with fluorescently labeled albumin: a novel method for non-invasive optical imaging of blood-brain barrier impairment after focal cerebral ischemia in mice. J Neurosci Meth 180(1):126–132 Kuhlmann MT, Kirchhof P, Klocke R, Hasib L, Stypmann J, Fabritz L, Stelljes M, Tian W, Zwiener M, Mueller M, Kienast J, Breithardt G, Nikol S (2006) G-CSF/SCF reduces inducible arrhythmias in the infarcted heart potentially via increased connexin43 expression and arteriogenesis. J Exp Med 203(1):87–97 Kunichika H, Peters B, Cotter B, Masugata H, Kunichika N, Wolf PL, DeMaria AN (2003) Visualization of risk-area myocardium as a high-intensity, hyperenhanced “hot spot” by myocardial contrast echocardiography following coronary reperfusion: quantitative analysis. J Am Coll Cardiol 42(3):552–557 Laitinen I, Saraste A, Weidl E, Poethko T, Weber AW, Nekolla SG, Leppanen P, Yla-Herttuala S, Holzlwimmer G, Walch A, Esposito I, Wester HJ, Knuuti J, Schwaiger M (2009) Evaluation of alphavbeta3 integrin-targeted positron emission tomography tracer 18F-galacto-RGD for imaging of vascular inflammation in atherosclerotic mice. Circ Cardiovasc Imaging 2(4):331–338 Levkau B, Schafers M, Wohlschlaeger J, von Wnuck Lipinski K, Keul P, Hermann S, Kawaguchi N, Kirchhof P, Fabritz L, Stypmann J, Stegger L, Flogel U, Schrader J, Fischer JW, Hsieh P, Ou YL, Mehrhof F, Tiemann K, Ghanem A, Matus M, Neumann J, Heusch G, Schmid KW, Conway EM, Baba HA (2008) Survivin determines cardiac function by controlling total cardiomyocyte number. Circulation 117(12): 1583–1593 Lindner JR (2009) Molecular imaging of cardiovascular disease with contrast-enhanced ultrasonography. Nat Rev Cardiol 6(7):475–481 Lu E, Wagner WR, Schellenberger U, Abraham JA, Klibanov AL, Woulfe SR, Csikari MM, Fischer D, Schreiner GF, Brandenburger GH, Villanueva FS (2003) Targeted in vivo labeling of receptors for vascular endothelial growth factor: approach to identification of ischemic tissue. Circulation 108(1):97–103 Ly HQ, Frangioni JV, Hajjar RJ (2008) Imaging in cardiac cellbased therapy: in vivo tracking of the biological fate of therapeutic cells. Nat Clin Pract Cardiovasc Med 5(2):S96–S102 Manka DR, Gilson W, Sarembock I, Ley K, Berr SS (2000) Noninvasive in vivo magnetic resonance imaging of injuryinduced neointima formation in the carotid artery of the apolipoprotein-E null mouse. J Magn Reson Imaging 12(5): 790–794 McAteer MA, Akhtar AM, von Zur Muhlen C, Choudhury RP (2010) An approach to molecular imaging of atherosclerosis, thrombosis, and vascular inflammation using microparticles of iron oxide. Atherosclerosis 209:18–27 Min JJ, Gambhir SS (2008) Molecular imaging of PET reporter gene expression. Handb Exp Pharmacol (185 Pt 2):277–303 Mulder WJ, Douma K, Koning GA, van Zandvoort MA, Lutgens E, Daemen MJ, Nicolay K, Strijkers GJ (2006) Liposomeenhanced MRI of neointimal lesions in the ApoE-KO mouse. Magn Reson Med 55(5):1170–1174

470 Mulder WJ, Griffioen AW, Strijkers GJ, Cormode DP, Nicolay K, Fayad ZA (2007a) Magnetic and fluorescent nanoparticles for multimodality imaging. Nanomedicine (Lond) 2(3): 307–324 Mulder WJ, Strijkers GJ, Vucic E, Cormode DP, Nicolay K, Fayad ZA (2007b) Magnetic resonance molecular imaging contrast agents and their application in atherosclerosis. Top Magn Reson Imaging 18(5):409–417 Nahrendorf M, Badea C, Hedlund LW, Figueiredo JL, Sosnovik DE, Johnson GA, Weissleder R (2007) High-resolution imaging of murine myocardial infarction with delayed-enhancement cine micro-CT. Am J Physiol Heart Circ Physiol 292(6):H3172– H3178 Nahrendorf M, Sosnovik DE, French BA, Swirski FK, Bengel F, Sadeghi MM, Lindner JR, Wu JC, Kraitchman DL, Fayad ZA, Sinusas AJ (2009) Multimodality cardiovascular molecular imaging, part II. Circ Cardiovasc Imaging 2(1):56–70 Pissarek M, Meyer-Kirchrath J, Hohlfeld T, Vollmar S, OrosPeusquens AM, Flogel U, Jacoby C, Krugel U, Schramm N (2009) Targeting murine heart and brain: visualisation conditions for multi-pinhole SPECT with (99 m)Tc- and (123) I-labelled probes. Eur J Nucl Med Mol Imaging 36(9): 1495–1509 Ramirez MS, Bankson JA (2007) A practical method for 2D multiple-animal MRI. J Magn Reson Imag 26(4):1162–1166 Ramirez MS, Ragan DK, Kundra V, Bankson JA (2007) Feasibility of multiple-mouse dynamic contrast-enhanced MRI. Magn Reson Med 58(3):610–615 Riemann B, Schafers KP, Schober O, Schafers M (2008) Small animal PET in preclinical studies: opportunities and challenges. Q J Nucl Med Mol Imaging 52(3):215–221 Ropinski T, Hermann S, Reich R, Schafers M, Hinrichs K (2009) Multimodal vessel visualization of mouse aorta PET/CT scans. IEEE Trans Vis Comput Graph 15(6):1515–1522 Rowland DJ, Cherry SR (2008) Small-animal preclinical nuclear medicine instrumentation and methodology. Semin Nucl Med 38(3):209–222 Schafers M, Riemann B, Kopka K, Breyholz HJ, Wagner S, Schafers KP, Law MP, Schober O, Levkau B (2004) Scintigraphic imaging of matrix metalloproteinase activity in the arterial wall in vivo. Circulation 109(21):2554–2559 Schafers KP, Reader AJ, Kriens M, Knoess C, Schober O, Schafers M (2005) Performance evaluation of the 32-module quadHIDAC small-animal PET scanner. J Nucl Med 46(6):996–1004 Schafers KP, Lang N, Stegger L, Schober O, Schafers M (2006) Gated listmode acquisition with the quadHIDAC animal PET to image mouse hearts. Z Med Phys 16(1):60–66 Schambach SJ, Bag S, Steil V, Isaza C, Schilling L, Groden C, Brockmann MA (2009) Ultrafast high-resolution in vivo volume-CTA of mice cerebral vessels. Stroke 40(4): 1444–1450 Scherrer-Crosbie M, Liel-Cohen N, Otsuji Y, Guerrero JL, Sullivan S, Levine RA, Picard MH (2000) Myocardial perfusion and wall motion in infarction border zone: assessment by myocardial contrast echocardiography. J Am Soc Echocardiogr 13(5):353–357 Scherrer-Crosbie M, Rodrigues AC, Hataishi R, Picard MH (2007) Infarct size assessment in mice. Echocardiography 24(1):90–96

M. Schäfers et al. Shoghi KI, Gropler RJ, Sharp T, Herrero P, Fettig N, Su Y, Mitra MS, Kovacs A, Finck BN, Welch MJ (2008) Time course of alterations in myocardial glucose utilization in the Zucker diabetic fatty rat with correlation to gene expression of glucose transporters: a small-animal PET investigation. J Nucl Med 49(8):1320–1327 Sinusas AJ, Bengel F, Nahrendorf M, Epstein FH, Wu JC, Villanueva FS, Fayad ZA, Gropler RJ (2008) Multimodality cardiovascular molecular imaging, part I. Circ Cardiovasc Imaging 1(3):244–256 Sosnovik DE, Schellenberger EA, Nahrendorf M, Novikov MS, Matsui T, Dai G, Reynolds F, Grazette L, Rosenzweig A, Weissleder R, Josephson L (2005) Magnetic resonance imaging of cardiomyocyte apoptosis with a novel magnetooptical nanoparticle. Magn Reson Med 54(3):718–724 Sosnovik DE, Nahrendorf M, Weissleder R (2007) Molecular magnetic resonance imaging in cardiovascular medicine. Circulation 115(15):2076–2086 Stegger L, Hoffmeier AN, Schafers KP, Hermann S, Schober O, Schafers MA, Theilmeier G (2006) Accurate noninvasive measurement of infarct size in mice with high-resolution PET. J Nucl Med 47(11):1837–1844 Stegger L, Heijman E, Schafers KP, Nicolay K, Schafers MA, Strijkers GJ (2009) Quantification of left ventricular volumes and ejection fraction in mice using PET, compared with MRI. J Nucl Med 50(1):132–138 Streif JU, Nahrendorf M, Hiller KH, Waller C, Wiesmann F, Rommel E, Haase A, Bauer WR (2005) In vivo assessment of absolute perfusion and intracapillary blood volume in the murine myocardium by spin labeling magnetic resonance imaging. Magn Reson Med 53(3):584–592 Tang TY, Muller KH, Graves MJ, Li ZY, Walsh SR, Young V, Sadat U, Howarth SP, Gillard JH (2009) Iron oxide particles for atheroma imaging. Arterioscler Thromb Vasc Biol 29(7):1001–1008 Thibault H, Lafitte S, Timperley J, Tariosse L, Becher H, Roudaut R, Dos Santos P (2005) Quantitative analysis of myocardial perfusion in rats by contrast echocardiography. J Am Soc Echocardiogr 18(12):1321–1328 Tiemann K, Weyer D, Djoufack PC, Ghanem A, Lewalter T, Dreiner U, Meyer R, Grohe C, Fink KB (2003) Increasing myocardial contraction and blood pressure in C57BL/6 mice during early postnatal development. Am J Physiol Heart Circ Physiol 284(2):H464–H474 Tipre DN, Fox JJ, Holt DP, Green G, Yu J, Pomper M, Dannals RF, Bengel FM (2008) In vivo PET imaging of cardiac presynaptic sympathoneuronal mechanisms in the rat. J Nucl Med 49(7):1189–1195 Tsui BM, Kraitchman DL (2009) Recent advances in smallanimal cardiovascular imaging. J Nucl Med 50(5):667–670 Tsujita Y, Kato T, Sussman MA (2005) Evaluation of left ventricular function in cardiomyopathic mice by tissue Doppler and color M-mode Doppler echocardiography. Echocardiography 22(3):245–253 van der Have F, Vastenhouw B, Ramakers RM, Branderhorst W, Krah JO, Ji C, Staelens SG, Beekman FJ (2009) U-SPECT-II: an ultra-high-resolution device for molecular small-animal imaging. J Nucl Med 50(4):599–605 Vanhove C, Lahoutte T, Defrise M, Bossuyt A, Franken PR (2005) Reproducibility of left ventricular volume and ejec-

31 Imaging in Cardiovascular Research tion fraction measurements in rat using pinhole gated SPECT. Eur J Nucl Med Mol Imaging 32(2):211–220 Vogel R, Indermuhle A, Reinhardt J, Meier P, Siegrist PT, Namdar M, Kaufmann PA, Seiler C (2005) The quantification of absolute myocardial perfusion in humans by contrast echocardiography: algorithm and validation. J Am Coll Cardiol 45(5): 754–762 Wagner S, Breyholz HJ, Faust A, Holtke C, Levkau B, Schober O, Schafers M, Kopka K (2006) Molecular imaging of matrix metalloproteinases in vivo using small molecule inhibitors for SPECT and PET. Curr Med Chem 13(23):2819–2838 Waldeck J, Hager F, Holtke C, Lanckohr C, von Wallbrunn A, Torsello G, Heindel W, Theilmeier G, Schafers M, Bremer C (2008) Fluorescence reflectance imaging of macrophagerich atherosclerotic plaques using an alphavbeta3 integrintargeted fluorochrome. J Nucl Med 49(11):1845–1851 Wehrl HF, Judenhofer MS, Wiehr S, Pichler BJ (2009) Preclinical PET/MR: technological advances and new perspectives in biomedical research. Eur J Nucl Med Mol Imaging 36(Suppl 1):S56–S68 Weissleder R, Lee AS, Khaw BA, Shen T, Brady TJ (1992) Antimyosin-labeled monocrystalline iron oxide allows detection of myocardial infarct: MR antibody imaging. Radiology 182(2):381–385 Welch MJ, Lewis JS, Kim J, Sharp TL, Dence CS, Gropler RJ, Herrero P (2006) Assessment of myocardial metabolism in diabetic rats using small-animal PET: a feasibility study. J Nucl Med 47(4):689–697 Wiesmann F, Ruff J, Hiller KH, Rommel E, Haase A, Neubauer S (2000) Developmental changes of cardiac function and mass assessed with MRI in neonatal, juvenile, and adult

471 mice. Am J Physiol Heart Circ Physiol 278(2): H652–H657 Wiesmann F, Frydrychowicz A, Rautenberg J, Illinger R, Rommel E, Haase A, Neubauer S (2002) Analysis of right ventricular function in healthy mice and a murine model of heart failure by in vivo MRI. Am J Physiol Heart Circ Physiol 283(3):H1065–H1071 Wu JC, Inubushi M, Sundaresan G, Schelbert HR, Gambhir SS (2002) Positron emission tomography imaging of cardiac reporter gene expression in living rats. Circulation 106(2): 180–183 Yagle KJ, Eary JF, Tait JF, Grierson JR, Link JM, Lewellen B, Gibson DF, Krohn KA (2005) Evaluation of 18F-annexin V as a PET imaging agent in an animal model of apoptosis. J Nucl Med 46(4):658–666 Yu Q, Leatherbury L, Tian X, Lo CW (2008) Cardiovascular assessment of fetal mice by in utero echocardiography. Ultrasound Med Biol 34(5):741–752 Zhang SH, Reddick RL, Piedrahita JA, Maeda N (1992) Spontaneous hypercholesterolemia and arterial lesions in mice lacking apolipoprotein E. Science 258(5081):468–471 Zhang S, Picard MH, Vasile E, Zhu Y, Raffai RL, Weisgraber KH, Krieger M (2005) Diet-induced occlusive coronary atherosclerosis, myocardial infarction, cardiac dysfunction, and premature death in scavenger receptor class B type I-deficient, hypomorphic apolipoprotein ER61 mice. Circulation 111(25): 3457–3464 Zhang Y, Ruel M, Beanlands RS, deKemp RA, Suuronen EJ, DaSilva JN (2008) Tracking stem cell therapy in the myocardium: applications of positron emission tomography. Curr Pharm Des 14(36):3835–3853

Imaging in Neurology Research I: Neurooncology

32

Yannic Waerzeggers, Parisa Monfared, Alexandra Winkeler, Thomas Viel, and Andreas H. Jacobs

Contents

32.1 Introduction

32.1 Introduction............................................................. 473

Over the last decades, the development of animal models of neurological disorders has progressed rapidly. Many in vitro and in vivo models of brain tumours have been developed and have proven to be important tools for our understanding of human cancer. Progression in neurosciences, especially in neurooncology, also increasingly demanded improvements in non-invasive imaging methodologies to provide non-invasive measures of a broad range of tumour-relevant parameters, both at the cellular and the molecular level. In recent years there has been a rapid increase in the variety of ways to non-invasively monitor animal tumour models. There has not only been a refinement in specialized hardware dedicated to small animal imaging that overcomes the limitations of spatial resolution and sensitivity associated with the use of clinical scanners for imaging small-sized laboratory animals (dedicated small animal computed tomography (CT) (Paulus et al. 2000), positron emission or single photon emission computed tomography (PET/SPECT) and magnetic resonance imaging (MRI) (Pichler et al. 2008) scanners), but there has also been a further development of new imaging techniques for non-invasive imaging on the cellular and sub-cellular level (optical imaging (OI), confocal microscopy, two-photon microscopy, fibred fluorescence microscopy, Raman spectroscopy and photoacoustic tomography (Pierce et al. 2008; Wang 2008)). Furthermore, the application of improved imaging probes for selective accumulation in tumours or for activation by tumour-specific molecules and of new imaging methodologies such as detection of reporter transgene expression in vivo had great impact on versatility, sensitivity and specificity of tumour imaging in living subjects. Moreover, as repetitive measurements

32.2 Brain Tumour Models............................................ 474 32.3 Imaging Through Transgene-Based Approaches.............................................................. 474 32.4 Imaging Tumour Growth and Metabolism.......... 476 32.5 Imaging Response to Therapy............................... 479 32.5.1 Apoptosis Imaging.................................................... 479 32.5.2 Angiogenesis Imaging.............................................. 481 32.5.3 Imaging Growth Factors........................................... 481 32.5.4 Imaging Signal Transduction Pathways.................... 484 32.6 Imaging Gene and Cell-Based Therapy................ 485 32.7 Imaging Tumour Chemo- and/or Radiosensitivity....................................................... 493 32.8 Conclusion............................................................... 495 References............................................................................ 496

Y. Waerzeggers, P. Monfared, A. Winkeler, T. Viel Laboratory for Gene Therapy and Molecular Imaging at the Max Planck Institute for Neurological Research with KlausJoachim-Zülch Laboratories of the Max Planck Society, University of Cologne, Gleuelerstrasse 50, 50931 Cologne, Germany A.H. Jacobs (*) Laboratory for Gene Therapy and Molecular Imaging at the Max Planck Institute for Neurological Research with KlausJoachim-Zülch Laboratories of the Max Planck Society, University of Cologne, Gleuelerstrasse 50, 50931 Cologne, Germany and European Institute for Molecular Imaging (EIMI), Westfälische Wilhelms University of Münster (WWU), Münster, Germany e-mail: [email protected]; [email protected]

F. Kiessling and B.J. Pichler (eds.), Small Animal Imaging, DOI: 10.1007/978-3-642-12945-2_32, © Springer-Verlag Berlin Heidelberg 2011

473

474

can be performed in the same animal over time, a dynamic picture of the progressive changes of biological parameters under investigation can be generated and statistical significance can be achieved using far less animals than with conventional laboratory experiments. These developments profoundly influenced our basic understanding of in vivo tumour biology and comprise a strong tool which should be used to test the in vivo efficacy of new therapeutic intervention strategies (e.g. molecular-targeted, gene and cell-based therapies) and to characterize factors that influence chemo- or radioresistance. This chapter will give an overview of the recent developments in the application of small animal imaging in neurooncology. Particular emphasis has been placed on the use of MRI, PET/SPECT and OI, since these approaches had the most significant impact on small animal cancer research in recent years (Table 32.1).

32.2 Brain Tumour Models Many in vitro and in vivo models of brain tumours have been developed. The ideal animal tumour models should be able to accurately recapitulate all aspects of human tumour physiology such as angiogenesis, tumour-stromal and tumour-host interaction and tumour environment and should also address the various factors that underlie the molecular and genetic lesions associated with tumour heterogeneity. The study of brain tumour models has increased our understanding of brain tumour initiation, formation, progression and metastasis and provides an excellent experimental system to discover novel therapeutic targets and test various therapeutic agents (Fomchenko and Holland 2006). Many tumour-derived cell lines have been established and they are often employed for screening of novel drugs in cell culture or xenograft experiments because of their ready availability and ease of use. Xenograft models, induced either by subcutaneous or by orthotopic (into native tumour sites) injection of primary tumour cells or tumour cell lines, represent the most frequent used in vivo cancer modelling system. However, both cell culture and xenograft model systems lack the stepwise genetic alterations thought to occur during tumour progression and do not recapitulate the genetic and cellular heterogeneity of primary tumours and the complex tumour-stroma

Y. Waerzeggers et al.

interaction. In contrast, genetically engineered mouse (GEM) models recapitulate more accurately the causal genetic events and subsequent molecular evolution in situ and give rise to tumour-stroma interactions resembling those of the native tumours and also harbour cellular subpopulations like cancer stem cells thought to be of central importance to the development, maintenance and drug resistance of brain cancer (Huse and Holland 2009). In recent years, several such mouse models have been developed using combinations of known glioma-associated genetic alterations, and for in depth information, several excellent reviews exist (Fomchenko and Holland 2006; Huse and Holland 2009). However, it should be kept in mind that up till now no true “humanized” animal tumour model has been developed and that, consequently, interventional results obtained in mouse models not necessarily will be translated to 100% in the human situation. Most of the cell lines used in xenograft cancer models are transgenic cell lines that exhibit overexpression, depression or complete inactivation of a gene of interest to provide insights in the molecular and functional impact of this specific gene on tumour pathophysiology. Moreover, in addition to this gene of interest, often a second reporter gene is introduced and allows for non-invasive phenotyping of the effects of the induced molecular alterations by molecular imaging techniques.

32.3 Imaging Through Transgene-Based Approaches Many recent advances associated with imaging tumours in small animal have arisen from the application of reporter transgenes. Reporter genes have been developed for optical, nuclear, and to a lesser extent, also for MR imaging modalities, and the choice of a reporter gene is dependent on the question to be addressed, the available imaging modality, the need for translationability to larger animals or humans and the desired throughput of the analyses (Gross and Piwnica-Worms 2006). Reporter genes can be used to visualize the levels of expression of particular exogenous and endogenous genes and several intracellular biologic phenomena such as specific signal transduction pathways, nuclear receptor activities, protein–protein interactions and

50 mm

50 mm

1–2 mm

1–2 mm

10–100 mm

Several mm

1 mm

CT (X-rays)

US (high-frequency sound)

PET (high-energy g-rays)

SPECT (low-energy g-rays)

MRI (radio-waves)

BLI (visible light)

FI (visible and NIR light)

<10 cm

Centimetres

No limit

No limit

No limit

Centimetres

No limit

Depth penetration

Seconds– minutes

Minutes

Minutes–hours

Minutes–hours

Minutes–hours

Minutes

Minutes

Time scale

P,M

M

A,P,M

P,M

P,M

A,P

A,P

Primary target interrogated

€

€

€€€

€€

€€€

€

€€

Cost

Yes

Yes

Yes

No

In development

Sensitivity 10-100 times lower than PET

High spatial resolution Low sensitivity, long acquisition and image and soft tissue processing times contrast, functional information, versatile Light emission prone to attenuation and scatter with increased tissue depth Excitation and emission light <600 nm prone to attenuation with increased tissue depth, autofluorescence Potential for multiplex imaging, high throughput, transgenebased approach confers versatility

High sensitivity, high throughput, transgenebased approach confers versatility

Radioisotopes have longer half-lives than those used in PET, multiple probes can be detected simultaneously

High sensitivity, quantitative, variety of available probes

Cyclotron needed, short-lived radioisotopes, low resolution, single-to-noise ratio

Yes

Poor soft tissue contrast

Yes

Clinical use

Main disadvantages

Difficult to image through bone or lungs

Primarily vascular and soft tissue imaging

Primarily lung and bone imaging of tumours metastases

Main advantages

The main advantages and disadvantages associated with each system are listed. Primary targets of the respective imaging system: A anatomical; P physiological; M molecular

Spatial resolution

Technique (and used energy window)

Table 32.1 Overview of the main non-invasive imaging systems commonly used to image tumours in rodents

32 Imaging in Neurology Research I: Neurooncology 475

476

post-translational protein modifications (Kang and Chung 2008). Regardless of the examined process, any strategy for imaging genetically encoded reporters is comprised of three major components: (i) a reporter gene that generates an imageable signal, (ii) a regulatory element governing the activity of the reporter gene and therefore generating contrast, and (iii) a detection device able to non-invasively sense and quantify the signal produced by the reporter gene within the intact cell or organism (Gross and Piwnica-Worms 2005). Genetically encoded reporters can produce signal (i) intrinsically by the reporter (e.g. fluorescent proteins or quantum dots), (ii) through enzymatic activation of an inactive substrate by the reporter (e.g. firefly luciferase that catalyses the light-producing reaction from the substrate D-luciferin in the presence of O2, Mg2+ and ATP; b-galactosidase that enzymatically cleaves an inactive paramagnetic chelate or quenched fluorophore into an active contrast agent), (iii) by enzymatic modification of an active (e.g. radiolabelled) substrate resulting in selective trapping in reporter cells (e.g. selective retention of [18F]-labelled 9-[4-fluoro-3-(hydroxymethyl) butyl]guanine, [18F]FHBG, or [124I]-labelled 2’-fluoro2’-deoxy-1-ß-d-arabinofuranosyl-5-iodouracil, [124I]FIAU, by the herpes simplex virus type 1 thymidine kinase enzyme HSV-1-TK), or (i.v.) by direct binding or import of an active substrate by intracellular or extracellular receptors and/or transporters (e.g. binding of radiolabelled somatostatin to somatostatin receptor type 2 [SSTR2]-expressing cells, accumulation of radioactive iodine in cells expressing the human thyroid sodium/iodide symporter (hNIS) gene or accumulation of superparamagnetic transferrin probe or iron in cells expressing an engineered human transferrin receptor or the metalloprotein ferritin, respectively) (Gross and Piwnica-Worms 2005; Gilad et al. 2007). A common feature of all reporter constructs is the cDNA expression cassette containing the reporter transgenes of interest. The versatility of reporter gene imaging results in part from the flexibility to tailor the expression cassette to an individual need. Regulation of the reporter genes can be achieved at the transcriptional level (e.g. a constitutively active or inducible promoter or an upstream cis-regulatory element) or at the post-transcriptional/translational level (e.g. a regulatory polypeptide sequence fused in frame with the reporter). Conditional elements regulating transgene expression can restrict their expression to a certain

Y. Waerzeggers et al.

time frame, a desired organ or a certain biological event (e.g. tetracycline and mifepristone responsive elements, Cre recombinase/lox, and Gal4-dependent transactivation) (Waerzeggers et al. 2009a). One major disadvantage of reporter gene imaging is the need to transduce or transfect the target tissue, which hampers its widespread clinical use. The cDNA containing vector can be used directly to transfect the cells, mostly facilitated by cationic lipid-based transfection agents or electroporation, or the vectors can be wrapped into a recombinant viral vector which is then used to transduce the cells. Several currently available vector types can be used (e.g. retrovirus, adenovirus, adeno-associated virus, lentivirus, herpes simplex virus). Reporter transgenes can be introduced ex vivo into xenografted or allografted cells, directly delivered to tissues by somatic gene transfer or cloned into the genomes of transgenic animals.

32.4 Imaging Tumour Growth and Metabolism Tumour growth and survival is dependent on the maintenance of cell energy metabolism and on protein and DNA synthesis and is governed by the adequate delivery of necessary metabolites (glucose, amino acids and nucleosides) through development of a tumour vascular network. Conventional PET and MR techniques have since long been the imaging techniques of choice for diagnosis and staging of patient with brain tumours by monitoring the above-mentioned mechanisms. The increased cell metabolism in tumour cells results in a strongly increased glucose uptake and metabolism (glycolysis) compared to normal tissue, and in the past PET imaging with [18F]-labelled glucose ([18F]FDG) has been used to diagnose or to assist in the evaluation of malignancy in patients with known or suspected abnormalities (Oriuchi et al. 2006). The molecular targets of FDG are glucose transporters, such as the insulin-dependent GLUT1 transmembrane transporter and hexokinase, both of which are overexpressed in many cancers. Once phosphorylated, FDG-6 phosphate is retained in cells in proportion to their glycolytic activity. [18F]FDG has been widely used for imaging glucose transporter expression and hexokinase activity and is commonly used for brain tumour diagnosis (Herholz et al. 2007). In clinical neurooncology,

32 Imaging in Neurology Research I: Neurooncology

[18F]FDG has been used to detect the metabolic differences between normal brain tissue, low-grade and high-grade gliomas, and radionecrosis. Moreover, increased intratumoural glucose consumption correlates with tumour grade, cell density, biological aggressiveness and patient survival in both primary and recurrent gliomas. However, due to the relatively high glucose metabolism in normal gray matter, clear tumour location and delineation with [18F]FDG PET is difficult and limitations related to its non-specific uptake in proliferating non-cancerous cells or tissues (such as inflammatory cells or granulation tissue) and to the high glucose metabolism of normal brain tissue reduce its detection sensitivity and compromise the ability of [18F]FDG PET to monitor the therapeutic response of intracranial lesions (Oriuchi et al. 2006; Herholz et al. 2007). Therefore, other PET and SPECT probes have been developed for tumour identification via a variety of different tumour-specific mechanisms. Because of the uncontrolled and accelerated growth of cancer, the process of protein synthesis in tumours is increased. As a consequence, the demand for amino acids, the building blocks of proteins, is increased. Amino acids that have been used for cancer imaging include [18F] or [11C]methionine (MET), [18F]fluoroethyl tyrosine (FET), and [18F]fluoromethyltyrosine (FMT). In neurooncology, radiolabelled methionine and tyrosine compounds have been shown to be more specific tracers in tumour detection, delineation and staging owing to their relatively low uptake in normal brain (Herholz et al. 2007; Singhal et al. 2007). As a high rate of cellular proliferation is a key feature of malignant tumours, in recent years much attention has been given to the search for cellular proliferation markers, the uptake of which would accurately reflect DNA synthesis. There are two main pathways involved in DNA synthesis: (i) the exogenous or “salvage” pathway, which utilizes DNA precursors, nucleic acids, from outside the cell that are phosphorylated by thymidine kinase 1 (TK1); and (ii) the endogenous pathway, in which intracellular molecules, such as uridine monophosphate, enter DNA synthesis after phosphorylation by thymidylate synthase (Sherley and Kelly 1988). A variety of nucleic acids have been radiolabelled as PET tracers for cellular proliferation such as 11 C-labelled thymidine and 18F-labelled 3¢-deoxy3¢-fluoro-thymidine (FLT) (Herholz et al. 2007). After its intravenous administration, [18F]FLT is distributed extracellularly and is subsequently transported into the

477

cytosol, where it is phosphorylated by TK1, a cytosolic enzyme that is expressed during the S phase of the cell cycle (Sherley and Kelly 1988). In its phosphorylated form, [18F]FLT is metabolically “trapped” in the cell, with essentially no incorporation into DNA. It has been demonstrated in many types of cancer, including glioblastoma, that [18F]FLT uptake in vivo is a measure of tumour proliferation activity and that there exists a significant correlation between [18F]FLT uptake and the in vitro proliferation marker Ki-67 (Ullrich et al. 2008). Bradbury et al. recently investigated whether [18F]FLT can be used to detect and assess proliferative activity in GEM high-grade glioma models (Bradbury et al. 2008). The authors characterized the intracranial gliomas by both static and dynamic small animal PET imaging of [18F]FLT uptake and developed and implemented a compartmental modelling approach, using a 3-compartment, 4-parameter model that permits the discrimination of tracer transport and delivery from tumour proliferation. The ability to distinguish between the delivery and retention components of radiotracer uptake may become particularly important in the evaluation of treatment responses. Without information on the relative contributions of the delivery and metabolic trapping of [18F]FLT to overall uptake, incorrect interpretations regarding tumour proliferation status can result, particularly under conditions in which delivery largely dictates overall tumour uptake; for example, reduced delivery of [18F]FLT to hypoperfused tumours may result in a spuriously low estimate of proliferation status. Conventional MRI techniques, such as T1- and T2-weighted imaging, contrast-enhanced T1-weighted imaging, dynamic contrast-enhanced (DCE) imaging and diffusion-weighted imaging (DWI), provide information on tumour localisation and extent, local blood– brain barrier (BBB) damage and brain invasiveness, regional blood flow and blood volume and tumour cellularity which, in the clinical setting, are known to be associated with glioma grade and prognosis (Herholz et al. 2007). Also, secondary changes associated with tumour growth, such as oedema, brain shift or hydrocephalus, can be determined on two- or three-dimensional MR images. Recently, McConville et al. demonstrated that MRI can be used to predict tumour grade and survival in a GEM model of glioma (McConville et al. 2007). In the Ntv-a model, approximately 100% of mice spontaneously develop gliomas by 3 weeks of age, with 30% of these tumours

478

displaying high-grade histologic features. T2-weighted and T1-weighted, gadolinium-enhanced MRI could distinguish between high- and low-grade tumours on the basis of their growth rate and contrast enhancement. However, care should be taken when interpreting contrast leakage as an indicator of BBB breakdown due to malignant degeneration as this sign can be absent in diffuse infiltrative tumour regions, non-specific or induced by therapeutic intervention (Herholz et al. 2007). Especially the treatment with antiangiogenic compounds can restore the BBB in angiogenic regions without concomitant tumour regression and underlines the need for alternatives to gadoliniumbased contrast-enhanced MRI to detect progressing tumour portions that are not associated with angiogenesis and thus BBB disruption, but are based on an angiogenesis-independent tumour growth, for instance, via co-option of pre-existent vasculature (Leenders et al. 2004). Although these infiltrative lesions cannot be detected with gadolinium-based contrast-enhanced MRI, the relatively low vascular volume in these tumours, as compared to the surrounding tissue, can be exploited to detect these lesions using blood pool contrast agents such as ultrasmall particles of iron oxide (USPIO) and to evaluate the response to antiangiogenic therapy (Claes et al. 2008). Furthermore, it has been shown that the rim of contrast enhancement at the margins of brain tumours detectable on MRI 24 h after intravenous injection of USPIOs correlates to the presence of iron-loaded macrophages and microglia and can be used for tumour delineation and evaluation of tumour aggressiveness as the level of contrast enhancement seems to correlate with tumour proliferation and tumour growth (Kremer et al. 2007). When conjugating such an MR-detectable iron oxide nanoparticle to an optically detectable NIRF fluorochrome, this multimodality approach permits the correlation of preoperative MR images to intraoperative optical images to guide accurate tumour resection (Kircher et al. 2003). Another technique to follow indirectly in vivo tumour angiogenesis as a necessary component of tumour expansion, invasion and possible metastasis is DCE MRI that enables the longitudinal investigation of changes in tumour vascular permeability, vascular density and vessel morphology. Recently, Veeravagu et al. demonstrated in an orthotopic murine (GL26) glioblastoma model that in vivo changes in blood vessel permeability, as shown by DCE MRI, correlate with histologic quantification of vascular density and vessel

Y. Waerzeggers et al.

calibre as well as with the molecular expression of angiogenic factors such as vascular endothelial growth factor (VEGF) and angiopoietins (ANG-1 and ANG-2) (Veeravagu et al. 2008). Finally, DWI can be used in cancer imaging to assess tumour cellularity and infiltration and to monitor response to therapy. The apparent diffusion coefficient (ADC) of water has been found to increase in the early phase of anticancer therapies. Treatment-induced killing of tumour cells in a 9L brain glioma model, leading to a decrease of cell density in tumour tissue, results in an increase of the ADC that may be explained by destruction of tumour cells, widening of the extracellular space and a consequent increase in extracellular, relatively mobile water (Chenevert et al. 1997). Moreover, the treatmentinduced increase in ADC values occurs as early as 4–5 days after therapy onset and precedes volumetric tumour changes (McConville et al. 2007). By depicting patterns of anisotropic diffusion, diffusion tensor imaging (DTI) and fibre tractography permit the visualization of the structural integrity and connectivity of neuronal fibres surrounding brain tumours and this approach can be used to discriminate between the compressive growth of C6 gliomas or the infiltrative growth of F98 gliomas and to evaluate the neuroprotective effect of a free radical trapping compound against invasive glioma growth (Asanuma et al. 2008). Although metabolic changes associated with intracranial tumour growth can be monitored with MRS and can be used to evaluate cell damage as early as 2–4 days after ganciclovir (GC) treatment in gene therapy of experimental gliomas (Hakumaki et al. 1999), most information on the metabolic state of intracranial gliomas can be gained with PET imaging, as stated above. Another, rapidly expanding, approach to image tumour load in intact animals is optical imaging, especially bioluminescence imaging (BLI). In general, BLI of tumour load is based on ATP-dependent conversion of luciferin into the light-emitting product oxyluciferin by luciferase in cancer cells transfected with the luciferase gene, usually that of Firefly (Fluc). The ATP, which is required for the enzyme reaction, is endogenously present in viable cancer cells and thus the bioluminescence signal is only derived from living cancer cells and not from necrotic areas in the tumour (Klerk et al. 2007). Several validation studies that correlated BLI signals with the amount of inoculated tumour cells, with tumour volume measurements using callipers, with MRI volume measurements, with tumour

479

32 Imaging in Neurology Research I: Neurooncology

weight measurements or with fluorescence-activated cell sorting (FACS) analysis of cancer cells in affected tissues indicate that BLI is useful to determine tumour load in the course of time, with each animal serving as its reference. However, BLI is less suited for the determination of absolute tumour mass because of quenching of bioluminescence by tissue components and for the determination of exact tumour location because of its limited spatial resolution (Klerk et al. 2007). Nevertheless, BLI is particularly useful for longitudinal follow-up of tumour growth during therapy intervention studies. Rehemtulla et al. (2000) investigated the ability of BLI to non-invasively quantitate the growth and therapeutic-induced cell kill of orthotopic rat brain tumours derived from 9L gliosarcoma cells genetically engineered to stably express Fluc. Quanti tative comparison of tumour cell death determined from serial MRI volume measurements and BLI photon counts following 1,3-bis(2-chloroethyl)1-nitrosourea (BCNU) treatment revealed a statistically highly significant correlation between both modalities, providing direct validation of BLI imaging as a powerful and quantitative tool for high-throughput assessment of antineoplastic therapies. In a similar approach, Szentirmai et al. evaluated the antiangiogenic treatment effect of adenoviruses encoding antiangiogenic soluble VEGF receptors in nude mice bearing U87MG glioma cells genetically engineered to express the firefly luciferase gene (Szentirmai et al. 2006).

after treatment only T2/FLAIR correlated significantly with histology. Tumour shrinkage is usually a late response and dependent on cell apoptosis and clearance of non-viable cells. By tracking changes in the molecular motion of water (designated as ADC), diffusion-weighted MRI gives insights into changes in tumour cellularity and membrane permeability that can occur well before changes in tumour volume or tumour contrast enhancement (McConville et al. 2007) and provides a sensitive means to assess dynamic therapy-induced responses, including the appearance of changes in drug sensitivity (Lee et al. 2006). Also, [18F]FDG PET has proven to be a relative early response marker in a variety of tumours (breast cancer, ovarian cancer, rectal cancer, oesophageal cancer, lung cancer), with significant detectable changes in [18F]FDG uptake after a few cycles or a few weeks of chemotherapy treatment (Pantaleo et al. 2008). However, not all tumours are rapidly dividing and glycolytic and many emerging targeted therapies are cytostatic or do not influence tumour glucose metabolism. As a result, other more specific imaging approaches are necessary that are also capable of evaluating directly the effects of novel targeted therapies and not only just secondary, relatively non-specific changes. Examples of such specific imaging approaches are the direct non-invasive evaluation of apoptosis, angiogenesis or growth factors and related signal transduction pathways.

32.5.1 Apoptosis Imaging 32.5 Imaging Response to Therapy Conventional PET and MR techniques have been used for a long time to evaluate response to therapy. Most commonly used is standard T1- and T2-weighted MR imaging that relies on the detection of morphological (volume) changes related to therapy. In a rat central nervous system lymphoma model, Jahnke et al. recently compared T2/fluid-attenuated inversion recovery (FLAIR) and gadolinium contrast-enhanced T1 MRI sequences for their ability to monitor response to chemotherapy by evaluating image-derived changes in tumour volume before and 1 week after treatment (Jahnke et al. 2009). In untreated control animals, tumour histological volumes correlated well with T2/ FLAIR and contrast-enhanced T1 images; however,

Apoptosis is an essential component of normal human growth and development, immunoregulation and tissue homeostasis by providing a means for elimination of redundant, damaged or diseased cells. Apoptosis may be initiated in response to cellular stresses and DNA damaging events, such as growth factor deprivation, hypoxia, heat, cold or chemical injury (Coppola et al. 2008). Furthermore, several genetic alterations associated with cancer, such as bcl-2 activation, MDM2 overexpression or p53 mutations, dysregulate apoptosis and facilitate neoplastic transformation. Apoptotic cell death can be initiated through an extrinsic pathway involving activation of cell surface death receptors or by an intrinsic pathway via the mitochondria (Kelloff et al. 2005). Both pathways lead to activation of initiator (e.g. caspase-1,-8,-10) and effector caspases

480

(e.g. caspase-3,-6,-7) that trigger a proteolytic cascade resulting in fragmentation of intracellular components and appearance of apoptotic morphology. One of the earliest effects of caspase activation is the disruption of the translocase system that normally maintains phosphatidylserine (PS) on the interior of the cell membrane. This results in the redistribution of PS to the outer membrane leaflet, where it serves as a signal to phagocytic cells to engulf and digest the membraneenclosed apoptotic cells. Because many current cancer therapies promote cell death by reinstituting the apoptotic pathways, the ability to detect apoptosis by direct imaging of caspase activity or PS expression would aid significantly in the development and evaluation of such apoptotic pathway modulators (Blankenberg 2008). Caspase peptide substrates containing either a nuclear (a 18F-labelled caspase-inhibiting analogue) or a bioluminescence label or a far-red or near-infrared (NIR) optical fluorochrome have been developed to detect apoptosis (Blankenberg 2008). Using firefly luciferase-based BLI in nude mice, Laxman et al. developed a caspase-cleavable reporter probe able to detect tumour apoptosis following chemotherapy (Laxman et al. 2002). In this study, a recombinant luciferase reporter molecule was developed in which luciferase activity was silenced via steric hindrance by the oestrogen receptor regulatory domain (ER). Inclusion of a protease cleavage site for caspase-3 (DEVD) between the luciferase domain and the silencing domain allowed for protease-mediated activation of the reporter molecule. In vivo studies using a human glioma cell line stably expressing the reporter molecule revealed that caspase-3 activity by activation on tumour necrosis factor-related apoptosis-inducing ligand (TRAIL) treatment could be imaged noninvasively by using BLI. Furthermore, the temporal pattern of caspase-3 activity in response to TRAIL treatment has been followed: sustained TRAIL administration only results in a transient induction of apoptosis, whereas concomitant treatment with 5-fluorouracil and TRAIL produces a synergistic and prolonged enhancement in apoptosis activation and a prolonged growth delay (Lee et al. 2007). In an attempt to develop an improved apoptosis reporter with increased signal to noise, the same group recently used the split firefly reporter strategy (Coppola et al. 2008). The developed reporter, ANLucBCLuc, constitutes a fusion of small interacting peptides (peptide A and peptide B) with the NLuc and CLuc fragments of luciferase and with the

Y. Waerzeggers et al.

caspase-3 cleavage site DEVD between pepANLuc (ANLuc) and pepBCLuc (BCLuc). During apoptosis, caspase-3 cleaves the reporter, enabling separation of ANLuc from BCLuc. A high-affinity interaction between peptide A and peptide B then restores luciferase activity by NLuc and CLuc complementation. Treatment of live cells and mice carrying D54 glioma tumour xenografts with chemotherapeutic agents such as temozolomide (an alkylating agent that induces DNA methylation) and perifosine (an alkylphospholipid that exerts its cytotoxic effect by interfering with Akt activation) resulted in a dose-dependent induction of bioluminescence activity, which correlated with activation of caspase-3 as determined by Western blot analysis and caspase-3 staining. Furthermore, treatment of mice-bearing glioma xenografts with combination therapy of temozolomide and radiation resulted in increased bioluminescence activity over individual treatments and increased therapeutic response (tumour shrinkage) due to enhanced apoptosis. Recently, a multimodality caspase-3 sensor has been developed that is composed of a fluorescence reporter (monomeric red fluorescent protein, mRFP1), a PET reporter (HSV-1-sr39 thymidine kinase, tTK) and firefly luciferase (Fluc), each joined by DEVD as peptide linker (Ray et al. 2008). In the fused form, all the three reporter proteins had markedly attenuated activity. However, upon cleavage by activated caspase-3 (induced by staurosporine), a significant gain in mRFP1, FLuc, and TK enzyme activity could be observed by FACS, enzyme-based assays and in vivo BLI and microPET imaging. PS exposure has been the most pursued target for the detection of cell death using molecular imaging methods (Blankenberg 2008; Neves and Brindle 2006). The 40-kDa vesicle-associated protein Annexin V (AnxV) has been the most widely used PS-targeting moiety. AnxV binds PS in a calcium-dependent manner and with high (nanomolar) affinity. Initially, AnxV was coupled to fluorescent dye molecules and used as an apoptosis detection reagent for fluorescence microscopy and flow cytometry. Subsequently, AnxV was coupled to a radionuclide (99mTc) and used to detect apoptosis non-invasively in animals and in the clinic using radionuclide imaging techniques. Recent studies in oncology suggest that a single scan 24–48 h after the start of treatment can identify patients with response after one course of chemotherapy. Other radionuclide derivatives of AnxV have been developed, including

481

32 Imaging in Neurology Research I: Neurooncology

AnxV labelled for SPECT with 123I or for PET labelled with 18F or 124I and 64Cu-labelled streptavidin for PET imaging after pre-targeting of PS with biotinylated annexin V. Photonic imaging methods have also been used to image AnxV labelled with fluorochromes or NIR fluorochromes, and Schellenberger et al. were the first to label AnxV with superparamagnetic iron oxide particles (SPIO) nanoparticles for MR detection (Schellenberger et al. 2003). Later, also Gd-containing AnxV-coated liposomes for positive or bimodal MR contrast have been developed as well as multimodality contrast agents for combined MRI and fluorescent imaging. More recently, studies have focused on the use of a smaller PS-targeting agent, synaptotagmin-I, which also binds PS with nanomolar affinity in a calciumdependent manner. This protein has been labelled with SPIO nanoparticles and Gd3+ chelates for MRI-based detection of PS exposure in vivo after chemotherapy.

32.5.2 Angiogenesis Imaging A variety of genetic anomalies that trigger gliomaassociated angiogenesis, such as overexpression of the growth factors VEGF, epidermal growth factor (EGF), platelet-derived growth factor (PDGF) and their receptors or chronic activation of the hypoxia-inducible transcription factor-1 (HIF-1), have been identified and can be measured directly or indirectly by noninvasive imaging techniques and used as read-outs for targeted tumour treatment. Conventional imaging techniques like MRI and PET focus on the measurement of physiologic parameters, such as blood flow, blood volume, vascular perfusion, permeability and/or structure, and represent the radiographic tools in current clinical trials of antiangiogenic therapy (Waerzeggers et al. 2009b). During the last years, intense research focused on VEGF/VEGFR-targeted molecular imaging and a wide variety of targeting molecules (peptides, proteins, antibodies and nanoparticles) have been labelled with various imaging labels (such as radioisotopes, fluorescent dyes, and microbubbles) for PET, SPECT, optical imaging or contrast-enhanced ultrasound imaging of tumour angiogenesis (Cai and Chen 2008). Hsu et al. (2007) used multimodality (BLI, MRI and PET) molecular imaging to determine the antiangiogenic and antitumour efficacies of a vasculature-targeting

fusion toxin (VEGF(121)/rGel) composed of the VEGF-A isoform VEGF(121) linked with a G(4)S tether to recombinant plant toxin gelonin (rGel) in an orthotopic glioblastoma mouse model. In this study, the level of target expression was monitored before therapy by [64Cu]-1,4,7,10-tetraazacyclododedaneN,N¢,N ,N ¢-tetraacetic acid (DOTA)-VEGF(121)/rGel PET, whereas [18F]FLT scans were obtained before and after treatment to evaluate VEGF(121)/rGel therapeutic efficacy. In VEGF(121)/rGel-treated mice a significant decrease in [18F]FLT uptake and peak BLI tumour signal intensities could be observed as compared to non-treated mice and these results were validated by histologic analysis. Also, expression of cell adhesion molecules, such as integrins, is significantly up-regulated during tumour growth and angiogenesis and avb3 expression has been correlated with tumour aggressiveness. avb3 integrin expression can be measured by targeted radiolabelled, paramagnetic and fluorescent molecules (cyclic arginine–glycine–aspartic acid RGD peptides) and this method can be used to selectively target suicide gene therapy or drug delivery (Cai and Chen 2008). Serganova et al. (2004) developed a dual reporter gene cassette to monitor non-invasively the dynamics and spatial heterogeneity of HIF-1-specific transcriptional activity in tumours and showed that HIF-1mediated activation of TKGFP (thymidine kinase green fluorescent protein) reporter gene expression in hypoxic tumour tissue can be non-invasively and repeatedly visualized in living mice using PET imaging with [ 18F]-2¢-fluoro-2¢-deoxy-1b-d-arabinofuranosyl-5ethyl-uracil (FEAU). Tumour cells transfected with this construct may be used to study how treatment modulates hypoxia and subsequent HIF-1-modulated gene expression.

32.5.3 Imaging Growth Factors The epidermal growth factor receptor (EGFR) is one of the most studied molecules as a target for cancer therapy (Brandes et al. 2008). EGFR is a cell surface receptor with tyrosine kinase (TK) activity that belongs to the c-erb family. Dysregulation of EGFR is associated with several key features of cancer, such as autonomous cell growth, inhibition of apoptosis, angiogenic potential, invasion and metastases (Fig. 32.1), and

482

Y. Waerzeggers et al.

aberrant EGFR pathway activation may be the result of different molecular abnormalities which include gene amplification, receptor mutations, growth factor overexpression or cross-link activation of downstream signalling pathways. EGFR has been shown to be differentially deregulated, overexpressed, mutated or amplified in many types of cancer, including in 40–60% of glioblastoma cases, and has been associated with a poor prognosis, especially when occurring in younger glioblastoma patients (Brandes et al. 2008). Two main categories of EGFR mutations exist: deletion or point mutations in the extracellular domain and somatic mutations in the TK domain (Brandes et al. 2008). The

type III EGFR mutant (EGFRvIII), which lacks the amino acid residues 6–273 in the extracellular domain, is the most common and is observed in 60–70% of EGFR-overexpressing glioblastomas. Since EGFRvIII lacks a large portion of the ligand-binding domain and hence is unable to bind either EGF or other EGFR ligands, it is constitutively active and can initiate downstream signalling because the deletion results in a conformational change that mimics the one induced by ligand binding in wild-type EGFR. These EGFRvIIIpositive tumours are reported to be associated with a worse prognosis and shorter life expectancy. The other category of EGFR mutations involves changes in the

Fig. 32.1 Schematic representation of relevant signal transduction pathways and cell cycle control pathways known to be dysregulated and involved in glioma initiation and growth. GF growth factor; GFR growth factor receptor; RTK receptor tyrosine kinase; EGF epidermal growth factor; VEGF vascular endothelial growth factor; PDGF platelet-derived growth factor and their respective receptors EGFR, VEGFR and PDGR; MAPK mitogen-activated protein kinase; ERK extracellular signal-regulated kinase; MEK MAPK/ERK kinase; PLC phospholipase C; PKC protein kinase

C; PI3K phosphatidylinositol-3-kinase; PTEN phosphatase and tensin homology deleted on chromosome 10; Akt protein kinase B; mTOR mammalian target of rapamycin; HIF-1 hypoxiainducible factor-1; INK4a/ARF inhibitor of kinase 4/alternative reading frame tumour suppressor genes; CDK4 cyclin-dependent kinase 4; MDM2 murine double minute 2 oncogene; p53 protein 53 transcription factor; Rb retinoblastoma tumour suppressor protein; E2F1 transcription factor (activator) from E2F family of transcription factors

32 Imaging in Neurology Research I: Neurooncology

TK domain that can influence the binding of TK inhibitors (TKIs). These mutants exhibit increased activation compared to wild-type EGFR, but at the same time are much more sensitive to inhibition by EGFRTKIs. However, mutations in the tyrosine kinase domain could not be shown in malignant gliomas so far (Brandes et al. 2008). Monoclonal antibodies against EGFR and small molecule inhibitors targeting tyrosine kinases that bind to the intracellular part of the receptor and thereby block receptor phosphorylation are the two major strategies for inhibiting the EGFR pathway and have been introduced in clinical practice for patients with malignant gliomas (Waerzeggers et al. 2009b). However, findings reported in studies evaluating EGFR inhibitors are surprisingly contradictory and no clear predictive role of receptor status on anti-EGFR drug response could be found (Brandes et al. 2008). These results indicate the need for the development and validation of non-invasive imaging techniques targeted to EGFR expression to predict which patients will likely respond to antiEGFR therapy and to monitor patient response to such personalized cancer management. Different imaging approaches have been developed to specifically detect EGFR expression, from OI modalities to SPECT and PET technologies, and two different labelling strategies have been studied: labelling small molecules such as TKIs and labelling monoclonal antibodies (Pantaleo et al. 2009; Gelovani 2008; Cai et al. 2008). A number of 11C, 18F and radioiodine-labelled EGFR kinase inhibitors that bind reversibly or irreversibly to the EGFR kinase ATP-binding pocket have been developed. The radiolabelled irreversible EGFR inhibitors demonstrate the greatest potential for derivatization into effective EGFR kinase imaging agents, as compared to those developed based on reversible inhibitors due to the rapid washout of the latter. Especially PET imaging with (E)-but-2enedioic acid [4-(3[124I]iodoanilino)-quinazolin-6y l ] - a m i d e - ( 3 - m o r p h o l i n - 4 - y l - p r o py l ) - a m i d e (morpholino[124I]IPQA) seems to be very promising as this tracer binds covalently to the ATP-binding site in activated EGFR, but not in inactive EGFR (Pal et al. 2006). In vitro studies demonstrated rapid accumulation and progressive retention post washout of morpholino[123I]IPQA in A431 cells and in U87MG human glioblastoma cells genetically modified to express EGFRvIII (U87DEGFR), but not in the wild-type U87MG cells (that have low kinase

483

activity) under serum-starved conditions. Also, the in vivo studies demonstrated a significant difference in the accumulation of the radiotracer in A431 carcinoma xenografts that reflects the EGFR expression and EGFR activity in comparison to K562 control xenografts. The authors concluded that PET imaging with morpholino-124I-IPQA should allow for identification of tumours with high EGFR signalling activity, such as brain tumours expressing the EGFRvIII mutant and NSCLC expressing gain-of-function EGFR mutants expected to have an increased sensitivity to EGFR-TKIs (Cai et al. 2008). However, it should be pointed out that PET imaging was not carried out in U87DEGFR human glioma xenograft mice. Also, pegylation of known irreversible compounds with various lengths of polyethyleneglycol (PEG) chains seems very promising due to increased stability and solubility of the radiolabelled tracers (Dissoki et al. 2007); however, in vivo imaging data are still lacking. Labelling monoclonal antibodies with radioisotopes represents another strategy for the in vivo visualisation of EGFR tumour expression, but it seems more complicated. In comparison to labelling TKIs, few studies have been conducted with the attempt of synthesizing monoclonal antibody-based probes (Pantaleo et al. 2009; Cai et al. 2008). As the natural ligand of EGFR, EGF has been labelled with 76Br or 68 Ga for direct targeted PET imaging or with 99mTc (Cornelissen et al. 2005) or 131I for tumour detection and evaluation of therapy response with SPECT imaging. Also, monoclonal antibodies (mAbs) against EGFR, used for anti-EGFR therapy such as cetuximab, have been labelled with positron emitters, e.g. 64Cu, or gamma-emitters, e.g. 99mTc and 111 In, with good tumour contrast and good correlation between tumour uptake and tumour EGFR expression level as measured by western blot. However, MRI with a ferromagnetic anti-EGFR mAb in athymic rats bearing EGFR-positive tumours revealed that the targeted contrast agent only gave modest EGFR-specific MR contrast in vivo (Suwa et al. 1998). Also, optical imaging studies on EGFR expression have been carried out with fluorescently labelled EGF or anti-EGFR mAbs (Diagaradjane et al. 2008). In an attempt to follow the role of growth factor receptor expression in glioma progression and to image therapeutic efficacy of targeted therapies, Arwert et al. genetically engineered glioma cells to

484

visualize the dynamics of EGFR and targeted therapies in real time in vivo (Arwert et al. 2007). The authors created LV reporter constructs containing fusions between EGFR or EGFRvIII and the bioluminescent-fluorescent fusion protein marker GFPRluc and used them to transduce Gli36 glioma cells expressing Fluc-DsRed2. In vivo imaging of mice bearing EGFR-GFP-Rluc/Fluc-DsRed2 or EGFRvIIIGFP-Rluc/Fluc-DsRed2 gliomas into the right frontal lobe revealed that there is a direct correlation between EGFR expression (Rluc activity) and glioma cell proliferation (Fluc activity) in the initial stages of glioma progression. Using the same tumour model, the authors also could show that EGFR-targeted therapies (short hairpin RNAs targeting the two variants of EGFR or a clinically used monoclonal EGFRantibody, cetuximab) result in a considerable reduction in glioma cell proliferation in culture and glioma burden in vivo that can be monitored in real time. Similar results have been obtained with HSV-1 amplicon vector-mediated shRNA targeting EGFR and evaluation by BLI (Saydam et al. 2005). Another tyrosine kinase receptor that is implicated in the malignant progression of glioblastomas is c-Met, the receptor for the hepatocyte growth factor (HGF). Expression of HGF has been shown to be associated with poor prognosis of malignant gliomas and both HGF and c-Met have emerged as key determinants of brain tumour growth and angiogenesis (Arrieta et al. 2002; Abounader and Laterra 2005). c-Met expression can be non-invasively imaged with MRI by specifically tagging the receptor with an anti-c-Met antibody linked to biotinylated Gd (gadolinium)-DTPA(diethylene triamine penta acetic acid)-albumin (anti-c-Met-Gd-DTPA-albumin) (Towner et al. 2008).

32.5.4 Imaging Signal Transduction Pathways Recent advances in preclinical research have led to the development of several targeted reporters to interrogate the activity of other receptors or elements in intracellular signalling cascades related to cell growth and proliferation. Direct non-invasive imaging of therapy-induced changes in the cell cycle regulatory pathways controlled by p53 or E2F offers further

Y. Waerzeggers et al.

strategies to develop and evaluate targeted therapies directed to specific cancer-related molecular alterations (Fig. 32.1). Our group recently constructed a Cis-E2F/ LucIRESTKGFP reporter system to non-invasively assess E2F-1-dependent transcriptional regulation in culture and in vivo (Monfared et al. 2008). In this reporter system the reporter genes firefly luciferase, HSV-1 thymidine kinase and green fluorescent protein are placed under control of an artificial cis-acting E2F-specific enhancer element. We could show that this reporter system is sensitive to non-invasively monitor various changes in cellular E2F-1 levels following DNA damage-induced up-regulation of E2F-1 activity (Fig. 32.2). Following retroviral transduction of U87 glioma cells, exposure to BCNU leads to increased E2F-1 expression levels in a dose- and time-dependent manner, which can be quantified by in vivo imaging in established xenografts. Activation of the transcription factor E2F-1 via alteration of the p16-cyclinD-Rb pathway is one of the key molecular events in the development of gliomas and represents a promising anticancer target. In a similar model system, Doubrovin et al. used the HSV-1 thymidine kinase reporter gene to image p53-dependent gene expression in an animal tumour xenograft model system, following DNA damage induced by the alkylating chemotherapeutic agent carmustine (Doubrovin et al. 2001). Damage-induced up-regulation of p53 transcriptional activity, measured by elevated transgene HSV-1-thymidine kinase enzyme activity in vivo, correlated with the expression of p53-dependent downstream genes. Another major target for anticancer drug development is the serine/threonine kinase Akt that mediates mitogenic and anti-apoptotic responses resulting from activation of multiple signalling cascades. Zhang et al. (2007) constructed a bioluminescence Akt reporter based on the split-luciferase technology. By the use of this reporter, the authors could monitor quantitatively and dynamically Akt activity (Akt phosphorylation) in cultured cells and tumour xenografts (human glioma cell line) in response to upstream signalling pathway modulators such as EGF or PI3K inhibitors. On the other hand, mTOR inhibitors acting downstream of Akt did not result in reporter bioluminescence alteration. These results indicate the utility of the reporter to determine the pharmacodynamics of compounds that modulate this signalling pathway.

32 Imaging in Neurology Research I: Neurooncology

Fig. 32.2 Imaging E2F-1 signal transduction. (a) E2F-regulated cells and negative and positive control cells were implanted as a set of four tumours in the back of different groups of experimental mice. Mice were followed over time by bioluminescence imaging (BLI) until tumours could be clearly visualized. Mice were then subjected to BCNU treatment (50%) or control treatment (50%), and repeat imaging was performed 24 h later. An increased luciferase signal was observed only in mice bearing E2F-1-regulated cells and not in mice bearing negative and posi-

32.6 Imaging Gene and Cell-Based Therapy

485

tive control cells. Colour scale luminescent signal intensity; blue least intense signal; red most intense signal. (b) Mean of the total bioluminescent signals emitted from E2F-1-regulated tumours and negative and positive control tumours in response to BCNU administration. Columns, mean of three independent experiments with n = 6 animals per group; bars SD. Significant differences are indicated by asterisk (p < 0.05) and double asterisk (p < 0.05) (modified with permission from Monfared et al. 2008)

that from the activity of the reporter gene (level of signal intensity), the activity of the therapeutic gene can be deduced. Several reporter gene systems have been In recent years, much effort has been made to increase developed for nuclear, optical and MR imaging. the efficacy of gene therapy for oncological applica- Especially, the herpes simplex virus type 1 thymidine tion. Especially, the development of effective non- kinase gene1-tk has been studied extensively for gene invasive in vivo imaging techniques for monitoring therapy purposes due to the fact that it can serve as suivector delivery and subsequent gene expression and cide gene and reporter gene simultaneously depending treatment efficacy has received particular attention over on the administered substrate (the prodrug GC or the the last few years (Waerzeggers et al. 2009a). For this radiolabelled ligands FHBG, FIAU or 2¢-fluoropurpose, a reporter gene is commonly linked to the 2¢deoxy-1b-d-arabionofuranosyl-5-ethyl-uracil, FEAU, therapeutic gene by a shared promoter, which results in respectively) (Kang and Chung 2008; Waerzeggers a proportional co-expression of both genes. This means et al. 2009a). Our group has successfully applied small

486

animal PET imaging to monitor gene expression and to evaluate gene therapy in vivo. Employing an “imagingguided” suicide gene therapy protocol, we showed in a subcutaneous glioblastoma model a response rate of 68% of tumours in vivo transduced with a HSV-1 amplicon vector proportionally co-expressing E. coli cytosine deaminase (cd) as a therapeutic gene and HSV-1-tk as a combined PET marker and therapeutic gene (Jacobs et al. 2007). In this study, MRI was used for tumour localization and PET, using the radiotracers [18F]FDG, [18F]FLT and [11C]MET, was used for identification of actively proliferating tumour tissue and guidance of targeted vector application (Fig. 32.3a). This is an important aspect in a gene therapy protocol as only viable tumour tissue is likely to respond to the introduced therapeutic genes, whereas necrotic tissue will show no benefit. These radiotracers displayed homogeneous uptake in small tumours and heterogeneous uptake in larger tumours. In the latter, no radiotracer uptake occurred in the central part of the tumours indicative of central necrosis. For image validation, transaxial PET images were co-registered with histology showing a correlation between the lack of tracer accumulation and the histological signs of necrosis, whereas positive tracer accumulation correlated with viable tumour tissue. Tumour-to-background ratios for the various radiotracers ranged from 1.9 to 8.0 in viable tumour tissue and from 0.4 to 1.2 in necrosis. In this study [18F]FLT was also used to determine the gene therapy-induced inhibition of proliferative activity of the tumour and thus therapeutic efficacy. Furthermore, [18F]FHBG was used as a marker for the exogenously introduced therapeutic (HSV-1-tk) gene expression and to predict response to therapy (Fig. 32.3a). Specific [18F]FHBG accumulation could localize the transduced tissue dose of therapeutic gene expression. Moreover, the primary transduction efficiency as measured by [18F]FHBG PET was correlated to the induced therapeutic effect as measured by [18F]FLT PET (Fig. 32.3b, c). Overall, this study showed the feasibility of a multimodal imaging approach for (i) identification of viable target tissue that might benefit from gene therapy, (ii) quantification of the transduction efficiency and (iii) monitoring of therapeutic efficacy. Wang et al. compared various PET tracers for their ability to assess response in tumours undergoing HSV-1-tk plus GCV prodrug activation gene therapy (Wang et al. 2006). The tracers evaluated interrogated therapy-induced changes in tumour glucose

Y. Waerzeggers et al.

metabolism ([18F]FDG), protein synthesis ([18F]FET) and cell proliferation/DNA synthesis (5-[18F]-fluoro29-deoxyuridine, [18F]FUdR). In addition, [123I]FIAU SPECT imaging was performed to evaluate HSV-1-tk gene expression. In agreement with the study of Jacobs et al., the authors found that [123I]FIAU SPECT imaging was the most reliable method for prediction of tumour response to GCV therapy, which was proportional to the magnitude of HSV-1-tk expression in tumour tissue. Furthermore, the authors showed that PET imaging with [18F]FUdR reliably visualizes proliferating tumour tissue and is most suitable for the assessment of responses in tumours undergoing HSV-1-tk plus GCV prodrug activation gene therapy, whereas PET imaging with [18F]FDG or [18F]FET can be used as additional “surrogate” biomarkers of the treatment response, although these radiotracers are less sensitive than [18F] FUdR for monitoring tumour responses to prodrug activation gene therapy with HSV-1-tk and GCV in this model. Especially, bioluminescent reporter genes have emerged as particularly useful for evaluation of therapeutic gene transfer and treatment response due to the operational ease, low cost and high throughput of OI. Söling and co-workers monitored suicide gene therapy in malignant glioma employing the prodrug activating system HSV-1-tk/GCV with BLI (Soling et al. 2004). The authors developed a fusion construct of HSV-1-tk and Fluc and showed that GCV-induced cytotoxicity in U87MG glioma xenografts stably expressing the fusion construct could be monitored by serial BLI and that bioluminescent signal intensities correlated to tumour volume changes measured with callipers. Especially, the ability to image two or more biological processes in a single animal greatly increases the utility of luciferase imaging for gene therapy purposes. Dual luciferase imaging takes advantage from the fact that the luciferases from Renilla and Firefly have different substrates, coelenterazine and d-luciferin, respectively, and thus can be imaged in the same living mouse with kinetics of light production being separable in time by separate injections of these two substrates (Waerzeggers et al. 2009a). Dual BLI has been used to monitor gene delivery via a therapeutic vector and to follow the effects of the therapeutic protein tumour necrosis factorrelated apoptosis-inducing ligand (TRAIL) in gliomas (Shah et al. 2003). Glioma cells stably expressing Fluc were implanted subcutaneously into nude mice, and the tumour growth was

32 Imaging in Neurology Research I: Neurooncology

Fig. 32.3 Imaging-guided gene therapy paradigm. (a) experimental protocol for identification of viable target tissue and assessment of vector-mediated gene expression in vivo in a mouse model with three subcutaneous gliomas. Row 1 localization of tumours is displayed by MRI. Row 2 the viable target tissue is displayed by [18F]FDG PET; note the signs of necrosis in the lateral portion of the left-sided tumour (arrow). Rows 3+4 following vector application into the medial viable portion of the tumour (arrow), the tissue dose of vector-mediated gene expression is quantified by [18F]FHBG PET. Row 3 shows an image acquired early after tracer injection, which is used for co-registration; row 4 displays a late image with specific tracer accumulation in the tumour that is used for quantification (adapted with permission from (Jacobs et al. 2007)). (b) Response to gene therapy correlates to therapeutic gene expression. The intensity of cdIREStk39gfp expression, which is equivalent to transduction efficiency

487

and tissue dose of vector-mediated therapeutic gene expression, is measured by [18F]FHBG PET (in %ID/g), and the induced therapeutic effect is measured by [18F]FLT PET (R = 0.73, p < 0.01). Therapeutic effect ([18F]FLT) was calculated as the difference between [18F]FLT accumulation after and before therapy (adapted with permission from (Jacobs et al. 2007)). (c) Relation between changes in volumetry and [18F]FLT uptake. Changes in tumour volume and [18F]FLT uptake were plotted for tumours grown in 11 nude mice. There is a strong correlation between volumetry and change in [18F]FLT uptake (R = 0.83) for those tumours responding to therapy (complete responders) and a weaker correlation (R = 0.57) for those tumours not responding to therapy (nonresponders). No correlation was found for those tumours where focal alterations of [18F]FLT uptake occurred, which did not lead to a reduction in overall tumour volume (partial responders; adapted with permission from (Jacobs et al. 2007))

488

Y. Waerzeggers et al.

Fig. 32.3 (continued)

monitored in vivo over time by luciferin administration and BLI. HSV amplicon vectors bearing the genes for TRAIL and Rluc were injected directly into these Fluc-positive gliomas allowing super imposition of gene delivery to the tumour by coelenterazine administration and BLI. In contrast to the mock-injected gliomas, in TRAIL-treated gliomas tumour regression over time could be monitored with BLI after luciferin administration. In a subsequent study, the same authors took advantage of the glioma tracking capabilities of neural progenitor cells (NPCs) to deliver therapeutic molecules to intracranial tumours (Shah et al. 2005). The authors engineered mouse NPCs to simultaneously express Fluc and deliver a secreted form of TRAIL (sTRAIL) (NPC-FL-sTRAIL) and by using a highly malignant human glioma model expressing Renilla luciferase (Rluc), they showed that intracranially implanted NPC-FL-sTRAIL expressing both firefly luciferase (Fluc) and sTRAIL migrated into the tumours and induced profound antitumour effects. Again, by the use of the dual luciferase imaging method, both NPC migration and tumour burden could be monitored in the same animal over time. For gene therapy protocols to be effective, the level and duration of transgene expression are important. Even more important are selective delivery and expression of the therapeutic transgenes in the tissue of interest, mostly cancer cells, with sparing of the surrounding normal cells to reduce side-effects and the possibility

of temporal induction of therapeutic gene expression to turn on/off therapeutic gene expression at the desired time point (e.g. when vectors maximally reached the target tissue or when the desired therapeutic effect has been reached). To demonstrate that noninvasive imaging of regulated expression of any type of gene after in vivo transduction by versatile vectors is feasible, our group generated regulatable HSV-1 amplicon vectors carrying hormone (mifepristone) or antibiotic (tetracycline) regulated promoters driving the proportional co-expression of two marker genes (a fluorescent gene [rfp or egfp] with either a PET [HSV1-tk] or an optical [firefly luciferase] imaging gene) (Winkeler et al. 2007). Inducible and regulated gene expression could be monitored by fluorescence microscopy in culture and by PET (Fig. 32.4) or BLI in vivo (Fig. 32.5). Moreover, several strategies have been developed and evaluated to selectively deliver and/or express transgenes in the tissue of interest. These strategies either utilize tissue- or tumour-specific promoters to drive transgene expression (transcriptional targeting) or use a variety of modifications on vector tropism (transductional targeting) (Waerzeggers et al. 2009a). For instance, Liang et al. coated an adenovirus expressing firefly luciferase from the CMV early promoter (AdCMVFluc) with a recombinant, bi-specific molecule containing the soluble portion of CAR (Coxsackie and Adenovirus receptor) fused to EGF. By this procedure, CAR-

32 Imaging in Neurology Research I: Neurooncology Fig. 32.4 In vivo imaging of regulated gene expression. Multimodal imaging ([18F] FLT-, [18F]FDG- and [18F] FHBG PET) of inducible gene expression by PET in nude mice bearing subcutaneous human Gli36dEGFR xenografts employing HSV-Switch-TG17 or HET6C-tk39. Mice were randomized to mifepristoneor doxycycline-treated (each group n = 7) and untreated (n-mif = 8; n-dox = 12) groups. Indicated xenografts (arrowhead) were injected with HSV-Switch-TG17 and HET6C-tk39, respectively, 48 h prior to PET imaging resulting in high accumulation of [18F]FHBG in HSV-Switch-TG17 or HET6C-tk39 infected and mifepristone- or doxycyclinetreated tumours as compared to some background activity in non-treated tumours (adapted with permission from (Winkeler et al. 2007))

489

490

Y. Waerzeggers et al.

Fig. 32.5 In vivo BLI of doxycycline-dependent gene expression over time. BLI of induced LUC expression and image validation by histology. Unit for all colour scales as well as the histogram on temporal analysis was defined as photons/second/ cm2/steradian (p/s/cm2/sr). (a) Temporal analysis of up- and down-regulation of LUC expression. HET-6C injection was performed intratumourally at day 0. Days where bioluminescent images were obtained are indicated at the upper right corner, days of doxycycline treatment at the top. (b) Quantitative analysis of luciferase signal (OFF–ON–OFF) in response to doxycycline. (c) Temporal analysis of up- and down-regulation of

LUC expression in the intracranial glioma model (OFF– ON–OFF–ON). Indicated are the days of tumour growth. (d) Image validation by histology. BLI of a mouse bearing a subcutaneous glioma stably expressing LUC on its left shoulder after in vivo transduction with HET6C-luc in the tumour on the right shoulder. Representative histological sections taken from the in vivo transduced tumour showing co-localization of eGFP expressed constitutively from the herpes viral immediate early 4/5 promoter, and RFP, expressed from the bi-directional regulated promoter (Scale bar overlay: 150 mm, exposure time: 0.5 s) (adapted with permission from Winkeler et al. 2007)

dependent adenovirus infection to normal tissue could be blocked and redirected to specific, EGFR expressing, target cells (Liang et al. 2004a, b). After systemic administration of coated AdCMVFluc, luciferase expression was reduced in the liver and was facilitated in tumour xenografts expressing high levels of the EGFR, as quantitatively demonstrated by BLI.

The use of neural stem cells (NSCs) for the treatment of brain tumours has attracted much interest in recent years (Waerzeggers et al. 2009a). The characteristics of NSCs that make them attractive vehicles for targeted delivery are their tropic behaviour towards neoplasms. Aboody et al. (2000) first documented that modified exogenous NSCs injected into the

32 Imaging in Neurology Research I: Neurooncology

contralateral hemispheres migrate over long distances to sites of gliomas in mice. Recently, several groups have reported promising results with extended survival or reduced tumour growth after NSC-based gene therapy in animal models of high-grade gliomas (Brekke et al. 2007). An intrinsic tumour-inhibition effect of NSCs has been observed after co-inoculation of NSCs with glioma cells or grafting into established gliomas. Also, due to the pluripotency of NSCs, their potential role may be to repair the damage caused by the brain tumours themselves and the neurological impairment that is frequently associated with traditional cancer treatment approaches (Brekke et al. 2007). Until recently, the in vivo study of particular populations of cells, such as NSCs, was mainly based on snapshot images from ex vivo histology. Thus, imaging techniques for in vivo longitudinal detection of the dynamics of these cells became desirable. Over the past years, the tumour targeting abilities of implanted neural stem and progenitor cells have been monitored non-invasively with various imaging modalities and the most extensive information has been gained with reporter genes and optical imaging. Also, cellular MR imaging techniques have been used increasingly and even PET imaging has been applied to detect stem cell migration and therapy in intracranial glioma models. By the use of BLI systems, Shah and co-workers simultaneously monitored both the migration of NSCs toward gliomas and the efficacy of sTRAIL on the glioma burden in real time by dual enzyme substrate (Rluc/Fluc) imaging (Shah et al. 2005). However, in general, BLI offers poor tissue penetration and poor spatial resolution. Miletic et al. (2007) together with our group used a multimodal imaging protocol combining multi-tracer PET and MRI to demonstrate that a subpopulation of bone marrow-derived mesenchymal stem cells can be used as tumour-infiltrating therapeutic cells against malignant glioma. The stem cells genetically engineered to express HSV-1-TK were injected into rat intracranial 9L gliomas. After transplantation, stem cell localization and distribution could be monitored non-invasively by means of the HSV-1-TK-specific PET radioligand [18F]FHBG. In addition, the therapeutic effect of GC treatment could be monitored sequentially by MRI and [11C]MET PET and strongly correlated with histological analysis (Fig. 32.6). Cellular imaging by magnetic resonance imaging (cellular MRI) provides another non-invasive dynamic method for evaluating the seeding, migration and homing of magnetically labelled NSCs (Modo et al. 2005), as well as an excellent soft tissue differentiation with a

491

high spatial resolution. This led Zhang et al. (2004) to successfully monitor NPCs labelled with SPIO in a rat gliosarcoma model using in vivo MRI as they infiltrated the tumour mass or as they tracked down invading tumour cells. Also, migration and incorporation of magnetically labelled NSCs into the angiogenic vasculature of brain tumours has been studied (Modo et al. 2005). Brekke et al. (2007) investigated the potential of cellular MR imaging to monitor in vivo the migration and infiltration of GRID-labelled murine NSCs (MHP36) from the seeding site to the tumour region in a rat glioma model using longitudinal multiparametric MR imaging. Additionally, the authors were able to demonstrate the therapeutic potential of the mere injection of NSCs through an MRI-based measurement of tumour growth and development of vasogenic oedema. Fulci et al. (2006) used cellular MRI to non-invasively monitor innate immune response (iron oxide nanoparticle-loaded macrophages) after glioma virotherapy in a syngeneic rat glioma model. Furthermore, the applied image protocol enabled the authors to image the increased efficiency of oncolytic virotherapy after preadministration of cyclophosphamide. This increased efficiency was credited to the immunosuppressive action of cyclophosphamide. Simultaneous tracking of transplanted stem cell fate and function non-invasively in vivo has gained increa sed attention as it can address many safety aspects of the use of these cells as targeting vehicles for gene therapy (stem cell distribution, normal or aberrant stem cell proliferation, extent and duration of transgene expression, response to therapy). To address this issue, Cao et al. (2007) stably transduced murine embryonic stem (ES) cells with double or triple fusion reporter gene constructs expressing monomeric RFP, firefly luciferase and truncated HSV-1-TK, respectively, and injected these cells subcutaneously into nude mice and tracked cell survival non-invasively by bioluminescence and PET imaging. This multimodality imaging approach was successful to monitor transplanted ES cell survival and proliferation in vivo and to assess the efficacy of suicide gene therapy as a back-up safety measure (death switch) against teratoma formation. Also, our group addressed the feasibility of multimodal imaging for non-invasive and early detection of the safety aspects related with the use of stem cells as delivery vectors for gene therapy. In this study, mouse NPCs genetically engineered to express HSV1-TKGFP and Fluc (C17.2-LITG) were injected in the brains of mice bearing highly malignant human gliomas, either within the tumours or within the

492

Fig. 32.6 Multimodal imaging of cell-based therapy. Repre sentative three-dimensional MRI (T1-weighted FLASH, echo time = 5 ms, repetition time = 70 ms, 60° pulse, resolution 121 × 121 × 242 mm, postadministration of gadopentetic acid [Gd-DTPA]), [11C]MET PET and [18F]FHBG-PET scans. Time points after tumour implantation: (a–c) 6 days, MRI before ganciclovir (GC) treatment; (d–f) 7–8 days, [11C]MET PET before GC treatment; (g–i) 12 days, [18F]FHBG PET during GC treatment; (j–l) 13 days, MRI during GC treatment; (m–o) 14 days, [11C]MET PET during GC treatment; (p) 21 days, MRI after GC

Y. Waerzeggers et al.

treatment; (q) 22 days, [11C]MET PET after GC treatment. The three different groups were treated as described: (a, d, g, j, m, p, q) with injection of bone marrow-derived tumour-infiltrating cells expressing the genes for thymidine kinase and green fluorescent protein (BM-TIC-tk-gfp) and GC treatment; (b, e, h, k, n) with BM-TIC-tk-gfp injection without GC treatment; (c, f, i, l, o) 9LDsRed only with GC treatment (daily intraperitoneal injections of 30 mg/kg GC for 10 days, starting at day 4 after BM-TIC-tk-gfp implantation) (adapted with permission from Miletic et al. 2007)

32 Imaging in Neurology Research I: Neurooncology

493

Resistance to chemoradiotherapy is an important cause of treatment failure in patients with cancer, including brain tumours, and one of the most significant obstacles in improving the efficacy of cancer treatment. Non-invasive imaging approaches to predict response to chemoradiotherapy would significantly aid in guiding more efficient clinical management at an early time point. Chemotherapeutic failure is a multifactorial phenomenon related to initial resistance intrinsic to the tumour cells, to resistance acquired by tumours during treatment

or to physiologic factors extrinsic to tumours (e.g. individual variations and pharmacokinetic behaviours of drugs). Moreover, resistance may be heterogeneous with respect to either different tumours in the same individual or even different cells in the same tumour (Hsueh et al. 2006). The term multidrug resistance (MDR) is commonly used to indicate an overexpression of the transmembrane protein P-glycoprotein (P-gp) encoded by the mdr1 gene. P-glycoprotein (P-gp) acts as an energydependent efflux transporter that efficiently enhances outward transport and/or prevents entry of compounds such as anticancer drugs, thereby resulting in decreased intracellular accumulation and thus decreased cytotoxicity. P-gp confers cross-resistance to unrelated drugs that differ widely with respect to molecular structure and target specificity, including many natural product agents (e.g. paclitaxel, vincristine, doxorubicin, verapamil) as well as new targeted anticancer agents (e.g. Gleevec, a kinase inhibitor) (Pichler et al. 2005). P-gp is not only expressed in tumour cells, but also in normal, nonmalignant tissue involved in drug excretion and absorption such as kidneys, liver and colon. Furthermore, P-gp is highly expressed in endothelial cells at the blood– brain interface and is an essential element of its barrier function and protects against neurotoxicity. The mechanism of mdr1 gene regulation in these tissues, particularly in response to xenobiotics and other environmental and physiologic stimuli, is not fully understood. Several radiopharmaceuticals have been used as general probes for functional imaging of MDR pumps by measuring the efflux of these probes via the P-gp and other membrane transporters. The most commonly applied tracers are [99mTc]sestamibi and [99mTc]

Fig. 32.7 BLI of aberrant stem cell migration. C17.2-LITG cells were injected into the left striatum of a non-glioma-bearing mouse and their behaviour over time was monitored with optical imaging. NPC migration in the direction of the cerebellar hemi-

spheres (red arrow) could be demonstrated 13 days after injection (red arrow); at 3 weeks, NPCs also localized at the level of the thoracolumbar spine (green arrow) (adapted with permission from Waerzeggers et al. 2008)

opposite brain hemisphere, or of mice not bearing intracranial gliomas (Waerzeggers et al. 2008). BLI could not only detect migration of injected NPCs towards gliomas growing in the contralateral hemisphere, but was also able to detect aberrant cell migration towards the cerebellum and the spine, even in animals not bearing intracranial gliomas (Fig. 32.7). Furthermore, BLI and [18F]FHBG-PET imaging showed excessive proliferation of intratumourally injected NPCs. These results underscore the importance of non-invasive imaging as safety switch in cellbased studies. Most importantly, multimodal imaging of these cells transplanted into pre-established gliomas was obtained by MRI, PET and OI demonstrating for the first time that cells can be detected and followed by multimodal imaging in a clinically applicable imaging protocol (Fig. 32.8).

32.7 Imaging Tumour Chemo- and/or Radiosensitivity

494

Y. Waerzeggers et al.

Fig. 32.8 Multimodal imaging of intracranially injected C17.2LITG cells. FeO-labelled C17.2-LITG cells were co-injected together with Gli36dEGFR human glioma cells (2 × 105 cells, 1:1) into the right striatum of a nude mouse and in vivo imaging

was performed by OI, PET and MRI to detect glioma growth ([18F]FLT PET and Gd-enhanced T1-weighted MRI) and stem cell survival and location (BLI, [18F]FHBG PET and T2*weighted MRI)

tetrofosmin for SPECT imaging and [11C]verapamil and [18F]fluoropaclitaxel for PET imaging and they can quantify P-gp activity to predict response to chemotherapy (Kelloff et al. 2005; Hsueh et al. 2006) and can be used to monitor P-gp inhibition by cyclosporine (Sasongko et al. 2005). Also, coelenterazine, the bioluminescent substrate of the reporter gene Renilla luciferase, is a well-known substrate for the multidrug transporter P-gp, thereby enabling P-gp transport activity to be monitored in intact cells and living animals with non-invasive BLI. This approach has been used to non-invasively image the reversal of MDR in mice (Pichler et al. 2004). In cultured living cells, stably transfected with a codon-humanized Renilla luciferase, it was shown that low baseline coelenterazine-mediated bioluminescence could be fully enhanced (reversed) to non-P-gp-matched control levels with potent and selective P-gp inhibitors. This study emphasizes the role of coelenterazine as a P-gp substrate, but at the same time raises concerns regarding the indiscriminative use of Renilla luci ferase and aequorin as reporters in intact cells and

transgenic animals, since P-gp-mediated alterations in coelenterazine permeability may impact results (Gross and Piwnica-Worms 2005). Inhibition of the activity or expression of P-gp represents a valuable means to circumvent the drug resistance of cancer cells and several approaches have been tested. In a study by Pichler et al., the effectiveness of a P-gp inhibition strategy using short hairpin RNA interference (shRNAi) against human MDR1 mRNA was tested and could be monitored by direct non-invasive BLI using coelenterazine (Pichler et al. 2005). Recently, Park et al. investigated the feasibility of a combination of chemo- and radioiodide therapy and non-invasive imaging with a dual therapeutic vector expressing MDR1 shRNAi and a hNIS (Park et al. 2008). The authors showed that shMDR-NIS-expressing cells showed a significant decrease in the expression of MDR1 messenger RNA and P-gp and that this inhibition enhanced the intracellular accumulation of paclitaxel, resulting in increased sensitivity to this chemotherapeutic agent. Furthermore, the shMDRNIS-expressing cells showed a significant increase of

495

32 Imaging in Neurology Research I: Neurooncology 125 I uptake and a marked decrease in cell survival rate after 131I treatment. Interestingly, the combination of doxorubicin and 131I displayed synergistic cytotoxicity that correlated with MDR1 inhibition and NIS expression in shMDR-NIS-expressing cells. In mice with shMDR-NIS-expressing tumour xenografts, small animal PET with 124I clearly visualized shMDR1-NIS-expressing tumours. To gain a better understanding of mdr1 transcriptional regulation, Gu and colleagues recently developed a knock-in mouse model that uses a firefly luciferase (Fluc) gene inserted by homologous recombination into the murine mdr1a genetic locus and that is flanked by loxP sites, so that Cre-mediated recombination is required to configure the Fluc gene directly under the control of the endogenous mdr1a promoter (Gu et al. 2009). The authors showed that this knock-in model can be used for non-invasive assessment of basal and xenobiotic-induced regulation of mdr1a.Fluc expression in real time. With the chosen vector design, which requires Cre-mediated recombination to achieve the Fluc knock-in, it should also be possible to control mdr1a.Fluc expression spatially by mating mdr1aflox mice with Cre-donor mice that express Cre in a tissue-restricted way. Therefore, this mouse model represents a unique tool to study the magnitude and kinetics of mdr1a induction under a variety of physiologic, pharmacologic, genetic, and environmental conditions. Since over 40 years, it is known that tumour cells at low oxygen tension are relatively radioresistant (Belkacemi et al. 2007). In recent years, more and more studies indicate that hypoxia can also induce the expression of specific genes and promote a more aggressive tumour phenotype. These findings suggest that procedures which could accurately measure the hypoxic fraction of individual tumours before, during and after therapy would permit valuable predictive information about tumour response and identification of subgroups of patients that would benefit from hypoxia-targeted therapy, such as hyperbaric oxygen treatment, neutron radiotherapy or hypoxic radiosensitizer administration. Like with chemoresistance, tumour radioresistance can be constitutional or acquired during treatment. The hypoxic fraction of individual tumours can be defined by microelectrode techniques or non-invasive imaging assays using different radiolabelled products of

the imidazole family. These markers are selectively reduced and accumulate in hypoxic cells and tissues. Most studied are 18F-labelled analogues of misonidazole (MISO) and the copper chelate Cu-diacetylbis(N4-methylthiosemicarbazone) (Cu-ATSM) and a significant correlation between tumour ligand uptake and radioresistance could be demonstrated (Belkacemi et al. 2007). Tumour hypoxia can also be measured indirectly with reporter genes sensitized to over-express genes being induced under hypoxic/ anoxic conditions, such as HIF-1. Hypoxic glioma tumour tissue has been imaged non-invasively and repeatedly in vivo through HIF-1-mediated activation of HSV-1-TKGFP reporter gene expression and PET imaging with 18F-labelled 2¢-fluoro-2¢deoxy1b-d-arabionofuranosyl-5-ethyl-uracil ([18F]FEAU) (Serganova et al. 2004). Furthermore, Wen et al. showed the equivalence between hypoxia detection by the intratumour distribution of [124I]FIAU or by [18F]MISO (Wen et al. 2004).

32.8 Conclusion Continuous refinements in available brain tumour models together with improvements in small animal imaging techniques have recently led to a rapid progress in a variety of ways by which tumour biology can be monitored non-invasively in living animals. Insights into the pathophysiological processes related to disease initiation and progression resulted in the identification of new molecular targets or treatment strategies, and functional and molecular imaging probes directed to disease-specific alterations have been developed and optimized. These advances greatly increased our knowledge of disease dynamics and refined the design of effective therapeutic interventions in a true translational manner. Ultimately, application of non-invasive imaging in animal brain tumour models will result in improved patient-tailored care by an individualized imaging-guided therapy approach. Acknowledgements Our work is supported in part by the Deutsche Forschungsgemeinschaft (DFG-Ja98/1-2), by the 6th FW EU grants EMIL (LSHC-CT-2004-503569), DiMI (LSHB-CT-2005-512146) and CliniGene NoE (LSHB-CT2006-018933) and by the Bundesministerium für Bildung und Forschung (MoBiMed 01EZ0811).

496

References Paulus MJ, Gleason SS, Kennel SJ, Hunsicker PR, Johnson DK (2000) High resolution X-ray computed tomography: an emerging tool for small animal cancer research. Neoplasia 2(1–2):62–70 Pichler BJ, Wehrl HF, Judenhofer MS (2008) Latest advances in molecular imaging instrumentation. J Nucl Med 49(suppl 2): 5S–23S Pierce MC, Javier DJ, Richards-Kortum R (2008) Optical contrast agents and imaging systems for detection and diagnosis of cancer. Int J Cancer 123(9):1979–1990 Wang LV (2008) Prospects of photoacoustic tomography. Med Phys 35(12):5758–5767 Fomchenko EI, Holland EC (2006) Mouse models of brain tumors and their applications in preclinical trials. Clin Cancer Res 12(18):5288–5297 Huse JT, Holland EC (2009) Genetically engineered mouse models of brain cancer and the promise of preclinical testing. Brain Pathol 19(1):132–143 Gross S, Piwnica-Worms D (2006) Molecular imaging strategies for drug discovery and development. Curr Opin Chem Biol 10(4):334–342 Kang JH, Chung JK (2008) Molecular-genetic imaging based on reporter gene expression. J Nucl Med 49(suppl 2):164S–179S Gross S, Piwnica-Worms D (2005) Spying on cancer: molecular imaging in vivo with genetically encoded reporters. Cancer Cell 7(1):5–15 Gilad AA, Winnard PT Jr, van Zijl PC, Bulte JW (2007) Developing MR reporter genes: promises and pitfalls. NMR Biomed 20(3):275–290 Waerzeggers Y, Monfared P, Viel T, Winkeler A, Voges J, Jacobs AH (2009a) Methods to monitor gene therapy with molecular imaging. Methods 48(2):146–160 Oriuchi N, Higuchi T, Ishikita T, Miyakubo M, Hanaoka H, Iida Y et al (2006) Present role and future prospects of positron emission tomography in clinical oncology. Cancer Sci 97(12): 1291–1297 Herholz K, Coope D, Jackson A (2007) Metabolic and molecular imaging in neuro-oncology. Lancet Neurol 6(8):711–724 Singhal T, Narayanan TK, Jain V, Mukherjee J, Mantil J (2007) (11)C-L-methionine positron emission tomography in the clinical management of cerebral gliomas. Mol Imaging Biol. 2008 Jan-Feb;10(1):1–18. Epub 2007 Oct 24 Sherley JL, Kelly TJ (1988) Regulation of human thymidine kinase during the cell cycle. J Biol Chem 263(17):8350–8358 Ullrich R, Backes H, Li H, Kracht L, Miletic H, Kesper K et al (2008) Glioma proliferation as assessed by FLT-PET in patients with newly diagnosed high grade glioma. Clin Cancer Res 14:2049–2055 Bradbury MS, Hambardzumyan D, Zanzonico PB, Schwartz J, Cai S, Burnazi EM et al (2008) Dynamic small-animal PET imaging of tumor proliferation with 3¢-deoxy-3¢-18Ffluorothymidine in a genetically engineered mouse model of high-grade gliomas. J Nucl Med 49(3):422–429 McConville P, Hambardzumyan D, Moody JB, Leopold WR, Kreger AR, Woolliscroft MJ et al (2007) Magnetic resonance imaging determination of tumor grade and early response to temozolomide in a genetically engineered mouse model of glioma. Clin Cancer Res 13(10):2897–2904

Y. Waerzeggers et al. Leenders WP, Kusters B, Verrijp K, Maass C, Wesseling P, Heerschap A et al (2004) Antiangiogenic therapy of cerebral melanoma metastases results in sustained tumor progression via vessel co-option. Clin Cancer Res 10(18 pt 1):6222–6230 Claes A, Gambarota G, Hamans B, van Tellingen O, Wesseling P, Maass C et al (2008) Magnetic resonance imaging-based detection of glial brain tumors in mice after antiangiogenic treatment. Int J Cancer 122(9):1981–1986 Kremer S, Pinel S, Vedrine PO, Bressenot A, Robert P, Bracard S et al (2007) Ferumoxtran-10 enhancement in orthotopic xenograft models of human brain tumors: an indirect marker of tumor proliferation? J Neurooncol 83(2):111–119 Kircher MF, Mahmood U, King RS, Weissleder R, Josephson L (2003) A multimodal nanoparticle for preoperative magnetic resonance imaging and intraoperative optical brain tumor delineation. Cancer Res 63(23):8122–8125 Veeravagu A, Hou LC, Hsu AR, Cai W, Greve JM, Chen X et al (2008) The temporal correlation of dynamic contrastenhanced magnetic resonance imaging with tumor angiogenesis in a murine glioblastoma model. Neurol Res 30(9): 952–959 Chenevert TL, McKeever PE, Ross BD (1997) Monitoring early response of experimental brain tumors to therapy using diffusion magnetic resonance imaging. Clin Cancer Res 3(9): 1457–1466 Asanuma T, Doblas S, Tesiram YA, Saunders D, Cranford R, Yasui H et al (2008) Visualization of the protective ability of a free radical trapping compound against rat C6 and F98 gliomas with diffusion tensor fiber tractography. J Magn Reson Imaging 28(3):574–587 Hakumaki JM, Poptani H, Sandmair AM, Yla-Herttuala S, Kauppinen RA (1999) 1H MRS detects polyunsaturated fatty acid accumulation during gene therapy of glioma: implications for the in vivo detection of apoptosis. Nat Med 5(11):1323–1327 Klerk CP, Overmeer RM, Niers TM, Versteeg HH, Richel DJ, Buckle T et al (2007) Validity of bioluminescence measurements for noninvasive in vivo imaging of tumor load in small animals. Biotechniques 43(1 suppl):7–13, 30 Rehemtulla A, Stegman LD, Cardozo SJ, Gupta S, Hall DE, Contag CH et al (2000) Rapid and quantitative assessment of cancer treatment response using in vivo bioluminescence imaging. Neoplasia 2(6):491–495 Szentirmai O, Baker CH, Lin N, Szucs S, Takahashi M, Kiryu S et al (2006) Noninvasive bioluminescence imaging of luciferase expressing intracranial U87 xenografts: correlation with magnetic resonance imaging determined tumor volume and longitudinal use in assessing tumor growth and antiangiogenic treatment effect. Neurosurgery 58(2): 365–372, discussion 365–372 Jahnke K, Muldoon LL, Varallyay CG, Lewin SJ, Brown RD, Kraemer DF et al (2009) Efficacy and MRI of rituximab and methotrexate treatment in a nude rat model of CNS lymphoma. Neuro Oncol 11:503–513 Lee KC, Hall DE, Hoff BA, Moffat BA, Sharma S, Chenevert TL et al (2006) Dynamic imaging of emerging resistance during cancer therapy. Cancer Res 66(9):4687–4692 Pantaleo MA, Nannini M, Maleddu A, Fanti S, Ambrosini V, Nanni C et al (2008) Conventional and novel PET tracers for imaging in oncology in the era of molecular therapy. Cancer Treat Rev 34(2):103–121

32 Imaging in Neurology Research I: Neurooncology Coppola JM, Ross BD, Rehemtulla A (2008) Noninvasive imaging of apoptosis and its application in cancer therapeutics. Clin Cancer Res 14(8):2492–2501 Kelloff GJ, Krohn KA, Larson SM, Weissleder R, Mankoff DA, Hoffman JM et al (2005) The progress and promise of molecular imaging probes in oncologic drug development. Clin Cancer Res 11(22):7967–7985 Blankenberg FG (2008) In vivo detection of apoptosis. J Nucl Med 49(suppl 2):81S–95S Laxman B, Hall DE, Bhojani MS, Hamstra DA, Chenevert TL, Ross BD et al (2002) Noninvasive real-time imaging of apoptosis. Proc Natl Acad Sci U S A 99(26):16551–16555 Lee KC, Hamstra DA, Bhojani MS, Khan AP, Ross BD, Rehemtulla A (2007) Noninvasive molecular imaging sheds light on the synergy between 5-fluorouracil and TRAIL/ Apo2L for cancer therapy. Clin Cancer Res 13(6):1839–1846 Ray P, De A, Patel M, Gambhir SS (2008) Monitoring caspase-3 activation with a multimodality imaging sensor in living subjects. Clin Cancer Res 14(18):5801–5809 Neves AA, Brindle KM (2006) Assessing responses to cancer therapy using molecular imaging. Biochim Biophys Acta 1766(2):242–261 Schellenberger EA, Bogdanov A Jr, Petrovsky A, Ntziachristos V, Weissleder R, Josephson L (2003) Optical imaging of apoptosis as a biomarker of tumor response to chemotherapy. Neoplasia 5(3):187–192 Waerzeggers Y, Jacobs AH, Monfared P, Viel T (2009) Molecular imaging in patient-tailored cancer therapies. Eur Pharm Rev Digital 3 Issue 3, p27–37 Cai W, Chen X (2008) Multimodality molecular imaging of tumor angiogenesis. J Nucl Med 49(suppl 2):113S–128S Hsu AR, Cai W, Veeravagu A, Mohamedali KA, Chen K, Kim S et al (2007) Multimodality molecular imaging of glioblastoma growth inhibition with vasculature-targeting fusion toxin VEGF121/rGel. J Nucl Med 48(3):445–454 Serganova I, Doubrovin M, Vider J, Ponomarev V, Soghomonyan S, Beresten T et al (2004) Molecular imaging of temporal dynamics and spatial heterogeneity of hypoxia-inducible factor-1 signal transduction activity in tumors in living mice. Cancer Res 64(17):6101–6108 Brandes AA, Franceschi E, Tosoni A, Hegi ME, Stupp R (2008) Epidermal growth factor receptor inhibitors in neuro-oncology: hopes and disappointments. Clin Cancer Res 14(4):957–960 Pantaleo MA, Nannini M, Maleddu A, Fanti S, Nanni C, Boschi S et al (2009) Experimental results and related clinical implications of PET detection of epidermal growth factor receptor (EGFr) in cancer. Ann Oncol 20(2):213–226 Gelovani JG (2008) Molecular imaging of epidermal growth factor receptor expression-activity at the kinase level in tumors with positron emission tomography. Cancer Metastasis Rev 27:645–653 Cai W, Niu G, Chen X (2008) Multimodality imaging of the HER-kinase axis in cancer. Eur J Nucl Med Mol Imaging 35(1):186–208 Pal A, Glekas A, Doubrovin M, Balatoni J, Namavari M, Beresten T et al (2006) Molecular imaging of EGFR kinase activity in tumors with 124I-labeled small molecular tracer and positron emission tomography. Mol Imaging Biol 8(5):262–277 Dissoki S, Aviv Y, Laky D, Abourbeh G, Levitzki A, Mishani E (2007) The effect of the [18F]-PEG group on tracer qualifi-

497 cation of [4-(phenylamino)-quinazoline-6-YL]-amide moiety–an EGFR putative irreversible inhibitor. Appl Radiat Isot 65(10):1140–1151 Cornelissen B, Kersemans V, Burvenich I, Oltenfreiter R, Vanderheyden JL, Boerman O et al (2005) Synthesis, biodistribution and effects of farnesyltransferase inhibitor therapy on tumour uptake in mice of 99mTc labelled epidermal growth factor. Nucl Med Commun 26(2):147–153 Suwa T, Ozawa S, Ueda M, Ando N, Kitajima M (1998) Magnetic resonance imaging of esophageal squamous cell carcinoma using magnetite particles coated with anti-epidermal growth factor receptor antibody. Int J Cancer 75(4):626–634 Diagaradjane P, Orenstein-Cardona JM, Colón-Casasnovas NE, Deorukhkar A, Shentu S, Kuno N et al (2008) Imaging epidermal growth factor receptor expression in vivo: pharmacokinetic and biodistribution characterization of a bioconjugated quantum dot nanoprobe. Clin Cancer Res 14(3): 731–741 Arwert E, Hingtgen S, Figueiredo JL, Bergquist H, Mahmood U, Weissleder R et al (2007) Visualizing the dynamics of EGFR activity and antiglioma therapies in vivo. Cancer Res 67(15):7335–7342 Saydam O, Glauser DL, Heid I, Turkeri G, Hilbe M, Jacobs AH et al (2005) Herpes simplex virus 1 amplicon vectormediated siRNA targeting epidermal growth factor receptor inhibits growth of human glioma cells in vivo. Mol Ther 12(5):803–812 Arrieta O, Garcia E, Guevara P, Garcia-Navarrete R, Ondarza R, Rembao D et al (2002) Hepatocyte growth factor is associated with poor prognosis of malignant gliomas and is a predictor for recurrence of meningioma. Cancer 94(12):3210–3218 Abounader R, Laterra J (2005) Scatter factor/hepatocyte growth factor in brain tumor growth and angiogenesis. Neuro Oncol 7(4):436–451 Towner RA, Smith N, Doblas S, Tesiram Y, Garteiser P, Saunders D et al (2008) In vivo detection of c-Met expression in a rat C6 glioma model. J Cell Mol Med 12(1):174–186 Monfared P, Winkeler A, Klein M, Li H, Klose A, Hoesel M et al (2008) Noninvasive assessment of E2F-1-mediated transcriptional regulation in vivo. Cancer Res 68(14):5932–5940 Doubrovin M, Ponomarev V, Beresten T, Balatoni J, Bornmann W, Finn R et al (2001) Imaging transcriptional regulation of p53dependent genes with positron emission tomography in vivo. Proc Natl Acad Sci U S A 98(16):9300–9305 Zhang L, Lee KC, Bhojani MS, Khan AP, Shilman A, Holland EC et al (2007) Molecular imaging of Akt kinase activity. Nat Med 13(9):1114–1119 Jacobs AH, Rueger MA, Winkeler A, Li H, Vollmar S, Waerzeggers Y et al (2007) Imaging-guided gene therapy of experimental gliomas. Cancer Res 67(4):1706–1715 Wang HE, Yu HM, Liu RS, Lin M, Gelovani JG, Hwang JJ et al (2006) Molecular imaging with 123I-FIAU, 18F-FUdR, 18F-FET, and 18F-FDG for monitoring herpes simplex virus type 1 thymidine kinase and ganciclovir prodrug activation gene therapy of cancer. J Nucl Med 47(7): 1161–1171 Soling A, Theiss C, Jungmichel S, Rainov NG (2004) A dual function fusion protein of Herpes simplex virus type 1 thymidine kinase and firefly luciferase for noninvasive in vivo imaging of gene therapy in malignant glioma. Genet Vaccines Ther 2(1):7

498 Shah K, Tang Y, Breakefield X, Weissleder R (2003) Real-time imaging of TRAIL-induced apoptosis of glioma tumors in vivo. Oncogene 22(44):6865–6872 Shah K, Bureau E, Kim DE, Yang K, Tang Y, Weissleder R et al (2005) Glioma therapy and real-time imaging of neural precursor cell migration and tumor regression. Ann Neurol 57(1):34–41 Winkeler A, Sena-Esteves M, Paulis LE, Li H, Waerzeggers Y, Ruckriem B et al (2007) Switching on the lights for gene therapy. PLoS One 2(6):e528 Liang Q, Yamamoto M, Curiel DT, Herschman HR (2004a) Noninvasive imaging of transcriptionally restricted transgene expression following intratumoral injection of an adenovirus in which the COX-2 promoter drives a reporter gene. Mol Imaging Biol 6(6):395–404 Liang Q, Dmitriev I, Kashentseva E, Curiel DT, Herschman HR (2004b) Noninvasive of adenovirus tumor retargeting in living subjects by a soluble adenovirus receptor-epidermal growth factor (sCAR-EGF) fusion protein. Mol Imaging Biol 6(6):385–394 Aboody KS, Brown A, Rainov NG, Bower KA, Liu S, Yang W et al (2000) Neural stem cells display extensive tropism for pathology in adult brain: evidence from intracranial gliomas. Proc Natl Acad Sci U S A 97(23):12846–12851 Brekke C, Williams SC, Price J, Thorsen F, Modo M (2007) Cellular multiparametric MRI of neural stem cell therapy in a rat glioma model. Neuroimage 37(3):769–782 Miletic H, Fischer Y, Litwak S, Giroglou T, Waerzeggers Y, Winkeler A et al (2007) Bystander killing of malignant glioma by bone marrow-derived tumor-infiltrating progenitor cells expressing a suicide gene. Mol Ther 15(7):1373–1381 Modo M, Hoehn M, Bulte JW (2005) Cellular MR imaging. Mol Imaging 4(3):143–164 Zhang Z, Jiang Q, Jiang F, Ding G, Zhang R, Wang L et al (2004) In vivo magnetic resonance imaging tracks adult neural progenitor cell targeting of brain tumor. Neuroimage 23(1):281–287 Fulci G, Breymann L, Gianni D, Kurozomi K, Rhee SS, Yu J et al (2006) Cyclophosphamide enhances glioma virotherapy by inhibiting innate immune responses. Proc Natl Acad Sci U S A 103(34):12873–12878

Y. Waerzeggers et al. Cao F, Drukker M, Lin S, Sheikh AY, Xie X, Li Z et al (2007) Molecular imaging of embryonic stem cell misbehavior and suicide gene ablation. Cloning Stem Cells 9(1):107–117 Waerzeggers Y, Klein M, Miletic H, Himmelreich U, Li H, Monfared P et al (2008) Multimodal imaging of neural progenitor cell fate in rodents. Mol Imaging 7(2):77–91 Hsueh WA, Kesner AL, Gangloff A, Pegram MD, Beryt M, Czernin J et al (2006) Predicting chemotherapy response to paclitaxel with 18F-Fluoropaclitaxel and PET. J Nucl Med 47(12):1995–1999 Pichler A, Zelcer N, Prior JL, Kuil AJ, Piwnica-Worms D (2005) In vivo RNA interference-mediated ablation of MDR1 P-glycoprotein. Clin Cancer Res 11(12):4487–4494 Sasongko L, Link JM, Muzi M, Mankoff DA, Yang X, Collier AC et al (2005) Imaging P-glycoprotein transport activity at the human blood-brain barrier with positron emission tomography. Clin Pharmacol Ther 77(6):503–514 Pichler A, Prior JL, Piwnica-Worms D (2004) Imaging reversal of multidrug resistance in living mice with bioluminescence: MDR1 P-glycoprotein transports coelenterazine. Proc Natl Acad Sci U S A 101(6):1702–1707 Park SY, Kwak W, Thapa N, Jung MY, Nam JO, So IS et al (2008) Combination therapy and noninvasive imaging with a dual therapeutic vector expressing MDR1 short hairpin RNA and a sodium iodide symporter. J Nucl Med 49(9): 1480–1488 Gu L, Tsark WM, Brown DA, Blanchard S, Synold TW, Kane SE (2009) A new model for studying tissue-specific mdr1a gene expression in vivo by live imaging. Proc Natl Acad Sci U S A 106(13):5394–5399 Belkacemi Y, Tsoutsou P, Magne N, Castadot P, Azria D (2007) Metabolic functional imaging for tumor radiosensitivity monitoring. Crit Rev Oncol Hematol 62(3):227–239 Wen B, Burgman P, Zanzonico P, O’Donoghue J, Cai S, Finn R et al (2004) A preclinical model for noninvasive imaging of hypoxia-induced gene expression; comparison with an exogenous marker of tumor hypoxia. Eur J Nucl Med Mol Imaging 31(11):1530–1538

Imaging in Neurology Research II: PET Imaging of CNS Disorders

33

Gjermund Henriksen and Alexander Drzezga

Contents 33.1 Specific Properties of the Brain with Regard to Imaging.................................................. 499 33.2 Specific Properties of Small Molecular Tracers..................................................................... 500 33.2.1 Uptake into the Brain................................................ 500 33.2.2 Lipophilicity............................................................. 500 33.2.3 Metabolism and Fate of the Label............................ 501 33.2.4 Mass Effects and Tracer Preparation........................ 501 33.3 Imaging CNS Transporters and Receptors in Rodent Models of Human Disease.................... 502 33.4 Imaging PBBR on Inflammation Sites in the CNS of Rodents....................................................... 502 33.5 Imaging Aggregation Pathology in Neurodegenration................................................... 502 33.5.1 Protein Aggregation as a Basic Pathology of Neurodegeneration.................................................... 502 33.5.2 Animal Models of Neurodegenerative Protein Aggregation.................................................. 506 33.5.3 Development of Tracers for In Vivo Imaging of b-Amyoid Plaque................................... 507 33.5.4 Imaging b-Amyloid Plaques in Mouse Models of AD........................................................... 508 33.6 Summary and Outlook........................................... 510 References............................................................................ 510

G. Henriksen (*) and A. Drzezga Department of Nuclear Medicine, Technische Universität München, Ismaninger Strasse 22, München, Germany e-mail: [email protected]

33.1 Specific Properties of the Brain with Regard to Imaging The imaging of physiologic and pathophysiologic processes in the brain is hampered by several factors. Starting with the anatomy, the brain can be regarded as the anatomically most complex organ in the human body. In addition to the complicated 3-dimensional structure, the functional anatomy of the brain is still poorly understood. Within the brain, a number of subconscious, autonomous and internal processes are controlled as well as the conscious processing of sensory information. Sensory motor functions as well as all cognitive processes such as memory, speech, spatial thinking, emotion and more complex constructs such as creativity or personality are all carried out in the brain. Thus, the brain can be considered as containing several different organs rather than representing one functional entity, and the term “central nervous system (CNS)” appears more appropriate than generalising the brain as a single organ. The different processes mentioned are controlled by a highly complex interaction of different types of signal-processing systems. Perfusion, metabolism, neuroendocrine factors, multiple transmitter/receptor systems and electrophysiologic neuronal conduction are involved in a complex interplay, which is only partly understood. Not only the understanding of the physiologic basics of brain function, but also that of the pathophysiology of many of diseases of the brain is still very limited. This applies to vascular and inflammatory disorders and brain tumours, and also particularly to the several neurodegenerative disorders that are affecting the society proportionally to the increased life expectancy of the population. Neurodegenerative disorders are often based on highly complex molecular pathologies. Their

F. Kiessling and B.J. Pichler (eds.), Small Animal Imaging, DOI: 10.1007/978-3-642-12945-2_33, © Springer-Verlag Berlin Heidelberg 2011

499

500

definite diagnosis is difficult because brain tissue is hardly accessible in vivo, e.g. for biopsy samples. In contrast to many inflammatory, vascular and malignant cerebral pathologies, the neurodegenerative pathologies usually do not disrupt the blood–brain barrier (BBB), which obstructs the access of therapeutic as well as diagnostic agents to the brain. Furthermore, cerebral compensation processes often do only allow clinical diagnosis of diseases in advanced stages, when symptoms become manifest. Thus, definite diagnosis of these disorders is often only possible by post-mortem histopathological examination. In consequence, treatment of neurodegenerative disorders is frequently particularly arduous, and therapies are still mostly symptomatic, i.e. a cure cannot be achieved. These facts underline the need for suitable diagnostic procedures, with imaging representing the most promising approach to assess intracerebral processes in vivo. On the other hand, they also explain why the development of new diagnostic and therapeutic means is particularly complicated and why small animal imaging may bear a great potential for the development and evaluation of new diagnostic as well as therapeutic approaches, particularly for neurodegenerative disorders. The advent of sophisticated technologies for imaging small animals offers an opportunity to circumvent some of the limitations mentioned above. In the following chapter, the development, application and evaluation of these techniques will be illustrated by examples of the selected applications of CNS tracers in rodent imaging studies of: (a) transporters and receptors, (b) inflammation and (c) neurodegenerative protein aggregation pathologies. Using point (c) as an example, the entire process from the development and evaluation of new tracers, selection of suitable animal models and performance of in vivo small animal studies will be extensively explained.

33.2 Specific Properties of Small Molecular Tracers 33.2.1 Uptake into the Brain The brain is separated from the circulatory system by a membrane composed of capillary endothelial cells, referred to as the BBB (Reese and Karnovsky 1967;

G. Henriksen and A. Drzezga

Brightman and Reese 1969). This diffusion barrier is formed by tight junctions between adjacent brain capillary endothelial cells with absence of fenestrations and low transcytosis. Under normal physiological conditions, diffusion of polar molecules between the circulation and brain interstitial fluid is restricted (for review, see Eyal et al. 2009). Transporters are located at the luminal and abluminal membranes of endothelial cells and choroid plexus epithelial cells and transfer several molecules, including amino acids, glucose and hormones from blood to brain, as well as in the reverse direction (for reviews, see Eyal et al. 2009; Grover and Bennet 2009). In general, targeting neurotransmitter systems and b-amyloid deposits with tracers for nuclear imaging after administration of a tracer to the peripheral vascular system requires diffusion of the tracer in question through an intact BBB.

33.2.2 Lipophilicity A passive diffusion of a tracer through the BBB requires a relatively small molecular weight (<450–500 g/mol) and a pronounced lipophilic character. For ensuring a high penetrability of the BBB, the number of H-bond donors is preferably £3 and number of H-bond acceptors preferably £6 (Silverman 2004). A preferred measure to indicate the lipophilicity of a compound is by the partition coefficient between octanol and phosphate buffered saline (Poct/PBS). Generally, the values of logPoct/PBS of the CNS tracers established for imaging with PET and SPECT are in the range 1.5–4 (Waterhouse 2003; Wong et al. 2003; Henriksen et al. 2008). The efficacy of diffusion across the BBB of structurally related compounds, e.g. a series of tracers, is a parabolic function of the logPoct/ value (Waterhouse 2003 and references cited PBS therein), with values of logPoct/PBS in the range of 2–3 resulting in the highest uptake. The optimum value of lipophilicity for reaching a high initial brain uptake varies among different compound classes. A higher lipophilicity, logPoct/PBS > ~4, usually restricts the initial brain uptake due to a lowered availability of the tracer in plasma as a result of high binding to plasma proteins and increased tracer uptake in the lung and liver. In the CNS, the degree of lipophilicity has an effect on the amount of non-specific binding and thus the signal to background ratio. In order to be practically

33 Imaging in Neurology Research II: PET Imaging of CNS Disorders

useful for quantification of the concentration of target, the time-dependent count rate information collected by the PET camera must be converted into figures relating to binding parameters. Since the signal detected by the PET camera represents total radioactivity, it includes specific and non-specific binding as well as free tracer in the tissue and a blood volume component. To designate the contributions from each of these parameters, reference has to be made either to the arterial concentration of the tracer (corrected for radioactive metabolites) or to a reference region within the brain with no or negligible specific binding of the tracer in question, hence providing a kinetic estimate of non-specific binding plus free tracer concentration (as well as a blood volume component), which is used as an input. A determination of the concentration of the target receptors in quantitative units relies on deriving the ratio of specifically bound tracer to the concentration of free tracer at equilibrium by compartmental modelling (for reviews see: Ichise et al. (2001); Gunn et al. (2002); Laruelle et al. (2002); Innis et al. (2007)). Generally, within a given compound class, the nonspecific binding to CNS increases with higher values of logP, although other, presently not identified, factors also control the amount of non-specific binding. The value of the free fraction of the tracer not bound to the target is estimated by the free fraction of tracer in plasma. Thus, tracer binding to plasma proteins must be reversible and measurable. A high fraction of free tracer in plasma binding allows for a more reliable determination of its value, which favours compounds of relatively low values of lipophilicity (logPoct/PBS < 3).

33.2.3 Metabolism and Fate of the Label For CNS tracers, a high metabolic stability of the tracer in the brain or at least a negligible ratio of metabolites to intact tracer in brain is preferable, although correction of radioactivity in the brain borne by metabolites can be performed (Halldin et al. 1995; Ma et al. 2003). The preferred properties of CNS tracers are (a) a low uptake in the brain of metabolites generated in the pheriphery, (b) high stability of the tracer in brain and, for determination of the metabolite-corrected plasma input function by tracer kinetic modelling, (c) the fraction of intact tracer in plasma

501

sufficiently high for allowing a reliable determination of its value.

33.2.4 Mass Effects and Tracer Preparation The concentration of a target in the CNS gives indications to the affinity required for obtaining a specific signal, useful for depiction and quantification. Neuro receptor targets successfully imaged in mice and rats have concentrations of ~10 fmol/mg protein and above (Sadée et al. 1982; Kung et al. 1989; Hall et al. 1990; Cunningham et al. 1991; Hume et al. 1998, 2007) as measured by in vitro ligand binding assays, which corresponds to estimated concentration in wet tissue (Kung and Kung 2005) of >~1 nmol/kg. Affinity constants of these tracers are in the range of 10 pM–1 nM. Currently, small animal PET scanners have a limited sensitivity for detection of radioactivity. The volume resolution element is reduced in size in going from human to mouse, from around 1 cm to 1 mm. To preserve the number of registered counts per image resolution element and thereby obtain a comparable signal to background level as that provided by clinical PET scanners, a significantly higher dose of radioactivity per unit body weight is administered for imaging mice and rats in small animal PET scanners. Despite the much smaller size of the animals under study, the amount of radioactivity injected is not proportional to the size of the animal as the number of counts per resolution element must remain approximately the same. This obviously leads to an administration of a significantly higher chemical amount of the compound in question per unit body weight for animal imaging that can be limiting for the validity of tracer principles; i.e. occupancy of less than 5% of the total available sites in order to ensure linearity of the analytical methods employed for calculation of the concentration of target receptors. Consideration of mass effects is particularly important for targets with a low concentration, as the use of high affinity tracers combined with restrictions to the achievable specific activity can easily lead to the tracer occupying a significantly higher fraction of available binding sites (for a detailed discussion of mass effects, see Kung and Kung 2005). This may in turn lead to a

502

diminished specific signal and is limiting for the applicability of tracer principles. For imaging in rodents, especially in mice, challenges are presented also for the formulation of the radiopharmaceutical. The weight and blood volume of a mouse are about 0.03% relative to that of an average human. To preserve the hemodynamic conditions in a mouse after an intravenous injection, the injected volume should be less than 0.2 mL. This limitation restricts the applicability of the routines established for preparation of a given tracer for application in humans, as they are normally formulated in a larger volume (>10 mL). Thus, imaging small animals with a PET tracer often requires either an isolation of the tracer from these solutions, i.e. by means of a solid-phase extraction and subsequent reformulation of the tracer in a small volume (<0.5–1 mL), or a separate preparation of the tracer. Further, even with a separate synthesis that is technically adapted to the imaging of small animals, the availability of tracers with shortlived radionuclides is limiting the turnover of experiments. With 11C as the label on CNS tracers with a high affinity to target, a full kinetic analysis may correspond to 2–3 half-lives of the radionuclide. This is limiting investigations with 11C-tracers – at practicable values of specific radioactivity – for targets of low concentration to one investigation from a single preparation unless more than one animal can be scanned simultaneously.

33.3 Imaging CNS Transporters and Receptors in Rodent Models of Human Disease In human Alzheimer’s disease, Parkinson’s disease (PD) and Huntington’s disease, the several mechanisms of signalling in the cerebral cortex are altered, although the causal role of alterations to these systems in the development of these diseases is still to be clarified. Further, although several rodent models of altered expression or function of receptors and transporters are now available (for reviews, see: Curtis et al. 2007; Heng et al. 2008; Jenner 2008), these models reproduce conditions and characteristics of the pathology associated with human disease, but generally differ in causes and mechanisms relative to that of human conditions. Despite these limitations, the evaluation of tracers for imaging of human diseases can benefit from an evaluation of the sensitivity of the tracer in question for detection of changes to the

G. Henriksen and A. Drzezga

availability of the specific target. An overview of selected findings from in vivo PET imaging of CNS tracers in rodents is presented in Table 33.1.

33.4 Imaging PBBR on Inflammation Sites in the CNS of Rodents At current, the role of activated microglia in the brain of patients affected by neurodegenerative disorders is not completely understood. Some studies point to the primary role of secretion of neurotoxic substances, eventually leading to neuronal death (GonzalezScarano and Baltuch 1999; Benveniste et al. 2001 and references cited therein). On the other hand, the phagocytosis of b-amyloid is held to be a primary mechanism for clearing Ab from the CNS (Wilcock et al. 2004; Masliah et al. 2005). Similar to the role of imaging b-amyloid plaques, the imaging of activated microglia may help to clarify pathomechanisms and evaluate experimental therapies. The expression of peripheral benzodiazepine receptors (PBBR) is significantly upregulated as a result of the activation of microglia. A recent study demonstrated a specific binding of the tracer R-[11C]PK11195, in APPSwe/PSEN1DeltaE9 mice correlating with immunohistochemical findings of labelled activated microglia (Venneti et al. 2009). Despite some limitations of R-[11C] PK11195 for PET investigations, these results indicate a potential of PBBR as a target for imaging in Tg models of AD. A moderate, but still significant increase in R-[11C]PK11195 binding has been observed in patients with AD compared with that found in healthy controls (Cagnin et al. 2001). PBBR imaging in rats with another tracer was reported recently (Chauveau et al. 2009).

33.5 Imaging Aggregation Pathology in Neurodegenration 33.5.1 Protein Aggregation as a Basic Pathology of Neurodegeneration In the recent years, considerable progress has been made in the identification of the pathological processes underlying neurodegenerative disorders.

503

33 Imaging in Neurology Research II: PET Imaging of CNS Disorders Table 33.1 Selected examples of in vivo PET imaging in rodents Tracer

O

11CH O 3

Cl

Primary target

Findings

References

Post-synaptic D2/D3 receptors

Increased binding in 6-hydroxydopamine-lesioned striatum of rats (PD model)

Pellegrino et al. (2007)

Reduced binding in quinolinic acid-treated rats (HD model)

Moresco et al. (2008)

Decreased binding in both ventral and dorsal striatum of rats after amphetamine treatment

Pedersen et al. (2007)

N

N H OH Cl [11C]raclopride

Increased binding in neurodegen- Sossi et al. (2009) eration levodopa-derived synaptic dopamine levels in 6-hydroxydopamine model of PD

CH3O H3CO

18F

Haloperidol 30 min before intravenous injection of [18F] fallypride blocked tracer accumulation in the striatum by >95%

Honer et al. (2004)

Reduced binding in the nucleus accumbens of highly impulsive rats (ADHD model)

Dalley et al. (2007)

Reduced binding in striatum of rats treated with methamphetamine

Tokunaga et al. (2009)

Increased binding in striatum of rats from dopamine depletion effected by reserpine and a-methyl-para-tyrosine

Seneca et al. (2008)

DAT.

Decreased binding in striatum of rats treated by 6-hydroxydopamine

Pellegrino et al. (2007)

DAT

Reduced binding in the 1-methyl-4-phenyl-1,2,3, 6-tetrahydropyridine (MPTP) mouse model of PD

Honer et al. (2006)

Post-synaptic D2/D3 receptors

O N

N H

[18F]fluoro-fallypride

Post-synaptic D2/D3 receptors

H311CO N H

HO HO

[11C]MNPA

H311C N

CO2CH3 F H [11C]β-CFT

18F

N CO2CH3

[18F]FECNT

Cl (continued)

504

G. Henriksen and A. Drzezga

Table 33.1 (continued) Tracer NH2

Primary target

Findings

References

SERT

Decreased binding in thalamus of rats from increased 5-HT concentration (induced by tranylcypromine)

Lundquist et al. (2007)

SERT

Dose dependent inhibition of 4-[18F]ADAM uptake by paroxetine and decreased 4-4-[18F]-ADAM uptake with 5,7-DHT lesioning

Ma et al. (2009)

m and k-Opioid receptors

Increased binding in the amygdala, hippocampus, somatosensory cortex of rats from noxious stimulus

Hjornevik et al. (2010)

mGluR5

Abolished binding in mGlu5Rknock-out mouse

Ametamey et al. (2006)

mGluR5

Increased binding contralaterally to 6-hydroxydopamine lesion in striatum, hippocampus, and cortex

Pellegrino et al. (2007)

GABAa

Reduced binding in rats subjected to the amygdala kindling model of epilepsy

Liefaard et al. (2009)

S CN H3C

N

11CH 3

[11C]DASB

N

NH2 S 18F

4-[18F]ADAM

N

H

H O

HO

CH3

CH3 OH

H311CO [11C]PEO

N

N

O11CH3

[11C]ABP688

N H311C

[11C]MPEP N

CO2Et

N N

F O

11CH 3

[11C]flumazenil

505

33 Imaging in Neurology Research II: PET Imaging of CNS Disorders Table 33.1 (continued) Tracer O N N

Primary target

Findings

References

PBBR on activated microglia

Binding correlated with activated microglia in APP/PS1 transgenic mice (AD model)

Venneti et al. (2009)

PBBR on activated microglia

Specific uptake in neuroinflammatory sites induced by AMPA administration to the striatum of rats

Chauveau et al. (2009)

11CH 3

Cl

(R)-[11C]PK11195

N N

O(CH2)218F

N O NEt2

[18F]DPA-714

Numerous types of pathologies are known to be involved in the development and progression of these disorders, such as neuronal dysfunction, inflammation, loss of number and function of synapses, changes regarding transmitter and receptor status and finally neuronal death and brain atrophy. However, today, the increased production, misfolding and pathological aggregation of peptides or proteins and their deposition in the brain are discussed as common causal factors, early involved in the pathogenesis of most neurodegenerative disorders (Selkoe 2003). As a consequence, Alzheimer’s disease and other forms of neurodegenerative disorders are today most commonly regarded as protein aggregation disorders. Apparently, diminished clearance or altered production of specific proteins and their subsequent misfolding may lead to aggregation of these proteins: first to soluble oligomers, then to protofibrils and finally to insoluble fibrillar aggregations in the brain (Alafuzoff 1992). Prominent examples of characteristic protein aggregations are the extracellular deposits of b-amyloid plaques and intracellular neurofibrillary tangles containing tau protein, both well-known hallmarks of AD, as well as a-synuclein-based Lewy bodies, found in Dementia with Lewy bodies (DLB) and PD. Finally, the DNA-binding protein-43 (TDP-43) has been recently identified to be the key aggregation pathology of tau-negative ubiquitin-positive lesions in patients with frontotemporal lobar degeneration (FTLD-U) (Neumann et al. 2006).

Although the proof of some of these protein aggregations in histopathological post-mortem examination of brain tissue can be regarded pathognomonical for some disorders (such as b-amyloid plaques and neurofibrillary tangles for AD), it has to be kept in mind that a considerable overlap between the pathology can be observed in different neurodegenerative disorders (for review, see Drzezga 2008). In the recent years, several genetic mutations have been associated particularly with familial forms of neurodegenerative disorders. Interestingly, these mutations were found exactly in genes encoding for the proteins, which in aggregated form are considered pathological hallmarks of these disorders. This includes mutations in genes encoding the amyloid precursor protein (APP). These mutations lead to altered affinities of the cleaving enzymes BACE1 and g-secretase to its substrate APP, leading to production of increased and/or more aggregative forms of b-amyloid (Radde et al. 2008). Furthermore, mutations in the presenilin-1 and -2 genes (PSEN1 and PSEN2) can also lead to familial AD. It has been discussed that the presenilins function as a part of the catalytic subunit of the g-secretase complex, which leads to cleavage of APP to b-amyloid. The majority of the PSEN1/2 mutations appear to shift the cleavage ratio towards the longer b-amyloid-variant, Ab42, which is more fibrillogenic than the shorter Ab40. These mutations are however found only in the rare familial forms of AD and not in the much more common sporadic forms. The strongest genetic factor associated with

506

sporadic AD identified so far is a polymorphism of the APOE-gene (the APOE e4-positive genotype), which leads to an increased susceptibility to the disease. Apart from AD, mutations in the gene encoding the tau-protein (MAPT) have been identified in FTD with Parkinsonism linked to chromosome 17 (FTDP17), mutations of the progranulin gene (PGRN) leading to a loss of its function in FTLD-U and mutations in the TARDBP-gene encoding the TDP-43 protein have been identified in amyotrophic lateral sclerosis (ALS) (Gotz and Ittner 2008). In patients with familial PD, mutations in the gene encoding a-synuclein and several other genes have been identified (Rocken stein et al. 2007). These findings of genetic mutations in neurodegenerative disorders have further strengthened the hypothesis that the mentioned protein aggregation pathologies do actually represent a primary causal event, initiating the neurodegenerative process of many disorders, and it has been speculated that other abnormalities commonly observed in neurodegeneration, such as changes in receptor or transmitter status, inflammation, neuronal dysfunction and death, represent downstream consequences, rather than the original trigger of disease (e.g. in the “amyloid hypothesis” of AD) (Hardy and Higgins 1992; Selkoe 2008). It has to be considered that sometimes the described mutations are found only in the relatively rare familial forms of neurodegeneration and not in the frequent sporadic forms, even though the pathological aggregations in the brain are often highly similar (e.g. in AD).

33.5.2 Animal Models of Neurodegenerative Protein Aggregation Whereas some pathologies of the brain can be readily mimicked in animals, e.g. cerebral ischemia, models of neurodegeneration are more difficult to generate. The finding that mutated genes can be identified as a causal factor leading to familiar forms of neurodegeneration has inspired the generation of transgenic animal models. These models represent the currently most elegant approach to induce progressive development of neurodegenerative core pathologies in animals without the need for invasive intervention in the living animal. Particularly for small animal imaging studies, the

G. Henriksen and A. Drzezga

numerous transgenic rodent models of neurodegeneration available today may be highly valuable. Among these transgenic animals are numerous mouse models of AD pathology, which overexpress some of the mutated forms of the APP that are causing the autosomal dominant inheritable forms of familial AD (Radde et al. 2008). These mutations lead to deposition of parenchymal and vascular amyloid deposits in the brains of the animals, similar to those detected in the brain of AD patients. In addition to these models, several double transgenic models exist, which, in addition to mutated APP, also express mutated PSEN1/2 and some of which are prone to more accelerated deposition of ß-amyloid in the animal brain (Holcomb et al. 1998). Whereas these transgenic mice display some behavioural features similar to AD such as memory loss, it has been critically mentioned that tau neurofibrillary tangles as the other pathological hallmark of AD do not develop in these animals, and no major neuronal loss is observed up to very advanced stages of plaque deposition. This shows that overexpression of mutated human APP and/or PSEN1/2 is not sufficient for simulating the full pathological picture of AD in mice. For an overview, see http://www.alzforum.org/ res/com/tra/. Apart from amyloid pathology, transgenic animals have been developed expressing human tau-mutations such as the MAPT mutation P301L, which is found in FTDP-17 (Gotz and Ittner 2008; Rockenstein et al. 2007). It has been shown that these models develop neurofibrillary tangles and nerve-cell dysfunction and loss in vivo. Also, they exhibit glial pathology and develop motor and behavioural disturbances (similarly to human tauopathies such as in FTDP-17) (Rockenstein et al. 2007). Several pathologic aspects detected in these models such as tau-aggregation, synapse loss or inflammation are also important pathological features of AD. Interestingly, in models carrying APP/PSEN mutations leading to b-amyloid deposition, no AD-typical development of neurofibrillary tangles has been found, vice versa in animal models expressing tau-mutations, no typical b-amyloid aggregation pathology has been observed. These findings raise questions about the causal interconnection between the two pathologies. However, recently, triple-transgenic animals have been developed expressing mutant forms of APP, PSEN 1 and Tau in one animal. These animals display b-amyloid deposition as well as tau-pathology and thus can be regarded to more closely resemble the pattern of

33 Imaging in Neurology Research II: PET Imaging of CNS Disorders

human AD pathology (Oddo et al. 2003a, b). Several other forms of mouse models of AD have been developed to prove/examine selective disease mechanisms. These include b-secretase (BACE)-deficient mice, APOE models, models of axonal transport function models etc. (Gotz and Ittner 2008). The deposition of a-synuclein in form of Lewy bodies has been associated with PD and DLB. Animal models have been developed, overexpressing wildtype a-synuclein under the regulatory control of the PDGF-b promoter (Rockenstein et al. 2007). Similarly to human pathology, these models have been demonstrated to develop motor deficits, dopaminergic loss and formation of inclusion bodies. Other mouse models have been established expressing human (h)a-synuclein. It has been discussed that a-synuclein expression in these Tg mice mimics some neuropathological alterations observed in DLB and other synucleopathies, such as multiple system atrophies (MSAs). Also models expressing mutant forms of the a-synuclein gene (such as the A53T mutation, which is found in rare hereditary forms of Parkinsons disease) have been developed and shown to develop inclusions, behavioural deficits and synaptic alterations with some differences depending on the promoter used (Rockenstein et al. 2007). Tg mice that express a-synuclein under the control of oligodendroglial promoters have been reported to demonstrate accumulation of a-synuclein in oligodendrocytes with demyelination and degeneration, especially in the spinal cord and, thus, to mimick some features of MSA (Rockenstein et al. 2007; Yazawa et al. 2005). The potential applications of these models in the context of molecular imaging of dementia are fourfold: (1) to test new molecular imaging methods in vivo, (2) to correlate the imaging in vivo signal with histopathological ex vivo/in vitro information, (3) to gain insights about pathomechanisms of the disease in vivo and (4) to evaluate the effect of experimental treatments directed towards prevention or reversal of b-amyloid deposition in the brain (e.g. b- and g-secretase inhibitors, vaccination strategies) (Citron 2004). The fact that many of these transgenic mouse models develop protein deposition pathologies in similar forms as observed in patients can be regarded as another proof of the significance of these mutated genes and the associated protein pathologies in the pathogenesis of the respective neurodegenerative disorders. However, it has to be kept in mind that the transgenic animals

507

may only represent incomplete models of disease, which only express one specific pathologic aspect of the actual disorder. Additionally, a distinct overlap of pathologies can be observed between different forms of neurodegeneration in humans. Frequently, the models are based on the hereditary forms of disease, which may show significant differences compared to the sporadic disease. Some of the transgenic animal models do show the hallmark pathology without the neurodegenerative consequences observed in humans. Most of the existing models should therefore rather be regarded as models of specific pathological features rather than of a particular disease. This fact can also represent an advantage for scientific experiments, not only a limitation (Radde et al. 2008). Another limitation can be found in the fact that soluble forms of the associated protein aggregations are currently considered to be more toxic than the corresponding deposits. The proof of existence and quantification of these soluble aggregations in the brain of transgenic animals is obviously not trivial. Finally, it should be emphasised that findings in animal models can not always be directly translated to humans. For example, therapeutic approaches have often been found to be working in animal models but showed less or no success when applied in humans. These limitations are applicable also to imaging studies. For example, the tracer [11C]PIB which is currently the most established tracer for PET imaging of betaamyloid plaques in humans was at first unsuccessful when tested in transgenic animal models of AD (Klunk et al. 2005).

33.5.3 Development of Tracers for In Vivo Imaging of b-Amyoid Plaque Early attempts to nuclear imaging tracers for b-amyloid plaque were based on Chrysamine G (CG) and Congo red (CR), which were known to possess a high affinity for b-amyloid fibrils (Klunk et al. 1995) and be useful in fluorescent staining of plaques in postmortem brain sections. The carboxylic function in these two structures is deprotonated at physiological pH, yielding ionic species, which prevents a high brain uptake in mice (Han et al. 1996; Zhen et al. 1999). Other, more chemically distant analogues of CG composed of styrylbenzene salicylic acids, such as X-34, provided similar results, i.e. high in vitro binding

508

G. Henriksen and A. Drzezga

affinity to aggregated Ab and low brain uptake in rodents (Styren et al. 2000; Klunk et al. 2002). A breakthrough in the field of in vivo imaging b-amyloid was realised by providing charge-neutral derivatives of Thioflavin-T. Two classes of compounds possessing a high brain uptake in rodents were provided through 2-[4¢-(methylamino)phenyl]benzothiazoles (BTAs) (Mathis et al. 2002, 2003) and amino-phenylimidazo[1,2-a]-pyridines (Kung et al. 2002; Zhuang et al. 2003). (for review, see Henriksen et al. 2008). Among the BTAs, N-[11C-methyl]6-OH-BTA-1 11 ([ C]PiB) was identified to have a suitable high affinity for b-amyloid plaques combined with a fast clearance from normal brain (Mathis et al. 2003), and [11C]PiB is currently the most thoroughly characterised PET tracer for Ab imaging in humans (Fig. 33.1).

33.5.4 Imaging b-Amyloid Plaques in Mouse Models of AD Tg mice overexpressing the same mutations as in familiar AD cases have been shown to develop b-amyloid deposits similar to those in AD brain and could be beneficial for preclinical investigations of in vivo uptake kinetics and determination of specific binding to Ab.

However, despite the advances in PET technology to image small animals, the in vivo detection of Ab in Tg mice has been challenging despite the demonstration of a high content of cortical Ab plaques in different transgenic mouse models by histopathological, biochemical and in vitro binding studies (Kawarabayashi et al. 2001; Klunk et al. 2005). By means of various ex vivo analyses, several PET tracers for b-amyloid plaque imaging were demonstrated to be taken up in brain regions of Tg mice containing b-amyloid plaque (Kung et al. 2002, 2004; Zhang et al. 2005; Klunk et al. 2005). In contrast, two early mPET studies concluded that Ab plaque imaging with [11C]PIB in transgenic mice was not feasible (Klunk et al. 2005; Toyama et al. 2005) and speculated that a lower concentration of high affinity binding sites in Tg mice as compared to humans was the major reason for their negative results. Indeed, the binding potential (BP) of [3H]PIB to Tg brain homogenates from APP/PS1 mouse models has been calculated to be approximately eight times lower compared to a human AD brain (BP = 78 vs. 636) (Klunk et al. 2005). Still, based on general experience with PET tracers, a BP > 10 should be highly sufficient for detecting a specific signal in vivo (Mathis et al. 2003). Thus, factors different from the low BP may have obscured the detection of potentially small differences

S N N Thioflavin-T H N

11CH 3 NH

HO

11CH 3

S

CH3

N N HO

H

PIB

SB-13

O

18F

3

BAY94,172 (florbetaben)

NC

18F O

18F O N BF-168

HO

C H3

S

18F

N

NH

N 3'F-PIB (flutemetamol)

Fig. 33.1 Compounds with affinity to b-amyloid plaques

H

N FDDNP

CN

509

33 Imaging in Neurology Research II: PET Imaging of CNS Disorders

in tracer retention between Tg and Ctl mice, probably due to methodological limitations and the lack of Tg mice suitable for mPET imaging of Ab plaques. Especially, a high variability of Ab plaque deposition in previous Tg mouse models may have resulted in low absolute differences between tracer retention in Tg and Ctl mice. Maeda and coworkers demonstrated that in vivo [11C]PIB imaging using mPET in a transgenic mouse model may be generally feasible (Maeda et al. 2007). However, this study was constrained by the necessity of using [11C]PIB with very high specific activity in mice of a very old age, limiting the reproducibility of this approach and the applicability, particularly for longitudinal follow-up studies. In a recent work from our laboratory, we selected a newly developed double transgenic mouse model that demonstrated excellent prerequisites for imaging Ab plaques in vivo (Willuweit et al. submitted). This APP/ PS1 mouse model allows breeding to homozygosity

a

b

Fig. 33.2 Example of interconnected imaging methods for depiction of b-amyloid plaques in a PS1/APP double transgenic mouse. (a) In vivo mPET/MRI with [11C]PIB. (b) Ex vivo analysis by means of digital autoradiography of a brain section from

and thus provides a measurable signal at young age. Furthermore, we established an individual MRI-guided region-of-interest (ROI) definition, implementing a reference region for the assessment of specific binding, similar to the strategies applied in human studies. We utilised [11C]PIB with a specific activity comparable to that routinely used in humans for PET imaging (Manook et al. submitted). In animals down to 9 months of age, we demonstrated specific [11C]PIB uptake in mPET corresponding to individual regional Ab plaque deposition, as verified by autoradiographical and neurohistological examination (Manook et al. submitted). In sum, a generally reproducible experimental set-up for small animal [11C]PIB imaging of amyloid plaques in a double transgenic mouse model of AD is now provided. The established protocol allows accurate quantification of regional tracer uptake even in young mice (<50% of their average life span) and thus may be suitable for follow-up studies (Fig. 33.2).

c

d

the same animal killed 1 h after administration of [3H]PIB. (c) In vitro fluorescence microscopy after Thioflavin-T staining of brain sections. (d) Immunofluorescence with anti-Ab antibodies; Ab1–40 (green) and Ab1–42 (red)

510

33.6 Summary and Outlook Pathologies of the CNS, particularly the various neurodegenerative disorders, represent an increasing burden of today’s society. Understanding, diagnosis and therapy of these disorders are still limited, mainly caused by two factors: (a) the complex nature of the diseases and (b) the specific properties of the brain such as the BBB, limiting access to brain tissue for scientific, diagnostic and therapeutic purposes. Modern small imaging technologies offer a unique opportunity to overcome several of the mentioned limitations at once: Specific small molecule tracers that are able to pass the BBB and show a favourable binding to the target in question can be labelled for PET imaging and used to determine changes in the presentation and function of receptors, transporters and enzymes as a function of pathological state, or by detecting the presence of unique markers of disease. However, an analysis of the tracer distribution alone does not reveal the underlying mechanisms responsible for a change in the specific binding of the tracer. The nuclear imaging methodology is just visualising the spatial and temporal distribution of the tracer; the quality of the information in the signal depends in full on the characteristics of the labelled tracer. Thus, a derivation of cellular and sub-cellular processes directly from a PET image is speculative, and data from complementary methods are required for validation of the PET data. Obtaining specific quantitative information on the expression of a target in question is a prerequisite for optimal planning of molecular or targeted therapies, as well as in the evaluation of treatment and for obtaining information on the mechanisms of action of these applications. Investigations of molecular mechanisms of the interactions between a tracer and target in question, and an independent analysis and a validation of the target as a disease specific marker would both benefit from the addition of information obtained by complementary investigations of tracertarget binding. Compared to other applications of nuclear imaging methods, e.g. in oncology, it is more complex to meet this requirement for CNS tracers due to limited access to tissue samples for verification and validation of the measured PET signal. State-of-the-art imaging technology and sophisticated animal models (e.g. transgenic) of different aspects of neurological/neurodegenerative disorders do now allow the correlation between the in vivo imaging

G. Henriksen and A. Drzezga

signal with ex vivo information, e.g. on density and distribution of the actual target. Particularly in the field of neuroscience, this may be highly valuable for improving the understanding of the imaging signal, to develop and improve new tracers meeting specific demands and to learn about the pathology in the animal model itself. Finally, it may help in developing optimised tools for diagnosis and therapy in humans. The combination of PET with structural imaging methods such as CT and MRI can already be regarded as mandatory for correlating the information of the distribution of the tracer to an anatomical region in the CNS. In the future, the field of CNS tracer development would strongly benefit from interconnecting to other small animal imaging methods, e.g. in vivo fluorescent techniques, multiphoton imaging, fluorescence tomography or fMRI for obtaining information on the kinetics of uptake at a higher level of resolution.

References Alafuzoff I (1992) The pathology of dementias: an overview. Acta Neurol Scand Suppl 139:8–15 Ametamey SM, Kessler LJ, Honer M et al (2006) Radiosynthesis and preclinical evaluation of 11C-ABP688 as a probe for imaging the metabotropic glutamate receptor subtype 5. J Nucl Med 47:698–705 Benveniste EN, Nguyen VT, O’Keefe GM (2001) Immunological aspects of microglia: relevance to Alzheimer’s disease. Neurochem Int 39:381–391 Brightman M, Reese T (1969) Junctions between intimately apposed cell membranes in the vertebrate brain. J Cell Biol 40:648–677 Brody DL, Holtzman DM (2008) Active and passive immunotherapy for neurodegenerative disorders. Annu Rev Neurosci 31:175–193 Cagnin A, Brooks DJ, Kennedy AM et al (2001) In-vivo measurement of activated microglia in dementia. Lancet 358:461–467 Chauveau F, Van Camp N, Dollé F et al (2009) Comparative evaluation of the translocator protein radioligands 11C-DPA713, 18F-DPA-714, and 11C-PK11195 in a rat model of acute neuroinflammation. J Nucl Med 50:468–476 Citron M (2004) Strategies for disease modification in Alzheimer’s disease. Nat Rev Neurosci 5:677–685 Cunningham VJ, Hume SP, Price GR et al (1991) Compartmental analysis of diprenorphine binding to opiate receptors in the rat in vivo and its comparison with equilibrium data in vitro. J Cereb Blood Flow Metab 11:1–9 Curtis MA, Faull RLM, Eriksson PS (2007) The effect of neurodegenerative diseases on the subventricular zone. Nature Rev Neurosci 8:712–723 Dalley JW, Fryer TD, Brichard L et al (2007) Nucleus accumbens D2/3 receptors predict trait impulsivity and cocaine reinforcement. Science 315:1267–1270

33 Imaging in Neurology Research II: PET Imaging of CNS Disorders Drzezga A (2008) Basic pathologies of neurodegenerative dementias and their relevance for state-of-the-art molecular imaging studies. Eur J Nucl Med Mol Imaging 35(suppl 1):S4–S11 Eyal S, Hsiao P, Unadkat JD (2009) Drug interactions at the bloodbrain barrier: fact or fantasy? Pharmacol Ther 123:80–104 Games D, Adams D, Alessandrini R et al (1995) Alzheimer-type neuropathology in transgenic mice overexpressing V717F beta-amyloid precursor protein. Nature 373:523–527 Gonzalez-Scarano F, Baltuch G (1999) Microglia as mediators of inflammatory and degenerative diseases. Annu Rev Neurosci 22:219–240 Gordon HL, MN McGowan E et al (1998) Accelerated Alzheimer-type phenotype in transgenic mice carrying both mutant amyloid precursor protein and presenilin 1 transgenes. Nat Med 4:97–100 Gotz J, Ittner LM (2008) Animal models of Alzheimer’s disease and frontotemporal dementia. Nat Rev Neurosci 9:532–544 Grover A, Bennet LZ (2009) Effect of drug transporters on the volume of distribution. AAPS J 11:250–261 Gunn RN, Gunn SR, Turkheimer FE et al (2002) Positron emission tomography compartmental models: a basis pursuit strategy for kinetic modeling. J Cereb Blood Flow Metab 22:1425–1439 Hall H, Wedel I, Halldin C et al (1990) Comparison of the in vitro receptor binding properties of N-[3H]methylspiperone and [3H]raclopride to rat and human brain membranes. J Neurochem 55:2048–2057 Halldin C, Swahn CG, Farde L et al (1995) Radioligand disposition and metabolism-key information in drug development. In: Comar D (ed) PET for drug development and evaluation. Kluwer, Boston, pp 55–61 Han H, Cho CG, Lansbury PT Jr (1996) Technetium complexes for the quantification of brain amyloid. J Am Chem Soc 118:4506–4507 Hardy JA, Higgins GA (1992) Alzheimer’s disease: the amyloid cascade hypothesis. Science 256:184–185 Heng MY, Detloff PJ, Albin RL (2008) Rodent genetic models of Huntington disease. Neurobiol Dis 32:1–9 Henriksen G, Yousefi BH, Drzezga A et al (2008) Development and evaluation of compounds for imaging of beta-amyloid plaque by means of positron emission tomography. Eur J Nucl Med Mol Imaging 35(suppl 1):S75–S81 Hjornevik T, Schoultz BW, Marton J et al (2010) “http://www. ncbi.nlm.nih.gov/pubmed/20662879” Spinal long-term potentiation is associated with reduced opioid neurotransmission in the rat brain. Clin Physiol Funct Imaging 30:285–293 Holcomb L, Gordon MN, McGowan E et al (1998) “http://www. ncbi.nlm.nih.gov/pubmed/9427614” Accelerated Alzheimertype phenotype in transgenic mice carrying both mutant amyloid precursor protein and presenilin 1 transgenes. Nat Med 4:97–100 Honer M, Brühlmeier M, Missimer J et al (2004) Dynamic imaging of striatal D2 receptors in mice using quad-HIDAC PET. J Nucl Med 45:464–470 Honer M, Hengerer B, Blagoev M, Hintermann S et al (2006) Comparison of [18F]FDOPA, [18F]FMT and [18F]FECNT for imaging dopaminergic neurotransmission in mice. Nucl Med Biol 33:607–614 Honer M, Stoffel A, Kessler LJ et al (2007) Radiolabeling and in vitro and in vivo evaluation of [18F]-FE-DABP688 as a

511

PET radioligand for the metabotropic glutamate receptor subtype 5. Nucl Med Biol 34:973–980 Hsiao K, Chapman P, Nilsen S et al (1996) Correlative memory deficits, Abeta elevation, and amyloid plaques in transgenic mice. Science 274:99–102 Huang Y, Bae SA, Zhu Z et al (2005) Fluorinated diaryl sulfides as serotonin transporter ligands: synthesis, structure-activity relationship study, and in vivo evaluation of fluorine-18- labeled compounds as PET imaging agents. J Med Chem 48: 2559–2570 Hume SP, Gunn RN, Jones T (1998) Pharmacological constraints associated with positron emission tomographic scanning of small laboratory animals. Eur J Nucl Med 25:173–176 Hume SP, Lingford-Hughes AR, Nataf V et al (2007) “http://www. ncbi.nlm.nih.gov/pubmed/17488881” Low sensitivity of the positron emission tomography ligand [11C]diprenorphine to agonist opiates. J Pharmacol Exp Ther 322:661–667 Ichise M, Meyer JH, Yonekura Y (2001) An introduction to PET and SPECT neuroreceptor quantification models. J Nucl Med 42:755–763 Innis RB, Cunningham VJ, Delforge J et al (2007) Consensus nomenclature for in vivo imaging of reversibly binding radioligands. J Cereb Blood Flow Metab 27:1533–1539 Jakob-Roetne R, Jacobsen H (2009) Alzheimer’s disease: from pathology to therapeutic approaches. Angew Chem Int Ed Engl 48:3030–3059 Jenner P (2008) Functional models of Parkinson’s disease: a valuable tool in the development of novel therapies. Ann Neurol 64(suppl 2):S16–S29 Kawarabayashi T, Younkin LH, Saido TC et al (2001) Agedependent changes in brain, CSF, and plasma amyloid (beta) protein in the Tg2576 transgenic mouse model of Alzheimer’s disease. J Neurosci 21:372–381 Klunk WE, Debnath ML, Pettegrew JW (1995) Chrysamine-G binding to Alzheimer and control brain: autopsy study of a new amyloid probe. Neurbiol Aging 16:541–548 Klunk WE, Bacskai BJ, Mathis CA (2002) Imaging Abeta plaques in living transgenic mice with multiphoton microscopy and methoxy-X04, a systemically administered Congo red derivative. J Neuropathol Exp Neurol 61:797–805 Klunk WE, Lopresti BJ, Ikonomovic MD et al (2005) Binding of the positron emission tomography tracer Pittsburgh compound-B reflects the amount of amyloid-beta in Alzheimer’s disease brain but not in transgenic mouse brain. J Neurosci 25:10598–10606 Kung MP, Kung HF (2005) Mass effect of injected dose in small rodent imaging by SPECT and PET. Nucl Med Biol 32:673–678 Kung HF, Pan S, Kung MP et al (1989) In vitro and in vivo evaluation of [123I]IBZM: a potential CNS D2 dopamine receptor imaging agent. J Nucl Med 30:88–92 Kung MP, Hou C, Zhuang ZP et al (2002) IMPY: an improved thioflavin-T derivative for in vivo labeling of beta-amyloid plaques. Brain Res 956:202–210 Kung MP, Hou C, Zhuang ZP et al (2004) Characterization of IMPY as a potential imaging agent for beta-amyloid plaques in double transgenic PSAPP mice. Eur J Nucl Med Mol Imaging 31:1136–1145 Laruelle M, Slifstein M, Huang Y (2002) Positron emission tomography: imaging and quantification of neurotransporter availability. Methods 27:287–299

512 Liefaard LC, Ploeger BA, Molthoff CF et al (2009) Changes in GABAA receptor properties in amygdala kindled animals: in vivo studies using [11C]flumazenil and positron emission tomography. Epilepsia 50:88–98 Lundquist P, Roman M, Syvänen S et al (2007) Effect on [11C] DASB binding after tranylcypromine-induced increase in serotonin concentration: positron emission tomography studies in monkeys and rats. Synapse 61:440–449 Ma Y, Kiesewetter DO, Lang L et al (2003) Determination of [18F]FCWAY, [18F]FP-TZTP, and their metabolites in plasma using rapid and efficient liquid-liquid and solid phase extractions. Nucl Med Biol 30:233–240 Ma KH, Huang WS, Kuo YY et al (2009) Validation of 4-[18F]-ADAM as a SERT imaging agent using micro-PET and autoradiography. Neuroimage 45:687–693 Maeda J, Ji B, Irie T et al (2007) Longitudinal, quantitative assessment of amyloid, neuroinflammation, and anti-amyloid treatment in a living mouse model of Alzheimer’s disease enabled by positron emission tomography. J Neurosci 27:10957–10968 Manook A, Henriksen G, Platzer S, Yousefi BH, Huisman M, Voss A, Settles M, Neff F, Busch R, Schoor M, Willuweit A, von der Kammer H, Velden J, Wester HJ, Schwaiger M, Drzezga A. Practicable in vivo microPET imaging of amyloid plaques in a new double transgenic mouse model of Alzheimer’s disease (Submitted to PLoS 09-PONE-RA08856R1-A) Masliah E, Hansen L, Adame A et al (2005) Abeta vaccination effects on plaque pathology in the absence of encephalitis in Alzheimer disease. Neurology 64:129–131 Mathis CA, Wang Y, Holt DP et al (2003) Synthesis and evaluation of 11C-labeled 6-substituted 2-arylbenzothiazoles as amyloid imaging agents. J Med Chem 46:2740–2754 Mathis CA, Bacskai BJ, Kajdasz ST et al (2002) “http://www. ncbi.nlm.nih.gov/pubmed/11814781” A lipophilic thioflavin-T derivative for positron emission tomography (PET) imaging of amyloid in brain. Bioorg Med Chem Lett 12:295–298 McGowan E, Sanders S, Iwatsubo T et al (1999) Amyloid phenotype characterization of transgenic mice overexpressing both mutant amyloid precursor protein and mutant presenilin 1 transgenes. Neurobiol Dis 6:231–244 Moresco RM, Lavazza T, Belloli S et al (2008) Quinolinic acid induced neurodegeneration in the striatum: a combined in vivo and in vitro analysis of receptor changes and microglia activation. Eur J Nucl Med Mol Imaging 35:704–715 Neumann M, Sampathu DM, Kwong LK et al (2006) “http:// www.ncbi.nlm.nih.gov/pubmed/17023659” Ubiquitinated TDP-43 in frontotemporal lobar degeneration and amyotrophic lateral sclerosis. Science 314:130–133 Oddo S, Caccamo A, Shepherd JD et al (2003a) Triple-transgenic model of Alzheimer’s disease with plaques and tangles: intracellular Abeta and synaptic dysfunction. Neuron 39: 409–421 Oddo S, Caccamo A, Kitazawa M et al (2003b) Amyloid deposition precedes tangle formation in a triple transgenic model of Alzheimer’s disease. Neurobiol Aging 24:1063–1070 Pedersen K, Simonsen M, Østergaard SD et al (2007) Mapping the amphetamine-evoked changes in [11C]raclopride binding in living rat using small animal PET: modulation by MAO-inhibition. Neuroimage 35:38–46

G. Henriksen and A. Drzezga Pellegrino D, Cicchetti F, Wang X et al (2007) Modulation of dopaminergic and glutamatergic brain function: PET studies on parkinsonian rats. J Nucl Med 48:1147–1153 Radde R, Duma C, Goedert M et al (2008) The value of incomplete mouse models of Alzheimer’s disease. Eur J Nucl Med Mol Imaging 35(suppl 1):S70–S74 Reese T, Karnovsky M (1967) Fine structural localization of a blood-brain barrier to exogenous peroxidase. J Cell Biol 34:207–217 Rockenstein E, Crews L, Masliah E (2007) Transgenic animal models of neurodegenerative diseases and their application to treatment development. Adv Drug Deliv Rev 59: 1093–1102 Sadée W, Perry DC, Rosenbaum JS (1982) [3H]diprenorphine receptor binding in vivo and in vitro. Eur J Pharmacol 81:431–440 Selkoe DJ (2003) Folding proteins in fatal ways. Nature 426:900–904 Selkoe DJ (2008) Soluble oligomers of the amyloid beta-protein impair synaptic plasticity and behavior. Behav Brain Res 192:106–113 Seneca N, Zoghbi SS, Skinbjerg M et al (2008) Occupancy of dopamine D2/3 receptors in rat brain by endogenous dopamine measured with the agonist positron emission tomography radioligand [11C]MNPA. Synapse 62:756–763 Silverman RB (2004) The organic chemistry of drug design and drug action, p65-66. Elsevier, London Sossi V, Dinelle K, Topping GJ et al (2009) Dopamine transporter relation to levodopa-derived synaptic dopamine in a rat model of Parkinson’s: an in vivo imaging study. J Neurochem 109:85–92 Sturchler-Pierrat C, Abramowski D, Duke M et al (1997) Two amyloid precursor protein transgenic mouse models with Alzheimer disease-like pathology. Proc Natl Acad Sci U S A 94:13287–13292 Styren SD, Hamilton RL, Styren GC et al (2000) X-34, a fluorescent derivative of Congo red: a novel histochemical stain for Alzheimer’s disease pathology. J Histochem Cytochem 48:1223–1232 Thal DR, Capetillo-Zarate E, Del Tredici K et al (2006) The development of amyloid b protein deposits in the aged brain. Sci Aging Knowledge Environ 8:re1 Tokunaga M, Seneca N, Shin RM et al (2009) Neuroimaging and physiological evidence for involvement of glutamatergic transmission in regulation of the striatal dopaminergic system. J Neurosci 29:1887–1896 Toyama H, Ye D, Ichise M et al (2005) PET imaging of brain with the beta-amyloid probe, [11C]6-OH-BTA-1, in a transgenic mouse model of Alzheimer’s disease. Eur J Nucl Med Mol Imaging 32:593–600 Venneti S, Lopresti BJ, Wang G et al (2009) PK11195 labels activated microglia in Alzheimer’s in vivo in a mouse model using PET. Neurobiol Aging 30:1217–1226 Waterhouse RN (2003) “http://www.ncbi.nlm.nih.gov/ pubmed/14667492” Determination of lipophilicity and its use as a predictor of blood-brain barrier penetration of molecular imaging agents. Mol Imaging Biol 5:376–389 Wang JL, Parhi AK, Oya S et al (2008) 2-(2¢-((Dimethylamino) methyl)-4¢-(3-[18F]fluoropropoxy)-phenylthio)benzenamine for positron emission tomography imaging of serotonin transporters. Nucl Med Biol 35:447–458

33 Imaging in Neurology Research II: PET Imaging of CNS Disorders Wilcock DM, Rojiani A, Rosenthal A et al (2004) Passive amyloid immunotherapy clears amyloid and transiently activates microglia in a transgenic mouse model of amyloid deposition. J Neurosci 24:6144–6151 Willuweit A, Velden J, Godemann R et al (2009). “http://www. ncbi.nlm.nih.gov/pubmed/19936202?itool=EntrezSystem2. PEntrez.Pubmed.Pubmed_ResultsPanel.Pubmed_ RVDocSum&ordinalpos=1” Early-onset and robust amyloid pathology in a new homozygous mouse model of Alzheimer’s disease. PLoS One 4:e7931 Wong DF, Pomper MG (2003) “http://www.ncbi.nlm.nih”.gov/ pubmed/14667490” Predicting the success of a radiopharmaceutical for in vivo imaging of central nervous system neuroreceptor systems. Mol Imaging Biol 5:350–362 Yazawa I, Giasson BI, Sasaki R et al (2005) Mouse model of multiple system atrophy alpha-synuclein expression in

513

oligodendrocytes causes glial and neuronal degeneration. Neuron 45:847–859 Zhang W, Oya S, Kung MP et al (2005) F-18 stilbenes as PET imaging agents for detecting beta-amyloid plaques in the brain. J Med Chem 48:5980–5988 Zhen W, Han H, Anguiano M et al (1999) Synthesis and amyloid binding properties of rhenium complexes: preliminary progress towards a reagent for SPECT imaging of Alzheimer’s disease brain. J Med Chem 42:2805–2815 Zhuang ZP, Kung MP, Wilson A et al (2003) Structure-activity relationship of imidazo[1, 2-a]pyridines as ligands for detecting beta-amyloid plaques in the brain. J Med Chem 46:237–243

Imaging in Neurology Research III: Focus on Neurotransmitter Imaging

34

Wynne K. Schiffer

Contents

34.1 Introduction

31.1 Introduction............................................................. 515

Small animal PET is one of the most recently developed and rapidly evolving areas of biomedical imaging, with applications ranging from drug development to the study of molecular events associated with normal and abnormal thought and behavior. The aim of this chapter is to describe how the development of techniques for measuring neurotransmitter concentrations has enabled neurochemical physiology and function (particularly of the brain) to be imaged directly using PET. This compares to traditional PET approaches that have used receptor imaging to measure the static concentration of proteins localized to a particular tissue in the brain. The development of neurotransmitter mapping approaches has opened up the possibility to study the brain in ways that were previously unachievable, with results that are universally comparable. PET is now a well-established modality both in clinical and experimental investigations of the body. In this chapter, we will assume familiarity with the basic principles of PET data acquisition and image formation. Those requiring a comprehensive description of PET theory are referred to the relevant chapters in this volume. This chapter takes the following format: after this introduction, the history and premise is described followed by theoretical and practical issues relevant to the design and implementation of preclinical neurotransmitter mapping studies. In the next section, the premise behind measuring neurotransmitter activity with PET is presented, and the sensitivity of PET to measure this phenomenon is introduced. The complexity of neurotransmission in biological systems is explored, and experimental paradigms, which have been designed to investigate this, are described and evaluated. In the following section, methods and

34.2 History and Premise............................................... 516 34.2.1 Neurochemical Transmission as a Fundamental Molecular Phenomenon................... 516 34.2.2 Clinical Uses of Neurotransmitter Activation Studies..................................................... 517 34.2.3 Relationship Between Extracellular Dopamine Levels and D2 Radiotracer Binding......... 520 34.3 Theoretical and Practical Issues............................ 523 34.3.1 Choice of a Ligand and Target Neurotransmitter System.......................................... 524 34.3.2 Kinetic Modeling of Chosen Radiotracer................. 526 34.3.3 Design of a Challenge Paradigm.............................. 529 34.3.4 Paired Bolus vs. Continuous Infusion...................... 529 34.3.5 Timing of Challenge................................................. 530 34.3.6 Behavioral Imaging Protocols in Small Animals with PET.................................................... 531 34.3.7 Limitations to Absolute Quantification.................... 533 34.3.8 Mass.......................................................................... 533 34.3.9 Data Corrections....................................................... 534 34.4 Conclusion.................................................................. 535 References............................................................................ 536

W.K. Schiffer Laboratory for Behavioral and Molecular Neuroimaging, Feinstein Institute for Medical Research North Shore – LIJ Health System, 350 Community Drive, Manhasset, NY 11030, USA [email protected]

F. Kiessling and B.J. Pichler (eds.), Small Animal Imaging, DOI: 10.1007/978-3-642-12945-2_34, © Springer-Verlag Berlin Heidelberg 2011

515

516

caveats for measuring neurotransmitter function in awake and anesthetized animals are described in detail and compared. It is hoped that these insights will help to define the mechanisms of neuropathology in psychiatric disease and allow evaluation of therapies aimed at treating these pathophysiological states.

34.2 History and Premise 34.2.1 Neurochemical Transmission as a Fundamental Molecular Phenomenon It is known that neurochemical transmission in biological systems is a complex phenomenon, influenced directly by both the presynaptic and postsynaptic architecture, and that its measurement can provide a large amount of information about the organization of the brain in normal and diseased states. Neurotransmitters are neither static nor isolated in their distribution. In fact, it is through interactions with other neurochemically distinct systems that the central nervous system (CNS) performs its vital role in sustaining life. Exclusive quantitative capabilities intrinsic to PET make this technology a suitable experimental tool to measure not only the regional distribution of specific receptors and their subtypes, but also the dynamic properties of neurochemical systems and their inherent influence on related neurotransmitter pathways. The ability to investigate dynamic properties in a noninvasive and reproducible manner provides a powerful tool that can extend our current knowledge of brain

W.K. Schiffer

function. Coupled with innovative paradigms including pharmacologic and behavioral manipulations, knowledge derived from PET studies can greatly advance our understanding of normal and abnormal brain function. The overriding power of PET is derived from its molecular specificity. This unique specificity has been used to image specific proteins and to quantify their number and kinetics. With the acquisition of dynamic data and the application of models, one can ask: How many receptors are there in a given volume of tissue? What is the rate of trapping or the rate at which a given molecule enters and exits a given tissue? More recently, PET researchers began using the behavior of radioactive molecules to infer the activity of natural or endogenous chemicals that are not radioactive. With this dynamic approach, one can ask: When a chemical in the brain is perturbed, how does its level change, and how do functionally related chemicals behave? This imaging by inference of molecules that are not radioactive is possible because of the predictable competition between endogenous and exogenous ligands for specific binding sites. The premise is that competition occurs between an injected radioligand and a neurotransmitter for binding to a limited population of receptors (Fig. 34.1). This premise is usually explored and defined under a theoretical framework referred to as the occupancy model (Laruelle 2000). This model predicts that challenges that increase synaptic neurotransmitter concentrations will result in increased receptor occupancy by a given neurotransmitter, and thus reduced availability of receptors for binding of the radiotracer. On the other hand, manipulations that decrease synaptic neurotransmitter concentrations will reduce receptor occupancy by the neurotransmitter, and increase receptor

Fig. 34.1 Illustration of the premise behind imaging dopamine with [11C]-raclopride. [11C]-raclopride provides an indirect index of synaptic dopamine levels because it competes with dopamine for occupancy of D2 receptor sites

517

34 Imaging in Neurology Research III: Focus on Neurotransmitter Imaging

availability to radiotracer binding. In practical terms, a subject is injected with a neurotransmitter receptor radioligand, such as a dopamine receptor ligand, after a pharmacological challenge such as amphetamine, which stimulates dopamine release. If dopamine release is stimulated, then more of it is released into the synapse. Released intrasynaptic dopamine competes for receptor sites with an appropriately chosen radioligand, usually [11C]-raclopride. Hence, there is a decrease in the amount of [11C]-raclopride binding observed. An alternate approach is to measure decreases in synaptic dopamine using compounds that inhibit its synthesis, such as alpha-methyl-paratyrosine (AMPT) – a drug that temporarily depletes dopamine stores from the presynaptic terminals. This results in increased radiotracer binding, as illustrated in Fig. 34.1. By comparing the specific binding at baseline with that during the challenge scan, one can demonstrate the dynamic activity of a subset of neurochemically distinct neurons: in this example, a dopaminergic pathway (Fig. 34.2).

Baseline

34.2.2 Clinical Uses of Neurotransmitter Activation Studies From a historical perspective, suggestions that certain dopamine receptor radiotracers may be used to image changes in dopamine levels began in 1989, with publication of ex vivo data demonstrating the sensitivity of D2 receptor radiotracers to changes in endogenous dopamine levels (Ross and Jackson 1989a, b; Seeman et al. 1989). Indications that this sensitivity could also be observed in vivo using PET rapidly followed, when increased displacement of the D2 tracer [18F]N-methylspiroperidol was observed following administration of the anticholinergic benztropine to baboons (Dewey et al. 1990). This finding was subsequently confirmed by applying the same technique to investigate amphetamine-induced DA release (Dewey et al. 1991). Landmark investigations in humans followed swiftly afterwards; data showing decreases in binding of the D2 receptor PET radiotracer [11C]-raclopride in

AMPT pretreat

AMPH pretreat

Dopamine [11C]-raclopride

Synaptic dopamine concentrations

[11C]-raclopride scan female papio anubis baboons, 16−19 kg

Fig. 34.2 Schematic representation of dopamine depletion and dopamine enhancement experiments. At baseline, an unknown proportion of dopamine D2 receptors are occupied by synaptic dopamine (shaded area), and only a fraction of receptors are unoccupied and available to binding by [11C]-raclopride (redtagged circles). After depletion of endogenous dopamine with

alpha-methyl-para-tyrosine (AMPT), more receptors are available to [11C]-raclopride binding. Enhancing endogenous dopamine with drugs such as amphetamine (AMPH) increases synaptic dopamine levels and decreases the available binding sites for [11C]-raclopride. Volumetric images were generated with the PMOD software package (www.pmod.com)

518

W.K. Schiffer

response to administration of amphetamine were published in 1992 (Farde et al. 1992), and similar results were subsequently obtained following administration of the DA reuptake inhibitor GBR-12909 or amphetamine (Dewey et al. 1993). This approach has been supported by a variety of protocols using [11C]-raclopride, some of which are summarized in Table 34.1. The development of PET radiotracers for the dopamine system has also made it possible to directly study disorders such as schizophrenia (Farde et al. 1986; Wong et al. 1986), Parkinson’s Disease (Brooks 1997), Tourette’s syndrome (Wong et al. 1998), drug addiction (Schiffer et al. 2007a), and ADHD (Jucaite et al. 2005; Krause et al. 2003; Rosa Neto et al. 2002) among others. More recently, this imaging approach has been applied to behavioral studies, where the challenge is not a drug but a behavioral task. In a hallmark PET experiment, Koepp and coworkers demonstrated that endogenous dopamine is released in the human striatum during a goal-directed motor task (Koepp et al. 1998a). Each of their volunteers received two [11C]-raclopride PET scans (total injected dose of 16–20 mCi), one during an experimental challenge condition (video game that involved learning to

navigate a tank for a monetary incentive) and one under baseline conditions (blank screen). Subjects played the video game from 10 min before to 50 min after the [11C]-raclopride injection (Koepp et al. 1998a). Binding of [11C]-raclopride to dopamine receptors in the striatum was significantly reduced during the video game compared with baseline, consistent with increased release and binding of dopamine to its receptors. Interestingly, the reduction of [11C]-raclopride binding in the striatum was positively correlated with the performance level during the video task and was greatest in the ventral striatum (Koepp et al. 1998a). This approach, namely correlating the amount of specific transmitter release with individual behavioral parameters, provides a powerful novel means for understanding the role of neurotransmitters in different behaviors. We have used [11C]-raclopride imaging in awake rodents to demonstrate a correlation between methamphetamine (METH)-induced increases in locomotion and decreases in [11C]-raclopride binding, and another relationship between dopamine release and the degree of craving in an animal model of cocaine abuse (Patel et al. 2008; Schiffer et al. 2009b). These studies are described in detail in the next section. Since the

Table 34.1 Pharmacological studies mapping dopamine transmission in humans with [11C]-raclopride Subjects

Infusiona

BP derivationb

% DBP

References

AMPH

14

B+CI

SRTM

−17.8 ± 13.8

Martinez et al. (2003)

AMPH

16

PB

SRTM

−10.6 ± 5.1

Oswald et al. (2005)

AMPH

7

PB

GA

−14.8 ± 11.5

Drevets et al. (2001)

AMPH

12

PB

SRTM

−13 ± 5

Cardenas et al. (2004)

6 h post

12

PB

SRTM

−18 ± 8.7

Cardenas et al. (2004)

Methylphenidate

7

PB

GA

−18.4 ± 8.7

Wang et al. (1999)

Nicotine

20

B + CI

SRTM

−25.9 ± 10.7

Brody et al. (2004)

Nicotine

10

PB

SRTM

−21.2 ± 4.7

Barrett et al. (2004)

Psilocybin

7

PB

GA

−19.9 ± 5.5

Vollenweider et al. (1999)

AMPT

6

PB

SRTM

+13.9 ± 5.9

Verhoeff et al. (2003)

AMPT

6

PB

SRTM

+18.5 ± 3.0

Verhoeff et al. (2001)

AMPT

14

B+CI

SRTM

+11.1 ± 4.4

Martinez et al. (2009)

Drug challenge Dopamine enhancement

Dopamine depletion

Tyrosine depletion 8 PB SRTM +11.8 ± 11.9 B + CI bolus plus constant infusion of [11C]-raclopride; PB paired bolus b BP binding potential; SRTM simplified reference tissue model; GA graphical analysis a

Leyton et al. (2004)

519

34 Imaging in Neurology Research III: Focus on Neurotransmitter Imaging

publication of this original finding (Koepp et al. 1998a), there have been a large number of studies in this field, employing a number of different approaches. Some of these are summarized in Table 34.2. As a new direction for clinical PET, animal studies employing similar measures can reveal much about the biology and the animal model itself. In principle, and given an appropriate radiotracer, neurotransmitter competition techniques can be applied to any neurochemical pathway. Radiotracers are currently being developed, or are already available, for imaging dopamine, serotonin (5-HT), acetylcholine, GABA, histamine, norepinephrine, and other synapses. However, they are not all sensitive to the levels of endogenous neurotransmitter. Only a few studies have evaluated the use of ligands to study other neurotransmitter systems. These include the noradrenergic (Tyacke et al. 2005), cholinergic (Carson et al. 1998; Nishiyama et al. 2001), opioid (Koepp et al. 1998b; Zubieta et al. 2003), and GABA systems (Frankle et al. 2009). Studies with [11C]-flumazenil, classically not regarded as sensitive to endogenous GABA, are now showing that a model other

than the occupancy model explains the sensitivity of this ligand to endogenous GABA (Frankle et al. 2009). Results from studies using serotonergic ligands such as [18F]-MPPF and [11C]-WAY-100635 are inconsistent with regard to sensitivity to endogenous serotonin. Phys iological changes in 5-HT levels did not induce measurable effects in humans (Rabiner et al. 2002), and studies in the rat, using the 5-HT-releasing agent and reuptake inhibitor fenfluramine, have shown that larger increases in 5-HT levels were not always effective either (Hume et al. 2001; Zimmer et al. 2002). Using the 5-HT2 antagonists [18F]-altanserin and [11C]-MDL 100907, no displacement was seen after citalopram (an SSRI) or fenfluramine-induced increases in 5-HT levels, respectively (Hirani et al. 2003; Pinborg et al. 2004). Binding of the serotonin transporter ligand [11C]-DASB was found to be decreased after increases in 5-HT (Ginovart et al. 2003). However, ligand binding was not affected after a decrease in 5-HT concentration (Talbot et al. 2005). Thus, functional neurochemical maps of the brain can be created, which complement more general functional maps that measure glucose metabolism or blood flow.

Table 34.2 Nonpharmacological (behavioral) studies mapping dopamine transmission in humans with [11C]-raclopride Task/stimulus

Subjects

Infusiona

BP derivationb

% DBP

Reference

Video game

8

PB

SRTM

−13.9 (p < 0.02)

Koepp et al. (1998a)

Meditation

8

PB

SRTM

−7.9 (p = 0.013)

Kjaer et al. (2002)

Foot movement

8

PB

GA

−30.1 (p < 0.05)

Ouchi et al. (2002)

Food stimulus

10

PB

GA

−7.5 (NS)

Volkow et al. (2002)

Motor task

8

PB

SRTM

−8.8 (p < 0.05)

Goerendt et al. (2003)

Food

7

PB

SRTM

−12.4 (p = 0.03)

Small et al. (2003)

Gambling

10

PB

SRTM

−8.7 (p < 0.01)

Zald et al. (2004)

Arithmetic task

5

PB

SRTM

−9.6 (p = 0.02)

Pruessner et al. (2004)

Arithmetic task

14

B+CI

Ratio

+0.8 (NS)

Montgomery et al. (2006)

Conditioned stimulus

9

PB

SRTM

−23.3 (p = 0.02)

Boileau et al. (2007)

Expectation of pain

4

B+CI

Ratio/GA

−11.0 (p < 0.02)

Scott et al. (2007)

Uncertain rest

13

PB

GA

+23.1 (p < 0.05)

Yoder et al. (2008)

Arithmetic task

7

PB

SRTM

−9.8 (p = 0.03)

Soliman et al. (2008)

Spatial memory

7

PB

SRTM

−9.0 (p < 0.01)

Sawamoto et al. (2008)

Spatial planning 10 B+CI Ratio −2.7 (NS) 11 B + CI bolus plus constant infusion of [ C]-raclopride; PB paired bolus b BP binding potential; SRTM simplified reference tissue model; GA graphical analysis a

Lappin et al. (2009)

520

34.2.3 Relationship Between Extracellular Dopamine Levels and D2 Radiotracer Binding The site of the neurochemical competition in question is the synapse. Synapses are specialized structures that allow one neuron to communicate with another. The transfer of chemical information takes anywhere from 2 ms for fast (e.g., glutamate) synapses to tens or hundreds of milliseconds (or occasionally many seconds) for slow (e.g., serotonin or muscarinic acetylcholine) synapses. As described above, neurotransmitter mapping paradigms rely on the competition between endogenous neurotransmitter and radiotracer for the same postsynaptic receptors. Using the example of drug-induced increases in neurochemical transmission, several conditions are assumed to exist during PET measurements. First, it is expected that endogenous neurotransmitter is released into the synapse when action potentials arrive at presynaptic terminals. It is also expected that the amount of endogenous neurotransmitter that is released from a drug must be greater than at rest, and the increased competition for receptors must be sufficient enough to cause a measurable decrease in radiotracer binding. Dopamine enhancement with amphetamine. Although the concept of competition between endogenous neurotransmitter and injected radioligand makes sense intuitively, rigorous scientific proof supporting the validity of this technique is required. One way these neurotransmitter activation studies have been validated is with parallel or simultaneous measures of dopamine levels using microdialysis. Microdialysis is an in vivo sampling technique that requires a relatively invasive surgical placement of a probe into a specific region of the brain, at the end of which a dialysis membrane collects fluid samples for analysis by high-pressure liquid chromatography (HPLC). Microdialysis samples can be collected every minute if necessary, but newer systems usually collect every 5–10 min. This collection rate dictates the amount of dialysate in the sample for analysis, as the flow through the dialysis membrane is usually fixed to be around 2 mL/min, giving only a 10 mL sample for a 5-min collection rate. This becomes important to small animal PET data because the microdialysis sampling times can be designed to coincide with PET collection frames. However, like PET, the smaller “bins” or collection times used, the noisier the data. Because of

W.K. Schiffer

the dimensions of microdialysis probes (usually ~200– 300 mg), microdialysis does not sample the neurotransmitter directly from the synapse but detects compounds relatively far away from the site of release. Therefore, it is important to remember that no matter how direct microdialysis is as a sampling technique, it only measures dopamine from outside the synapse, in an area referred to as the extrasynaptic space. As mentioned above, PET measures mostly dopamine fluctuations inside the synapse. With activation, the extrasynaptic space may experience a larger change in dopamine levels than does the synaptic space (discussed in (Fisher et al. 1995)). Nevertheless, for the most part, it is widely held that extrasynaptic (or “extracellular fluid,” (ECF) sometimes abbreviated, ECF) neurotransmitter concentrations reflect those inside the synapse. In Fig. 34.3, a male Sprague Dawley rat is pictured inside a microPET R4 system. This animal is undergoing a microdialysis experiment inside the small animal scanner. Dopamine levels are sampled from the striatum prior to and during a baseline scanning session, followed immediately by a second scan where amphetamine was administered intravenously with the second [11C]-raclopride injection (these studies are called “coinjection” experiments, as opposed to pretreating the animal subjects just prior to the second scan). After the second PET scan and before the animal is removed from the PET gantry, a small amount (30 mCi) of

Fig. 34.3 Combined microPET/microdialysis experiments allow direct measurements of dopamine levels simultaneously with [11C]-raclopride displacement. These types of studies provide an index of amount of dopamine release required to produce measurable changes in [11C]-raclopride binding

34 Imaging in Neurology Research III: Focus on Neurotransmitter Imaging

radiotracer is flushed through the microdialysis probes, and a short image is acquired. This facilitates ROI placement on the PET images in precisely the same location of the striatum that coincides with the microdialysis samples. In Fig. 34.4, combined microPET and microdialysis studies illustrate the effects of changing dopamine levels on [11C]-raclopride binding.

Fig. 34.4 Radioactivity distribution images of [11C]-raclopride in the rat striatum during a combined PET/microdialysis study. In (a), a baseline scan shows [11C]-raclopride binding at baseline, before dopamine stimulation. In (b), a challenge scan with AMPH coadministered with [11C]-raclopride significantly reduces binding. In (c), 30 mCi of radiotracer infused through the microdialysis probes prior to moving the animal out of the gantry facilitates ROI placement. In (d), the probe images are overlaid on the radioactivity distribution image. Note the lack of harderian gland activity in the challenge scan. Unpredictable glandular uptake of radiotracers can impact small animal PET experiments, where the glands are close to the brain

521

The time activity of [11C]-raclopride and dopamine from experiments such as that pictured above is demonstrated in Fig. 34.5. This illustrates how increasing dopamine activity can affect the time activity of [11C]-raclopride binding. Several laboratories have combined PET with microdialysis; two of which use primates in clinical PET systems (Breier et al. 1997;

522

W.K. Schiffer

Fig. 34.4 (continued)

Dopamine levels (% of baseline)

[11C]-raclopride binding (%ID/cc)

0.7

0.7 BASE STR BASE CB

0.6 0.5

0.5

0.4

0.4

0.3

0.3

0.2

0.2

0.1

0.1

0.0

0.0

1000 900 800 700 600 500 400 300 200 100 0 10

20

30

40

50

Time (in minutes)

AMPH CB

1000 900 800 700 600 500 400 300 200 100 0

ECF DA

0

AMPH STR

0.6

60

0

10

20

30

40

50

60

Time (in minutes)

Fig. 34.5 AMPH increases dopamine levels and decreases the availability of binding sites for [11C]-raclopride. [11C]-raclopride time activity curve (TAC) data from the striatum (STR) and cerebellum (CB) during a baseline scan (BASE) and after an AMPH challenge. Microdialysis traces are shown below their respective TACs. Quantitation of these studies is described in Fig. 34.6.

Importantly, AMPH-induced dopamine release peaks and declines to baseline within 30 min, while it has been shown that [11C]-raclopride binding potential (BP) is decreased as long as 5 h after AMPH in a rodent, and 12 h in a human (Cardenas et al. 2004 #20345)

Tsukada et al. 2002) while our studies are performed in rodents using small animal PET (Morris et al. 2008; Schiffer et al. 2003, 2005). While in a PET scanner, animal subjects are injected with a drug, which changes dopamine levels in the

synapse. The dopamine receptor binding is assessed with PET while the extracellular concentration of dopamine is simultaneously measured by microdialysis. One goal of these types of studies is to estimate the sensitivity of [11C]-raclopride as an assay for dopamine

34 Imaging in Neurology Research III: Focus on Neurotransmitter Imaging

fluctuations. Sensitivity, in this case, means the amount of dopamine required to cause measurable changes in radiotracer binding. This information permits inferences to be made about the magnitude of dopamine responses in clinical PET studies where microdialysis is not feasible (for example, see (Marenco et al. 2004)). The quantitation of this data and the arrival at a measure of [11C]-raclopride binding potential (BP) is described below. The increase in dopamine levels from amphetamine relative to the decrease in [11C]-raclopride BP can be estimated that for every 1% decrease in [11C]-raclopride BP, there is a 50% increase in dopamine levels. This is consistent with results in primates, which found that for every one percent decrease in [11C]-raclopride BP, there was a 44% increase in extracellular dopamine (Breier et al. 1997; Laruelle et al. 1997) or a 100% increase in extracellular dopamine (Tsukada et al. 1999a). Dopamine depletion with AMPT. Previous studies have shown that acute depletion of endogenous dopamine increases the availability of binding sites for [11C]-raclopride and other benzamide radiotracers (see Table 34.2). The reduction in endogenous dopamine increases the percentage of receptors available to bind to the radiotracer by reducing the pool of receptors occupied by dopamine. Reserpine and AMPT are pharmacological agents commonly used in animal models to induce dopamine depletion. Reserpine acts by blocking the vesicular monoamine transporter, which leads to less dopamine contained in presynaptic vesicles, leading to reduced levels of synaptic dopamine (Quinn et al. 1959). AMPT diminishes dopamine by inhibiting tyrosine hydroxylase, the first of the two enzymes involved in the synthesis of dopamine from the amino acid tyrosine (Sjoerdsma et al. 1965). It has been shown with both of these drugs that decreases in dopamine lead to a subsequent reduction in dopamine levels and an increase in [11C]-raclopride or 3H-raclopride binding ranging from 11 to 60% (Ginovart et al. 1997; Park and Park 2000). Tissue concentrations of dopamine decrease sharply over 1 h after the administration of AMPT and level off around 40% of baseline levels (Park and Park 2000). Microdialysis studies show a similar reduction to 40–60% of baseline levels (Laruelle et al. 1997; Watanabe et al. 2005). A paradigm has been developed to use AMPT in both PET and single photon emission computed tomography (SPECT) studies (Laruelle et al. 1997) to image the percent of D2 receptors occupied by endogenous

523

dopamine. It has been demonstrated that the tissue level of AMPT determines the degree of enzymatic inhibition, and the duration for which this tissue level is maintained determines the magnitude of DA depletion (Spector 1966). In this way, investigators can estimate the levels of endogenous dopamine that was depleted. These studies have shown that when D2 receptors are unmasked with AMPT, increased radiotracer binding can be observed in schizophrenic patients (Abi-Dargham et al. 2000) and cocaine addicts (Martinez et al. 2009). AMPT is preferred over other dopamine depletion strategies such as reserpine because of its acute, short lasting effects, which do not lead to compensatory increases in D2 receptor density (Laruelle et al. 1997). The effects of AMPT on [11C]-raclopride binding are shown in Fig. 34.6. Partial depletion of dopamine increases the availability of binding sites for [11C]-raclopride and other benzamide radioligands. In Fig. 34.7, the radioactivity concentrations and associated microdialysis traces are shown at baseline and after an acute dose of AMPT.

34.3 Theoretical and Practical Issues Neurotransmitter mapping with PET is based on the principle that labeled compounds can be designed to compete with endogenous neurochemical activity at receptor sites on both pre and postsynaptic terminals. In this, radiotracers can be targeted at a number of physiological functions, from presynaptic release and reuptake to postsynaptic uptake. In the midst of disease states, this information provides vital insight into the development of potential pharmacologic interventions. Further, neurochemical mapping with PET can be carefully controlled, so that depending on the kinetic properties of the tracer molecule, varying degrees of competition interfere with normal neurotransmitter activity and occupy all receptor sites, or less competitively occupy only a few of the receptor sites. Several concepts are fundamental to understanding the general foundation of mapping dynamic neurochemical interactions with PET. These are the choices of a ligand and target neurotransmitter system, kinetic modeling of chosen radiotracer, and the design of a challenge paradigm. From this, we can assay the effects of pharmaceuticals on interrelated neurotransmitter systems as well as gather information about regular and irregular endogenous states.

524

W.K. Schiffer

Fig. 34.6 Parametric images of [11C]-racloride BP at baseline and following 250 mg/kg AMPT. Parametric images and all image processing were performed using the PMOD software package

34.3.1 Choice of a Ligand and Target Neurotransmitter System The freedom implied by the ability to label a variety of compounds might lead one to raise the possibility of labeling the neurotransmitters themselves, but this is impossible for several reasons. Primarily, neurotransmitters do not cross the blood–brain barrier (BBB). Because labeled compounds must be injected or inhaled, BBB permeability is necessary. Second, the binding affinity of all amino acid and amine neurotransmitters to their specific receptors is low. In this, the fate of neurotransmitters themselves cannot be controlled (i.e., presynaptic reuptake or enzymatic

degradation), but the action of a molecule designed to bind to a specific receptor can be reliably predicted depending on the kinetics of the labeled molecule. Therefore, it is necessary to use molecules that bind to the receptor under investigation with a known affinity. The choice of a ligand should be clinically relevant, and its actions should be well documented. In this, ligands should have a known affinity for specific receptor subtypes to minimize the possibility of confounding effects or untraceable radioligand activity from promiscuous ligand delivery. Second, ligands should have a biological half-life long enough to remain in systemic circulation for the duration of the scanning period (although kinetic modeling parameters can

525

34 Imaging in Neurology Research III: Focus on Neurotransmitter Imaging

[11C]-raclopride binding (%ID/cc)

0.8

0.8 BASE STR BASE CB

0.7 0.6

0.5

0.4

0.4

0.3

0.3

0.2

0.2

0.1

0.1

0.0

0.0 ECF DA

Dopamine levels (% of baseline)

140

120

120

100

100

80

80

60

60

40

40

20

20

0

0

20

40

60

80

Time (in minutes)

AMPT CB

0.6

0.5

140

AMPT STR

0.7

0

0

20

40

60

80

Time (in minutes)

Fig. 34.7 AMPT decreases dopamine and increases the availability of binding sites for [11C]-raclopride. [11C]-raclopride TAC data from the striatum (STR) and cerebellum (CB) during a baseline scan (dynamic scan 1, DY1) and after an AMPT chal-

lenge (dynamic scan 2, DY2). Microdialysis traces are shown below their respective TACs. Radioactivity concentrations are calculated as the percentage of total injected dose per cc of tissue. Quantitation of these studies is described in Fig. 34.9

compensate for this). Finally, it is important that ligands have a conformation that can be readily labeled while retaining BBB permeability. The development of adequate PET radiotracers for targets in the CNS requires that a large number of chemical and radiochemical parameters as well as pharmacological and biological properties are fulfilled. The candidate radiotracer must be suitable for high specific activity (SA) labeling in order to avoid a significant mass of stable compound. A too-low SA may compromise the specific binding of the radiotracer as it leads to occupation of a non-negligible fraction of the target receptors. In addition, it potentially leads to toxic effects. A radiopharmaceutical that is useful for the imaging of CNS receptors after intravenous injection must possess a relevant brain uptake. Except for tracers that are actively transported across the BBB, the latter necessitates that a ligand is sufficiently lipid-soluble to permit its passive diffusion into the brain. This implies an appropriate lipophilicity (logPoct/PBS in the range 1.5–4) (Waterhouse 2003), a low

molecular weight (<500 g/mol), and an absence of active transport mechanism for the tracer out of the brain. Furthermore, tracer binding to plasma proteins must be reversible and measurable. Even though quantitative tracer kinetic modeling in PET allows using ligands whose metabolite(s) cross the BBB by correcting the dynamic plasma data for the brain uptake of metabolite(s), and a high metabolic stability in blood or at least a negligible ratio of metabolites to intact tracer in brain is preferable. Thus, the criteria for selection of lead compounds may be (a) the absence of significant amounts of labeled metabolites in blood, (b) a low BBB penetration of metabolites into the brain, (c) high stability in brain tissue, (d) fast washout of metabolites after formation from the brain, and (e) low binding of metabolites in the brain. Extensive efforts have been made in the field of radiotracer development for the in vivo investigation of biochemical processes involved in dopaminergic neurotransmission. Overall, the dopaminergic tracer with the most important clinical role is 6-[18F]

526

fluoro-l-3,4-dihydroxyphenylalanine (6-[18F]fluoroDOPA), which allows for dopamine synthesis in vivo to be studied. In contrast to 6-[18F]FDOPA, 6-[18F] fluoro-m-l-tyrosine (6-[18F]FMT) is not a substrate for catechol-O-methyltransferase (COMT) and has therefore been suggested for a more selective quantification of amino acid decarboxylase. 11C- and 18 F-labeled derivatives of tropane such as [18F] FECNT (Davis et al. 2003) and [11C]PE2I (Halldin et al. 2003 and Hirvonen et al. 2008) have shown promise for quantification of the rate of dopamine reuptake via the dopamine reuptake transporter. Among the identified subclasses of postsynaptic dopamine receptors (D1-like: D1, D5; D2-like: D2–4), the development of tracers suitable for D2 imaging has been a subject of intensive research in PET imaging. The D2/ D3 binding tracer [11C]raclopride was among the very first PET ligands available (Ehrin et al. 1985; Farde et al. 1985). In recent years, efforts to develop alternatives to [11C]raclopride have been conducted (for reviews, see (Elsinga et al. 2006) and (de Paulis 2003)), motivated by the advantages of 18F-labeled tracers, which provide more time to investigate the effect of pharmacological challenges and/or stimulation on the D2/(D3) receptor occupancy, as well as to provide tracers with higher affinity in order to allow depiction and quantification of extrastriatal (cortical) regions that have a lower receptor density. The D2/D3 antagonists [18F]fluoro-fallypride (Cropley et al. 2008; Mukherjee et al. 1995, 2002), [18F] fluoro-desmethoxy-fallypride (Mukherjee et al. 1996; Siessmeier et al. 2005), and [11C]FLB457 (Asselin et al. 2007; Halldin et al. 1995; Montgomery et al. 2007; Olsson et al. 2004) have been evaluated in humans to study extrastriatal dopamine D2/(D3) receptors. More recently, the development of D2/D3 radiotracers with agonist properties has been addressed. These radiotracers were hypothesized to allow a better differentiation of high-affinity (G-protein coupled) state of the receptors from the low-affinity (G-protein decoupled) state than achievable with D2/D3 receptor antagonist radiotracers and, therefore, to facilitate the measurement of changes in tracer binding, which are related to changes in the endogenous concentration of dopamine. 11C-(+)-4-propyl-9-hydroxynaphthoxazine ([11C]-(+)-PHNO) has been shown to distribute in the brain differentially than the frequently used D2/D3 antagonist tracers (Ginovart et al. 2006; Graff-Guerrero et al. 2008; Wilson et al. 2005) as well as relative to that of another candidate D2(D3) radiotracer with agonistic properties: (−)-N-[11C]propyl-norapomorphine ([11C]

W.K. Schiffer

NPA) (Narendran et al. 2004, 2005). Some studies have indicated that [11C]-(+)-PHNO is more sensitive to displacement from d-amphetamine-mediated-induced displacement of dopamine as compared to [11C]raclopride (Willeit et al. 2008). On the other hand, [11C]-(+)-PHNO is a D3 preferring agonist, which may account for its regional brain distribution and its potentially higher displacement by endogenous release of dopamine compared to that of [11C]raclopride and [11C]NPA (Narendran et al. 2006). However, a recent study has shown that [11C](+)-PHNO- and [3H]raclopride do not differ in their response to agonist challenge and thus not support predictions of the validity of an in vivo D2 high/low model (McCormick et al. 2008).

34.3.2 Kinetic Modeling of Chosen Radiotracer It is important to remember that the analysis of imaging data involves the use of statistical techniques that simplify (and sometimes oversimplify), while producing images that are inherently visually compelling. PET images are a composite of various signals – only one of which relates to the radiotracer of interest. In order to isolate the component of the PET signal that is due to the phenomenon of interest, we introduce a mathematical model relating the dynamics of the tracer molecule – and all its possible states – to the resultant PET image. These models allow us to extract information on binding, or delivery, or any hypothesized process, as being distinct from all the other processes that contribute to the PET signal. Because these models, like all models representing biological events, are necessarily simplified descriptions that can not fully account for all potentially relevant factors, the experimental procedures must be designed to minimize possible errors arising from limitations and/or imperfections of the kinetic model. In other words, by optimizing the experimental protocol, we can minimize the dependence of our PET measurements on the kinetic model and thus obtain a signal that more closely relates to the biology in question. Binding potential. The most commonly used measure for the quantification of dopamine release in PET imaging is a change in radiotracer BP. BP is a unitless measure of tracer binding to receptors of interest at equilibrium. It is defined as: BP = Bavail / K D ,

34 Imaging in Neurology Research III: Focus on Neurotransmitter Imaging

where Bavail is the available receptor density (the total number of unoccupied receptors), and KD is the radioligand equilibrium dissociation constant. BP can also be thought of as the ratio of “bound” to “free” radiotracer in a given region of the brain. It is important to note that this definition of BP refers to the “true” in vitro measure of Bavail/KD. In PET, measurements of BP must be made with respect to a reference concentration (the “free tracer”), leading to multiple definitions or interpretations of BP. A recent consensus on nomenclature was developed to distinguish between the various definitions of BP (Innis et al. 2007). The use of the different versions of BP depends on the experiment and the method used to calculate it. No matter which version of BP used, one must remember that it is not a direct measure of dopamine concentrations, but rather an index of receptor availability. The only methods of direct dopamine measurements from the brain are microdialysis and voltammetry. As detailed above, several groups have used a combined approach to demonstrate the feasibility of measuring dopamine activity indirectly with PET (Breier et al. 1997; Schiffer et al. 2005; Tsukada et al. 1999a). These studies have shown that we can rely on the change in BP as an indirect measure of dopamine release. One assumption is that changes in BP due to pharmacological or nonpharmacological challenges are the result of transient changes in endogenous dopamine levels, and not changes in receptor number or affinity of receptors for the radiotracer. BP becomes a powerful tool to quantifying dopamine fluctuations with PET when calculated in two different experimental conditions; rest and activation (Fisher et al. 1995). Specifically, the BP from a baseline condition is compared to the BP from an activation condition in which changes in dopamine levels are expected as the result of some behavioral of pharmacological challenge. As described below, with data from microdialysis studies, the percent change in BP can then be used as a measure of how much dopamine concentrations changed as a result of the experimental intervention, relative to baseline dopamine levels. Determination of BP is versatile. Because there are different ways to calculate BP, researchers can choose the best method for their experimental design, computational abilities, and desired research outcomes. BP calculations are also based on many assumptions, the violation of which can invalidate the measurement. Since BP is dependent on many factors including the

527

scanner used and its calibration, the method used for calculation, and the tracer used, direct comparisons of BP across studies and scanners are usually not possible. The change in BP, or DBP, is the most appropriate measure for comparing results between studies (as in Table 34.1). Simulation studies have shown that DBP is sensitive to the timing and kinetics of the radiotracer and the endogenous ligand, as discussed below. Nevertheless, it is important to remember that the only information that BP provides about dopamine is whether it increases or decreases in response to an experimental challenge. Here, we focus on the use of graphical methods to calculate BP; however, another popular approach with small animal PET is probably the simplified reference tissue model (SRTM). We use the graphical approach for practical reasons because it is amenable to imaging awake, behaving animals; however, we have shown that in traditional animal PET experiments (anesthetized animals with a full time activity curve (TAC)), the two quantitation methods give identical results (Alexoff et al. 2003). Graphical methods rely on the rearrangement of radiotracer kinetics to yield a simple, linear equation whose slope and intercept provide information on the kinetic parameters of the radiotracer. Graphical techniques have been described for both reversibly and irreversibly binding radiotracers (Gunn et al. 2001; Logan et al. 1990; Patlak et al. 1983). For reversibly binding radiotracers, the graphical technique developed by Logan et al. (1991) proposes that after some time t, the plot of striatal to cerebellar integral is linear with slope equal to the distribution volume (DV) of the radiotracer (Logan et al. 1991). This is depicted in Fig. 34.8 for the data presented above, in Fig. 34.5. This method is also sensitive to decreases in dopamine levels, such as those induced by pretreatment with AMPT. Logan et al. (1991) have described the applicability and limitations of using the graphical method for determining decreases in endogenous dopamine. The graphical analysis applied to increasing [11C]-raclopride BP is depicted in Fig. 34.9. The DV is the concentration ratio of the radiotracer in the tissue to that in the plasma (Innis et al. 2007). It can also be thought of as the capacity of the tissue to bind to the radiotracer (Logan 2003) or a linear function of free receptor concentration in the tissue (Logan et al. 1996). Using a graphical analysis with a reference region from the same scan, the ratio of DVROI to DVREF yields the distribution volume ratio (DVR;

528 BASE STR BASE CB AMPH STR AMPH CB

0.7 0.6 0.5 0.4 0.3 0.2 0.1

b

10

20

30

40

50

BASE AMPH

100

0.0 0

120

ò0TSTR(dt)/STR(T)

[11C]-raclopride binding (%ID/cc)

a

W.K. Schiffer

80

y = 2.73x - 20.9

60 40 20 0

60

0

10

Time (in minutes)

BASE CB AMPH STR AMPH CB

0.4 0.3 0.2 0.1 0.0

0

20

40

60

40

50

80

Time (in minutes)

which is the distribution volume ratio (DVR). Here, the DVR for the baseline scan is 3.41 and for the AMPH challenge scan is 2.73. BP = DVR−1, so the measured BP for the baseline condition is 2.41 and after AMPH, 1.73. This represents a 28.2% decrease in [11C]-raclopride binding

b

200

BASE AMPH

ò0TSTR(dt)/STR(T)

[11C]-raclopride binding (%ID/cc)

BASE STR

0.7 0.5

30

ò0

0.8 0.6

20

TCB(dt)/STR(T)

Fig. 34.8 Graphical analysis of [11C]-raclopride time activity data from Fig. 34.5. In (a), TACs from the striatum (STR) and cerebellum (CB) are corrected for the injected dose of radiotracer and plotted for the baseline scan (BASE, grey symbols) and AMPH challenge scan (AMPH, black symbols). In (b), a line of best fit is drawn over the linear portion of the graph, the slope of

a

y = 3.41x - 32.1

y = 3.6x - 30.9

150 y = 4.14x - 94.2

100

50 0

0

20

40

60

80

TCB(dt)/STR(T)

ò0

Fig. 34.9 Quantitation of [11C]-raclopride time activity data from Fig. 34.7 using a graphical method with the cerebellum as the reference region. In (a), TACs from the striatum (STR) and cerebellum (CB) are corrected for the injected dose of radiotracer and plotted for the baseline scan (BASE, grey symbols) and AMPT challenge scan (AMPT, black symbols). In (b), a line of

best fit is drawn over the linear portion of the graph, the slope of which is the distribution volume ratio (DVR). Here, the DVR for the baseline scan is 3.6 and for the AMPT scan it is 4.14. BP = DVR−1, so the measured BP for the baseline condition is 2.6 and after AMPT, 3.14. This represents a 20.1% increase in [11C]-raclopride binding

(Logan 2003)). Although the DV is a ratio of concentrations (and hence dimensionless), it is called a volume because it equals the volume of blood that contains the same activity as 1 mL of tissue (Carson 1996). For example, DV = 2 means that at equilibrium, the tracer in tissue is twice as concentrated as that in blood. That is, 2 mL of blood would have the same quantity of radioactivity as 1 mL of tissue. BP is related to the DVR through the equation (Logan et al. 1990):

The Logan graphical technique is easy to employ and computationally quick and does not rely on a compartment model, thus removing any problems with model selection. The use of a reference region instead of plasma input also avoids the need for arterial sampling. However, this technique has the distinct disadvantage of being susceptible to noise (Slifstein and Laruelle 2000). PET data are inherently noisy, especially in studies using isotopes with short half-lives such as carbon-11. By the end of the scanning period, the signal-to-noise ratio (SNR) is particularly low due

DVR - 1 = BAvail / K D = BP.

34 Imaging in Neurology Research III: Focus on Neurotransmitter Imaging

to radiotracer decay, an effect that is magnified in small animal imaging studies where the injected doses are low to avoid mass effects, described in the next section. Such noise has been shown to cause an underestimation of the DV and thus a lower calculated BP than what is true (Slifstein and Laruelle 2000). Additionally, underestimated DVs may have increased sensitivity to changes in blood flow (Logan 2003). Fitting methods have been developed to provide some relief for bias due to noise (Logan et al. 2001). In summary, BP is an invaluable metric in PET quantification and can be calculated through multiple methods. It is through this metric that we can measure DA release, and BP is a common endpoint for functional PET experiments in both animals and humans.

34.3.3 Design of a Challenge Paradigm The uses of neurotransmitter mapping in clinical PET studies are being demonstrated in an increasingly wide variety of applications. As these studies are extended to small animal imaging, appropriate design of challenge paradigms becomes crucial. In this next section, we will address the choice of route of radiotracer administration in a paired bolus or bolus plus continuous infusion protocol. Next, we will address the timing of the challenge relative to the start of the scan – where, in time, should the challenge occur relative to scanning for optimal sensitivity? How can this affect the measured outcome? The effects of anesthesia are briefly addressed in the next section, where we discuss paradigms that permit simultaneous measurements of brain activity and behavior in freely moving animals with PET, although a thorough review of the impact of various anesthetics and anesthetic regimens is beyond the scope of this chapter. Finally, we will address several limitations to absolute quantification of small animal PET data, including mass and quantitation issues, and how these can impact the outcome of neurotransmitter mapping experiments.

34.3.4 Paired Bolus vs. Continuous Infusion Several methodological and analytical approaches have been applied in [11C]-raclopride studies of dopamine release following behavioral challenges that

529

have different practical and methodological advantages and disadvantages. There have been two experimental protocols widely applied to study neurotransmitter changes. The first is based on a comparison of PET signals from two bolus injections of radiotracer (these are typically referred to as “paired bolus” studies). One bolus study is performed while the neurotransmitter is at steady state; a second bolus study is carried out while the endogenous neurotransmitter is undergoing an induced change (for example, an amphetamine challenge). An alternative experimental approach also involves perturbation of the neurotransmitter, but it is done only after the radiotracer has been artificially brought to a steady state in the tissue by a combination of bolus plus continuous infusion. Here, [11C]-raclopride may be administered by an initial bolus followed by constant infusion (B + CI) to maintain radiotracer equilibrium, during which both baseline and challenge data are collected (Carson et al. 1997; Watabe et al. 2000). As discussed above, the PET approach to detecting dopamine release is based on the estimation of changes in the concentration of available neuroreceptor sites (Bavail), which occur in response to the associated changes in the local neurotransmitter concentration. The kinetic behavior of [11C]-raclopride is in turn dependent on Bavail and is linear at tracer concentrations. This enables the determination of a BP, which equals the ratio of the specifically bound radioligand over the free concentration of radioligand in the brain at equilibrium. In vitro, in the absence of competing ligands, BP is equal to the density of radiotracer binding sites (Bmax) divided by the radiotracer affinity (KD) (Mintun et al. 1984). In practice, in PET studies, BP is defined as the ratio at equilibrium between specifically bound tracer and that in the free and nonspecifically bound compartments. Changes in BP in these studies are usually assumed to reflect changes in Bavail, rather than in the KD for the radiotracer, and changes in BP are assumed to reflect changes in endogenous neurotransmitter release. BP is an equilibrium concept but may be estimated from dynamic PET studies when a suitable reference region that is devoid of specific radiotracer binding sites is available. An input function describing the time course of the delivery of the radiotracer to the tissue is required for quantification of dynamic studies, but the plasma input function may be substituted by the tracer time course in the reference region itself. For [11C]-raclopride, the cerebellum may be used (Logan et al. 1996). The BI

530

technique confers significant advantage as, once equilibrium is reached, BP may be calculated as the ratio of the concentration of the radiotracer in the ROI to the concentration of the radiotracer in the reference region: (BP = (CROI−CREF)/CREF). This approach has the advantage of being quantitatively simple compared to the analysis methods applied to dynamic bolus studies (Carson 2000). It is also thought that the B + CI method is less sensitive to local or global changes in blood flow, since once equilibrium has been established, the constant levels of radiotracer in the plasma avoid any confounding effects of blood flow on specific binding (Carson 2000; Carson et al. 1993, 1997; Endres and Carson 1998; Endres et al. 1997). With paired bolus studies, transient equilibrium may be assumed when maximal values for specific binding are obtained (Farde et al. 1989); this occurs approximately 20–25 min after the bolus injection of [11C]-raclopride (Ito et al. 1998). In contrast to the B + CI approach, equilibrium is not sustained as the radiotracer begins to washout of the tissue. For these studies, BP must be derived as described above using model-based methods such as graphical analysis (Logan et al. 1990) or a compartmental kinetic analysis such as SRTM (Lammertsma and Hume 1996). With specific reference to measurement of dopamine release, both the graphical and STRM methods have been criticized on the basis that they assume that dopamine levels achieve a steady state for the duration over which BP is measured (Alpert et al. 2003), whereas it is clear from the microdialysis plots that it is not. Nevertheless, the sensitivity of these approaches has been specifically compared: following administration of amphetamine (Carson et al. 1997) or nicotine (Marenco et al. 2004) to primates, paired bolus and B + CI approaches have roughly equivalent power to detect alterations in dopamine levels. Practical advantages of paired bolus designs are that a continuous infusion system and protocol are not necessary. Also, paired bolus designs facilitate the study of freely moving animals, as continuous measurements cannot be made from animals outside tomograph (Patel et al. 2008; Schiffer et al. 2009b). The advantages of B + CI studies are quantitative simplicity and, more importantly, the requirement of a single radiotracer delivery for both baseline and challenge scans. This is especially pertinent to small animal imaging experiments, where dedicated radiotracer deliveries may not be possible; small animal

W.K. Schiffer

investigators must “piggyback” off of radiotracer deliveries intended for clinical experiments.

34.3.5 Timing of Challenge An important factor may be the shape and timing of the dopamine release curve in comparison to the radiotracer time–activity curve. Graphical analysis following bolus administration of [18F]-Nmethylspiroperidol showed that change in uptake rate is maximal for large DA peaks and slow DA clearance (Logan et al. 1991). Similar results have been obtained for the dual condition, single scan BI approach; changes in specific binding following amphetamine challenge correlate with both the height of the DA pulse (nM) and the DA clearance rate (per min), and tightest correlations are obtained when the change in specific binding is correlated against the integral of the DA pulse (mM min) (Endres et al. 1997). It is unclear at present whether DA curves obtained under all physiological stimuli will be sufficient to produce significant radiotracer displacement using this technique. Simulations performed by Morris et al. (1995) for the paired bolus approach suggest that BP changes may be maximized when the activation task is performed over a long time period and commenced on or before radiotracer administration. Similar results were obtained by Logan et al. (1991), where the largest change in [18F]-Nmethylspiroperidol uptake rate occurred when the task commenced simultaneously with radiotracer injection, a finding also replicated in [11C]-raclopride simulations of Endres and Carson (1998). Yoder et al. (2004) have further demonstrated that the change in BP can be affected by the timing of the dopaminergic response in relation to the timing of [11C]-raclopride concentration following bolus administration. Here, greater changes in BP were detected if the onset of the dopaminergic response occurred just before [11C]-raclopride administration (Yoder et al. 2004). Furthermore, the magnitude of change in BP reflected not only the magnitude of dopamine release (area under the curve) but also differences in the shape of the dopamine pulse, with blunt curves producing larger changes in BP for a given amount of dopamine released (Yoder et al. 2004). When using the paired bolus approach, it is therefore recommended that tasks start just prior to radiotracer administration and continue for a significant duration of the scan.

34 Imaging in Neurology Research III: Focus on Neurotransmitter Imaging

34.3.6 Behavioral Imaging Protocols in Small Animals with PET While FDG has been a mainstay for imaging patterns of brain metabolic activation with PET (for review, see (Schiffer et al. 2007b)), the use of neurotransmitterspecific radiotracers such as [11C]-raclopride has been largely limited to ex vivo studies (Wadenberg et al. 2001), or studies where the behavior is measured in separate animals or on separate days as the imaging experiment (Hume et al. 1995). We have developed an imaging paradigm in awake, freely moving animals that permits concurrent measurements of animal behavior and [11C]-raclopride binding (Patel et al. 2008; Schiffer et al. 2009a). This simple paradigm is possible due to the well-established kinetic profile of [11C]-raclopride and its predictable response to changes in endogenous dopamine. An illustration of the paradigm is depicted in Fig. 34.10. In this paradigm, freely moving animals receive the radiotracer through a previously placed jugular vein catheter in their home cage, during which they have absolutely no restrictions on their range or degree of movement. Following 30 min of radiotracer uptake, animals are anesthetized and scanned within the tomograph for 25 min. To test this paradigm, we first characterized the range of responses that can be measured by using high mass, low SA injections and low mass, high SA injections (Patel et al. 2008). Next, we demonstrated that these measurements were reproducible in the same animal scanned on different days. Having established this range and variability, we performed a series of studies using pharmacological and

Behavioral Imaging Protocol Awake and freely moving (25-30 min)

[11C]-raclopride

anesthesia

PET scan (25min)

EOS

patel et al., Neuroimage (2008) 41:1051-66

Fig. 34.10 Protocol for imaging freely moving animals with [11C]-raclopride. In this protocol, radiotracer uptake occurs away from the tomograph, while the animal is freely moving. Anesthesia occurs with an intravenous dose of injectable anesthetic, followed by a short PET acquisition

531

behavioral challenges. Specifically, we demonstrated that this method reliably captures increases in synaptic dopamine (decreases in striatal [11C]-raclopride binding) following a METH challenge and that this increase can be attenuated by increasing GABA levels following administration of the active enantiomer of gamma vinyl-GABA [S(+)-GVG]. Finally, we demonstrated that handling-induced stress in awake animals also reduced [11C]-raclopride binding (Patel et al. 2008), and cues associated with cocaine administration also significantly reduce [11C]-raclopride binding (Schiffer et al. 2009b). In this latter experiment, we were able to show that the change in [11C]-raclopride binding correlated with the craving exhibited by each animal subject, such that those animals that did not prefer the cocaine had smaller changes in BP in the cocaine environment compared to animals that showed a strong preference for the drug (Schiffer et al. 2009b). The use of a video camera and specialized software packages such as TopScan (Cleversys, Inc. USA) permits off-line analysis of behaviors. These software packages, most of which are designed for phenotyping mice, can quantitate an extraordinary range of behaviors, from freezing behavior to seizures to whisker twitches. Figure 34.11 shows the activity of one animal during [11C]-raclopride uptake at baseline, and then during [11C]-raclopride uptake when a small amount of cocaine (5 mg/kg) was added to the injection medium. It is evident in the figure that a dark background was used – the bedding in the animal cages is actually for designed aquariums, and mimics the bedding in the animal home cage except that it is black. This permits facile analysis by the motion detection software. In contrast to clinical PET imaging, PET data collection in small animals must involve immobilization of the animal, which is usually attained by anesthesia of the animal during the PET scanning procedure. Several issues have to be considered to minimize the confounding influence of anesthesia on tracer kinetics and distribution. The effects of anesthesia on radiotracer activity depend on both the radiotracer and the anesthetic. Ideally, anesthesia should be administered in a controllable manner and at a superficial level. For this reason, gaseous anesthetics, which provide highly controlled anesthesia that reverses rapidly, are the most common anesthetic regimen in small animals. On the other hand, injectable anesthetics like ketamine permit the

532

W.K. Schiffer

Fig. 34.11 Measures of locomotion during [11C]-raclopride uptake alone and when coinjected with cocaine (5 mg/kg). It is important to place the animals in an area of high contrast, such as the black background depicted above. This allows motion detection software programs to easily discern the activity of the animals. This data can be analyzed off-line and correlated with the imaging measures

opportunity to anesthetize many animals for studies where radiotracer administration is staggered. However, ketamine is a vasodilator that decreases systolic and dialostic blood pressure as well as heart rate in rats (Rodrigues et al. 2006), and it also significantly decreases cerebral vascular resistance (Takeshita et al. 1972). Some anesthetics are also known to target neurotransmitter receptors or transporters in a direct or an indirect manner. Isoflurane, for example, interacts directly with the GABAA receptor and was also shown to induce changes in the dopaminergic system in nonhuman primates, possibly by modulating presynaptic dopamine transporter availability (Opacka-Juffry et al. 1991; Tsukada et al. 1999b; Votaw et al. 2004). In primates, mechanical ventilation associated with anesthesia influenced the uptake of [11C]-cocaine, where the same anesthetic regimen in spontaneously breathing animals did not affect [11C]-cocaine uptake (Schiffer et al. 2001). Both have shown effects on radiotracer uptake. For example, the gaseous anesthetic halothane profoundly affected the response to amphetamine measured with [11C]-raclopride, where ketamine did not (Ginovart et al. 2002; Hassoun et al. 2003). We have shown that [11C]-raclopride uptake in the striatum and cerebellum of awake, freely moving animals was considerably lower than [11C]-raclopride uptake in anesthetized animals, roughly five times lower in both regions (Patel et al. 2008). This is consistent with other studies in rodents demonstrating significantly less radiotracer

uptake in awake vs. anesthetized states (Honer et al. 2006; Momosaki et al. 2004; Schiffer et al. 2007b). It was suggested that a faster rate of peripheral metabolism might underlie lower uptake in awake mice (Honer et al. 2006); however, in our studies, at least with [11C]-raclopride, there is no difference in the rate of metabolism between the awake and anesthetized condition (Patel et al. 2008). If the culprit were blood flow, then an anesthesia-induced increase in blood flow would facilitate the delivery of [11C]-raclopride from plasma to brain tissue, resulting in an increase in brain uptake. In fact, PET studies using [15O]-H2O to measure regional cerebral blood flow (rCBF) in cats demonstrated that anesthetic doses of ketamine increased rCBF compared to the awake state (Hassoun et al. 2003). Most anesthetics have significant impact on the respiratory and cardiovascular systems, and consequently, tracer distribution and elimination will be influenced in vivo by the use of anesthesia. Crosscomparisons between anesthetized and nonanesthetized animals using alternative techniques (such as postmortem tissue sampling) must be performed to assess the effect of anesthesia on tracer accumulation. Fluctuations in the anesthetic level among various scans are supposed to increase the interindividual variation in tracer uptake, kinetics, and metabolism and must be kept at a minimum by standardization of anesthetic regimens and precise monitoring of physiological and hemodynamic parameters. Homeostasis of fluid, electrolyte,

34 Imaging in Neurology Research III: Focus on Neurotransmitter Imaging

and acid/base balance during anesthesia should be considered, especially for long-term scanning. In addition, body temperature must be controlled and adjusted with an appropriate heating device.

34.3.7 Limitations to Absolute Quantification Data analyses typically assume that tissue radioactivity is accurately measured in absolute units, i.e., nCi/ mL. In reality, it is not so simple. Radiotracer mass can have a significant effect on the measured BP. Attenuation of radiation by tissue, scatter, randoms, deadtime, and spatial resolution effects all degrade the accuracy of PET measurements. For the most part, image reconstruction software provided by the manufacturer provides a means to correct for most of these effects; however, spatial resolution effects are less accurately corrected for in most small animal studies.

34.3.8 Mass As the physical size of the animal decreases, the relative mass in a tracer dose becomes more significant, resulting in the possibility that the radiotracer itself may perturb the system under investigation. Mice and rats used in typical small animal studies weigh between 20–30 and 200–300 g, respectively. The “standard man,” on the other hand, weighs 70,000 g, several thousand times more than a mouse and several hundred times more than a rat. In round numbers, the linear dimensions of organs in these animals are ten or more times smaller in each dimension than in human and large primate subjects (Green et al. 2001; Jagoda et al. 2004). It follows that if imaging studies in these animals are to be equivalent to human and large primate studies, the spatial resolution of an animal scanner must be about ten or more times better than a human scanner. “Equivalent,” in this case, means that the animal organ is visualized on the animal scale with the same relative acuity as the human organ is imaged on the human scale. The ratio of pixel and/or voxel volumes in a PET image varies according to the ratio of body masses, so if the precision of measurement of voxel activity is to be the same as in the equivalent

533

human study, the same absolute amount of radioactivity must be delivered to the small animal voxel. This requirement, in turn, implies that the same absolute amount of drug must be given to the animal as to the human, since it is the drug that carries the radioactivity to that voxel. However, if a human dose is given to an animal, the concentration of drug in the animal may be thousands of times greater than in the human since the animal is thousands of times less massive. The possibility exists, therefore, that such large concentrations could, in some cases, cause pharmacological effects in small animals rather than act as tracers. This is highly likely when the target protein population is limited in number or changing with time, such as a saturable receptor population. A pharmacologically active dose of radioligand would violate the fundamental premise that unites the activity of a radioactive molecule to a physiologic process. That is, tracer kinetic modeling by definition requires that the radioligand be delivered in a tracer dose. In addition, changes in BP associated with changes in neurotransmitter activity are predicated on the basic assumption that the radiotracer itself does not alter these levels. Combined with recent experimental data suggesting that the BP declines with mass doses as low as 1.4 or as high as 5.0 nmol/kg (Alexoff et al. 2003; Myers and Hume 2002; Opacka-Juffry et al. 1998; Schiffer et al. 2005), there appears to be a condition under which the influence of mass is more significant (i.e., lower sensitivity cameras, low SA radiotracer synthesis). In order to define this relationship, we combined microPET imaging using 11C-raclopride with microdialysis measures of ECF dopamine (DA), under the hypothesis that excessive mass of radiotracer would be measurable as overflow of dopamine into the extracellular space, a common finding after D2 receptor saturation. In addition, we performed a series of studies in which a known mass of raclopride was microinfused into one striatum prior to a high SA systemic injection of 11 C-raclopride. This single-injection approach provided a high and low SA region of radiotracer binding in the same animal during the same scanning session. Our data demonstrate that the BP declines above 3.5 pmol/mL (0.35 mg), with an ED50 of 8.55 ± 5.62 pmol/mL. These data also provide evidence that BP may be compromised by masses of raclopride below 2.0 pmol/mL (0.326 mg). Increases in ECF DA were produced by mass doses of raclopride over

534

3.9 pmol/mL (0.329 mg) with an ED50 of 8.53 ± 2.48 pmol/mL. Taken together, it appears that an optimal range of raclopride mass exists between 2.0 and 3.5 pmol/mL, around which the measured BP can be compromised by system sensitivity, endogenous DA, or excessive competition with unlabeled compound. In reality, any amount of mass should alter the 11 C-raclopride BP – the issue is to define the limits of this range when both instrument size and animal size are dramatically reduced relative to the human PET data used to originally develop these tracer kinetic models (i.e., (Logan et al. 1997)). Radiochemical synthesis is routinely designed to produce compounds with SA appropriate for humans and large primate imaging studies. Unfortunately, the smaller physical size of a rodent means that there is an intricate balance between injecting a low enough dose to prevent a pharmacological response while maintaining sufficient counting statistics. This is most evident in Table 34.1, where the lowest mass doses of radiotracer do not necessarily give higher measures of BP. It is likely that this is due to increased noise produced by lower counting statistics. This tradeoff has been the focus of previous investigations and discussions (Alexoff et al. 2003; Hume et al. 1998; Myers and Hume 2002). Perhaps Hume’s 1995 experiments (a clinical ECAT 953B) would have required a higher SA injection in order to obtain measures equivalent to those reported using dedicated small animal tomographs (Meikle et al. 2000). As discussed previously, measured BP can also be influenced by the tomograph. When Hume’s pioneering studies using the ECAT 953B were followed by studies using dedicated animal tomographs, there was a progressive increase in the measured BP (from an average of 0.77 with the RATPET camera (Hume et al. 1996)) and to 1.9 using the same camera and different time points to 2.77 using the quad-HIDAC; (Houston et al. 2004)). This was attributed directly to the higher resolution of each tomograph. These studies support the notion that physical characteristics that are unique to a dedicated small animal tomograph may contribute to the variability in findings from different laboratories.

34.3.9 Data Corrections Accurate determination of radioactivity concentration using PET requires application of several corrections

W.K. Schiffer

to coincidence lines of response recorded by the camera. These corrections include subtraction of random coincidences, correction for deadtime losses, subtraction of scattered coincidences, correction for photon attenuation, and normalization for varying individual detector efficiencies. Finally, the PET camera must be calibrated against a known radioactivity concentration. If all necessary corrections are not applied and/or validated, calibration should be done with a phantom approximating the geometry of each imaging experiment (e.g., primate brain vs. mouse body). Sometimes, practical corrections for photon attenuation and scatter are not available from the PET manufacturer. Therefore, a scatter correction algorithm should be implemented, and its effect on measurements of radiotracer uptake and receptor availability in the target system should be evaluated independently. Scatter and attenuation. Scatter is known to have an appreciable effect on 3D PET data in the human brain (Cherry and Huang 1995). Given that the diameter of the adult rodent head is small compared to the mean free path of a 511 keV photon in tissue (~10 cm), radiotracer imaging in the rodent brain was expected to give small animal PET data with relatively small scatter fractions. However, the lack of septa in small animal tomographs and the large axial field-of-view (FOV) increase the probability that sinogram scatter fractions might be appreciable. Region-of-interest (ROI) data have shown a significant effect of scatter, especially when derived from brain regions of lower accumulation of tracer radioactivity like the cerebellum, where mean ROI pixel values from scatter-corrected images decreased as much as 20% compared to uncorrected image data (Alexoff et al. 2003). Further, there can be a significant increase in sinogram background from the beginning of the study to the end. Scatter fractions determined from early time frames (<40 min) have been measured at ~40% in the rat head, similar to those reported by Knoess et al. (2003) for a rat phantom using the same energy window. However, at 60 min, scatter fractions have been found to be significantly higher (~65% (Alexoff et al. 2003). Thus, the effects of scatter on small animal PET data can be significant. Because they affect regions with low radiotracer activity (or low counts) more than regions with high counts, scatter effects can dramatically influence the quantitation of data, which relies on cerebellar or other reference regions. Further, because scatter effects vary over time, becoming more significant

34 Imaging in Neurology Research III: Focus on Neurotransmitter Imaging

toward the end of the scan (where quantitation from TACs occurs most frequently), they can lead to significant errors in quantitative accuracy. While the absolute attenuation of deep rodent brain regions can be calculated to be significant (approaching 20–40%), the relative variation across the rodent head is more likely to be closer to 10% since all parts of the image are attenuated to some extent. In fact, Yoa et al. (2000) have detected nonuniformity due to attenuation in the rat brain to be ~12%. Image attenuation corrections determined previously for the striatum and cerebellum in adult Sprague Dawley rats using a custom-built animal PET camera (Myers et al. 1996) were found to be the same (16%). Since most data analysis methods with dopaminergic radiotracers rely on the ratio of a receptor-rich region (striatum) with a region used to estimate nonspecific binding (cerebellum), a constant attenuation correction factor for the striatum and cerebellum could be ignored. To confirm this, the effect of attenuation correction on graphical analysis of [11C]-raclopride binding in the rat striatum was evaluated and had no significant effect on measures of specific binding (Alexoff et al. 2003). Spillover and partial volume. Components of spillover (contamination from surrounding tissue) and partial volume (loss of true tissue concentration) can influence the quantitation of small animal PET data, especially in models of neurodegenerative disease. In animal species with nictitating membranes, such as rodents, the harderian glands located just behind the eyes are a noteworthy contributor to spillover artifacts. They lie deep within the orbit and encircle the bulbus oculi and optic nerve on the medial, superior, and inferior sides. Although harderian glands are reported to have a range of incompletely described photoreceptive, endocrine, and immunologic functions, they appear to serve primarily as an exocrine secreting accessory to the lacrimal gland for eye lubrication (Watanabe 1980). The problem of radioactive spillover from the harderian glands during the use of FDG as a PET tracer in rats and mice brain studies has been reported by multiple groups (Brammer et al. 2007; Fukuyama et al. 1998; Kuge et al. 1997a, b; Moore et al. 2000; Zovein et al. 2004). Although the specific mechanism by which FDG is incorporated into the glands is unclear, one hypothesis is that glucose plays a rate-limiting role in lipid synthesis (Watanabe 1980). Kuge et al. (1997a, b) were the first to describe this image-degrading phenomenon in rats. It has been

535

observed that [11C]-raclopride and most dopamine receptor and transporter radiotracers are taken up by the Harderian glands, while [18F]-DOPA is not (Hume et al. 1996). While harderian gland spillover effects may not influence measures of radioactivity concentrations from the rat, they may be problematic for ROI over cortical and subcortical structures from the mouse. For example, we have shown differential harderian uptake in mice with epilepitic seizures (Mirrione et al. 2007), which dramatically influences methods for spatial normalization of mouse FDG PET data. Reference region selection. Small animal PET studies are typically analyzed by ROIs drawn manually on the PET image. However, in the absence of any anatomical information, ROIs that are based on the observed radioactivity distribution can be misleading, especially in regions designated as “reference regions” due to their low density of target receptors. In this regard, dopamine transporter ligands appear to be especially vulnerable, due to a greater degree of nonspecific binding in the cerebellum compared to dopamine D2 receptor ligands. In fact, Hume demonstrated that a blocking injection of the dopamine transport ligand, RTI-121 (unlabeled) led to greater initial extraction and retention of [11C]-RTI-121 in the brain relative to controls (Hume et al. 1996). This emphasizes the need for parallel blocking (pretreatment with unlabeled compound) studies to ensure that the reference region accurately reflects nonspecific binding. Further, if one notices a greater standard deviation in the mean BP measurements for the blocked studies, this usually reflects variations in plasma input and can result in variations in the availability of a compound to bind.

34.4 Conclusion Until recently, it has only been possible to use PET for quantitative brain imaging in humans and nonhuman primates. Advances in PET technology have now created opportunities to image brains of smaller animals, such as rats and mice. With small animal PET, therapeutic strategies for CNS intervention can now be modeled in the rodent and evaluated in vivo imaging prior to further evaluation in nonhuman primate models and human disease. In this way, small animal PET studies provide a unique continuum of research methods from the rodent to the nonhuman primate that will ultimately

536

contribute to imaging molecular biology-based interventions in humans. These integrated molecular- and system-based assessments give insight into molecular mechanisms while validating and optimizing methodologies for subsequent human applications. Although neuroscientists have sought to quantify the physically meaningful parameters of functional imaging, the intricacies of biological tissue on a cellular scale, and the numerous experimental difficulties encountered, have complicated the simple measurement and interpretation of these quantities. Nevertheless, it is clear that neurotransmitter measurements provide a powerful tool for investigating neurochemical function, and displacement studies provide a readily understood, albeit simplistic, assessment of tissue function. As techniques develop further, and our understanding of the relationships between these measurements and the tissue structure and function improves, this method offers great promise in extending our knowledge of both normal tissues and tissue that has, or is suffering, damage. Since PET is noninvasive, the measurements may be repeated to follow neurochemical changes over the time scales of seconds, minutes, days, or even years. The value of this information is enhanced by the additional ability of simultaneous behavioral measures, to corroborate imaging results. In addition, metabolic measurements can provide important biochemical information related to tissue function. These capabilities have ensured that neurotransmitter imaging with PET has found extensive application in animal models of disease and human studies.

References Abi-Dargham A, Rodenhiser J, Printz D, Zea-Ponce Y, Gil R, Kegeles LS, Weiss R, Cooper TB, Mann JJ, Van Heertum RL, Gorman JM, Laruelle M (2000) Increased baseline occupancy of D2 receptors by dopamine in schizophrenia. Proc Natl Acad Sci USA 97(14):8104–8109 Alexoff DL, Vaska P, Marsteller D, Gerasimov T, Li J, Logan J, Fowler JS, Taintor NB, Thanos PK, Volkow ND (2003) Reproducibility of 11C-raclopride binding in the rat brain measured with the microPET R4: effects of scatter correction and tracer specific activity. J Nucl Med 44(5):815–822 Alpert NM, Badgaiyan RD, Livni E, Fischman AJ (2003) A novel method for noninvasive detection of neuromodulatory changes in specific neurotransmitter systems. Neuroimage 19(3):1049–1060 Asselin MC, Montgomery AJ, Grasby PM, Hume SP (2007) Quantification of PET studies with the very high-affinity

W.K. Schiffer dopamine D2/D3 receptor ligand [11C]FLB 457: re-evaluation of the validity of using a cerebellar reference region. J Cereb Blood Flow Metab 27(2):378–392 Barrett SP, Boileau I, Okker J, Pihl RO, Dagher A (2004) The hedonic response to cigarette smoking is proportional to dopamine release in the human striatum as measured by positron emission tomography and [11C]raclopride. Synapse 54(2):65–71 Boileau I, Dagher A, Leyton M, Welfeld K, Booij L, Diksic M, Benkelfat C (2007) Conditioned dopamine release in humans: a positron emission tomography [11C]raclopride study with amphetamine. J Neurosci 27(15):3998–4003 Brammer DW, Riley JM, Kreuser SC, Zasadny KR, Callahan MJ, Davis MD (2007) Harderian gland adenectomy: a method to eliminate confounding radio-opacity in the assessment of rat brain metabolism by 18F-fluoro-2-deoxy-Dglucose positron emission tomography. J Am Assoc Lab Anim Sci 46(5):42–45 Breier A, Su TP, Saunders R, Carson RE, Kolachana BS, de Bartolomeis A, Weinberger DR, Weisenfeld N, Malhotra AK, Eckelman WC, Pickar D (1997) Schizophrenia is associated with elevated amphetamine-induced synaptic dopamine concentrations: evidence from a novel positron emission tomography method. Proc Natl Acad Sci USA 94(6):2569–2574 Brody AL, Olmstead RE, London ED, Farahi J, Meyer JH, Grossman P, Lee GS, Huang J, Hahn EL, Mandelkern MA (2004) Smoking-induced ventral striatum dopamine release. Am J Psychiatry 161(7):1211–1218 Brooks DJ (1997) PET and SPECT studies in Parkinson’s disease. Baillieres Clin Neurol 6(1):69–87 Cardenas L, Houle S, Kapur S, Busto UE (2004) Oral D-amphetamine causes prolonged displacement of [11C] raclopride as measured by PET. Synapse 51(1):27–31 Carson RE (1996) Mathematical modeling and compartmental analysis. In: Harbert JC, Eckelman WC, Neumann RD (eds) Nuclear medicine diagnosis and therapy. Thieme, New York Carson RE (2000) PET physiological measurements using constant infusion. Nucl Med Biol 27(7):657–660 Carson RE, Yan Y, Daube-Witherspoon ME, Freedman N, Bacharach SL, Herscovitch P (1993) An approximation formula for the variance of PET region-of-interest values. IEEE Trans Med Imaging 12(2):240–250 Carson RE, Breier A, de Bartolomeis A, Saunders RC, Su TP, Schmall B, Der MG, Pickar D, Eckelman WC (1997) Quantification of amphetamine-induced changes in [11C] raclopride binding with continuous infusion. J Cereb Blood Flow Metab 17(4):437–447 Carson RE, Kiesewetter DO, Jagoda E, Der MG, Herscovitch P, Eckelman WC (1998) Muscarinic cholinergic receptor measurements with [18F]FP-TZTP: control and competition studies. J Cereb Blood Flow Metab 18(10):1130–1142 Cherry SR, Huang SC (1995) Effects of scatter on model parameters in 3D PET studies of the human brain. IEEE Trans Med Imaging 42:1174–1179 Cropley VL, Innis RB, Nathan PJ, Brown AK, Sangare JL, Lerner A, Ryu YH, Sprague KE, Pike VW, Fujita M (2008) Small effect of dopamine release and no effect of dopamine depletion on [18F]fallypride binding in healthy humans. Synapse 62(6):399–408

34 Imaging in Neurology Research III: Focus on Neurotransmitter Imaging Davis MR, Votaw JR, Bremner JD, Byas-Smith MG, Faber TL, Voll RJ, Hoffman JM, Grafton ST, Kilts CD, Goodman MM (2003) Initial human PET imaging studies with the dopamine transporter ligand 18F-FECNT. J Nucl Med 44(6):855–861 de Paulis T (2003) The discovery of epidepride and its analogs as high-affinity radioligands for imaging extrastriatal dopamine D(2) receptors in human brain. Curr Pharm Des 9(8):673–696 Dewey SL, MacGregor RR, Brodie JD, Bendriem B, King PT, Volkow ND, Schlyer DJ, Fowler JS, Wolf AP, Gatley SJ (1990) Mapping muscarinic receptors in human and baboon brain using [N-11C-methyl]-benztropine. Synapse 5(3):213–223 Dewey SL, Logan J, Wolf AP, Brodie JD, Angrist B, Fowler JS, Volkow ND (1991) Amphetamine induced decreases in (18F)-N-methylspiroperidol binding in the baboon brain using positron emission tomography (PET). Synapse 7(4):324–327 Dewey SL, Smith GS, Logan J, Brodie JD, Fowler JS, Wolf AP (1993) Striatal binding of the PET ligand 11C-raclopride is altered by drugs that modify synaptic dopamine levels. Synapse 13(4):350–356 Drevets WC, Gautier C, Price JC, Kupfer DJ, Kinahan PE, Grace AA, Price JL, Mathis CA (2001) Amphetamine-induced dopamine release in human ventral striatum correlates with euphoria. Biol Psychiatry 49(2):81–96 Ehrin E, Farde L, de Paulis T, Eriksson L, Greitz T, Johnstrom P, Litton JE, Nilsson JL, Sedvall G, Stone-Elander S et al (1985) Preparation of 11C-labelled Raclopride, a new potent dopamine receptor antagonist: preliminary PET studies of cerebral dopamine receptors in the monkey. Int J Appl Radiat Isot 36(4):269–273 Elsinga PH, Hatano K, Ishiwata K (2006) PET tracers for imaging of the dopaminergic system. Curr Med Chem 13(18):2139–2153 Endres CJ, Carson RE (1998) Assessment of dynamic neurotransmitter changes with bolus or infusion delivery of neuroreceptor ligands. J Cereb Blood Flow Metab 18(11):1196–1210 Endres CJ, Kolachana BS, Saunders RC, Su T, Weinberger D, Breier A, Eckelman WC, Carson RE (1997) Kinetic modeling of [11C]raclopride: combined PET-microdialysis studies. J Cereb Blood Flow Metab 17(9):932–942 Farde L, Ehrin E, Eriksson L, Greitz T, Hall H, Hedstrom CG, Litton JE, Sedvall G (1985) Substituted benzamides as ligands for visualization of dopamine receptor binding in the human brain by positron emission tomography. Proc Natl Acad Sci USA 82(11):3863–3867 Farde L, Hall H, Ehrin E, Sedvall G (1986) Quantitative analysis of D2 dopamine receptor binding in the living human brain by PET. Science 231(4735):258–261 Farde L, Eriksson L, Blomquist G, Halldin C (1989) Kinetic analysis of central [11C]raclopride binding to D2-dopamine receptors studied by PET–a comparison to the equilibrium analysis. J Cereb Blood Flow Metab 9(5):696–708 Farde L, Nordstrom AL, Wiesel FA, Pauli S, Halldin C, Sedvall G (1992) Positron emission tomographic analysis of central D1 and D2 dopamine receptor occupancy in patients treated with classical neuroleptics and clozapine. Relation to extrapyramidal side effects. Arch Gen Psychiatry 49(7):538–544

537

Fisher RE, Morris ED, Alpert NM, Fischman AJ (1995) In vivo imaging of neuromodulatory synaptic transmission using PET: a review of relevant neurophysiology. Hum Brain Mapp 3:24–34 Frankle WG, Cho RY, Narendran R, Mason NS, Vora S, Litschge M, Price JC, Lewis DA, Mathis CA (2009) Tiagabine increases [11C]flumazenil binding in cortical brain regions in healthy control subjects. Neuropsychopharmacology 34(3):624–633 Fukuyama H, Hayashi T, Katsumi Y, Tsukada H, Shibasaki H (1998) Issues in measuring glucose metabolism of rat brain using PET: the effect of harderian glands on the frontal lobe. Neurosci Lett 255(2):99–102 Ginovart N, Farde L, Halldin C, Swahn CG (1997) Effect of reserpine-induced depletion of synaptic dopamine on [11C] raclopride binding to D2-dopamine receptors in the monkey brain. Synapse 25(4):321–325 Ginovart N, Hassoun W, Le Cavorsin M, Veyre L, Le Bars D, Leviel V (2002) Effects of amphetamine and evoked dopamine release on [11C]raclopride binding in anesthetized cats. Neuropsychopharmacology 27(1):72–84 Ginovart N, Wilson AA, Meyer JH, Hussey D, Houle S (2003) [11C]-DASB, a tool for in vivo measurement of SSRIinduced occupancy of the serotonin transporter: PET characterization and evaluation in cats. Synapse 47(2):123–133 Ginovart N, Galineau L, Willeit M, Mizrahi R, Bloomfield PM, Seeman P, Houle S, Kapur S, Wilson AA (2006) Binding characteristics and sensitivity to endogenous dopamine of [11C]-(+)-PHNO, a new agonist radiotracer for imaging the high-affinity state of D2 receptors in vivo using positron emission tomography. J Neurochem 97(4):1089–1103 Goerendt IK, Messa C, Lawrence AD, Grasby PM, Piccini P, Brooks DJ (2003) Dopamine release during sequential finger movements in health and Parkinson’s disease: a PET study. Brain 126(pt 2):312–325 Graff-Guerrero A, Willeit M, Ginovart N, Mamo D, Mizrahi R, Rusjan P, Vitcu I, Seeman P, Wilson AA, Kapur S (2008) Brain region binding of the D2/3 agonist [11C]-(+)-PHNO and the D2/3 antagonist [11C]raclopride in healthy humans. Hum Brain Mapp 29(4):400–410 Green MV, Seidel J, Vaquero JJ, Jagoda E, Lee I, Eckelman WC (2001) High resolution PET, SPECT and projection imaging in small animals. Comput Med Imaging Graph 25(2):79–86 Gunn RN, Gunn SR, Cunningham VJ (2001) Positron emission tomography compartmental models. J Cereb Blood Flow Metab 21(6):635–652 Halldin C, Farde L, Hogberg T, Mohell N, Hall H, Suhara T, Karlsson P, Nakashima Y, Swahn CG (1995) Carbon-11FLB 457: a radioligand for extrastriatal D2 dopamine receptors. J Nucl Med 36(7):1275–1281 Halldin C, Erixon-Lindroth N, Pauli S, Chou YH, Okubo Y, Karlsson P, Lundkvist C, Olsson H, Guilloteau D, Emond P, Farde L (2003) [(11)C]PE2I: a highly selective radioligand for PET examination of the dopamine transporter in monkey and human brain. Eur J Nucl Med Mol Imaging 30(9):1220–1230 Hassoun W, Le Cavorsin M, Ginovart N, Zimmer L, Gualda V, Bonnefoi F, Leviel V (2003) PET study of the [11C]raclopride binding in the striatum of the awake cat: effects of anaesthetics and role of cerebral blood flow. Eur J Nucl Med Mol Imaging 30(1):141–148

538 Hirani E, Sharp T, Sprakes M, Grasby P, Hume S (2003) Fenfluramine evokes 5-HT2A receptor-mediated responses but does not displace [11C]MDL 100907: small animal PET and gene expression studies. Synapse 50(3):251–260 Hirvonen J, Johansson J, Teras M, Oikonen V, Lumme V, Virsu P, Roivainen A, Nagren K, Halldin C, Farde L, Hietala J (2008) Measurement of striatal and extrastriatal dopamine transporter binding with high-resolution PET and [11C] PE2I: quantitative modeling and test-retest reproducibility. J Cereb Blood Flow Metab 28(5):1059–1069 Honer M, Hengerer B, Blagoev M, Hintermann S, Waldmeier P, Schubiger PA, Ametamey SM (2006) Comparison of [18F] FDOPA, [18F]FMT and [18F]FECNT for imaging dopaminergic neurotransmission in mice. Nucl Med Biol 33(5):607–614 Houston GC, Hume SP, Hirani E, Goggi JL, Grasby PM (2004) Temporal characterisation of amphetamine-induced dopamine release assessed with [11C]raclopride in anaesthetised rodents. Synapse 51(3):206–212 Hume SP, Opacka-Juffry J, Myers R, Ahier RG, Ashworth S, Brooks DJ, Lammertsma AA (1995) Effect of L-dopa and 6-hydroxydopamine lesioning on [11C]raclopride binding in rat striatum, quantified using PET. Synapse 21(1):45–53 Hume SP, Lammertsma AA, Myers R, Rajeswaran S, Bloomfield PM, Ashworth S, Fricker RA, Torres EM, Watson I, Jones T (1996) The potential of high-resolution positron emission tomography to monitor striatal dopaminergic function in rat models of disease. J Neurosci Methods 67(2):103–112 Hume SP, Gunn RN, Jones T (1998) Pharmacological constraints associated with positron emission tomographic scanning of small laboratory animals. Eur J Nucl Med 25(2):173–176 Hume S, Hirani E, Opacka-Juffry J, Myers R, Townsend C, Pike V, Grasby P (2001) Effect of 5-HT on binding of [(11)C] WAY 100635 to 5-HT(IA) receptors in rat brain, assessed using in vivo microdialysis nd PET after fenfluramine. Synapse 41(2):150–159 Innis RB, Cunningham VJ, Delforge J, Fujita M, Gjedde A, Gunn RN, Holden J, Houle S, Huang SC, Ichise M, Iida H, Ito H, Kimura Y, Koeppe RA, Knudsen GM, Knuuti J, Lammertsma AA, Laruelle M, Logan J, Maguire RP, Mintun MA, Morris ED, Parsey R, Price JC, Slifstein M, Sossi V, Suhara T, Votaw JR, Wong DF, Carson RE (2007) Consensus nomenclature for in vivo imaging of reversibly binding radioligands. J Cereb Blood Flow Metab 27(9):1533–1539 Ito H, Hietala J, Blomqvist G, Halldin C, Farde L (1998) Comparison of the transient equilibrium and continuous infusion method for quantitative PET analysis of [11C] raclopride binding. J Cereb Blood Flow Metab 18(9):941–950 Jagoda EM, Vaquero JJ, Seidel J, Green MV, Eckelman WC (2004) Experiment assessment of mass effects in the rat: implications for small animal PET imaging. Nucl Med Biol 31(6):771–779 Jucaite A, Fernell E, Halldin C, Forssberg H, Farde L (2005) Reduced midbrain dopamine transporter binding in male adolescents with attention-deficit/hyperactivity disorder: association between striatal dopamine markers and motor hyperactivity. Biol Psychiatry 57(3):229–238 Kjaer TW, Bertelsen C, Piccini P, Brooks D, Alving J, Lou HC (2002) Increased dopamine tone during meditation-induced change of consciousness. Brain Res Cogn Brain Res 13(2):255–259

W.K. Schiffer Knoess C, Siegel S, Smith A, Newport D, Richerzhagen N, Winkeler A, Jacobs A, Goble RN, Graf R, Wienhard K, Heiss WD (2003) Performance evaluation of the microPET R4 PET scanner for rodents. Eur J Nucl Med Mol Imaging 30(5):737–747 Koepp MJ, Gunn RN, Lawrence AD, Cunningham VJ, Dagher A, Jones T, Brooks DJ, Bench CJ, Grasby PM (1998a) Evidence for striatal dopamine release during a video game. Nature 393(6682):266–268 Koepp MJ, Richardson MP, Brooks DJ, Duncan JS (1998b) Focal cortical release of endogenous opioids during readinginduced seizures. Lancet 352(9132):952–955 Krause KH, Dresel SH, Krause J, la Fougere C, Ackenheil M (2003) The dopamine transporter and neuroimaging in attention deficit hyperactivity disorder. Neurosci Biobehav Rev 27(7):605–613 Kuge Y, Minematsu K, Hasegawa Y, Yamaguchi T, Mori H, Matsuura H, Hashimoto N, Miyake Y (1997a) Positron emission tomography for quantitative determination of glucose metabolism in normal and ischemic brains in rats: an insoluble problem by the Harderian glands. J Cereb Blood Flow Metab 17(1):116–120 Kuge Y, Miyake Y, Minematsu K, Yamaguchi T, Hasegawa Y (1997b) Effects of extracranial radioactivity on measurement of cerebral glucose metabolism by rat-PET with [18F]-2-fluoro-2-deoxy-D-glucose. J Cereb Blood Flow Metab 17(11):1261–1262 Lammertsma AA, Hume SP (1996) Simplified reference tissue model for PET receptor studies. Neuroimage 4(3 pt 1):153–158 Lappin JM, Reeves SJ, Mehta MA, Egerton A, Coulson M, Grasby PM (2009) Dopamine release in the human striatum: motor and cognitive tasks revisited. J Cereb Blood Flow Metab 29(3):554–564 Laruelle M (2000) Imaging synaptic neurotransmission with in vivo binding competition techniques: a critical review. J Cereb Blood Flow Metab 20:423–451 Laruelle M, Iyer RN, al-Tikriti MS, Zea-Ponce Y, Malison R, Zoghbi SS, Baldwin RM, Kung HF, Charney DS, Hoffer PB, Innis RB, Bradberry CW (1997) Microdialysis and SPECT measurements of amphetamine-induced dopamine release in nonhuman primates. Synapse 25(1):1–14 Leyton M, Dagher A, Boileau I, Casey K, Baker GB, Diksic M, Gunn R, Young SN, Benkelfat C (2004) Decreasing amphetamine-induced dopamine release by acute phenylalanine/ tyrosine depletion: A PET/[11C]raclopride study in healthy men. Neuropsychopharmacology 29(2):427–432 Logan J (2003) A review of graphical methods for tracer studies and strategies to reduce bias. Nucl Med Biol 30(8):833–844 Logan J, Fowler JS, Volkow ND, Wolf AP, Dewey SL, Schlyer DJ, MacGregor RR, Hitzemann R, Bendriem B, Gatley SJ, al e (1990) Graphical analysis of reversible radioligand binding from time-activity measurements applied to [N-11Cmethyl]-(−)-cocaine PET studies in human subjects. J Cereb Blood Flow Metab 10(5):740–747 Logan J, Dewey SL, Wolf AP, Fowler JS, Brodie JD, Angrist B, Volkow ND, Gatley SJ (1991) Effects of endogenous dopamine on measures of [18F]N-methylspiroperidol binding in the basal ganglia: comparison of simulations and experimental results from PET studies in baboons. Synapse 9(3):195–207

34 Imaging in Neurology Research III: Focus on Neurotransmitter Imaging Logan J, Fowler JS, Volkow ND, Wang GJ, Ding YS, Alexoff DL (1996) Distribution volume ratios without blood sampling from graphical analysis of PET data. J Cereb Blood Flow Metab 16(5):834–840 Logan J, Volkow ND, Fowler JS, Wang GJ, Fischman MW, Foltin RW, Abumrad NN, Vitkun S, Gatley SJ, Pappas N, Hitzemann R, Shea CE (1997) Concentration and occupancy of dopamine transporters in cocaine abusers with [11C] cocaine and PET. Synapse 27(4):347–356 Logan J, Fowler JS, Volkow ND, Ding YS, Wang GJ, Alexoff DL (2001) A strategy for removing the bias in the graphical analysis method. J Cereb Blood Flow Metab 21(3):307–320 Marenco S, Carson RE, Berman KF, Herscovitch P, Weinberger DR (2004) Nicotine-induced dopamine release in primates measured with [11C]raclopride PET. Neuropsycho pharmacology 29(2):259–268 Martinez D, Slifstein M, Broft A, Mawlawi O, Hwang DR, Huang Y, Cooper T, Kegeles L, Zarahn E, Abi-Dargham A, Haber SN, Laruelle M (2003) Imaging human mesolimbic dopamine transmission with positron emission tomography. Part II: amphetamine-induced dopamine release in the functional subdivisions of the striatum. J Cereb Blood Flow Metab 23(3):285–300 Martinez D, Greene K, Broft A, Kumar D, Liu F, Narendran R, Slifstein M, Van Heertum R, Kleber HD (2009) Lower level of endogenous dopamine in patients with cocaine dependence: findings from PET imaging of D(2)/D(3) receptors following acute dopamine depletion. Am J Psychiatry 166(10):1170–1177 McCormick PN, Kapur S, Seeman P, Wilson AA (2008) Dopamine D2 receptor radiotracers [(11)C](+)-PHNO and [(3)H]raclopride are indistinguishably inhibited by D2 agonists and antagonists ex vivo. Nucl Med Biol 35(1):11–17 Meikle SR, Eberl S, Fulton RR, Kassiou M, Fulham MJ (2000) The influence of tomograph sensitivity on kinetic parameter estimation in positron emission tomography imaging studies of the rat brain. Nucl Med Biol 27(6):617–625 Mintun MA, Raichle ME, Kilbourn MR, Wooten GF, Welch MJ (1984) A quantitative model for the in vivo assessment of drug binding sites with positron emission tomography. Ann Neurol 15(3):217–227 Mirrione MM, Schiffer WK, Fowler JS, Alexoff DL, Dewey SL, Tsirka SE (2007) A novel approach for imaging brainbehavior relationships in mice reveals unexpected metabolic patterns during seizures in the absence of tissue plasminogen activator. Neuroimage 38(1):34–42 Momosaki S, Hatano K, Kawasumi Y, Kato T, Hosoi R, Kobayashi K, Inoue O, Ito K (2004) Rat-PET study without anesthesia: anesthetics modify the dopamine D1 receptor binding in rat brain. Synapse 54(4):207–213 Montgomery AJ, Mehta MA, Grasby PM (2006) Is psychological stress in man associated with increased striatal dopamine levels? A [11C]raclopride PET study. Synapse 60(2):124–131 Montgomery AJ, Asselin MC, Farde L, Grasby PM (2007) Measurement of methylphenidate-induced change in extrastriatal dopamine concentration using [11C]FLB 457 PET. J Cereb Blood Flow Metab 27(2):369–377 Moore AH, Osteen CL, Chatziioannou AF, Hovda DA, Cherry SR (2000) Quantitative assessment of longitudinal metabolic changes in vivo after traumatic brain injury in the adult rat using FDG-microPET. J Cereb Blood Flow Metab 20(10):1492–1501

539

Morris ED, Fisher RE, Alpert NM, Rauch SL, Fischman AJ (1995) In vivo imaging of neuromodulation using positron emission tomography: Optimal ligand characteristics and task length for detection of activation. Hum Brain Mapp 3:35–55 Morris ED, Normandin MD, Schiffer WK (2008) Initial comparison of ntPET with microdialysis measurements of methamphetamine-induced dopamine release in rats: support for estimation of dopamine curves from PET data. Mol Imaging Biol 10(2):67–73 Mukherjee J, Yang ZY, Das MK, Brown T (1995) Fluorinated benzamide neuroleptics–III. Development of (S)-N-[(1allyl-2-pyrrolidinyl)methyl]-5-(3-[18F]fluoropropyl)-2, 3-dimethoxybenzamide as an improved dopamine D-2 receptor tracer. Nucl Med Biol 22(3):283–296 Mukherjee J, Yang ZY, Brown T, Roemer J, Cooper M (1996) 18F-desmethoxyfallypride: a fluorine-18 labeled radiotracer with properties similar to carbon-11 raclopride for PET imaging studies of dopamine D2 receptors. Life Sci 59(8):669–678 Mukherjee J, Christian BT, Dunigan KA, Shi B, Narayanan TK, Satter M, Mantil J (2002) Brain imaging of 18F-fallypride in normal volunteers: blood analysis, distribution, test-retest studies, and preliminary assessment of sensitivity to aging effects on dopamine D-2/D-3 receptors. Synapse 46(3):170–188 Myers R, Hume S (2002) Small animal PET. Eur Neuropsychopharmacol 12(6):545–555 Myers R, Hume SP, Ashworth S, Lammertsma A, Bloomfield PM, Rajeswaran S, Jones T (1996) Quantification of dopamine receptors and transporter in rat striatum using a small animal PET scanner. In: Myers R, Cunningham VJ, Bailey D, Jones T (eds) Quantification of brain function using PET. Academic, London, pp 12–15 Narendran R, Hwang DR, Slifstein M, Talbot PS, Erritzoe D, Huang Y, Cooper TB, Martinez D, Kegeles LS, Abi-Dargham A, Laruelle M (2004) In vivo vulnerability to competition by endogenous dopamine: comparison of the D2 receptor agonist radiotracer (−)-N-[11C]propyl-norapomorphine ([11C] NPA) with the D2 receptor antagonist radiotracer [11C]-raclopride. Synapse 52(3):188–208 Narendran R, Hwang DR, Slifstein M, Hwang Y, Huang Y, Ekelund J, Guillin O, Scher E, Martinez D, Laruelle M (2005) Measurement of the proportion of D2 receptors configured in state of high affinity for agonists in vivo: a positron emission tomography study using [11C] N-propyl-norapomorphine and [11C]raclopride in baboons. J Pharmacol Exp Ther 315(1):80–90 Narendran R, Slifstein M, Guillin O, Hwang Y, Hwang DR, Scher E, Reeder S, Rabiner E, Laruelle M (2006) Dopamine (D2/3) receptor agonist positron emission tomography radiotracer [11C]-(+)-PHNO is a D3 receptor preferring agonist in vivo. Synapse 60(7):485–495 Nishiyama S, Tsukada H, Sato K, Kakiuchi T, Ohba H, Harada N, Takahashi K (2001) Evaluation of PET ligands (+)N-[(11) C]ethyl-3-piperidyl benzilate and (+)N-[(11)C]propyl-3-piperidyl benzilate for muscarinic cholinergic receptors: a PET study with microdialysis in comparison with (+)N-[(11) C]methyl-3-piperidyl benzilate in the conscious monkey brain. Synapse 40(3):159–169 Olsson H, Halldin C, Farde L (2004) Differentiation of extrastriatal dopamine D2 receptor density and affinity in the human brain using PET. Neuroimage 22(2):794–803

540 Opacka-Juffry J, Ahier RG, Cremer JE (1991) Nomifensineinduced increase in extracellular striatal dopamine is enhanced by isoflurane anaesthesia. Synapse 7(2):169–171 Opacka-Juffry J, Ashworth S, Ahier RG, Hume SP (1998) Modulatory effects of L-DOPA on D2 dopamine receptors in rat striatum, measured using in vivo microdialysis and PET. J Neural Transm 105(4–5):349–364 Oswald LM, Wong DF, McCaul M, Zhou Y, Kuwabara H, Choi L, Brasic J, Wand GS (2005) Relationships among ventral striatal dopamine release, cortisol secretion, and subjective responses to amphetamine. Neuropsychopharmacology 30(4):821–832 Ouchi Y, Yoshikawa E, Futatsubashi M, Okada H, Torizuka T, Sakamoto M (2002) Effect of simple motor performance on regional dopamine release in the striatum in Parkinson disease patients and healthy subjects: a positron emission tomography study. J Cereb Blood Flow Metab 22(6):746–752 Park MH, Park EH (2000) Synaptic concentration of dopamine in rat striatal slices in relationship to [3H]raclopride binding to the dopamine D2 receptor. Arch Pharm Res 23(4):360–366 Patel VD, Lee DE, Alexoff DL, Dewey SL, Schiffer WK (2008) Imaging dopamine release with positron emission tomography (PET) and (11)C-raclopride in freely moving animals. Neuroimage 41(3):1051–1066 Patlak CS, Blasberg RG, Fenstermacher JD (1983) Graphical evaluation of blood-to-brain transfer constants from multiple-time uptake data. J Cereb Blood Flow Metab 3(1):1–7 Pinborg LH, Adams KH, Yndgaard S, Hasselbalch SG, Holm S, Kristiansen H, Paulson OB, Knudsen GM (2004) [18F] altanserin binding to human 5HT2A receptors is unaltered after citalopram and pindolol challenge. J Cereb Blood Flow Metab 24(9):1037–1045 Pruessner JC, Champagne F, Meaney MJ, Dagher A (2004) Dopamine release in response to a psychological stress in humans and its relationship to early life maternal care: a positron emission tomography study using [11C]raclopride. J Neurosci 24(11):2825–2831 Quinn GP, Shore PA, Brodie BB (1959) Biochemical and pharmacological studies of RO 1-9569 (tetrabenazine), a nonindole tranquilizing agent with reserpine-like effects. J Pharmacol Exp Ther 127:103–109 Rabiner EA, Messa C, Sargent PA, Husted-Kjaer K, Montgomery A, Lawrence AD, Bench CJ, Gunn RN, Cowen P, Grasby PM (2002) A database of [(11)C]WAY-100635 binding to 5-HT(1A) receptors in normal male volunteers: normative data and relationship to methodological, demographic, physiological, and behavioral variables. Neuroimage 15(3):620–632 Rodrigues SF, de Oliveira MA, Martins JO, Sannomiya P, de Cassia Tostes R, Nigro D, Carvalho MH, Fortes ZB (2006) Differential effects of chloral hydrate- and ketamine/xylazine-induced anesthesia by the s.c. route. Life Sci 79(17):1630–1637 Rosa Neto P, Lou H, Cumming P, Pryds O, Gjedde A (2002) Methylphenidate-evoked potentiation of extracellular dopamine in the brain of adolescents with premature birth: correlation with attentional deficit. Ann N Y Acad Sci 965:434–439 Ross SB, Jackson DM (1989a) Kinetic properties of the accumulation of 3H-raclopride in the mouse brain in vivo. Naunyn Schmiedebergs Arch Pharmacol 340(1):6–12 Ross SB, Jackson DM (1989b) Kinetic properties of the in vivo accumulation of 3H-(−)-N-n-propylnorapomorphine in mouse brain. Naunyn Schmiedebergs Arch Pharmacol 340(1):13–20

W.K. Schiffer Sawamoto N, Piccini P, Hotton G, Pavese N, Thielemans K, Brooks DJ (2008) Cognitive deficits and striato-frontal dopamine release in Parkinson’s disease. Brain 131(pt 5):1294–1302 Schiffer WK, Carter P, Alexoff DL, Fowler JS, Logan J, Shea C, Gerasimov M, Brodie J, Dewey S (2001) Manipulations in mechanical ventilation influence the transport constant (K-1) of the reversible radiotracer C-11-cocaine. J Nucl Med 42(5):909 Schiffer WK, Alexoff D, Dewey SL (2003) Microdialysis probe placement produces marked ipsilateral hypometabolism in the absence of similar changes in dopamine receptor activity and extracellular dopamine concentrations. In: Proceedings of the XXIst International Symposium on Cerebral Blood Flow, Metabolism and Function, Calgary, Alberta, Canada Schiffer WK, Alexoff DL, Shea C, Logan J, Dewey SL (2005) Development of a simultaneous PET/microdialysis method to identify the optimal dose of 11C-raclopride for small animal imaging. J Neurosci Methods 144(1):25–34 Schiffer WK, Liebling CN, Patel V, Dewey SL (2007a) Targeting the treatment of drug abuse with molecular imaging. Nucl Med Biol 34(7):833–847 Schiffer WK, Mirrione MM, Dewey SL (2007b) Optimizing experimental protocols for quantitative behavioral imaging with 18F-FDG in rodents. J Nucl Med 48(2):277–287 Schiffer WK, Liebling CN, Reiszel C, Brodie JD, Dewey SL (2009a) Cue-induced dopamine release predicts cocaine preference: PET studies in freely moving rodents. J Neurosci 29:6176–6185 Schiffer WK, Liebling CN, Reiszel C, Hooker JM, Brodie JD, Dewey SL (2009b) Cue-induced dopamine release predicts cocaine preference: positron emission tomography studies in freely moving rodents. J Neurosci 29(19):6176–6185 Scott DJ, Stohler CS, Koeppe RA, Zubieta JK (2007) Timecourse of change in [11C]carfentanil and [11C]raclopride binding potential after a nonpharmacological challenge. Synapse 61(9):707–714 Seeman P, Guan HC, Niznik HB (1989) Endogenous dopamine lowers the dopamine D2 receptor density as measured by [3H]raclopride: implications for positron emission tomography of the human brain. Synapse 3(1):96–97 Siessmeier T, Zhou Y, Buchholz HG, Landvogt C, Vernaleken I, Piel M, Schirrmacher R, Rosch F, Schreckenberger M, Wong DF, Cumming P, Grunder G, Bartenstein P (2005) Parametric mapping of binding in human brain of D2 receptor ligands of different affinities. J Nucl Med 46(6):964–972 Sjoerdsma A, Engelman K, Spector S, Udenfriend S (1965) Inhibition of catecholamine synthesis in man with alphamethyl-tyrosine, an inhibitor of tyrosine hydroxylase. Lancet 2(7422):1092–1094 Slifstein M, Laruelle M (2000) Effects of statistical noise on graphic analysis of PET neuroreceptor studies. J Nucl Med 41(12):2083–2088 Small DM, Jones-Gotman M, Dagher A (2003) Feeding-induced dopamine release in dorsal striatum correlates with meal pleasantness ratings in healthy human volunteers. Neuroimage 19(4):1709–1715 Soliman A, O’Driscoll GA, Pruessner J, Holahan AL, Boileau I, Gagnon D, Dagher A (2008) Stress-induced dopamine release in humans at risk of psychosis: a [11C]raclopride PET study. Neuropsychopharmacology 33(8):2033–2041 Spector S (1966) Inhibitors of endogenous catecholamine biosynthesis. Pharmacol Rev 18(1):599–609

34 Imaging in Neurology Research III: Focus on Neurotransmitter Imaging Takeshita H, Okuda Y, Sari A (1972) The effects of ketamine on cerebral circulation and metabolism in man. Anesthesiology 36(1):69–75 Talbot PS, Frankle WG, Hwang DR, Huang Y, Suckow RF, Slifstein M, Abi-Dargham A, Laruelle M (2005) Effects of reduced endogenous 5-HT on the in vivo binding of the serotonin transporter radioligand 11C-DASB in healthy humans. Synapse 55(3):164–175 Tsukada H, Nishiyama S, Kakiuchi T, Ohba H, Sato K, Harada N (1999a) Is synaptic dopamine concentration the exclusive factor which alters the in vivo binding of [11C]raclopride?: PET studies combined with microdialysis in conscious monkeys. Brain Res 841(1–2):160–169 Tsukada H, Nishiyama S, Kakiuchi T, Ohba H, Sato K, Harada N, Nakanishi S (1999b) Isoflurane anesthesia enhances the inhibitory effects of cocaine and GBR12909 on dopamine transporter: PET studies in combination with microdialysis in the monkey brain. Brain Res 849(1–2):85–96 Tsukada H, Miyasato K, Kakiuchi T, Nishiyama S, Harada N, Domino EF (2002) Comparative effects of methamphetamine and nicotine on the striatal [(11)C]raclopride binding in unanesthetized monkeys. Synapse 45(4):207–212 Tyacke RJ, Robinson ES, Lalies MD, Hume SP, Hudson AL, Nutt DJ (2005) Estimation of endogenous noradrenaline release in rat brain in vivo using [3H]RX 821002. Synapse 55(2):126–132 Verhoeff NP, Kapur S, Hussey D, Lee M, Christensen B, Psych C, Papatheodorou G, Zipursky RB (2001) A simple method to measure baseline occupancy of neostriatal dopamine D2 receptors by dopamine in vivo in healthy subjects. Neuropsychopharmacology 25(2):213–223 Verhoeff NP, Christensen BK, Hussey D, Lee M, Papatheodorou G, Kopala L, Rui Q, Zipursky RB, Kapur S (2003) Effects of catecholamine depletion on D2 receptor binding, mood, and attentiveness in humans: a replication study. Pharmacol Biochem Behav 74(2):425–432 Volkow ND, Wang GJ, Fowler JS, Logan J, Jayne M, Franceschi D, Wong C, Gatley SJ, Gifford AN, Ding YS, Pappas N (2002) “Nonhedonic” food motivation in humans involves dopamine in the dorsal striatum and methylphenidate amplifies this effect. Synapse 44(3):175–180 Vollenweider FX, Vontobel P, Hell D, Leenders KL (1999) 5-HT modulation of dopamine release in basal ganglia in psilocybin-induced psychosis in man – a PET study with [11C] raclopride. Neuropsychopharmacology 20(5):424–433 Votaw JR, Byas-Smith MG, Voll R, Halkar R, Goodman MM (2004) Isoflurane alters the amount of dopamine transporter expressed on the plasma membrane in humans. Anesthesiology 101(5):1128–1135 Wadenberg ML, Soliman A, VanderSpek SC, Kapur S (2001) Dopamine D(2) receptor occupancy is a common mechanism underlying animal models of antipsychotics and their clinical effects. Neuropsychopharmacology 25(5):633–641 Wang GJ, Volkow ND, Fowler JS, Logan J, Pappas NR, Wong CT, Hitzemann RJ, Netusil N (1999) Reproducibility of repeated measures of endogenous dopamine competition with [11C]raclopride in the human brain in response to methylphenidate. J Nucl Med 40(8):1285–1291 Watabe H, Endres CJ, Breier A, Schmall B, Eckelman WC, Carson RE (2000) Measurement of dopamine release with continuous infusion of [11C]raclopride: optimization and signal-to-noise considerations. J Nucl Med 41(3):522–530

541

Watanabe M (1980) An autoradiographic, biochemical, and morphological study of the harderian gland of the mouse. J Morphol 163(3):349–365 Watanabe S, Fusa K, Takada K, Aono Y, Saigusa T, Koshikawa N, Cools AR (2005) Effects of alpha-methyl-p-tyrosine on extracellular dopamine levels in the nucleus accumbens and the dorsal striatum of freely moving rats. J Oral Sci 47(4):185–190 Waterhouse RN (2003) Determination of lipophilicity and its use as a predictor of blood-brain barrier penetration of molecular imaging agents. Mol Imaging Biol 5(6):376–389 Willeit M, Ginovart N, Graff A, Rusjan P, Vitcu I, Houle S, Seeman P, Wilson AA, Kapur S (2008) First human evidence of d-amphetamine induced displacement of a D2/3 agonist radioligand: A [11C]-(+)-PHNO positron emission tomography study. Neuropsychopharmacology 33(2):279–289 Wilson AA, McCormick P, Kapur S, Willeit M, Garcia A, Hussey D, Houle S, Seeman P, Ginovart N (2005) Radiosynthesis and evaluation of [11C]-(+)-4-propyl-3, 4, 4a, 5, 6, 10b-hexahydro-2H-naphtho[1, 2-b][1, 4]oxazin-9 -ol as a potential radiotracer for in vivo imaging of the dopamine D2 high-affinity state with positron emission tomography. J Med Chem 48(12):4153–4160 Wong DF, Wagner HNJ, Tune LE, Dannals RF, Pearlson GD, Links JM, Tamminga CA, Broussolle EP, Ravert HT, Wilson AA, al e (1986) Positron emission tomography reveals elevated D2 dopamine receptors in drug-naive schizophrenics. Science 234(4783):1558–1563 Wong DF, Ricaurte G, Grunder G, Rothman R, Naidu S, Singer H, Harris J, Yokoi F, Villemagne V, Szymanski S, Gjedde A, Kuhar M (1998) Dopamine transporter changes in neuropsychiatric disorders. Adv Pharmacol 42:219–223 Yoa R, Seidel J, Johnson CA (2000) Performance characteristics of the 3-D OSEM algorithm in the reconstruction of small animal PET images. Ordered-subsets expectation-maximization. IEEE Trans Med Imaging 19:798–804 Yoder KK, Wang C, Morris ED (2004) Change in binding potential as a quantitative index of neurotransmitter release is highly sensitive to relative timing and kinetics of the tracer and the endogenous ligand. J Nucl Med 45(5):903–911 Yoder KK, Kareken DA, Morris ED (2008) What were they thinking? Cognitive states may influence [11C]raclopride binding potential in the striatum. Neurosci Lett 430(1):38–42 Zald DH, Boileau I, El-Dearedy W, Gunn R, McGlone F, Dichter GS, Dagher A (2004) Dopamine transmission in the human striatum during monetary reward tasks. J Neurosci 24(17):4105–4112 Zimmer L, Mauger G, Le Bars D, Bonmarchand G, Luxen A, Pujol JF (2002) Effect of endogenous serotonin on the binding of the 5-hT1A PET ligand 18F-MPPF in the rat hippocampus: kinetic beta measurements combined with microdialysis. J Neurochem 80(2):278–286 Zovein A, Flowers-Ziegler J, Thamotharan S, Shin D, Sankar R, Nguyen K, Gambhir S, Devaskar SU (2004) Postnatal hypoxic-ischemic brain injury alters mechanisms mediating neuronal glucose transport. Am J Physiol Regul Integr Comp Physiol 286(2):R273–R282 Zubieta JK, Ketter TA, Bueller JA, Xu Y, Kilbourn MR, Young EA, Koeppe RA (2003) Regulation of human affective responses by anterior cingulate and limbic mu-opioid neurotransmission. Arch Gen Psychiatry 60(11):1145–1153

Imaging in Oncology Research

35

Wolfgang A. Weber and Fabian Kiessling

Contents 35.1

Introduction............................................................. 543

35.2

Major Applications of Imaging in Oncology Research................................................. 544 Tumor Localization and Quantification of Tumor Mass.............................................................. 544 Imaging of Gene and Protein Expression................. 547 Characterization of the Tumor Microenvironment.................................................... 550 Studies of Tumor Cell Proliferation and Metabolism............................................................... 555

35.2.1 35.2.2 35.2.3 35.2.4 35.3 35.3.1 35.3.2 35.3.3 35.3.4 35.3.5

Typical Protocols for Imaging Studies.................. 557 Optical Imaging........................................................ 557 Ultrasound................................................................ 559 mCT........................................................................... 559 MRI........................................................................... 559 Nuclear Imaging....................................................... 560

References............................................................................ 561

W.A. Weber (*) Department of Nuclear Medicine, University Hospital Freiburg Hugstetterstrasse 55, 79106 Freiburg, Germany e-mail: [email protected] F. Kiessling Department of Experimental Molecular Imaging, RWTH-Aachen, University Clinics, Helmholtz-Institute for Biomedical Engineering, Pauwelsstrasse 20, 52074 Aachen, Germany

35.1 Introduction Imaging has been extensively used to monitor the development and progression of cancer in laboratory animals. Imaging also plays an increasing role in monitoring tumor response to therapeutic interventions. Using modern imaging technologies, tumor volumes can be measured not only in subcutaneous, but also in orthotopic models allowing for noninvasive serial studies of the treated animals. Furthermore, animals can be screened for the development and progression of metastases. Using imaging techniques, it is also feasible to study noninvasively the microenvironment of the tumor tissue, such as regional perfusion, vascular permeability, or oxygen tension. In addition to these studies of tumor size, morphology, and functional status, small animal imaging is increasingly used for the development of new diagnostic and therapeutic agents, since imaging studies can greatly reduce the number of animals necessary to study the biodistribution of experimental compounds. Depending on the specific scientific question, the tumor model, and the location of disease, tumors are best studied by ultrasound, optical imaging, CT, MRI, SPECT, PET, or a combination of these techniques. Given these many applications and different techniques, it would be far beyond the scope of this chapter to provide a comprehensive overview of the literature on small animal imaging in oncology research. Instead, we focus on techniques that have been used to study the following key areas: (i) tumor localization and quantification of tumor mass, (ii) imaging gene and protein expression, (iii) characterization of the tumor microenvironment, (iv) studies of tumor cell proliferation and metabolism, and (v) monitoring therapeutic interventions. The use and limitations of imaging in

F. Kiessling and B.J. Pichler (eds.), Small Animal Imaging, DOI: 10.1007/978-3-642-12945-2_35, © Springer-Verlag Berlin Heidelberg 2011

543

544

these five areas are discussed for various organ systems. Accordingly, we do not intend to provide an overview on developing cutting-edge imaging technologies, but focus on techniques that have been thoroughly studied and whose applications and limitations have been critically evaluated. The aim is to provide the reader an overview of the features of malignant tumors that can be studied on a routine basis in small animal imaging laboratories. Furthermore, practical guidelines for imaging protocols are provided and known “pitfalls” are discussed.

35.2 Major Applications of Imaging in Oncology Research 35.2.1 Tumor Localization and Quantification of Tumor Mass In subcutaneous tumors, repeated measurements of tumor diameters continue to play a key role in oncology research. However, the limitations of the subcutaneous tumor model are well recognized and orthotopic tumor models are increasingly used in many areas of cancer research. Most orthotopic tumors are not palpable and their size cannot be measured with calipers or similar techniques. Imaging thus becomes necessary to localize the tumors and determine their growth over time. For many applications of orthotopic tumor models, a large number of animals need to be imaged repeatedly. Thus, imaging technologies need to be fast and robust. Furthermore, imaging should pose only minimal stress to the animals. Ideally, imaging should also be highly sensitive and allow for accurate measurements of small lesions. Currently, no single imaging technology does meet all these needs in all areas of the body. In the following sections the use and limitations of different imaging technologies are discussed.

35.2.1.1 Optical Imaging Bioluminescence imaging of tumors expressing luci ferases allows for high throughput and low-cost studies in mice. A key advantage of bioluminescence imaging is that no signal is produced until the substrate/enzyme interaction of a luciferase and luciferin

W.A. Weber and F. Kiessling

occurs. Consequently, there is a very low background luminescence level in most animals allowing highly sensitive detection of tumor cells that are located closely to the body surface. It has been demonstrated that the intensity of the bioluminescence signal in general correlates with the tumor mass. Furthermore, a signal is only produced by viable cells allowing investigators to differentiate between necrotic and viable tumors. However, sensitivity and spatial resolution quickly degrade for more deep-seated lesions. This is caused by significant absorption of light as well as due to scatter of the emitted photons. The minimum tumor size that can be detected by optical imaging is therefore not only dependent on the used imaging equipment, but also on luciferase expression levels and the location of the tumor. Furthermore, the positioning of the animal in the imaging system can have a significant impact on the measured light signal (Cui et al. 2008). There is therefore no general answer to the question “how many tumor cells can be detected by bioluminescence imaging.” Using state of the art equipment, as little as 10–100 superficially located cells with high luciferase expression levels can be detected by whole body bioluminescence imaging (Rabinovich et al. 2008). Since lesion detectability depends on multiple factors, it is necessary to validate the correlation between the in vivo light signal and tumor cell numbers by pilot studies with tissue sampling at different time points during tumor development. Fluorescent reporter genes such as green fluorescence protein (gfp), red fluorescent protein (rfp), and their various mutants have been used extensively to tag tumor cells and to study their migration and metastasis formation. In this context, mutants that show fluorescence emission in or nearby the near infrared (NIR) range (such as “tomato,” “plum,” and “cherry”) provide better access to tissues that are more deeply localized in the body. Expression of these fluorescent proteins can be imaged at the cellular and subcellular level by advanced microscopy techniques including confocal and multiphoton laser microscopy (MPM) (Denk et al. 1990). MPM can be used for in vivo imaging with a maximum tissue penetration of about 500 mm. This technique, which depends on the simultaneous absorption of two infrared photons by a fluorophore resulting in spatially localized fluorescence excitation, is capable of collecting fluorescence images with a spatial resolution of less than 1 mm in living animals (Dunn and

545

35 Imaging in Oncology Research

Sutton 2008). In addition to providing better penetration and resolution than previous methods of in vivo microscopy, the use of infrared light also makes MPM significantly less toxic for the studied tissues (Andresen et al. 2009). Using fluorescent probes, it is easy to label both tumor cells and the cells that interact with them. MPM systems can acquire 1–2 images per second allowing for real-time of cellular motility, invasiveness and metastasis formation. Subcutaneously implanted tumors are easily accessible for MPM after an incision in the overlying skin and have been most extensively studied. Using a surgical access, tumors have also been imaged in several organs (Kienast et al. 2010). Whole body fluorescence imaging has been used to study macroscopic tumors in mice. In contrast to bioluminescence imaging, no substrates needs to be injected for fluorescence imaging. Thus, the signal is not dependent on distribution and delivery of luciferase substrates (Hoffman 2005). Detection of small tumors fluorescence imaging is, however, in many circumstances limited by autofluorescence of normal tissues or bowel content. In addition, the limited tissue penetration of the green light emitted by gfp makes visualization of deeper lesions challenging. rfp provides better tissue penetration, but is still limited by autofluorescence. In comparative studies, whole body bioluminescence imaging provided a better signal-to-noise ratio than fluorescence imaging. Using currently available imaging technologies, gfp imaging is thus preferable for ex vivo and in vivo microscopy, whereas bioluminescence imaging is more sensitive for whole body in vivo imaging. Multimodality reporter genes have been generated that allow bioluminescence, fluorescence, and PET imaging (Ray et al. 2007). Using such reporter genes, cell culture and in vivo microscopy studies can be performed using fluorescence imaging, whereas tumor growth and metastasis formation can be monitored in mice by bioluminescence imaging. The PET reporter gene offers the opportunity for studies in larger animals and eventually translational research in patients (Kang and Chung 2008). An alternative optical imaging approach for monitoring the progression of cancer in mice is fluorescence imaging with activatable agents. In contrast to imaging with gfp, no genetic modifications of the studied tumor cells are necessary. Since this imaging technique relies on the activity of specific enzymes, it is discussed in Sect. 35.2.2.1 below.

35.2.1.2 Ultrasound Ultrasound imaging enables an easy and reliable size determination of subcutaneous tumors. In comparison to caliper measurements, it is of advantage that a third dimension can be assessed. Modern high frequency ultrasound devices using motor driven holders of the transducer generate 3D data sets that can be used for semiautomated determination of tumor volumes. This results in a highly accurate determination of tumor volumes. Also, tumors deep inside the animal body like liver tumors or experimental prostate cancers can be localized by ultrasound. However, the access to these areas of the body by ultrasound is much more difficult and often the tumors cannot be visualized properly. Therefore, ultrasound imaging may be used to check for the existence of orthotopic tumors in deep organs and to estimate their growth, but for the accurate determination of growth kinetics, other imaging modalities like MRI or mCT may be preferred.

35.2.1.3 mCT mCT enables imaging with very high spatial resolution, but usually the tissue contrast is low. This can make visualization of subcutaneous and orthotopic tumors difficult. Contrast agents usually have to be applied. Unfortunately, most clinical contrast agents do not circulate long enough in small animals to accumulate sufficiently in experimental tumors. Therefore, high contrast agent volumes have to be injected, which are not well tolerated by the animal, particularly if they are administered repeatedly. For orthotopic liver tumors, iodinated liposomes (please see Chap. 11) are helpful, which enhance the contrast of the normal liver tissue and thus enable the delineation of the tumors as darker spot. A clear indication for the use of mCT imaging are tumors in lungs and bones (Li et al. 2006). Nevertheless, the exposure of animals and tumors to X-rays has to be considered carefully in the study planning (see Chap. 10). Particularly in longitudinal studies, X-ray exposure can decrease the health status of the animal and affect tumor growth. Please also note that for the imaging of lung and liver tumors, respiratory gating should be applied, which can be done prospectively or retrospectively (Dinkel et al. 2008). Depending on the scanner

546

type and the chosen spatial resolution, measurement time per animal is in the range of seconds up to more than 30 min.

35.2.1.4 MRI MRI is favorably suited for tumor staging and monitoring in small animals. This holds true for both subcutaneous and orthotopic tumors. The minimal detectable tumor size depends on the field strength of the MR-scanner. Please note that for the detection of submillimeter-sized tumors, clinical MR-scanners usually will not be sufficient. For the determination of tumor size, simple protocols with T2-weighted sequences can be sufficient. In this context, the imaging of lung tumors is more difficult than imaging of tumors in other areas of the body. If liver or lung tumors are in the focus of research, respiratory gating should be applied, which is usually done prospectively. Usually, tumor size determination by MRI requires 10–20 min per animal, of which 5–15 min are pure scan time (the remaining time is needed for placing the animal in the MR-coil, for the adjustment of the MR-coil, and for arranging the measurement position).

35.2.1.5 Nuclear Imaging Techniques PET and SPECT imaging can provide high contrast images of tumors expressing a variety of reporter genes including herpes simplex thymidine kinase (HSV-tk), sodium iodide symporter (NIS), and norepinephrine transporter (NET) (Kang and Chung 2008). Since gamma radiation is only minimally absorbed in mice, the depth of the tumor has only a negligible influence on detectability. Images are 3-dimensional allowing accurate localization of the tumor, especially when PET or SPECT studies are combined with CT or MRI. Overall, PET and SPECT are less sensitive than bioluminescence imaging. Typically about 100,000–1 million tumor cells can be reliably detected by reporter gene imaging with small animal PET (Johnson et al. 2009). Furthermore, the costs are considerably higher as short-lived radiopharmaceuticals need to be produced for each imaging study. A key advantage of nuclear imaging techniques is that tumor can also be detected based on the increased

W.A. Weber and F. Kiessling

expression of endogenous receptors, transporters, or enzymes. Therefore, it is not necessary to genetically modify tumor cells for PET or SPECT imaging studies. This allows, for example, studies of spontaneous tumor development in transgenic animals. In this context PET imaging with fluorodeoxyglucose (FDG) is of particular interest as this radiopharmaceutical is extensively used clinically for imaging of cancer patients. Thus, FDG is readily available for animal studies at most many small animal imaging facilities located at university hospitals. When using FDG for such studies, several methodological aspects need to be considered. First of all, the PET scanner basically measures activity concentrations within the body of the animal. These can be expressed as counts/voxel or as Bq/mL if the scanner is appropriately calibrated. In order to standardize the results obtained from different experiments, activity concentrations are frequently expressed as percentage of the injected dose per milliliter tissue (%ID/mL) or as standardized uptake values (SUV). All these parameters are not directly related to tumor size or mass, because all of them describe the concentration of FDG within the tumor. In practice, there is a correlation between the PET signal intensity and tumor size, because of partial volume effects. The typical spatial resolution of small animal PET scanners is in the range of 2 mm. Thus, the true activity concentration is significantly underestimated in lesions that have a diameter of less than 4 mm. Accordingly, SUVs or other quantitative parameters will increase during tumor growth. However, the relationship between tumor size and PET signal is not linear and depends on the image reconstruction and postprocessing protocols. Furthermore, FDG uptake rates may change during tumor progression. Hypoxia, for example, may increase FDG transport and metabolism due to activation of HIF-1. Therefore, increasing FDG uptake over time cannot be interpreted as a direct reflection of tumor growth. Finally, the normal distribution of FDG needs to be considered. The rapid renal excretion of FDG makes imaging of tumors in the pelvis challenging as the high radioactivity concentration in the bladder causes artifacts in the PET images. Imaging of tumors in the brain is limited by the physiologic FDG uptake of normal gray matter and of the Harderian glands.

547

35 Imaging in Oncology Research

35.2.2 Imaging of Gene and Protein Expression The expression of genes and proteins has been assessed by genetically encoded reporters and by ligands that specifically bind to the protein of interest. In the reporter gene approach, a reporter gene is put under the control of a promoter/enhance element that regulates the expression of the gene of interest. Alternatively, a fusion gene is generated that consists of the targeted gene and the reporter gene (Kang and Chung 2008). In vivo reporter gene imaging was initially developed for monitoring cancer gene therapy (Tjuvajev et al. 1996). However, subsequent research has demonstrated that this approach is very versatile and allows for a variety of different applications. The reporter gene approach is fairly general and can in principle be used to study the expression of any gene. In contrast, the number of specific ligands for molecular imaging is still very limited. Development of ligands with the required high affinity and specificity for the targeted protein is a complex and slow process. In addition to binding with high affinity to the target, ligands need to be metabolically stable and demonstrate only low levels of unspecific binding. Furthermore, the clearance and distribution in normal organs need to be optimized. If clearance is too fast, the amount of ligand bound to the target may be too low for imaging. Conversely, a slow clearance can result in poor contrast between tumors and normal organs. For the reporter gene approach, only a limited number of ligands or substrates need to be developed that meet these criteria. In contrast, the process of ligand development needs to be repeated for each protein of interest if the protein is directly imaged by a specific ligand. On the other hand, the reporter gene approach has also specific limitations. First of all, cells need to be transfected with the reporter gene, which may confound the expression of the target protein. Second, the expression of the target protein can be influenced by posttranslational regulatory mechanisms. These generally will not be captured by reporter gene imaging. Third, the strength of the promoter may be insufficient for in vivo reporter gene imaging. This represents a frequent limitation for reporter gene imaging. Amplification strategies have been developed to overcome these limitations (Iyer et al. 2001), but these require more complex reporter gene constructs and need to be tested for the individual application.

35.2.2.1 Optical Imaging Optical imaging with genetically encoded reporters has been used by several groups to monitor gene expression in mice. A well-studied example is the use of bioluminescence imaging of hypoxia-inducible factor 1 (Hif-1). In order to image Hif-1 activity, several groups have created a reporter gene based on the fusion between the HIF-1a-ODD (oxygen dependent degradation) domain and luciferases. The resulting fusion protein (ODD-Luc) is degraded in the presence of oxygen, whereas hypoxia leads to stabilization of the protein. The ODD-Luc signal is therefore responsive to hypoxia and allows for noninvasive imaging of HIF-1 stabilization and degradation in real time (Harada et al. 2005; Moroz et al. 2009). A mouse that ubiquitously expresses an ODD-Luc reporter has been generated (Safran et al. 2006). Hif-1 activity has also been imaged by transcriptional reporters based on the binding of HIF-1 to hypoxia-response elements (HRE) which drive reporter genes (Serganova et al. 2004). Both approaches have been used to monitor Hif-1 levels and the expression Hif-1-dependent genes during various interventions (Ou et al. 2009). In recent years a variety of ligands and enzyme substrates have been developed for optical imaging of gene expression. A key enabling technology for this development has been NIR fluorochromes with excitation and emission maxima in the NIR between 700 and 900 nm and high quantum yield (Kovar et al. 2007). Commercially available imaging agents include ligands for the alpha-v beta-3 integrin, the epidermal growth factor receptor (EGFR), and phosphatidyl serine. In addition, various protease substrates can be used for optical imaging (Montet et al. 2005). These protease substrates are activated after cleavage by cathepsins (Kirsch et al. 2007) or matrix metalloproteinases (MMPs) (Bremer et al. 2001). Since a signal is only generated in this activated state, this class of imaging contrast agents is also referred to as “smart probes” or “molecular beacons.” These probes are based on a quenching–dequenching mechanism. The probes are optically silent in their native (quenched) state and become highly fluorescent after enzyme-mediated release of fluorochromes (Mahmood and Weissleder 2003).The most extensively studied groups of enzymes are the cathepsins, a group of enzymes which degrade polypeptides. Several fluorescent imaging agents are commercially available for studies of cathepsin activity.

548

Imaging of cathepsin activity allows for tumor detection without having to genetically engineer the cancer cells. Quantum dots (QDs) represent another class of imaging probes that is currently extensively studied for imaging of cancer in murine models. QDs faciliate “multiplexed imaging,” since QSs of different sizes emit a strong fluorescence signal of different wave lengths. Furthermore, excitation of multiple-colored QDs is feasible with a common excitation source (Bentolila et al. 2009; Michalet et al. 2005). QDs have been coated with a variety of targeting agents including antibodies against prostate-specific membrane antigen (PSMA) (Gao et al. 2004), epidermal growth factor (EGF) for targeting the EGFR (Diagaradjane et al. 2008), and argine–glycine– asparate (RGD) peptides for targeting the alpha-v beta-3 integrin (Cai et al. 2007). Using planar and tomographic devices, pM concentrations of fluorescent QDs can be detected in superficially located tumors (de la Zerda et al. 2009), but the signal intensity rapidly decreases with the depth of the lesion. Similarly, a submillimeter resolution can be achieved for superficial lesions, whereas resolution of planar images degrades to more than 1 cm at a depth of more than 5 mm (de la Zerda et al. 2009). These factors need to be considered when interpreting the results of optical imaging studies in mice and when using the optical signal to quantify the expression of proteins over time. Fluorescent labeling of proteins, peptides, and QDs is in many cases straightforward. However, each of these imaging probes needs to be carefully validated, since tumor uptake is not only dependent on the expression of the target, but is also influenced by unspecific binding, blood clearance, and metabolism (Kovar et al. 2007). For QDs, the fast blood clearance and high liver uptake still represent a significant limitation for in vivo imaging. Furthermore, unspecific uptake of QDs and other nanoparticles due to the enhanced permeability of intratumoral blood vessels needs to be considered (Gao et al. 2004). For example, EGF coated and uncoated QDs showed similar tumor uptake at 1 h postinjection. Subsequently, however, targeted QDs were accumulated by the tumor tissue, whereas untargeted QDs were washed out. At 24 h postinjection, the fluorescence signal had returned to baseline value for targeted and untargeted QDs (Diagaradjane et al. 2008). This illustrates that pharmacokinetic analysis may be needed to assess the expression of EGFR in the tumor tissue by nanoparticles.

W.A. Weber and F. Kiessling

35.2.2.2 MRI While ultrasound and CT do not play a relevant role for reporter gene imaging, there are few examples for the use of MRI. However, it should be noted that these are still highly experimental and may not be ready for routine use in preclinical research yet. Three concepts are mostly followed: 1. Modulation of the relaxivity of lanthanide complex by enzymatic cleavage of the probe. For example, in Xenopus embryos, which were genetically modified to express b-galactosidase, the cleavage of galactose from a gadolinium-containing chelate complex improved the interaction of the lanthanide with surrounding protons, thus leading to a positive MR signal (Louie et al. 2000). 2. Modification of the cellular iron uptake and storage. By overexpressing the iron storage molecule ferritin, a decrease in T2 and T2* relaxation time can be achieved, which means that a negative contrast of the engineered cells is given (Modo and Bulte 2007). The expression of these genes can be put under control of tetracycline, which allows to switch the contrast on and of depending on the application of the antibiotic drug. 3. Heterologous proteins that act as chemical exchange saturation transfer agents (CEST; see Chap. 13). Gilad and coworkers constructed cells that expressed a lysine-rich protein. The magnetization transfer between the bulk water and the macromolecular protons resulted in a detectable CEST effect (Gilad et al. 2007).

35.2.2.3 Nuclear Imaging Techniques Radiolabeled reporter probes have been used to study gene expression in a variety of tumor models. Compared to optical imaging approaches, the key advantage of PET and SPECT is the ability to generate 3-dimensional quantitative images that are considerably less influenced by the depth of the studied lesion. On the other hand, PET and SPECT imaging is more expensive than optical imaging as short-lived radiolabeled probes need to be synthesized for each experiment. Using similar techniques as described above for optical imaging, PET with radiolabeled substrates of HSV1-tk has been used to study the expression of

35 Imaging in Oncology Research

Hif1-regulated genes (Serganova et al. 2004). Another example is imaging of genes regulated by the tumor suppressor p53. In these studies the HSV1-tk (Doubrovin et al. 2001) or NIS (Kim et al. 2005) were expressed under the control of an artificial enhancer p53 responsive element (p53RE). Using this reporter gene construct, the upregulation of p53 activity after DNA imaged could be assessed noninvasively in cell cultures and in tumor xenografts (Doubrovin et al. 2001; Kim et al. 2005). Radiolabeled antisense oligonucleotide probes have been used to directly image endogenous gene expression at the transcription level. This imaging approach is based on short oligonucleotide sequences that bind to complementary segments of target mRNA or DNA. Scintigraphic images of endogenous gene expression using radiolabeled antisense oligonucleotides have been reported (Lendvai et al. 2009; Hnatowich 1999). However, there are formidable biologic barriers for the efficient intracellular delivery of oligonucleotides. Furthermore, the low concentrations of target mRNA or DNA limit the signal. Therefore, antisense imaging still remains experimental (Lendvai et al. 2009). There is a variety of radiolabeled ligands for imaging of proteins expressed on the cell surface. These are based on antibodies, antibody fragments, small proteins, and peptides. A thorough discussion of these various ligands is beyond the scope of this chapter. Instead, some examples are discussed for each class of ligands to illustrate the strengths and limitations of the different approaches. Radiolabeling of antibodies provides a very general approach to image the expression of cell surface molecules. However, the slow blood clearance of antibodies frequently limits the contrast between tumors and surrounding normal tissues. Furthermore, antibodies can accumulate unspecifically within tumors due to the leakiness of the intratumoral blood vessels. The specificity of the signal should therefore be confirmed by unspecific control antibodies. Since imaging needs to be performed several days after injection, radiolabeling with long-lived radioisotopes is generally necessary. This can cause logistic problems, as radioactive waste needs to be stored for longer periods of time. Radioisotopes used for labeling antibodies for PET imaging include iodine-124 (half-life 4.18 days), zirconium-89 (half-life 3.27 days). Antibody fragments of various sizes have also been used for PET imaging.

549

Antibody fragments are cleared faster from the blood stream, but also frequently show lower tumor uptake. Antibody engineering allows for optimization of pharmacokinetic parameters for imaging (Olafsen et al. 2005). For antibodies with relatively fast binding and excretion, copper-64 (half-life 0.53 days) or even fluorine-18 (half-life 110 min) can also be used (Wu and Senter 2005). Targets that have been successfully imaged with radiolabeled antibodies or antibody fragments include, among others, carbonic-anhydrase IX (CA-IX), and HER2, carcinoembryonicantigen (CEA), prostate stem cell antigen (PSCA), and PSMA (Wu 2009). Peptide hormone receptors are overexpressed on many malignant tumors. Radiolabeled peptides can be used to image the expression of these receptors (Reubi and Maecke 2008). Well-studied examples include somatostatin (Ginj et al. 2008) and bombesin receptors (Mansi et al. 2009). Radiolabeled peptides have also been extensively used to image the expression of alpha-v beta-3 integrins on cancer cells. A variety of ligands based on the RGD sequence has been tested (Haubner et al. 2001; Chen 2006). While several integrins bind to the RGD sequence, cyclic pentapeptides with a specific sterical structure allow for specific imaging of alpha-v beta-3 integrins (Haubner et al. 2001; Chen 2006). These radiolabeled peptides can be used to study the expression of alpha-v beta-3 integrins and to monitor therapeutic interventions targeting this integrin (Fig. 35.1). When interpreting these studies, one needs to consider that the relative contribution of integrins on the surface of cancer cells as compared to integrins expressed on activated endothelial cells will depend on the specific model studied. Furthermore, binding of the ligands to other integrin subtypes, such as alpha-v beta-5 integrins, has not been comprehensively analyzed. Relatively small differences in the structure of the ligand can markedly affect the specificity of target binding (Zannetti et al. 2009). When using radiolabeled RGD peptides for studies of angiogenesis, it is therefore necessary to correlate in pilot studies the measured uptake of the radiolabeled ligands with the expression of the alpha-v beta-3 integrin as determined by immunohistochemical techniques. Radiolabeled small molecules allow imaging of intracellular targets with PET or SPECT. Imaging of intracellular targets is more challenging than of

550

W.A. Weber and F. Kiessling

Fig. 35.1 Imaging the expression of alpha-v beta-3 integrins with a Cu-64 labeled RGD-peptide. The untreated U87 glioblastoma demonstrates high uptake and retention of the tracer as shown by the PET/CT images and the time activity curves. Treatment with the alpha-v beta-3 ligand c(RGDfV) causes a marked decrease in uptake and rapid washout of the radiolabeled RGD peptide

extracellular targets, since the image probe needs to cross the plasma membrane in order to reach the target. Then the probe has to be retained in cells expressing the target, whereas it should wash out rapidly from cells not expressing the target. Because of these challenges, considerably less intracellular than extracellular targets have been successfully imaged with SPECT or PET. Fluorine-18 fluoroestradiol (FES) binds with affinity to the estrogen receptor alpha and allows noninvasive imaging of this target with PET (Sundararajan et al. 2007). Conversely, androgen receptors can be studied with PET and 16beta-18F-fluoro-5alpha-dihydrotestosterone (FDHT) (Larson et al. 2004). Recently, the levels of activated caspase-3/7 have been imaged with a fluorine-18labeled isatin sulfonamide ([18F]ICMT-11) (Nguyen et al. 2009).

35.2.3 Characterization of the Tumor Microenvironment The tumor microenvironment has an important influence on the growth rate, invasiveness, and metastatic potential of cancer cells (Gatenby and Gillies 2008). Noninvasive studies to characterize the microenvironment are therefore of great interest. Parameters that can be studied by imaging include perfusion, vascular

permeability, oxygen tension, and pH. Noninvasive imaging of these parameters allows monitoring of therapeutic interventions that aim to modulate these factors.

35.2.3.1 Optical Imaging Using intravascular fluorescent dyes, the fractional vascular volume can be estimated by fluorescence tomography (Montet et al. 2007). This parameter can be used to monitor therapeutic interventions targeting intratumoral blood vessels. Intratumoral angiogenesis leads to expression of alpha-v beta-3 integrins on the surface of activated endothelial cells. Expression of this integrin can be imaged with fluorescent compounds based on the RGD sequence or nonpeptidic ligands (Chen et al. 2004; Kimura et al. 2009). However, alpha-v beta-3 integrin expression is not specific for endothelial cells. In fact, a variety of cancer cells also express significant amounts of alpha-v beta-3 integrins. Peptide and other small molecule-based imaging probes cannot be used to differentiate between expression of alpha-v beta-3 on activated endothelial cells and tumor cells. Imaging with nanoparticles may allow specific imaging of integrins expressed on endothelial cells (Cai et al. 2007).

35 Imaging in Oncology Research

551

Ultrasound imaging is favorably suited to characterize tumor neovascularization (Kiessling et al. 2009). Without the need for contrast agents, Doppler can be used to visualize neovascularization. However, even when using high frequency ultrasound scanners, the flow-dependant Doppler methods only display larger vessels above 100 mm in diameter. For the visualization of tumor microvascularization, ultrasound contrast agents should be used. Since the larger vessels with higher flow usually are more mature, the combination of nonenhanced Doppler with contrast-enhanced ultrasound scans can be used to investigate the maturity of the tumor vascularization and to monitor vascular maturation during antiangiogenic treatments (Palmowski et al. 2008a) (Fig. 35.2). Using ultrasound contrast agents, the relative blood volume (rBV) can easily be determined by performing an automated 3D-scan of the tumor and determining the amount of vascularized voxels in relation to all tumor voxels. Nevertheless, it should be noted that this leads to an overestimation of rBV due to partial volume effects. Using contrast-enhanced ultrasound, a reduction or rBV could be visualized after few hours

far before a reduction in tumor volume occurred. Blood flow, blood velocity, and perfusion in tumors can be determined using replenishment kinetics, meaning that during a stable microbubble concentration in the blood, a microbubble-destructive pulse is applied and the replenishment is recorded (Wei et al. 1998). Then, perfusion can be estimated from the initial upslope of the curve. Also, using replenishment imaging, antiangiogenic therapy effects could be determined early and before a tumor size change occurred. The use of molecular ultrasound imaging to characterize tumor angiogenesis is emerging. Among others, the overexpression of VEGFR2, E-Selectin, alpha-v beta-3-integrin, ICAM, and endoglin has been studied using targeted microbubbles in different tumor models (Kiessling et al. 2009). In this context, both phospholipid and hard-shell microbubbles (e.g., polycyanoacrylate based) have been used. In most cases coupling of the biotinylated antibodies and peptides was possible via streptavidin on the microbubble surface. Please note that biotinylated molecular ultrasound contrast agents are commercially available for animal research. Since microbubbles have a considerably short blood half life and since they can be destroyed in the tumor with high energy ultrasound pulses, the sequential

Fig. 35.2 Contrast-enhanced and noncontrast-enhanced multiple image alignments of an A431-tumor-xenograft during antiangiogenic treatment. While the tumor appears highly vascularized in the contrast enhanced scan, noncontrast-enhanced Doppler images (showing blood vessels with higher blood flow) mostly display vessels at the tumor periphery and prominent vascular bundles.

Histologically, these vessels are larger and more mature. During therapy a rapid collapse of vascularization in the tumor center can be observed by contrast-enhanced imaging. In contrast, therapy effects on the “more mature vessel fraction” displayed by the noncontrast-enhanced Doppler are less pronounced (bar: 1 mm) (image taken from Palmowski et al. 2008a)

35.2.3.2 Ultrasound

552

investigation of different molecular targets within the same session is feasible. This has been done in a study of Palmowski and colleagues, who longitudinally investigated the expression of VEGFR2 and alpha-v beta-3-integrin in squamous cell carcinomas treated with a matrix-metalloproteinase inhibitor and the expression of ICAM-1 and alpha-v beta-3-integrin in prostate cancers treated with a heavy ion therapy (Palmowski et al. 2008b, 2009). When using molecular ultrasound imaging for therapy monitoring, it is important to measure the rBV as an internal reference and to form ratios between the accumulation of sitespecific microbubbles and vascularization. By this, it is possible to distinguish a change in vessel density from a change in the molecular profile of vessels (e.g., during maturation).

35.2.3.3 CT Dynamic contrast-enhanced CT can be used to determine functional vascular parameters in tumors. Depending on the scan protocol, postprocessing can be performed using first-pass analysis (e.g., using the Miles model; (Miles 2002)) to determine rBV and perfusion or using two (or more) compartment models to derive data about rBV, perfusion, and vessel permeability (Brix et al. 1999). However, when intending to perform perfusion CT studies, keep in mind that beam hardening artifacts coming from the contrast agent bolus within large arteries and veins can significantly disturb the assessment of the concentration time curve in the tumor. Furthermore, the tumor enhancement can be low (usually only 20–60 HU) and a good contrast-to-noise ratio is required to get a reliable curve fit. Please also carefully consider that the X-ray dose applied in dynamic contrast-enhanced mCT-scans is considerably high and can be critically high if several imaging sessions are performed within a short period of time.

35.2.3.4 MRI MRI offers a broad variety of techniques to characterize the tumor microenvironment and only the most frequently used ones will be mentioned in this book chapter. DCE MRI is well established to characterize tumor neovascularization. Postprocessing is mostly done by using two-compartment models (mostly the models of

W.A. Weber and F. Kiessling

Brix and Tofts and modifications thereof) (Kiessling et al. 2007). Semiquantitative descriptors of rBV, perfusion, and vessel permeability can be obtained and their capability to characterize tumor angiogenesis has been shown frequently. For example, using the permeability weighted “exchange rate constant,” higher vascular maturation in squamous cell carcinomas after PDGF transfection could be demonstrated (Lederle et al. 2010). Also, the response of tumors to different kind of treatments like radiotherapy, chemotherapy, and antiangiogenic therapy has been demonstrated convincingly (Fig. 35.3). In line with the findings observed by ultrasound, changes in vascularization often preceded the change of tumor volume. In this context, parameters of rBV and vessel permeability/perfusion (such as ktrans and kep, which derive from two-compartment models) were identified as sensitive markers. Superiority of the one or the other marker seems to highly depend on the tumor model used. Nevertheless, for the accurate quantification of these functional vascular parameters, it is recommended to measure the arterial input function in a large vessel (e.g., the aorta). The use of MR contrast agents of higher molecular weight (e.g., gadoliniumloaded albumin or gadolinium-containing polymers) further facilitates the correct assessment of the signal time course by slowing down the contrast agent exchange with the interstitial space and thus enabling better separation of perfusion and vessel permeability. Susceptibility weighted steady state MRI is an alternative to determine the rBV. Here, long circulating ultrasmall superparamagnetic iron oxide nanoparticles USPIO are injected and the T2*-relaxation times are recorded before and after contrast agent injection. Bremer and colleagues showed that this method can be used to catch the vascular breakdown following the administration of vascular disruption agents with high congruency to histology (Persigehl et al. 2007). Combining T2- and T2*-relaxometry before and after administration of USPIO (or other MR-contrast agents) enables the determination of the mean vessel size in tissues by a method called vessel size imaging (Tropres et al. 2001). Using vessel size imaging, changes in the vascular architecture (e.g., maturation) occurring during antiangiogenic therapy can be assessed (Zwick et al. 2009). However, up to now, results obtained are controversial and seem to highly depend on the used tumor model. If a MR scanner with sufficiently high field strength (>4.7 T) is available, vessel maturity and

35 Imaging in Oncology Research

553

Fig. 35.3 Example for a DCE MRI study to track tumor response to antiangiogenic therapy. Color-coded parameter maps of “Amplitude” (a marker correlating with the relative blood volume) of a VEGFR2-antibody(DC101)-treated and an untreated tumor are shown before (0) and 1, 2, and 4 days after the start of treatment. Two days after the start of antibody treatment, spots occurred in the tumor center, which did not show

any contrast enhancement (arrows). After 4 days, no central enhancement was found and thus no amplitudes could be calculated in central tumor areas anymore (arrows). During these 4 days of treatment, no significant differences were observed in tumor volume between the treated and the untreated tumors yet (image taken from Kiessling et al. 2004)

vascular volume fraction can also be investigated with “blood oxygenation dependant imaging (BOLD; see Chap. 6.1),” which was proven by the group of Michal Neeman, who used it to investigate vessel responsiveness and maturation in ovarian tumor xenografts (Gilead et al. 2004). Diffusion-weighted imaging (DWI) is a MRI method to investigate the motion of protons in tissues and it is based on the Brownian motion. The higher the cell density, the more will be the diffusion barriers and the lower will be the diffusion. This leads to high signal intensities in diffusion-weighted images and to low “apparent diffusion coefficient (ADC)” values. Thus, DWI is an indirect measure of cellularity. In preclinical studies DWI can favorably be used to characterize tumors in nonmoving organs (e.g., bones and the brain). There are few reports on the use of DWI to monitor the efficacy of chemo- and antiangiogenic therapy in rodents; however, results obtained were controversial. Different cell types contribute to the development of a malignant stroma. Among those are macrophages, endothelial progenitor cells, and (myo-)fibroblasts. In order to investigate migration of these cells to tumors, they can be labeled in vitro by superparamagnetic iron

oxide nanoparticles. In this context, since the cellular uptake of dextran-coated commercial SPIO and USPIO usually is relatively low, transfection agents (e.g., protamine sulfate) should be used to support internalization. Alternatively, cell penetrating peptides such as HIV-tat may be bound to the surface of iron oxide nanoparticles. Imaging usually is performed with T2*-weighted gradient echo sequences. However, it is recommended to additionally perform T2 or T2* relaxometry. Besides the subjective visual inspection, this enables quantification of the signal changes after injection of the labeled cells. Please note that for most in vivo cell tracking experiments (except for the case that a very high number of labeled cells accumulate in the tumors), high field MR-scanners (<4.7 T) or additional gradient inserts in clinical MR-scanners are required. Additionally to the imaging of cells supporting the formation of a malignant stroma, cell tracking can be used to monitor the success of cell-based treatments. For example, Kircher et al. successfully used MRI to monitor the accumulation of magnetically labeled, activated T-cells in melanoma xenografts and to monitor tumor response (Kircher et al. 2003) (Fig. 35.4). Loading cells in vitro with USPIO or other MR contrast agents is considerably easy as compared to

554

W.A. Weber and F. Kiessling

Fig. 35.4 Intratumoral distribution of sequentially administered OT-I CD8+ T cells. Mice were implanted with B16melanoma (left) and OVA-transfected B16 melanoma cells (right). Ten days later, 107 OT-I CD8+ T cells were labeled with CLIO-HD and adoptively transferred into the recipient (0 h). Sequential adoptive transfers of 107 CLIO-HD-labeled OT-I CD8+ T cells were administered to the same mouse at 48 and 96 h. MR imaging at 8.5 T was performed 12 h after each adoptive transfer of CLIO-HD-labeled OT-I CD8+ T cells; (a–d) axial slices through the mouse thighs at (a), before adoptive transfer; (b) 12 h after first adoptive transfer; (c) 12 h after

second adoptive transfer (60 h); (d) 12 h after third adoptive transfer (108 h); indicating the heterogeneity of OT-I CD8+ T cell recruitment to the B16-OVA tumor with sequential injections. (e–h) Three-dimensional-rendering of the B16-OVA tumor at (f), 16 h (red, first injection); (g) 60 h (green, second injection); (h) 108 h (blue, third injection). (e) Color-coded compilation of (f–h). It can clearly be seen by MRI that the activated OT-I CD8+T cells selectively home to the OVAtransfected tumor and inhibit its progression (image taken from Kircher et al. 2003)

in vivo targeting. Since the sensitivity of MRI for contrast agents is relatively low, liposomal compounds including thousands of gadolinium chelates or USPIO have to be bound to the targeting moiety to achieve a visible (or measurable) contrast in the target tissue. Note that binding one or two gadolinium chelates to an antibody or a peptide usually is not sufficient. Large imaging probes, unfortunately, often go along with unfavorable pharmacokinetic behavior, high RES uptake, and long-lasting unspecific deposition in the interstitial space. Using this kind of MR imaging probes, targets at the endothelium can be reached most easily and this may be one expla nation why many groups prefer to image markers of angiogenesis targets like alpha-v beta-3-integrins. There are only few reports on using molecular MR-imaging to track therapy response. Nevertheless, Mulder and colleagues showed the proof of principle and demonstrated a reduced accumulation of avb3targeting liposomes to tumor xenografts after antiangiogenic therapy (Mulder et al. 2007).

Nevertheless, with some exceptions, molecular MRI is a playground for groups being highly experienced in the design of imaging compounds, but presently, it cannot be recommended as a standard tool for biomedical research in a small animal imaging unit.

35.2.3.5 Nuclear Imaging Techniques Several radiolabeled imaging probes have been developed to image tumor hypoxia with PET (Serganova et al. 2006). The most extensively studied imaging probe for hypoxia imaging is fluorine-18 labeled fluoromisonidazole (FMISO). Other imaging probes include fluorine-18 labeled fluoroazomycin arabinoside (FAZA) and copper-64 labeled copper(II)-diacetyl-bis(N(4)methylthiosemicarbazone (ATSM) (Serganova et al. 2006). These probes provide a higher contrast than FMISO, but their ability to image hypoxia has been less thoroughly validated than for FMISO. On a macroscopic level, FMISO uptake has been found to correlate

35 Imaging in Oncology Research

well with other markers of hypoxia, such as oxygen probe measurements, CA-IX expression, and pinomidazole binding (O’Donoghue et al. 2005). FMISO PET cannot, however, resolve regional difference in hypoxia on a microscopic level.

35.2.4 Studies of Tumor Cell Proliferation and Metabolism Unregulated proliferation is a common characteristic of malignant tumors. Therefore, imaging of cellular proliferation is of high interest for oncological research. Proliferation is closely linked to metabolism, since proliferation requires DNA synthesis, which again is dependent on the synthesis of nucleotides. In recent years there has also been renewed interest in the altered tumor glucose metabolism of cancer cells. In normal mammalian cells, glycolysis is inhibited by the presence of oxygen, which allows mitochondria to oxidize pyruvate to CO2 and H2O. This inhibition of glycolysis is termed the “Pasteur effect,” after Louis Pasteur, who first demonstrated that glucose flux was reduced by the presence of oxygen. Conversion of glucose to lactic acid in the presence of oxygen is known as aerobic glycolysis and was reported by Otto Warburg at the beginning of the twentieth century as a specific metabolic abnormality of cancer cells (Kroemer and Pouyssegur 2008). Recent studies have now indicated that several oncogenes involved in the development and progression of common human cancers also play an important role in the regulation of glycolysis. For example, unregulated activity of the serine-threonine kinase Akt has been shown to increase glucose uptake of tumor cells as well as increase resistance to apoptosis. The oncogene c-myc, a transcription factor, directly binds numerous glycolytic genes (hexokinase 2, enolase, and lactate dehydrogenase A) and activates their expression. The “glycolytic phenotype” is increasingly considered as necessary for tumor progression, since it provides the necessary building blocks for the synthesis of nucleosides and also allows cancer cells to grow in a hypoxic environment (Kroemer and Pouyssegur 2008; Gatenby and Gillies 2004; Vander Heiden et al. 2009). In addition to increased glycolytic activity, overexpression of choline kinase (ChoK) is frequently

555

observed in many malignant tumors. Experimental studies have indicated that various oncogenic sig naling pathways including phosphatidylinositol3-kinase (PI3K) and HER2 lead to increased FAS expression (Van de Sande et al. 2002). Several oncogenes such as ras, src, raf, and mos increase choline kinase activity when expressed in mouse fibroblast (Aoyama et al. 2004). Likewise, transfection of human mammary epithelial cells with the erbB2 oncogene has been reported to cause a significant increase in phosphocholine levels (Ramirez de Molina et al. 2004). Choline kinase activity and phosphocholine levels were generally not well correlated with proliferation rates (Bhakoo et al. 1996), suggesting that activation of choline kinase cannot be explained alone by the cell membrane synthesis of proliferating cells. Another metabolic abnormality of cancer cells are the accelerated rates of amino acid transport by the transport systems A and L (McGivan and PastorAnglada 1994). Mammalian cells take up amino acids by a set of transport molecules with overlapping substrate specificity. Amino acid transport systems have been functionally characterized by affinity for specific amino acids, sodium dependency, and sensitivity to inhibitors before the actual transporter proteins had been identified. Based on functional characteristics, 13 major amino acid transports system can be differentiated (McGivan and Pastor-Anglada 1994). These include the sodium-dependent systems A, ASC, N, Gly, B0, B, B0,+, XAG−, b and the sodium independent systems L, y+, b0,+, xC− (McGivan and Pastor-Anglada 1994). For imaging purposes, system L is particularly interesting, since several imaging probes are transported by this system. System L mediates sodium independent transport of aromatic and branched chain amino acids. There are currently four known proteins that mediate l type amino acid transport, designated LAT1–LAT4. LAT1 is physiologically expressed in many human tissues including the brain where it mediates transport of neutral amino acids across the blood brain barrier (Verrey 2003). Following transport across the cell membrane amino acids can enter protein synthesis or multiple other anabolic and catabolic processes. However, during the relatively short time interval that can be studied by PET, amino acid transport generally appears to be the dominant factor for the uptake of most clinically used radiolabeled amino acids (Ishiwata et al. 1993).

556

35.2.4.1 Optical Imaging There are few fluorescent probes for studies of tumor cell metabolism and proliferation. The chemical properties of small molecules such as nucleosides, glucose, or amino acid are significantly altered by the fluorescent dyes which have a larger molecular weight than the labeled substrates themselves. A fluorescent glucose analog for in vivo imaging is commercially available (Kovar et al. 2009). This compound is accumulated in experimental tumors. However, the uptake mechanisms are not well understood. Since the compound is not taken up by the brain, it is not clear whether its uptake actually reflects glucose metabolism (Kovar et al. 2009).

35.2.4.2 MRI Metabolism and indirect markers of cell proliferation in tumors can be investigated by MR-spectroscopy (see Chap. 14). In this context, choline is typically upregulated in tumors and can be seen as an indicator of rapid membrane turnover, which is a typical feature for proliferative tissues. In proton MR spectra, choline can be identified even with clinical MR-scanners, and in experimental tumors, it has been shown to negatively correlate with therapy response, e.g., under radiotherapy. In this context, creatine often is used as an internal standard and ratios between choline and creatine are formed. Another prominent peak derives from lipids. Increasing amounts of free lipids have been shown to correlate with tumor aggressiveness in orthotopic liver tumors (Foley et al. 2001) and experimental prostate cancers. Using 31P-MR-spectroscopy, the intracellular pH-value of tumor tissues can be investigated. However, it is important to note that in tumors the extracellular pH-value usually is decreased, but due to upregulation of proton pumps in tumor cells the intracellular pH-value in tumors can be normal or even increased. Using high-field MR scanners, many more interesting metabolites such as N-acetylaspartate, lactate, and taurine can be separated, but it would go beyond the scope of this chapter to discuss them all in detail. In addition to spectroscopy, the pH value can also be obtained using pH-sensitive CEST-probes (see Chap. 14), which, however, are not commercially available yet and still in the process of evaluation.

W.A. Weber and F. Kiessling

These will mostly measure the extracellular pH value and thus may be used complementary to phosphorspectroscopy.

35.2.4.3 Nuclear Imaging Techniques For imaging of tumor cell proliferation, thymidine and thymidine analogs are of special interest, since thymidine is the only nucleoside which is exclusively in corporated into DNA, but not RNA. Radiolabeled thymidine has therefore been used since many years to study cellular proliferation in vitro. However, thymidine is too unstable to be used for routine in vivo imaging of tumor proliferation. By far the most extensively studied probe for imaging of cellular proliferation is, therefore, the thymidine analog 3¢-deoxy-3¢-18Ffluorothymdine (FLT) (Shields et al. 1998). Similar to thymidine, FLT is transported across the cell membrane by nucleoside transporters (Krohn et al. 2005). Intracellularly, FLT and thymidine are phosphorylated by thymidine kinase 1 (TK-1). In contrast to thymidine, however, FLT is not incorporated into DNA. Studies have indicated that TK-1 activity is the key factor determining the amount of FLT uptake by cancer cells. In normal cells and most cancer cells, TK-1 activity is strongly upregulated during S phase, while there is only little activity during the other phases of the cell cycle. Studies have shown that FLT flux as measured by dynamic PET studies as well as FLT uptake at a fixed time postinjection is reasonably well correlated with histopathologic markers of tumor cell proliferations, such as the Ki-67 labeling index (Krohn et al. 2005). Tumor glucose metabolism can be imaged by PET with the glucose analog fluorine-18 fluorodeoxyglucose (FDG). Following intravenous injection, FDG is transported across the cell membrane by sodium-independent facilitative glucose transporters (Gluts). In malignant tumors Glut-1 is frequently overexpressed, but expression of Glut-3, and more recently, Glut-12 has also been reported in some tumor types (Macheda et al. 2005). Unlike glucose, FDG is not a substrate for the sodium-dependent glucose transporters found in the tubulus system of the kidneys. As a consequence, FDG is not reabsorbed after glomerular filtration, but excreted with the urine. This contributes to the rapid clearance of FDG from the blood stream which is important for imaging metabolically active tissues

557

35 Imaging in Oncology Research

with high contrast. Intracellularly, FDG and glucose are phosphorylated by hexokinase to glucose-6-phosphate and FDG-6 phosphate, respectively. Glucose-6 phosphate is then further metabolized to fructose1,6-biphosphate and enters glycolysis. Alternatively, glucose-6-phosphate enters the pentose phosphate pathway and is eventually converted to ribose-5-phosphate, which can serve as a building block for DNA and RNA synthesis. In contrast to this complex metabolic fate of glucose-6-phosphate, FDG-6-phosphate cannot be further metabolized in the glycolytic pathway because the fluorine atom at the C2 position prevents FDG-6P from further degradation. While FDG undergoes only the first two steps of glucose metabolism (transport and phosphorylation by hexokinase), it is important to note that FDG flux, nevertheless, reflects exogenous glucose utilization of cancer cells, as long as patients are imaged at steadystate conditions. At steady-state conditions, the concentrations of the various metabolites produced during glycolysis are constant and the net glucose flux across the cell membrane equals total exogenous glucose utilization. Consequently, uptake rates of FDG are not only dependent on the activity of glucose transporters and hexokinase, but also on the activity of downstream molecules. Radiolabeled choline ([11C]choline) (Hara et al. 1998) and the choline analogs ([18F]choline and [18F] fluoroethylcholine) (DeGrado et al. 2001; Hara et al. 2002) have been used to study choline metabolism by PET. In patients radiolabeled choline and choline analogs are rapidly and intensely accumulated in a variety of tumors, including prostate cancer, gliomas, nonsmall cell lung cancer, and esophageal cancer. In mice, however, uptake of choline and choline analogs has been found to be low by many investigators. The reasons for this unexpectedly low uptake are currently not well understood. Radiolabeled methionine (l-[11C-metyl]-methionine, MET), tyrosine (l-[1-11C]tyrosine), and various tyrosine analogs including l-[2-18F]fluorotyrosine, O-(2[18Ffluoroethyl])-l-tyrosine (FET), l-[3-18F]fluoro-amethyl tyrosine (FMT), and l-[3-123I]iodo-a-methyl tyrosine (IMT) represent the most commonly used amino acid tracers (Plathow and Weber 2008). A comprehensive review of these various imaging probes is beyond the scope of this review and only certain common characteristics are addressed here. Cell culture studies indicate that l-type amino acid transport

appears to be a common mechanism for tumor uptake of MET, the tyrosine/phenylalanine analogs (Langen et al. 2006). Other amino acid transporters also contribute to a varying, but overall lesser degree, to the total cellular uptake of these amino acid tracers. Consistent with these in vitro experiments, in vivo studies in tumor-bearing animals and patients with brain tumors have shown a close correlation between the uptake of MET and IMT, and MET and FET (Langen et al. 2006). The main applications of PET scans with radiolabeled substrates are monitoring of tumor progression (see Sect. 35.2.1) and evaluating tumor response to therapy (Fig. 35.5). A series of studies has indicated that FLT uptake decreases very rapidly in response to radiotherapy, cytotoxic chemotherapy (Barthel et al. 2003), and various protein kinase inhibitors (Wei et al. 2008; Ullrich et al. 2008). In some (Barthel et al. 2003; Ullrich et al. 2008), but not all of these studies (Wei et al. 2008), changes in FLT uptake better reflected the effects of therapy than change in FDG uptake. Some protein kinase inhibitors such as gefitinib and rapamycin cause very rapid changes in FDG uptake, since their targets are involved in regulating tumor glucose use (Wei et al. 2008).

35.3 Typical Protocols for Imaging Studies In the following please note that the estimated examination times are based on our experience and highly depend on the imaging device, the chosen spatial resolution, and the protocol.

35.3.1 Optical Imaging For fluorescence imaging with genetically encoded proteins, no special animal preparation is required. If imaging tumors close to the abdominal cavity, one should consider that chlorophyll, which is often present in animal chow, absorbs at 655 and 411 nm, and fluoresces at 673 nm and thus producing strong signal in the abdominal cavity. For optimal fluorescent imaging performance, purified food formulation that does

558

W.A. Weber and F. Kiessling

Fig. 35.5 Monitoring tumor response to therapy with PET and bioluminescence imaging. Treatment of U87 human glioblastoma xenograft causes a marked growth inhibition (left). After 3 days of rapamycin therapy, there is a marked decrease in tumor proliferation and metabolism as shown by FLT- and FDG-PET

studies. However, there is no apparent change in the bioluminescence signal indicating that the tumor cells are still viable at the time of PET imaging. Thus, the decrease of the PET signals specifically reflects changes in tumor cell proliferative activity and is not caused by cell death

not contain plant products may be used (Kovar et al. 2007). The nude mouse is an ideal choice for NIR optical imaging, because fur interferes with imaging by blocking, absorbing, and scattering light. Tumor uptake of fluorescent dyes and image contrast are dependent on the time after injection as well as the route of administration (i.v. or i.p). Furthermore, the dose of the fluorescent dye can affect the measured signal. High doses may saturate specific target binding, whereas signal-to-noise ratios may be unfavorable for low doses. Pilot studies are frequently necessary to optimize the injected dose and the timing of optical scans in order to ensure sufficient clearance of background activity and maximize specific tumor uptake. The intensity of the bioluminescence signal is strongly dependent on the time after luciferin injection. For tumors expressing firefly luciferase, the signal typically reaches a peak within 20–30 min after i.p. injection of 100 mg/kg luciferin (Cui et al. 2008) and

then decreases gradually. Time to peak is shorter for i.v. injection (Keyaerts et al. 2008). The time to peak light signal varies between different cell lines and is also influenced by the size of the tumor. By mechanisms that are currently not fully understood, anesthesia can markedly influence the intensity signal. In one study the bioluminescence signal was more than threetimes stronger under anesthesia with ketamine/xylazine than under anesthesia with isoflurane (Cui et al. 2008). Given the large number of potentially confounding factors, it is recommended to perform pilot studies in order to define the time to peak light output and the duration of the peak signal for each tumor model studied. For planar fluorescence and bioluminescence imaging, one should also consider that the signal is strongly depth-dependent. Therefore, imaging animals in the prone and supine position of the animal can be helpful to visualize lesions.

559

35 Imaging in Oncology Research

35.3.2 Ultrasound Tumor size determination can be performed automated in 3D or by hand in two dimensions and will take about 1 min/animal. High frequency Doppler scans of subcutaneous tumors may be used to get a rough orientation on tumor vascularization; however, by this, microvessels will not be captured (2 min/animal). Contrast-enhanced scan may be applied subsequently, which can be performed in 2D or 3D. Since catheterization of the animals will be required, the usual examination time increases to about 10–15 min/animal. The same holds true for molecular ultrasound imaging (approximately 20 min/animal). Here, usually one scan is done before microbubble injection for tumor size determination and orientation. Then, 5–20 min after microbubble injection a low MI scan is used to assess the amount of microbubbles in the tissue. Subsequently, a microbubble-destructive pulse is applied followed by a second scan after few seconds, which only captures the nonstationary microbubble fraction. The amount of stationary microbubbles can be obtained by subtracting the signal levels before and after the destructive pulse. Alternatively, SPAQ (sensitive particle acoustic quantification (Reinhardt et al. 2005)) scans before and after microbubble injection (approximately 10 min/ animal) may be used to quantify the amount of stationary microbubbles.

35.3.3 mCT For tumor staging, it is usually sufficient to image mice with a resolution of 50–100 mm. Using dedicated animal scanner, this should not take more than 1–10 min. Eventually, respiratory gating has to be performed (e.g., for lung metastases). We recommend retrospective gating (intrinsic or extrinsic) since the animals suffer from intubation (required for prospective gating). Gating can increase the scan time by a factor of 3–5.

35.3.4 MRI MRI offers a broad variety of scan options, but also is time intensive. Imaging protocols have to be adapted

to the needed outcome parameters, but should be kept as small as possible to minimize the scan time (Modo and Bulte 2007).

35.3.4.1 Tumor Size (~15 min/Animal) Tumor size determination usually can be performed with a short protocol including T2-weighted sequences, which usually give the best tumor contrast. Protocol: • Localizer • Sagittal or coronal T2-weighted (spin-echo or turbo spin-echo) sequence (2D or 3D) • Transversal T2-weighted (spin-echo or turbo spinecho) sequence

35.3.4.2 Tumor Size and Morphology (~25 min/ Animal; Catheterization Required) Additional information on tumor size can be obtained without the need to inject contrast agents by including DWI and MR-spectroscopy. Please note that examination times significantly increase if animals have to be catheterized. Protocol: • Localizer • Sagittal or coronal T2-weighted (spin-echo or turbo spinecho) sequence • Transversal T2-weighted (spin-echo or turbo spinecho) sequence • T1-weighted spin-echo or gradient echo sequence before and after contrast agent injection • Eventually DWI (5–10 additional minutes) • Eventually MR-spectroscopy (10–15 additional minutes)

35.3.4.3 Tumor Vascularization (30 min/Animal; Catheterization Required) For the assessment of tumor vascularization, DCE MRI is frequently used. In this context, it is recommended to perform a T1-relaxometry measurement prior to the dynamic scan to enable the conversion of

560

the signal intensity time curve into a concentrations time curve. A T1-weighted sequence may be applied at the end to get a well-contrasted high signal-to-noise image of the tumor. Protocol: • Localizer • Sagittal or coronal T2-weighted (spin-echo or turbo spin-echo) sequence • Transversal T2-weighted (spin-echo or turbo spinecho) sequence • T1-weighted sequence • T1 relaxometry (to enable later conversion of signal intensities in concentrations) • DCE MRI scan (contrast agent injection) • T1-weighted sequence

35.3.4.4 Molecular Imaging (30–40 min/Animal) For molecular MR imaging, it is most important to get quantitative data. For this purpose, relaxometric measurements should be performed in combination with high-resolved morphological images at high signalto-noise levels. If possible, all sequences should be in identical orientation and with identical geometry. Please also take care of suitable controls, e.g., receptor positive and negative tumors (in the same mouse), negative control probes (e.g., with a nonbinding peptide or antibody), and competition experiments. Protocol: • Localizer • T2-weighted (spin-echo or turbo spin-echo) sequence • T1- T2- or T2*-relaxometry • Spatially high-resolved T2*- or T1-weighted sequence

35.3.5 Nuclear Imaging Imaging protocols for PET and SPECT imaging are obviously very much dependent on the radiotracer used and the animal model studied. Nevertheless, some general considerations apply to most types of studies. First of all, PET and SPECT can be acquired as “static”

W.A. Weber and F. Kiessling

or “dynamic” scans. In a static scan the distribution of radioactivity at one point in time is imaged. Imaging is usually performed when uptake in the tumor has reached a plateau and activity has sufficiently cleared from surrounding normal tissues. Furthermore, the radiolabeled probe still needs to emit sufficient signal at the time of imaging. Typically, imaging can be performed during 1–2 physical half-lives of the radiolabel, i.e., 2–4 h for fluorine-18, 1–2 h for gallium-68, and 12–24 h for copper-64. When tumor uptake of the used radiotracer does not reach a plateau during the time frame that is available for imaging, it is very important that all studies are performed at the same time postinjection. Otherwise, there may be significant variability in the measured radiotracer uptake values (Fueger et al. 2006). In a dynamic scan, imaging is started immediately after tracer injection and a series of images is acquired over a time period of 1–2 h. Using dynamic imaging, one can therefore measure the amount of the radiotracer taken up by the tumor over time. In addition, the clearance of the tracer from the circulation can be determined. Using these data and compartmental modeling or other tracer kinetic approaches, one can calculate tumor uptake rates of the studied radiopharmaceutical. In contrast to uptake values, uptake rates are no longer time-dependent. Furthermore, tracer kinetic modeling allows for corrections of differences in the clearance of the tracer from the circulation. Thus, dynamic imaging can reduce the inter and intra-animal variability of quantitative parameters derived from SPECT or PET studies. In some cases it is even possible to correct for differences in perfusion and vascular permeability by compartmental modeling of dynamic PET studies. However, dynamic imaging has also limitations. Dynamic imaging takes considerably more time than static imaging. For example, a typical static FDG scan of a mouse can be acquired in 5–10 min, whereas a dynamic study requires 40–60 min. Consequently, only relatively few animals can be imaged in one session. Furthermore, the stress for the animal is higher, especially when repeated PET scans have to be performed for treatment monitoring. Tracer kinetic modeling to correct the influence of perfusion and vascular permeability is not always feasible. For example, it may be challenging to measure the clearance of the radiotracer from the blood due to the presence of radiolabeled metabolites. Considerable experience and mathematical knowledge are also required in order to

561

35 Imaging in Oncology Research

obtain reliable corrections. Tracer kinetic modeling is therefore not routinely performed at many sites. When using receptor ligands for PET or SPECT imaging, it is important to consider the mass of the tracer injected. Although scintigraphic techniques are extremely sensitive, radiolabeling of the tracer is frequently not highly efficient. Therefore, a considerable amount of unlabeled tracer may be injected into the animal. This unlabeled tracer can partially block binding of the radiolabeled tracer to the receptor. On the one hand, this can limit image contrast and lesion detectability. On the other hand, it may cause errors in quantitative measurements if the amount of unlabeled tracer varies between different studies. The effect of unlabeled tracer on tracer uptake is often called the “mass effect.” It is important to know the specific activity (the amount of radioactivity per mol injected tracer) of the injected tracers in order to judge whether the target may be partially blocked. In this context one also needs to consider that for short-lived radioisotopes the specific activity decreases rapidly over time. For example, if a gallium-68-labeled tracer is injected 3 h after synthesis, its specific activity will be eight times lower. Consequently, eight times more mass is injected which may reduce the measured signal and image contrast. For metabolic tracers such as FDG, anesthesia and animal preparation can have a profound effect on tumor uptake and image contrast. For example, ketamine/ xylazine, a commonly used anesthesia for rodents, induces marked hyperglycemia in mice. Since glucose competes with FDG for uptake and phosphorylation, this can markedly decrease tumor FDG uptake. Isoflurane anesthesia has a less pronounced, but still measurable impact on glucose levels, which is highly dependent on the studied mouse strain. In some strains isoflurane has been shown to cause mild hyperglycemia, whereas in other hypoglycemia has been observed (Fueger et al. 2006; Flores et al. 2008). Isoflurane anesthesia during the uptake period can considerably stimulate myocardial FDG uptake by mechanisms that are currently not understood. When mice are not warmed before and after FDG injection, thermoregulatory brown adipose tissue becomes activated resulting in very intense FDG uptake in the neck and upper thoracic area of mice, which can severely affect tumor detection in this area (Fueger et al. 2006). For a typical FDG-PET study, the mice cages are, therefore, put on a temperature-controlled heating pad

about 30 min prior to the planned FDG injection (Flores et al. 2008). Mice are anesthetized prior to FDG injection, since injection in conscious mice results in high uptake by skeletal muscles. Overnight fasting may increase tumor FDG uptake, but also results in weight loss, especially when repeated imaging studies are performed. Most investigators inject FDG in a tail vein, but FDG is rapidly absorbed following i.p. injection. At 1 h p.i. tumor, FDG uptake is comparable after i.v. and i.p. injection (Fueger et al. 2006). However, small amounts of FDG may still be present in the abdomen and can interfere with the detection of intraabdominal lesions. Static PET imaging is performed about 60 min postinjection. Using current systems about 10 MBq, FDG are injected per mouse and images are acquired for 5–10 min. Most investigators do not use attenuation correction for the PET scan, since the effect on the measured tracer uptake is small in mice. Using such a protocol, tumors as small as 2 mm can be visualized by FDG-PET (Woo et al. 2008). However, metabolic activity significantly varies across different tumor models and markedly influences the detection limits. Test-retest reproducibility of FDG-PET scans acquired according to standardized protocol is good with a coefficient of variation of repeated measures in untreated animals of about 15% (Dandekar et al. 2007).

References Andresen V, Alexander S, Heupel WM, Hirschberg M, Hoffman RM, Friedl P (2009) Infrared multiphoton microscopy: subcellular-resolved deep tissue imaging. Curr Opin Biotechnol 20:54–62 Aoyama C, Liao H, Ishidate K (2004) Structure and function of choline kinase isoforms in mammalian cells. Prog Lipid Res 43:266–281 Barthel H, Cleij MC, Collingridge DR, Hutchinson OC, Osman S, He Q et al (2003) 3¢-deoxy-3¢-[18F]fluorothymidine as a new marker for monitoring tumor response to antiproliferative therapy in vivo with positron emission tomography. Cancer Res 63:3791–3798 Bentolila LA, Ebenstein Y, Weiss S (2009) Quantum dots for in vivo small-animal imaging. J Nucl Med 50:493–496 Bhakoo KK, Williams SR, Florian CL, Land H, Noble MD (1996) Immortalization and transformation are associated with specific alterations in choline metabolism. Cancer Res 56:4630–4635 Bremer C, Tung CH, Weissleder R (2001) In vivo molecular target assessment of matrix metalloproteinase inhibition. Nat Med 7:743–748

562 Brix G, Bahner ML, Hoffmann U, Horvath A, Schreiber W (1999) Regional blood flow, capillary permeability, and compartmental volumes: measurement with dynamic CT–initial experience. Radiology 210:269–276 Cai W, Chen K, Li ZB, Gambhir SS, Chen X (2007) Dualfunction probe for PET and near-infrared fluorescence imaging of tumor vasculature. J Nucl Med 48:1862–1870 Chen X (2006) Multimodality imaging of tumor integrin alphavbeta3 expression. Mini Rev Med Chem 6:227–234 Chen X, Conti PS, Moats RA (2004) In vivo near-infrared fluorescence imaging of integrin alphavbeta3 in brain tumor xenografts. Cancer Res 64:8009–8014 Cui K, Xu X, Zhao H, Wong ST (2008) A quantitative study of factors affecting in vivo bioluminescence imaging. Luminescence 23:292–295 Dandekar M, Tseng JR, Gambhir SS (2007) Reproducibility of 18F-FDG microPET studies in mouse tumor xenografts. J Nucl Med 48:602–607 de la Zerda A, Bodapati S, Teed R, Schipper ML, Keren S, Smith BR et al (2009) A comparison between time domain and spectral imaging systems for imaging quantum dots in small living animals. Mol Imaging Biol 11:127–133 DeGrado TR, Coleman RE, Wang S, Baldwin SW, Orr MD, Robertson CN et al (2001) Synthesis and evaluation of 18F-labeled choline as an oncologic tracer for positron emission tomography: initial findings in prostate cancer. Cancer Res 61:110–117 Denk W, Strickler JH, Webb WW (1990) Two-photon laser scanning fluorescence microscopy. Science 248:73–76 Diagaradjane P, Orenstein-Cardona JM, Colon-Casasnovas NE, Deorukhkar A, Shentu S, Kuno N et al (2008) Imaging epidermal growth factor receptor expression in vivo: pharmacokinetic and biodistribution characterization of a bioconjugated quantum dot nanoprobe. Clin Cancer Res 14: 731–741 Dinkel J, Bartling SH, Kuntz J, Grasruck M, Kopp-Schneider A, Iwasaki M et al (2008) Intrinsic gating for small-animal computed tomography: a robust ECG-less paradigm for deriving cardiac phase information and functional imaging. Circ Cardiovasc Imaging 1:235–243 Doubrovin M, Ponomarev V, Beresten T, Balatoni J, Bornmann W, Finn R et al (2001) Imaging transcriptional regulation of p53-dependent genes with positron emission tomography in vivo. Proc Natl Acad Sci U S A 98:9300–9305 Dunn KW, Sutton TA (2008) Functional studies in living animals using multiphoton microscopy. ILAR J 49:66–77 Flores JE, McFarland LM, Vanderbilt A, Ogasawara AK, Williams SP (2008) The effects of anesthetic agent and carrier gas on blood glucose and tissue uptake in mice undergoing dynamic FDG-PET imaging: sevoflurane and isoflurane compared in air and in oxygen. Mol Imaging Biol 10:192–200 Foley LM, Towner RA, Painter DM (2001) In vivo image-guided (1)H-magnetic resonance spectroscopy of the serial development of hepatocarcinogenesis in an experimental animal model. Biochim Biophys Acta 1526:230–236 Fueger BJ, Czernin J, Hildebrandt I, Tran C, Halpern BS, Stout D et al (2006) Impact of animal handling on the results of 18F-FDG PET studies in mice. J Nucl Med 47:999–1006 Gao X, Cui Y, Levenson RM, Chung LW, Nie S (2004) In vivo cancer targeting and imaging with semiconductor quantum dots. Nat Biotechnol 22:969–976

W.A. Weber and F. Kiessling Gatenby RA, Gillies RJ (2004) Why do cancers have high aerobic glycolysis? Nat Rev Cancer 4:891–899 Gatenby RA, Gillies RJ (2008) A microenvironmental model of carcinogenesis. Nat Rev Cancer 8:56–61 Gilad AA, McMahon MT, Walczak P, Winnard PT Jr, Raman V, van Laarhoven HW et al (2007) Artificial reporter gene providing MRI contrast based on proton exchange. Nat Biotechnol 25:217–219 Gilead A, Meir G, Neeman M (2004) The role of angiogenesis, vascular maturation, regression and stroma infiltration in dormancy and growth of implanted MLS ovarian carcinoma spheroids. Int J Cancer 108:524–531 Ginj M, Zhang H, Eisenwiener KP, Wild D, Schulz S, Rink H et al (2008) New pansomatostatin ligands and their chelated versions: affinity profile, agonist activity, inte rnalization, and tumor targeting. Clin Cancer Res 14: 2019–2027 Hara T, Kosaka N, Kishi H (1998) PET imaging of prostate cancer using carbon-11-choline. J Nucl Med 39:990–995 Hara T, Kosaka N, Kishi H (2002) Development of (18) F-fluoroethylcholine for cancer imaging with PET: synthesis, biochemistry, and prostate cancer imaging. J Nucl Med 43:187–199 Harada H, Kizaka-Kondoh S, Hiraoka M (2005) Optical imaging of tumor hypoxia and evaluation of efficacy of a hypoxia-targeting drug in living animals. Mol Imaging 4:182–193 Haubner R, Wester HJ, Weber WA, Mang C, Ziegler SI, Goodman SL et al (2001) Noninvasive imaging of alpha(v) beta3 integrin expression using 18F-labeled RGD-containing glycopeptide and positron emission tomography. Cancer Res 61:1781–1785 Hnatowich DJ (1999) Antisense and nuclear medicine. J Nucl Med 40:693–703 Hoffman RM (2005) The multiple uses of fluorescent proteins to visualize cancer in vivo. Nat Rev Cancer 5:796–806 Ishiwata K, Kubota K, Murakami M, Kubota R, Sasaki T, Ishii S et al (1993) Re-evaluation of amino acid PET studies: can the protein synthesis rates in brain and tumor tissues be measured in vivo? J Nucl Med 34:1936–1943 Iyer M, Wu L, Carey M, Wang Y, Smallwood A, Gambhir SS (2001) Two-step transcriptional amplification as a method for imaging reporter gene expression using weak promoters. Proc Natl Acad Sci U S A 98:14595–14600 Johnson M, Karanikolas BD, Priceman SJ, Powell R, Black ME, Wu HM et al (2009) Titration of variant HSV1-tk gene expression to determine the sensitivity of 18F-FHBG PET imaging in a prostate tumor. J Nucl Med 50:757–764 Kang JH, Chung JK (2008) Molecular-genetic imaging based on reporter gene expression. J Nucl Med 49(suppl 2): 164S–179S Keyaerts M, Verschueren J, Bos TJ, Tchouate-Gainkam LO, Peleman C, Breckpot K et al (2008) Dynamic bioluminescence imaging for quantitative tumour burden assessment using IV or IP administration of D: -luciferin: effect on intensity, time kinetics and repeatability of photon emission. Eur J Nucl Med Mol Imaging 35:999–1007 Kienast Y, von Baumgarten L, Fuhrmann M, Klinkert WE, Goldbrunner R, Herms J et al (2010) Real-time imaging reveals the single steps of brain metastasis formation. Nat Med 16:116–122

35 Imaging in Oncology Research Kiessling F, Farhan N, Lichy M, Vosseler S, Heilmann M, Krix M, Bohlen P, Miller DW, Mueller MA, Semmler W, Fusenig NE, Delorme S (2004) Dynamic contrast enhanced magnetic resonance imaging rapidly indicates vessel regression in human squamous cell carcinomas grown in nude mice caused by VEGF-receptor 2 blockade with DC101. Neoplasia 6:213–223 Kiessling F, Jugold M, Woenne EC, Brix G (2007) Non-invasive assessment of vessel morphology and function in tumors by magnetic resonance imaging. Eur Radiol 17:2136–2148 Kiessling F, Huppert J, Palmowski M (2009) Functional and molecular ultrasound imaging: concepts and contrast agents. Curr Med Chem 16:627–642 Kim KI, Chung JK, Kang JH, Lee YJ, Shin JH, Oh HJ et al (2005) Visualization of endogenous p53-mediated transcription in vivo using sodium iodide symporter. Clin Cancer Res 11:123–128 Kimura RH, Cheng Z, Gambhir SS, Cochran JR (2009) Engineered knottin peptides: a new class of agents for imaging integrin expression in living subjects. Cancer Res 69:2435–2442 Kircher MF, Allport JR, Graves EE, Love V, Josephson L, Lichtman AH et al (2003) In vivo high resolution threedimensional imaging of antigen-specific cytotoxic T-lymphocyte trafficking to tumors. Cancer Res 63: 6838–6846 Kirsch DG, Dinulescu DM, Miller JB, Grimm J, Santiago PM, Young NP et al (2007) A spatially and temporally restricted mouse model of soft tissue sarcoma. Nat Med 13:992–997 Kovar JL, Simpson MA, Schutz-Geschwender A, Olive DM (2007) A systematic approach to the development of fluorescent contrast agents for optical imaging of mouse cancer models. Anal Biochem 367:1–12 Kovar JL, Volcheck W, Sevick-Muraca E, Simpson MA, Olive DM (2009) Characterization and performance of a near-infrared 2-deoxyglucose optical imaging agent for mouse cancer models. Anal Biochem 384:254–262 Kroemer G, Pouyssegur J (2008) Tumor cell metabolism: cancer’s Achilles’ heel. Cancer Cell 13:472–482 Krohn KA, Mankoff DA, Muzi M, Link JM, Spence AM (2005) True tracers: comparing FDG with glucose and FLT with thymidine. Nucl Med Biol 32:663–671 Langen KJ, Hamacher K, Weckesser M, Floeth F, Stoffels G, Bauer D et al (2006) O-(2-[18F]fluoroethyl)-L-tyrosine: uptake mechanisms and clinical applications. Nucl Med Biol 33:287–294 Larson SM, Morris M, Gunther I, Beattie B, Humm JL, Akhurst TA et al (2004) Tumor localization of 16beta-18F-fluoro5alpha-dihydrotestosterone versus 18F-FDG in patients with progressive, metastatic prostate cancer. J Nucl Med 45: 366–373 Lederle W, Linde N, Heusel J, Bzyl J, Woenne EC, Zwick S et al (2010) Platelet-derived growth factor-B normalizes micromorphology and vessel function in vascular endothelial growth factor-A-induced squamous cell carcinomas. Am J Pathol 176:981–994 Lendvai G, Estrada S, Bergstrom M (2009) Radiolabelled oligonucleotides for imaging of gene expression with PET. Curr Med Chem 16:4445–4461 Li XF, Zanzonico P, Ling CC, O’Donoghue J (2006) Visualization of experimental lung and bone metastases in live nude mice by X-ray micro-computed tomography. Technol Cancer Res Treat 5:147–155

563 Louie AY, Huber MM, Ahrens ET, Rothbacher U, Moats R, Jacobs RE et al (2000) In vivo visualization of gene expression using magnetic resonance imaging. Nat Biotechnol 18:321–325 Macheda ML, Rogers S, Best JD (2005) Molecular and cellular regulation of glucose transporter (GLUT) proteins in cancer. J Cell Physiol 202:654–662 Mahmood U, Weissleder R (2003) Near-infrared optical imaging of proteases in cancer. Mol Cancer Ther 2:489–496 Mansi R, Wang X, Forrer F, Kneifel S, Tamma ML, Waser B et al (2009) Evaluation of a 1, 4, 7, 10-tetraazacyclododecane-1, 4, 7, 10-tetraacetic acid-conjugated bombesin-based radioantagonist for the labeling with single-photon emission computed tomography, positron emission tomography, and therapeutic radionuclides. Clin Cancer Res 15:5240–5249 McGivan JD, Pastor-Anglada M (1994) Regulatory and molecular aspects of mammalian amino acid transport. Biochem J 299(pt 2):321–334 Michalet X, Pinaud FF, Bentolila LA, Tsay JM, Doose S, Li JJ et al (2005) Quantum dots for live cells, in vivo imaging, and diagnostics. Science 307:538–544 Miles KA (2002) Functional computed tomography in oncology. Eur J Cancer 38:2079–2084 Modo M, Bulte J (2007) Cellular and molecular MR imaging. CRC, Boca Raton, pp 209–210 Montet X, Ntziachristos V, Grimm J, Weissleder R (2005) Tomographic fluorescence mapping of tumor targets. Cancer Res 65:6330–6336 Montet X, Figueiredo JL, Alencar H, Ntziachristos V, Mahmood U, Weissleder R (2007) Tomographic fluorescence imaging of tumor vascular volume in mice. Radiology 242:751–758 Moroz E, Carlin S, Dyomina K, Burke S, Thaler H, Blasberg R et al (2009) Real-time imaging of HIF-1a stabilization and degradation. Plos One 4:e5077 Mulder WJ, van der Schaft DW, Hautvast PA, Strijkers GJ, Koning GA, Storm G et al (2007) Early in vivo assessment of angiostatic therapy efficacy by molecular MRI. FASEB J 21:378–383 Nguyen QD, Smith G, Glaser M, Perumal M, Arstad E, Aboagye EO (2009) Positron emission tomography imaging of druginduced tumor apoptosis with a caspase-3/7 specific [18F]-labeled isatin sulfonamide. Proc Natl Acad Sci U S A 106:16375–16380 O’Donoghue JA, Zanzonico P, Pugachev A, Wen B, Smith-Jones P, Cai S et al (2005) Assessment of regional tumor hypoxia using 18F-fluoromisonidazole and 64Cu(II)-diacetyl-bis(N4methylthiosemicarbazone) positron emission tomography: Comparative study featuring microPET imaging, Po2 probe measurement, autoradiography, and fluorescent microscopy in the R3327-AT and FaDu rat tumor models. Int J Radiat Oncol Biol Phys 61:1493–1502 Olafsen T, Kenanova VE, Sundaresan G, Anderson AL, Crow D, Yazaki PJ et al (2005) Optimizing radiolabeled engineered anti-p185HER2 antibody fragments for in vivo imaging. Cancer Res 65:5907–5916 Ou G, Itasaka S, Zeng L, Shibuya K, Yi J, Harada H et al (2009) Usefulness of HIF-1 imaging for determining optimal timing of combining bevacizumab and radiotherapy. Int J Radiat Oncol Biol Phys 75:463–467 Palmowski M, Huppert J, Hauff P, Reinhardt M, Schreiner K, Socher MA et al (2008a) Vessel fractions in tumor xenografts

564 depicted by flow- or contrast-sensitive three-dimensional high-frequency Doppler ultrasound respond differently to antiangiogenic treatment. Cancer Res 68:7042–7049 Palmowski M, Huppert J, Ladewig G, Hauff P, Reinhardt M, Mueller MM et al (2008b) Molecular profiling of angiogenesis with targeted ultrasound imaging: early assessment of antiangiogenic therapy effects. Mol Cancer Ther 7:101–109 Palmowski M, Peschke P, Huppert J, Hauff P, Reinhardt M, Maurer M et al (2009) Molecular ultrasound imaging of early vascular response in prostate tumors irradiated with carbon ions. Neoplasia 11:856–863 Persigehl T, Bieker R, Matuszewski L, Wall A, Kessler T, Kooijman H et al (2007) Antiangiogenic tumor treatment: early noninvasive monitoring with USPIO-enhanced MR imaging in mice. Radiology 244:449–456 Plathow C, Weber WA (2008) Tumor cell metabolism imaging. J Nucl Med 49(suppl 2):43S–63S Rabinovich BA, Ye Y, Etto T, Chen JQ, Levitsky HI, Overwijk WW et al (2008) Visualizing fewer than 10 mouse T cells with an enhanced firefly luciferase in immunocompetent mouse models of cancer. Proc Natl Acad Sci U S A 105: 14342–14346 Ramirez de Molina A, Banez-Coronel M, Gutierrez R, Rodriguez-Gonzalez A, Olmeda D, Megias D et al (2004) Choline kinase activation is a critical requirement for the proliferation of primary human mammary epithelial cells and breast tumor progression. Cancer Res 64:6732–6739 Ray P, Tsien R, Gambhir SS (2007) Construction and validation of improved triple fusion reporter gene vectors for molecular imaging of living subjects. Cancer Res 67:3085–3093 Reinhardt M, Hauff P, Briel A, Uhlendorf V, Linker RA, Maurer M et al (2005) Sensitive particle acoustic quantification (SPAQ): a new ultrasound-based approach for the quantification of ultrasound contrast media in high concentrations. Invest Radiol 40:2–7 Reubi JC, Maecke HR (2008) Peptide-based probes for cancer imaging. J Nucl Med 49:1735–1738 Safran M, Kim WY, O’Connell F, Flippin L, Gunzler V, Horner JW et al (2006) Mouse model for noninvasive imaging of HIF prolyl hydroxylase activity: assessment of an oral agent that stimulates erythropoietin production. Proc Natl Acad Sci U S A 103:105–110 Serganova I, Doubrovin M, Vider J, Ponomarev V, Soghomonyan S, Beresten T et al (2004) Molecular imaging of temporal dynamics and spatial heterogeneity of hypoxia-inducible factor-1 signal transduction activity in tumors in living mice. Cancer Res 64:6101–6108 Serganova I, Humm J, Ling C, Blasberg R (2006) Tumor hypoxia imaging. Clin Cancer Res 12:5260–5264 Shields AF, Grierson JR, Dohmen BM, Machulla HJ, Stayanoff JC, Lawhorn-Crews JM et al (1998) Imaging pro-

W.A. Weber and F. Kiessling liferation in vivo with [F-18]FLT and positron emission tomography. Nat Med 4:1334–1336 Sundararajan L, Linden HM, Link JM, Krohn KA, Mankoff DA (2007) 18F-Fluoroestradiol. Semin Nucl Med 37:470–476 Tjuvajev JG, Finn R, Watanabe K, Joshi R, Oku T, Kennedy J et al (1996) Noninvasive imaging of herpes virus thymidine kinase gene transfer and expression: a potential method for monitoring clinical gene therapy. Cancer Res 56:4087–4095 Tropres I, Grimault S, Vaeth A, Grillon E, Julien C, Payen JF et al (2001) Vessel size imaging. Magn Reson Med 45: 397–408 Ullrich RT, Zander T, Neumaier B, Koker M, Shimamura T, Waerzeggers Y et al (2008) Early detection of erlotinib treatment response in NSCLC by 3¢-deoxy-3¢-[F]-fluoro-Lthymidine ([F]FLT) positron emission tomography (PET). PLoS One 3:e3908 Van de Sande T, De Schrijver E, Heyns W, Verhoeven G, Swinnen JV (2002) Role of the phosphatidylinositol 3¢-kinase/PTEN/Akt kinase pathway in the overexpression of fatty acid synthase in LNCaP prostate cancer cells. Cancer Res 62:642–646 Vander Heiden MG, Cantley LC, Thompson CB (2009) Understanding the Warburg effect: the metabolic requirements of cell proliferation. Science 324:1029–1033 Verrey F (2003) System L: Heteromeric exchangers of large, neutral amino acids involved in directional transport. Pflugers Archiv 445:529–533 Wei K, Jayaweera AR, Firoozan S, Linka A, Skyba DM, Kaul S (1998) Quantification of myocardial blood flow with ultrasound-induced destruction of microbubbles administered as a constant venous infusion. Circulation 97:473–483 Wei LH, Su H, Hildebrandt IJ, Phelps ME, Czernin J, Weber WA (2008) Changes in tumor metabolism as readout for mammalian target of rapamycin kinase inhibition by rapamycin in glioblastoma. Clin Cancer Res 14:3416–3426 Woo SK, Lee TS, Kim KM, Kim JY, Jung JH, Kang JH et al (2008) Anesthesia condition for (18)F-FDG imaging of lung metastasis tumors using small animal PET. Nucl Med Biol 35:143–150 Wu AM (2009) Antibodies and antimatter: the resurgence of immuno-PET. J Nucl Med 50:2–5 Wu AM, Senter PD (2005) Arming antibodies: prospects and challenges for immunoconjugates. Nat Biotechnol 23:1137–1146 Zannetti A, Del Vecchio S, Iommelli F, Del Gatto A, De Luca S, Zaccaro L et al (2009) Imaging of alpha(v)beta(3) expression by a bifunctional chimeric RGD peptide not cross-reacting with alpha(v)beta(5). Clin Cancer Res 15:5224–5233 Zwick S, Strecker R, Kiselev V, Gall P, Huppert J, Palmowski M et al (2009) Assessment of vascular remodeling under antiangiogenic therapy using DCE-MRI and vessel size imaging. J Magn Reson Imaging 29:1125–1133

Imaging in Immunology Research

36

Jason T. Lee, Evan D. Nair-Gill, Brian A. Rabinovich, Caius G. Radu, and Owen N. Witte

Contents 36.1 The Immune System: Cell Types and Functions................................................................. 565 36.2 Whole Body Imaging to Monitor T-Cell Trafficking and Function....................................... 567 36.3 Imaging Cell Trafficking by Direct Labeling of Immune Cells...................................... 567

36.4 Reporter Gene Imaging for Visualizing the Immune System................................................ 567 36.4.1 Methods to Introduce Reporter Genes into Cells . ..................................................... 568 36.4.2 Bioluminescent Reporter-Based Imaging (BLI) of the Immune System.................................... 569 36.4.3 PET Reporter Gene Imaging ................................... 570 36.4.4 Functional Imaging of the Immune System ............ 574 36.5 Imaging Limitations and Improvements.............. 579 36.5.1 Instrumentation......................................................... 579 36.5.2 PET Data Analysis.................................................... 579 36.6 Concluding Remarks.............................................. 580

J.T. Lee and C.G. Radu Department of Molecular and Medical Pharmacology, David Geffen School of Medicine at UCLA, Los Angeles, CA, USA and Crump Institute for Molecular Imaging, Los Angeles, CA, USA E.D. Nair-Gill Department of Molecular and Medical Pharmacology, David Geffen School of Medicine at UCLA, Los Angeles, CA, USA B.A. Rabinovich Department of Experimental Diagnostic Imaging, University of Texas, MD Anderson Cancer Center, Houston, TX, USA O.N. Witte (*) Department of Molecular and Medical Pharmacology, David Geffen School of Medicine at UCLA, Los Angeles, CA, USA; Department of Microbiology, Immunology and Molecular Genetics, David Geffen School of Medicine at UCLA, Los Angeles, CA, USA and Howard Hughes Medical Institute, University of California, Los Angeles, CA 90095-1662, USA e-mail: [email protected]

References............................................................................ 580

36.1 The Immune System: Cell Types and Functions The vertebrate immune system is composed of innate and adaptive components that confer general and specific protection against a wide range of pathogens, respectively. This introduction will summarize the major innate and adaptive cell types and their functions during an immune response. A more comprehensive description of the development of various immune cell types and their effector functions can be found in Murphy et al. (2008).

F. Kiessling and B.J. Pichler (eds.), Small Animal Imaging, DOI: 10.1007/978-3-642-12945-2_36, © Springer-Verlag Berlin Heidelberg 2011

565

566

A successful immune response requires the coordinated functional differentiation and migration of innate and adaptive immune cells (leukocytes) within different tissue environments in the body. Innate leukocytes such as macrophages, neutrophils, and natural killer (NK) cells respond quickly to the presence of pathogens by infiltrating infected tissues and mounting a nonspecific response. Macrophages and neutrophils phagocytose microbes and infected cells and produce proinflammatory cytokines including IFN-gamma and TNF-alpha. In general, infected cells upregulate expression of stress-induced major histocompatibility complex class I (MHC-I)-like molecules (e.g., ligands for the activation receptor, NKG2D), but display lower levels of “self” classic MHC-I (ligands for inhibitory receptors) (Vivier et al. 2002; Raulet 2003; Biron 1997). NK cells express an array of activation and inhibitory receptors that react to changes in the expression of MHC class I and its homologues on infected cells and either directly lyse these cells and/or produce inflammatory cytokines. Thus, the innate response involves the direct clearance of infected cells and the production of cytokines that prime subsequent adaptive immunity. The latter is mediated mainly by T cells and B cells which migrate to infected tissues 48–72 h after the innate response and priming by antigen presenting cells (APCs). Macrophages and dendritic cells are APCs required for priming of adaptive immunity. APCs phagocytose pathogen-derived proteins in peripheral tissues and then migrate to local lymph nodes. Here, APCs present antigenic peptide fragments in a MHC-restricted fashion to CD4+ and CD8+ T-cell clones which express T-cell receptors (TCRs) specific for the antigenic peptide. Binding of a TCR to its cognate antigenic peptide/ MHC complex, together with appropriate costimulation, results in T-cell activation and massive proliferation (clonal expansion) within secondary lymphoid structures such as spleen and lymph nodes. CD8+ cytotoxic T lymphocytes (CTLs) are responsible for killing intracellular pathogens by lysing infected cells. Following activation and clonal expansion, CTLs egress from secondary lymphoid organs (e.g., spleen and lymph nodes) and infiltrate peripheral tissues seeking target cells that express cognate antigens. After elimination of the pathogen, most CTLs generated during the response are trapped in the liver and die through apoptosis. Such immunoregulation is critical for the maintenance of T-cell homeostasis and

J.T. Lee et al.

the prevention of autoimmunity. From the population of CTL, a subset of memory cells persists in a quiescent, nonproliferative state. Subsequent invasion of the same pathogen results in activation and expansion of these memory cells with faster kinetics than during the primary response. The memory response results in rapid elimination of the pathogen, often before any sign of infection can be detected (Murali-Krishna et al. 1998). CD4+ T cells, or helper T cells (Th), have many functions during an immune response. Through direct cell– cell contacts and the production of proinflammatory cytokines such as IL-2, IFN-gamma, and TNF-alpha, they augment the function of innate immune cells in clearing pathogens and increase the cytotoxicity of CTL. Th cell production of IL-4 stimulates antigen-specific B cells to produce antibodies. Th cells clonally expand earlier than CTL during an immune response and the absolute number of antigen-specific Th cells is typically less than CTLs (Schepers et al. 2002). As with CTL, a subset of memory Th cells survive that confer long-term protection against the initiating pathogen. In addition to the immune system’s primary function of protecting the host from infection, innate and adaptive leukocytes have been implicated in the initiation, progression, and maintenance of tumors as well as their elimination. The mechanisms for both scenarios are complex and not fully elucidated. Infiltration of tumors by eosinophils, macrophages, and regulatory T cells under chronic inflammatory conditions establishes a wound-like immunosuppressive microenvironment that supports tumor survival and growth (Mantovani et al. 2008). Endogenous antitumor immunity is less well documented, but there is evidence NK cells and T cells may play a role in tumor surveillance, particularly against virally transformed cells. In patient biopsy samples, the presence of NK cells and T cells in solid tumors are associated with improved outcome (Galon et al. 2006). Therapeutic interventions including tumor antigen-based vaccines and adoptive tumor-specific T cell immunotherapy have shown promise as immunebased antitumor strategies. With the exception of the eyes and the gonads, the immune system surveys every tissue in the body. It is by definition a highly dynamic network of multiple cell types that migrate within and between tissues. These attributes make the immune system well suited for live animal imaging. To date, most immunological studies using live animal imaging have focused on T cell

567

36 Imaging in Immunology Research

localization because T cells are easy to isolate and manipulate ex vivo, expand to large numbers, and are a central player in most immune responses. Defects in T-cell function at the level of activation, clonal expansion, effector differentiation, or memory formation are implicated in autoimmune disorders, cancer, and chronic infections. Thus, it is important to measure T-cell function at multiple sites within the body to understand how their activity can be modulated in disease. This chapter will focus on whole body imaging techniques that have allowed experimental interrogation of T cell-mediated immune responses.

36.2 Whole Body Imaging to Monitor T-Cell Trafficking and Function Classic methods for studying T cell-mediated immunity include serial measurements of T cell numbers and cytokine production in blood samples and various tissues using flow cytometric and immunohistochemistry (reviewed in Clay et al. (2001)).While providing valuable information about the activity of specific cell populations, these measurements do not illuminate the spatial and temporal kinetics that dictate the outcome of immune responses in vivo and are hindered by animal to animal variation. Whole body imaging techniques enable measurements of the complex dynamics of the immune network. Multiple imaging modalities such as magnetic resonance imaging (MRI), bioluminescence imaging (BLI), positron emission tomography (PET), and single photon emission tomography (SPECT) have been adapted for small animal imaging. These modalities are further explained in this book. Here, we will focus on the application of BLI, PET, and SPECT to visualize T cell trafficking and function.

vivo via incubation with a radiolabeled probe, and infused into recipient animals which can then be serially imaged. This technique was used to visualize trafficking of monocytes, dendritic cells, and CD8+ T cells. In one characteristic study, adoptively transferred splenocytes were monitored for up to 20 h using PET after ex vivo labeling with the lipophilic redox-active carrier molecule pyruvaldehyde-bis(N4-methylthiosemicarbazone labeled with 64Copper (64Cu-PSTM) (Adonai et al. 2002). Cells infused via the tail vein into recipient animals migrated from the lungs to the spleen in less than 4 h postinfusion. More recently, Pittet et al. (2007) used SPECT to visualize the infiltration of transgenic CTLs that express a TCR specific for the influenza hemagglutinin peptide (HA512–520 in the context of the H2KdMHC molecule) into HA+ tumors after adoptive transfer. Tumor infiltrating T cells could be visualized as early as 2 h postadoptive transfer. SPECT signals localized to the central region of the HA+ tumor. In contrast, a diffuse signal was observed in the margins of tumors that did not express the HA antigen. Imaging of T cells through direct labeling techniques can be useful in evaluating parameters that influence T cell trafficking in vivo. However, there are specific limitations of this technique for studying immune responses. Loading cells with the radiolabeled probe may have toxic effects on those cells (Balaban et al. 1987). In addition, death of labeled cells in vivo, followed by the release of probe into the extracellular space, may result in nonspecific labeling of other cell types. Importantly, the radiolabeled probe will be diluted with replication of the labeled populations leading to a loss of signal over time. This is particularly concerning when performing longitudinal imaging studies of rapidly dividing cell populations such as activated T lymphocytes (Balaban et al. 1986, 1987; Botti et al. 1997), which, when antigen-stimulated, have been shown to divide every 5 h (Hwang et al. 2006).

36.3 Imaging Cell Trafficking by Direct Labeling of Immune Cells

36.4 Reporter Gene Imaging for Visualizing the Immune System Early studies of in vivo trafficking of various immune cell types used direct labeling of cells ex vivo prior to in vivo nuclear imaging with PET or SPECT (Matsui et al. 2004; Paik et al. 2002; Ridolfi et al. 2004). Cells of interest are isolated from donor animals, labeled ex

One approach to circumvent the limitations of direct labeling is to introduce genetically encoded reporters into T cell populations of interest. Reporter genes

568

allow labeled cells to be detected in vivo either via intrinsic fluorescence (e.g., red fluorescent proteins) or following administration of reporter-specific small molecule probes that are radiolabeled (for PET and SPECT imaging) or catabolized by enzymes that lead to the emission of photons in the form of bioluminescence. This section will describe different methods to introduce reporter genes into T cells and then present key studies that use BLI and PET reporters to visualize the immune system in different disease contexts.

36.4.1 Methods to Introduce Reporter Genes into Cells Efficient delivery of reporter genes to target immune cells is critical to the success of a reporter system. This section will describe three such methods: (i) in vitro transduction (viral gene transfer) of target cells of interest, (ii) in vivo delivery of reporter genes, and (iii) germline transmission of reporter constructs.

36.4.1.1 In Vitro Transduction of Target Cells with Reporter Genes Reporter gene labeling using in vitro methods involves isolation of cells from the host animal followed by viral transduction of a reporter construct. Although viral transduction of target cells generally provides long-term stable protein expression, the use of these systems requires certain considerations. Adenoviruses are highly efficient at mediating gene transfer, but do not integrate their genetic cargo into the cellular genome. This can result in relatively transient transgene expression. Because adenovirus-associated proteins must be produced for expression of the transgene, there is the risk of making transduced cells highly immunogenic and susceptible to rejection (Muruve 2004). Newer adenovirus-based vectors may circumvent these limitations by exploiting the Epstein–Barr virus (EBV) episome (Gallaher et al. 2009). These hybrid adenovirus-EBV vectors facilitate long-term expression of a reporter gene in vivo. Lentiviral or retroviral constructs have been designed to only express the transgene of interest following genomic integration, thus reducing immunogenicity of

J.T. Lee et al.

transduced cells. The integration of lentiviral and retroviral DNA can occur almost anywhere in the genome, potentially disrupting expression of important genes. In human trials, cases of induced leukemia have been noted following retroviral delivery of adenosine deaminase to stem cells in immunodeficient patients (Kohn et al. 2003). Reporter gene constructs can be engineered to simultaneously express multiple reporters. A fluorescent protein (Shaner et al. 2005; Snapp 2009) or heterologous cell surface marker enables detection and purification of transduced cells using fluorescenceactivated cell sorting (FACS). Inclusion of secreted proteins such as alkaline phosphatase can measure the persistence of transferred cells with simple detection assays of blood or urine samples (Yang et al. 1997). Certain immune-mediated disease models, such as graft-versus-host disease (GVHD), are sensitive to in vitro handling of T cells (Beilhack et al. 2005). One method to circumvent this obstacle is to transduce hematopoietic stem cells with the reporter construct. Following a bone marrow transplant into irradiated recipient animals, the immune system will be reconstituted such that the reporter is expressed in all immune cell types. These cells can then be directly transplanted into secondary recipients to achieve the desired disease model without extensive in vitro handling steps.

36.4.1.2 In Vivo Targeted Delivery of Reporter Genes In vivo reporter gene delivery approaches can target cells or tissues of interest using viruses or bound to antibodies or ligands that bind cell-specific molecules, thus avoiding ex vivo handling of cells. To examine somatic cell labeling in mice, Liang et al. (2002a, b) used adenoviral-based vectors to deliver the Herpes simplex virus 1-sr39tk (HSV1-sr39tk) reporter gene to the liver which is the tropism of adenoviruses. They repetitively and quantitatively monitored gene expression over a three-month period with the radiolabeled substrate 9-(4-18F-Fluoro-3-[hydroxymethyl]butyl)guanine (18F-FHBG). For tumor targeting, modifications have centered on generating adenoviruses expressing modified coat proteins specific for tumor-associated molecules and increasing adenovirus resident time in circulation, avoiding immune destruction, and manipulating vector tropism. The latter avoids unwanted

569

36 Imaging in Immunology Research

expression of reporter genes in the liver. Xiong et al. (2006) used adenoviral particles chemically coupled to the cyclic RGD peptide motif using polyethylene glycol to target aVb3-expressing tumors resulting in lower nonspecific 18F-FHBG signal in the liver. Techniques such as these can be adopted for T cell targeting and lineage-specific studies. A lentiviral vector engineered to specifically bind to the dendritic cell surface protein DC-SIGN allowed targeted transduction and maturation of these cells and induction of antibody response in vivo (Yang et al. 2008). Targeting specific cell types in vivo with viral vectors in immuno competent animals could induce an antiviral host immune response. Host immune surveillance may diminish the delivery of reporter genes to target cells and reduce imaging sensitivity. Less immunogenic targeting techniques including the use of cDNA conjugated to nanoparticles, liposomes, and membrane-penetrating peptides are currently being developed (reviewed in Rettig and Rice (2007)).

36.4.1.3 Generating Reporter Transgenic Animals Creation of transgenic animals expressing reporter genes (reporter transgenics) can be accomplished with random genomic integration using conventional transgenic approaches, virus-mediated transduction of stem cells (Pfeifer et al. 2001), or bacterial artificial chromosome (BAC) techniques (Lee et al. 2003). Nonrandom gene integration takes advantage of homologous recombination for controlled site-specific integration into the genome. Lentiviral transgenics allow for rapid evaluation of reporter expression and enables promoter-specific reporter expression (Lois et al. 2002). Reporter transgenics allow for the continuous expression of the gene throughout an animal’s lifetime as well as in subsequent generations with the advantage of monitoring cell localization in various stages of immunity. Reporter transgenics have been used to monitor interleukin-4 (IL-4) producing T-cell trafficking patterns at various stages of immune responses (Mohrs et al. 2001). Transgenic mice with firefly luciferase expression driven by response elements for immune cytokines such as interleukin-2 (IL-2) are now in commercial production (Caliper CD1-Tg(Il2-luc)Xen). Similar approaches may be applied to other types of reporter genes.

Efficient retroviral and lentiviral transductions of leukocytes require in vitro mitogenic stimulation, which can limit subsequent in vivo survival and potentially alter normal function and hinder imaging of activation-related events. A major advantage to reporter transgenics is the expression of reporter genes in naïve leukocytes. Expression of an imaging reporter during embryonic development, however, may alter the function of the developing leukocytes. The use of conditional (tissue-specific) and inducible transgenic technologies (Branda and Dymecki 2004; Zhu et al. 2002) should alleviate much of these concerns by allowing investigators to initiate reporter gene expression at any time.

36.4.2 Bioluminescent Reporter-Based Imaging (BLI) of the Immune System BLI depends on the expression of a luciferase enzyme that can catalyze light-emitting reactions by oxidizing an exogenously supplied photoreactive substrate. Photon emissions from these oxidation reactions are detected with cooled charge-coupled device (CCD) cameras. An important consideration for BLI is the wavelength of light generated by the catalytic reaction. Wavelengths in the red and near infrared range are capable of penetrating through tissue at greater efficiency than shorter wavelengths. For mouse models involving deep tissue imaging, it is important that the emission maximum of the luciferase used be greater than 600 nm. Target cells are engineered to express the luciferase, injected into an organism, and tracked noninvasively with CCD imaging. Most applications are performed using firefly (Photinus pyralis) or Renilla (Renilla reniformis) luciferases (FLuc and RLuc, respectively). These enzymes catabolze d-luciferin and coelenterazine, respectively, and have been used for whole body studies of immune cell trafficking (Yaghoubi et al. 2007; Costa et al. 2001). Gaussia princeps luciferase (GLuc), a coelenterazine metabolizing enzyme which is normally secreted, has been humanized and modified such that it is sequestered in the endoplasmic reticulum (ER). This construct provides approximately a 200-fold increase in sensitivity compared to RLuc in vivo (Tannous et al. 2005).

570

Recently, a red shifted mutant of RLuc (RLuc8.6–535) has been generated. In our hands, this variant is superior to GLuc and is not hindered by the flash kinetics that is observed when imaging with GLuc (Rabinovich et al., unpublished data). Because RLuc and GLuc use different substrates than FLuc, it is possible to perform dual-reporter imaging with a combination of these reporters. Optical imaging with BLI has been used to visualize localization of autoreactive T cells to target tissues during autoimmune responses such as rheumatoid arthritis (Yaghoubi et al. 2007) and experimental autoimmune encephalomyelitis (EAE) (Costa et al. 2001). BLI has also been useful in defining key parameters required for effective hematopoietic reconstitution (Abdel-Azim et al. 2008; Wang et al. 2003) and is the standard imaging technology used to evaluate cellular factors that control the onset, manifestation, and progression of GVHD ((Beilhack et al. 2005; Nishimura et al. 2008), reviewed in Negrin and Contag (2006)). Bioluminescent reporter gene constructs have undergone several improvements in order to enhance gene expression in immune cells and augment signal intensity. Santos et al. (2009) generated a membrane-bound version of GLuc (memGLuc) by coupling the CD8 transmembrane domain to GLuc’s carboxy terminus. MemGluc could be transduced with high efficiency and, unlike native FLuc, high expression levels into activated primary T cells. Moreover, although RLuc is readily expressed in primary T cells, memGLuc generated a signal that was 30–100 times stronger, similar to that observed with ER-bound GLuc and slightly lower than that observed with RLuc8.6–535 (Rabinovich et al., unpublished data). Until recently, BLI was only able to detect a lower limit of tens of thousands of T cells. In the clinical setting, however, T-cell infiltration into tumors has been measured to be less than 0.005% per gram of tumor. Therefore, it is imperative to develop technology with much lower limits of detection. Rabinovich et al. (2008) modified the FLuc DNA sequence by codon optimization, removal of cryptic splice sites, and by changing the retroviral delivery system to develop an enhanced FLuc (effLuc) vector (Fig. 36.1a). EffLuc has a lightgenerating capacity that is more than 100 times greater than FLuc. Furthermore, as few as three cells implanted subcutaneously and less than 3 × 104 adoptively transferred T cells in mice were detected (Fig. 36.1b). OT-1 TCR transgenic T cells (specific for the chicken

J.T. Lee et al.

ovalbumin peptide, SIINFEKL, in the context of H2Kb) were transduced with effLuc and injected intravenously into tumor-bearing mice. BLI of mice with established Ova− and Ova+ tumors indicated T cell localization and infiltration of tumor in an antigen-specific manner (Fig. 36.1c). Importantly, by adjusting the number of adoptively transferred effLuc+ OT-1 T cells, it was possible to achieve similar infiltration of tumor-specific T cells (i.e., cells per gram tumor) in this mouse model as is seen in the clinic. BLI is well suited for experimental imaging of T cell-mediated immune responses in small animals. T cells can be engineered quickly, the system is inexpensive and highly sensitive, and the imaging instrumentation is easy to use. However, light produced by FLuc, RLuc, and GLuc is highly attenuated by tissue depth. Currently, FLuc and RLuc8.6–535 are the most suitable luciferin and coelenterazine metabolizing enzymes, respectively, because a portion of their emission spectra fall beyond 600 nm (FLuc = 570–680 nm, RLuc8.6– 535 = 500–620 nm). Tissue-associated signal attenuation prohibits imaging of deep-seated lesions and underscores the need for bioluminescent reporters that emit in the near infrared range. Other limitations of BLI in imaging the immune system include its poor spatial resolution and the lack of methods for robust signal quantification. Most commonly used BLI systems are not tomographic, making it difficult to accurately assess the anatomical location of signal. Newer optical imaging machines, however, may eliminate some of these obstacles (Zhang et al. 2008; Ntziachristos et al. 2005).

36.4.3 PET Reporter Gene Imaging A PET reporter gene encodes a protein in the target cell that facilitates the specific accumulation of a probe carrying a positron-emitting isotope. PET reporters can be divided into three general types: (i) enzyme-based reporters; (ii) transport-mediated reporters, and (iii) receptorbased reporters (Fig. 36.2). Enzymatic reporters biochemically modify radiolabeled probes and trap them within cells. The most common enzyme reporters are kinases that, through phosphorylation, add an electronegative phosphate group to the probe, preventing its efflux across the plasma membrane. Phosphorylation-mediated trapping can result in high-levels of probe accumulation and allow signal amplification.

571

36 Imaging in Immunology Research Fig. 36.1 Luciferase optimization for enhanced light generation. (a) Modifications to the firefly luciferase-encoding retroviral vector pMSCVffLuc-pIRES2-Thy1.1 (v-ffLuc, top construct) include codon optimization, removal of cryptic splice sites, and retroviral modification to produce the enhanced ffLuc (effLuc, bottom construct). (b) BLI comparison of v-ffLuc and v-effLuc and lower limit of detection of effLuc in vivo. (c) BLI of cell transfer-based immunotherapy model for the rejection of preexisting EL4 Ova− (left flank) and EG.7 Ova+ (right flank) tumors. Images were acquired 5 days after adoptive cell transfer of 3 × 104 OT-1 T cells (adapted from Rabinovich et al. 2008)

a

5’ LTR

Psi

ffLuc (pGL3)

EMCV IRES Thy1.1

5’ LTR

Psi

effLuc

EMCV IRES Thy1.1 WHV Pre 3’ LTR

3’ LTR

b

v-ffLuc

v-effLuc

c

In the transporter-based systems, translocation of a probe across the cell membrane is mediated by membrane transport proteins. A large and diverse pool of potential reporter genes exists, including those in the solute carrier family of membrane transport proteins (Hediger et al. 2004). For receptor binding, cells are labeled in a stoichiometric manner dependent on the surface expression of the receptor. Drawbacks of these

t echniques include variability inherent in metabolic activity of enzymes, lack of intracellular trapping mechanisms from transported probes, and lack of signal amplification in stoichiometric receptor binding. Antibody reporters have also been studied as potential reporter genes (Wei et al. 2008). Here, antibody fragments are expressed in cells of interest and targeted by radiolabeled hapten probes (Fig. 36.2).

572

J.T. Lee et al. 18

F-FHBG 18

F-FESP

P HSV1-tk P transporter Bo

nz

yl D

D2R nucleus

Y-AABD

OT 86 A O NH

Na+ I

124

coupled transporter

substitution at position 168, showed purine, but not pyrimidine, phosphorylation activity (Balzarini et al. 2006). Used in tandem, these mutants potentially allow for differential imaging of various activities or cell types in the same subject.

Na+ 124 I

DAbR1

Fig. 36.2 PET reporter gene/reporter probe imaging. Clockwise from top left: enzyme-based system – probes are trapped within cells by a biochemical modification such as phosphorylation (i.e., 18F-FHBG and Herpes Simplex Virus 1 – thymidine kinase). Receptor-based system – the probe binds to cell surface receptors in a stoichiometric manner (i.e., 18F-fluoro-ethyl-spiperone (18F-FESP) and the dopamine-2 receptor). Transporter-mediated system – probe accumulates in cells through its facilitated movement across the cell membrane (i.e., iodine-124 (124I) and the human sodium iodide symporter). Antibody-based system – the hapten probe binds to cell surface-anchored antibody fragment (i.e., 86yttrium-(S)-2-(4-acrylamidobenzyl)-DOTA (86Y-AABD) and the DOTA antibody reporter 1 (DAbR1))

36.4.3.1 The Herpes Simplex Virus Thymidine Kinase PET Reporter Gene System Reporter imaging studies of T cell-mediated immunity have largely been based on the viral protein Herpes Simplex Virus 1 thymidine kinase (HSV1-tk). Unlike human TK1, which only phosphorylates thymidine, HSV1-tk phosphorylates both acycloguanosine analogs and thymidine. To generate a viral TK reporter with greater specificity for acycloguanosine analogs, the HSV1-tk gene was subjected to semirandom mutagenesis. The resulting HSV1-sr39tk (Black et al. 1996; Gambhir et al. 2000) is characterized by a twofold increase in uptake of 18F-pencyclovir compared to the original HSV1-tk. Likar et al. (2009) recently developed additional mutants containing an arginine-to-glutamine substitution at position 176 (HSV1-R176Qsr39tk) and an alanine-to-tyrosine substitution at position 167 in combination with five amino acid substitutions (HSV1-A167Ysr39tk). The former was shown to be more specific for the pyrimidine-based reporter probe 18F-FEAU, while the latter is more specific for the acycloguanosine-based 18 F-FHBG. Variant HSV1-tk-A168H with amino acid

36.4.3.2 Visualizing Immune Responses with HSV-1-tk-Based Reporters PET reporter imaging using HSV1-tk or HSV1-sr39tk has been used to study T cell-mediated antitumor responses. Koehne et al. (2003) transduced CTLs specific for EBV with the HSV1-tk gene. The pyrimidine analog reporter probe 2¢-fluoro-2¢-deoxy-1b-darabinofuranosyl-5-124I-iodouracil (124I-FIAU) was used to monitor T cell trafficking in mice bearing MHCmatched EBV+ xenograft tumors. 124I-FIAU accu mulation was detectable in MHC-I-matched EBV+ tumors, indicating that this system is a viable technique for visualizing T cells using PET. Dubey et al. (2003) measured the dynamics of a memory T cell response against an established autochthonous tumor. Immunocompetent mice injected with the Moloney murine sarcoma and leukemia virus (MSV/MuLV) complex develop rhabdomyosarcomas that are rejected by CTL specific for viral gag and env proteins (Milan et al. 1999). Secondary transfer of HSV1-sr39tk-transduced T cells from these mice into immunodeficient sarcoma-bearing mice allowed 18 F-FHBG PET imaging of sarcoma-specific T cells at tumor sites. An increase in the 18F-FHBG accumulation over time was detected by sequential PET imaging. In contrast, 18F-FHBG accumulation was low in control mice receiving HSV1-sr39tk-expressing T cells from immunologically naïve donors who were not exposed to the MSV/MuLV complex. Su et al. (2006) used PET reporter imaging to compare the trafficking patterns of primary and secondary T cell antitumor responses. Naïve and memory OT-1 TCR transgenic T cells were isolated, activated ex vivo, and transduced with the HSV1sr39tk reporter gene. These T cells were then adoptively transferred into immunodeficient mice bearing ovalbumin-expressing tumors. Longitudinal PET imaging indicated a greater infiltrate of memory vs. primary activated T cells at the tumor site and secondary lymphoid organs. This result correlated with antitumor activity.

573

36 Imaging in Immunology Research

Using multimodality imaging, Shu et al. (2005) studied primary immune responses to tumor challenge in a bone marrow chimeric mouse model using a triple fusion reporter gene system. Hematopoietic stem cells transduced with HSV1-sr39tk coupled to green fluorescent protein and RLuc (Ray et al. 2004) were engrafted into lethally irradiated recipient C57BL/6 mice. BLI was used to follow reconstitution of the immune system. Signals were obtained in the thymus, spleen, and long bones, corroborating flow cytometric evidence for reconstitution of T cells, B cells, and myeloid cells. Mice were challenged with MSV/MuLV and developed highly immunogenic rhabdomyosarcomas that were rejected by cytotoxic CD8+ T cells with aid from helper CD4+ T cells. Longitudinal PET reporter imaging with 18F-FHBG indicated localization of T cells to the tumor and draining lymph nodes (DLNs) with peak signal activity occurring on day 10 postchallenge. Furthermore, immunosuppression by dexamethasone abolished the 18F-FHBG signals in the DLNs, suggesting that this approach can be used to monitor immunomodulatory therapies. A significant challenge in cancer immunotherapy is breaking tolerance to self-antigens expressed by tumor cells, so that an effective cell-based antitumor immune response may occur. Unless a tumor-associated antigen (TAA) is also a neoantigen or an autoimmune response is initiated, immune-reactivity to “self” TAA is extremely low. One strategy is to artificially break tolerance by adoptively transferring

preactivated genetically engineered TAA-specific T cells into cancer patients. Such a strategy has shown some promise for treatment of patients with metastatic melanoma (Johnson et al. 2009). Shu et al. (2009) used PET reporter imaging to longitudinally monitor injected T cells in a mouse model of adoptive T cell immunotherapy developed by Overwijk et al. (2003) for the rejection of B16 melanoma cells expressing the self TAA gp10025–33 (Fig. 36.3). CD8+ T cells specific for gp10025–33 were isolated from donor TCR transgenic mice and were transduced with the PET reporter gene HSV1-sr39tk. gp10025–33-specific T cells were intravenously injected into sublethally irradiated C57BL/6 mice bearing preestablished B16 tumors (gp10025–33+) in combination with high-dose systemic IL-2 (HD IL-2; required to promote T cell survival and cytotoxicity) with or without gp10025–33 peptide-pulsed dendritic cells (DCs) to induce T cells expansion (Overwijk et al. 2003). T cell trafficking and localization were determined by longitudinal 18 F-FHBG PET/CT imaging (Fig. 36.4). 18F-FHBG signals at the sites of tumor and lymph nodes were detected by PET imaging as early as day 3 postadoptive transfer and over the course of 3 weeks with significant variability between individual mice (Fig. 36.5). DC vaccination and HD IL-2 treatment increased the overall 18F-FHBG signals from HSV1sr39tk+ T cells in multiple lymph nodes as well as in the spleen. Ex vivo flow cytometric analysis using a yellow fluorescent coreporter (seYFP (Ntziachristos Recipient

Fig. 36.3 Schematic of an advanced murine adoptive cell transfer model for treatment of human diseases such as cancer. T cells are harvested from donor mice and expanded in vitro. Prior to adoptive cell transfer, mice undergo lymphodepletion (XRT, 500 cGy). Expanded donor T cells are adoptively transferred into recipient mice along with a regimen of dendritic cell vaccination and high-dose systemic interleukin-2. Post-ACT monitoring of adoptive T cell trafficking can be facilitated using BLI or PET imaging. DC dendritic cell; IL-2 interleukin-2

Donor

Melanoma

Lymphodepletion/ myeloablation

• DC vaccination • IL-2 administration

1. Antigenspecific T lymphocytes are harvested from donor mouse

2. Culture of tumor-specific T cells.

106 cells

574

J.T. Lee et al.

a

LTR HSV1-sr39tk-IRES-eYFP LTR

Pmel-1TCR transgenic mouse

Transduce lymphocytes with retrovius

Adoptive transfer +/-DC vacc. HD IL-2 Sublethally-irradiated B16 melanoma bearing mouse

b

d

Time (days) -8

µPET/CT imaging with 18 F-FHBG

-1

0 1

3

5

7

8 10 12 14 15 17 19 21

Event Day 0: AT DC + IL−2

Day 7: DC + IL−2

Day 14: DC + IL−2

B16 melanoma challenge Sublethal irradiation F-FHBG PET

18

c

Fig. 36.4 Longitudinal study of T cell persistence using 18 F-FHBG imaging of combined immunotherapy against melanoma. (a) Schematic for combined immunotherapy study with whole body irradiation (500 cGy), DC vaccination, high-dose IL-2, and adoptive T cell transfer. (b) Time course of treatment and imaging. (c) Tumor response to full immunotherapeutic treatment compared to mice receiving T cells only. (d) 18F-FHBG

microPET imaging on day 3 postadoptive transfer (post-AT) comparing mice with combined immunotherapy, no AT and lymphodepletion only. T tumor; A axillary LN; B brachial LN; SP spleen; GI gastrointestinal tract; BL bladder; %ID/g percent injected dose of PET probe per gram of tissue (adapted from Shu et al. 2009)

et al. 2005)) showed a good correlation between 18 F-FHBG signals and numbers of gp10025–33 specific TCR transgenic T cells present in the lymph nodes and spleen. This study demonstrated the feasibility of using PET reporter imaging for quantitative temporal evaluation of homing, expansion, and persistence of adoptively transferred T cells to tumor sites. However, HSV1-sr39tk-independent 18F-FHBG signals were observed in large tumors from mice only receiving lymphodepletion (i.e., no T cells), suggesting that endogenous TK was trapping 18F-FHBG. Newer HSV1-tk substrates, such as 18F-2¢-fluoro2¢deoxy-1b-d-arabinofuranosyl-5-ethyl-uracil (18F-FEAU) (Yaghoubi et al. 2007; Tseng et al. 2006; Serganova et al. 2004; Ponomarev et al. 2007; Buursma et al. 2006; Alauddin et al. 2007), may have lower

background than 18F-FHBG and should be evaluated in this model.

36.4.4 Functional Imaging of the Immune System Reporter imaging studies described in the previous sections monitor T cell trafficking. They do not directly describe cellular processes such as expression of endogenous genes, activity of endogenous proteins, or alterations in metabolism of these cells across various states of immunity and anatomical locations. This section will detail the use of whole body imaging modalities to measure such activities.

575

36 Imaging in Immunology Research

a

Day 1

Day 3

Day 5

E

Day 8

E

Day 10

E

E

E

E

1.3

E

C

B B Day 12

C

A

C T

T

T B B

Day 15

B

BB

Day 17

Day 19

%ID/g

A

T B

Day 21

A T

T

T

T

A

B

B

B

Mouse 1

Mouse 2

Mouse 3

Mouse 4 E

E

1.3

C

C B

E

B

T

B

T

Mouse 5

T A

A

Mouse 6

B

T

Mouse 7 E

E

B T

B

B

T

%ID/g

b

0.3

B T 0.3

Fig. 36.5 Serial imaging and replicate studies of combined adoptive transfer immunotherapy reveal patterns of dynamic and individual T cell trafficking to lymph nodes and tumor sites. Mice were challenged with gp10025–33 self-antigen-expressing melanoma and treated with a regiment of lymphodepletion (X-ray therapy, 500 cGy) and T cell adoptive transfer on day 0, and gp10025–33/DC vaccination and high-dose IL-2 administration on days 0, 7, and 14. Each microPET/CT scan was performed 60 min after injection of 18F-FHBG reporter probe. (a)

Longitudinal scans of representative mouse showing dynamic HSV1-sr39tk transduced pmel-1 T cell patterns. Red arrows indicate gp10025–33/DC vaccination and high-dose IL-2 administration. (b) Scans of replicate mice demonstrate variability in cell trafficking and accumulation of 18F-FHBG indicative of T cell localization. T tumor; A axillary LN; B brachial LN; C cervical LN; E eye; %ID/g percent injected dose of PET probe per gram of tissue (adapted from Shu et al. 2009)

576

36.4.4.1 Imaging Endogenous Gene Expression and Molecular Interactions in Immune Cells PET reporter gene imaging has the potential to measure transcriptional activity and specific molecular interactions critical to immune function in vivo. Ponomarev et al. (2001) used PET to image activated T cells. The HSV1-tk PET reporter was placed under control of an enhancer specific for nuclear factor of activated T cells (NFAT) in Jurkat cells, an immortal T cell line. NFAT is a transcription factor that regulates the expression of several genes downstream of the TCR (reviewed in Crabtree (1999)). NFAT activity of Jurkat T cells was monitored by 124I-FIAU microPET before and after activation in vivo by intravenous administration of antihuman CD3 and CD28 antibodies. PET signals from Jurkat T cells expressing the PET reporter under the control of NFAT and injected subcutaneously were similar to those obtained with Jurkat T cells constitutively expressing HSV1-tk. These results correlated with flow cytometric analyses of CD69 levels, a marker of early T cell activation. Cellular promoters are often too weak to drive sufficient reporter gene expression to meet lower limits of detection. The two-step transcriptional activation (TSTA (Iyer et al. 2005; Ilagan et al. 2006)) system circumvents this limitation by placing a potent transcriptional activator under control of the cellular promoter of interest. Promoter activation by transcriptional machinery during gene induction leads to the expression of the transcriptional activator, which in turn binds in trans to a synthetic promoter that drives expression of the reporter gene. This strategy has allowed investigators to elucidate key signaling mechanisms during prostate cancer progression (Sato et al. 2005; Johnson et al. 2005). Although this has not been applied specifically to monitoring T cells, it has potential for robust measurements of gene expression critical to T cell function in a noninvasive, site-specific manner throughout the body. PET may allow visualization of protein–protein interactions in vivo in dynamic T cell processes during antigenic activation, proliferation, homing, and persistence. Luker et al. (2002) monitored the interaction between T antigen of Simian virus-40 (TAg) and p53 in vivo. In this study, doxycycline treatment of mice induced simultaneous expression of two fusion proteins in xenograft tumors from engineered HeLa cell

J.T. Lee et al.

lines: (a) TAg fused to the VP16 transactivation domain and (b) p53 fused to the Gal4 DNA binding domain (Gal4-DBD). The p53-Gal4-DBD fusion bound to the Gal4 promoter on a reporter construct containing mNLSsr39tk, a variant of the HSV1-sr39tk reporter gene. Interaction of TAg with p53 brought the VP16 transcriptional activator in close proximity to the promoter complex leading to expression of mNLS-sr39tk. 18F-FHBG PET imaging of tumor-bearing mice quantified the amount of TAg/p53 interaction which was shown to be dependent on the dose of doxycycline. Ray et al. (2002) designed a similar system, this time using BLI to monitor the interaction of MyoD (a skeletal muscle transcription factor) and Id (a protein inhibitor of DNA binding) via the VP16-Gal4 transcription system. Successful application of this technique to T cells may enable noninvasive quantification of critical immune response-regulated protein–protein interactions. Ongoing investigations into reporter complementation and reconstitution strategies could also provide novel ways of visualizing protein and cellular interactions in vivo (Paulmurugan and Gambhir 2006, 2007). In these strategies reporter protein fragments are fused to proteins of interest with the interaction of these proteins leading to recovery of reporter function and, in turn, reporter signal.

36.4.4.2 Imaging T Cell Function Using PET Probes of Cellular Metabolism Over the last decade, it has become increasingly clear that the function of diverse immune cell types including lymphocytes and macrophages is critically dependent on their ability to modulate cellular metabolism in response to external stimuli. Regulation of glycolysis through HIF-1a is essential for innate immune cell function in inflammatory hypoxic environments (Cramer et al. 2003). Enhanced glycolytic activity is also observed in T and B cells during activation and clonal expansion (Frauwirth et al. 2002; Woodland et al. 2008). Thus, measuring immune cell metabolic activity in vivo should provide insight into immune function. The PET probe, 18F-2-fluorodeoxyglucose (18F-FDG), measures glycolytic activity (reviewed in Phelps et al. (2000)) and is extensively used in the clinic for diagnosis, staging, and treatment monitoring of cancer and other disorders. 18F-FDG is an analog of glucose and,

577

36 Imaging in Immunology Research

upon transport into cells, is phosphorylated and trapped by the initial and rate-limiting enzyme, hexokinase. Cells that exhibit high glycolytic activity phosphorylate and retain greater quantities of 18F-FDG and manifest stronger PET signals. Activated lymphocytes exhibit upregulated glycolysis and 18F-FDG has been used to image the magnitude of inflammation in models of autoimmunity, including the EAE mouse model of multiple sclerosis. Using 18F-FDG PET, it was possible to measure an increase in glycolytic activity in areas of neuroinflammation that contained dense infiltrates of activated lymphocytes. Treatment with the immunosuppressive drug dexamethasone diminished 18F-FDG uptake in these areas and limited disease progression (Radu et al. 2007) (Fig. 36.6). In murine colitis disease models, a positive correlation between disease activity and PET signal was observed using intestinal imaging with 18F-FDG PET and a contrast-based CT isocontour method. This suggested that mucosal 18F-FDG signal was associated with activated but not quiescent CD4+ T cells (Brewer et al. 2008). Additional applications of 18F-FDG include studies of rheumatoid arthritis (Matsui et al. 2009) and GVHD (Stelljes et al. 2008). Glucose transporter and hexokinase activities are not limited to lymphocytes and innate immune cells

within inflammatory foci, but also observed in the brain, heart, and a variety of tumor types. These “baseline” 18F-FDG signals limit the ability to differentiate lymphocyte infiltration from surrounding tissue in these regions. Our group has sought to develop new probes which complement 18F-FDG in visualizing and assessing T cell-mediated immunity. Whereas most tissues rely on de novo production of deoxyribonucleoside precursors for DNA synthesis, activated T cells also utilize the nucleoside salvage pathway (NSP) which recycles extracellular deoxyribonucleosides from the extracellular milieu (Fig. 36.7a). We employed a differential screening strategy to identify small molecule probes that would preferentially be transported, phosphorylated, and retained in activated T cells via the NSP (Radu et al. 2008) (Fig. 36.7b). 18F-FAC (1-(2¢-deoxy-2¢-18F-fluoroarabino-furanosyl) cytosine) is an analog of the chemotherapy prodrug gemcitabine and a substrate for deoxycytidine kinase, a rate-limiting enzyme in nucleoside salvage. 18F-FAC PET signals were observed in the spleen, thymus, and lymph nodes in C57Bl/6 mice. This approach enabled the detection of localized immune activation in the MSV/ MuLV model of antitumor immunity. Moreover, consistent with the pathology associated with systemic autoimmunity observed in faslpr mice, the accumulation of 18F-FAC indicated enlargement of the spleen, MOG35-55/CFA/PTX

H

16

B

16 H

H

SC

SC

Fig. 36.6 18F-FDG PET/CT imaging in experimental autoimmune encephalomyelitis. Mice were immunized with the self neuro-antigen MOG35–55 in complete Freund’s adjuvant (CFA). Mice also received injections with pertussis toxin (PTX) on days 0 and 2. Left: data were analyzed by drawing 3D regions of interest (ROI) corresponding to each vertebrae using coregis-

BL SC

3

BL SC

16

SC

SC

SC 1

B

BF

BF

BL

microCT

B

Score 3 day 19

%ID/g

H Cervical C1-8 (9mm3/ROI) Thoracic T1-13 (9-27mm3/ROI) Lumbar L1-6 (27mm3/ROI)

16

%ID/g

B

%ID/g

3D ROI

Score 1 day 16

%ID/g

Score 0 day 12

Unimmunized

Sagittal

4

BL

4

SC

tered CT images; right: sagittal PET/CT overlays show increasing intensity of PET signal in the spinal cord during early disease progression (clinical score 1). SC spinal column; BF brown fat; BL bladder; H heart; B brain; %ID/g percent injected dose of PET probe per gram of tissue (adapted from Radu et al. 2007)

578

J.T. Lee et al. treated

untreated

13.9

B T SP

GI

SP

GI

% ID/g

LN

BL BL

0.8

Fig. 36.7 F-FAC PET imaging of a model of systemic autoimmunity. (a) 18F-FAC and gemcitabine are metabolized similarly via the deoxycytidine nucleoside salvage pathway. (b) Chemical structure of 18F-FAC, an analog of deoxycytidine. (c) PET scans of representative C57BL/6J wild-type mouse (left) and B6. MRL-Faslpr/J mouse model of autoimmunity (right) after 60 min injection of 18F-FAC. PET imaging reveals increased spleen and lymph node signals reflective of expected increased lymphoid mass in the same tissues due to the autoimmune phenotype. dA deoxyadenosine; dC deoxycytidine; dG deoxyguanosine; dT deoxythymidine; MP monophosphate; DP diphosphate; TP triphosphate; ENT equilibrative nucleoside transporter; CDA cytidine deaminase; DCTD dCMP deaminase; TS thymidylate synthetase; TK1 thymidine kinase 1; TK2 thymidine kinase 2; DGK deoxyguanosine kinase; DNA deoxyribonucleic acid; BM bone narrow; LN lymph node; T thymus; SP spleen; GI gastrointestinal tract; BL bladder; %ID/g percent injected dose per gram of tissue (adapted from Radu et al. 2008) 18

thymus, and lymph nodes when injected into these mice (Fig. 36.7c). In the gp100/pmel-1 antigen-specific adoptive cell therapy model of melanoma, 18FFAC was able to monitor correlates of tumor response to therapy such as increased spleen size (Laing et al., unpublished data, Fig. 36.8). However, this approach could not visualize tumor infiltrating T cells. A direct comparison of 18F-FAC and 18F-FDG accumulation in immune cell subtypes isolated from different tissues during an antitumor immune response demonstrated 18F-FAC was associated with the cellular proliferation of innate leukocytes and T cells, while 18F-FDG was mostly retained in CD11bhi cells consisting of macrophages and granulocytes

Fig. 36.8 18F-FAC microPET/CT imaging of adoptive cell transfer immunotherapy. Image analysis of a representative C57Bl/6 mouse treated (right) with combined immunotherapy shows increased PET signal from 18F-FAC accumulation in the spleen as compared to untreated control (left). B bone marrow; T thymus; SP spleen; LN lymph node; GI gastrointestinal tract; BL bladder; %ID/g percent injected dose per gram of tissue (Laing et al., unpublished data)

(Nair-Gill et al. 2010). This finding shows that PET probes for different metabolic pathways provide distinct information about immune cell composition and function in vivo.

36.4.4.3 Future Directions for PET Imaging Functional imaging of the immune system using metabolic PET probes can provide site-specific information about immune activity within an organism during an immune response. This technology, however, is not cell-specific. Metabolic pathways such as glycolysis and nucleoside salvage are likely to be differentially regulated in multiple immune and other cell types in a manner that depends on the type of stimulus. Coupling the reporter imaging approaches described earlier in this chapter to metabolic PET imaging could identify site-specific changes in immune cell function within an organism. These types of investigations could better serve to evaluate the effects of immune-modulatory therapies targeting specific aspects of immune function.

36 Imaging in Immunology Research

579

36.5 Imaging Limitations and Improvements

challenges that contribute toward the ability of PET to visualize and quantify low concentrations of a desired signal. These advances impact photon sensitivity and spatial, energy, and temporal resolutions which affect what Levin calls “molecular sensitivity,” or the ability to detect, visualize, and quantify low concentrations of molecular probe interacting with a molecular target. The achieved molecular sensitivity is dependent on several key factors including detector geometric arrangement and efficiency of photon capture, time resolution, and the windowing of photon energies accepted. Time and energy windowing may allow for higher capture of photon activity, but also increases background and acceptance of random and scattered coincidences that are not true events. Furthermore, issues such as spillover and partial volumes due to poor image resolutions can lead to inaccurate signal differentiation between adjacent regions (Phelps 2004). Sensitivity of small animal imaging systems is important, particularly for imaging regions of the immune system with relatively low cell numbers such as lymph nodes and heterogeneous organs or tissues with low density of immune cells. In studies performed with preclinical PET scanners with sensitivity of approximately 4%, imaging of adoptively transferred T cells had limits of detection of approximately 10% of total cells in spleen and 3% in lymph nodes (Shu et al. 2009). MicroPET systems are improving with current sensitivities reaching approximately 7% indicating enhancement of the lower limit of detection. Additionally, Goertzenet al. (2007) demonstrated the possibility of visualizing sources less than 1 kBq (27 nCi) in lutetium oxyorthosilicate-based animal PET systems by optimizing energy windowing. Thus, higher sensitivity in combination with reduction in intrinsic noise from background radioactivity may allow for imaging of anatomically small or heterogeneous ROI. Similar improvements in clinical PET scanners will be critical to the translation of the aforementioned results.

Increasing sensitivity of imaging by improving limit of detection parameters will be critical to expand the use of molecular imaging in immunology. We have discussed in earlier sections modifications that can be made to reporter gene/probe systems to enhance their sensitivity. Here we will briefly present optimizations in imaging instrumentation and postimaging analyses.

36.5.1 Instrumentation The intrinsic spatial resolution, which defines the ability of an imaging system to accurately separate adjacent objects, is a critical limitation in the detection and quantification of lymphocyte populations using PET. Current microPET scanners have a lower limit resolution of approximately 1 mm3. At this threshold, difficulty exists in distinguishing lymphoid regions of interest from adjacent anatomical areas. Furthermore, it is not possible to visualize organ substructures such as follicles or peri-arterial lymphoid sheaths that are characteristic of large secondary lymphoid organs such as the spleen. Tissue contrast enhancement depends strongly on both molecular probe design and image acquisition and processing. Improvements of spatial, energy, and temporal resolutions in PET systems are needed in order to advance molecular imaging capabilities of small animal PET for the study of the immune system (Levin 2008). Coregistration of microPET with microCT provides the sensitivity of PET and high spatial resolution of CT, enabling anatomical segregation of PET signals in region-of-interest (ROI) analysis. High spatial resolution imaging modalities such as microCT and dedicated small animal MRI devices have resolutions of 50 and 25 mm, respectively (reviewed in Massoud and Gambhir (2003)). Emerging imaging platforms such as PET/MRI demonstrate a combination of enhanced resolution and sensitivity (Catana et al. 2006; Judenhofer et al. 2008). MRI is superior to CT in resolving soft tissue detail and, during an immune response, can define the occurrence of microvascular disruption and pathology (Denis et al. 2004). In a detailed analysis of PET imaging technology, Levin (2008) discusses the various improvements and

36.5.2 PET Data Analysis A quantitative analysis of PET imaging for beta islet cells (1–2% of pancreatic mass) suggested that the discrimination of beta cells from surrounding exocrine cells requires the former to retain approximately 1,000

580

times more probe (Sweet et al. 2004). Thus, analytic tools capable of adjusting for the difference in cellular retention are required to decouple the signal emanating from the cells of interest from the surrounding tissue. In immunocompetent wild-type mice, T cells constitute greater than 95 and 30% of the thymus and spleen, respectively. Although T cells comprise ~40–50% of lymph nodes, if total cellularity falls below the resolution of imaging modalities, it would be difficult to visualize and accurately quantify the lymphocyte population. Correlation of PET signal to cell numbers was studied by Su et al. (2004). Cells expressing HSV1-sr39tk were injected directly into established vascularized tumor models. 18F-FHBG PET signals were quantified and normalized to the total injected dose and the mass of the tissue on a per-cell basis. Signal was confirmed to be cell number-dependent with a lower limit of detection of 106 cells per 100 mL. Shu et al. (2009) indicated that the lower limit of PET detection in a model of adoptive CD8+ T cell immunotherapy was determined to be 104 T cells in lymph nodes with 3% of total cells being HSV1-sr39tk+ and 7 × 105 in the spleen with 10% being HSV1-sr39tk+, with cell density potentially explaining differences between tissues. It will be necessary to quantitatively define the thresholds for resolution and sensitivity of imaging scanners and specificity of probes in order to visualize small or heterogeneous subpopulations of immune cells. Further aspects that affect the ability of PET to measure specific in vivo processes involve probe metabolism and postimaging data analyses. Undesired endogenous proteins may function to transport or metabolize probes that are analogs of these substrates, thereby decreasing specificity. Furthermore, endogenous substrates may compete for protein targets and reduce signal. Kinetic studies of probes and their metabolites using high performance liquid chromatography, liquid chromatography/mass spectrometry, and tracer kinetic modeling (Mankoff et al. 1998; Huang et al. 1980; Green et al. 2004) may be employed to differentiate the contributions of these factors to overall signals in ROI analysis and quantify in vivo dynamics. Kinetic models describing endogenous immune cell trafficking and metabolism under quiescent conditions are necessary for comparison to active immunity and manipulations thereof. With the development of 18F-FAC, we seek to employ this technique for delineating intrinsic factors that differentiate probe retention in various lymphoid organs.

J.T. Lee et al.

36.6 Concluding Remarks Whole body imaging technologies for small animal studies have provided critical platforms for noninvasive monitoring of the immune system in vivo. BLI, PET, CT, and MRI technologies provide invaluable tools for studying the intact immune system and its response to various perturbations throughout the body. Improvements at all levels of research from genetic enhancements of reporter genes to probe development to imaging and analysis technology are critical to the continued use of live animal imaging to answer immunological questions. Thus, there is a great need for multidisciplinary collaborations among immunologists, physicists, engineers, and mathematicians to further this field. Successes to date are indicative of the possibilities of molecular imaging in visualizing and monitoring the intricacies of the immune system. Acknowledgments We acknowledge all colleagues who contributed to the work presented here. We acknowledge Andrew Tran for helping with figure preparation and Barbara Anderson for helping with manuscript preparation. JTL is supported by the In Vivo Cellular and Molecular Imaging Centers (ICMIC) Developmental Project Award NIH 5 P50 CA86306. ENG is supported by US National Institutes of Health T32 GM08042 UCLA Medical Scientist Training Program and Interdisciplinary Training in Virology and Gene Therapy T32 A1065067. CGR is supported by National Cancer Institute Grant No. 5U54 CA119347, ICMIC Developmental Project Award NIH P50 CA86306, the CIRM Tools and Technology Grant and the Dana Foundation. ONW is an Investigator of the Howard Hughes Medical Institute.

References Abdel-Azim H et al (2008) Expansion of multipotent and lymphoid-committed human progenitors through intracellular dimerization of Mpl. Blood 111:4064–4074 Adonai N et al (2002) Ex vivo cell labeling with 64Cu-pyruvaldehydebis(N4-methylthiosemicarbazone) for imaging cell trafficking in mice with positron-emission tomography. Proc Natl Acad Sci U S A 99:3030–3035 Alauddin MM et al (2007) In vivo evaluation of 2¢-deoxy -2¢-[(18)F]fluoro-5-iodo-1-beta-D: -arabinofuranosyluracil ([(18)F]FIAU) and 2¢-deoxy-2¢-[ (18)F]fluoro-5-ethyl-1beta-D: -arabinofuranosyluracil ([ (18)F]FEAU) as markers for suicide gene expression. Eur J Nucl Med Mol Imaging 34:822–829 Balaban EP, Simon TR, Sheehan RG, Frenkel EP (1986) Effect of the radiolabel mediator tropolone on lymphocyte structure and function. J Lab Clin Med 107:306–314

36 Imaging in Immunology Research Balaban EP, Simon TR, Frenkel EP (1987) Toxicity of indium-111 on the radiolabeled lymphocyte. J Nucl Med 28: 229–233 Balzarini J et al (2006) Engineering of a single conserved amino acid residue of herpes simplex virus type 1 thymidine kinase allows a predominant shift from pyrimidine to purine nucleoside phosphorylation. J Biol Chem 281:19273–19279 Beilhack A et al (2005) In vivo analyses of early events in acute graft-versus-host disease reveal sequential infiltration of T-cell subsets. Blood 106:1113–1122 Biron CA (1997) Activation and function of natural killer cell responses during viral infections. Curr Opin Immunol 9:24–34 Black ME, Newcomb TG, Wilson H-MP, Loeb LA (1996) Creation of drug-specific hepes simplex virus type 1 thymidine kinase mutant for gene therapy. Proc Natl Acad Sci U S A 93:3525–3529 Botti C et al (1997) Comparison of three different methods for radiolabelling human activated T lymphocytes. Eur J Nucl Med Mol Imaging 24:497–504 Branda CS, Dymecki SM (2004) Talking about a revolution: the impact of site-specific recombinases on genetic analyses in mice. Dev Cell 6:7–28 Brewer S et al (2008) Molecular imaging of murine intestinal inflammation with 2-deoxy-2-[18F]fluoro-D-glucose and positron emission tomography. Gastroenterology 135:744–755 Buursma AR et al (2006) 18F-FEAU as a radiotracer for herpes simplex virus thymidine kinase gene expression: in-vitro comparison with other PET tracers. Nucl Med Commun 27:25–30 Catana C et al (2006) Simultaneous acquisition of multislice PET and MR images: initial results with a MR-compatible PET scanner. J Nucl Med 47:1968–1976 Clay TM, Hobeika AC, Mosca PJ, Lyerly HK, Morse MA (2001) Assays for monitoring cellular immune responses to active immunotherapy of cancer. Clin Cancer Res 7:1127–1135 Costa GL et al (2001) Adoptive immunotherapy of experimental autoimmune encephalomyelitis via T cell delivery of the IL-12 p40 subunit. J Immunol 167:2379–2387 Crabtree GR (1999) Generic signals and specific outcomes: signaling through Ca2+, calcineurin, and NF-AT. Cell 96:611–614 Cramer T et al (2003) HIF-1alpha is essential for myeloid cellmediated inflammation. Cell 112:645–657 Denis MC, Mahmood U, Benoist C, Mathis D, Weissleder R (2004) Imaging inflammation of the pancreatic islets in type 1 diabetes. Proc Natl Acad Sci U S A 101:12634–12639 Dubey P et al (2003) Quantitative imaging of the T cell antitumor response by positron-emission tomography. Proc Natl Acad Sci U S A 100:1232–1237 Frauwirth KA et al (2002) The CD28 signaling pathway regulates glucose metabolism. Immunity 16:769–777 Gallaher SD, Gil JS, Dorigo O, Berk AJ (2009) Robust in vivo transduction of a genetically stable Epstein-Barr virus episome to hepatocytes in mice by a hybrid viral vector. J Virol 83:3249–3257 Galon J et al (2006) Type, density, and location of immune cells within human colorectal tumors predict clinical outcome. Science 313:1960–1964 Gambhir SS et al (2000) A mutant herpes simplex virus type 1 thymidine kinase reporter gene shows improved sensitivity for imaging reporter gene expression with positron emission tomography. Proc Natl Acad Sci U S A 97:2785–2790

581 Goertzen AL, Suk JY, Thompson CJ (2007) Imaging of weaksource distributions in LSO-based small-animal PET scanners. J Nucl Med 48:1692–1698 Green LA et al (2004) A tracer kinetic model for 18F-FHBG for quantitating herpes simplex virus type 1 thymidine kinase reporter gene expression in living animals using PET. J Nucl Med 45:1560–1570 Hediger MA et al (2004) The ABCs of solute carriers: physiological, pathological and therapeutic implications of human membrane transport proteinsIntroduction. Pflugers Arch 447:465–468 Huang SC et al (1980) Noninvasive determination of local cerebral metabolic rate of glucose in man. Am J Physiol 238: E69–E82 Hwang LN, Yu Z, Palmer DC, Restifo NP (2006) The in vivo expansion rate of properly stimulated transferred CD8+ T cells exceeds that of an aggressively growing mouse tumor. Cancer Res 66:1132–1138 Ilagan R et al (2006) Imaging mitogen-activated protein kinase function in xenograft models of prostate cancer. Cancer Res 66:10778–10785 Iyer M et al (2005) Non-invasive imaging of a transgenic mouse model using a prostate-specific two-step transcriptional amplification strategy. Transgenic Res 14:47–55 Johnson M et al (2005) Micro-PET/CT monitoring of herpes thymidine kinase suicide gene therapy in a prostate cancer xenograft: the advantage of a cell-specific transcriptional targeting approach. Mol Imaging 4:463–472 Johnson LA et al (2009) Gene therapy with human and mouse T-cell receptors mediates cancer regression and targets normal tissues expressing cognate antigen. Blood 114: 535–546 Judenhofer MS et al (2008) Simultaneous PET-MRI: a new approach for functional and morphological imaging. Nat Med 14:459–465 Koehne G et al (2003) Serial in vivo imaging of the targeted migration of human HSV-TK-transduced antigen-specific lymphocytes. Nat Biotechnol 21:405–413 Kohn DB, Sadelain M, Glorioso JC (2003) Occurrence of leukaemia following gene therapy of X-linked SCID. Nat Rev Cancer 3:477–488 Lee GR, Fields PE, Griffin TJ, Flavell RA (2003) Regulation of the Th2 cytokine locus by a locus control region. Immunity 19:145–153 Levin CS (2008) New imaging technologies to enhance the molecular sensitivity of positron emission tomography. Proc IEEE 96:439–467 Liang Q et al (2002a) Monitoring adenoviral DNA delivery, using a mutant herpes simplex virus type 1 thymidine kinase gene as a PET reporter gene. Gene Ther 9:1659–1666 Liang Q et al (2002b) Noninvasive, repetitive, quantitative measurement of gene expression from a bicistronic message by positron emission tomography, following gene transfer with adenovirus. Mol Ther 6:73–82 Likar Y et al (2009) PET imaging of HSV1-tk mutants with acquired specificity toward pyrimidine- and acycloguanosine-based radiotracers. Eur J Nucl Med Mol Imaging 36: 1273–1282 Lois C, Hong EJ, Pease S, Brown EJ, Baltimore D (2002) Germline transmission and tissue-specific expression of transgenes delivered by lentiviral vectors. Science 295:868–872

582 Luker GD et al (2002) Noninvasive imaging of protein-protein interactions in living animals. Proc Natl Acad Sci U S A 99:6961–6966 Mankoff DA et al (1998) Kinetic analysis of 2-[carbon-11]thymidine PET imaging studies: compartmental model and mathematical analysis. J Nucl Med 39:1043–1055 Mantovani A, Allavena P, Sica A, Balkwill F (2008) Cancerrelated inflammation. Nature 454:436–444 Massoud TF, Gambhir SS (2003) Molecular imaging in living subjects: seeing fundamental biological processes in a new light. Genes Dev 17:545–580 Matsui K, Wang Z, McCarthy TJ, Allen PM, Reichert DE (2004) Quantitation and visualization of tumor-specific T cells in the secondary lymphoid organs during and after tumor elimination by PET. Nucl Med Biol 31:1021–1031 Matsui T et al (2009) Inflammatory cytokines and hypoxia contribute to 18F-FDG uptake by cells involved in pannus formation in rheumatoid arthritis. J Nucl Med 50:920–926 Milan G et al (1999) Dissecting the immune response to moloney murine sarcoma/leukemia virus-induced tumors by means of a DNA vaccination approach. J Virol 73:2280–2287 Mohrs M, Shinkai K, Mohrs K, Locksley RM (2001) Analysis of type 2 immunity in vivo with a bicistronic IL-4 reporter. Immunity 15:303–311 Murali-Krishna K et al (1998) Counting antigen-specific CD8 T cells: a reevaluation of bystander activation during viral infection. Immunity 8:177–187 Murphy KP, Travers P, Walport M, Janeway C (2008) Janeway’s immunobiology. Garland Science, New York Muruve DA (2004) The innate immune response to adenovirus vectors. Hum Gene Ther 15:1157–1166 Nair-Gill et al (2010) PET probes for distinct metabolic pathways have different cell specificities during immune responses in mice. 120(6):2005–2015 Negrin RS, Contag CH (2006) In vivo imaging using bioluminescence: a tool for probing graft-versus-host disease. Nat Rev Immunol 6:484–490 Nishimura R et al (2008) In vivo trafficking and survival of cytokine-induced killer cells resulting in minimal GVHD with retention of antitumor activity. Blood 112: 2563–2574 Ntziachristos V, Ripoll J, Wang LV, Weissleder R (2005) Looking and listening to light: the evolution of whole-body photonic imaging. Nat Biotechnol 23:313–320 Overwijk WW et al (2003) Tumor regression and autoimmunity after reversal of a functionally tolerant state of self-reactive CD8+ T cells. J Exp Med 198:569–580 Paik JY, Lee KH, Byun SS, Choe YS, Kim BT (2002) Use of insulin to improve [18 F]fluorodeoxyglucose labelling and retention for in vivo positron emission tomography imaging of monocyte trafficking. Nucl Med Commun 23:551–557 Paulmurugan R, Gambhir SS (2006) An intramolecular folding sensor for imaging estrogen receptor-ligand interactions. Proc Natl Acad Sci U S A 103:15883–15888 Paulmurugan R, Gambhir SS (2007) Combinatorial library screening for developing an improved split-firefly luciferase fragment-assisted complementation system for studying protein-protein interactions. Anal Chem 79:2346–2353 Pfeifer A et al (2001) Transduction of liver cells by lentiviral vectors: analysis in living animals by fluorescence imaging. Mol Ther 3:319–322

J.T. Lee et al. Phelps ME (2000) PET: the merging of biology and imaging into molecular imaging. J Nucl Med 41:661–681 Phelps ME (2004) PET: molecular imaging and its biological applications. Springe, New York Pittet MJ et al (2007) In vivo imaging of T cell delivery to tumors after adoptive transfer therapy. Proc Natl Acad Sci U S A 104:12457–12461 Ponomarev V et al (2001) Imaging TCR-dependent NFATmediated T-cell activation with positron emission tomography in vivo. Neoplasia 3:480–488 Ponomarev V et al (2007) A human-derived reporter gene for noninvasive imaging in humans: mitochondrial thymidine kinase type 2. J Nucl Med 48:819–826 Rabinovich BA et al (2008) Visualizing fewer than 10 mouse T cells with an enhanced firefly luciferase in immunocompetent mouse models of cancer. Proc Natl Acad Sci U S A 105:14342–14346 Radu CG, Shu CJ, Shelly SM, Phelps ME, Witte ON (2007) Positron emission tomography with computed tomography imaging of neuroinflammation in experimental autoimmune encephalomyelitis. Proc Natl Acad Sci U S A 104: 1937–1942 Radu CG et al (2008) Molecular imaging of lymphoid organs and immune activation by positron emission tomography with a new [18F]-labeled 2¢-deoxycytidine analog. Nat Med 14:783–788 Raulet DH (2003) Roles of the NKG2D immunoreceptor and its ligands. Nat Rev Immunol 3:781–790 Ray P et al (2002) Noninvasive quantitative imaging of proteinprotein interactions in living subjects. Proc Natl Acad Sci U S A 99:3105–3110 Ray P, De A, Min JJ, Tsien RY, Gambhir SS (2004) Imaging trifusion multimodality reporter gene expression in living subjects. Cancer Res 64:1323–1330 Rettig GR, Rice KG (2007) Non-viral gene delivery: from the needle to the nucleus. Expert Opin Biol Ther 7:799–808 Ridolfi R et al (2004) Evaluation of in vivo labelled dendritic cell migration in cancer patients. J Transl Med 2:27 Santos EB et al (2009) Sensitive in vivo imaging of T cells using a membrane-bound Gaussia princeps luciferase. Nat Med 15:338–344 Sato M et al (2005) Functionality of androgen receptor-based gene expression imaging in hormone refractory prostate cancer. Clin Cancer Res 11:3743–3749 Schepers K et al (2002) Differential kinetics of antigen-specific CD4+ and CD8+ T cell responses in the regression of retrovirus-induced sarcomas. J Immunol 169:3191–3199 Serganova I et al (2004) Molecular imaging of temporal dynamics and spatial heterogeneity of hypoxia-inducible factor-1 signal transduction activity in tumors in living mice. Cancer Res 64:6101–6108 Shaner NC, Steinbach PA, Tsien RY (2005) A guide to choosing fluorescent proteins. Nat Methods 2:905–909 Shu CJ et al (2005) Visualization of a primary anti-tumor immune response by positron emission tomography. Proc Natl Acad Sci U S A 102:17412–17417 Shu CJ et al (2009) Quantitative PET reporter gene imaging of CD8+ T cells specific for a melanoma-expressed self-antigen. Int Immunol 21:155–165 Snapp EL (2009) Fluorescent proteins: a cell biologist’s user guide. Trends Cell Biol 19:649–655

36 Imaging in Immunology Research Stelljes M et al (2008) Clinical molecular imaging in intestinal graft-versus-host disease: mapping of disease activity, prediction, and monitoring of treatment efficiency by positron emission tomography. Blood 111:2909–2918 Su H, Forbes A, Gambhir SS, Braun J (2004) Quantitation of cell number by a positron emission tomography reporter gene strategy. Mol Imaging Biol 6:139–148 Su H, Chang DS, Gambhir SS, Braun J (2006) Monitoring the antitumor response of naive and memory CD8 T cells in RAG1−/− mice by positron-emission tomography. J Immunol 176:4459–4467 Sweet IR, Cook DL, Lernmark A, Greenbaum CJ, Krohn KA (2004) Non-invasive imaging of beta cell mass: a quantitative analysis. Diabetes Technol Ther 6:652–659 Tannous BA, Kim DE, Fernandez JL, Weissleder R, Breakefield XO (2005) Codon-optimized Gaussia luciferase cDNA for mammalian gene expression in culture and in vivo. Mol Ther 11:435–443 Tseng JC et al (2006) Tumor-specific in vivo transfection with HSV-1 thymidine kinase gene using a Sindbis viral vector as a basis for prodrug ganciclovir activation and PET. J Nucl Med 47:1136–1143 Vivier E, Tomasello E, Paul P (2002) Lymphocyte activation via NKG2D: towards a new paradigm in immune recognition? Curr Opin Immunol 14:306–311 Wang X et al (2003) Dynamic tracking of human hematopoietic stem cell engraftment using in vivo bioluminescence imaging. Blood 102:3478–3482 Wei LH et al (2008) Engineered antibody fragments with infinite affinity as reporter genes for PET imaging. J Nucl Med 49:1828–1835

583 Woodland RT et al (2008) Multiple signaling pathways promote B lymphocyte stimulator dependent B-cell growth and survival. Blood 111:750–760 Xiong Z et al (2006) Imaging chemically modified adenovirus for targeting tumors expressing integrin alphavbeta3 in living mice with mutant herpes simplex virus type 1 thymidine kinase PET reporter gene. J Nucl Med 47: 130–139 Yaghoubi SS, Creusot RJ, Ray P, Fathman CG, Gambhir SS (2007) Multimodality imaging of T-cell hybridoma trafficking in collagen-induced arthritic mice: image-based estimation of the number of cells accumulating in mouse paws. J Biomed Opt 12:064025 Yang TT, Sinai P, Kitts PA, Kain SR (1997) Quantification of gene expression with a secreted alkaline phosphatase reporter system. Biotechniques 23:1110–1114 Yang L et al (2008) Engineered lentivector targeting of dendritic cells for in vivo immunization. Nat Biotechnol 26: 326–334 Zhang Q, Yin L, Tan Y, Yuan Z, Jiang H (2008) Quantitative bioluminescence tomography guided by diffuse optical tomography. Opt Express 16:1481–1486 Zhu Z, Zheng T, Lee CG, Homer RJ, Elias JA (2002) Tetracyclinecontrolled transcriptional regulation systems: advances and application in transgenic animal modeling. Semin Cell Dev Biol 13:121–128

Appendix Relevant Physiological Standard Parameters of Rodents

Relevant Physiological Standard Parameters

Recommendations of the GV-SOLAS Blood Collection

Mouse

Nude mouse

Rat

Body weight (g)

20–35

20–35

250–600

Body temperature (°C)

36–38

37.3

37.5–39

Respiratory frequency (breathing/min)

60–230

84–230

60–170

Species

Blood volumea ml/animalb ml/kg b.w.

Hematocrit % b.w. (%)

Heart frequency (beat/min)

300–800

300–800

300–500

Mouse (25 g)

1.7

74 (70–80)

6.6

42 (33–50)

Blood pressure (mmHg)

133–160/ 102–110

133–160/ 102–110

116/90

Rat (300 g)

19

64 (50–70)

6.4

46 (40–61)

Fertility (days)

28–49

56–70

50–72

30

75 (65–90)

7.5

44 (37–50)

Urine (ml/ animal/day)

1–3

1–3

10–15

Guinea pig (400 g)

Homing 20–24 temperature (°C)

24–28

18–24

Homing humidity (%)

40–65

50–70

See Tables A.1–A.4. Table A.1 Average blood volume

50–70

7.8 78 (65–80) 7.8 51 (39–59) Golden hamster (100 g) a Differences are possible regarding race, strain and age; the percentage amount is less in obese animals than in normal weighted b Estimated blood volume at the specified body weight in column 1 of this table

The hematologic parameters are highly species- and strain-specific and can be obtained from the commercial breeders, e.g., Charles River, Jackson Laboratories or Harlan Laboratories.

F. Kiessling and B.J. Pichler (eds.), Small Animal Imaging, DOI: 10.1007/978-3-642-12945-2, © Springer-Verlag Berlin Heidelberg 2011

585

586

Appendix

Table A.2 Parameters of blood collection Species

Examples of total blood volume collectiona Singleb (max. 10%) (ml) Dailyc (1%) (ml)

Finald (ml)

Mouse (25 g)

0.17

0.02

0.7–1.0

Rat (300 g)

1.9

0.2

10

Guinea pig (400 g)

3.0

0.3

15

Golden hamster (100 g) 0.7 0.07 3.5 The blood collection parameters are only valid for adult, healthy animals and can be less for freshly manipulated, diseased, old or stressed animals b A recovery phase for at least 2–3 weeks necessary c After daily blood collection at maximum for 2 weeks a recovery phase for at least 2–3 weeks necessary d Only under anesthetic; approx. 50% of the whole blood volume collectable (determined empirically) a

Table A.3 Techniques for blood collection in mice Blood collection Quantity and frequency

Location

Specifics

1. Small amount (0.02–0.04 ml)

(a) Tail vein

(b) Infants only up to the end of the suckle period

(b) Amputation of tail tip

(c) Under anesthetic (e.g., isoflurane) by using a glass capillary with an outer diameter of 0.8 mm

(c) Retro-bulbar (d) V. facialis (e) V. saphena 2. Single maximum blood volume (see Table A.2 as well)

(a) Retro-bulbar

(a) Identical to (1c)

(b) Tail vein

(b) 24–26-G needle

(c) Veins angle

(c) Anesthetic; 23–25-G needle

(d) V. facialis 3. Repeated collection (see Table A.2 as well)

(a) Tail vein (b) V. facialis (c) V. saphena (d) Retro-bulbar

4. Final collection (exsanguinations)

Use alternating blood collection locations!

(d) Anesthetic, only two blood collections per eye in intervals of 2 weeks

(a) Cardial

(a), (b) and (c) anesthetic

(b) Aorta

(c) Also possible without anesthetic

(c) V. cava (d) Decapitation

587

Appendix Table A.4 Techniques for blood collection in rats Blood collection Quantity and frequency

Location

Specifics

1. Small amount (0.04–0.05 ml)

(a) Tail vein

(a) 23–26-G needle

(b) Amputation of tail tip

(b) Infants only up to the end of the suckle period

(c) V. saphena 2. Single maximum blood volume (see Table A.2 as well)

3. Repeated collection (see Table A.2 as well)

(a) Tail vein

(a) 23–26-G needle

(b) Retro-bulbar

(b) Under anesthetic (e.g., isoflurane) by using a glass capillary up to an outer diameter of 0.9 mm

(c) Veins angle

(c) Anesthetic; 21–25-G needle

(a) Tail vein

(b) Anesthetic, polyethylene catheter, outer diameter of 0.61 mm, port system

(b) Permanent catheter in V. jugularis (c) Retro-bulbar

(c) Anesthetic, only two blood collections per eye in intervals of 2 weeks

(d) V. saphena 4. Final collection (exsanguinations)

(a) Cardial

(a), (b) and (c) anesthetic

(b) Aorta

(a) 20–21-G needle

(c) V. cava

(d) Also possible without anesthetic

(d) Decapitation

Criteria of Maximum Tumor Burden The tumor burden should not exceed 5% of the body weight of animals used for the passage of tumors. Furthermore, the tumor burden should be limited to a maximum of 10% of the body weight of the animals in therapeutic studies. Animals have to be sacrificed by Table A.5 Criteria of maximum tumor burden allowed Type of tumor

Maximum Volumes for Injection

Maximum diameter (cm) Mouse Rat

Requirements for injectable solutions are:

Single tumor

1.5

3.0

Multiple tumors (sum of all single tumors)

3.0

6.0

• • • •

Skin tumor

2.0

3.0

Subcutaneouslya

1.5

3.0

Two tumors

Each 1.5

Each 3.0

Autochthonic tumors

Tumors transplanted

Intramuscularly 0.5 Localized in back and flank regions

a

reaching the mentioned maximum tumor diameter or in cases that those criteria are met such as tumor ulceration, severe invasiveness resulting in pain and loss of vital functions or loss of body weight at a maximum of 20% (see Table A.5).

1.0

Isoton Neutral pH (pH 7.0–7.3) Body temperature Concentrations, chemical compositions or physical characteristics which do not induce general damages or local irritations

Hyper- or hypotonic solutions or solutions consisting of nonphysiologic pH-values are able to cause severe pain and tissue damages (e.g., in the case of perivascular injection) or lead to erythrocyte damages (hemolysis) (see Tables A.6 and A.7).

588

Appendix

Table A.6 Volume parameters for local injection of solutions Species

ml per injection site Intradermal

Intramuscular

µl/animala Intracerebro-ventriculara

Mouse

0.02 (27 G)

0.03 (27 G)

3 (28 G)

Rat

0.05 (27 G)

0.1 (23–25 G)

5 (28 G)b

Guinea pig

0.10 (27 G)

0.25 (25 G)

5 (28 G)b

Golden hamster 0.02 (27 G) 0.05 (24–25 G) A very slow injection speed is required (2–3 min in mouse and rat) b Through a 22-G guidance cannula

3 (28 G)b

a

Table A.7 Volume parameters for other injection sites Species

ml/kg b.w. Subcutaneous

Intraperitoneal

Intravenousa,b Bolus

G

Mouse

10 (25 G)

20 (25–27 G)

5

26–28

Rat

10 (25 G)

10 (23–25 G)

5

25–27

Guinea pig

10 (23–25 G)

10 (23–25 G)

5

25–27

Golden hamster 10 (23–25 G) 10 (23–25 G) 5 25–27 i.v. injection time: minimum of 1 min up to 2.5 min b Infusion taking longer than 2 hours should not exceed the maximum infusion volume < 10% of the total blood volume a

References Lab animal blood collection recommendations of the Society for Laboratory Animal Science (GV-SOLAS), http://www. gv-solas.de/auss/tie/tie_blutentnahme09.pdf Recommended criteria regarding a premature immolation of tumor burden mice and rats of the Society for Laboratory

Animal Science (GV-SOLAS). http://www.gv-solas.de/auss/ tie/tie_vorz.toetung2009.pdf Recommended maximum injection volumes in lab animals of the Society for Laboratory Animal Science (GV-SOLAS). http:// www.gv-solas.de/auss/tie/Injektionsvol_August_2010.pdf

Index

A Activatable probes, 282, 283, 285–288 ADC. See Apparent diffusion coefficient Adenovirus, 568 Aerosol, 101 Affine transformations, 371, 373 ALARA principle, 129, 133 Albumin-coated bubbles, 220 Alignment, 318, 322–324, 326, 327, 330, 333, 336, 338 Alpha-methylparatyrosine, 517 ALS. See Amyotrophic lateral sclerosis Alternative hypothesis, 376, 377 Alzheimer’s disease, 502, 505 Amphetamine (AMPH), 517–518, 520–523, 526, 528–530, 532 Amphiphilic Gd(III) complex, 174 a-Amyloid, 500, 502, 505–508 b-Amyloid plaques, 502, 505, 507–509 Amyloid precursor protein (APP), 502, 505, 506, 508, 509 Amyotrophic lateral sclerosis (ALS), 506 Anaesthesia, 94, 100, 101, 104, 106, 108, 113, 114 Analgesia, 83, 84, 90, 91 Analgetics, 89–91 Anesthesia, 30, 33–35, 37, 41–43, 45, 50, 55, 83–89, 434, 435 Angiogenesis, 171, 172, 224, 226, 431, 432, 438–443, 445, 452, 457, 460, 465, 474, 478, 479, 481, 484 Angiography, 126, 127 Animal facility, 61–63, 66–67 Animal handling, 161–162 Animal preparation, 434–435 Animal use issues, 30, 32–33 Annexin, 460 Annexin V (AnxV), 465, 480, 481 Annihilation quanta, 231, 232, 235 Anova. See One way analysis of variance Anti-angiogenic therapy, 224, 225, 478, 481 Antibody, 171 Antibody-based probes, 483 Antigen presenting cells (APC), 566 AnxV. See Annexin V APC. See Antigen presenting cells APD. See Avalanche photodiode Apolipoprotein E-knock-out (ApoE-/-), 452 Apoptosis, 449, 452, 460, 465, 479–481, 486 APP. See Amyloid precursor protein Apparent diffusion coefficient (ADC), 157, 162, 163

Arginine-glycine-aspartic acid (RGD), 465, 467 Arterial spin labeling MRI, 439 ASL. See Aterial spin labelling Assessment bias, 73, 74, 80 Asthma, 49, 51 Aterial spin labelling (ASL), 459 Atherosclerosis, 449–453, 456–458, 460, 461, 465–467 Atherosclerotic plaques, 188, 223, 449, 450, 456, 458, 460, 465, 467 Atom, 152, 153 Atropin, 114 Attenuation, 234, 235 correction, 296–297, 305, 309 Autonomous nervous system, 463–465 Autoradiography, 247–259, 317, 318, 327, 330, 331, 333–340, 342 Avalanche photodiode (APD), 233, 299, 301–303 Avidin, 171, 172, 221 B BACE-deficient mice, 507 BAL. See Broncho-alveolar lavage Barbiturate, 465 anesthetics, 434 Barium sulfate, 142 BBB. See Blood-brain barrier B cell, 566, 573, 576 Benzodiazepine receptors, 502 Beta emitters, 232 Beta-galactosidase, 418 Binding affinity, 247–253, 259–261 Binding potential, 518, 519, 522, 523, 526–529 Bioluminescence, 438, 442, 443, 445 Bioluminescence imaging (BLI), 420, 432, 475, 478–481, 484–486, 488, 490, 491, 493, 494 Biomarkers, 17, 18, 23–26 Biopsies, 208, 214 Biosafety, 29–33, 37, 41, 42, 44, 45 Biotin, 171, 172, 184, 221, 227 BLI. See Bioluminescence imaging Block detectors, 233 “Blockface” photographs, 318, 327, 330–332, 336, 340 Blocking, 74, 76, 78, 79 Blood-brain barrier (BBB), 477, 478, 500, 510, 524, 525 Blood flow, 210, 212–214, 222–226 Blood oxygenation, 430

F. Kiessling and B.J. Pichler (eds.), Small Animal Imaging, DOI: 10.1007/978-3-642-12945-2, © Springer-Verlag Berlin Heidelberg 2011

589

590 Blood oxygenation level dependent (BOLD), 430 Blood pool agents, 169–171, 176 Blood pool imaging, 220 Blood sampling, 356–357, 586–587 Blood volume, 222–226, 585 B-mode, 208 Body surface area (BSA), 369 Body temperature, 84, 86, 434–435 Bone densitometry, 125–127 Bonferroni correction, 377 Brain and liver, 196–197 Brain development, 423, 427, 432 Brain tumour, 240, 473, 476, 478, 479, 483, 484, 490, 491, 493 models, 474, 495 Breast cancer, 442 Broncho-alveolar lavage (BAL), 49, 50, 52–53 Bronchoconstriction, 54, 55 BSA. See Body surface area BTAs. See 2-[4¢-(methylamino)phenyl]benzothiazoles Budgeting, 37, 44, 45 Buprenorphine, 90 C 11 C, 502–505, 508 11 C-acetate, 238, 242, 463 Caliper measurement, 442 Cancer, 207, 208, 210–212, 214 Cardiac chamber, 426, 429 Cardiovascular, 207, 208, 214–216 Carprofen, 89 Caspase, 465, 479–480 “Cast” model, 452, 458 Catheter, 100, 107–111, 113, 114 CCD. See Charge coupled devices Cell-based therapy, 485–493 Cell tracking, 421, 430, 466–467 Cell trafficking, 567, 569, 572–575, 580 Cellular labeling, 175, 188 Cellular MRI, 491 CE MRI. See contrast-enhanced (CE) MRI Central nervous system (CNS), 421, 423, 428, 499–510 tracers, 500–502, 510 Cerebral ischaemia, 21 Challenge paradigm, 523, 529 Charge coupled devices (CCD), 271–277 11 C-HED-PET, 465 Chemical exchange saturation transfer (CEST), 165, 166, 176–182, 184 Chemical structures, 283, 286, 287 Chemoradiotherapy, 493 Chemotherapy, 224, 225 Chicken, 418 Cholestasis, 24 Chronic obstructive pulmonary disease (COPD), 19, 22, 49 Chronic venous cannulation, 108 13 C-hyperpolarized pyruvate, 187 c-Met, 484 11 C-MET, 240 CNS. See Central nervous system Coil, 152, 154, 158, 159, 161 Colitis disease model, 577

Index Collimators, 232–234, 236 Colloidal solutions, 143 Combinatorial chemistry, 248–249 Compartmental modelling, 390, 477 Computed tomography (CT), 125–138, 364–372, 374, 375, 418, 420–422, 426, 429, 432–433, 453 Computer infrastructure, 30, 38–39, 45 Contrast agent, 165–190, 207, 208, 210, 214 Contrast enhanced (CE) MRI, 421, 423, 430, 431 COPD. See Chronic obstructive pulmonary disease Co-registration, 579 Coronary artery, 450, 461 Coronary microcirculation, 456 Corpus luteum, 438 [11C]PIB, 507–509 Cross-correlation, 336, 340 CT. See Computed tomography mCT. See Micro-computed tomography CTDI. See CT dose index CT dose index (CTDI), 132 CTL. See Cytotoxic T lymphocytes Cu-ATSM. See Cu-diacetyl-bis(N4-methylthiosemicarbazone) Cu-diacetyl-bis(N4-methylthiosemicarbazone) (Cu-ATSM), 495 Cytokine, 566, 567, 569 Cytotoxic T lymphocytes (CTLs), 566, 567, 572 D Data analysis, 363–378 Data documentation, 405–413 Data storage, 406–411 Data structure, 79–80 DCE-MRI. See Dynamic contrast-enhanced magnetic resonance imaging DCIS. See Ductal carcinoma in situ Depth of interaction (DOI), 234, 235 Desflurane, 86–88 Detection efficiency, 233–234 Developmental imaging, 417–435 DICOM. See Digital imaging and communications Differential uptake ratio (DUR), 369 Diffusion, 160, 162, 163 Diffusion tensor imaging (DTI), 421, 423, 428, 432, 478 Diffusion weighted imaging (DWI), 477, 478 Digital imaging and communications (DICOM), 364, 405, 408–413 header, 407, 409, 410, 412 Digital subtraction angiography (DSA), 126 Dilative cardiomyopathy, 452 Dimeric non-ionic contrast media, 142, 143 Distribution volume (DV), 527–529 Distribution volume ratio (DVR), 527, 528 DNA synthesis, 476, 477, 486 DNP. See Dynamic nuclear polarization DOI. See Depth of interaction Dopamine depletion, 517, 519, 523 Dopamine enhancement, 517, 518, 520 Doppler, 208–210, 212–216, 421, 422, 424, 425, 429, 452, 455, 456 US imaging, 223, 438–439, 441, 445 Dorsal metatarsal vein, 106–108

591

Index Dose, 126, 129, 131–134, 137 D2 radiotracer binding, 516–517, 520–523, 526, 533 3D reconstruction of a post-mortem volume, 326, 327, 329 3D registration, 321, 338–339 Drop-out, 74–78, 80 Drug development, 212, 214 Drug discovery and development, 17, 18, 27 Drug distribution, 22 DSA. See Digital subtraction angiography DTI. See Diffusion tensor imaging Dual energy CT, 125, 136 Ductal carcinoma in situ (DCIS), 442 DUR. See Differential uptake ratio DV. See Distribution volume DVR. See Distribution volume ratio DWI. See Diffusion weighted imaging Dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI), 439, 441, 442, 445, 478, 552 Dynamic micro-CT, 134–135 Dynamic nuclear polarization (DNP), 184, 186–188 Dysprosium EOB DTPA, 142, 144, 145 E EAE. See Experimental autoimmune encephalomyelitis EBV. See Epstein-barr virus ECF-CM. See Extracellular fluid contrast media Echocardiography (ECG), 435, 450, 453–456, 458, 462, 464 Edema, 49–52 Efficacy, 17, 18, 20, 23 EGFR. See Epidermal growth factor receptor Elastase, 53–54 Elastic transformation, 340 Embryogenesis, 418, 432 Embryonic development, 418, 420, 427, 429, 432 Embryonic heart, 423, 424, 427, 429, 430, 432 Emphysema, 49, 53–54 Endoglin, 224, 225 Endometriosis, 438, 442–445 Endometriotic lesion, 437, 443–446 Endpoint specification, 72–73 Enteral administration, 96–100 Epidermal growth factor receptor (EGFR), 481–484, 489, 490, 494 kinase inhibitors, 483 Epi-illumination multi-fluorescence microscopy, 440, 443 Epstein-Barr virus (EBV), 568, 572 Ergonomic flow, 41–42 E-selectin, 172 Experimental autoimmune encephalomyelitis (EAE), 570, 577 Extracellular fluid contrast media (ECF-CM), 141, 142, 145, 149 Extra-striatal dopamine, 526 F 19 F, 166, 188–190 FBP. See Filtered back projection Fenestra VC, 142, 144, 147 Fentanyl, 90 Ferritin, 430–431, 476 FET. See Fluoroethyl tyrosine Fetal development, 417, 420, 427–430, 434, 435 Fetal identification, 435

F-FAC (1-(2¢-deoxy-2¢-18F-fluoroarabino-furanosyl) cytosine), 577, 578, 580 18 F-FDG, 237–239, 461–464, 467, 476, 477, 479, 486, 487, 489 576–578 [18F]FHBG, 476, 486, 487, 489, 491–494 18 F-FLT, 239 18 F-fluoride, 238, 240–241 [18F]fluoroethylcholine, 241 9-(4-[18F]fluoro-3-hydroxymethylbutyl)-guanine, 476 18 F-FMISO, 240 18 F-galacto-RGD, 465 Fibre tractography, 478 Fibrin targeted nanoparticles, 188 Fiducial markers, 330, 333–335 Fiducial registration error (FRE), 372 Filtered back projection (FBP), 234 Firefly luciferase (Fluc), 476, 479, 480, 484, 488, 491, 495 Fish, 417–418 Fixation by perfusion, 324–326 18 F-labelled analogues of misonidazole (MISO), 495 Flat-detector, 128 FLT. See Fluoro-thymidine Fluc. See Firefly luciferase Fluorescein-isothiocyanate (FITC)-labeled dextran, 439, 440, 443 Fluorescence mediated tomography (FMT), 364, 366–368, 370–373 Fluorescence reflectance imaging (FRI), 366–368 Fluorescent dyes, 281–284, 286, 287 Fluorescent particles, 286, 287 Fluorescent proteins, 432, 476, 480, 481, 484, 492, 567–568, 573 reporters, 418 2¢-fluoro-2¢deoxy-1b-d-arabionofuranosyl-5-ethyl-uracil, FEAU, 485, 495 Fluoroethyl tyrosine (FET), 477 [18F]Fluoro-fallypride, 526 9-(4-fluoro-3-hydroxymethylbutyl) guanine (FHBG), 485 Fluoromethyltyrosine (FMT), 477 Fluoro-thymidine (FLT), 477 FMT. See Fluorescence mediated tomography Follicle stimulating hormone (FSH), 438 Fractional uptake rate (FUR), 369 FRE. See Fiducial registration error Frequency-encoding, 155–157 FRI. See Fluorescence reflectance imaging Frog, 417–418, 430 FSH. See Follicle stimulating hormone Full-width-half-max (FWHM), 365, 369, 375 FUR. See Fractional uptake rate FWHM. See Full-width-half-max 18

G Gadolinium chelate, 442 Gadolinium-diethylenetriaminepentaacetate (Gd-DTPA), 458, 459 Gd-BOPTA (Multihance), 166, 167, 170 Gd-DOTA (Dotarem), 166, 167, 170 Gd-EOB-DTPA (Eovist), 166, 167 Gd-HPDO3A (Prohance), 166, 167 Gd-DTPA (Magnevist), 166, 167, 172 Gd-DTPA-BMA (Omniscan), 166, 167 68 Ga-DOTATOC, 241–242 b-Galactosidase, 476

592 Gamma counters, 353–357 G-APD, 302 Gastric emptying, 24 Gastro-intestinal contrast media, 141, 142 Gated SPECT, 461 Gating, 423, 428, 429, 435 Gaussian (‘normal’) distribution, 73 Gaussia princeps luciferase (GLuc), 569, 570 Gd, 166–174, 178, 181, 189, 190 Geiger-mode diodes, 302 Gene, 474–495 reporter imaging, 466 therapy, 478, 481, 485–488, 491 transfer, 476, 486 Genotyping, 431, 435 Glioma, 474, 477–488, 490, 491, 493–495 GLuc. See Gaussia princeps luciferase Glucose metabolism, 239 Gluteus muscle, 103 GLUT1 transmembrane transporter, 496 Gold nanoparticles, 142, 145, 146 Gonadotropin, 438, 441 Gradient, 152 coil, 155, 156, 158–160 Graft-versus-host disease (GVHD), 568, 570, 577 Growth factor, 478, 479, 491–494 Gynecologic cancer, 438, 442 Gynecologic oncology, 438, 442 Gynecology, 437–446 Gyromagnetic ratio, 183, 187 H Halogens, 253–256 Halothane, 86–88, 90 Harderian glands, 385, 521, 535 Harmonic imaging, 221–222 HE. See Hematoxylin-and-eosin-staining He-3, 184 Heart, 207, 208, 214, 215 Heart failure, 449, 452, 455, 459, 460, 462–463, 466 Helper T cell, 566 Hematopoietic stem cell, 568, 573 Hematoxylin-and-eosin-staining (HE), 341, 342 Hemoglobin, 174, 184 Hepatic lesion, 223 Hepatocellular carcinoma, 208 Hepatocellular or tissue-specific CM, 141, 142, 144–145, 148, 149 Hepatocyte growth factor (HGF), 484 HER-2, 172 Herpes simplex virus type 1 thymidine kinase enzyme (HSV-1TK), 466, 476, 486, 488, 491, 495, 572–574, 576 Hexokinase, 476 HGF. See Hepatocyte growth factor HIF-1a, 576 High-grade glioma models, 477 High throughput, 159 Hind limb model, 452 Hollow teflon rods, 340 Hounsfield units (HU), 366 H215O-water, 463

Index HSA. See Human serum albumin HSV-1-TK. See Herpes simplex virus type 1 thymidine kinase enzyme HU. See Hounsfield units Human serum albumin (HSA), 169–171, 174 Huntington’s disease, 502 Hyperpolarised 3He, 22 Hyperpolarization techniques, 183–187 Hyperpolarized molecules, 166, 182–188 Hyperthermia, 176 Hypothermia, 84–85 Hypoxia, 238, 240, 245, 479, 481, 482, 495 Hypoxia-inducible transcription factor (HIF) HIF-1, 481, 482, 495 I IACUC. See Institutional Animal Care and Use Committee ICAM-1, 224, 457 IFN-gamma, 566 * I-gluco-RGD, 465 5-[124I]iodo-29-fluoro-1-b-d-arabinofuranosyluracil, [124I]FIAU, 476 Image fusion, 293–312 Image quality, 364, 368, 370 Image reconstruction, 233–235 Image registration, 321, 323, 324, 327, 333, 334, 338, 371–372 Image resizing, 370–371 Imaging, 29–45 data handling, 405–407, 410, 411 modalities, 4–9, 11–13 123 I-MIBG, 243 123 I-MIBG-SPECT, 465 Imidazole family, 495 Immune cells, 565–570, 576–580 Immune system, 565–580 Immunology research, 565–580 Immunotherapy, 566, 571, 573–575, 578, 580 Individually ventilated cages (IVC), 61, 63, 65–67 Inflammation, 224, 226–227 Infrastructure requirements, 30, 37–41 Inhalant anesthetics, 434 Inhalation, 100, 101, 113 Injected activity, 384–386 Injection, 93–95, 99–114, 208, 211, 212, 214 111 In-oxine, 467 Institutional Animal Care and Use Committee (IACUC), 64 Integrin (avb3), 171, 172 Integrine, 457 “Intensity diffusion,” 365, 374 Inter-slice intensity normalization, 336–337 Intradermal (intracutaneous) injection, 101, 588 Intragastric administration, 98–100 Intramuscular injection, 101–103, 588 Intraperitoneal injection, 94, 100, 104–105, 588 Intravenous injection, 100, 101, 104–111, 588 In vitro evaluation, 260–263 In vitro transduction, 568 In vivo evaluation, 263–265 In vivo/post mortem registration, 339–342 111 In zevalin, 244–245

Index Iodinated contrast media, 142–143 Iodinated triglyceride emulsion (IT-LE), 144, 148 Iodine-containing lipid, 142, 144 Iodine-containing micelles, 142, 145 5-iodo-1-(2-deoxy-2-fluoro-b-d-arabinofuranosyl) uracil (FIAU), 485 Iron oxide particles, 165, 166, 174–176, 430 Isoflurane, 86–88, 90, 434 ISO surfaces, 374 Iterative reconstruction, 234 IT-LE. See Iodinated triglyceride emulsion IVC. See Individually ventilated cages K Ketamine, 86, 88–91, 434 Kidney perfusion, 222, 226, 245 Kinetic modeling, 523–529 Kolmogorov-Smirnov test, 376–377 Kruskal-Wallis test, 377 L Labeling chemistry, 281, 287–290 Labeling strategies, 247, 253–260 Laboratory protocols, 290–291 Lab protocol, 445–446 Lactate, 187 Lanthanides, 176, 179, 181 Larmor equation, 154, 155 Larmor frequency, 154–156 Laryngoscope, 113 Lateral marginal vein, 106–107 LBM. See Lean body mass Lean body mass (LBM), 369 Lentiviral, 568, 569 Leukocyte, 566, 569, 578 LH. See Luteinizing hormone Lillifors correction, 377 Linear correlation, 377–378 Linear regression, 378 Line-of-response, 232, 234, 235 Lipidic nanoparticles, 171 LipoCEST, 180–182 Lipophilicity (log P), 500–501, 525 Liposomal encapsulation of iodinated contrast media, 142 Liposomes, 142–144, 146–148, 171, 174–176, 180–182 Liver metastases, 212 Local anesthetics, 90 Logan, 528 Luciferase, 443–445 Luciferin, 445, 476, 478, 486, 488 Lung inflammation, 50–53 Lung oedema, 19 Luteinizing hormone (LH), 438 Lymph nodes, 210 Lymphoma, 239, 244, 245 M Macromolecular contrast media, 142 Macromolecular Gd(III)-complexes, 169, 173 Macromolecules, 143, 144 Macrophage, 566, 576, 578

593 Magnetic resonance imaging (MRI), 4, 6–9, 11, 13, 15–17, 19, 21–26, 49–55, 151–168, 171–173, 175–177, 183, 184, 186–190, 364, 366, 368–370, 418–421, 423–432, 450, 458–460, 462–464, 473–475, 477–479, 481, 483, 484, 486, 487, 491–494 coils, 203–204 contrast agents, 160, 165–168 sequences, 199–202 spectroscopy, 195–204 Magnetic resonance microscopy (MRM), 151, 152, 154, 156, 160–164 Magnetic resonance spectroscopic imaging (MRSI), 163 Magnetization, 151–161 Magnetoliposome, 176 Major Histocompatibility Complex class I (MHC-I) like molecules, 566 Make-test cycle, 17, 20–21 Manganese enhanced MRI (MEMRI), 421, 423, 428–429 Mann-Whitney test, 377 Mann-Whitney-U test, 377 Mann-Whitney-Wilcoxon test, 377 Mass doses, 529, 533–534 MCAo. See Middle cerebral artery occlusion mCT. See Micro-computed tomography MDR. See Multidrug resistance Measurement schedule, 70, 76, 77 Mechanical index (MI), 221–223, 225, 227, 228 Medetomidine, 90 Melanoma, 212 Meloxicame, 89, 90 MEMRI. See Manganese enhanced MRI MET. See Methionine Metabolic stability, 247, 249, 251–253, 263, 264 Metabolite analysis, 357 Metal complexe, 165, 166, 178 Metallo-proteinases (MMPs), 465 Metals, 256–260 Metatarsal vein, 106–108 Methionine (MET), 477 2-[4¢-(methylamino)phenyl]benzothiazoles (BTAs), 508 MHC-I like molecules. See Major Histocompatibility Complex class I (MHC-I) like molecules MI. See Mechanical index Microbubbles, 214, 220, 221, 226, 445 Micro-computed tomography (mCT), 126–130, 132, 134–137, 420, 432, 433 MikroCT, 453 Microdialysis, 520–523, 525, 527, 530, 533 Micro-focus tube, 125, 129 Microglia, 502, 505 Microinjection, 418, 423, 430 Micro-magnetic resonance imaging (mMRI), 420, 423 MicroPET, 233, 574–576, 578, 579 Midazolam, 86, 91 Middle cerebral artery occlusion (MCAo), 452 M-mode, 421, 422, 424, 429 MMPs. See Metallo-proteinases mMRI. See Micro-magnetic resonance imaging Mn, 166–174 MnDPDP, 167, 168 Model characterization, 21–23

594 Molecular imaging, 171–172, 176, 178, 183, 190 Molecular targeting, 457–460 Molecular ultrasound, 220, 226 Monomeric non-ionic contrast media, 142, 146 Monte Carlo methods, 132, 451–453 MPEG-iodolysine, 142, 145 MRI. See Magnetic resonance imaging MRM. See Magnetic resonance microscopy MRSI. See Magnetic resonance spectroscopic imaging MR-spectroscopy, 556, 559 MSA. See Multiple system atrophies 99m Tc-MAG3, 245 99m Tc-MIBI, 244 99m Tc-microspheres, 242–243 99m Tc-pertechnetate mTOR, 482, 484 Multi channel plates, 270–271 Multidrug resistance (MDR), 493–495 Multimodal imaging, 293–312 devices, 370 Multiparameter imaging, 308–309 Multiparity, 417 Multiple comparisons, 377 Multiple system atrophies (MSA), 507 Muscle and subcutaneous tumor, 197–199, 202 Myocardial blood flow, 241, 242, 456–457 Myocardial infarction, 215, 449–453, 457, 459, 460, 462, 464–467 Myocardial innervation, 463–465 Myocardial perfusion, 457, 459, 461–463 Myocardial sympathetic innervation, 243 N N-ammonia, 241 Nanoparticles, 142, 145 Natural killer (NK) cell, 566 Near infrared imaging, 272, 273 Neural progenitor cells (NPCs), 488, 491, 493 Neural stem cells (NSCs), 490, 491 Neurochemical competition, 516, 519, 520, 523 Neurodegenerative disorders, 499, 500, 502, 505–507, 510 Neuroendocrine tumour, 242 Neurooncology, 473–495 Neuroreceptor target, 501 Neurotransmitter activation, 517–520 NK cell. See Natural killer (NK) cell NMR. See Nuclear magnetic resonance Noble gas, 184 Non-DICOM image formats, 406, 410 Non-linear transformation, 318, 322, 324, 327, 328, 339 Nonparametric tests, 376 Non-targeted, 282–284, 287 NPCs. See Neural progenitor cells NSCs. See Neural stem cells NSP. See Nucleoside salvage pathway Nuclear imaging/optical, 299 Nuclear magnetic resonance (NMR), 165, 166, 168, 176, 177, 182–187 Nucleoside salvage pathway (NSP), 577, 578 Null hypothesis, 375–378

13

Index O OI. See Optical imaging Oil emulsions of ethiodol and iodine, 142, 145 Oncology, 210–211, 343 One-sided t-test, 78–79 One way analysis of variance (Anova), 377 Optical imaging (OI), 267–268, 275, 277, 417, 418, 458, 461, 467, 473, 474, 478, 481, 483, 486, 491, 493, 494 Optical probes, 281–291 Optical projection tomography (OPT), 420 Optical properties, 282, 286–287 Optical tomography, 276, 277 Oral administration, 96–98 Ordered subsets expectation maximization (OSEM), 234 Oro-endotracheal intubation, 113–114 Orthotopic, 208, 210 OSEM. See Ordered subsets expectation maximization Osmolality, 142, 143 Osmotic minipump, 111 Osteoarthritis, 23 Osteoporosis, 129 Outlier, 80–81 Ovarian cancer, 442 Ovary, 438–441, 445 P PACS. See Picture archiving and communication system Paired t-test, 376 ParaCEST, 178–180 Paraganglioma, 243 Para-hydrogen induced polarization (PHIP), 184–186 Parallax error, 235 Parallel imaging, 458 Paramagnetic, 160 blood pool contrast agents, 458 liposome, 458 Parametric test, 376 Parenteral administration, 94, 95, 100–101, 111–114 Parkinson’s disease (PD), 502, 507 Partial volume, 535 Partial volume effect (PVE), 235, 298, 364–365, 383–384 PAT. See Photoacoustic tomography PBS. See Phosphate buffered saline PD. See Parkinson’s disease Pellet, 97, 100, 111 Penis vein, 107–108 Pentobarbital, 89 Peptides, 247–265 Percent-injected dose per cubic-centimeters (%ID/cc), 522, 525, 528 Percutaneous absorption, 100 Perfluorinated gas, 220 Perfluorocarbon (PFC), 179, 188–190 Perfusion, 450–452, 456–457, 459, 461 Permeability, 438, 439, 441, 442, 445 Personnel, 30–32, 34–39, 42, 44, 45 PET. See Positron emission tomography PET and SPECT quantitation, 355–357 PET/CT, 295, 297, 299, 309 PET/MRI, 299–309, 579 PFC. See Perfluorocarbon

595

Index P-glycoprotein (P-gp), 493–494 pH, 95, 96, 103 Phaeochromocytoma, 243 Phage display libraries, 248, 249 Phantom, 131–134, 137 Pharmacokinetics, 247, 249, 250, 252–253, 256, 257, 259–262, 264 Pharmacology, 17, 18, 21–25 Phased array coil, 159 Phase-encoding, 156, 157 PHIP. See Para-hydrogen induced polarization Phosphate buffered saline (PBS), 94 Phosphatidylserine (PS), 480, 481 Phospholipid, 220, 221 Photoacoustic tomography (PAT), 277–278 Photo multiplier tubes (PMT), 233, 234, 270–271 Photon attenuation, 382–383 pH value, 556 Physiological saline, 94 Picture archiving and communication system (PACS), 407–411, 413 Pig, 451 Pinhole collimators, 232–233, 236 Pituitary hyperplasia, 49 Placenta, 441, 442, 445 Placental pathology, 438, 441–442 Placental perfusion, 441, 442, 445 Plaque, 449, 450, 452, 453, 456, 458, 461, 465 PMT. See Photo multiplier tubes Point spread function (PSF), 365, 375 Polycystic ovarian syndrome, 438 Polyethylene glycol (PEG) coating agents, 142, 144, 146, 147 Poly-l-lysine, 144, 145 Polymer, 220 Polymer-coated Bi2S3, 145 Polymer-stabilized bubble, 220 Positron emission tomography (PET), 4–9, 11, 13–17, 20, 22, 25, 26, 231–236, 364, 368–370, 373, 374, 453, 458, 461–467, 499–510 reporter gene imaging, 570–574, 576 tracers, 237–242, 245 Post-processing, 129–130, 136–137 Post-synaptic receptors, 516, 520, 523, 526 Power (1-b), 70–71, 78, 81 Power analysis, 376 Power Doppler, 210, 212, 213, 216 Power modulation, 221–222 Pregnancy, 419–423, 426, 429–435, 437, 438, 441, 442 Presenilin, 505 Probes, 4, 6–8, 11, 14, 293, 294, 299, 307, 309, 312 Prolactinoma, 49 Proliferation, 239 Prostate cancer, 211, 212, 241–243 Proton MRI, 54 PS. See Phosphatidylserine P-selectin-binding microbubbles, 226–227 P-selectine, 457 PSF. See Point spread function Pulse inversion, 221–222 PVE. See Partial volume effect

Q Quadriceps muscle, 102 Quail, 418, 419 Quantification, 225, 379–386 Quantitative image analysis, 379–380 Quantum dot-based nanoparticles, 171 Quantum dots, 476 R [11C]Raclopride, 516–523, 525–535 Radiation dose, 7, 8, 420–421, 429 Radiation use, 29–31, 35–38, 41, 45 Radio frequency (RF) pulse, 153–157, 159 Radioisotopes, 231 Radiometals, 253, 256–260, 262, 263 Radiopharmaceuticals, 247–265 Radiotracer, 247–265 Radiowave frequency (RF) coil, 153–155, 158, 159, 161 Randomization, 74, 76–79 Rank correlation, 377–378 Rational drug design, 248 Reactive dyes, 284–285, 287, 290 Reconstruction, 133–135 Redox potential, 172–174 Regional blood volume, 456–457 Region of interest (ROI), 365, 366, 369, 371–374 definition, 372–373 evaluation, 373 transfer, 373 Relaxation, 155, 158, 160, 162–163, 165–176, 183, 184, 188–190 Relaxometry, 366 REM. See Resource equation method Rendering, 130, 136–138 Reporter gene, 418, 421, 430–432, 435, 474, 476, 481, 484–486, 491, 494, 495, 567–580 Reporter transgenes, 473, 474, 476 Reporter transgenic animal, 569 Reproductive cycle, 417 RES. See Reticuloendothelial system Resolution, 5, 7, 9 Resonance frequency, 152, 154, 156, 159 Resource equation method (REM), 78, 79 Respiratory gating, 458, 461 Reticuloendothelial system (RES), 171, 175 Retroviral constructs, 568 RF coil. See Radiowave frequency coil RF pulse. See Radio frequency pulse RGD, 172 Rheumatoid arthritis, 570, 577 ROI. See Region of interest Rotate, scale, translate (RST) transformation, 371 3R principles, 47–56, 61 S Safety aspects, 491 Sample size estimation, 73, 78–79 Sample size formula, 78 Scanhead, 208, 209, 211 Scatter, 232, 235 Scintigraphic imaging, 461–463 Scintillator crystal, 231, 233, 234

596 Security, 30, 38, 39, 43–45 Sedation, 84, 88–91 Segmentation, 366, 372–375 Seldiger vessel cannulation technique, 114 Selectine, 457 Selectins, 224, 227 Sensitivity, 8, 11, 232–236 Sevoflurane, 86–88, 90 Signed rank test, 376 Significance level (the nominal a-level), 70–71, 78, 375–377 Simplified reference tissue model (SRTM), 518, 519, 527, 530 Single photon emission computed tomography (SPECT), 4, 6–9, 11, 17, 22, 23, 231–236, 364, 369–370, 373–375, 450, 453, 461–467 tracer, 242–245 Sinogram, 234 Site requirements, 30–32 Site-specific ultrasound contrast agents, 224–225 Skeletal development, 418, 429 Skeletal pathology, 240 Skeletogenesis, 420, 432–433 Slice selection, 156, 157 Sodium chloride, 94, 95, 100, 112 Somatostatin receptor, 476 Spatial resolution, 4–8, 232–236 Spearman’s rank correlation, 378 Specific activity, 235–236 Specific pathogen free (SPF), 66–67 Speckle, 215 SPECT. See Single photon emission computed tomography SPECT/CT, 295, 296 SPF. See Specific pathogen free Spill-over, 535 Spin-lattice relaxation, 154, 160 Spin-lattice relaxation time T1, 154–155, 160, 162–163 Spins, 151–157, 160, 161, 163 Spin-spin relaxation time T2, 154, 155, 160, 162–163 SPIO. See Superparamagnetic iron oxide particles SrB1-/-/ApoE-R61h/h mice, 452 SRTM. See Simplified reference tissue model Staff training, 38, 42–43 Standardized uptake value (SUV), 369, 370, 385, 386 Statistical evaluation, 71, 73, 75–77, 79, 80 Statistical tests, 374–378 Rank correlation, 377–378 Wilcoxon Rank Sum test, 377 Wilcoxon Signed Rank test, 376 Streptavidin, 221, 224, 226–227 Stroke, 449, 450, 452, 459 Student’s t-test, 376, 377 Study design, 79, 80 Subcutaneous injection, 94, 101, 103–105 Subcutaneous tumor, 210, 213, 214 Sublingual administration, 96 vein, 106, 108, 109 Superparamagnetic, 160 Superparamagnetic iron oxide (SPIO) particles, 174, 175, 481, 492 Surface coils, 159

Index Surgical anaesthesia, 85 Survivin knock-out mouse, 452 SUV. See Standardized uptake value T T1, 165–174, 178, 183, 185, 187 T2, 153–155, 160, 162, 163, 165, 174–176, 178, 183 TAC. See Thoracal aortic constriction Tachyarrhythmia, 449–451 Tail vein, 107, 109, 110 Target confidence, 17–20 Target distribution, 18 Target registration error (TRE), 372 Target validation, 18, 20 Tattooing, 435 Tau-protein, 505, 506 T cell, 189, 566–570, 572–578, 580 trafficking, 567, 569, 572–575 Technetium, 253, 257–258 T2/FLAIR, 479 Thermoluminescent dosimeters (TLD), 132 Thoracal aortic constriction (TAC), 452 Thyroid diseases, 244 Tight junction, 500 Tissue perfusion, 223 TLD. See Thermoluminescent dosimeters TNF-alpha, 566 Topical application, 100–101 Toxicology, 17, 20, 23–24 Tracer evaluation, 260–265 Tracer kinetic modeling, 525, 533, 534 Tracer optimization, 249–253, 260, 262 Tracking cells, 175 Training and education, 42–44 Transducer, 208, 209, 216 Transfection, 476 Transferrin, 476 Transformation methods, 311–312 Transgenic animal models, 506, 507, 510 Transmission CT, 4, 6–9, 11 Transmitter, 499, 505, 506 Transverse magnetization, 153, 154 TRE. See Target registration error T1 recovery, 153, 154 T1 relaxation, 153–155, 160, 163 T2-susceptibility agents, 174–176 Tumor, 162–163, 187, 223–225, 228, 566–568, 570, 572–578, 580 doubling time, 77 hypoxia, 13, 15, 16 perfusion, 210, 212, 214, 224, 226 size, 211 two compartment model, 552 U UCA. See Ultrasound contrast agents Ultrasmall superparamagnetic iron oxide (USPIO), 175, 478 Ultrasound (US), 4, 6–8, 11, 207–216, 438, 364, 368, 373, 441, 442, 418, 420, 422, 423, 426–428, 453–458 imaging, 438–439, 441, 442, 444–446 Ultrasound biomicroscopy (UBM), 420–423, 426–430

597

Index Ultrasound contrast agents (UCA), 219–228 US. See Ultrasound User education, 38, 42–44 USPIO. See Ultrasmall superparamagnetic iron oxide Uterine leiomyoma, 442 U-test, 376–377

Ventricular anatomy and function, 453–455, 458–459 Vessel density, 224, 225 av-integrin, 445 Virotherapy, 491 Viscosity, 143, 145 Volume coils, 159

V Vaccine, 566 VAPs. See Vascular access ports Variability, 70, 71, 73, 76, 78, 79, 81 Vascular access ports (VAPs), 100, 108 Vascular anatomy, 453, 455–456, 458 Vascular cell adhesion molecule (VCAM), 457 Vascular dyes, 283–284 Vascular endothelial (VE) cadherin, 43 Vascular endothelial growth factor (VEGF), 465 Vascular imaging, 223, 224 Vascular inflammation, 465 Vascular space contrast media, 141–147, 149 avb3 integrin, 224, 225, 457, 465, 569 VCAM. See Vascular cell adhesion molecule VCAM-1, 224 VE cadherin. See Vascular endothelial (VE) cadherin VEGF. See Vascular endothelial growth factor VEGF-R2, 213, 214, 224, 225

W Water solubility, 141–143, 145, 148 Whole-body autoradiography, 348–350, 352 Wilcoxon-Mann-Whitney test, 377 Wilcoxon Rank Sum test, 377 Wilcoxon Signed Rank test, 376 Wire-induced endothelial disruption, 452 X Xe-129, 184 Xenograft, 437, 441, 442 tumor, 162, 474 Xenopus, 418 X-ray, 17, 24, 25, 125–138 doses, 132, 296 Xylazine, 89, 90, 434 Z Zebrafish, 214, 417–419