Vascular Development (Novartis Foundation Symposium 283)

VASCULAR DEVELOPMENT The Novartis Foundation is an international scientific and educational charity (UK Registered Ch...

Author: Novartis Foundation

42 downloads 718 Views 4MB 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

VASCULAR DEVELOPMENT

The Novartis Foundation is an international scientific and educational charity (UK Registered Charity No. 313574). Known until September 1997 as the Ciba Foundation, it was established in 1947 by the CIBA company of Basle, which merged with Sandoz in 1996, to form Novartis. The Foundation operates independently in London under English trust law. It was formally opened on 22 June 1949. The Foundation promotes the study and general knowledge of science and in particular encourages international co-operation in scientific research. To this end, it organizes internationally acclaimed meetings (typically eight symposia and allied open meetings and 15–20 discussion meetings each year) and publishes eight books per year featuring the presented papers and discussions from the symposia. Although primarily an operational rather than a grant-making foundation, it awards bursaries to young scientists to attend the symposia and afterwards work with one of the other participants. The Foundation’s headquarters at 41 Portland Place, London W1B 1BN, provide library facilities, open to graduates in science and allied disciplines. Media relations are fostered by regular press conferences and by articles prepared by the Foundation’s Science Writer in Residence. The Foundation offers accommodation and meeting facilities to visiting scientists and their societies. Information on all Foundation activities can be found at http://www.novartisfound.org.uk

Novartis Foundation Symposium 283

VASCULAR DEVELOPMENT

Copyright © Novartis Foundation 2007 Published in 2007 by John Wiley & Sons Ltd, The Atrium, Southern Gate, Chichester PO19 8SQ, UK National 01243 779777 International (+44) 1243 779777 e-mail (for orders and customer service enquiries): [email protected] Visit our Home Page on http://eu.wiley.com All Rights Reserved. No part of this book may be reproduced, stored in a retrieval system or transmitted in any form or by any means, electronic, mechanical, photocopying, recording, scanning or otherwise, except under the terms of the Copyright, Designs and Patents Act 1988 or under the terms of a licence issued by the Copyright Licensing Agency Ltd, 90 Tottenham Court Road, London W1T 4LP, UK, without the permission in writing of the Publisher. Requests to the Publisher should be addressed to the Permissions Department, John Wiley & Sons Ltd, The Atrium, Southern Gate, Chichester, West Sussex PO19 8SQ, England, or emailed to [email protected], or faxed to (+44) 1243 770620. This publication is designed to provide accurate and authoritative information in regard to the subject matter covered. It is sold on the understanding that the Publisher is not engaged in rendering professional services. If professional advice or other expert assistance is required, the services of a competent professional should be sought. Other Wiley Editorial Offices John Wiley & Sons Inc., 111 River Street, Hoboken, NJ 07030, USA Jossey-Bass, 989 Market Street, San Francisco, CA 94103-1741, USA Wiley-VCH Verlag GmbH, Boschstr. 12, D-69469 Weinheim, Germany John Wiley & Sons Australia Ltd, 33 Park Road, Milton, Queensland 4064, Australia John Wiley & Sons (Asia) Pte Ltd, 2 Clementi Loop #02-01, Jin Xing Distripark, Singapore 129809 John Wiley & Sons Canada Ltd, 6045 Freemont Blvd, Mississauga, Ontario L5R 4J3, Canada Wiley also publishes its books in a variety of electronic formats. Some content that appears in print may not be available in electronic books. Novartis Foundation Symposium 283 x + 250 pages, 46 figures, 3 tables

Anniversary Logo Design: Richard J. Pacifico British Library Cataloguing in Publication Data A catalogue record for this book is available from the British Library ISBN 978-0-470-03428-6 Typeset in 10½ on 12½ pt Garamond by SNP Best-set Typesetter Ltd., Hong Kong. Printed and bound in Great Britain by T. J. International Ltd, Padstow, Cornwall. This book is printed on acid-free paper responsibly manufactured from sustainable forestry, in which at least two trees are planted for each one used for paper production.

Contents

Symposium on Vascular development, held at the Novartis Foundation, London, 13–15 June 2006 Editors: Derek J. Chadwick (Organizer) and Jamie Goode This symposium is based on a proposal made by Eckhard Lammert and Adam Wilkins Christer Betsholtz Chair’s introduction

1

Maria Grazia Lampugnani and Elisabetta Dejana The control of endothelial cell functions by adherens junctions 4 Discussion 13 Maike Schmidt, Ann De Mazière, Tanya Smyczek, Alane Gray, Leon Parker, Ellen Filvaroff, Dorothy French, Suzanne van Dijk, Judith Klumperman and Weilan Ye The role of Eg fl7 in vascular morphogenesis 18 Discussion 28 Max Levin, Andrew J. Ewald, Martin McMahon, Zena Werb and Keith Mostov A model of intussusceptive angiogenesis 37 Discussion 42 Tomáš Kucˇera, Jan Eglinger, Boris Strilic´ and Eckhard Lammert Vascular lumen formation from a cell biological perspective 46 Discussion 56 Christopher J. Drake, Paul A. Fleming and W. Scott Argraves The genetics of vasculogenesis 61 Discussion 71 Steven Suchting, Catarina Freitas, Ferdinand le Noble, Rui Benedito, Christiane Bréant, Antonio Duarte and Anne Eichmann Negative regulators of vessel patterning 77 Discussion 80 v

vi

CONTENTS

Taija Mäkinen and Kari Alitalo Lymphangiogenesis in development and disease 87 Discussion 98 Irene Noguera-Troise, Christopher Daly, Nicholas J. Papadopoulos, Sandra Coetzee, Pat Boland, Nicholas W. Gale, Hsin Chieh Lin, George D. Yancopoulos and Gavin Thurston Blockade of Dll4 inhibits tumour growth by promoting non-productive angiogenesis 106 Discussion 121 Georg Breier, Alexander H. Licht, Anke Nicolaus, Anne Klotzsche, Ben Wielockx and Zuzana Kirsnerova HIF in vascular development and tumour angiogenesis 126 Discussion 133 Karina Yaniv, Sumio Isogai, Daniel Castranova, Louis Dye, Jiro Hitomi and Brant M. Weinstein Imaging the developing lymphatic system using the zebrafish 139 Discussion 148 Frances High and Jonathan A. Epstein Signalling pathways regulating cardiac neural crest migration and differentiation 152 Discussion 161 Ralf H. Adams Investigation of the angiogenic programme with tissue-specific and inducible genetic approaches in mice 165 Discussion 171 Gary K. Owens Molecular control of vascular smooth muscle cell differentiation and phenotypic plasticity 174 Discussion 191 Andrea Lundkvist, Sunyoung Lee, Luisa Iruela-Arispe, Christer Betsholtz and Holger Gerhardt Growth factor gradients in vascular patterning 194 Discussion 201 Deborah A. Freedman, Yasushige Kashima and Kenneth S. Zaret Endothelial cell promotion of early liver and pancreas development 207 Discussion 216

CONTENTS

vii

Jörg Wilting, Kerstin Buttler, Jochen Rössler, Susanne Norgall, Lothar Schweigerer, Herbert A. Weich and Maria Papoutsi Embryonic development and malformation of lymphatic vessels 220 Discussion 227 Joaquim Miguel Vieira, Quenten Schwarz and Christiana Ruhrberg Role of the neuropilin ligands VEGF164 and SEMA3A in neuronal and vascular patterning in the mouse 230 Discussion 235 Final discussion 238 Tracheal tube development in Drosophila Closing remarks 240 Contributor index 242 Subject index 244

238

Participants

Ralf H. Adams Vascular Development Laboratory, Cancer Research UK London Research Institute, 44 Lincoln’s Inn Fields, London WC2A 3PX, UK Hellmut G. Augustin Department of Vascular Biology & Angiogenesis Research, Tumour Biology Center, Breisacher Str 117, Freiburg, D-79106, Germany Christer Betsholtz (Chair) Laboratory of Vascular Biology, Division of Matrix Biology, House A3, Plan 4, Department of Medical Biochemistry and Biophysics, Scheeles vag 2, Karolinska Institutet, SE-171 77 Stockholm, Sweden Georg Breier Institute of Pathology, Technische Universität Dresden, Fetscherstr 74, 01307 Dresden, Germany Jamie Davies Genes & Development IDG, Anatomy Building, University of Edinburgh, Teviot Place, Edinburgh EH10 5HF, UK Elisabetta Dejana FIRC Institute of Molecular Oncology, Via Adamello 16, I-20139 Milan, Italy Christopher J. Drake Cardiovascular Developmental Biology Center, Department of Cell Biology, Medical University of South Carolina, 173 Ashley Avenue, Charleston, SC 29425, USA Anne Eichmann Institut National de la Santé et de la Recherche Medicale U833, Collège de France, 11 Place Marcelin Berthelot, 75005 Paris, France Jonathan A. Epstein Department of Cell and Developmental Biology, Cardiovascular Institute, University of Pennsylvania, 954 BRB II, 421 Curie Boulevard, Philadelphia, PA 19104, USA Deborah A. Freedman Cell and Developmental Biology Program, Fox Chase Cancer Center, 333 Cottman Avenue, Philadelphia, PA 19111, USA viii

PARTICIPANTS

ix

Holger Gerhardt Vascular Biology Laboratory, Cancer Research UK London Research Institute, 44 Lincoln’s Inn Fields, London WC2A 3PX, UK Frances High University of Pennsylvania, BRB II/III Room 949, 421 Curie Boulevard, Philadelphia, PA 19104, USA Jan Kitajewski Department of Pathology, Irving Cancer Research Center 217B, Columbia University Medical Center, 1130 St Nicholas Avenue, New York, NY 10032, USA Eckhard Lammert Max Planck Institute of Molecular Cell Biology and Genetics, Pfotenhauerstraße 108, 01307 Dresden, Germany Keith E Mostov Department of Anatomy; Program in Cell Biology; and Department of Biochemistry and Biophysics, University of California School of Medicine, Genentech Hall, Room N212B, Mail Code 2140, 600 16th Street, San Francisco, CA 94158-2517, USA Taija Mäkinen Lymphatic Development Laboratory, Cancer Research UK London Research Institute, 44 Lincoln’s Inn Fields, London WC2A 3PX, UK Gary K. Owens Department of Molecular Physiology & Biological Physics, University of Virginia School of Medicine, 415 Lane Road, PO Box 801394, Room 1322 Medical Research Building 5, Charlottesville, VA 22908, USA Christiana Ruhrberg Institute of Ophthalmology, University College London, 11–43 Bath Street, London EC1V 9EL, UK Masabumi Shibuya Division of Genetics, Institute of Medical Science, University of Tokyo, 4-6-1 Shirokane-dai, Minato-ku, Tokyo, 108-8639, Japan Claire L. Shovlin BHF Cardiovascular Medicine Unit, National Heart and Lung Institute, Imperial College Faculty of Medicine, Hammersmith Hospital, Du Cane Road, London W12 0NN, UK Anne E. Uv Institute of Biomedicine, University of Göteborg, Medicinaregatan 9A, Box 440, 40530 Göteborg, Sweden Neil A. Vargesson Section of Molecular and Cellular Medicine, Division of Biomedical Sciences, Faculty of Natural Sciences, Imperial College London,

x

PARTICIPANTS

Workspace D, 1st Floor, Sir Alexander Fleming Bldg, South Kensington, London SW7 2AZ, UK Brant M. Weinstein Section on Vertebrate Organogenesis, Laboratory of Molecular Genetics, NICHD, National Institutes of Health, Building 6B, Room 309, 6 Center Drive, Bethesda, MD 20892, USA Adam Wilkins BioEssays, 10/11 Tredgold Lane, Napier Street, Cambridge CB1 1HN, UK Jörg Wilting Zentrum für Kinderheilkunde und Jugendmedizin, Pädiatrie I, Forschungslabor, Georg-August-Universität Göttingen, Robert-Koch-Strasse 40, 37075 Göttingen, Germany George D. Yancopoulos Regeneron Pharmaceuticals, 777 Old Saw Mill River Road, Tarrytown, NY 10591, USA Weilan Ye Tumor Biology and Angiogenesis Department, Genentech Inc., Mail Stop 230A, Building 15, Room 152015, 1 DNA Way, San Francisco, CA 94080, USA

Chair’s introduction Christer Betsholtz Laboratory of Vascular Biolog y, Division of Matrix Biolog y, Department of Medical Biochemistry and Biophysics, Scheeles vag 2, Karolinska Instituet, SE-171 77 Stockholm, Sweden

The overall theme of this symposium is vascular development, a subject that has progressed quite rapidly over recent years. It is also a broad subject: often it is described as encompassing just two steps, vasculogenesis and angiogenesis, but it is clear that this division might be somewhat artificial, and moreover, that multiple, complex phenomena are encompassed by these terms. Vasculogenesis involves the differentiation of vascular cells from non-vascular cells (mesodermal or mesenchymal), the assembly of these cells into vessels, and sometimes the incorporation of circulating precursors (bone marrow-derived endothelial progenitor cells) into the vasculature. Angiogenesis is even more complicated, involving a number of different morphogenetic phenomena in the vasculature, such as sprouting of new shoots, circumferential growth, the proliferation and enlargement of vessels, splitting of vessels (intussusception) and remodelling. This last term is often used in the literature as a description of the transition of an immature plexus to a refi ned hierarchically organized pattern of vasculature. Branch regression is also included in remodelling. The cellular and molecular processes collected under the simple terms ‘angiogenesis’, ‘vasculogenesis’ and ‘remodelling’ are extremely complicated and we therefore need to diversify our nomenclature in order to defi ne and focus our studies, and know what we are discussing. Just focusing on one of these processes, sprouting in angiogenesis, it becomes immediately clear that even this process is quite complicated. Hypoxia acting via hypoxia-responsive transcriptional machinery leads to the induction of signalling molecules that set off an angiogenic response. This response involves (1) the destabilization and degradation of vascular wall components (the matrix and the dissociation of mural cells that stabilize the vasculature), followed by (2) activation of the endothelial cells, their migration and proliferation to form a sprout, (3) the branching of the sprout and connection of the branches into a communicating and functional vessel network, and finally (4) the stabilization of the newly formed vessels by basement membrane deposition and maturation, and by the recruitment and incorporation of new mural cells. 1

2

BETSHOLTZ

We are beginning to appreciate that there is functional subspecialization among the vascular cells. This is relatively recent. An example is what occurs at the tip of the vascular sprout, where the leading cell, the tip cell, distinguishes itself morphologically and genetically from the trailing lumen-forming endothelial cells, the stalk cells. There are also mural cells: those associated with the sprouts and subsequent microvessels are referred to as pericytes. These cells are associated with the sprout from the onset of sprouting. Functional subspecialization among the vascular cells, and even within the endothelial compartment, is perhaps something that has been underappreciated so far. It may underlie the unequal cellular responses to the extracellular cues stimulating sprouting, as well as different types of intracellular signalling programmes in the different cell types. The cells are also connected to each other by cell adhesion molecules and junctional complexes; there is likely to be intense cell–cell communication via cell adhesion molecule-associated signalling mechanisms and juxtacrine or paracrine signalling. One of the emerging themes in vascular development is that the processes that are being uncovered at the growing vessel tips follow the same developmental principles that operate in other organogenesis processes. We have observed attractive signalling, repulsive signalling, cell–cell signalling leading to the establishment of cell boundaries, and we have noticed quite striking similarities to what has been observed for the development of the insect trachea. Remarkably, these processes of tubular sprouting are also similar to a process of subcellular sprout extension, namely neurite extension and guidance. Vascular development is interesting in its own regard, but one of the major driving forces for scientific progress in this area is the therapeutic promise. There are diseases which could benefit from the enhanced formation of new blood vessels if we learn how to stimulate their formation, such as infarction, fractures, thrombosis and even male baldness. There are other diseases where blood vessel formation is an intrinsic part of the pathological problem. Cancer is the most obvious example, but there is also clinical success with anti-angiogenic treatment in the area of ocular diseases. What is happening in the therapeutic area? Currently, there is just one validated target for anti-angiogenic therapies, and there are no clinically validated proangiogenic therapies. The anti-angiogenic target is vascular endothelial growth factor (VEGF), for which there are currently a few inhibitors in clinical use. Those of us who have been in the angiogenesis field for a while have, however, witnessed a long list of failures in human clinical trials of candidate anti-angiogenic molecules. Endostatin and angiostatin are well known examples, but there are many more that haven’t made it into the clinics. It’s reasonable to ask why molecules that act potently as anti-angiogenic factors in vitro or in animal models, fail in human clinical trials. One of the answers may be that the complexity of vascular development is enormous, and that we are still suffering from a quite profound

CHAIR’S INTRODUCTION

3

lack of insight into the molecular mechanisms of angiogenesis. For example we know very little about the molecular targets of some of the anti-angiogenic drugs that were entered into clinical trials some years ago. In this meeting we will discuss recent data on the guidance of endothelial cells in angiogenic sprouting and vessel remodelling. We will further discuss cell–cell communication at endothelial cell junctions, polarization of endothelial cells and lumenization of the vessel. Mural cells were in the shadows for a long time but they have recently been highlighted: a couple of papers will look at these. We have a second vascular system to consider—the lymphatics. I am glad that there are representatives of the field of lymphangiogenesis present at this symposium. We are also realizing that the developing vasculature is not just responding to surrounding cues, but also delivering developmentally important cues to the organs that it invades (in addition to oxygen and nutrients). Thus, vessels appear to fulfi l inductive functions during organogenesis. Reciprocal inductive signalling is a well known principle in epithelial–mesenchymal interactions, and it appears to also take place between blood vessels and surrounding cell types. Technology is important for studying the vasculature, and experimental genetics has become a major technology for advancing this field. Two participants at this symposium will discuss new developments that push the limits of mouse genetics. Finally, the symposium will have a translational component, as we will discuss vascular drug targets, new and old.

The control of endothelial cell functions by adherens junctions Maria Grazia Lampugnani* and Elisabetta Dejana*†1 IFOM, FIRC Institute of Molecular Oncolog y, * Mario Negri Institute for Pharmacological Research and † Department of Biomolecular Sciences and Biotechnologies, Faculty of Sciences, University of Milan, Milan, Italy

Abstract. Cell to cell junctions are important regulators of endothelial responses both in quiescent and angiogenic vessels. Endothelial cells express tight and adherens junctional structures. Although different in their specific molecular composition, these junctional complexes present a relatively similar general arrangement. Both types of junctions are formed by transmembrane adhesive proteins that bind homophilically to identical proteins on an adjacent cell and start a sequence of signalling events. Signal transmission is mediated by interaction with cytoplasmic and transmembrane partners. Adherens junctions are ubiquitous along the vascular tree. In these structures adhesion is mediated by VE-cadherin and its intracellular partners. In vitro and in vivo data show that VE-cadherin is required for endothelial integrity in quiescent vessels and for the correct organization of new vessels. VE-cadherin regulates endothelial functions through different mechanisms that include: (i) direct activation of signalling molecules such as PI3 kinase and Rac, to sustain survival and organization of the actin cytoskeleton; (ii) regulation of gene transcription, possibly modulating the nuclear level of transcription co-factors such as β -catenin and p120; (iii) formation of complexes with growth factor receptors, such as the type 2 receptor of VEGF (VEGFR2) and modulation of their signalling properties. 2007 Vascular development. Wiley, Chichester (Novartis Foundation Symposium 283) p 4–17

Cell–cell contacts control critical endothelial functions both in quiescent conditions, and in activated situations, such as inflammation and angiogenesis. Junctional proteins restrain cell migration, inhibit proliferation and apoptosis, and contribute to the maintenance of apical–basal polarity. In general, therefore, junctional signals should counteract angiogenesis and should be inhibited when vessels are induced to proliferate. Indeed, angiogenesis is accompanied by increased vessel permeability (Eliceiri et al 1999) and reduction of endothelial barrier function accompanies most inflammatory situations (Weis et al 2004a). 1

This paper was presented at the symposium by Elisabetta Dejana, to whom correspondence should be addressed. 4

ENDOTHELIAL CELL TO CELL JUNCTIONS AND VE-CADHERIN

5

A somehow intuitive consequence of these observations is that cell to cell junctions need to be weakened to allow the vessels to grow and be strengthened to maintain the endothelium in a quiescent state. An important aspect to define is the molecular mechanism through which this regulation takes place. Molecular architecture of endothelial junctions Various junctional structures have been identified at endothelial cell to cell contacts. Tight junctions, located in the most apical position toward the vessel lumen, exert a strong control on solute permeability and cell trafficking. These structures are present only in endothelia that need to maintain a strict control of permeability, such as brain microcirculation or large vessel endothelia. Adherens junctions on the other hand are ubiquitous along the vascular tree (Liebner et al 2006). Although the molecular composition of the different types of junctions varies, in general they are formed by both transmembrane and cytoplasmic components. At junctions, dimeric adhesive proteins bind to other identical dimers present on a nearby cell. The result is the lateral clustering of the adhesive molecules at cell to cell contacts, forming a zipper-like structure along the cell periphery (Dejana 2004). The recognition/adhesive information is delivered inside the cell by cytoplasmic and transmembrane partners. Junctions therefore behave as true signalling complexes (Liebner et al 2006). VE-cadherin and adherens junctions The major transmembrane component of endothelial adherens junctions is vascular endothelial (VE-) cadherin. This member of the cadherin family of adhesive receptors is expressed specifically in endothelial cells from the time of the earliest endothelial differentiation in the embryo. The extracellular domain of this singlepass transmembrane protein comprises five homologous repeats. The minimal functional unit of VE-cadherin is a homodimer that recognizes as a ligand, in a calcium-dependent way, an identical dimer present on a nearby cell. This starts a process of clustering of VE-cadherin molecules at the apposing cell membranes, which is sustained by both cis- and trans-interactions between cadherin molecules of the same cell and on the opposing cell, respectively (Gumbiner 2005). From the junction, signals are transmitted with the cooperation of interacting partners such as the intracellular molecules collectively known as catenins: α -, β -, γ -catenin/ plakoglobin and p120. β -catenin, plakoglobin and p120 bind strongly to defined sequences of the VE-cadherin cytoplasmic domain (Table 1), while α -catenin binds indirectly through the bridge of β - or γ -catenin. Other cytoplasmic and transmembrane proteins form complexes with VE-cadherin and contribute to the modulation of its adhesive and signalling activities (Table 1).

TABLE 1

VE-cadherin interactors 6

Transmembrane

Region of VE-cadherin involved in the binding

VEGFR2

n.d., β -catenin required

Dep1

n.d., β -catenin required

VE-PTP

membrane-proximal extracellular domain EC5

Cytoplasmic β -catenin

Induced

+ (in vivo)

+ (in vitro) +

↑

+

n.d.

↓

+

↑

+

↓

+

↑

Tiam Csk Src

n.d., β -catenin required phosphotyrosine 685 n.d.

+

SHP2

n.d., β -catenin required

+

Shc

cytoplasmic domain (aa703–784 in human VE-cadherin)

+

+

+

↑ ↑

Functional consequence Regulation of VEGFR2 internalization, mitogenic and antiapoptotic activities (1,2) endothelial barrier-function (3,4) Down-regulation of VEGFR2, Tyr-phosphorylation, internalization and mitogenic response (1,2) Reduced phosphorylation of VEGFR2 in response to VEGF, increased VE-cadherin-mediated barrier integrity, defective vasculogenesis (5,6) Association with VE-cadherin required to control vessel morphogenesis (7) Association with VE-cadherin contributes to junction stabilization (8) Involved in the maintenance of endothelial barrier function (9) Protection from apoptosis activated by VE-cadherin clustering (7) Regulation of actin and focal contact organization (10) Involved in density-dependent inhibition of cell growth (11) Mediates VEGF-induced VE-cadherin phosphorylation in vivo and in vitro (12) Down-regulation of catenin’s tyrosine phosphorylation. Target of thrombin (13) Possibly involved in the regulation of response to VEGF (14)

(1) Lampugnani et al 2003, (2) Lampugnani et al 2006, (3) Weis et al 2004a, (4) Weis et al 2004b, (5) Nawroth et al 2002, (6) Baumer et al 2006, (7) Carmeliet et al 1999, (8) Lampugnani et al 1995, (9) Iyer et al 2004, (10) Lampugnani et al 2002, (11) Baumeister et al 2005, (12) Lambeng et al 2005, (13) Ukropec et al 2000, (14) Zanetti et al 2002. Some of the reported interactions are direct and the molecular domains of VE–cadherin involved have been identified, while others remain to be determined (n.d., not determined). These interactions can be constitutive or induced by cell stimulation or clustering of VE–cadherin at cell to cell contacts as in confluent cultures. VE–cadherin partners may transfer intracellular signals and influence cell behaviour in a way dependent on cell confluency.

LAMPUGNANI & DEJANA

p85/PI3K

p120

↑

+

cytoplasmic domain (aa703–784 in human VE-cadherin) cytoplasmic domain (aa703–784 in human VE-cadherin) cytoplasmic domain (aa621–702 in human VE-cadherin) n.d., β -catenin required

γ -catenin

Modulated by cell confluence

Constitutive

ENDOTHELIAL CELL TO CELL JUNCTIONS AND VE-CADHERIN

7

VE-cadherin and endothelial functions Data collected in several in vitro and in vivo models indicate that VE-cadherin expression and clustering strongly modulate endothelial cell behaviour (Carmeliet et al 1999, Corada et al 1999, 2001, 2002, Liao et al 2002, Weis et al 2004a, Crosby et al 2005). The involvement of VE-cadherin in the organization of new vascular structures is clearly indicated by the effect of homozygous null mutation of the VE-cadherin gene. In the mouse embryo this mutation induces a lethal phenotype at around 9.5 dpc due to major defects in the organization of the vascular tree (Carmeliet et al 1999). Among the various examples that illustrate the requirement of VE-cadherin for the formation of a new endothelial network, we report in Fig. 1 an ex vivo model of organ culture. The still avascular allantois of an 8.0 dpc mouse embryo can be explanted and maintained in culture. After about 18 h endothelial cells differentiate and form a vascular network (Fig. 1 a–e) (Drake & Fleming 2000). If the allantois is explanted from embryos with homozygous null mutation of the VE-cadherin gene, endothelial cells differentiate (see PECAM1and Fli1-positive cells in Fig. 1) (Crosby et al 2005), but they fail to organize in a three-dimensional structure (Fig. 1 f–l). Beyond illustrating the biological role of VE-cadherin, this experimental model is also useful to screen for the antiangiogenic activity of antibodies to VE-cadherin (Crosby et al 2005). Conversely, the importance of VE-cadherin in preservation of vascular integrity is underlined by the study of the effects of a blocking monoclonal antibody, BV13, directed to the EC1 region of the extracellular domain of the protein (Corada et al 2002). In vivo administration of this antibody strongly increases vascular permeability in several organs of adult mice (Corada et al 1999). Consistently, a monoclonal antibody, BV9, directed to the extracellular domain of human VE-cadherin enhances the permeability of monolayers of human endothelial cells (Corada et al 2001). While these data support the concept that VE-cadherin is a key element in the control of endothelial integrity, they also imply that targeting VE-cadherin for therapeutic intervention may produce severe systemic effects on quiescent vasculature. However, another monoclonal antibody, E4G10, directed to a different epitope (May et al 2005) was found to target VE-cadherin specifically at angiogenic sites, leaving resting vessels unperturbed (Liao et al 2002). VE-cadherin is in a different conformational state in resting versus growing endothelia and E4G10 recognizes an epitope of VE-cadherin that is accessible only in activated angiogenic vessels, while it is masked in quiescent endothelia. VE-cadherin signalling pathways VE-cadherin may modulate endothelial behaviour through different pathways (see Table 1). We have described the association to VE-cadherin of the p85 subunit of

8

LAMPUGNANI & DEJANA

VE positive

a

8.0 dpc mouse embryo

VE null

f

b

d

c

e

g

i

h

l

FIG. 1. VE-cadherin null allantois explants fail to undergo normal vascular morphogenesis in culture. Allantoises (small panel on the left) explanted from 8.0 dpc embryos can be maintained in culture for 16–24 hours. In this time frame endothelial cells differentiate, (see PECAM1 [green fluorescence, not visible in this grey scale reproduction, but please see colour figure at http://www.novartisfound.org.uk/dejana_figure1.jpg] and Fli1 [red fluorescence] positive cells) and organize in a vascular like network (panel a). The allantois from a homozygous null VE-cadherin mutant embryo contains a comparable number of PECAM1 (green fluorescence) and Fli1 (red fluorescence) positive endothelial cells. However, the cells fail to form stable vascular structures and only form sparse aggregates (f ). Small panels on the right (b–e, wildtype, g-l, VE-null) show independent magnifications (b,c and g,h 20×, and d,e and i,l 40×, respectively) of the 10 × magnification fields on the left (a,f ).

phosphatidylinositol 3 (PI3) kinase in confluent cells. This results in activation of Akt and enhanced resistance to apoptosis (Carmeliet et al 1999). The C-terminal truncated VE-cadherin, unable to bind β -catenin, does not recruit p85 and does not protect cells from apoptosis. VE-cadherin, again only when clustered in stable junctions, induces recruitment at cell–cell contacts of Tiam, a Rac-specific guanosine exchange factor. The effect is activation of Rac, inhibition of Rho and reorganization of actin cytoskeleton and adhesion plaques (Lampugnani et al 2002). Tyrosine kinases and phosphatases have been reported to form complexes with VE-cadherin. Src kinase is constitutively associated to VE-cadherin and mediates VEGF-induced VE-cadherin phosphorylation (Lambeng et al 2005). This may contribute to the maintenance of VEGF-stimulated angiogenic processes.

ENDOTHELIAL CELL TO CELL JUNCTIONS AND VE-CADHERIN

9

C-terminal src kinase, csk, binds via its SH2 domain to the phosphorylated tyrosine 685 of VE-cadherin (Baumeister et al 2005). The functional consequence of such an interaction is inhibition of cell proliferation. Through its most membraneproximal extracellular domain, VE-cadherin associates with the receptor-typevascular endothelial protein tyrosine phosphatase (VE-PTP) (Nawroth et al 2002). This reduces tyrosine phosphorylation of VE-cadherin, reduces cell layer permeability, and most importantly, influences a correct vasculogenesis in the embryo (Baumer et al 2006, Nawroth et al 2002). The non-receptor tyrosine phosphatase SHP2 associates with VE-cadherin via β -catenin (Ukropec et al 2000). Thrombin induces phosphorylation of SHP2 and release of SHP2 from the complex. This is accompanied by tyrosine phosphorylation of the VE-cadherin-associated catenins ( β -, γ -, p120) and may be part of the mechanism through which thrombin increases endothelial permeability. The transmembrane phosphatase Dep-1 (density enhanced phosphatase 1) is recruited by VE-cadherin and modulates the responses to VEGFR-2 (see below, Lampugnani et al 2003). It was also found that VE-cadherin regulates gene transcription. This has been observed by comparison of the transcriptome of endothelial cells genetically ablated of VE-cadherin and that of the same cell line gene transduced to express VE-cadherin. 85 and 66 genes have been found up-regulated and down-regulated, respectively, by the presence of VE-cadherin clustered at cell to cell contacts. One reasonable mechanism through which VE-cadherin can exert its control on gene transcription is through sequestration at junctions of transcriptional co-factors, such as β -catenin and p120 (Park et al 2005). This would reduce the nuclear level of these factors and transcriptional regulation of their target genes. While this mechanism may be operating for genes presenting in the promoter regions at least one putative β -catenin/Tcf-Lef-1 binding domain, the precise molecular mechanism of such an effect is still unclear. In addition, the pathways through which VE-cadherin can control genes not under putative β -catenin control remain to be defined. VE-cadherin can also actively cross-talk with receptors for growth factors. This has been analysed in detail for VEGFR2, a classical modulator of endothelial function (Lampugnani et al 2003). When VE-cadherin is clustered at cell–cell contacts, as is the case in mature quiescent vessels and in confluent cultures, the two proteins can form a complex and, as a consequence, receptor signalling is re-directed from mitogenesis to survival (see below and Fig. 2). The complex between VE-cadherin and VEGFR2 observed in vivo is constitutive and transiently disrupted by treatment with VEGF (Weis et al 2004b). While in cell culture VEGF induces the formation of the complex (Zanetti et al 2002). A possible explanation of such discrepancy is the effect of blood flow as a modulator of endothelial functions. Although the molecular organization of the complex between VE-cadherin and VEGFR2 remains to be defined in detail, we have shown that β -catenin is required

10

LAMPUGNANI & DEJANA

Confluent VE-positive cells

** Akt

Survival

Sparse or VE-null cells

P42/44 MAP

kinases

Proliferation

FIG. 2. VE-cadherin clustering at cell to cell contacts directs VEGFR2 signalling. In confluent cells, as in mature quiescent vessels, VE-cadherin clustered at cell–cell contacts can form a complex with VEGFR2. The complex forms in response to VEGF stimulation which induces tyrosine phosphorylation of the receptor (represented as ** in the figure) (Shibuya & ClaessonWelsh 2006). β -catenin is indispensable for the complex to form (Lampugnani et al 2003) and the transmembrane tyrosine-phosphatase Dep1 is recruited in the complex. Association with VE-cadherin reduces VEGFR2 internalization and allows its dephosphorylation by Dep1 (Lampugnani et al 2003). The fi nal outcome is that the internalized receptor is poorly phosphorylated and transfers a weak mitogenic signal from intracellular compartments (Lampugnani et al 2006). In the absence of VE-cadherin, VEGFR2 is highly phosphorylated in response to VEGF (Lampugnani et al 2003), is efficiently internalized and can signal more effectively from intracellular compartments (Lampugnani et al 2006).

ENDOTHELIAL CELL TO CELL JUNCTIONS AND VE-CADHERIN

11

for its formation. The biochemical consequence of its interaction with VEcadherin is that VEGFR2 is less tyrosine-phophorylated following stimulation with VEGF. This results in reduced activation of MAPK and reduced proliferation (Lampugnani et al 2003). Somewhat surprisingly, under the same experimental conditions, the anti-apoptotic response to VEGF is enhanced (Carmeliet et al 1999). The inhibition of tyrosine phosphorylation of VEGFR2 is due to the recruitment of the complex of the transmembrane phosphatase density-enhanced phosphatase 1 (Dep1). This transmembrane phosphatase has been reported to preferentially dephosphorylate particular tyrosine residues of Met receptor (Palka et al 2003) and PDGFβ receptor (Kovalenko et al 2000). If a similar site selectivity also operates for VEGFR2 it may represent the basis of the decreased proliferation and enhanced anti-apoptosis signalled by VEGFR2 in confluent VE-cadherin positive cells (Lampugnani et al 2003). Dep1 could preferentially dephosphorylate the tyrosines involved in recruiting mitogenic mediators (such as PLCγ ) (Takahashi et al 2001), leaving intact the phosphotyrosines that sustain the antiapoptotic signal (Shibuya & Claesson-Welsh 2006). A further consequence of the formation of a complex between VE-cadherin and VEGFR2 is that the internalization of the activated receptor is inhibited. If VE-cadherin is not expressed or not clustered at cell–cell contacts VEGFR2 can enter intracellular compartments to a greater extent and in a highly phosphorylated form (Lampugnani et al 2006). In this active state, VEGFR2 can activate PLCγ, which can sustain mitogenesis from intracellular compartments. This process is strongly inhibited in VE-cadherin positive cells in which the receptor is retained more effectively at the membrane and, due to Dep1 activity, is actively dephosphorylated. As a consequence, the mitogenic signal from the internalized receptor is much weaker (Lampugnani et al 2006). The implications of these results are that: (i) signalling from VEGFR2 is under a junctional control; (ii) the formation of a complex between VE-cadherin and VEGFR2 limits its mitogenic signalling; (iii) internalization is a way for the receptor to escape inactivation and signal more effectively and for a longer time; and (iv) any system that modulates the recruitment of VEGFR2 to the complex with VE-cadherin may effectively modulate the activity of this receptor. Conclusions It appears that endothelial cell–cell junctions are the association sites of a complex array of transmembrane and cytoplasmic molecules. These molecules functionally interact and the result of such cross-talk is a fine tuning of the endothelial behaviour. Cell–cell junctions represent possible targets of therapeutic intervention since modulation of junction adhesion and signalling may strongly influence endothelial growth, apoptosis and permeability.

12

LAMPUGNANI & DEJANA

Acknowledgements This work was supported by the Associazione Italiana per la Ricerca sul Cancro, the European Community (QLRT-2001-02059, Integrated Project Contract No LSHG-CT-2004-503573; NoE MAIN 502935; NoE EVGN 503254).

References Baumeister U, Funke R, Ebnet K, Vorschmitt H, Koch S, Vestweber D 2005 Association of Csk to VE-cadherin and inhibition of cell proliferation. EMBO J 24:1686–1695 Baumer S, Keller L, Holtmann A et al 2006 Vascular endothelial cell-specific phosphotyrosine phosphatase (VE-PTP) activity is required for blood vessel development. Blood 107:4754– 4762 Carmeliet P, Lampugnani MG, Moons L et al 1999 Targeted deficiency or cytosolic truncation of the VE-cadherin gene in mice impairs VEGF-mediated endothelial survival and angiogenesis. Cell 98:147–157 Corada M, Mariotti M, Thurston G et al 1999 Vascular endothelial-cadherin is an important determinant of microvascular integrity in vivo. Proc Natl Acad Sci USA 96:9815–9820 Corada M, Liao F, Lindgren M et al 2001 Monoclonal antibodies directed to different regions of vascular endothelial cadherin extracellular domain affect adhesion and clustering of the protein and modulate endothelial permeability. Blood 97:1679–1684 Corada M, Zanetta L, Orsenigo F et al 2002 A monoclonal antibody to vascular endothelialcadherin inhibits tumor angiogenesis without side effects on endothelial permeability. Blood 100:905–911 Crosby CV, Fleming PA, Argraves WS et al 2005 VE-cadherin is not required for the formation of nascent blood vessels but acts to prevent their disassembly. Blood 105:2771–2776 Dejana E 2004 Endothelial cell-cell junctions: happy together. Nat Rev Mol Cell Biol 5:261–270 Drake CJ, Fleming PA 2000 Vasculogenesis in the day 6.5 to 9.5 mouse embryo. Blood 95:1671–1679 Eliceiri BP, Paul R, Schwartzberg PL, Hood JD, Leng J, Cheresh DA 1999 Selective requirement for Src kinases during VEGF-induced angiogenesis and vascular permeability. Mol Cell 4:915–924 Gumbiner BM 2005 Regulation of cadherin-mediated adhesion in morphogenesis. Nat Rev Mol Cell Biol 6:622–634 Kovalenko M, Denner K, Sandstrom J 2000 Site-selective dephosphorylation of the plateletderived growth factor beta-receptor by the receptor-like protein-tyrosine phosphatase DEP1. J Biol Chem 275:16219–16226 Lambeng N, Wallez Y, Rampon C et al 2005 Vascular endothelial-cadherin tyrosine phosphorylation in angiogenic and quiescent adult tissues. Circ Res 96:384–391 Lampugnani MG, Corada M, Caveda L et al 1995 The molecular organization of endothelial cell to cell junctions: differential association of plakoglobin, beta-catenin, and alpha-catenin with vascular endothelial cadherin (VE-cadherin). J Cell Biol 129:203–217 Lampugnani MG, Zanetti A, Breviario F et al 2002 VE-cadherin regulates endothelial actin activating Rac and increasing membrane association of Tiam. Mol Biol Cell 13:1175–1189 Lampugnani MG, Zanetti A, Corada M et al 2003 Contact inhibition of VEGF-induced proliferation requires vascular endothelial cadherin, beta-catenin, and the phosphatase DEP-1/CD148. J Cell Biol 161:793–804 Lampugnani MG, Orsenigo F, Gagliani MC, Tacchetti C, Dejana E 2006 Vascular endothelial cadherin controls VEGFR-2 internalization and signaling from intracellular compartments. J Cell Biol 174:593–604

ENDOTHELIAL CELL TO CELL JUNCTIONS AND VE-CADHERIN

13

Liao F, Doody JF, Overholser J et al 2002 Selective targeting of angiogenic tumor vasculature by vascular endothelial-cadherin antibody inhibits tumor growth without affecting vascular permeability. Cancer Res 62:2567–2575 Liebner S, Cavallaro U, Dejana E 2006 The multiple languages of endothelial cell-to-cell communication. Arterioscler Thromb Vasc Biol 26:1431–1438 May C, Doody JF, Abdullah R et al 2005 Identification of a transiently exposed VE-cadherin epitope that allows for specific targeting of an antibody to the tumor neovasculature. Blood 105:4337–4344 Nawroth R, Poell G, Ranft A et al 2002 VE-PTP and VE-cadherin ectodomains interact to facilitate regulation of phosphorylation and cell contacts. EMBO J 21:4885–4895 Palka HL, Park M, Tonks NK 2003 Hepatocyte growth factor receptor tyrosine kinase met is a substrate of the receptor protein-tyrosine phosphatase DEP-1. J Biol Chem 278:5728– 5735 Park JI, Kim SW, Lyons JP et al 2005 Kaiso/p120-catenin and TCF/beta-catenin complexes coordinately regulate canonical Wnt gene targets. Dev Cell 8:843–854 Shibuya M, Claesson-Welsh L 2006 Signal transduction by VEGF receptors in regulation of angiogenesis and lymphangiogenesis. Exp Cell Res 312:549–560 Takahashi T, Yamaguchi S, Chida K, Shibuya M 2001 A single autophosphorylation site on KDR/Flk-1 is essential for VEGF-A-dependent activation of PLC-gamma and DNA synthesis in vascular endothelial cells. EMBO J 20:2768–2778 Ukropec JA, Hollinger MK, Salva SM, Woolkalis MJ 2000 SHP2 association with VE-cadherin complexes in human endothelial cells is regulated by thrombin. J Biol Chem 275:5983– 5986 Weis S, Cui J, Barnes L, Cheresh D 2004a Endothelial barrier disruption by VEGF-mediated Src activity potentiates tumor cell extravasation and metastasis. J Cell Biol 167:223–229 Weis S, Shintani S, Weber A et al 2004b Src blockade stabilizes a Flk/cadherin complex, reducing edema and tissue injury following myocardial infarction. J Clin Invest 113:885–894 Zanetti A, Lampugnani MG, Balconi G et al 2002 Vascular endothelial growth factor induces SHC association with vascular endothelial cadherin: a potential feedback mechanism to control vascular endothelial growth factor receptor-2 signaling. Arterioscler Thromb Vasc Biol 22:617–622

DISCUSSION Lammert: Perhaps connecting the sprouts with the cadherins would be an interesting avenue: can we understand how stalk cells proliferate, weakening their junctions, and how this can be explained by VE-cadherin being upstream of VEGF receptor, for example? Betsholtz: That’s a good suggestion. It is an interesting possibility that the junctions are structurally different at the sprout sites. Dejana: One would expect that confluent cells remain relatively insensitive to VEGF activation. However, it is probably the local concentration of VEGF that makes the difference. There might be a concentration threshold, and cells would start to migrate once this is passed. VEGF can induce phosphorylation of VEcadherin and reduce the strength of adhesion at junctions. At the right concentration of VEGF, junctions would become weaker and endothelial cells would release themselves from inhibition of growth, and then start to migrate.

14

LAMPUGNANI & DEJANA

Davies: Most of the things you talked about were global responses of a cell to the loss of VE-cadherin expression, such as internalization of receptor and so forth. Do you think everything will be global, or do you have any evidence that part of a cell can develop differently to another part? For example, if one part of a cell has lost expression, it can go into its lamellipodium-forming mode and forge forward, whereas the other side of the cell knows that it is still connected and just needs to pull its neighbours with it. A cell at one position has neighbours on one side but not on the other. One part of the membrane would be responding differently. Dejana: Let’s imagine a wound in the vessel wall. The cells at the front are released from contact inhibition of growth, and would migrate much more effectively. Cells at the migrating front present part of their membrane free, i.e. not in contact with neighbouring cells. VEGF receptors, present in that area, maintain a high response to migration and proliferation. In contrast, receptors at the back where the cells are still in contact with other cells are less responsive. This phenomenon may dictate the direction of movement. Betsholtz: Jamie Davies, it sounds to me that you are describing the tip cell, which is partially detached at one side and reacts to VEGF with a migratory response. But it still remains attached to the vasculature at the trailing end of the cell. Davies: It could be the tip cell, or even, if this is true of other cadherins, such as N-cadherin, could be modulated by loss of connection with pericytes. Gerhardt: We have recently looked at the distribution of VE-cadherin in the growing sprout. In the retinal model, the tip cell seems to have weakened junctions at its contact point with the stalk cell. It also relocates a lot of the VE-cadherin to the sprouting front, to the tips of the fi lopodia (unpublished results). We know that the response to stimulation of VEGFR-2 also differs between the tip cells and the stalk cell. The question I find interesting in the in vivo situation is why the tip cell doesn’t respond with proliferation. If you inject VEGF into the retina, almost the entire vascular plexus starts to proliferate. All the cells grow except the tip cells. But these cells retain stable VE-cadherin contacts. If you want to make an analogy to cell culture, I would consider these cells as being confluent. They start to proliferate, not migrate. Dejana: The concentration of VE-cadherin at junctions doesn’t tell the whole story. The most important parameter is the state of phosphorylation of these proteins. When we add histamine or other permeability increasing agents, we don’t see cell retraction if we don’t use very high concentrations: we just see phosphorylation of VE-cadherin at the junctions. The overall organization of junctions may change depending on the functional state of the cells. Maybe junctions of the cells at the tip allow their migration but not proliferation. Drake: We have compared tip cells with angioblasts. What we have concluded is the following. Angioblasts do not express VE-cadherin. In the tip cell, VEcadherin is only expressed at the site where the tip cell and stalk cell share a

ENDOTHELIAL CELL TO CELL JUNCTIONS AND VE-CADHERIN

15

common cell-cell junction. Physically, tip cells appear to adhere tightly to the extracellular matrix. I think cell matrix adhesion is what is inhibiting VE-cadherin expression in both angioblasts and tip cells. When one moves back to the stalk cell, which expresses VE-cadherin, the cells are engaged in cell–cell versus cell– matrix interactions. Dejana: Our view is more concentrated on VE-cadherin. Certainly, interaction with the matrix is important. There are a series of signals from integrin to VEcadherin and vice versa. VEGFR2, can also interact with integrins, there might be a better interaction of this receptor with integrins while the cells are growing and migrating, while the receptor would interact with cadherins better when the cells are stabilized. Augustin: What is the relevance of your model in the context of sprouting angiogenesis? I think we would all agree that the tip cells have altered junctional complexes, and that subconfluent cells reflect some of the properties of an activated endothelial cell. Yet, I am not sure whether the two are the same. We have done a lot of transcriptomic profi ling of confluent and subconfluent cells. Essentially, we can validate 100% of the genes expressed by subconfluent endothelial cells in the carotid denudation model, whereas we find that only few of the genes expressed by subconfluent cells are actually involved in angiogenesis. Lateral cell migration as it occurs in vivo following a carotid denudation is distinctly different in terms of function and phenotype versus the three-dimensional sprouting angiogenesis in a complex matrix, where we see very distinct matrix-dependent gene expression profi les. Dejana: It might be different, because cell migration is more related to wound repair. In sprouting cells junctions may undergo very subtle modifications. These two models might have something in common but be quite different in nature. Augustin: You have determined the VE-cadherin-dependent transcriptome of endothelial cells. If you were to validate those genes you presented as candidates, in which model would they validate? Dejana: The general hypothesis is that VE-cadherin induces stabilization signals. Therefore the genes up- or down-regulated by this molecule should also be up- or down-regulated in confluent, stabilized monolayers. Owens: I have a question about cell autonomous functions of the cadherins. If you make a chimeric knockout mouse where half of the cells lack the VEcadherins, do they populate blood vessels or undergo apoptosis? Dejana: We haven’t looked at that, and I’m not aware of anyone else who has. However, VE-cadherin is a way through which cells communicate, if you do not have VE-cadherin you may interrupt communication. So, VE-cadherin null cells may die either because they do not express the molecule (cell autonomous) or because they cannot get in touch with the other endothelial cells.

16

LAMPUGNANI & DEJANA

Owens: It would be interesting to know if VE-cadherin null embryonic stem cells are incapable of differentiating into endothelial cells within chimeric knockout mice, but capable of forming other cell lineages. In addition, such experiments may provide insights about the importance of intercellular interactions via cadherins in conferring information to cells about who their neighbours are and where they are at. You spoke extensively about homotypic EC–EC interactions. What is known about heterotypic cell–cadherin interactions and vascular development? I recall seeing ultrastructural evidence that adherens-type junctions between mural cells (smooth muscle cells and pericytes) and endothelial cells are highly abundant during development, but tend to become much rarer later on. Dejana: In vivo in the brain there is a contribution from N-cadherin, which is expressed on astrocytes, pericytes and smooth muscle cells. The interesting concept is that N-cadherin can interact with the fibroblast growth factor (FGF) receptor. This type of interaction may have a different meaning in terms of intracellular signalling. It may be important for the guidance of the endothelium, and FGF has been implicated in this role. It might also be important in the remodelling and stabilization of the vessels. This is still speculation because there aren’t many molecular data. With regard to your other point, you are saying that even in the absence of VE-cadherin there may be other mechanisms that determine the endothelial phenotype. We can’t exclude this. What we know so far is that in the absence of VE-cadherin, or when the junctions are broken using blocking antibodies, or when we use RNAi for VE-cadherin, we could validate most of the genes that we found to be modulated by the absence of this molecule. In vivo if we give blocking antibodies to VE-cadherin, we kill the animal in 24 h. There is a recent paper (Luo et al 2005) on the knockout of N-cadherin specifically in endothelial cells. The embryo died within 10.5 d of development. They claim this is due to the fact that in the absence of N-cadherin there is down-regulation of VE cadherin. Gerhardt: The data that we generated on N-cadherin were from the chicken system. In that system we found that N-cadherin was distributed on the abluminal side of the endothelium in contact with the pericytes (Gerhardt et al 2000). Blocking antibodies lead to a detachment of the pericytes, particularly in the brain parenchyma. Interestingly, we found that the intracellular components that associated with N- and VE-cadherin are different (Liebner et al 2000). It appeared that the junction between endothelium and pericyte only has β -catenin, and we couldn’t find the prominent association with γ -catenin, at this time indicating that it was a very loose type of junction. Now it might be good to look at other associations that might be important. Dejana: A nice idea is that N-cadherin may be important for the guidance of vessels. Gerhardt: We haven’t been able to study this in detail.

ENDOTHELIAL CELL TO CELL JUNCTIONS AND VE-CADHERIN

17

Drake: In studying de novo blood vessel formation, with Elisabetta Dejana, we looked at the VE-cadherin nulls; the null embryos were able to form lumenized blood vessels. There was a recent paper that was a follow-up of an earlier Ncadherin knockout which showed that like the VE-cadherin nulls, the N-cadherin nulls also were able to form lumenized blood vessels. Whatever is directing the polarity of endothelial cells required for lumen formation is, therefore, an open question. References Luo Y, Radice GL 2005 N-cadherin acts upstream of VE-cadherin in controlling vascular morphogenesis. J Cell Biol 169:29–34 Gerhardt H, Wolburg H, Redies C 2000 N-cadherin mediates pericytic–endothelial interaction during brain angiogenesis in the chicken. Dev Dyn 218:472–479 Liebner S, Gerhardt H, Wolburg H 2000 Differential expression of endothelial β -catenin and plakoglobin during development and maturation of the blood–brain and blood–retina barrier in the chicken. Dev Dyn 217:86–98

The role of Eg fl7 in vascular morphogenesis Maike Schmidt*, Ann De Mazière¶, Tanya Smyczek*, Alane Gray†, Leon Parker¶, Ellen Filvaroff‡, Dorothy French§, Suzanne van Dijk¶, Judith Klumperman¶ and Weilan Ye1* * Tumor Biolog y and Angiogenesis Department and † Molecular Biolog y Department, ‡ Molecular Oncolog y Department, § Patholog y Department, Genentech Inc. 1 DNA Way, South San Francisco, CA 94080, USA and ¶ Cell Microscopy Center, Department of Cell Biolog y and Institute for Biomembranes, University Medical Center, Utrecht, 3584C, Utrecht, The Netherlands

Abstract. EGFL7 was identified by a number of groups as a putative secreted factor produced by the vascular endothelial cells (ECs). In a recent publication, we showed that EGFL7 regulates midline angioblast migration in zebrafish embryos—a key step in vascular tubulogenesis. In this study, we further characterized the zebrafish vasculature in the Eg fl7 knockdown embryos at the ultrastructural level, and found that malformation of axial vessels is indeed due to the accumulation of angioblasts and aberrant connection among themselves, but not abnormal interaction between ECs and other cell types. Using in vitro biochemical assays, we demonstrated that EGFL7 is tightly associated with the extracellular matrix (ECM), and it supports EC migration either as a single factor or in combination with other ECM molecules. In order to evaluate if the biological function of EGFL7 is evolutionarily conserved, we generated Eg fl7 knockout mice and analysed vascular development in a number of tissues. We found that vascular coverage of a given tissue is reduced or delayed, and vascular morphogenesis is defective in the Eg fl7 mutant mice. Taken together, we conclude that EGFL7 provides a proper microenvironment for endothelial cell migration, thereby enabling accurate patterning. Our study indicates that the molecular composition of the ECM influences vascular morphogenesis. 2007 Vascular development. Wiley, Chichester (Novartis Foundation Symposium 282) p 18–36

Actively growing tissues such as embryos and tumours require adequate blood supply. They satisfy this need by producing pro-angiogenic factors, which promote new blood vessel formation from existing vessels (angiogenesis), or from progenitor cells (vasculogenesis) (Carmeliet & Jain 2000, Folkman 1997, Hanahan 1997). Although the initiation of new vessel formation is often triggered by mitogenic 1

This paper was presented at the symposium by Weilan Ye, to whom correspondence should be addressed. 18

EGFL7

19

factors that act directly on endothelial cells (ECs) or their progenitors, the construction of a functional vascular network is a complex biological process, which involves all or many of the following steps: (a) EC and supporting cell differentiation and proliferation; (b) EC and supporting cell migration; (c) EC coalescence into cord-like structures; (d) transition from cord to tube (tubulogenesis) (Hogan & Kolodziej 2002); (e) formation of secondary vessels from existing ones (angiogenesis) (Burri & Djonov 2002, de Waal & Leenders 2005); (f) transition from a primitive vascular plexus to a complex network (remodeling) (Carmeliet & TessierLavigne 2005) and (g) recruitment of perivascular supporting cells (Jain 2003, Betsholtz et al 2005). To date, a significant number of molecules have been shown to regulate vasculogenesis and angiogenesis (Alva & Iruela-Arispe 2004, Coultas et al 2005, Davis & Yancopoulos 1999, Ferrara et al 2003, Klagsbrun & Eichmann 2005, Yancopoulos et al 1998). What we learn from studying these factors not only greatly enhances our understanding of vascular biology, but also leads to the development of therapeutic strategies to treat angiogenesis disorders (Ferrara et al 2004, Ferrara & Kerbel 2005, Folkman 2003). However, our knowledge regarding the complex vascular development process is still fragmented; likewise, our means to leverage the angiogenesis process toward medical benefits are limited. Advancements in these two areas should be facilitated by the discovery and characterization of additional angiogenesis regulators. In an effort to identify novel molecules involved in the regulation of vascular development, we performed a whole-mount in situ hybridization screen to identify factors that are expressed in and around the mouse embryonic vasculature. Out of ∼900 genes examined, about two dozen are expressed in the vasculature. One of the vascular-associated genes is Eg fl7, which encodes a ∼30 kDa putative secreted factor with an N-terminal EMI domain, two centrally located EGF-like domains, and a leucine- and valine-rich C-terminus that has some similarities to a coiled-coil domain. We and other groups found that Eg fl7 is specifically expressed in vascular endothelial cells, and its expression is up-regulated during active vascular growth (Campagnolo et al 2005, Fitch et al 2004, Parker et al 2004, Soncin et al 2003). When Eg fl7 is knocked down using a morpholino antisense oligo in zebrafish embryos, we found that midline angioblast migration along the dorsal ventral axis is blocked, and tubulogenesis of axial vessels does not occur (Parker et al 2004). These findings suggest that EGFL7 plays an important role in regulating vascular development. EGFL7 is required for proper positioning of endothelial cells To further understand the mechanism by which EGFL7 influences axial vessel morphogenesis in the zebrafish embryos, we carried out ultrastructural analysis on cross-sections of zebrafish embryos at the 30 somite stage using electron

20

SCHMIDT ET AL

microscopy. We asked if the lack of EC migration along the dorsal–ventral (D–V) axis is due to altered interaction between ECs and the surrounding cells, and we found that ECs do not form abnormal connections with their surrounding cells in the Eg fl7 knockdown (KD) embryos (Fig. 1). Consistent with what we observed using deconvolution microscopy, we saw that ECs accumulate in the space between the hypocord and the gut tube in the KD embryos, whereas ECs in the control oligo injected embryos have readily separated along the D–V axis and assembled into the aorta and posterior cardinal vein at this age. Strikingly, aberrant EC–EC connections are frequently observed in the KD embryos, as adherent junctions are seen across the space where vascular lumen would normally form (Fig. 1 C and D). The abnormal connections are quantified and summarized in Table 1. These

FIG. 1. Ultrastructural analysis of the zebrafish axial vessels. Overview of the vascular area transversely sectioned immediately posterior to the yolk ball-yolk ball extension transition in control (A) and Eg fl7 knockdown (B) zebrafish embryos. In the knockdown embryo, the lumen of the axial vein is obliterated (C) by a cell junction (D, arrow). A, dorsal aorta; V, axial vein; N, notochord; H, hypochordal cell(s); E, endoderm; S, somite; K, pronephric duct. Scale bar: (A, B) 10 µm; (C) 2 µm; (D) 0.5 µm.

EGFL7

21

TABLE 1 Quantitative electron microscopic analysis of vascular defects in the Eg fl7 knockdown (KD) embryos

Hypochord–endoderm distance ( µm) Number of sections with cell junctions crossing aorta and/or vein lumen Number of cell junctions crossing aorta and/or vein lumen Number of junctional interfaces in aorta and vein endothelium Number of junctional interfaces in aorta endothelium Number of junctional interfaces in vein endothelium Number of hypochord-aorta appositions Number of endoderm-vein appositions

Wild-type (n = 13)

Eg fl7 KD (n = 12)

P

43.0 ± 4.2 0 out of 13

28.0 ± 4.3 5 out of 12

<<0.001

0±0

0.9 ± 1.6

<0.05

16.2 ± 2.7

14.1 ± 2.6

N.S.

8.9 ± 2.1

9.8 ± 2.3

N.S.

7.3 ± 2.2

4.3 ± 2.2

<0.01

1.5 ± 1.2 2.1 ± 2.4

2.3 ± 1.7 2.4 ± 2.0

N.S. N.S.

N.S., not significant

findings indicate that the primary defect in the KD vasculature is the inappropriate positioning of ECs in the midline, which results in the formation of aberrant connections, thereby preventing the subsequent assembly of ECs into vascular tubes. EGFL7 regulates vascular morphogenesis in mouse embryos Since the amino acid sequence and expression pattern of EGFL7 is evolutionarily conserved (Parker et al 2004), we went on to investigate if its function is also conserved in mammals by analysing an Eg fl7 knockout mouse line generated via homologous recombination. In this line, a recombination cassette encoding the neo-resistance gene and β -galactosidase reporter gene replaces exons 5 to 7 in the Eg fl7 locus. The inserted cassette creates a premature stop codon and produces a truncated protein with an incomplete EMI domain. The wild-type EMI domain has two highly conserved stretches and is predicted to have a folding pattern that depends on all the cysteines (Callebaut et al 2003, Doliana et al 2000), we believe that the truncated EMI domain is unstable as the most conserved C-terminal portion unique to EMI domains is deleted. Mice lacking the functional EGFL7 protein were born with a substantially altered Mendelian ratio (Table 2). Genotype analysis of neonatal mice indicates that this is due to embryonic lethality. Wholemount immunohistochemistry (IHC) analysis using a series of endothelial cell and mural cell markers reveals an apparent

22

SCHMIDT ET AL

TABLE 2 Summary of the genotypic analysis of the Eg fl7 knockout mouse line Age/percentage of total animals/genotype

+/+

+/−

−/−

Total # of animals

E9.5 to E11.5 P3 to P8 Adult and neonates

28% 29% 32%

47% 61% 56%

25% 11% 12%

255 198 475

C

A

B

F

D

-/-

+/+

+/+

+/+

G

E

-/-

+/+ H

-/-

-/-

FIG. 2. Vascular phenotype in the Eg fl7 knockout mice. Cranial vessels of an Eg fl7 homozygous knockout E10.5 embryo (B) and a wild-type littermate (A) are revealed by wholemount endoglin staining. Vascular smooth muscle cell recruitment is evaluated by wholemount α -smooth muscle actin staining in a pair of wild-type (C and E) and knockout (D and F) E10.5 embryos, and in the cardiac tissues from a pair of wild-type (G) and knockout (H) adult mice. E and F are enlarged views of the intracarotid arteries (arrows) shown in C and D. Sections in G and H are counterstained with haematoxylin. Eg fl7 genotypes are indicated at the bottomright in each panel.

vascular remodelling defect around embryonic day 10.5 (E10.5) in about half of the Eg fl7 −/− embryos (17 out of 35 knockout embryos show vascular abnormality). Since the cranial vasculature is highly stereotypic, we chose this region to describe the vascular phenotype. In the wild-type E10.5 embryos, a set of subcutaneous large veins and the intracarotid artery (ica) run between the ventral and medial midbrain, and then elaborate into a microvascular bed covering the dorsal half of the midbrain (Fig. 1A). In the Eg fl7 −/− embryos, intact vessels are missing, and a cohort of disorganized EC clusters develop in this region (Fig. 2B). Similar EC

EGFL7

23

clusters are also seen in the microvascular bed (data not shown). These isolated EC clusters that are unable to align and connect into intact vessels reveal a problem in the proper positioning of ECs, a defect similar to what we have observed in the zebrafish knockdown embryos. In the Eg fl7 −/− cranial vasculature, the divergence in vessel diameters indicates that vascular remodelling did occur, but the remodelling process is defective. Vascular remodelling involves extensive local movements of ECs and establishment of new EC–EC connections. The aberrant remodelling phenotype in the Eg fl7 −/− vasculature suggests that EGFL7 is required for EC local migration and/or proper EC connection. Since EGFL7 (also called VE-statin) is reported to inhibit vascular smooth muscle cell (vSMC) migration in vitro (Soncin et al 2003), we examined whether vSMC recruitment is affected in the Eg fl7 knockout embryos. At E10.5, vSMC recruitment in the ica as indicated by α smooth muscle actin ( α SMA) IHC appears to be comparable between the Eg fl7 +/+ and −/− littermates (Fig. 2 C–F). In the subset of knockout mice that survived to adulthood, we analysed vSMCs in many organs, and found no obvious defect at the gross morphological level in the knockout mice (Fig. 2G–H). However, ultrastructural analysis is required to find out if EGFL7 is involved in regulating proper EC–vSMC interaction. Aside from the phenotypes described above, the Eg fl7 knockout mouse embryos are morphologically normal. The phenotype in the Eg fl7 mutant fish and mouse embryos plus the highly restricted expression pattern suggests that Eg fl7 functions within the vasculature. EGFL7 regulates tumour angiogenesis in adult mice About 50 % of the Eg fl7 −/− mice survive to adulthood and appear to be normal at the gross morphological level. This enables us to investigate the role of EGFL7 in pathological angiogenesis. To this end, we implanted Lewis Lung Carcinoma cells (Simpson-Herren et al 1974) subcutaneously into age matched Eg fl7 +/+ and −/− female mice that had been backcrossed to C57/BL6 mice for five generations, waited until tumours reached the size of ∼10 × 10 mm, then analysed tumour angiogenesis by FITC-lectin perfusion and subsequent staining of an EC marker ICAM2 on tumour sections. Nine to 10 days after tumour cell injection, tumours in all wild-type mice (n = 6) as well as 3 out of 5 Eg fl7 −/− mice reached the target size and were fully vascularized and perfused (Fig. 3A–F). However, tumours in 2 out of the 5 Eg fl7−/− mice did not reach the same size until day 15, and were not vascularized (Fig. 3H–I). Interestingly, tumour cells in these two mice were still viable judging from the positive DAPI staining (Fig. 3G), presumably by subsisting on the oxygen and nutrients supplied by the adjacent skin vasculature. In tumours that are vascularized, the patterns of lectin and ICAM2 suggest that many tumour vessels are products of angiogenic sprouting from the vessels in the overlying skin.

24

SCHMIDT ET AL Egfl7 +/+ littermates tumor

Egfl7 -/- littermates

D skin

tumor

G skin

tumor DAPI

A skin

day15

day9

E

H

C

F

I ICAM2 staining

B

Perfused FITC-lectin

day9

FIG. 3. Pathological angiogenesis defect in the Eg fl7 knockout mice. Lewis Lung Carcinoma cells were implanted into age matched Eg fl7 wild-type (A–C) and knockout (D–I) adult females that had been backcrossed to C57BL/6 for five generations. Viable tumours beneath the skin are indicated by DAPI staining (A, D, G), vascular perfusion of the skin and tumour was evaluated by tail vein injection of FITC-conjugated tomato lectin (B, E, H). Tumour vascularization was measured by ICAM2 immunofluorescent staining on the same sections (C, F, I). Time of tissue harvest is indicated in the bottom-right in panels A, D and G. ‘Day9’ means 9 days after tumour cell injection.

Lack of tumour angiogenesis in a subset of Eg fl7 −/− mice indicates that EGFL7 may play a role in regulating short distance migration of ECs from the skin into the tumour. The reason for partial penetrance of the aforementioned phenotypes is unclear. We speculate that gene redundancy might play a role. The closely related gene Eg fl 8 is expressed in a subset of vessels in the embryos and some tumours, and the subcellular localization of EGFL8 is similar to that of EGFL7 (data not shown). Therefore, it is conceivable that Eg fl 8 could partially compensate for the loss of Eg fl7. Analysis of Eg fl7::Eg fl 8 double knockout mice should provide valuable information to address this question. EGFL7 is an extracellular matrix associated protein To understand the underlying molecular mechanism of the vascular phenotype observed in the Eg fl7 knockdown fish and knockout mice, we carried out a series

EGFL7 A

MW 1

25 2

3

4

5

6

EGFL7-HA / DAPI

B

Lysates

CM

Transfected CEF cells, stained without detergent

EGFL7

E

D

C

Fibronectin

F

Merged

G

ECM deposited by HUVEC after cell removal.

EGFL7 CD31

H

MCH66 tumor section

FIG. 4. EGFL7 is associated with the extracellular matrix. (A) EGFL7 cannot be detected in the conditioned medium when overexpressed in epithelial cells. Anti-FLAG-tag western blot using 293 cell lysates (lanes 1–3) or conditioned medium (CM, lanes 4–6). 293 cells were transfected with an expression vector alone (lanes 1 and 4), or a control secreted factor interleukin 17 with FLAG tag (lanes 2 and 5), or full length EGFL7 with FLAG tag (lanes 3 and 6). (B–D) Anti-HA immunofluorescent staining of chicken embryonic fibroblasts (CEF) transfected with full length human EGFL7 tagged with HA and cultured for 17 (B), 26 (C) and 90 hours (D). Staining was performed in the absence of detergent. (E–G) Matrix laid down by HUVEC cells was stained with an anti-human EGFL7 mab (E) and anti-fibronectin mab (F) after cells were removed from the plate by EDTA treatment. Significant overlap between EGFL7 and fibronectin is indicated in (G). (H) A MCH66 xenograft tumour section is double stained with an endothelial membrane marker CD31 (red in original colour version) and EGFL7 (green).

of in vitro assays to investigate the biochemical properties of EGFL7. Although the amino acid sequence predicts EGFL7 to be a secreted factor, we could not detect EGFL7 in the conditioned medium when we overexpressed the full-length protein in epithelial cell lines such as human embryonic kidney (HEK-293) cells (Fig. 4A) or Chinese Hamster Ovary (CHO) cells (data not shown), suggesting that EGFL7 is either inefficiently secreted, or is unstable in culture medium. Interestingly, low levels of EGFL7 can be detected when it is overexpressed in fibroblast cell lines such as chicken embryonic fibroblasts (CEF) or NIH3T3 cells (data not shown). Since fibroblasts deposit much more extracellular matrix (ECM) than epithelial cells, and EGFL7 contains three domains that are often found in ECM molecules, including the EMI domain and two calcium-binding subtypes of EGF domains, we speculated that EGFL7 is secreted and quickly incorporated into the ECM. Indeed, when we expressed the HA-tagged full-length EGFL7 in CEF and then stained the cells with an anti-HA antibody in the absence of membrane

26

SCHMIDT ET AL

permeablizing agent, we found that EGFL7 can be detected outside of the cells. In addition, EGFL7 appears to become more and more fibrous over time, a pattern typical of ECM molecules (Fig. 4B–D). To rule out that this is an overexpression artefact, we stained EGFL7 produced endogenously by human umbilical vein endothelial cells (HUVECs), and found that it is present in the ECM in a pattern that significantly overlaps with that of fibronectin (Fig. 4E–G). Furthermore, the localization of EGFL7 in tumour tissues is consistent with an ECM-associated molecule as it is always surrounding the endothelial membrane marker CD31 with a fibrous appearance (Fig. 4H). Taken together, we believe that EGFL7 is not a classic diffusible factor, but is likely a component of an ECM macromolecule. EGFL7 supports endothelial cell migration Since many ECM molecules are known to support cell migration, we went on to investigate if EGFL7 plays a similar role in vitro. In a Boyden Chamber transwell migration assay (Malinda et al 1997), BSA, EGFL7, fibronectin, or vitronectin were coated on both sides of the transwell before HUVECs were seeded on the upper side of the well and then cultured with regular medium in the top chamber and VEGF-containing medium in the bottom chamber. In response to VEGF stimulation, significant numbers of HUVECs migrate across the wells to the under sides when they are supported by EGFL7, fibronectin, or vitronectin, respectively (Fig. 5B–D, F–H), whereas only a small number of cells are capable of migrating on BSA coated wells (Fig. 5A, E). This experiment indicates that EGFL7 has the ability to support EC migration, similar to other known classic ECM substrates. However, the efficiency of HUVEC migration on EGFL7 appears to be poorer than on other substrates (Fig. 5I), a phenomenon that is surprising because EGFL7 is required for adequate migration in vivo. Since the extracellular accumulation of EGFL7 depends on a cell type that produces abundant ECM, and the distribution of EGFL7 coincides with other ECM components (Fig. 4), we reasoned that the natural function of EGFL7 should be evaluated in the context of a mixed ECM substrate. Indeed, we saw that HUVEC migration on a mixed substrate consisting of EGFL7 and fibronectin is comparable to, if not better than, fibronectin alone (Fig. 5J). This finding suggests that EGFL7 is probably incorporated into an ECM macromolecule, and the molecular composition of such an ECM provides the optimal environment for EC migration in vivo. In summary, we have shown that EGFL7 is specifically produced by endothelial cells, and becomes associated with other ECM molecules. The unique ECM complex containing EGFL7 regulates adequate endothelial cell migration and/or proper orientation, thereby enabling the formation of correct EC connections and the subsequent establishment of a patterned vascular network. This function is

EGFL7

27 BSA

EGFL7

Fibronectin

Vitronectin

A

B

C

D

E

F

G

H

5 hours

15 hours

I

HUVEC Transwell Migration assay 6000 5000 4000

5 hours

J

HUVEC transwell migration on different substrates 1600 1400 1200

15 hours

1000

3000

800

2000

600 400

1000

200

0

BSA [5µg/ml]

FN [1µg/ml]

VN [1µg/ml]

EGFL7 [5µg/ml]

0

0.5µg/ml FN

3µg/ml FN

5µg/ml FN

Substrates

4µg/ml FN + 1µg/ml EGFL7

Substrates

FIG. 5. EGFL7 supports EC migration. (A–H) HUVECs are allowed to migrate over a period of 5 (A–D) or 15 (E–H) hours across membranes coated with BSA (A, E), EGFL7 (B, F), fibronectin (C, G) and vitronectin (D, H). Cytogreen staining is used to indicate the populations of HUVECs that reach the underside of the transwell. Shown here is one example of multiple experiments, and only one sample out of three is shown per condition. (I) The experiment shown in A–H is quantified and plotted. FN, fibronectin; VN, vitronectin. Error bars = SEM from triplicate wells. ( J) Numbers of HUVECs migrate across membranes coated with different concentrations of fibronectin alone or in combination with EGFL7 are quantified. Varying concentrations of EGFL7 have been tested and the results were similar. Error bars = SEM from triplicate wells.

evolutionarily conserved. Although ECMs are usually composed of multiple molecules, and many ECM constituents appear to have overlapping functions (Jain 2003, Kalluri 2003, Stupack & Cheresh 2002), our study exemplifies the importance of fine-tuning the ECM composition in the regulation of vascular development. References Alva JA, Iruela-Arispe ML 2004 Notch signaling in vascular morphogenesis. Curr Opin Hematol 11:278–283 Betsholtz C, Lindblom P, Gerhardt H 2005 Role of pericytes in vascular morphogenesis. EXS 115–125 Burri PH, Djonov V 2002 Intussusceptive angiogenesis–the alternative to capillary sprouting. Mol Aspects Med 23:S1–27

28

SCHMIDT ET AL

Callebaut I, Mignotte V, Souchet M, Mornon JP 2003 EMI domains are widespread and reveal the probable orthologs of the Caenorhabditis elegans CED-1 protein. Biochem Biophys Res Commun 300:619–623 Campagnolo L, Leahy A, Chitnis S et al 2005 EGFL7 is a chemoattractant for endothelial cells and is up-regulated in angiogenesis and arterial injury. Am J Pathol 167:275–284 Carmeliet P, Jain RK 2000 Angiogenesis in cancer and other diseases. Nature 407:249–257 Carmeliet P, Tessier-Lavigne M 2005 Common mechanisms of nerve and blood vessel wiring. Nature 436:193–200 Coultas L, Chawengsaksophak K, Rossant J 2005 Endothelial cells and VEGF in vascular development. Nature 438:937–945 Davis S, Yancopoulos GD 1999 The angiopoietins: Yin and Yang in angiogenesis. Curr Top Microbiol Immunol 237:173–185 de Waal RM, Leenders WP 2005 Sprouting angiogenesis versus co-option in tumor angiogenesis. EXS 65–76 Doliana R, Bot S, Bonaldo P, Colombatti A 2000 EMI, a novel cysteine-rich domain of EMILINs and other extracellular proteins, interacts with the gC1q domains and participates in multimerization. FEBS Lett 484:164–168 Ferrara N, Kerbel RS 2005 Angiogenesis as a therapeutic target. Nature 438:967–974 Ferrara N, Gerber HP, LeCouter J 2003 The biology of VEGF and its receptors. Nat Med 9:669–676 Ferrara N, Hillan KJ, Gerber HP, Novotny W 2004 Discovery and development of bevacizumab, an anti-VEGF antibody for treating cancer. Nat Rev Drug Discov 3:391–400 Fitch MJ, Campagnolo L, Kuhnert F, Stuhlmann H 2004 Egfl7, a novel epidermal growth factor-domain gene expressed in endothelial cells. Dev Dyn 230:316–324 Folkman J 1997 Angiogenesis and angiogenesis inhibition: an overview. EXS 79:1–8 Folkman J 2003 Angiogenesis inhibitors: a new class of drugs. Cancer Biol Ther 2:S127–133 Hanahan D 1997 Signaling vascular morphogenesis and maintenance. Science 277:48–50 Hogan BL, Kolodziej PA 2002 Organogenesis: molecular mechanisms of tubulogenesis. Nat Rev Genet 3:513–523 Jain RK 2003 Molecular regulation of vessel maturation. Nat Med 9:685–693 Kalluri R 2003 Basement membranes: structure, assembly and role in tumour angiogenesis. Nat Rev Cancer 3:422–433 Klagsbrun M, Eichmann A 2005 A role for axon guidance receptors and ligands in blood vessel development and tumor angiogenesis. Cytokine Growth Factor Rev 16:535–548 Malinda KM, Goldstein AL, Kleinman HK 1997 Thymosin beta 4 stimulates directional migration of human umbilical vein endothelial cells. FASEB J 11:474–481 Parker LH, Schmidt M, Jin SW et al 2004 The endothelial-cell-derived secreted factor Egfl7 regulates vascular tube formation. Nature 428:754–758 Simpson-Herren L, Sanford AH, Holmquist JP 1974 Cell population kinetics of transplanted and metastatic Lewis lung carcinoma. Cell Tissue Kinet 7:349–361 Soncin F, Mattot V, Lionneton F et al 2003 VE-statin, an endothelial repressor of smooth muscle cell migration. EMBO J 22:5700–5711 Stupack DG, Cheresh DA 2002 ECM remodeling regulates angiogenesis: endothelial integrins look for new ligands. Sci STKE 2002:PE7 Yancopoulos GD, Klagsbrun M, Folkman J 1998 Vasculogenesis, angiogenesis, and growth factors: ephrins enter the fray at the border. Cell 93:661–664

DISCUSSION Dejana: Although you get the same amount of fibronectin, is the organization of the matrix the same? The organization of the matrix may be strongly

EGFL7

29

influenced, and this in turn may influence the direction of the movement of blood vessels. Ye: We need to further evaluate fibronectin organization in vivo by electron microscopy. However, in vitro, we have shown that wild-type and knockout endothelial cells migrate equally well on fibronectin coated on the inside of a Boyden chamber, indicating that fibronectin assembly is not affected under this condition. Although the presentation of fibronectin is artificial in vitro, it has been shown that a cell can influence how fibronectin is assembled under such condition. Dejana: Is VEGF inducing the production of matrix proteins? Can it induce endothelial cells to release and organize fibronectin? Ye: VEGF increases the level of Eg fl7 message one- to twofold. I don’t know the answer to fibronectin release and organization. Dejana: As soon as endothelial cells attach they will release fibronectin and organize their own matrix. It may be that the growth factor influences the organization of the endogenous matrix. Drake: It would be worth looking at the fibronectin receptor α5β1. Ye: We did; but have not seen any obvious difference between the wild-type and knockout tissues. Drake: There are recent papers, for example that of Wijelath et al (2002), suggesting that α5β1 integrin interacts with the FLK-1 receptor. This is interesting as there could be α5β1 present but for some reason it can’t form a complex that enhances VEGF signalling. Adams: You mentioned you haven’t done the electron microscopy yet. It would be interesting to look at the ultrastructure of those tubes, and in particular the cell shape. Many branching morphogenesis processes involve cell elongation. This could be important. Ye: That’s an interesting point. Our observation in zebrafish embryos was that the angioblasts in the Eg fl7 knockdown embryos failed to form squamous shapes, therefore we hypothesize that EGFL7 might have some signalling properties to guide endothelial cells to spread along a defined orientation. We paid particular attention to the knockout mice, surprisingly, most of the endothelial cells do assume squamous shapes even in the multilayered vascular front in the Eg fl7 knockout retina. This was evaluated by confocal microscopy. Adams: I would really look at the ultrastructure to see how those lumens look. Ye: I agree! I forgot to mention that when we look at the vascular migration front in the Eg fl7 knockout retina, we sometimes see more red blood cells outside of the vessels. We suspect that lumen formation is irregular, but we don’t know the exact nature of this phenotype. It seems that this problem can eventually resolve itself. Augustin: When there is a new molecule there are hundreds of obvious questions, such as what is it binding to and what are the receptors? Is the presentation on the endothelial cells altered in the null cells? What I find exciting about this molecule

30

SCHMIDT ET AL

is that it is a molecule that is made by endothelial cells themselves. This supports the notion that endothelial cells are professional blood vessel makers. They receive some outside stimulus, and then they know everything about making blood vessels. What do you know about the regulation of EGFL expression? Is hypoxia the driving force of EGFL expression? You showed the deposition in the matrix, but given the down-regulation in the adult, this appears to be a transient phenomenon. How long is it active when you go further back into the retinal layer? When does it disappear? You said it doesn’t affect neural cell recruitment, but what is the relationship between EGFL deposition and organization of neural cells? Lastly, do you have any idea about the receptor? Ye: It is exciting. We have been going after the receptor. We see binding to two different integrins, but interestingly we can see binding on cells that express these integrins, but we have not yet seen them binding to purified protein. This suggests that it might not be a simple story of binding. Perhaps integrins anchor EGFL7 to endothelial cells, and EGFL7 binds additional factors on the endothelial cells. With regard to the regulation of Eg fl7 expression, we did an in vitro experiment in which we stimulated HUVECs with a number of known angiogenic factors (e.g. VEGF, FGF2, HGF) and found that all of them increased the expression of Eg fl7 moderately. We think, along the same lines that you have proposed, that endothelial cells are professional blood vessel makers. They know what they should do when they enter a certain differentiation state. EGFL7 is probably part of the provisional matrix that only needs to be transiently expressed when endothelial cells are migrating in a collective fashion. Eg fl7 message level decreased in P12 retina in the NFL vessels, which are more mature and have tight mural cell coverage. We hypothesize that Eg fl7 expression is linked to a specific endothelial cell differentiation status, regardless of the external signals. Shibuya: You showed that this molecule is not only expressed in a physiological angiogenic setting, but also in tumour angiogenesis. Yet we know that the tumour vasculature is somewhat disorganized. So if we suppress this molecule in pathological conditions, will it suppress tumour angiogenesis, where patterning isn’t important? Ye: I agree that patterning isn’t an important factor in tumour angiogenesis, but there are probably two ways we can exploit a molecule. The first is to exploit its overall biological function (how it controls the creation of a vascular network). The second is a defined cellular property. In this case, we believe that Eg fl7 has a role in regulating the collective migration of endothelial cells. Under certain circumstances you may need rapid sprouting angiogenesis to support tumour growth. In this context, EGFL7 might be a useful target. There are data to support this hypothesis. We did a syngeneic tumour experiment in the knockout mice, and found that tumours in 40% of the knockout mice did not vascularize when they reached the size of about 6–8 mm in surface diameter, whereas all the same sized

EGFL7

31

tumours in the wild-type hosts were vascularized. In addition, it took almost twice as long for these non-vascularized tumours to reach the same size as those in the wild-type hosts. Gerhardt: I am pleased to see that you used the retina system. It’s great for looking at vascular development, but it has its problems. I was surprised to see that you found no change in VEGF levels, although large parts of the retina were still avascular. Ye: The hyaloid vessel regression is delayed significantly. We don’t think this is a cause of the delayed migration, because delayed migration occurs as early as P2, when the hyaloid vessels are still intact. We believe that delayed hyaloid vessel regression is a consequence of delayed retinal vascular development. Gerhardt: This is difficult, because as soon as the hyaloids don’t regress, a whole range of other problems come into play. For example, in your adult phenotype with the structural similarity to retinopathy, if the hyaloid vessels persist what they do is to send off small sprouts into the retina in the peripheral region, leading to multiple vascular layers. These sprouts do not follow the astrocytic network as closely as the normal vessels do. They cause changes in flow distribution, and a whole series of complications that look exactly like what you have presented. Maybe it would be interesting to look at the mechanism of hyaloid regression Ye: We are aware of the complication of the hyaloid vessels. In known examples where hyaloid vessel invasion occurred, the hyaloid vessels caused disruption of retinal architecture. It is quite a different kind of alteration compared to the phenotype in the Eg fl7 knockout. In a way, we can’t absolutely exclude the possibility of the contribution of hyaloid vessels to the multiple layers, because of the lack of definitive markers to distinguish these two populations of vessels. But the hyaloid vessels are much brighter for isolectin B4, and we found that the multilayered vessels in the Eg fl7 knockout retina are much dimmer for isolectin B4. On the basis of this limited evidence we think that it is not the hyaloid vessel diving down into the retina; it is the retinal vessels that generated the pattern. In addition, we also saw a similar phenotype (i.e. lack of planar restriction on the vasculature) in other tissues, such as the peritonea. Betsholtz: For how long do the hyaloids persist? Ye: We can still detect them at P11. They eventually regress in the adult. Kitajewksi: Can you interpret the phenotype as an inability of the endothelial cells to migrate along the VEGF gradient? Do genetic alterations of the VEGF gradient in vivo, such as those that Christer Betsholtz has done, exhibit similar phenotypes? Ye: The GFAP-PDGFA transgenic line generated in Dr Betsholtz’s group does have a thickened NFL vascular layer (Gerhardt et al 2003), similar to what we saw in our knockout line. In this transgenic line, the astrocytic network density is increased, which is the likely cause of increased NFL vascular thickness. In the

32

SCHMIDT ET AL

Eg fl7 knockout line, we did not see alteration in the astrocytic network, nor did we see alteration in VEGF distribution. Kitajewski: And the intrinsic migration is OK? Ye: Yes, on the basis of an in vitro migration assay using ECs isolated from wildtype and knockout tissues. Ruhrberg: You showed a dramatic delay in endothelial cell migration. Interestingly, the observation that there is partial recovery suggests that not all factors controlling migration are missing. Could the phenotype you describe represent a lack of migration along a vectorial path, i.e. from the centre to periphery, and be related to a mechanical problem? Holger Gerhardt has shown that the fi lopodial processes of endothelial cells follow the underlying astrocyte processes (Ruhrberg et al 2002, Gerhardt et al 2003). During this migration, there is probably not just signalling from the astrocytes to the endothelial cells, but also feedback from the growing endothelium to the astrocytes, and the two systems might therefore organize each other. So even though fibronectin is not grossly disturbed, might there be subtle reorganization, perhaps a change in orientation of the fibrils, which could be detected by electron microscopy? We know that VEGF provides the signal to stimulate outgrowth of vessels, but VEGF is not the force that pulls them. For the migration process to occur, mechanical traction is needed, and one possibility is that the system you are investigating provides this by helping vessels to organize their own fibronectin rails. Ye: That is a good point. We are doing analyses to pursue thoughts along these lines. Wilting: You mentioned data where you have been looking at abnormal angiogenesis in tumours. You see that EGFL is present there. On the one hand you say that during normal development it is needed for vessel development. On the other hand, you will never see normal vessels in a tumour. How do you reconcile this discrepancy? Ye: In the context of a tumour, endothelial cells are probably overwhelmed by the unchecked production of angiogenic factors. In general, there is overproduction of VEGF in a tumour. The over stimulation of tumour vasculature by VEGF could mask the activity of other factors, such as EGFL7, which is a player downstream of the initial angiogenic stimuli to fine-tune the vascular pattern. Owens: One way to get around this is to look at some more subtle models of arterialisation, such as hypoxia. Have you done this? Ye: Not yet. Owens: I think what is important is that this looks like a vascular patterning defect, but Christer Betsholtz and Holger Gerhardt have shown that precise spatiotemporal control of these factors is critical for normal vascular patterning. The corkscrew pattern seen in your mutants that survive really looks like

EGFL7

33

dysregulated vascular patterning that is common when we dump high concentrations of angiogenic substances onto these beds. It is likely that appropriate patterning is dependent on very precise spatiotemporal regulation of the concentration of angiogenic substances, and that global perturbations will not allow us to really understand how the process is really regulated. Wilkins: I am intrigued by the recovery process in the knockout, and its molecular biology. The simplest explanation would be that there is up-regulation of another molecule that takes its place, perhaps a paralogue. Ye: There is a homologue for EGFL7, which is EGFL8. However, we didn’t see a broad vascular expression of EGFL8. We checked the Eg fl7 knockout but did not see up-regulation of Eg fl 8. Currently, we have not found a strong candidate that could compensate for EGFL7. Betsholtz: What is the situation in fish? Is there a homologue for EGFL8 in fish? Ye: No. Betsholtz: Could this explain why the fish phenotype is so strong? Ye: That is what we thought. It is surprising: usually a fish has more because of the gene duplication event. We couldn’t find EGFL8 in fish. Lammert: I have a question about the fine localization of EGFL7. You focused mainly on the sprouts. Have you looked at the aorta, where it forms and assembles? You saw EGFL7 all around the cells first, then they intermingle, and when they just form the lumen do you see it inside the lumen or only outside? Ye: In all the mouse tissues that we have analyzed, we have not seen EGFL7 on the luminal side of a vessel. Initially, we thought that EGFL7 was providing a polarizing signal that enables the endothelial cells to flatten out and spread on the basal surface. But somewhat disappointingly, we found that endothelial cells are still able to flatten out and form tubes in the mouse Eg fl7 knockout tissues, though the tubular network is somewhat disorganized. What controls the polarity of endothelial cells is the big missing piece in our understanding of vascular tube formation. Dejana: How are the junctions in these cells? Ye: We have not done a thorough analysis on this yet. Epstein: Why do the half that die, die? Are there early patterning defects in the major vessels? Can you say something more about the epicardial defects that you mentioned? Ye: We speculate that it is because of the vascular defects because we haven’t seen any defect outside of the vasculature. Although there is circulation in the knockout embryos, development of some vessels is delayed. For example, as I mentioned in the talk, many large vessels that perfuse the dorsal midbrain are missing at E10.5, this could change the flow dynamics. As for the epicardium, we found that vascularization at E11.5 to E12.5 is delayed.

34

SCHMIDT ET AL

Wilting: It is not the perfusion that is missing, but as soon as there is a major vessel missing in the embryo there is an enormous increase in peripheral resistance that the tiny heart of the embryo won’t be able to overcome. Epstein: Do the hearts of the dead embryos look different? Ye: We haven’t looked closely. They appear anatomically normal. Drake: If the pump isn’t pumping, then you are not getting perfusion of the system. A quick screen for this would be to look at whether blood is moving through these major vessels. Shovlin: Coming back to the question of partial penetrance, which mouse strain did you look at? Ye: We have examined the knockouts in three genetic backgrounds: 129: C57BL/6 mixed background, backcrossed to BL6 at advanced levels (currently > N9), and in pure 129 (by crossing the chimera to 129). We have not seen significant changes in different backgrounds. Vargesson: What kills the zebrafish embryo? Ye: There is no circulation in the Eg fl7 knockdown zebrafish embryos, and they died around day 5. Zebrafish embryos can survive without circulation for about 5 days. The mutant vasculature is profoundly abnormal: besides lack of perfusion in the axial vessels, ISV sprouting is not sustained due to lack of flow. We see ISV sprouting at 24 hours, but they disappeared later. Adams: You first postulated that this might be a polarization molecule. Have you abandoned this hypothesis now? Ye: In the retina that seems to be the case. But we can’t conclude this yet. Gerhardt: One thing that should be mentioned is that migration of endothelial cells is quite different in the retina to other places because of the matrix. In the retina fibronectin is strongly produced by astrocytes, and laid down as a structure. A recent paper (Uemura et al 2006) looked at the TLX knockouts. In that mouse there is a similar phenotype in that there is a delay of the formation of the vascular plexus. The astrocytes form, but the authors conclude that the astrocytes fail to assemble the fibronectin matrix properly. Is it the matrix that is laid down by cells that guide the vessel? Is this the reason why your vessels don’t find their way into the retinal periphery? Do you think that EGFL7 is needed in the context of fibronectin? One of your images suggested that the astrocytic fibronectin matrix looks abnormal. Ye: In the image the astrocytic fibronectin matrix in the Eg fl7 knockout retina looks different because of increased thickness of the vascular bed. If we scan the knockout and wild-type retina with the same thickness, then the difference disappears. Gerhardt: We have knocked out fibronectin in the astrocytes and this causes a phenotype that looks similar to yours (unpublished data).

EGFL7

35

Ye: In my talk, we showed evidence to indicate that EGFL7 is a component of a special matrix that requires fibronectin. Your result further supports this hypothesis. Epstein: Are the adult null females fertile? Ye: Yes, but with smaller litter size. Adams: We have briefly touched on cell adhesion, and integrins have been mentioned. Have you compared the attachment, spreading and migration of wild-type cells on substrates containing EGFL7 and those which don’t? This might tell you something about the role of this molecule. Ye: We have tried mixing fibronectin with EGFL7 and looking at HUVEC migration on the mixed substrate, but didn’t see a clear tendency. Palecek et al (1997) showed that a biphasic relationship exists between substratum adhesiveness versus epithelial cell migration speed. We couldn’t reproduce this with endothelial cells (HUVECs), migration speed does not decrease even at exceedingly high concentration of fibronectin. Ruhrberg: Should endothelial cell junctions be considered in understanding the EGFL7 phenotype? Dejana: I think it might explain some of your phenotype if junctions are not correctly organized. The other point is the lack of regression of the hyaloid. Ye: We suspect delayed hyaloid regression is due to altered oxygen tension. It is an area I am still trying to understand. The fact that we saw dramatic delay at P2 and yet the hyaloid vessels have not regressed at this time point would suggest that delayed retinal vascular development would have some impact on hyaloid vessels regression. Wilting: Have you ever seen retinal cells detaching from the tips and migrating as single cells? This is a problem having a cell layer growing with the cells in front but not detaching. In embryos, we get the impression that there are sometimes cells that are detaching and perhaps integrating into the system again. Ye: At P11 we see individual endothelial cells detach. Gerhardt: If you expose animals to hyperoxia during the phase where the retinal plexus develops, endothelial proliferation ceases and the tip cells migrate off as clusters. Ruhrberg: I would like to mention the phenotype of the VEGF188 mice, which only express VEGF188 rather than a combination of several different VEGF isoforms. The VEGF188 mice don’t have an overall problem with forming a retinal vasculature, but they have a specific problem forming arteries (Stalmans et al 2002). In these mice, the hyaloid vasculature persists, presumably to compensate for the lack of arteries and oxygenation in the retina. Something similar may happen in the EGFL7 knockout mice.

36

SCHMIDT ET AL

References Gerhardt H, Golding M, Fruttiger M et al 2003 VEGF guides angiogenic sprouting utilizing endothelial tip cell fi lopodia. J Cell Biol 161:1163–1177 Palecek SP, Loftus JC, Ginsberg MH, Lauffenburger DA, Horwitz AF 1997 Integrin-ligand binding properties govern cell migration speed through cell-substratum adhesiveness. Nature 385:537–540 (erratum: 1997 Nature 388:210) Ruhrberg C, Gerhardt H, Golding M et al 2002 Spatially restricted patterning cues provided by heparin-binding VEGF-A control blood vessel branching morphogenesis. Genes Dev 16:2684–2698 Stalmans I, Ng YS, Rohan R et al 2002 Arteriolar and venular patterning in retinas of mice selectively expressing VEGF isoforms. J Clin Invest 109:327–336 Uemura A, Kusuhara S, Wiegand SJ, Yu RT, Nishikawa S 2006 Tlx acts as a proangiogenic switch by regulating extracellular assembly of fibronectin matrices in retinal astrocytes. J Clin Invest 116:369–377 Wijelath ES, Murray J, Rahman S et al 2002 Novel vascular endothelial growth factor binding domains of fibronectin enhance vascular endothelial growth factor biological activity. Circ Res 91:25–31

A model of intussusceptive angiogenesis Max Levin*†, Andrew J. Ewald*, Martin McMahon‡, Zena Werb* and Keith Mostov*†1 Departments of * Anatomy, † Biochemistry and Biophysics, and ‡ Cancer Research Institute and Department of Cellular and Molecular Pharmacolog y, University of California, San Francisco, CA 94143, USA

Abstract. There are two types of angiogenesis, sprouting angiogenesis and intussusceptive angiogenesis. Sprouting angiogenesis, the outgrowth of a new branch from an existing vessel, is a process that has been well characterized using numerous model systems. Intussusceptive angiogenesis, splitting of an existing vessel, starts with the formation of an intraluminal vascular pillar. The cellular interactions and molecular mechanisms regulating vascular pillar formation are largely unknown. An increased understanding of intussusceptive angiogenesis requires the development of observable, manipulable model systems to study vascular pillar formation. We have established a cell culture model of vascular pillar formation in an endothelial cell–smooth muscle cell co-culture system. Vascular pillar formation was inhibited by broad spectrum matrix metalloproteinase inhibitors. 2007 Vascular development. Wiley, Chichester (Novartis Foundation Symposium 283) p 37–45

Vascular networks are built and function in the embryo, but are continually remodelled and refined during fetal and postnatal development. There are two major mechanisms for remodelling vascular networks: vasculogenesis and angiogenesis. Vasculogenesis is the formation of de novo blood vessels and occurs by aggregation of endothelial precursor cells to vascular cords that subsequently form tubes containing lumens. Angiogenesis occurs in two different ways, by sprouting angiogenesis (outgrowth of a new vessel branch) and by intussusceptive angiogenesis (longitudinal splitting of vessels) (Fig. 1) (Patan 2004, Carmeliet 2005). For more than 120 years, sprouting angiogenesis was the only known form of angiogenesis. Intussusceptive angiogenesis was fi rst described by Burri and co-workers in 1986 (Caduff et al 1986). Using electron microscopy, they demonstrated that the massive growth of the rat lung vasculature after birth occurs almost exclusively through 1

This paper was presented at the symposium by Keith Mostov, to whom correspondence should be addressed. 37

38

LEVIN ET AL

FIG. 1. Sprouting angiogenesis and intussusceptive angiogenesis. In sprouting angiogenesis, a new vessel branch grows from an existing vessel. In intussusceptive angiogenesis, an intraluminal vascular pillar grows though the lumen. The pillar subsequently widens, thus splitting the vessel.

intussusception. Since this landmark discovery, intussusception has been demonstrated in most organs during development as well as in tumours (Burri & Djonov 2002). In contrast to intussusceptive angiogenesis, sprouting angiogenesis is a relatively well characterized process. Sprouting angiogenesis starts with the degradation of the vascular basement membrane followed by outgrowth of a new vessel branch. The cellular and molecular mechanisms regulating sprouting angiogenesis have been elucidated using a range of in vitro (e.g. cell culture assays), ex vivo (e.g. the rat aortic ring assay) and in vivo models (e.g. the postnatal mouse retina). However, the importance of intussusceptive angiogenesis may be underestimated. The vascular pillars that initiate intussusception were first visualized by electron microscopy and vascular casts. Standard methodology does not normally distinguish if a new vessel has formed by intussusception or sprouting. Therefore, new vessels are frequently assumed to be formed from sprouting. Intussusceptive angiogenesis has remained enigmatic largely due to the lack of flexible model systems. Intussuceptive angiogenesis starts with the formation of

IN VITRO INTUSSUSCEPTIVE ANGIOGENESIS

39

vascular pillars. The pillar is a slender endothelial cell cylinder, frequently with smooth muscle cells/pericytes inside, that stretches, through the lumen, from one part of the vascular wall to another. Subsequently, the pillar widens as a result of ingrowth of smooth muscle cells/pericytes, fibroblasts and extracellular material (Burri & Djonov 2002). In small vessels, a widening pillar splits one vessel into two parallel vessels. In larger vessels, a network of widening pillars remodels a single vessel into a network of smaller vessels (Patan 2004). How endothelial cells and smooth muscle cells/pericytes interact to form the vascular pillar is not known. Furthermore, the molecular mechanisms regulating intussusception have not been elucidated. We have established a cell culture model to study intussusceptive angiogenesis. With this model, key questions about intussusceptive angiogenesis may be addressed. How do endothelial and smooth muscle cells interact to form pillars? Which molecules regulate intussusceptive angiogenesis? Results and discussion Our initial aim was to develop a co-culture model to study endothelial cell–smooth muscle cell interactions during vascular tube formation. However, we subsequently observed that, not only tubes, but also vascular pillars could be induced. When endothelial cells are plated at low density; endothelial tubes are formed. When plated at high density; vascular pillars are formed (Fig. 2). In our model, endothelial cells are grown on top of smooth muscle cells. The smooth muscle cells are seeded first and allowed to form approximately two cell layers. Then endothelial cells are added on top of the smooth muscle cells. Thus, the model allows direct interaction between the two cell types. If endothelial cells are plated at low density, the cells aggregate into cords that subsequently form tubes containing lumens (Fig. 3A). The tubes have a vessel like morphology; approximately 10 µm in diameter and with a distinct basement membrane. If endothelial cells are plated at high density, an endothelial monolayer forms on top of the smooth muscle cells. The endothelial monolayer is separated from the smooth muscle cell layers by a basement membrane. In this way, a flat in vitro model of a vascular wall is formed. This ‘vascular wall’ subsequently undergoes folding and vascular pillars are formed in an intussusceptive-like manner (Fig. 3B and C). The pillars stretch, through the medium, from one part of the ‘vascular wall’ to another. Similar to what has been described in vivo, the pillars are covered by an intact endothelial layer with a basement membrane facing the inside of the pillar. Also similar to in vivo, smooth muscle cells can be detected inside the pillars. To investigate molecular mechanisms involved in intussusceptive angiogenesis, we performed screening experiments using our cell culture model. To this end, we

40

LEVIN ET AL

FIG. 2. Tube and pillar formation in vitro. Endothelial cells (ec) form tubes when grown at low density on top of smooth muscle cells (smc) (left panel). Endothelial cells at high density form a confluent monolayer (Day 2), the monolayer subsequently remodels to form pillars that extend through the media (Day 7) (right panel).

used growth factors and inhibitors known to be important in angiogenesis and vasculogenesis. Preliminary results from these screening experiments suggest that broad spectrum matrix metalloproteinase (MMP) inhibitors reduce intussusceptive pillar formation. Interestingly, MMP inhibition did not seem to have an effect on vascular tube formation. MMPs are a group of 23 different enzymes important in matrix degradation and tissue remodelling. MMPs have diverse biological roles; e.g. degradation of basement membranes and extracellular matrix, promoting cell migration through tissues and proteolytic processing of bioactive molecules (Hotary et al 2000, Heissig et al 2003, Mott & Werb 2004). The flexibility of our new model makes it possible to investigate which of these biological roles are important for vascular pillar formation. For example, cellular migrations and cellcell interactions in vascular pillars may be directly visualized using high resolution 3D confocal microscopy of fluorescently labelled co-cultures (i.e. green fluorescent endothelial cells and red fluorescent smooth muscle cells).

IN VITRO INTUSSUSCEPTIVE ANGIOGENESIS

41

FIG. 3. Confocal sections of tube (A, A′), pillar (B, B′), and endothelial monolayer below pillar (C, C′). The tubes grow partly embedded in the smooth muscle cell layer (A, A′). The apical side of the endothelial cells faces the lumen and the basal side is separated from smooth muscle cells by a basement membrane (A′). The pillars (B, B′) run through the medium above the endothelial monolayer (C, C′) that grows on top of the smooth muscle cells. In pillars, the apical side of the endothelial cells faces the medium and the basal side has a basement membrane facing the inside of the pillar (B′). Smooth muscle cells can frequently be detected inside pillars.

Conclusions Intussusceptive angiogenesis is likely to be important in cancer and in the development of many organs. Owing to the difficulty of studying intussusceptive angiogenesis it is likely that vascular remodelling processes attributed to sprouting angiogenesis may in fact be the result of intussusceptive angiogenesis. Work with the proposed in vitro model of intussusceptive angiogenesis is likely to yield data that are broadly significant in understanding intussusceptive angiogenesis and its many roles. Acknowledgements This study was supported by funds from the NIH (AI53194 to KM and ZW). ML was supported by Postdoctoral Fellowships from the Sweden America Foundation and the Swedish Heart and Lung Foundation. AJE was supported by an NIH NRSA Institutional Fellowship (HL-007731) and by the California Breast Cancer Research Program (11FB-0015).

42

LEVIN ET AL

References Burri PH, Djonov V 2002 Intussusceptive angiogenesis–the alternative to capillary sprouting. Mol Aspects Med 23:S1–27 Caduff JH, Fischer LC, Burri PH 1986 Scanning electron microscope study of the developing microvasculature in the postnatal rat lung. Anat Rec 216:154–164 Carmeliet P 2005 Angiogenesis in life, disease and medicine. Nature 438:932–936 Heissig B, Hattori K, Friedrich M, Rafi i S, Werb Z 2003 Angiogenesis: vascular remodeling of the extracellular matrix involves metalloproteinases. Curr Opin Hematol 10:136–141 Hotary K, Allen E, Punturieri A, Yana I, Weiss SJ 2000 Regulation of cell invasion and morphogenesis in a three-dimensional type I collagen matrix by membrane-type matrix metalloproteinases 1, 2, and 3. J Cell Biol 149:1309–1323 Mott JD, Werb Z 2004 Regulation of matrix biology by matrix metalloproteinases. Curr Opin Cell Biol 16:558–564 Patan S 2004 Vasculogenesis and angiogenesis. Cancer Treat Res 117:3–32

DISCUSSION Betsholtz: Is the starting point really a monolayer of endothelial cells? Or is it a flat bilayer, or tube of cells? Mostov: It looks like a monolayer of endothelial cells on top of two cell layers of smooth muscle cells. It is therefore similar to a flat vascular wall with the luminal side facing the cell culture medium; it is not a tube. We plate the smooth muscle cells first and then put the endothelial cells on top of them. Betsholtz: That might not necessarily be equivalent to intussusception, where the starting point is a tube. There are several models in which tube formation occurs from either dispersed or monolayer-seeded endothelial cells. Mostov: Correct, but we see the starting point as a flat vascular wall and then vascular pillars form; the first step in intussusceptive angiogenesis. The pillars stretch through the medium, from one part of the vascular wall to another. Importantly, the pillars do not have a lumen (they are not tubes), the inside contains the basement membrane and smooth muscle cells (more like an inverted tube). The pillar formation in our system is similar to what Patan and co-workers have described in the wall of a large vein in an experimental tumour model (Patan et al 2001). Usually, models of tube formation are in 3D culture, embedded in ECM. Betsholtz: Could Hellmut Augustin comment, who has had a lot of experience with 3D cultures? Augustin: I can comment on this model because we have done quite a few of the same experiments. We have developed a three dimensional co-culture model of endothelial cells and smooth muscle cells. In this model, smooth muscle cells form the core of a spheroid which is covered by a surface layer of endothelial cells. In this scenario, endothelial cells organize with the smooth muscle cells in a quasi-3D arrangement that mimics the assembly of the normal vessel wall in an inside–out orientation. The smooth muscle cells control the phenotype of the quiescent

IN VITRO INTUSSUSCEPTIVE ANGIOGENESIS

43

endothelium. In contrast, in a two dimensional endothelial cell–smooth muscle cell sandwich assay, endothelial cells in our hands don’t like to adhere to a monolayer of smooth muscle cells. You need a dense smooth muscle cell layer, because if endothelial cells contact the plastic they will push the smooth muscle cells aside and segregate into islands. You can play with this system to better mimic the in vivo situation in which endothelial cells and smooth muscle cells are separated by an extracellular matrix. We have covered the smooth muscle cell monolayer with different matrix components. This increases adhesion of endothelial cells, and then you see some formation of tubes or cords. Even then experience shows that single endothelial cells are very unhappy. The first thing they do is to align and make cell–cell contacts, which is the minimum requirement to render them responsive to growth factors. I would question whether alignment is the same as an invasive sprouting phenotype. We have done a lot of transcriptome analyses in 2D and 3D systems. In the 2D system, be it on Matrigel or on smooth muscle cells, we get a lot of genes that are related to stress responses or apoptosis, but we don’t find many that have to do with the angiogenic program. In contrast, 3D sprouting angiogenesis originating from matrix-embedded endothelial cell spheroids yields a transcriptomic fi ngerprint that very much corresponds to the gene expression profi le of angiogenic endothelial cells in vivo. Mostov: Max Levin tells me that what he sees is dependent on the cell density. At lower densities he sees what appear to be tubes forming. It is only at very high densities of plated cells that we see these folds. We see many more tubes at lower densities. He does this in a 96 well plate format, where he sees more of the folds. Augustin: Generally speaking, I would assume that if you titrate the number of smooth muscle cells you would see similar effects, in that the endothelial cells prefer to adhere to the matrix or plastic. Mostov: Under these conditions most of the endothelial cells are not adhering to the plastic. There is a basement membrane component that is laid down between the two cell types that is being remodelled. Lammert: How do you measure this intussusception? Is it that the junctions are mainly remodelled? Mostov: That’s an interesting question. From what has been published, one of the characteristics of intussusceptive angiogenesis is that the endothelial cells have pillars with smooth muscle cells within them (Burri & Djonov 2002). There is an active migration. Whether at the end, when the fusion of the cell layers is consummated, this is a matter of simple switching or something more complex is unknown. Davies: Once you have made your pillar, and it is elongating, junctional switching is supposed to happen at the end of the process. It is a bit like a zip fastener. Sybill Patan wrote a nice review on intussusceptive branching (Patan 2005).

44

LEVIN ET AL

Shovlin: Going back to your model, you mentioned that you had performed this in an immortalized cell line. Have you looked to see whether this is just an effect of the immortalization process? Mostov: We have only used the cell line Martin developed. He used primary human dermal cells and immortalized them with telomerase. He has claimed that these are reasonably normal cells (Venetsanakos et al 2002). Augustin: You used dermal microvascular cells. These are now known to be chimeric populations of blood endothelial cells and lymphatic endothelial cells. It would be interesting to separate these cells and ask the question about which cell population is contributing to the process. Wilting: I’d like to comment on the beginning of your talk, where you mentioned the history of intussusceptive microvascular growth. Dr Patan and Dr Djonov are students of Professor Burri, an anatomist in Bern, who recently retired. He started with this concept in the lung and the chorioallantoic membrane of chicken embryos. However, if you go back into the literature you will fi nd a paper from 1900 where Professor Minot from Boston was already saying that there is growth by sinusoids (Minot 1900). This intussusceptive microvascular growth is under-represented in the literature, as you point out. Sprouting takes place in some specific places, but there is intussusception in most parenchymal organs such as liver and pancreas. Lammert: With regard to the basement membrane to the collagen IV staining that you showed, it seemed to me that the stain was within the tubes. Is that correct? Mostov: It is correct that the collagen IV staining is inside, but the structures are pillars, not tubes. The inside of pillars consists of basement membrane and smooth muscle cells, not a lumen. However, when endothelial cells are grown at low density in the same system, they form tubes. These tubes have a lumen on the inside and a basement membrane on the outside. Lammert: I think it is very interesting if it proves to be within the lumen, particularly with regard to many invertebrates where the first vessels develop from mesothelial cells, and these vessels have lumenal basement membrane. From there, a more sophisticated tubular system might have evolved with the basement membrane being outside the tubes. Mostov: Mark Krasnow works on Drosophila tracheal formation and has found a chitin ECM within the lumen (Devine et al 2005). He is looking hard for a vertebrate analogue. References Burri PH, Djonov V 2002 Intussusceptive angiogenesis—the alternative to capillary sprouting. Mol Aspects Med 23:S1–27 Devine WP, Lubarsky B, Shaw K, Luschnig S, Messina L, Krasnow MA 2005 Requirement for chitin biosynthesis in epithelial tube morphogenesis. Proc Natl Acad Sci USA 102: 17014–17019

IN VITRO INTUSSUSCEPTIVE ANGIOGENESIS

45

Minot CS 1900 On a hitherto unrecognized form of blood circulation without capillaries, in the organs of vertebrates. Proc Boston Soc Nat Hist 29:185–215 Patan S 2005 How is the branching of animal blood vessels implemented? In: Davies JA (ed) Branching morphogenesis. Springer Patan S, Tanda S, Roberge S, Jones RC, Jain RKK, Munn LL 2001 Vascular morphogenesis and remodeling in a human tumor xenograft: blood vessel formation and growth after ovariectomy and tumor implantation. Circ Res 89:732–739 Venetsanakos E, Mirza A, Fanton C, Romanov SR, Tlsty T, McMahon M 2002 Induction of tubulogenesis in telomerase-immortalized human microvascular endothelial cells by glioblastoma cells. Exp Cell Res 273:21–33

Vascular lumen formation from a cell biological perspective Tomáš Kucˇera1, Jan Eglinger1, Boris Strilic´1 and Eckhard Lammert 2 Max Planck Institute of Molecular Cell Biolog y and Genetics, Pfotenhauerstr. 108, D-01307 Dresden, Germany

Abstract. Endothelial cells are equipped with the intrinsic ability to form tubes and sprouts with a central lumen. However, the mechanisms that allow endothelial cells to form a lumen are largely unknown. We would like to discuss critically current models of vascular lumen formation and point to many unexplored and open questions. We briefly present what vascular researchers can learn from the formation of other cell systems with a lumen, such as cysts formed by Madin-Darby Canine Kidney (MDCK) cells as well as tracheae formed by Drosophila epithelial cells. In addition, we point to a number of questions that need to be addressed to understand better the cell biology that drives the formation of a vascular lumen. 2007 Vascular development. Wiley, Chichester (Novartis Foundation Symposium 283) p 46–60

More than 25 years ago, Folkman and Haudenschild demonstrated that vascular endothelial cells have all the information required for developing multicellular capillary tubes with a lumen (Folkman & Haudenschild 1980). The authors observed that capillary tubes develop from a cylindrical vacuole within an endothelial cell, and they speculated that this vacuole initiates the formation of a multicellular capillary lumen. Further insights into the formation of vascular tubes came with the discovery of specific growth factors such as the vascular endothelial growth factors (VEGF) that induce the tube formation process (Senger et al 1983, Ferrara et al 1996, Karkkainen et al 2004). In addition, junctional proteins such as the vascular endothelial cadherin, VE-cadherin, were shown to stabilize vascular tubes (Carmeliet et al 1999, Breier et al 1996). Recently, a novel extracellular protein, EGFlike domain 7, Egfl7, was discovered and shown to be required for the formation of a vascular lumen (Parker et al 2004). 1

These authors equally contributed to this manuscript. This paper was presented at the symposium by Eckhard Lammert, to whom correspondence should be addressed. 2

46

VASCULAR LUMEN FORMATION

47

Despite this remarkable progress, however, many basic questions about the tube formation process remain unanswered. To give an example, it is still not known how endothelial cells form a luminal or apical plasma membrane. Existing plasma membranes could rearrange to become the luminal plasma membrane of vascular tubes. Alternatively, the luminal plasma membrane could be formed de novo by lipid synthesis. Another open question is the fi lling of the vascular lumen. Here, we describe three different models of vascular tube formation. In addition, we briefly describe findings on formation of other tubes that could be applied to the formation of vascular tubes. Current models of vascular tube formation The following specific models have been proposed for the de novo formation of vascular tubes: (1) Vacuole formation and coalescence (Fig. 1) (2) Wrapping around extracellular space (Fig. 2) (3) Cell death and phagocytosis (Fig. 3)

FIG. 1. Vacuole formation and coalescence. (A, B) Endothelial cells, embedded in 3D extracellular matrix (grey bars), invaginate part of their basal plasma membrane. (C, D) The resulting vacuoles subsequently fuse with each other and fi nally open. (D, E) The open ends of the vacuoles adhere to the open ends of vacuoles of neighbouring cells and thereby create a lumen. Thick lines depict the membrane that is, or becomes, luminal plasma membrane. The nuclei are shown in grey.

48

KUCˇERA ET AL

FIG. 2. Wrapping around extracellular space. (A, B) Endothelial cells fi rst aggregate and then start to elongate to give rise to open multicellular structures. (C) Finally, the elongated cell sheets close to give rise to a vascular tube.

FIG. 3. Cell death and phagocytosis. (A, B) Endothelial cells, located in the middle of a multicellular sprout, undergo apoptosis. (C) The endothelial cells that form the outer cell layer partially clear the future lumen by phagocytosis of the apoptotic cells. (C, D) The phagocytic vacuoles subsequently fuse with the luminal plasma membrane and contribute to its expansion (small arrows).

Vacuole formation and coalescence (Fig. 1) This model proposes that the lumen is created by the formation of vacuoles inside endothelial cells (Folkman & Haudenschild 1980). The vacuoles within adjacent endothelial cells coalesce with each other to give rise to a continuous multicellular lumen. The model explains the observation of seamless endothelial cells in

VASCULAR LUMEN FORMATION

49

capillary tubes, which are characterized by the absence of autocellular junctions and are often found at branching points of capillaries (Bar et al 1984). In recent years, Davis and co-workers characterized the formation of vacuoles and tubes in more detail (Bayless & Davis 2002, Davis & Camarillo 1996). The authors induce vacuole and tube formation in human umbilical vein endothelial cells (HUVECs) by embedding the cells in a three-dimensional extracellular matrix in the presence of angiogenic growth factors and phorbol ester. Their data show that vacuoles are formed in endothelial cells by macropinocytic endocytosis, a process in which large areas of the plasma membrane are incorporated into the cell along with fluid-phase markers. The intracellular vacuoles that arise from this process open and either fuse with vacuoles formed in adjacent cells or close again to form a unicellular lumen with autocellular junctions. It was shown that formation of vacuoles and subsequent tube formation was dependent on integrins that interact with the surrounding extracellular matrix (Davis & Camarillo 1996). In addition, vacuole and lumen formation requires Cdc42 and Rac1 GTPases (Bayless & Davis 2002). The model implies that the luminal plasma membrane is derived from a macropinocytic up-take of the basal plasma membrane, which faces the extracellular matrix. Hence, existing plasma membrane becomes luminal plasma membrane. The model explains the role of vacuoles and why fluid-phase markers are found in the vacuoles of endothelial cells. However, the proposed model does not address a number of issues, e.g. how endothelial cells compensate for the loss of basal plasma membrane, when they remove it from the basal side and add it to the luminal side. Other questions are raised when it comes to the opening of the vacuoles. Is the opening directed towards adjacent endothelial cells? And why does the open vacuole not collapse after its opening? How is the tube stabilized? Moreover, it is unclear how apical–basal polarity is established in these tubes. Wrapping around extracellular space (Fig. 2) Several transmission electron microscopic studies on the formation of the early dorsal aorta as well as vessels of several organs and wound-healing tissues question the general applicability of the previous model, because they fail to provide evidence for vacuoles within endothelial cells during vascular lumen formation (Parker et al 2004, Hirakow & Hiruma 1983). Instead, a mechanism has been suggested by which endothelial cells elongate and surround an extracellular space that subsequently becomes the vascular lumen (Hirakow & Hiruma 1983). The model explains why vacuole formation is not a general prerequisite for vascularization and why many vessels are initially open and not closed. The model also explains the flat shape of endothelial cells and the requirement of cytoskeletal regulators, such as Cdc42, for the vascularization process (Connolly et al 2002).

50

KUCˇERA ET AL

In addition, the model is in line with the finding that most capillary endothelial cells are not seamless, but have at least one autocellular junction. However, it is unclear why endothelial cells form a tube, rather than non-tubular tissue structures. Moreover, the model does not explain how polarity is established in these tubes. Cell death and phagocytosis (Fig. 3) There are in vitro and in vivo observations (Tertemiz et al 2005, Meyer et al 1997, Fierlbeck et al 2003), suggesting that the vascular lumen is created by apoptotic death of endothelial cells that are located in the middle of multicellular cords. It was proposed that adjacent endothelial cells phagocytose the apoptotic cells. The resulting phagocytic vacuoles could subsequently exocytose towards the middle of the cord to ensure that the space, previously occupied by the dying cells, becomes the luminal space. According to this model, dying endothelial cells define the lumen, while the extracellular matrix defines the abluminal (or basal) side of the tube-forming endothelial cells. Thus, this model explains how a vascular lumen is formed and how polarity is established. However, the model is difficult to apply to the formation of capillaries, because these small vessels lack the inner mass of endothelial cells that should undergo apoptosis. Learning from MDCK cysts Endothelial and most epithelial tissues share the intrinsic ability to form lumenized structures enclosed by a monolayer of cells. Madin-Darby Canine Kidney (MDCK) is an epithelial cell line that forms cysts in response to extracellular matrix that models the architecture of polarized epithelial tissues (Zegers et al 2003). Their smooth basal plasma membrane is adjacent to a basement membrane, while their apical plasma membrane, characterized by the presence of microvilli, is adjacent to the luminal fluid (Fig. 4D). In addition, a number of proteins have been identified that specifically localize to either the apical or basal plasma membrane of polarized MDCK cells. By addition of hepatocyte growth factor (HGF), MDCK cysts develop into branching tubes (Montesano et al 1991). Despite some similarity between MDCK cysts/tubes and vascular tubes, endothelial cells differ from MDCK cells. For example, endothelial cells adopt a flat morphology and have a less structured arrangement of tight junctions and adherens junctions. Moreover, most endothelial tight junctions (with the exception of those forming the blood–brain barrier) do not seal the plasma membranes to prevent fluid exchange (Bazzoni & Dejana 2004). Endothelial cell plasma membranes can also be locally fused in the form of fenestrae, which do not exist in

VASCULAR LUMEN FORMATION

51

FIG. 4. MDCK cyst formation. MDCK cysts can be generated in vitro in two ways, namely in collagen type I (upper panel) or in suspension culture (lower panel). (A–D) MDCK cysts are usually generated in a collagen type I gel. (B) At the basal plasma membrane, adjacent to the collagen type I gel, MDCK cells form a basement membrane. (C, D) Apical plasma membrane is subsequently delivered to the cell–cell contacts in the form of vesicles, where it generates a luminal plasma membrane with microvilli. (E) In suspension culture, MDCK cells initially form a lumen that is fi lled with basement membrane components. (F, G) The cells subsequently deliver apical plasma membrane to their abluminal cell surface. (G→D) These inverted cysts establish a normal polarity upon embedding in collagen type I. Grey bars, basement membrane components. Grey shading, collagen type I.

epithelial cells. In addition, only a few microvilli are found on the apical plasma membrane of endothelial cells. Thus, findings on MDCK cells can aid, but not replace, research on vascular lumen formation. When MDCK cells are embedded in type I collagen, they direct apical membranes to the sites of cell–cell contact, as shown by the apical marker protein gp135 (Ojakian & Schwimmer 1988). They thereby create a nascent lumen that subsequently grows by directed exocytosis of further apical membrane (Fig. 4C). Interestingly, when grown in suspension culture, MDCK cysts have an inverted polarity (Fig. 4E–G). Their apical plasma membrane faces the outside, while extracellular matrix components fi ll the lumen of the cyst. When these cysts are embedded in type I collagen, the cells change their polarity, degrade the central extracellular matrix and create an apical membrane towards the lumen (Wang et al 1990a, b). When the establishment of cell–cell contacts is prevented, single MDCK cells accumulate apical markers in an intracellular compartment, termed vacuolar apical compartment (VAC) (Vega-Salas et al 1988). As soon as cell–cell contacts are formed, this compartment is secreted to give rise to an intercellular lumen,

52

KUCˇERA ET AL

indicating the importance of cellular contacts for lumen formation. The VAC might be similar to the vacuoles, which are occasionally found in endothelial cells during vascular tube formation. To explain the architecture of epithelial tubes and cysts, it was proposed that each cell in an epithelial tissue requires three distinct surfaces (O’Brien et al 2002). The adjacent extracellular matrix, cells and lumen characterize the basal, lateral and apical surfaces, respectively. The sensing of these interactions is accomplished by different cell surface receptors. Matrix detection requires integrins and dystroglycan receptors that are found on the basal surface. For signalling of cell–cell contacts, cadherins and other proteins play a role. For sensing the non-adhesive luminal surface, a small pool of integrins is present at the luminal surface (Zuk et al 1989), and an absence of integrin signalling may define the identity of the apical surface. According to this model, if a cell does not sense all three surface types, it tries to establish the missing surface by vesicular delivery to the cell–cell contacts, or by migration. If the missing surfaces cannot be established, the cell undergoes apoptosis. It is possible that similar rules apply to the formation of vascular tubes and thus explain how several mechanisms can give rise to vascular tubes. Learning from fly tracheae The tracheal system of insects and the vascular system of vertebrates are tubular systems that deliver oxygen to all tissues and organs within the body. In insects, oxygen is transported as gas, whereas in vertebrates, haemoglobin of red blood cells carries oxygen. Similar to the vascular system, the tracheal system contains tubes in different shapes, such as seamless tubes as well as tubes with one or more junctions. Unlike vascular tubes, however, no mural cells surround the tracheae. Instead, tracheal epithelial cells have a stiff extracellular chitin-containing cuticle, which stabilizes the tubes from the luminal side (Noirot et al 1979). Recently, it was shown that proteins and fi lamentous polysaccharides are secreted into the developing tracheal lumen (Fig. 5), where they are required for remodelling and shaping the developing tube ( Jazwinska et al 2003, Devine et al 2005, Tonning et al 2005). Interestingly, some of the secreted proteins contain a ZP (zona pellucida) domain ( Jazwinska et al 2003), which is also part of the endoglin protein, required for normal vascular tube formation (Li et al 1999). Mutations in the human endoglin gene cause fragile and convoluted blood vessels, thus leading to hereditary haemorrhagic telangiectasia type 1 (HHT1) (Gallione et al 1998). The polysaccharide chitin was shown to accumulate in the tracheal lumen by forming a cylindrical structure (Fig. 5B) (Devine et al 2005, Tonning et al 2005). As the lumen grows, the cylinder simultaneously expands. Loss of chitin leads to constrictions and dilations of the tracheal tubular system. Other genes have been

VASCULAR LUMEN FORMATION

53

FIG. 5. Tracheal tube formation. Chitin and apically secreted proteins in tracheal tube formation. (A) Before tube expansion/remodelling begins, a thin extracellular matrix lines the luminal plasma membrane. Subsequently, chitin and matrix are secreted towards the lumen. (B) During tube morphogenesis, a fibrous chitin-containing cylinder expands along with the surrounding epithelial cells. (C) When the lumen reaches its desired size, the expansion stops and the luminal material is completely removed. Chitin is used to build up the apical cuticle. The cuticle consists of an inner chitinous procuticle and a protein/lipid-containing outer epicuticle.

54

KUCˇERA ET AL

discovered, which have an impact on the structure of chitin, and consequently on the shape and size of the tracheal tubes (Tonning et al 2005, Luschnig et al 2006). It is possible that luminal extracellular matrix proteins and polysaccharides also play a role in the development of vascular tubes, since amorphous material was found in many capillary tubes (Folkman & Haudenschild 1980). In addition, the newly discovered Egfl7 protein could be initially luminal (Parker et al 2004). However, more detailed microscopic analyses are required to localize this protein in developing vascular tubes. Exploring vascular tube formation in vitro Despite differences in shape and function, common principles seem to be involved in the formation of all cellular tubes. Lubarsky & Krasnow (2003) therefore proposed a unifying model of tube formation, in which formation of an apical plasma membrane plays a central role during formation of all tubes. In order to look at the basic principles governing vascular tube formation, we are currently working with in vitro vascular tube formation assays. We are able to generate vascular tubes with morphology similar, but not identical, to capillaries found in vivo. Because tube formation is an intrinsic feature of endothelial cells, the in vitro assays allow us to investigate this particular feature. We are interested in the vesicular transport in endothelial cells, and we would like to explore the origin and content of vesicles that are targeted to the developing vascular lumen. Acknowledgements This work was sponsored by DFC grant Sonderforschungsbereich SFB 655 (Dresden): ‘Cells into tissues: stem and progenitor commitment and interactions during tissue formation’.

References Bar T, Guldner FH, Wolff JR 1984 ‘Seamless’ endothelial cells of blood capillaries. Cell Tissue Res 235:99–106 Bayless KJ, Davis GE 2002 The Cdc42 and Rac1 GTPases are required for capillary lumen formation in three-dimensional extracellular matrices. J Cell Sci 115:1123–1136 Bazzoni G, Dejana E 2004 Endothelial cell-to-cell junctions: molecular organization and role in vascular homeostasis. Physiol Rev 84:869–901 Breier G, Breviario F, Caveda L et al 1996 Molecular cloning and expression of murine vascular endothelial-cadherin in early stage development of cardiovascular system. Blood 87: 630–641 Carmeliet P, Lampugnani MG, Moons L et al 1999 Targeted deficiency or cytosolic truncation of the VE-cadherin gene in mice impairs VEGF-mediated endothelial survival and angiogenesis. Cell 98:147–157 Connolly JO, Simpson N, Hewlett L, Hall A 2002 Rac regulates endothelial morphogenesis and capillary assembly. Mol Biol Cell 13:2474–2485

VASCULAR LUMEN FORMATION

55

Davis GE, Camarillo CW 1996 An alpha 2 beta 1 integrin-dependent pinocytic mechanism involving intracellular vacuole formation and coalescence regulates capillary lumen and tube formation in three-dimensional collagen matrix. Exp Cell Res 224:39–51 Devine WP, Lubarsky B, Shaw K et al 2005 Requirement for chitin biosynthesis in epithelial tube morphogenesis. Proc Natl Acad Sci USA 102:17014–17019 Ferrara N, Carver-Moore K, Chen H et al 1996 Heterozygous embryonic lethality induced by targeted inactivation of the VEGF gene. Nature 380:439–442 Fierlbeck W, Liu A, Coyle R, Ballermann BJ 2003 Endothelial cell apoptosis during glomerular capillary lumen formation in vivo. J Am Soc Nephrol 14:1349–1354 Folkman J, Haudenschild C 1980 Angiogenesis in vitro. Nature 288:551–556 Gallione CJ, Klaus DJ, Yeh EY et al 1998 Mutation and expression analysis of the endoglin gene in hereditary hemorrhagic telangiectasia reveals null alleles. Hum Mutat 11:286–294 Hirakow R, Hiruma T 1983 TEM-studies on development and canalization of the dorsal aorta in the chick embryo. Anat Embryol (Berl) 166:307–315 Jazwinska A, Ribeiro C, Affolter M 2003 Epithelial tube morphogenesis during Drosophila tracheal development requires Piopio, a luminal ZP protein. Nat Cell Biol 5:895–901 Karkkainen MJ, Haiko P, Sainio K et al 2004 Vascular endothelial growth factor C is required for sprouting of the fi rst lymphatic vessels from embryonic veins. Nat Immunol 5:74–80 Li DY, Sorensen LK, Brooke BS et al 1999 Defective angiogenesis in mice lacking endoglin. Science 284:1534–1537 Lubarsky B, Krasnow MA 2003 Tube morphogenesis: making and shaping biological tubes. Cell 112:19–28 Luschnig S, Batz T, Armbruster K, Krasnow MA 2006 Serpentine and vermiform encode matrix proteins with chitin binding and deacetylation domains that limit tracheal tube length in Drosophila. Curr Biol 16:186–194 Meyer GT, Matthias LJ, Noack L, Vadas MA, Gamble JR 1997 Lumen formation during angiogenesis in vitro involves phagocytic activity, formation and secretion of vacuoles, cell death, and capillary tube remodelling by different populations of endothelial cells. Anat Rec 249:327–340 Montesano R, Schaller G, Orci L 1991 Induction of epithelial tubular morphogenesis in vitro by fibroblast-derived soluble factors. Cell 66:697–711 Noirot C, Smith DS, Cayer ML, Noirot-Timothee C 1979 The organization and isolating function of insect rectal sheath cells: a freeze-fracture study. Tissue Cell 11:325–336 O’Brien LE, Zegers MM, Mostov KE 2002 Opinion: Building epithelial architecture: insights from three-dimensional culture models. Nat Rev Mol Cell Biol 3:531–537 Ojakian GK, Schwimmer R 1988 The polarized distribution of an apical cell surface glycoprotein is maintained by interactions with the cytoskeleton of Madin-Darby canine kidney cells. J Cell Biol 107:2377–2387 Parker LH, Schmidt M, Jin SW et al 2004 The endothelial-cell-derived secreted factor Egfl7 regulates vascular tube formation. Nature 428:754–758 Senger DR, Galli SJ, Dvorak AM et al 1983 Tumor cells secrete a vascular permeability factor that promotes accumulation of ascites fluid. Science 219:983–985 Tertemiz F, Kayisli UA, Arici A, Demir R 2005 Apoptosis contributes to vascular lumen formation and vascular branching in human placental vasculogenesis. Biol Reprod 72:727–735 Tonning A, Hemphala J, Tang E et al 2005 A transient luminal chitinous matrix is required to model epithelial tube diameter in the Drosophila trachea. Dev Cell 9:423–430 Vega-Salas DE, Salas PJ, Rodriguez-Boulan E 1988 Exocytosis of vacuolar apical compartment (VAC): a cell-cell contact controlled mechanism for the establishment of the apical plasma membrane domain in epithelial cells. J Cell Biol 107:1717–1728 Wang AZ, Ojakian GK, Nelson WJ 1990a Steps in the morphogenesis of a polarized epithelium. I. Uncoupling the roles of cell-cell and cell-substratum contact in establishing plasma membrane polarity in multicellular epithelial (MDCK) cysts. J Cell Sci 95:137–151

56

KUCˇERA ET AL

Wang AZ, Ojakian GK, Nelson WJ 1990b Steps in the morphogenesis of a polarized epithelium. II. Disassembly and assembly of plasma membrane domains during reversal of epithelial cell polarity in multicellular epithelial (MDCK) cysts. J Cell Sci 95:153–165 Zegers MM, O’Brien LE, Yu W, Datta A, Mostov KE 2003 Epithelial polarity and tubulogenesis in vitro. Trends Cell Biol 13:169–176 Zuk A, Matlin KS, Hay ED 1989 Type I collagen gel induces Madin-Darby canine kidney cells to become fusiform in shape and lose apical-basal polarity. J Cell Biol 108:903–919

DISCUSSION Davies: You used a dextran as a marker. But if the cell were releasing this randomly to all of its plasma membrane, just to offload its vesicles, would you notice that it ever got outside the cell? It’s obvious that it would get trapped in the lumen because it is sealed in. But if it were being released on the basal side of the cell, would you still see it hanging around or would it just diffuse into your bulk culture medium? Lammert: We are using a diluted matrigel. The Lucifer yellow dye goes away. The dextran is always a bit sticky, so I would say it goes away to some extent, but also sticks a bit. It is a macromolecule that gets trapped in extracellular matrix (ECM). Davies: So you would see a difference between the vesicles being offloaded randomly versus being targeted. Lammert: Yes. Ye: Your cholera toxin B labelling suggests that the membrane is turning over, and the old membrane always ends up on the lumenal side. The new membrane is on the outside. Can your result be explained by a directional turnover of the plasma membrane? Lammert: Cholera toxin B is labelling parts of the lumen rather than the lumenal plasma membrane. It is not a simple taking in of membrane and then just adding it to the other side. In general, with endothelial cells we find that many of the membrane dyes quickly end up in the degradative endocytic compartment. It is quite difficult to label endothelial membrane with lipophilic dyes because the membranes are so rapidly endocytosed. Ye: So could there be something in the lumenal matrix that is stabilizing the vesicles from the cells? Lammert: Yes, that is what we think. Vargesson: You suggested four models: can’t all four be active at once? Does it have to be just one? How does this happen in epithelial morphogenesis? Is there one model for lumen formation in epithelial tubes? Mostov: Many of these things are observed in different systems. Neurulation is a big wrapping around. One of the things I would like to know is what are the common principles of these four different models? There is plasticity: there are in vitro culture models that can be shifted from one to another.

VASCULAR LUMEN FORMATION

57

Wilkins: Another way of putting this is that these mechanisms do exist, and different ones are used in different circumstances. Mostov: In some cases we know a lot more about in vitro, in others we know more about in vivo. In epithelial morphogenesis you can see things like neurulation, which is a wrapping-around process. It is still controversial as to whether these vacuolar-type fusion processes take place in epithelia in vivo, but this has been seen for a long time in different in vitro or pathological circumstances. In epithelial cells the apical surface differs in composition very distinctly from the basolateral surface. New membrane components are made and may be added or substituted into the apical surface. This could happen at different stages. If you have intracellular vacuoles, there could be substitution by newly made proteins in the Golgi going to them and other proteins being removed from them. You see very large vacuoles in this George Davis-sort of model, and these large vacuoles can also be seen in experimental manipulation of epithelial cells. This is essentially vesicular transport from the Golgi or endosomes to what becomes luminal apical membrane. In some cases these vesicles might fuse to form macroscopic vesicles. Wilting: Is there a role for growth factors in lumen formation? VEGF knockout embryos have defective lumen formation in the dorsal aorta. Is proliferation required for lumen formation? Lammert: That’s an interesting question. We haven’t done a knockout or knock down of the VEGF receptor. We have added VEGF and FGF to our assays. We then see a higher efficiency of tube formation. It would be interesting to know the role of VEGF. Does it play a role in matrix production, or endocytosis, or exocytosis? Wilting: What can we learn about lumen formation from in vitro studies? Lumen formation by vacuoles came up with the first in vitro studies by Judah Folkman in the 1970s. To the best of my knowledge, no one has described lumen formation by vacuoles in vivo. I have always considered this to be an in vitro artefact. You show collagen IV within the lumen in the vessel, but this is a typical basal membrane protein. As long as you cannot clearly demarcate what is luminal and what is abluminal in vitro, to me this looks like an inside-out vessel. Lammert: We are trying to create a model that is consistent with our experimental data. The unifying model that the apical membrane formation is the driving force for a developing lumen is not consistent with what we see. Even though there are many tubes and cavities where the apical membrane is the lumenal membrane, it is not always the case. The blastocyst is a good example. It is a cavity and the basement membrane is on the luminal side. We inject the ES cells into this lumen. The apical side and also the junctions are towards the outside of the blastocyst. The same is true for MDCK cells, when these cells are grown in a suspension culture. The MDCK cells develop cysts with an inverted polarity as shown by Nelson and colleagues (Wang et al 1990). The first vessels that are observed in invertebrates

58

KUCˇERA ET AL

similarly have a lumenal basement membrane. The first vascular lumen seemed to have formed by the coelomic mesothelium: the apical side of the mesothelium faces the coelomic cavity, while its basolateral side with basement membrane faces the vascular lumen. This creates a vascular lumen with an inverted polarity. In answer to the idea that vacuoles are an in vitro artefact, I don’t think this is the case and I would like to refer you to a recent study (Kamei et al 2006). The reason why we went for an in vitro assay is that it gives us a simple system that is likely to reveal basic principles. Betsholtz: Has anyone looked at collagen IV deposition during yolk sac vascular development? This could perhaps be an easy way to address this question. Wilting: In the yolk sac there are blood islands, which is another model of lumen formation. There are cells inside that will become blood cells and cells outside that will become endothelial cells. You don’t need apoptosis of the inside cells. Eichmann: One of the things seen in quail–chick chimeras is that migrating angioblasts get incorporated into tubes. Have you seen this in your models? Could you add extra cells to see whether they will become incorporated into an alreadyformed tube? Lammert: So what you see is that there are blood cells being incorporated during tube formation? Drake: From the mesoderm there is a birth of an angioblast. Then the next phase is an aggregation of two to three angioblasts. We don’t have clear lumenal markers, but we can do confocal optical sections. At that point the angioblasts have no cell–cell adhesion markers. PECAM and CD34 are present and there is the first sense of lumen formation. It can be from three to nine cells which build this small vascular segment. We don’t see other cells join. It would be extremely difficult without a dye, but when we optically section, if there is a lumen and there is some point in the optical section where two labels are seen, this is indicative of a lumen. Eichmann: In quail–chick chimeras there are some cells that are incorporated somehow. I don’t know how this actually happens. Drake: Those cells come through the circulation and join the lumenized vessel. We labelled the blood islands before they were blood islands and asked whether they contributed circulating cells to angioblasts, and indeed they did. They come through the circulation, through the lumenal side, and join the endothelium. Ruhrberg: From a cell biological perspective, the model you described looks like a vesicle is endocytosed, only then to be recycled. Have you stained with markers to characterize these vesicles? The reason I am asking is that Elisabetta Dejana described that VEGFR2 is being internalised, but not necessarily degraded. There is evidence to suggest that VEGFR2 internalization occurs in a pathway that is distinct from the EGF pathway (Gampel et al 2006). The EGF-containing vesicles are labelled with markers for the endocytotic and lysosomal pathway.

VASCULAR LUMEN FORMATION

59

Lammert: We are looking into this. The compartment where the albumin gold particles and the dextran are found is a degradative pathway. However, this pathway is at some stage used for contributing to the tube-formation process. Over time we see an accumulation of lysosomal proteins as well as degradative proteins in the dextran cholera toxin-labelled compartments. Ruhrberg: Are proteins such as VEGFR2 endocytosed and degraded or recycled? Lammert: We haven’t looked at this yet. Shovlin: What does this tell us about the development of polarity in these cells? If polarity wasn’t established, presumably these vesicles would never fuse. Are any markers of polarity present? Lammert: There is basal polarity, which is marked by laminins and collagen IV. In the matrigel tubes we see that the basal ECM markers are lumenal. This is consistent with the primitive tubes people have seen in some invertebrates, and the dorsal vessel of the Drosophila heart. In our case, there is still a signal missing that establishes polarity after the initial tube formation. With our matrigel model we are able to explain the tube structure and why endocytosed material is ending up in the lumen. But with our in vitro data we are still unable to explain the step that creates polarity. Dejana: What is the role of the 3D matrix? Could it be that when the cells are in contact with the medium (apical side) and with their matrix (basal side) there is a directional trafficking of vesicles from the apical to the basal side and from the basal to the apical. In a 3D matrix the apical side becomes a basal side so the cells receive a different type of signal. The vesicles get into the centre of the cells because they cannot find a direction since the cells are not polarized. In the end these vesicles would fuse because they cannot be directed. The matrigel is certainly not a typical 3D matrix, but it is a common experience that tubes are formed when endothelial cells are put in a 3D matrix. When the endothelial cells are migrating in a 3D matrix they will be forming tubes. When they are in large vessels they are already polarized and tubes cannot be formed. Lammert: We have also repeated these experiments in a sandwich. It works. Dejana: It is a matter of how the cells are oriented, because you create only basal membrane, and they have to recreate the apical membrane inside. Lammert: That is what we know from the MDCK cells. Augustin: Expanding on the issue of polarity, a simple experiment would be to stimulate these tubes for a short period with tumour necrosis factor (TNF) α after establishment, and then to look at where inducible adhesion molecules are expressed a few hours later. I would be concerned if ICAM-1 and VCAM-1 are not expressed luminally in this model. Lammert: We haven’t done the TNFα experiment yet.

60

KUCˇERA ET AL

Yancopoulos: I wanted to respond to some of the questions about growth factors and how these may contribute to these processes. A couple of these mechanisms might have shared mediators or molecules. I was struck by the potential role of the Notch pathway here. During embryonic development in the Dll4 knockout the aorta has defects in the lumen formation, and the picture that we published looked like your cartoons in the model where cells coalesce, come around and close off. If we section up and down along the aorta, we see what looks like an intact endothelial circle that is perfectly closed, and other places where it is opened and is not connected. There are other places with collections of endothelial cells having no shape. Interestingly, the venous side of the circulation in the same animals looks as if lumen formation is totally normal. Dll4 seems to be playing a role: it is clearly induced dramatically by VEGF. In the yolk sac of the Dll4 knockouts, instead of a normal capillary plexus there are two sheets of endothelium almost unconnected. This could be more of an intussusception type of model, with defects in how the sheets would normally connect to form discrete vascular structures. A growth factor pathway may be contributing in a variety of mechanisms to lumen formation. Kitajewski: It has been published that the right balance of Notch activity, in the in vitro cord formation models, is required for proper capillary-like plexus formation. Too little, and the cells coalesce and don’t form cords well; too much and the cords blow apart. The cells don’t come together properly. This may be similar to the effects seen in animals. Mostov: In the epithelial lumen formation work, the availability of good markers for developing apical and basal membrane has been critical to the field. What do we really know about polarity markers in endothelial cells? Betsholtz: Not that much. Gp135 is one. Mostov: Gp135 seems to have a unique mechanism for polarization, and it may not be reliable. References Gampel A, Moss L, Jones MC, Brunton V, Norman JC, Mellor H 2006 VEGF regulates the mobilization of VEGFR2/KDR from an intracellular endothelial storage compartment. Blood 108:2624–31 Kamei M, Saunders WB, Bayless KJ, Dye L, Davis GE, Weinstein BM 2006 Endothelial tubes assemble from intracellular vacuoles in vivo. Nature 442:453–456 Wang AZ, Ojakian GK, Nelson WJ 1990 Steps in the morphogenesis of a polarized epithelium. I. Uncoupling the roles of cell-cell and cell-substratum contact in establishing plasma membrane polarity in multicellular epithelial (MDCK) cysts. J Cell Sci 95:137–151

The genetics of vasculogenesis Christopher J. Drake, Paul A. Fleming and W. Scott Argraves Department of Cell Biolog y and Anatomy, Medical University of South Carolina, Charleston, SC 29425, USA

Abstract. To identify genes important to the process of vasculogenesis, we have used a novel meta-analysis approach to evaluate retrospectively the embryonic vascular anomalies observed in over 100 mouse gene knockout studies. Through application of this method, termed Approach for Ranking of Embryonic Vascular Anomalies (AREVA), 12 genes were determined to be critical to vasculogenesis. Importantly, when the 12 genes were considered with respect to VEGF–VEGFR signalling, an integrated network centreing on the ShcA/Ras/Raf/Mek/Erk pathway became apparent. Herein, we discuss how the 12 vasculogenesis-critical genes influence specific stages in the process of vasculogenesis. 2007 Vascular development. Wiley, Chichester (Novartis Foundation Symposium 283) p 61–76

In the 11 years since the genetic ablation of the murine vascular endothelial growth factor (VEGF) receptor 2/Flk1 gene was reported (Shalaby et al 1995), numerous other genetic mutants have been described that display vascular anomalies that directly or indirectly have been implicated in the misregulation of VEGF signalling (Argraves & Drake 2005). However, an integrated signalling pathway that mediates key stages in the process of vasculogenesis has not come to light from these studies. Through retrospective analysis of murine knockouts having vascular anomalies, our recent study (Argraves & Drake 2005) has identified a set of 12 genes that are critical to vasculogenesis. By functional categorization of the genes, a canonical VEGF signalling pathway has emerged. In the present paper we discuss the process by which the genes were identified, their relationship to VEGF signalling and the impact of their genetic deficiency on discrete phases of vasculogenesis. The critical role of VEGF signalling in vasculogenesis The first indications that the VEGF–VEGF receptor signalling pathway played a central role in mediating vasculogenesis came from mouse gene-targeting studies that examined the consequence of disrupting genes for the VEGF receptors. Deletion of the VEGF receptor, Flk1/VEGFR2, resulted in arrested 61

62

DRAKE ET AL

vasculogenesis due to a failure in the generation of endothelial precursor cells (angioblasts) (Shalaby et al 1995). Deletion of Flt1/VEGFR1 also resulted in vasculogenesis anomalies but the defects were less severe than that observed in Flk1 knockouts (Fong et al 1995). These studies were followed by targeted deletion of the VEGFA gene which was both heterozygous and homozygous embryonic lethal (Ferrara et al 1996, Carmeliet et al 1996). This was a remarkable outcome, considering that other VEGF family members (VEGFC) are likely still expressed in the heterozygous embryos. Identification of genes critical to vasculogenesis Considering the various cell behaviours that constitute vasculogenesis, it was striking that of the 100 targeted gene deletions exhibiting vascular anomalies, only three of those genes had been reported to be critical to vasculogenesis; VEGFR2/ Flk1, fibronectin and cytochrome P450 reductase/Cpr (Shalaby et al 1995, George et al 1997, Otto et al 2003). To evaluate the possibility that additional genes were in fact critical to vasculogenesis, we reassessed the reported vascular defects in the 100 targeted gene deletion studies. To do this, we developed the Approach for Ranking of Embryonic Vascular Anomalies (AREVA) (Argraves & Drake 2005). AREVA was designed to tabulate a numeric score based on the state of vascular morphogenesis at three developmental stages and at three specific sites of vasculogenesis (i.e. yolk sac, endocardium and paired dorsal aortae). The scores, positive for vascular defects and negative for normal blood vessels, were temporally weighted such that the earlier a defect occurred, the higher the score. Based on AREVA analysis, 12 genes (Fibronectin (George et al 1997), Flt1/VEGFR1 (Fong et al 1995), Flk1/VEGFR2 (Shalaby et al 1995), α5 integrin (Francis et al 2002), Tek/Tie2 (Dumont et al 1994, Sato et al 1995), VE-cadherin (Carmeliet et al 1999, Gory-Faure et al 1999), VEGFA (Ferrara et al 1996, Carmeliet et al 1996), Connexin 45 (Kruger et al 2000), ShcA (Lai & Pawson 2000), Cytochrome P450 reductase (Otto et al 2003), CD148/DEP-1 (Takahashi et al 2003) and EphrinB2 (Wang et al 1998, Adams et al 1999, Gerety & Anderson 2002), were identified as being critical to vasculogenesis. Functional characterization of vasculogenic genes reveals relationships to VEGF signalling As a result of expanding the number of vasculogenic genes from three to 12, functional relationships between the entire vasculogenic gene set became apparent. For example, 10 of the 12 genes (Fibronectin, VEGFR1/Flt1, VEGFR2/Flk1, α5 integrin, VE-cadherin, VEGFA, ShcA, Cytochrome P450 reductase, CD148/ DEP-1 and EphrinB2) were clearly linked to the regulation of the VEGF–VEGFR signalling pathway. Support for this linkage are the following findings: (1) cyto-

GENETICS OF VASCULOGENESIS

63

chrome P450 reductase has been shown to promote the expression of both VEGF and Flk1 (Otto et al 2003); (2) VEGF–VEGFR signalling augments fibronectin (FN) expression and thus, FN-mediated signalling via the fibronectin-binding integrin receptor, α5β1 (Kazi et al 2004); (3) CD148/DEP-1 negatively regulates VEGF–VEGFR signalling (Takahashi et al 2003); and (4) EphrinB2 has also been shown to suppress VEGF signalling (Kim et al 2002, Kertesz et al 2006). A number of other genes whose AREVA scores were not sufficient to categorize them as critical to vasculogenesis are also functionally related to the VEGF signalling pathway. These include Tal1/SCL (Shivdasani et al 1995, Robb et al 1995), Notch 1 (Swiatek et al 1994), presenilins 1 and 2 (Donoviel et al 1999) and neuropilins 1 and 2 (Takashima et al 2002). Relationships of these genes to the VEGF signalling pathway include: (1) VEGF up-regulates Tal1/SCL expression (Giles et al 2005); (2) neuropilins 1 and 2 are VEGF receptors (Takashima et al 2002); and (3) Notch 1 and presenilins 1 and 2 are key components of the Notch 1 signalling pathway which, upon activation, down-regulates expression of Flk1 (Taylor 2002). It is possible that through further phenotypic analysis of these knockout embryos, new information will emerge that will increase their AREVA scores and place the genes among the set of genes critical to vasculogenesis. VEGF signalling via the ShcA/Ras/Raf/Mek/Erk signalling cascade In assessing functional relationships between the vasculogenesis genes, it also became evident that they act to regulate an integrated network involving VEGF and its receptors via the ShcA/Ras/Raf/Mek/Erk cascade (Fig. 1). For example, VEGF/Flk1 signalling induces phosphorylation of Shc isoforms and the formation of a Shc–Grb2 complex (Kroll & Waltenberger 1997). Similarly, FN ligation with its receptor, α5β1, leads to activation of Shc via FAK and Src phosphorylation (Schlaepfer et al 1998). VEGF, presumably acting via Flk1, also acts to induce Shc association with VE-cadherin (Zanetti et al 2002). EphrinB2 ligation to one of its receptors inhibits VEGF-induced Ras/mitogen-activated protein kinase (MAPK) activation (Kim et al 2002). While genetic knockout studies strongly support the importance of the ShcA/Ras/Raf/Mek/Erk cascade in VEGF signalling, the interpretations must be reconciled with biochemical findings which implicate PKC-Raf-MEK-MAPK as the cascade responsible for mediating VEGF signalling (Zachary 2003, Shibuya & Claesson-Welsh 2006). Involvement of vasculogenesis critical genes in the regulation of cell behaviours/processes during vasculogenesis There are numerous discrete events and cellular behaviours that occur during the process of vasculogenesis (Fig. 2). These include, (1) lineage formation (i.e. the birth of angioblasts/endothelial cell progenitors from undifferentiated mesoderm)

64

DRAKE ET AL

FIG. 1. Diagram depicting the 12 vasculogenesis-critical genes in the context of canonical signalling pathways. Genes highlighted in grey are those genes previously categorized by AREVA as playing critical roles in vasculogenesis (Argraves & Drake 2005). Flt1 is displayed twice since there are distinct functions of both membrane and soluble forms of Flt1.

(Giles et al 2005); (2) the migration of angioblasts (Argraves et al 2004); (3) the aggregation of angioblasts into cord-like structures (Drake et al 1997); (4) lumen formation (Drake et al 1992); (5) network formation (Drake et al 1995, 2000, Vernon et al 1995), (6) vascular fusion (i.e. the generation of larger diameter vessels from primary networks); and (7) blood vessel stabilization (Argraves et al 2002, Crosby et al 2005). In the following sections we will discuss the contribution of the vasculogenic genes to these processes. In many cases, a relationship between a vasculogenic gene and a discrete aspect of vasculogenesis is clearly evident; however, there are aspects of vasculogenesis in which there is no obvious relationship with a gene in the set (e.g. lumen formation). This implies that there are other genes whose functions are critical in these stages of vasculogenesis.

GENETICS OF VASCULOGENESIS

65

FIG. 2. Stages in the process by which angioblasts assemble to form nascent blood vessels and primary vascular networks. Shown in A–E are progressive stages in vasculogenesis occurring in quail embryos double labelled to detect Tal1/SCL protein expression in the nucleus (intense white labelling in A–E) and QH1 antigen expression on the cell surface (brighter grey labelling surrounding Tal1/SCL positive nuclei in B–E). Below the panels is a series of cartoons summarizing salient morphological events occurring in each of the panels.

The formation of the endothelial lineage VEGFA/Flk1 signalling is required to generate the endothelial lineage, as is evidenced by the complete lack of angioblasts in Flk1 null mice (Shalaby et al 1995) and the fact that deletion of Flt1, which is now viewed as a VEGF antagonist, enhanced endothelial lineage commitment (Fong et al 1999). While homozygous and heterozygous VEGFA knockouts are both early embryonic lethal (Ferrara et al 1996, Carmeliet et al 1996), it remains to be formally established that the deficiency impacts the lineage. How VEGF–VEGFR signalling may act to generate the endothelial lineage is suggested by the recent study of Giles et al (2005) in which newly gastrulating mesoderm was exposed to elevated levels of VEGF. This treatment induced Flk1-positive mesodermal cells to ectopically express Tal1, a helix-loop-helix transcription factor known to be expressed by angioblasts (Drake et al 1997). This finding suggests that Tal1 may be a downstream target of VEGF– VEGFR signalling that acts in lineage formation. Another of the vasculogenesis genes, cytochrome P450 reductase (cpr450), also appears to exert its effects at the level of lineage formation. This is suggested by the fact that the phenotype of cpr450 nulls (Otto et al 2003) closely resembles that of Flk1 nulls (Shalaby et al 1995). Based on depressed levels of both VEGFA and Flk1 observed in cpr450 nulls, cpr450 likely acts to positively regulate both VEGFA and Flk1 via its affects on retinoic acid (RA) homeostasis. A relationship between RA homeostasis, VEGF and vasculogenesis is also supported by our studies in which vascular network formation was disrupted in RA-deficient quail embryos due to insufficient angioblast numbers (LaRue et al 2004).

66

DRAKE ET AL

Angioblast and endothelial cell proliferation and survival Once the endothelial lineage is formed, angioblasts coalesce and differentiate to form endothelial cells. The proliferation of endothelial cells is a key component of the process of vasculogenesis, which is mediated principally through the action of VEGF. Of the 12 vasculogenesis gene knockouts, only one study (DEP1/ CD148) evaluated the consequences in terms of endothelial cell proliferation. Mice deficient in DEP1/CD148 exhibit increased VEGF levels and endothelial cell numbers attributed to cell proliferation (Takahashi et al 2003). A consequence of augmented endothelial cell proliferation in these mice was the formation of large sinusoidal vessels. This pathological process of sinusoidal blood vessel formation is analogous to the normal process of vascular fusion by which large vessels and vascular sinuses form during vasculogenesis. This process came to light in studies evaluating the consequences of elevated VEGFA levels on vasculogenesis. Administration of VEGF to early stage quail embryos induced small diameter vessels to fuse and form larger diameter vessels. Normally, this process is tightly regulated and underlies the formation of large diameter embryonic blood vessels including the dorsal aorta, endocardium and sinus venosus (Drake & Little 1995, 1999). Of the 12 genes designated as being critical to vasculogenesis, targeted deletion of 9 of these genes (FN, α5 integrin, VEGFR1/Flt1, connexin 45, VEGFA, Tek/Tie2, ShcA, CD148/DEP1 and EphrinB2) leads to the formation of sinusoidal blood vessels in the yolk sac. There are several mechanisms by which the deletion of a vasculogenesis gene could cause hypervascular fusion. The first is that the gene is a negative regulator of VEGF signalling via the ShcA/Ras/Raf/Mek/Erk signalling cascade leading to endothelial cell proliferation. For example, since connexin 45 can suppress both Raf and Erk activation (Stains & Civitelli 2005), deletion of connexin 45 can be expected to lead to increased flux through the VEGF/Flk1/ShcA/Ras/Raf/Mek/ Erk signalling cascade leading to increased numbers of endothelial cells. Additionally, since Ephrin B2 acts to suppress Ras, its deletion would also be expected to lead to increased flux through the pathway. Another possible mechanism by which a vasculogenesis gene could lead to hypervascular fusion is that the gene deficiency may lead to a reduction in the number of angioblasts. Deficiency in the angioblast numbers would be expected to lead to inadequate vascularization of embryonic tissues resulting in hypoxia and leading to up-regulation of VEGF expression. For example, haploinsufficiency of VEGFA may prevent generation of sufficient numbers of angioblasts leading to inadequate circulation and hypoxia-induced expression of VEGFA from the wild-type allele. Angioblast motility and filopodial formation/function The extent to which individual cell movement is critical to the process of vasculogenesis is not well understood. At least one phase in vasculogenesis in which

GENETICS OF VASCULOGENESIS

67

movement appears to be important is the coalescence of angioblasts into aggregates that elongate. Two vasculogenesis genes, FN and the α5 subunit of its integrin receptor, are linked to cell adhesion and motility. However, at fi rst glance, the phenotype of mice deficient in the expression of these genes (i.e. swollen vessels and reduced vessel complexity, Francis et al 2002, George et al 1997) do not reflect a defect in cell motility. The aberrant phenotypes are consistent with augmented VEGF expression and formation of sinusoidal blood vessels, a defining characteristic of the phenomenon of VEGF-induced hypervascular fusion (Drake & Little 1995, 1999). One connection between VEGFA-induced hypervascular fusion and cell motility is that endothelial cells engaged in hypervascular fusion exhibit enhanced protrusive activity (Drake & Little 1995, 1999). This protrusive activity is rather ill defined. Some of the protrusive activity may be likened to lamellipodia/ pseudopodia, which are associated with motility, while other protrusions may be likened to fi lopodia, which are characteristic features of specialized endothelial cells known as tip cells. Tip cells engage in the process of vascular network expansion and fi lopodia are associated with guidance functions (Eichmann et al 2005). While there is ample literature implicating FN and α5β1 in cell motility, the involvement of these proteins in guidance activity of fi lopodia remains to be elucidated. One of the 12 vasculogenesis genes, EphrinB2, has been implicated in the guidance activity of tip cells (Heroult et al 2006). Like the α5β1 and FN nulls, EphrinB2 nulls also display hypervascular fusion. Based on these findings, it can be hypothesized that α5β1, FN and EphrinB2 act to suppress excessive protrusive activity that would otherwise lead to hypervascular fusion. The concept that α5β1 might act to suppress excessive VEGF signalling is consistent with several other findings. First, Wijelath et al (2002) have shown a direct physical association between Flk1 and α5β1. Similarly, Borges et al (2000) have shown that αvβ3 directly interacts with Flk1. Furthermore, Reynolds et al (2002) showed that mice lacking β3 integrin exhibited increased Flk1 expression and were more responsive to exogenous VEGF in terms of angiogenic blood vessel formation. The interrelationships between these proteins and the VEGF–VEGFR signalling pathway are made all the more interesting by the fact that this pathway appears not only to be regulated by integrins, but itself regulates expression of FN (Kazi et al 2004). Stabilization of nascent blood vessels The final stage of vasculogenesis involves the process by which nascent endothelial tubes become stabilized. At least two of the vasculogenesis genes, VE-cadherin and DEP1/CD148, appear critical for this process. In a recent study done in collaboration with Dr Elisabetta Dejana (FIRC Institute of Molecular Oncology, Milan, Italy), we evaluated the role of VE-cadherin in vasculogenesis. Remarkably,

68

DRAKE ET AL

blood vessels formed in VE-cadherin null embryos and then rapidly disassembled. The findings demonstrated that VE-cadherin was not required for the initial assembly of blood vessels, but was required for the maintenance/stability of blood vessels (Crosby et al 2005). Reduced VEGF signalling appears to accompany vascular maturation. This is supported by findings showing that while VEGF antagonist treatment elicits pronounced effects on early phases of vasculogenesis, it has little effect on the morphology of matured embryonic blood vessels (Argraves et al 2002). Emerging evidence indicates that VE-cadherin may be playing a role in suppressing the VEGF signalling pathway as blood vessels mature. Grazia Lampugnani et al (2003) showed that the process by which VEGF-induced endothelial cell proliferation is suppressed through cell–cell contact requires VE-cadherin, DEP1/CD148 and β -catenin. In light of evidence that VE-cadherin can interact with DEP1/ CD148 and Flk1, it is reasonable to speculate that VE-cadherin expression in maturing blood vessels leads to dephosphorylation of Flk1, resulting in suppression of VEGF signalling. Summary Through retrospective analysis of embryonic vascular anomalies resulting from targeted deletion of over 100 murine genes, a group of genes was identified whose functions are critical to vasculogenesis. Functional categorization of these genes reveals an integrated signalling network, the essential purpose of which appears to be control of VEGF signalling via the ShcA/Ras/Raf/Mek/Erk signalling cascade. Deficiency in the genes can lead to decreased or increased flux through this cascade that can disrupt key events during the process of vasculogenesis ranging from failure to form adequate numbers of angioblasts to inability to stabilize nascent blood vessels. Acknowledgements This work was supported by NIH grants HL57375 (CJD) and HL061873 (WSA).

References Adams RH, Wilkinson GA, Weiss C et al 1999 Roles of ephrinB ligands and EphB receptors in cardiovascular development: demarcation of arterial/venous domains, vascular morphogenesis, and sprouting angiogenesis. Genes Dev 13:295–306 Argraves WS, Drake CJ 2005 Genes critical to vasculogenesis as defi ned by systematic analysis of vascular defects in knockout mice. Anat Rec A Discov Mol Cell Evol Biol 286:875–884 Argraves WS, Larue AC, Fleming PA, Drake CJ 2002 VEGF signaling is required for the assembly but not the maintenance of embryonic blood vessels. Dev Dyn 225:298–304

GENETICS OF VASCULOGENESIS

69

Argraves KM, Wilkerson BA, Argraves WS et al 2004 Sphingosine-1-phosphate signaling promotes critical migratory events in vasculogenesis. J Biol Chem 279:50580–50590 Borges E, Jan Y, Ruoslahti E 2000 Platelet-derived growth factor receptor beta and vascular endothelial growth factor receptor 2 bind to the beta 3 integrin through its extracellular domain. J Biol Chem 275:39867–39873 Carmeliet P, Ferreira V, Breier G et al 1996 Abnormal blood vessel development and lethality in embryos lacking a single VEGF allele. Nature 380:435–439 Carmeliet P, Lampugnani MG, Moons L et al 1999 Targeted deficiency or cytosolic truncation of the VE-cadherin gene in mice impairs VEGF-mediated endothelial survival and angiogenesis. Cell 98:147–157 Crosby CV, Fleming PA, Argraves WS et al 2005 VE-cadherin is not required for the formation of nascent blood vessels but acts to prevent their disassembly. Blood 105:2771–2776 Donoviel DB, Hadjantonakis AK, Ikeda M et al 1999 Mice lacking both presenilin genes exhibit early embryonic patterning defects. Genes Dev 13:2801–2810 Drake CJ, Little CD 1995 Exogenous vascular endothelial growth factor induces malformed and hyperfused vessels during embryonic neovascularization. Proc Natl Acad Sci USA 92:7657–7661 Drake CJ, Little CD 1999 VEGF and vascular fusion: implications for normal and pathological vessels. J Histochem Cytochem 47:1351–1356 Drake CJ, Davis LA, Little CD 1992 Antibodies to beta 1-integrins cause alterations of aortic vasculogenesis, in vivo. Dev Dyn 193:83–91 Drake CJ, Cheresh DA, Little CD 1995 An antagonist of integrin alpha v beta 3 prevents maturation of blood vessels during embryonic neovascularization. J Cell Sci 108:2655–2661 Drake CJ, Brandt SJ, Trusk TC, Little CD 1997 TAL1/SCL is expressed in endothelial progenitor cells/angioblasts and defi nes a dorsal-to-ventral gradient of vasculogenesis. Dev Biol 192:17–30 Drake CJ, LaRue A, Ferrara N, Little CD 2000 VEGF regulates cell behavior during vasculogenesis. Dev Biol 224:178–188 Dumont DJ, Gradwohl G, Fong GH et al 1994 Dominant-negative and targeted null mutations in the endothelial receptor tyrosine kinase, tek, reveal a critical role in vasculogenesis of the embryo. Genes Dev 8:1897–1909 Eichmann A, Le Noble F, Autiero M, Carmeliet P 2005 Guidance of vascular and neural network formation. Curr Opin Neurobiol 15:108–115 Ferrara N, Carver-Moore K, Chen H et al 1996 Heterozygous embryonic lethality induced by targeted inactivation of the VEGF gene. Nature 380:439–442 Fong GH, Rossant J, Gertsenstein M, Breitman ML 1995 Role of the Flt-1 receptor tyrosine kinase in regulating the assembly of vascular endothelium. Nature 376:66–70 Fong GH, Zhang L, Bryce DM, Peng J 1999 Increased hemangioblast commitment, not vascular disorganization, is the primary defect in flt-1 knock-out mice. Development 126:3015–3025 Francis SE, Goh KL, Hodivala-Dilke K et al 2002 Central roles of alpha5beta1 integrin and fibronectin in vascular development in mouse embryos and embryoid bodies. Arterioscler Thromb Vasc Biol 22:927–933 George EL, Baldwin HS, Hynes RO 1997 Fibronectins are essential for heart and blood vessel morphogenesis but are dispensable for initial specification of precursor cells. Blood 90:3073–3081 Gerety SS, Anderson DJ 2002 Cardiovascular ephrinB2 function is essential for embryonic angiogenesis. Development 129:1397–1410 Giles PB, Candy CL, Fleming PA et al 2005 VEGF directs newly gastrulated mesoderm to the endothelial lineage. Dev Biol 279:169–178 Gory-Faure S, Prandini MH, Pointu H et al 1999 Role of vascular endothelial-cadherin in vascular morphogenesis. Development 126:2093–2102

70

DRAKE ET AL

Grazia Lampugnani M, Zanetti A, Corada M et al 2003 Contact inhibition of VEGF-induced proliferation requires vascular endothelial cadherin, beta-catenin, and the phosphatase DEP-1/CD148. J Cell Biol 161:793–804 Heroult M, Schaffner F, Augustin HG 2006 Eph receptor and ephrin ligand-mediated interactions during angiogenesis and tumor progression. Exp Cell Res 312:642–650 Kazi AS, Lotfi S, Goncharova EA et al 2004 Vascular endothelial growth factor-induced secretion of fibronectin is ERK dependent. Am J Physiol Lung Cell Mol Physiol 286: L539–545 Kertesz N, Krasnoperov V, Reddy R et al 2006 The soluble extracellular domain of EphB4 (sEphB4) antagonizes EphB4-EphrinB2 interaction, modulates angiogenesis, and inhibits tumor growth. Blood 107:2330–2338 Kim I, Ryu YS, Kwak HJ et al 2002 EphB ligand, ephrinB2, suppresses the VEGF- and angiopoietin 1-induced Ras/mitogen-activated protein kinase pathway in venous endothelial cells. FASEB J 16:1126–1128 Kroll J, Waltenberger J 1997 The vascular endothelial growth factor receptor KDR activates multiple signal transduction pathways in porcine aortic endothelial cells. J Biol Chem 272:32521–32527 Kruger O, Plum A, Kim JS et al 2000 Defective vascular development in connexin 45-deficient mice. Development 127:4179–4193 Lai KM, Pawson T 2000 The ShcA phosphotyrosine docking protein sensitizes cardiovascular signaling in the mouse embryo. Genes Dev 14:1132–1145 LaRue AC, Argraves WS, Zile MH, Drake CJ 2004 A critical role for retinol in the generation/differentiation of angioblasts required for embryonic blood vessel formation. Dev Dyn 230:666–674 Otto DM, Henderson CJ, Carrie D et al 2003 Identification of novel roles of the cytochrome p450 system in early embryogenesis: effects on vasculogenesis and retinoic acid homeostasis. Mol Cell Biol 23:6103–6116 Reynolds LE, Wyder L, Lively JC et al 2002 Enhanced pathological angiogenesis in mice lacking beta3 integrin or beta3 and beta5 integrins. Nat Med 8:27–34 Robb L, Lyons I, Li R et al 1995 Absence of yolk sac hematopoiesis from mice with a targeted disruption of the scl gene. Proc Natl Acad Sci USA 92:7075–7079 Sato TN, Tozawa Y, Deutsch U et al 1995 Distinct roles of the receptor tyrosine kinases Tie-1 and Tie-2 in blood vessel formation. Nature 376:70–74 Schlaepfer DD, Jones KC, Hunter T 1998 Multiple Grb2-mediated integrin-stimulated signaling pathways to ERK2/mitogen-activated protein kinase: summation of both c-Src- and focal adhesion kinase-initiated tyrosine phosphorylation events. Mol Cell Biol 18:2571–2585 Shalaby F, Rossant J, Yamaguchi TP et al 1995 Failure of blood-island formation and vasculogenesis in Flk-1-deficient mice. Nature 376:62–66 Shibuya M, Claesson-Welsh L 2006 Signal transduction by VEGF receptors in regulation of angiogenesis and lymphangiogenesis. Exp Cell Res 312:549–560 Shivdasani RA, Mayer EL, Orkin SH 1995 Absence of blood formation in mice lacking the T-cell leukaemia oncoprotein tal-1/SCL. Nature 373:432–434 Stains JP, Civitelli R 2005 Gap junctions regulate extracellular signal-regulated kinase signaling to affect gene transcription. Mol Biol Cell 16:64–72 Swiatek PJ, Lindsell CE, del Amo FF, Weinmaster G, Gridley T 1994 Notch1 is essential for postimplantation development in mice. Genes Dev 8:707–719 Takashima S, Kitakaze M, Asakura M et al 2002 Targeting of both mouse neuropilin-1 and neuropilin-2 genes severely impairs developmental yolk sac and embryonic angiogenesis. Proc Natl Acad Sci USA 99:3657–3662 Takahashi T, Takahashi K, St John PL et al 2003 A mutant receptor tyrosine phosphatase, CD148, causes defects in vascular development. Mol Cell Biol 23:1817–1831

GENETICS OF VASCULOGENESIS

71

Taylor PC 2002 VEGF and imaging of vessels in rheumatoid arthritis. Arthritis Res 4 Suppl 3:S99–107 Vernon RB, Lara SL, Drake CJ et al 1995 Organized type I collagen influences endothelial patterns during spontaneous angiogenesis in vitro: planar cultures as models of vascular development. In Vitro Cell Dev Biol Anim 31:120–131 Wang HU, Chen ZF, Anderson DJ 1998 Molecular distinction and angiogenic interaction between embryonic arteries and veins revealed by ephrin-B2 and its receptor Eph-B4. 93:741–753 Wijelath ES, Murray J, Rahman S et al 2002 Novel vascular endothelial growth factor binding domains of fibronectin enhance vascular endothelial growth factor biological activity. Circ Res 91:25–31 Zachary I 2003 VEGF signalling: integration and multi-tasking in endothelial cell biology. Biochem Soc Trans 31:1171–1177 Zanetti A, Lampugnani MG, Balconi G et al 2002 Vascular endothelial growth factor induces SHC association with vascular endothelial cadherin: a potential feedback mechanism to control vascular endothelial growth factor receptor-2 signaling. Arterioscler Thromb Vasc Biol 22:617–622

DISCUSSION Ruhrberg: You have treated VEGF as one gene. However, given how complex the VEGF pathway is, it may be interesting to further dissect the VEGF pathway with respect to the phenotypes of mice expressing selective VEGF isoforms only and add them to your list, particularly the phenotypes of the VEGF188 mice, which expresses VEGF188 only. For example, it would be interesting to know if there are any genes that modulate VEGF isoform expression or their matrix localization. Interestingly, there is excess vessel sprouting during angiogenesis in the VEGF188 mice, and brain vessels appear hyperbranched and very thin (Ruhrberg et al 2002). However, the yolk sac phenotype is completely different in these mice: yolk sac vessels appear hyperfused, as if the excess sprouts have fused (Ng 2007). When the VEGF188 mouse was first published (Stalmans et al 2002), it was mentioned that half of embryos were embryonic lethal, and that the others had a milder phenotype, surviving to term; some VEGF188 mice even survived to adulthood, albeit with impaired artery formation in the retina. The lethal mutants were later found to suffer from defective aortic arch remodelling (Stalmans et al 2003), but the phenotype of a hyperfused yolk sac vasculature has not yet been investigated fully and merits further investigation. Interestingly, the findings in the VEGF188 mice are very different from the phenotypes of the VEGF120 mutants, in which endothelial cells assemble into fewer, but larger vessels comprised of more endothelial cells per vessel diameter, and this is the same in the brain and the yolk sac (Ruhrberg et al 2002). Drake: The NIH has a great reluctance to fund work to characterize mice that have already been characterized. I would like to bring these mice in but unless it is through a collaboration, I can’t do this. The VEGF mutants are perplexing. Why would the lack of VEGF generate a hyperfused phenotype?

72

DRAKE ET AL

Ruhrberg: In different environments, VEGF isoforms may do different things. During vasculogenesis, primitive vessel networks assemble from single cell precursors, and disrupted isoform signalling may have a different outcome than in situations of angiogenesis, when vessels sprout from established vessel networks. Davies: You have this set of genes, which as well as doing their ‘professional’ roles, also impinge on the VEGF pathway. They cause your hyperfusion. Are you able to rescue them by either a pharmacological or genetic intervention, which reduces the amount of VEGF supplied? Drake: I would love to do this. To test this, we should be able to rescue the hyperfusion phenotype with a VEGF antagonist. The soluble form of Flt would be effective at achieving this. This would tone down the VEGF signalling. Then, perhaps we could gain a better understanding of the developmental role of connexin 45. I’d like to do this but I might not have the opportunity. DEP1/CD148 is a phosphatase which de-phosphorylates the Flk receptor. If we had this null mouse, we could see whether it is possible to normalize the vessels by antagonizing VEGF. This is possible. Lammert: Is the allantois also a site of haematopoiesis in addition to vasculogenesis? Drake: Yes. However, in our in vitro studies using dissected allantoides, the site of haematopoiesis which is at the base of the allantois is not included in the explant. In the embryo, the yolk sac blood vessels form with associated haematopoiesis but even within the yolk sac region, blood vessels immediately inward from the blood islands form without associated haematopoiesis. The blood vessels just form de novo. Blood vessel formation without associated haematopoiesis also occurs intraembryonically (i.e. the dorsal aortae). In contrast, the coronary blood vessels form with associated haematopoiesis. Why and what regulates de novo blood vessel formation with and without haematopoiesis is a perplexing issue. Ye: In many other systems, Delta/Notch signalling is known to influence cell fate specification. In the Dll4 knockout mice, endothelial cell differentiation appears to be altered. In the knockout, you probably get a lot more immature endothelial cells that are much more responsive to VEGF. In your gene list you might have some genes that influence how far the endothelial cells can differentiate, and lack of differentiation might render them hyper-responsive to VEGF. This could be the reason why you get continuous proliferation. Janet Rossant’s group showed that the dorsal aorta is altered in the Dll4 knockout mice, many endothelial cells in the dorsal aorta seem to have lost their large arterial identity and form sprouts instead. Drake: If VEGF increases endothelial cell protrusive activity, the result will be the formation of a large blood vessel/vascular sinus. In dorsal aortae development, the primary networks of vessels that are the primordia of this vessel respond to VEGF signalling, resulting in an exquisite localized formation of tip cells. The increased tip cell activity leads to the formation of new blood vessels and the loss

GENETICS OF VASCULOGENESIS

73

of the avascular spaces. Examining the normal process of vasculogenesis, it is clear that invasion of avascular spaces by tip cells is tightly regulated. Elevated levels of VEGF165 deregulate this control resulting in the invasion of the avascular spaces by tip cells. When invading tip cell-lead vascular spouts join together, a new blood vessel is formed and the avascular space is decreased. When repeated, the entire avascular space is lost, being replaced by a large blood vessel. Ye: I have a hypothesis that lumen formation is related to a certain differentiation status of an endothelial cell. If we can somehow get a handle on this, it could be a set of genes expressed at this time that is responsible. In your system, is it possible to isolate endothelial cells from embryos at different stages? You might find a population that can form lumen versus a population that can’t. Yancopoulos: I didn’t quite understand your explanation of the relationship between hyperfusion in the yolk sac and the aorta situation. Can you comment on this? You see defects in the yolk sac, but you also have these defects in the aorta. The aortal defects don’t look like they are hyperfusion defects to me: are you relating the two? Drake: I am trying to explain why the deletion of so many genes would give a similar phenotype—vascular hyperfusion—of the yolk sac vessels. I am not sure how (or even whether) this translates to the dorsal aorta. I see large lumen formation as a common mechanism of VEGF-induced hyperfusion that is regulated by both endothelial cell proliferation and protrusive activity. Yancopoulos: I guess you are postulating two different things. Somehow these gene knockouts are resulting in up-regulation of VEGF, or they are up-regulating the pathway by some other mechanism. For example, the molecules themselves might normally act as negative regulators of the process to help restrain what the VEGF is doing. Dll4 might be an example of such a negative regulator whose knockout results in this type of phenotype. Drake: The overriding theme is that VEGF, as a potent mediator of neovascularization, requires that the VEGF signalling pathway be continually suppressed. Yancopoulos: Consistent with this, in the developing eye we can inject Dll4 blockers and then see what looks like hyperfusion, just as you get in the yolk sac. It seems as if there are almost two sheets. If we then add VEGF trap blocker on top of that, we can normalize it. This supports what you are saying. We haven’t measured the VEGF levels, so we can’t say that the VEGF levels are up or whether it is a negative regulator. Owens: There seems to be overwhelming evidence that VEGF is a key regulator of this process. However, there seems to be a layer of complexity that we haven’t discussed, which is the VEGF isoforms, and the key importance of very precise temporal-spatial distribution of these VEGF isoforms. Clearly, you need the right amount of VEGF activity and activation at just the right time. When VEGF is added in large quantity, this process is likely to be highly dysregulated. In the

74

DRAKE ET AL

smooth muscle field, for the last two decades many of us have been preoccupied by the molecular regulation of lineage, i.e. what are the molecular controls involved in going from an embryonic stem cell to a fully mature differentiated smooth muscle cell? The characteristics that distinguish endothelium are polarity and the ability to form lumens. So where exactly in the scheme of development of endothelial cells from embryonic stem cells is VEGF acting? As an outsider, I think of the cells expressing VEGF receptors in the blood islands. But you showed a situation where there were some cells within these islands that initially seemed to respond to VEGF. They then exploded and took over the entire process. So a critical question is what is upstream of the VEGF receptors (e.g. Flk)? Drake: We are now working with ES cells to resolve this. To induce Flk1 expression in undifferentiated ES cells, we add BMP4, suggesting that at least in vitro, BMP4 can act to up-regulate Flk1. Once Flk1 expression is initiated, VEGF can be used to induce the expression of Tal1. This is not to say that Sonic or Indian hedgehog do not act in these pathways; these are just the results we get using a protocol that works. Owens: As I recall, BMP4 wasn’t one of your 12 genes. Correct? Drake: The 12 genes analysed were ones that were fundamentally involved in the regulation of the cells of the endothelial lineage. There were a lot of genes that when deleted, were embryonic lethal that we couldn’t include in our study, many of which acted in the regulation of the mesodermal lineage, the lineage from which endothelial cells are derived. We had to have something to analyze. If we didn’t even get endothelial cells, as might be the case for BMP4, there would be little to study in the context of vasculogenesis. Owens: So what really distinguishes endothelial cells from other cells is the ability to form lumens and their distinct polarity. When during development from ES cells does polarity first appear in the endothelial lineage? Drake: It is with the expression of CD34 and CD31. This is the first overt sign we can see. Lammert: We stress the polarity in vascular tubes. But the situation of vascular tubes might be different from many epithelial tubes. In the columnar epithelium there are tight junctions, which prevent the membranes from swapping. The proteins on the apical side can’t go to the basal side and vice versa. In endothelial cells, with the exception of the blood–brain barrier, this is not necessarily the case. In capillaries there are fenestrae, which are a direct connection between the apical side and the basal side. It might even be possible that there is a membrane flow between the apical and basal sides. Drake: It wasn’t that these were an indication of polarity. Those are the first cell–cell adhesion molecules that we could see emerge. This doesn’t mean that there is polarity. When you are an angioblast, though, you are not adhering to another cell.

GENETICS OF VASCULOGENESIS

75

Owens: So is the distinguishing feature of an endothelial cell merely the ability to form lumens, and that there is no polarity? Augustin: I disagree. Endothelial cells of whatever calibre form a fully polarised monolayer. I refer to two adhesion molecules: to the best of my knowledge, abluminal expression of ICAM1 and VCAM1 has never been observed. During angiogenic activation, polarity may be lost, but in a quiescent vessel of whatever calibre, it is a strictly polarized monolayer. There is no free flow of membrane molecules between the luminal and abluminal aspect of a quiescent endothelial cell. Adams: You presented a diagram with signalling pathways. It is helpful to think about the downstream signalling pathways, but it is difficult to show what the important signalling pathways are in vivo. Another connection between different genes that may be involved in the same process could be that they regulate each other. Finally, with regard to EphrinB2 function, we have new data that this molecule is controlling cell shape changes. We haven’t gone back to the early lethality and looked in the yolk sac, but this could also explain what is taking place at these stages. Drake: Why would this kill the mouse? It is hard to imagine why the lack of connexin 45 or EphrinB2 would lead to early embryonic lethality. Wilting: Have you discovered Tal1 and Fli1 in the developing lymphatic endothelium? Drake: No. Wilkins: You showed stages of blood development, with long bands of gene activity that stretch across many individual stages. I am assuming these are just patterns of expression of the genes, and not necessarily periods in which you know the gene is required for each of these steps? Drake: That is correct. This is just localization of expression. Shovlin: I’m interested in potential common mechanisms that could lead to your VEGF effect. Has sufficient work been done on the endothelial cells that survive the 8.5 day stage to see whether they have undergone compensatory changes to survive, having had specific genes deleted? I ask, because for endoglin, contradictory data arise when endothelial cells that have been derived from endoglin null mice are examined for expression of the TGFβ receptor ALK5 (TβRI), compared with wild-type endothelial cells that have been exposed to endoglin siRNA. Endoglin null cell survival depended on up-regulation of ALK5, whereas exposure of wild-type cells to endoglin siRNA led to down-regulation of ALK5. Similar mechanisms could operate here. Drake: We have an excellent method for sorting out both angioblasts and endothelial cells from mouse embryos. Although we don’t have these mice, it would be interesting to isolate these cells from null embryos and determine if they differ from normal cells.

76

DRAKE ET AL

References Ng Y-S 2007 The biology of vascular endothelial growth factor isoforms. In: Ruhrberg C (ed) VEGF and development. Landes Bioscience, in press Ruhrberg C, Gerhardt H, Golding M et al 2002 Spatially restricted patterning cues provided by heparin-binding VEGF-A control blood vessel branching morphogenesis. Genes Dev 16:2684–98 Stalmans I, Ng YS, Rohan R et al 2002 Arteriolar and venular patterning in retinas of mice selectively expressing VEGF isoforms. J Clin Invest 109:327–36 Stalmans I, Lambrechts D, De Smet F et al 2003: a modifier of the del22q11 (DiGeorge) syndrome? Nat Med 9:173–82

Negative regulators of vessel patterning Steven Suchting, Catarina Freitas, Ferdinand le Noble, Rui Benedito*, Christiane Bréant, Antonio Duarte* and Anne Eichmann Inserm U833, Collège de France, 11 Place Marcelin Berthelot, 75005 Paris, France and *Technical University of Lisbon, Portugal

Abstract. Blood vessels and nerves are structurally similar, complex branched networks that require guidance to ensure their proper positioning in the body. Recent studies have demonstrated that specialized endothelial cells, resembling axonal growth cones, are located at the tips of growing capillaries. These endothelial tip cells guide outgrowing capillaries in response to gradients of extracellular matrix-bound vascular endothelial growth factor (VEGF). Here we show that endothelial tip cell formation and vessel branching are negatively regulated by the Notch ligand Delta-like 4 (Dll4). Heterozygous deletion of Dll4 or pharmacological inhibition of Notch signalling using γ -secretase inhibitor revealed a striking vascular phenotype, with greatly increased numbers of fi lopodia-extending endothelial tip cells and increased expression of tip cell marker genes compared to controls. Filopodia extension in Dll4 +/− retinal vessels required VEGF and was inhibited when VEGF signalling was blocked. While VEGF expression was not significantly altered in Dll4 +/− retinas, Dll4 +/− vessels showed increased expression of VEGF Receptor 2 and decreased expression of VEGF Receptor 1 compared to wildtype, suggesting that they could be more responsive to VEGF stimulation. In addition, expression of Dll4 in wildtype tip cells was itself decreased when VEGF signalling was blocked, indicating that Dll4 may act downstream of VEGF as a ‘brake’ on VEGFmediated angiogenic sprouting. Taken together, these data reveal Dll4 as a novel negative regulator of vascular sprouting and vessel branching that is required for normal vascular network formation during development. 2007 Vascular development. Wiley, Chichester (Novartis Foundation Symposium 283) p 77–86

The vertebrate vascular system is comprised of a highly branched network of arteries, capillaries and veins that penetrates all body tissues. To build a functional network during development, endothelial cells (ECs) must navigate through tissue corridors and precisely project to their targets. Recent studies have demonstrated that specialized endothelial cells, resembling axonal growth cones, are located at the tips of growing capillaries. These ‘tip cells’ are lumen-less ECs that extend numerous fi lopodia into the environment and guide outgrowing capillaries in response to gradients of extracellular matrix-bound vascular endothelial growth factor (VEGF). Tip cells also find and create connections with adjacent sprouts, and so generate a functional vascular network. Several families of axon guidance molecules, including Semaphorins, Slits and Netrins have also been implicated in 77

78

EICHMANN

vessel pathfinding and network formation, suggesting that vessels and axons use similar guidance cues to regulate their navigation through the body. We have studied the axon guidance cue Netrin and its receptor UNC5B during blood vessel development. Among the different Netrin receptors, we have found only the UNC5B receptor to be expressed in the vascular system. UNC5B expression was observed in arterial endothelium and in endothelial tip cells, suggesting a role for UNC5B in tip cell guidance. Inactivation of the Unc5b gene in mice led to significantly increased abnormal branching of capillaries. Treatment of endothelial cells with the ligand Netrin 1 resulted in tip cell fi lopodial retraction. This effect was abolished in Unc5b-deficient mice, suggesting that Netrin 1 mediates repulsive guidance of capillary tip cells through UNC5B signalling and may help direct vessels to their targets in development. Although capillary branching in Unc5b mutants was increased, not all endothelial cells in the capillary plexus of these mutants extended fi lopodia, suggesting the existence of additional negative regulators of vessel branching. One candidate molecule that potentially regulates vessel branching is Delta-like 4 (Dll4). Dll4 is a transmembrane ligand for Notch receptors that is expressed in arterial blood vessels and sprouting endothelial cells. Notch signalling controls cell fate specification in a variety of cell contexts during embryonic and postnatal development. Genetic deletion of multiple components of the Notch pathway have revealed a critical role for Notch in vascular development. Deletion of a single Dll4 allele in mice results in early embryonic death (from E9.5) associated with major defects in vascular remodelling in the yolk sac and embryo. Haploinsufficiency within the vascular system has previously been observed only for VEGF, suggesting that appropriate dosage of both of these genes is critical for correct vascular development. We compared E10.5 heterozygous Dll4 embryos carrying a b -galactosidase ( b gal ) reporter gene insertion to heterozygous Unc5b b gal reporter embryos, which show similar vascular expression to Dll4 +/− but with normal vascular development. In Dll4 +/− embryos the major arteries such as the internal carotid artery appeared normal, but displayed vastly increased numbers of smaller arterial branches compared to Unc5b +/− embryos. To exclude the possibility that the defects observed in vessel branching were secondary to other defects in the embryo, such as disrupted blood flow, we cultured aortic explants from Dll4 +/− and wild-type mice in collagen gels. Sprouting from Dll4 +/− aortas occurred earlier and more profusely than from wild-type aortas, indicating that the Dll4 +/− vessel defect is intrinsic to ECs. To establish the underlying cause of the Dll4 +/− vessel branching defects we used high-resolution confocal microscopy on isolectin B4 -stained Dll4 +/− and wild-type E11.5 hindbrains. Staining for b gal in Dll4 +/− hindbrains showed expression in virtually all ECs (data not shown). Vessel branching was significantly increased in Dll4 +/− compared to wild-type hindbrains. Strikingly, vessels from Dll4 +/− hindbrains extended numerous fi lopodia from the whole length of the vessel surface,

NEGATIVE REGULATORS

79

whereas vessels from wild-type hindbrains extended only few filopodia at scattered points. The profusion of fi lopodia in Dll4 +/− hindbrains suggested that most, if not all, ECs in the hindbrain vessels were acting like tip cells, leading to increased connections between adjacent vessels (i.e. branching). To further evaluate whether loss of Dll4 leads to increased endothelial tip cell formation we analysed postnatal retinal vascular development of surviving Dll4 +/− and wild-type pups (approximately one-third of Dll4 +/− embryos on an outbred CD1 background survive to birth). The retinal vasculature at P4–6 allows for simultaneous visualization of angiogenic sprouting at the vascular front (where most endothelial tip cells are located), and remodelling of the nascent vasculature within the vascular plexus. As with hindbrains, Dll4 +/− retinal vessels showed severe patterning defects, forming a hyper-branched, hyperfused plexus behind the vascular front. Numerous fi lopodia extended from Dll4 +/− vessels both at the vascular front and also within the vascular plexus in both arterial and venous zones. In contrast, wild-type vessels extended few fi lopodia in regions away from the vascular front. Several genes are expressed at high levels in endothelial tip cells in the retina, including Pdg fb and Unc5b. Compared to wild-type, Dll4 +/− retinal vessels expressed Pdg fb and Unc5b over an expanded area, especially in the hyper-fused plexus. Thus, vessels from Dll4 +/− retinas display genetic as well as morphological (fi lopodia) and behavioural (hyperfused vessels) indicators of an expansion in the number of ECs that have a ‘tip cell’ phenotype. This suggests that Dll4 normally functions to suppress tip cell formation in growing vessels. To confirm that the Dll4 +/− vascular defects were consistent with a disruption of Notch signalling, we performed pharmacological disruption of Notch signalling in wild-type retinas using the γ -secretase inhibitor DAPT. DAPT treatment produced vessel abnormalities similar to the Dll4 +/− phenotype, including a hyperfused plexus and numerous ectopic fi lopodia extending from vessels. The expression of tip cell marker genes, including Pdg fb and Dll4 itself, was also upregulated in DAPT-treated retinal vessels, indicating that sprouting cannot be suppressed when Notch signalling is blocked, even in the presence of increased Dll4. The vascular defects observed in Dll4 +/− mice are thus consistent with a disruption in Notch signalling. Likely candidate Notch receptors include Notch1 and Notch4, both of which are expressed in retinal vessels. Normal angiogenic sprouting in the retina requires local gradients of VEGF. If VEGF stimulation is blocked, tip cell fi lopodia retract and progression of the vascular sprout halts. To confirm that fi lopodia extension in Dll4 +/− vessels still required VEGF signal, we injected eyes with receptor blocking antibodies against VEGFR2 (Flk1), the major mediator of VEGF function, and VEGFR1, a decoy receptor that negatively modulates VEGF signalling. VEGFR2-blocking antibody induced fi lopodia retraction in wild-type and Dll4 +/− retinas whereas VEGFR1blocking antibody had no significant effect. Injection of soluble VEGFR1 (sFlt1,

80

EICHMANN

VEGF sink) also blocked fi lopodial extension in Dll4 +/− vessels. Thus, absence of a Dll4 allele does not alter the requirement for VEGF and VEGFR2 activation for tip cell fi lopodia extension. Activated Notch signalling in ECs has been shown to down-regulate the expression of VEGFR2 in vitro, suggesting that excessive sprouting in Dll4 +/− vessels could be due to an increased response to normal levels of VEGF signal, perhaps via altered VEGF receptor expression. Thus, we asked whether loss of a Dll4 allele would lead to increased VEGFR2 expression. We performed qPCR analysis on isolated retinal ECs and found increased Veg fr2 expression in Dll4 +/− compared to wild-type. ISH also showed an expanded domain of Veg fr2 expression in Dll4 +/− retinal vessels compared to wild-type, especially in the region of the hyperfused plexus. In addition, we observed decreased Veg fr1 expression in Dll4 +/− retinal vessels by both ISH and qPCR. Thus, Dll4 +/− retinal vessels express higher levels of VEGFR2 and lower levels of VEGFR1, potentially increasing their responsiveness to VEGF. Such alterations in VEGF receptor levels could therefore provide a rational explanation for the increased sprouting observed in Dll4 +/− retinas. Taken together, our data suggest a model in which Dll4, expressed in endothelial tip cells, inhibits the angiogenic response of adjacent ECs to VEGF stimulation, most likely via Notch signalling. This would allow for an asymmetric cellular response to VEGF stimulation during vascular sprouting by allowing some ECs to respond to a local VEGF gradient by forming a sprout while, through upregulation of Dll4 expression, inhibiting adjacent cells from also forming sprouts. When even a single Dll4 allele is absent, or when Notch signalling is blocked, this suppression is lost, resulting in increased sprout formation and tip cell filopodia. This provides an elegant negative feedback mechanism intrinsic to ECs to control their response to VEGF and suggests that vascular network formation is coordinated by VEGF and Dll4/Notch signalling. DISCUSSION Davies: You are seeing your effect in a heterozygote. In your lateral inhibition model all of the cells will be heterozygotic. I would naively expect this not to have much effect on lateral inhibition, because it is not as if the neighbours can express normal levels and fight back. Eichmann: The neighbours still express Dll4, but at half the levels. Apparently, the dosage of the protein is very important. If Notch signalling is blocked completely, or there is half the dosage of Dll4, the result is that phenotype. Davies: Are there other Notch/Delta-type signalling systems where a heterozygote has such a dramatic effect? Kitajewski: Yes. Notch2 heterozygotes show low penetrance phenotypes. A few other Notch knockouts display a low penetrance heterozygous lethality. Whether they all relate to this concept of phenotypes related to lateral inhibition is not known.

NEGATIVE REGULATORS

81

Wilkins: In fact, all the Drosophila Notch mutants were picked up as haploinsufficients. Lammert: I have learned from Holger Gerhardt and Christer Betsholtz that tip cells normally don’t divide, and they don’t form a lumen. In your case, do you still have a lumen? I guess you have a real vessel. Eichmann: We haven’t assayed for perfusion of these vessels, so I can’t tell you whether or not they have a lumen. We have looked for proliferation, and there is a slight increase. The vascular plexus looks more dense. I think this increase is probably a secondary effect which, on its own, couldn’t account for this spectacular phenotype. Yancopoulos: We have a similar finding. We have perfusion in the retina and see good perfusion throughout the network. Interestingly enough, it is a matter of degree. These are heterozygotes. We have similar things to what you described for UNC5B, where you looked at the matrigel and tumour models. We have done the Dll4 inhibition in tumour models using a biological reagent that we think more completely blocks than occurs in the heterozygote. There we see hypervascular networks, but the majority of the network is not perfused. It may be a matter of degree: if we completely inhibit we get a little bit of excess activity and they can still form lumen and a perfused network. But if we completely ablate it, we get a really primitive network. When we do this, even though we have a great increase in vascular density, because of the perfusion problem the tumours are much more hypoxic, and are prevented from growing. If we do the reverse and increase Dll4 activity with a biological reagent, we get the same phenotype you described in UNC5B with shaved vessels. If we block the Dll4/Notch pathway, we get increased sprouting because we are blocking a negative regulator. If you promote it you are getting increased negative interaction and feedback, resulting in straight vasculatures which don’t inhibit tumour growth. The tumour is still able to grow with a reduced network, but this reduced activity is enough to promote pretty good tumour growth. Adams: I can imagine that if you lower Dll4 expression, the threshold of Notch signalling may not be enough. How does this correlate with the normal expression domain of Dll4? Eichmann: What we see is that in the heterozygote there is expanded tip cell marker expression. This is also found for Dll4. Betsholtz: An expanded zone of expression of Dll4 might represent a ‘historic imprint’ of reduced Notch signalling during the entire process of sprouting across the retina. We have done experiments together with Holger’s group in which we used γ -secretase inhibitors to acutely inhibit Notch signalling. In this situation the effect on sprouting was observed only at the sprouting margin, where Dll4 is expressed, but no change was observed in the Dll4 expression. Kitajewski: Marcus Fruttiger published that there is heterogenous Dll4 expression in the capillary plexus of the retina, when analysed by in situ hybridization.

82

EICHMANN

Eichmann: It is hard to correlate this at the single-cell level with a cell that makes fi lopodia. We have tried hard to do this. Kitajewski: It might express a hyper-tip phenotype throughout the capillary plexus, that is, the plexus may be full of fi lopodia-containing cells. Eichmann: One of the things we are lacking is the ability to analyse this in real time. Currently, we can only take snapshots. Betsholtz: Dll4 might not be expressed only in tip cells. But I don’t see this as a problem in the context of the model we are discussing. Inhibition of tip cell formation likely needs to occur through lateral signalling between stalk cells, not only between tip and stalk. The zone of the plexus that is reached by the VEGF released from astrocytes is likely broader than just the tip cell and its immediate neighbour, so there is likely a need for protection against ectopic sprouting also at more proximal locations in the plexus. Shibuya: Dll4 looks to be closely connected to VEGFR2. Is the UNC5B system also connected or is it totally independent? Eichmann: We don’t know yet. Adams: Is the perfusion you see leading to sufficient oxygen? Eichmann: It is possible that if there is this extensive fi lopodial formation, there is reduced perfusion and that the retina would get hypoxic. This may contribute to VEGF up-regulation and proliferation. Gerhardt: We have contributed to the study by Christer Betsholtz by doing more extensive in situ hybridization. Short-term inhibition of Notch signalling with the γ -secretase inhibitor DAPT doesn’t change VEGF acutely, but as soon as there is an extensive fi lopodial phenotype in the plexus, VEGF begins to be expressed all the way down to the central areas. There are also some changes in arterial venous patterning resulting from long-term treatment, where arteries and veins essentially grow on top of each other. In this type of phenotype everything goes wrong because VEGF is not regulated properly. In terms of the expression of Dll4, we need to remember that this is a dynamic process that occurs during vessel development. Early on, where there is sprouting activity only at the front, this is a nicely confined region of Dll4 expression. In many images, the second wave of angiogenesis is just beginning, with new sprouts forming in the central retina and a strong induction of new Dll4-positive tip cells occurs in that region. These are also the regions, even after short-term γ -secretase inhibition, where we see new ectopic sprouts forming. In order to understand the phenotype, we need to look at vessel changes in the context of Dll4 expression. Drake: We are having a nomenclature problem. The tip cell has a distinct phenotype and extends fi lopodia, which among other things, VEGF induces. When you elevate VEGF signalling, you can induce fi lopodia formation in endothelial cells of blood vessels that have lumens; however, the presence of filopodia doesn’t

NEGATIVE REGULATORS

83

make this endothelial cell a tip cell. The other problem is that we have observed angioblasts that have fi lopodia; these cells also can’t be called tip cells as they are only one cell and are thus, the tip of nothing. What we are seeing with the VEGF pathway is what was first elucidated in the tip cell—the regulation of fi lopodia formation and dynamics. Eichmann: In the Dll4 mice we see a certain number of tip cell marker genes, such as Unc5b. In the Dll4 heterozygotes these are all overexpressed throughout the plexus. In the Unc5b knockout, where there is a similar morphological phenotype and more fi lopodia, tip cell marker genes are apparently not overexpressed. It is not enough just to have fi lopodia; there must be a gene expression pattern that goes with it. Ruhrberg: I have a ‘chicken or egg’ question about VEGFR2 expression in the Dll4 heterozygotes: Does the up-regulation of VEGFR2 cause excess tip cells, or is VEGFR2 up-regulation providing another marker that reveals excess tip cells? Eichmann: Although we see higher expression of VEGFR2 at the vascular front compared to the plexus, we don’t find expression of this receptor to be exclusive to tip cells. Ruhrberg: What I meant is that VEGFR2 expression may be another marker that visualizes excess tip cells, because tip cells up-regulate VEGFR2. Alternatively, could it be that you see higher levels of VEGFR2, because there are fewer stalk cells? Stalk cells, like cells in a resting plexus, normally down-regulate VEGFR2, presumably because they no longer need to respond to VEGF as exquisitely as tip cells. Perhaps the vessels are in a more immature stage of development in the Dll4 heterozygotes? Drake: It doesn’t have to go down. All that is needed is a decrease in Flk1 signalling which can be achieved by the dephosphorylation of Flk1. VE-cadherin in association with the tyrosine phosphatase DEP-1/CD148 could achieve this. We have looked at Flk1 expression in the tip cells and there is no overt change compared to other endothelial cells. There is, however, a huge difference in the response to VEGF. Depending on dose, we can get just tip cells to respond but no other endothelial cells, even though they express Flk1. Eichmann: In vivo it does go down. Drake: I like the idea that a VE-cadherin/CD148 complex acts to suppress Flk1 signalling. Yancopoulos: In your Dll4 experiment, do you actually see sprouting and filopodia extensions from arteries themselves, or only from the capillary sides near the arteries? Eichmann: Just from the capillary side. The artery is fine. The first order side branch is also normal but then afterwards smaller branches appear to be composed entirely of fi lopodia-extending cells.

84

EICHMANN

Yancopoulos: We see the same thing, in that there isn’t sprouting from the artery. You could say it is just a dosage thing because this is a heterozygote knockout, but even when we add pharmacological blockers we don’t get sprouting from there. From seeing all these models in the retina, when you do these oxygen-induced retinopathy models, you never see sprouting coming from the arterial side. I remember being struck by how high the Dll4 is in the arteries, and thinking that this must be the negative regulator there. Chris Drake, did you say that just the presence of support cells completely controls sprouting and changes the character of the vessels? Drake: No. To go back to the beginning, VEGF-mediated vascular hyperfusion causes nascent blood vessels to fuse resulting in the formation of a larger diameter blood vessel or, with continued fusion, a vascular sinus. In the case of the dorsal aortae, a network of vessels is formed first and then the vessels undergo fusion to form the aortae. At this stage of development, our studies have shown that these vessels lack smooth muscle cells. As the vessel matures it becomes less sensitive to VEGF/Flk signalling and is invested with pericytes or smooth muscle cells. Perhaps endothelial cell behaviour following investment comes under the control of pericytes and smooth muscle cells. Yancopoulos: I’m sure you see this all the time in OIR models and so forth, but the arteries become tortuous. This is because there is endothelial cell proliferation within the vessel and it has nowhere to go, so it just has to get kinked. They are still responsive to VEGFR because they are proliferating, so it is not like the smooth muscle is creating a barrier. So why aren’t they sprouting? Betsholtz: The obvious possibility is that the arteries are surrounded by another type of mural support cell than the veins and capillaries. Presumably, the extracellular matrix or basement membrane composition around the different type of vessels might be different as well. Wilting: There is one example of sprouting arterial endothelial cells in a pathological situation, during arteriosclerosis and plaque formation. If you provide a new matrix to these cells you can find arterial endothelial cells growing into this matrix. Yancopoulos: Someone showed that when PDGF is injected into the eye then the support cells come off because they get disoriented. This might be an interesting combination to try: knock the cells off and see whether you get sprouting from the denuded endothelium. Ruhrberg: Perhaps an answer to the question about whether arterial cells can sprout could be found in the aortic ring assay. In this assay, there is sprouting from aortic explants, but I don’t know if these sprouts originate from the aorta or the vasa vasorum. Gerhardt: It may depend on the species. In the rat aortic ring model, people think sprouting occurs from the vasa vasoris. Someone told me that the mouse doesn’t have vasa vasoris due to the small calibre of the aorta.

NEGATIVE REGULATORS

85

Ruhrberg: Maybe this relates to the fact that you have to add exogenous VEGF to mouse aortic rings, but not to rat aortic rings to stimulate sprouting. Perhaps you could address the question of arterial sprouting with the aortic ring assay? Gerhardt: You asked about the retinopathy model. The fi rst thing that is seen in arterioles is this elongation and tortuosity. In extreme cases the arterioles do respond with some kind of protrusive activity, but it is not fi lopodia. It is lamellipodia that they make. A matrix component would be a prime candidate for restricting what type of protrusions the endothelial cells can make. In extreme cases there are small balls of endothelial cells that make lamellipodia-type structures. Betsholtz: We have looked at pericytes in long-term γ -secretase-treated retinas. There are lots of them present. It’s difficult to say if they occur in abnormal densities. They have a different morphology, but the whole vasculature has a different shape. I think we can conclude that loss of pericytes is not responsible for the hypersprouting phenotype. Yancopoulos: A related observation. Some of you may have seen some work we did with Donald McDonald where we gave angiopoietins to a whole animal. We could actually get enlargement-proliferation of venous vessels but not arterial vessels in the same vascular bed. It is a different growth factor, different signalling system and different response, but in this case it is proliferation and venous enlargement, and is seen only on the venous side. Whatever is controlling the artery affects its response to angiopoietin as well. Lammert: Since we are entering a more general discussion, Anne Uv, you were the first to show that chitins are found in the lumen of the insect trachea, and are an essential part of the structure. What parallels do you see, following the discussion of tip cells in vascular biology, between the trachea and the vessels that we are studying? Uv: There are a lot of parallels. The respiratory organ of Drosophila is used as a model system for tubulogenesis. It is a network of tubes that extends throughout the fly body, and delivers oxygen directly to target tissues. Early tracheal development uses a similar molecular mechanism as early lung development, but the later sprouting of tubes to form fine branches appears more analogous to blood vessel formation. When it comes to Notch and Delta, these are needed to specify the tip cells in the tracheal system. They help determine whether the tip cell will go on spreading towards target tissues, or whether it will fuse. I wonder: in the growing tip cell in your system, what determines whether it will continue growing towards the target or fuse with another capillary? Of course, there are identified molecules for pathfinding of tracheal branches, for example, FGF, which acts as a chemoattractant, and molecules that are also involved in neuronal pathfinding. When it comes to tube formation, we can see different mechanisms for forming a lumen for different types of branches. With all these different models you

86

EICHMANN

present, could they correlate with different processes? Could it be that vasculogenesis occurs by one way in one situation and differently in another? Betsholtz: To comment further on the parallels between the vertebrate vasculature and the Drosophila trachea, the tracheal sprout has blind-ended terminal cells, predestined to form pervading blind-ending tubes, as well as fusion cells that are predestined to fuse and anastomose with other tracheal branches. The tip cell of the blood vascular sprout has just one destiny, which is to anastomose with other branches. Thus, the blood vessel tip cell may be analogous to the tracheal fusion cell. There are no blunt-ended blood vessels, but there are blunt-ended lymph vessels, and hence the lymph vessel tip cell may correspond to the tracheal terminal cell. Uv: Do the cells change fate? Is there a guidance identity first, and then it switches to a fusion identity? Owens: That’s a good question. Perhaps for transient periods blunt-ended vessels exist, but they are unlikely to exist as a stable structure. I have a comment about the stability of large arteries. Many of the large vessels are genetically programmed. You can disrupt them entirely; you can have a mutation in zebrafish where there is no beating heart or pressurized arteriolar system, and the vessels will still form down to a certain branch level, at which point haemodynamic forces seem to become very important. I think the normal artery is a fairly stable structure although it remains plastic, in the sense that if it is injured it has a remarkable ability to repair itself and regenerate the original structure. In any case, my point is that when you are looking at what influence neural cells have on vascular development and patterning, it is very important to define where you are in the vascular network since there is a good possibility that in a different location, mechanisms may work within different vascular beds and branch levels. Ruhrberg: Would it be useful if we had an agreed method to score novel vascular phenotypes? We could then organise different vessel mutants into pathways much better. If this were done at the outset with every novel mouse mutant, it would be easier to identify pathways. Betsholtz: I think this would be extremely helpful. Drake: I would guess that the people who knocked out neuropilin didn’t expect a vascular phenotype. Ruhrberg: Yes, that is true. But if they had had a list of suggestions for standardized analyses and a protocol of how to perform them, they might have described the vascular phenotype in more detail. Drake: What is the impetus to get someone to look at the vascular phenotype, other than being a good citizen? I review a lot of knockout papers, and people are generally now doing a better job of characterizing the phenotype but it is a lot to ask of them to conduct an in-depth analysis of the vascular phenotype.

Lymphangiogenesis in development and disease Taija Mäkinen1 and Kari Alitalo* Max Planck Institute of Neurobiolog y, Department of Molecular Neurobiolog y, Am Klopferspitz 18, 82152 Martinsried, Germany and * Molecular/Cancer Biolog y Laboratory and Ludwig Institute for Cancer Research, Biomedicum Helsinki, University of Helsinki, P.O.B. 63 (Haartmaninkatu 8), 00014 Helsinki, Finland

Abstract. Lymphatic vessels are important for the maintenance of normal tissue fluid balance, for immune surveillance and adsorption of digested fats. In spite of their important functions in physiological as well as in various pathological conditions, including tumour metastasis, lymphoedema and inflammation, the lymphatic vessels have not received as much attention as the blood vessels, and the mechanisms regulating their development and growth have been poorly understood. However, recent studies using mouse genetic tools and primary lymphatic endothelial cell cultures have greatly increased our understanding of how the lymphatic endothelial cells differentiate, how lymphatic vessel growth is regulated and how the remodelling of the lymphatic vasculature into a functional vessel network consisting of capillaries and collecting vessels occurs. Furthermore, these studies have also provided mechanistic insights into the processes involved in pathological lymphangiogenesis. 2007 Vascular development. Wiley, Chichester (Novartis Foundation Symposium 283) p 87–105

Lymphatic vascular system The main function of the lymphatic vasculature is to maintain normal tissue fluid balance by returning interstitial fluid to the cardiovascular system. In addition, the lymphatic vasculature is an important part of the immune system and involved in absorption and transport of digested fats in the intestine. Lymphatic vessels start in the peripheral connective tissue as blind-ended capillaries, which collect excess extravasated tissue fluid. Lymphatic capillaries converge and unite into larger collecting vessels, pass to the lymph nodes, and return the fluid to the venous circulation via the two collecting trunks, the thoracic and the right lymphatic ducts. The lymphatic capillaries have a discontinuous basement membrane and overlapping endothelial cell junctions, and they lack pericytes and smooth muscle cells (SMCs), which make them highly permeable to large macromolecules. In contrast, collecting 1

Current address: Lymphatic Development Laboratory, Cancer Research UK London Research Institute, 44 Lincoln’s Inn Fields, London WC2A 3PX, UK 87

88

MÄKINEN & ALITALO

lymphatic vessels have sparse SMC coverage, which helps in pumping lymph forward, and numerous, irregularly located valves, which prevent lymph backflow. Lymphatic endothelial cell differentiation The development of embryonic lymphatic vessels starts after the establishment of the blood vasculature. According to a theory proposed by Florence Sabin a century ago, lymphatic endothelial cells (LECs) differentiate and sprout from the major veins in jugular and perimesonephric areas to form primitive lymph sacs, from which the vessels grow further by centrifugal sprouting (see Fig. 1). Homeodomain transcription factor Prox1 has been found to be a critical regulator of LEC differentiation. Targeted disruption of Prox1 in mice leads to arrest of LEC budding and failure in lymphatic vessel development, while the development of blood vasculature is not affected (Wigle & Oliver 1999). An as yet unknown signal induces polarized expression of Prox1 in the cardinal vein, which leads to up-regulation of lymphatic specific genes, such as vascular endothelial growth factor receptor 3 (VEGFR3) (Wigle et al 2002). In Prox1-deficient embryos the lymphatic specific gene expression was not induced. Instead, the mutant cells continued expressing blood vascular markers, suggesting that the cells were not committed to the lymphatic lineage (Wigle et al 2002). In contrast, ectopic expression of Prox1 in blood

FIG. 1. Development of lymphatic vasculature during embryogenesis. Lymphatic endothelial cells differentiate and sprout from the major veins and form a lymphatic capillary plexus. Further maturation involves remodelling of the vasculature into a hierarchical vessel network. Molecules involved in different processes in lymphatic development are shown below each developmental event. Adapted from Karpanen & Makinen (2006).

LYMPHANGIOGENESIS

89

vascular endothelial cells (BECs) induced expression of lymphatic-specific genes and resulted in the down-regulation of blood vascular genes (Petrova et al 2002, Hong et al 2004a). These studies suggest a role for Prox1 as a fate-determining factor for the LECs, and provide evidence supporting the venous origin of lymphatic vessels. Recent discovery of lymphatic vessels in zebrafish and highresolution imaging of lymphatic vessel development in living animals provides further support for the development of the primitive lymphatic vessels by sprouting from the veins (Yaniv et al 2006). Grafting experiments have suggested that allantoic and paraxial/somitic mesoderm of the avian wing bud have the potential to differentiate into lymphatic endothelium, suggesting that differentiation can also occur from mesodermal precursor cells (Wilting et al 2000, Papoutsi et al 2001). In addition, Prox1-positive precursor cells, lymphangioblasts, appear to contribute to lymphatic vessel formation in Xenopus tadpoles (Ny et al 2005). In mouse embryos, scattered LYVE1- and Prox1-positive mesenchymal cells are found in the dermatome prior to the presence of lymphatic vessels. It is possible that these cells represent lymphangioblasts, which integrate into growing lymphatic vessels (Buttler et al 2006). However, the inductive signal for the in situ differentiation of LECs is unknown. Regulation of the growth of lymphatic vessels Vascular endothelial growth factors (VEGFs) and VEGF receptor tyrosine kinases VEGFR3, one of the genes up-regulated by Prox1 in lymphatic endothelium, has been found to be a major regulator of lymphangiogenesis. The two known ligands for VEGFR3, VEGFC and VEGFD, can induce lymphangiogenesis in vivo, and stimulate proliferation, migration and survival of LECs in vitro ( Jeltsch et al 1997, Veikkola et al 2001, Makinen et al 2001b). After proteolytic processing, VEGFC and VEGFD are also ligands for the major regulator of blood vascular endothelia, VEGFR2 (Joukov et al 1997, Achen et al 1998), and the fully processed mature forms are also weakly angiogenic (Cao et al 1998, Byzova et al 2002). VEGFC deficiency leads to failure in the migration and proliferation of the LECs from the cardinal vein, and as a result the Veg fc gene-targeted embryos do not develop lymphatic vessels (Karkkainen et al 2004). Furthermore, VEGFC haploinsufficiency or insufficient VEGFR3 signalling due to inactivating mutations in the kinase domain result in hypoplastic lymphatic vascular networks, indicating that normal levels of this growth factor/receptor are essential for proper lymphatic development (Karkkainen et al 2001, 2004). In contrast, VEGFD is dispensable for lymphatic as well as blood vascular development in vivo, and the only defect observed in Veg fd-deficient mice was the reduction of lymphatic vessels around lung bronchioles (Baldwin et al 2005). Continuous signalling via VEGFR3 is required not only for the growth but also for the survival of lymphatic endothelial cells and for the maintenance of the

90

MÄKINEN & ALITALO

vasculature. Inhibition of VEGFR3 signalling using a soluble VEGFR3 protein, which acts as a ligand trap, leads to apoptosis of LECs and regression of lymphatic vessels during embryogenesis (Makinen et al 2001a) and during the first postnatal weeks (Karpanen et al 2006). However, in adult mice VEGFR3 inhibition specifically blocks the growth of new lymphatic vessels while the survival or function of pre-existing vessels is not affected (Karpanen et al 2006, Pytowski et al 2005). These results suggest that lymphatic vessels undergo a postnatal maturation process, in a similar fashion to what has previously been observed in the blood vessels, after which they are independent of ligand-induced VEGFR3 signalling. Although the molecular nature of the maturation process remains to be elucidated, additional survival signals may be received from the extracellular matrix, which can promote LEC survival in vitro via integrin signalling (Makinen et al 2001b, Zhang et al 2005). VEGF is best known for its involvement in regulation of angiogenesis. However, in addition to its strong angiogenic effect, when highly expressed in the skin in transgenic mice, VEGF can induce enlargement of lymphatic vessels. Furthermore, VEGF can promote lymphangiogenesis during wound healing and in tumours (Nagy et al 2002, Hong et al 2004b, Hirakawa et al 2005). The lymphangiogenic effect is caused, at least partly, by VEGF-induced recruitment of monocytes and macrophages, which can produce lymphangiogenic growth factors (Cursiefen et al 2004, Schoppmann et al 2002). However, VEGFR2, the receptor for VEGF, is expressed in at least some lymphatic vessels (Saaristo et al 2002), and VEGF can promote proliferation and migration of cultured LECs (Makinen et al 2001b, Hirakawa et al 2005), suggesting that VEGF/VEGFR2 may also directly stimulate lymphatic endothelium. VEGFR2 may also regulate lymphangiogenesis via formation of heterodimeric receptor complexes with VEGFR3 (Dixelius et al 2003). Other lymphangiogenic growth factors There is a growing list of growth factors shown to be capable of stimulating lymphatic vessel growth and/or LEC proliferation. Three members of the angiopoietin family (Ang1, Ang2, Ang3/4) are ligands for the Tie1 and/or Tie2 receptor tyrosine kinases, which are expressed both in blood vascular and lymphatic endothelia (Iljin et al 2002, Morisada et al 2005, Tammela et al 2005). Ang1 can activate both Tie receptors in heteromeric receptor complexes (Saharinen et al 2005). In contrast, Ang2 can mediate either an agonistic or an antagonistic function in Tie2 signalling (Maisonpierre et al 1997, Gale et al 2002). During the formation and stabilization of the blood vasculature Ang2 seems to act as a Tie2 antagonist (Maisonpierre et al 1997), while an agonistic function of Ang2 is required for proper patterning of the lymphatic vasculature (Gale et al 2002). On

LYMPHANGIOGENESIS

91

the other hand, overexpression of Ang1 stimulated lymphatic endothelial cell proliferation and promoted vessel enlargement and generation of new sprouts (Morisada et al 2005, Tammela et al 2005). Other growth factors with lymphangiogenic potential include fibroblast growth factor 2 (FGF2), platelet-derived growth factors (PDGFs) and hepatocyte growth factor (HGF) (Kubo et al 2002, Cao et al 2004, Kajiya et al 2005). While the effect of FGF2 is indirect due to up-regulated expression of VEGFC and VEGFD by blood vascular endothelial and periendothelial cells (Kubo et al 2002, Chang et al 2004), PDGFs and HGF may act directly on the lymphatic endothelium (Cao et al 2004, Kajiya et al 2005). Based on the observation that their receptors were up-regulated in newly formed vessels and in activated lymphatic endothelia during wound healing in inflamed skin and in tumour-associated lymphatic vessels, PDGFs and HGF were suggested to play a role in pathological lymphangiogenesis (Cao et al 2004, Kajiya et al 2005). Separation of blood and lymphatic vascular networks The developing lymphatic vasculature needs to be separated from the pre-existing blood vessel network, and the only sites where the vessels retain direct contacts are at the junction where the thoracic and right lymphatic ducts join the subclavian veins. The molecular mechanisms that prevent blood and lymphatic vessels from fusing together are currently poorly understood. Loss of haematopoietic intracellular signalling molecules Syk or SLP76 was shown to lead to arteriovenous shunting and mixing of blood and lymphatic endothelial cells, which resulted in haemorrhaging and perinatal death (Abtahian et al 2003). Since Syk and SLP76 were found to be expressed only in haematopoietic cells and not on blood or lymphatic endothelium, the authors suggested a non-cell-autonomous mechanism where signals from circulating haematopoietic cells regulate the separation of blood and lymphatic vascular networks. However, the signals originating from haematopoietic cells are currently unknown. Remodelling of the lymphatic vasculature Remodelling of the blood vasculature is known to play a critical role in the development of a functional blood vessel network. Signalling mediated via Eph receptor tyrosine kinases and their transmembrane ligands, Ephrin Bs, have been implicated in the remodelling of the arterial–venous plexus during cardiovascular development. More recently, Eph–ephrin signalling was found to play an important role in the postnatal lymphatic maturation processes. Ephrin B ligands have intrinsic signalling capacities; their cytoplasmic domains can be phosphorylated on tyrosine residues and they have a C-terminal motif for binding of PDZ domain containing

92

MÄKINEN & ALITALO

proteins. Mouse mutants deficient in the PDZ binding motif of ephrinB2 develop normal blood vasculature, but display chylothorax and fail to remodel their lymphatic vasculature into a hierarchically organized vessel network consisting of lymphatic capillaries and collecting lymphatic vessels (Makinen et al 2005). In addition, the formation of luminal valves in the collecting vessels is defective in the mutant mice. The functional failure of the lymphatic system in ephrinB2 mutant mice demonstrates that postnatal remodelling is a critical process during the establishment of a normal lymphatic vascular network. Pericytes and SMCs have important roles during normal development of blood and lymphatic vessels. Angiopoietin 2 mice display lymphatic dysfunction due to disorganization and hyperplasia of collecting lymphatic vessels with poorly associated SMCs (Gale et al 2002), while excessive coverage of lymphatic vessels by SMCs in FOXC2 null mice leads to abnormal patterning and failure in luminal valve formation (Petrova et al 2004). These studies suggest that the SMCinteraction can regulate the behaviour of the endothelial cells and indicate the importance of SMCs in valve morphogenesis. Ectopic SMC coverage, together with disturbed valve formation, is also observed in the lymphatic vessels of the mice lacking ephrinB2 PDZ binding motif (Makinen et al 2005). The appearance of the lymphatic defects prior to the acquisition of SMC coverage suggests a cellautonomous function of ephrinB2 in the lymphatic endothelium rather than in SMCs, but ephrinB2 apparently has additional functions in the SMCs. Deletion of ephrinB2 specifically in the pericytes and vascular SMCs leads to ectopic recruitment of SMCs into lymphatic capillaries, although this does not seem to result in defective lymphatic function (Foo et al 2006).

Lymphatic vessels in pathological conditions Lymphoedema Malfunction of the lymphatic system rarely results in life-threatening diseases, however, failure of lymph transport can lead to lymphoedema, which is characterized by chronic and progressive swelling of the extremities due to interstitial accumulation of proteins and associated fluid. Lymphoedema may be an inherited disease (primary lymphoedema) or it may be caused when the lymphatic vessels or tissues are obstructed or damaged due to infection, radiation therapy or surgery (secondary or acquired lymphoedema). Primary lymphoedemas are a group of disorders that develop due to hypoplasia or hyperplasia of the lymphatic vessels and that can be associated with additional malformations in other organs. A congenital, autosomal dominant form of lymphoedema (Milroy disease) is caused by kinase inactivating mutations in the VEGFR3 gene (Ferrell et al 1998, Karkkainen et al 2000, Irrthum et al 2000). On the other hand, mutations identified in

LYMPHANGIOGENESIS

93

genes encoding for the transcription factors FOXC2 and Sox18 underlie the genetic causations of lymphoedema-distichiasis and hypotrichosis-lymphedematelangiectasia, respectively (Fang et al 2000, Finegold et al 2001, Irrthum et al 2003). The mouse mutants for VEGFR3, FOXC2 and Sox18 genes also show lymphatic abnormalities, allowing their use for further studies as models for lymphoedema, and for testing of new therapeutic strategies (Karkkainen et al 2001, Kriederman et al 2003, Pennisi et al 2000). Tumour lymphangiogenesis and lymphatic metastasis Metastatic spread of tumour cells to distant organs, either via direct invasion or via blood or lymphatic vessels, is a major cause of death in cancer patients. While the importance of angiogenesis for the growth and metastasis of solid tumours has been well recognized, recent studies suggest that malignant tumours can also activate lymphangiogenesis and metastasize through the lymphatic system. Clinical studies indicate that expression of VEGFC and VEGFD by the tumour cells correlates with high lymphatic vessel density in the vicinity of and/or inside the tumour and provides a prognostic indicator of the metastatic potential (reviewed in Achen et al 2005). In transgenic and xenotransplanted mice, overexpression of VEGFC or VEGFD induces tumour lymphangiogenesis and promotes intralymphatic tumour growth and formation of lymph node metastases (Karpanen et al 2001, Stacker et al 2001, Skobe et al 2001, Mandriota et al 2001). Inhibition of lymphatic metastasis using soluble VEGFR3 protein, or neutralizing antibodies against the ligand or the receptor, demonstrates the critical role of the VEGFR3 signalling pathway in lymphatic metastasis and makes it a promising target for anticancer therapeutics aimed at limiting metastatic spreading (Shimizu et al 2004, He et al 2002, Karpanen et al 2001, Stacker et al 2001). Inflammatory diseases Lymphatic vessels are involved in regulating inflammatory responses by decreasing tissue oedema and by transporting leukocytes from the inflammation sites to secondary lymphoid organs. Migration of antigen-presenting dendritic cells (DCs) through lymphatic vessels to draining lymph nodes is a key step in the initiation of an adaptive immune response. Upon inflammation, the changes in the lymphatic network within the activated lymph node and in the peripheral tissue enhance DC migration, which suggests an active role of lymphatic vasculature in DC mobilization (Angeli et al 2006). Macrophages are suggested to have a dual role in inflammation-induced lymphangiogenesis by secreting lymphangiogenic growth factors VEGFC and -D, which stimulate the growth of existing lymphatic endothelial cells and by transdifferentiating to lymphatic endothelial cells, which

94

MÄKINEN & ALITALO

can incorporate into the lymphatic endothelium (Maruyama et al 2005, Schoppmann et al 2002). Recent new discoveries reveal the active involvement of the lymphatic system in the maintenance of alloreactive immune responses. Rejected kidney transplants were found to contain a massive increase in the amount of lymphatic vessels as compared to normal kidneys, and these vessels produced the secondary lymphoid chemokine (SLC/CCL21), which further attracted CCR7+ lymphocytes and DCs (Kerjaschki et al 2004). In chronic airway inflammation, lymphangiogenesis prevented mucosal oedema, but whereas the newly formed blood vessels regressed after antibiotic treatment the lymphatic vessels persisted (Baluk et al 2005). Therapeutic strategies against lymphatic disorders? Both the stimulation of lymphatic vessel formation (e.g. in lymphoedema) or the inhibition of lymphangiogenesis (e.g. to prevent tumour lymphangiogenesis and subsequent lymphatic metastasis) could provide useful therapeutic approaches. Encouraging results in pro-lymphangiogenic therapy have been obtained by using VEGFC therapy in animal models for lymphoedema (Karkkainen et al 2001, Saaristo et al 2002, Szuba et al 2002). Furthermore, VEGFC gene transfer stimulated the formation of a functional lymphatic vessel network after surgical incision, suggesting important implications for the treatment of secondary lymphoedema (Saaristo et al 2004). On the contrary, inhibition of VEGFC/D–VEGFR3 mediated signalling via administration of soluble receptor protein or neutralizing antibodies against the ligand or the receptor inhibited lymphatic metastasis in experimental mouse models, suggesting that inhibition of VEGFR3 signalling may provide an important strategy for blocking tumour metastasis in human patients (Shimizu et al 2004, He et al 2002, Karpanen et al 2001, Stacker et al 2001). The metastasis of tumour cells is a complex process that requires multiple sequential events, such as cell detachment, intravasation, adherence in the target organ vasculature as well as extravasation, and therefore additional factors besides VEGFC are apparently involved in regulating lymphatic metastasis (He et al 2002, Cao et al 2004). Targeting of both tumour lymphangiogenesis and tumour cell invasion may be required in order to accomplish a complete inhibition of lymphatic metastasis (He et al 2002). Identification and functional characterization of new molecular players involved in physiological and pathological lymphangiogenesis are therefore prerequisites for the development of novel therapeutic approaches. Conclusions During the past decade the identification of lymphatic specific markers and growth factors has enabled detailed studies of the lymphatic system, and experiments using

LYMPHANGIOGENESIS

95

genetically modified mice have greatly increased our understanding of the mechanisms of normal lymphatic development. Since physiological and pathological lymphangiogenesis appear to be regulated by similar signalling pathways, understanding the basic biology has also provided novel insights into the pathogenesis of diseases involving lymphatic vessels, such as lymphoedema, inflammation and lymphatic metastasis. The challenge for future studies is to identify new molecular players and to increase our understanding of the mechanisms of lymphangiogenesis, which may reveal new therapeutic targets and help to develop novel therapies against lymphatic disorders. Acknowledgements KA is supported by grants from NIH (5 R01 HL075183-02) and European Union (LYMPHANGIOGENOMICS, LSHG-CT-2004-503573). TM is supported by postdoctoral fellowships from the European Molecular Biology Organization and the Human Frontier Science Program Organization.

References Abtahian F, Guerriero A, Sebzda E et al 2003 Regulation of blood and lymphatic vascular separation by signaling proteins SLP-76 and Syk. Science 299:247–251 Achen MG, Jeltsch M, Kukk E et al 1998 Vascular endothelial growth factor D (VEGF-D) is a ligand for the tyrosine kinases VEGF receptor 2 (Flk1) and VEGF receptor 3 (Flt4). Proc Natl Acad Sci USA 95:548–553 Achen MG, McColl BK, Stacker SA 2005 Focus on lymphangiogenesis in tumor metastasis. Cancer Cell 7:121–127 Angeli V, Ginhoux F, Llodra J et al 2006 B cell-driven lymphangiogenesis in inflamed lymph nodes enhances dendritic cell mobilization. Immunity 24:203–215 Baldwin ME, Halford MM, Roufail S et al 2005 Vascular endothelial growth factor D is dispensable for development of the lymphatic system. Mol Cell Biol 25:2441–2449 Baluk P, Tammela T, Ator E et al 2005 Pathogenesis of persistent lymphatic vessel hyperplasia in chronic airway inflammation. J Clin Invest 115:247–257 Buttler K, Kreysing A, von Kaisenberg CS et al 2006 Mesenchymal cells with leukocyte and lymphendothelial characteristics in murine embryos. Dev Dyn 235:1554–1562 Byzova TV, Goldman CK, Jankau J et al 2002 Adenovirus encoding vascular endothelial growth factor-D induces tissue-specific vascular patterns in vivo. Blood 99:4434–4442 Cao Y, Linden P, Farnebo J et al 1998 Vascular endothelial growth factor C induces angiogenesis in vivo. Proc Natl Acad Sci USA 95:14389–14394 Cao R, Bjorndahl MA, Religa P et al 2004 PDGF-BB induces intratumoral lymphangiogenesis and promotes lymphatic metastasis. Cancer Cell 6:333–345 Chang LK, Garcia-Cardena G, Farnebo F et al 2004 Dose-dependent response of FGF-2 for lymphangiogenesis. Proc Natl Acad Sci USA 101:11658–11663 Cursiefen C, Chen L, Borges LP et al 2004 VEGF-A stimulates lymphangiogenesis and hemangiogenesis in inflammatory neovascularization via macrophage recruitment. J Clin Invest 113:1040–1050 Dixelius J, Makinen T, Wirzenius M et al 2003 Ligand-induced vascular endothelial growth factor receptor-3 (VEGFR-3) heterodimerization with VEGFR-2 in primary lymphatic endothelial cells regulates tyrosine phosphorylation sites. J Biol Chem 278:40973–40979

96

MÄKINEN & ALITALO

Fang J, Dagenais SL, Erickson RP et al 2000 Mutations in FOXC2 (MFH-1), a forkhead family transcription factor, are responsible for the hereditary lymphedema-distichiasis syndrome. Am J Hum Genet 67:1382–1388 Ferrell RE, Levinson KL, Esman JH et al 1998 Hereditary lymphedema: evidence for linkage and genetic heterogeneity. Hum Mol Genet 7:2073–2078 Finegold DN, Kimak MA, Lawrence EC et al 2001 Truncating mutations in FOXC2 cause multiple lymphedema syndromes. Hum Mol Genet 10:1185–1189 Foo SS, Turner CJ, Adams S et al 2006 Ephrin-B2 controls cell motility and adhesion during blood-vessel-wall assembly. Cell 124:161–173 Gale NW, Thurston G, Hackett SF et al 2002 Angiopoietin-2 is required for postnatal angiogenesis and lymphatic patterning, and only the latter role is rescued by Angiopoietin-1. Dev Cell 3:411–23 He Y, Kozaki K, Karpanen T et al 2002 Suppression of tumor lymphangiogenesis and lymph node metastasis by blocking vascular endothelial growth factor receptor 3 signaling. J Natl Cancer Inst 94:819–825 Hirakawa S, Kodama S, Kunstfeld R et al 2005 VEGF-A induces tumor and sentinel lymph node lymphangiogenesis and promotes lymphatic metastasis. J Exp Med 201:1089–1099 Hong YK, Foreman K, Shin JW et al 2004a Lymphatic reprogramming of blood vascular endothelium by Kaposi sarcoma-associated herpesvirus. Nat Genet 36:683–685 Hong YK, Lange-Asschenfeldt B, Velasco P et al 2004b VEGF-A promotes tissue repairassociated lymphatic vessel formation via VEGFR-2 and the alpha1beta1 and alpha2beta1 integrins. FASEB J 18:1111–1113 Iljin K, Petrova TV, Veikkola T et al 2002 A fluorescent Tie1 reporter allows monitoring of vascular development and endothelial cell isolation from transgenic mouse embryos. FASEB J 16:1764–1774 Irrthum A, Karkkainen MJ, Devriendt K, Alitalo K, Vikkula M 2000 Congenital hereditary lymphedema caused by a mutation that inactivates VEGFR3 tyrosine kinase. Am J Hum Genet 67:295–301 Irrthum A, Devriendt K, Chitayat D et al 2003 Mutations in the transcription factor gene SOX18 underlie recessive and dominant forms of hypotrichosis-lymphedema-telangiectasia. Am J Hum Genet 72:1470–1478 Jeltsch M, Kaipainen A, Joukov V et al 1997 Hyperplasia of lymphatic vessels in VEGF-C transgenic mice. Science 276:1423–1425 Joukov V, Sorsa T, Kumar V et al 1997 Proteolytic processing regulates receptor specificity and activity of VEGF-C. EMBO J 16:3898–3911 Kajiya K, Hirakawa S, Ma B, Drinnenberg I, Detmar M 2005 Hepatocyte growth factor promotes lymphatic vessel formation and function. EMBO J 24:2885–2895 Karkkainen MJ, Ferrell RE, Lawrence EC et al 2000 Missense mutations interfere with VEGFR-3 signalling in primary lymphoedema. Nat Genet 25:153–159 Karkkainen MJ, Saaristo A, Jussila L et al 2001 A model for gene therapy of human hereditary lymphedema. Proc Natl Acad Sci USA 98:12677–12682 Karkkainen MJ, Haiko P, Sainio K et al 2004 Vascular endothelial growth factor C is required for sprouting of the fi rst lymphatic vessels from embryonic veins. Nat Immunol 5:74–80 Karpanen T, Makinen T 2006 Regulation of lymphangiogenesis–from cell fate determination to vessel remodeling. Exp Cell Res 312:575–583 Karpanen T, Egeblad M, Karkkainen MJ et al 2001 Vascular endothelial growth factor C promotes tumor lymphangiogenesis and intralymphatic tumor growth. Cancer Res 61: 1786–1790 Karpanen T, Wirzenius M, Makinen T et al 2006 Lymphangiogenic growth factor responsiveness is modulated by postnatal lymphatic vessel maturation. Am J Pathol 169:708–18

LYMPHANGIOGENESIS

97

Kerjaschki D, Regele HM, Moosberger I et al 2004 Lymphatic neoangiogenesis in human kidney transplants is associated with immunologically active lymphocytic infi ltrates. J Am Soc Nephrol 15:603–612 Kriederman BM, Myloyde TL, Witte MH et al 2003 FOXC2 haploinsufficient mice are a model for human autosomal dominant lymphedema-distichiasis syndrome. Hum Mol Genet 12:1179–1185 Kubo H, Cao R, Brakenhielm E et al 2002 Blockade of vascular endothelial growth factor receptor-3 signaling inhibits fibroblast growth factor-2-induced lymphangiogenesis in mouse cornea. Proc Natl Acad Sci USA 99:8868–8873 Maisonpierre PC, Suri C, Jones PF et al 1997 Angiopoietin-2, a natural antagonist for Tie2 that disrupts in vivo angiogenesis. Science 277:55–60 Makinen T, Jussila L, Veikkola T et al 2001a Inhibition of lymphangiogenesis with resulting lymphedema in transgenic mice expressing soluble VEGF receptor-3. Nat Med 7:199–205 Makinen T, Veikkola T, Mustjoki S et al 2001b Isolated lymphatic endothelial cells transduce growth, survival and migratory signals via the VEGF-C/D receptor VEGFR-3. EMBO J 20:4762–4773 Makinen T, Adams RH, Bailey J et al 2005 PDZ interaction site in ephrinB2 is required for the remodeling of lymphatic vasculature. Genes Dev 19:397–410 Mandriota SJ, Jussila L, Jeltsch M et al 2001 Vascular endothelial growth factor-C-mediated lymphangiogenesis promotes tumour metastasis. EMBO J 20:672–682 Maruyama K, Ii M, Cursiefen C et al 2005 Inflammation-induced lymphangiogenesis in the cornea arises from CD11b-positive macrophages. J Clin Invest 115:2363–2372 Morisada T, Oike Y, Yamada Y et al 2005 Angiopoietin-1 promotes LYVE-1-positive lymphatic vessel formation. Blood 105:4649–4656 Nagy JA, Vasile E, Feng D et al 2002 Vascular permeability factor/vascular endothelial growth factor induces lymphangiogenesis as well as angiogenesis. J Exp Med 196:1497–1506 Ny A, Koch M, Schneider M et al 2005 A genetic Xenopus laevis tadpole model to study lymphangiogenesis. Nat Med 11:998–1004 Papoutsi M, Tomarev SI, Eichmann A et al 2001 Endogenous origin of the lymphatics in the avian chorioallantoic membrane. Dev Dyn 222:238–251 Pennisi D, Gardner J, Chambers D et al 2000 Mutations in Sox18 underlie cardiovascular and hair follicle defects in ragged mice. Nat Genet 24:434–437 Petrova TV, Makinen T, Makela TP et al 2002 Lymphatic endothelial reprogramming of vascular endothelial cells by the Prox-1 homeobox transcription factor. EMBO J 21:4593–4599 Petrova TV, Karpanen T, Norrmen C et al 2004 Defective valves and abnormal mural cell recruitment underlie lymphatic vascular failure in lymphedema distichiasis. Nat Med 10:974– 981 Pytowski B, Goldman J, Persaud K et al 2005 Complete and specific inhibition of adult lymphatic regeneration by a novel VEGFR-3 neutralizing antibody. J Natl Cancer Inst 97:14–21 Saaristo A, Veikkola T, Tammela T et al 2002 Lymphangiogenic gene therapy with minimal blood vascular side effects. J Exp Med 196:719–730 Saaristo A, Tammela T, Timonen J et al 2004 Vascular endothelial growth factor-C gene therapy restores lymphatic flow across incision wounds. FASEB J 18:1707–1709 Saharinen P, Kerkela K, Ekman N et al 2005 Multiple angiopoietin recombinant proteins activate the Tie1 receptor tyrosine kinase and promote its interaction with Tie2. J Cell Biol 169:239–243 Schoppmann SF, Birner P, Stockl J et al 2002 Tumor-associated macrophages express lymphatic endothelial growth factors and are related to peritumoral lymphangiogenesis. Am J Pathol 161:947–956

98

MÄKINEN & ALITALO

Shimizu K, Kubo H, Yamaguchi K et al 2004 Suppression of VEGFR-3 signaling inhibits lymph node metastasis in gastric cancer. Cancer Sci 95:328–333 Skobe M, Hawighorst T, Jackson DG et al 2001 Induction of tumor lymphangiogenesis by VEGF-C promotes breast cancer metastasis. Nat Med 7:192–198 Stacker SA, Caesar C, Baldwin ME et al 2001 VEGF-D promotes the metastatic spread of tumor cells via the lymphatics. Nat Med 7:186–191 Szuba A, Skobe M, Karkkainen MJ et al 2002 Therapeutic lymphangiogenesis with human recombinant VEGF-C. FASEB J 16:1985–1957 Tammela T, Saaristo A, Lohela M et al 2005 Angiopoietin-1 promotes lymphatic sprouting and hyperplasia. Blood 105:4642–4648 Veikkola T, Jussila L, Makinen T et al 2001 Signalling via vascular endothelial growth factor receptor-3 is sufficient for lymphangiogenesis in transgenic mice. EMBO J 20:1223– 1231 Wigle JT, Oliver G 1999 Prox1 function is required for the development of the murine lymphatic system. Cell 98:769–778 Wigle JT, Harvey N, Detmar M et al 2002 An essential role for Prox1 in the induction of the lymphatic endothelial cell phenotype. EMBO J 21:1505–1513 Wilting J, Papoutsi M, Schneider M, Christ B 2000 The lymphatic endothelium of the avian wing is of somitic origin. Dev Dyn 217:271–278 Yaniv K, Isogai S, Castranova D et al 2006 Live imaging of lymphatic development in the zebrafish. Nat Med 12:711–716 Zhang X, Groopman JE, Wang JF 2005 Extracellular matrix regulates endothelial functions through interaction of VEGFR-3 and integrin alpha5beta1. J Cell Physiol 202:205–214

DISCUSSION Lammert: I am curious about the Ephrin B receptors. What Ephrin B receptors are expressed, and can you also do experiments with Fc fusion proteins that can activate reverse signalling and induce different changes than the ones induced by the dominant negative ephrinB2 form? Mäkinen: The only Ephrin B receptor from which I have conclusive results is EphB4, which is expressed in the primary lymphatic capillary plexus, and after the remodelling it is expressed both in the lymphatic capillaries and collecting vessels, unlike ephrinB2, which after the remodelling stage becomes restricted to the collecting lymphatic vessels. For sure, there are other Ephrin B receptors expressed, for example in the smooth muscle cells. There could be signalling from the smooth muscle cells to the endothelial cells in the collecting vessels. Lammert: What is the mechanism of the signalling? Mäkinen: In the remodelling process that occurs in the primary plexus we see the defect before the smooth muscle cells are present. I would think that in this case ephrinB2-mediated signalling is cell autonomous in the lymphatic endothelial cells, which start sprouting but show a defect in the sprout elongation. At that stage EphB4 is expressed in the same cells. However, at present we don’t know what is the effector molecule that mediates the signalling downstream of ephrinB2 in lymphatic endothelial cells.

LYMPHANGIOGENESIS

99

Betsholtz: Is there really no blood vascular phenotype in these mutants? Mäkinen: In older animals the blood vessels sometimes become tortuous, but this could be a secondary effect due to lymphatic dysfunction. At this point, we can’t rule out that there is nothing wrong with the blood vasculature. We haven’t done a careful analysis of the blood vasculature, and we haven’t studied pathological conditions, but superficially, the hierarchical organization of the vasculature for example in the skin looks normal. However, the mice show specific defects in the lung blood vessels, and this phenotype is currently being analysed in detail by George Wilkinson, in Rüdiger Klein’s laboratory. Adams: I find this work exciting, and it fits well with studies that we have done on loss-of-function and gain-of-function, indicating that ephrinB2 is promoting the sprout formation and cell shape changes in the sprouting process. It raises a puzzle: how does the EphB4 expression that is all over the lymphatics match with ephrinB2 expression? The situation must be different, compared with blood vessels where you have this arterial venous expression. Whether the mechanism is trans signalling between two adjacent cells (the conventional mechanism) or a different mechanism (we now have quite nice evidence that ephrinB2 is acting in a cell autonomous fashion, mediating cell shape changes even without contacting other cells) it is still an interesting question, but it will be challenging to investigate this in vivo. Betsholtz: Have you established endothelial cell cultures from these mutants? It would be interesting to follow the migratory properties of single cells, for example. Mäkinen: I am now in the process of doing this. It will be interesting to see whether we can stimulate them for example to sprout. Adams: Are the defects that you see in the mutant mice directly caused by ephrinB2, or is there an arrest in lymphangiogenesis before differentiation? Mäkinen: Some of the defects we see may be indirect, but I think there is also a direct effect, for example on the valve formation. I sometimes see abnormal valve structures forming in the ephrinB2 mutants. It is not known how the valve process occurs in the first place. In the mutants I occasionally see constrictions starting to form, but the valve that forms is completely abnormal. I would assume that this is a direct effect caused by disrupted ephrinB2 signalling. Dejana: One can summarize the problem in the ephrinB2 mutant as a defect in maturation of the larger lymphatics. Would it be correct to say that it is a defect in the maturation and cell specification of these vessels? Mäkinen: The earliest defect we see is the defect in sprouting. We can’t be sure whether there is some signal that is given after the remodelling has occurred, so that the vessels then become specified as capillaries and collecting vessels. In this case lack of remodelling in the ephrinB2 mutants may cause a developmental arrest and the subsequent maturation events would not occur. An alternative is that ephrinB2

100

MÄKINEN & ALITALO

directly controls the behaviour of endothelial cells in the collecting vessels, and that the defects in sprouting versus cell specification are two distinct events. Dejana: Is there any information about the intracellular partners that may link through the PDZ binding domain of ephrinB2? Mäkinen: We analysed the expression pattern of known ephrinB2-interacting PDZ domain-containing proteins and looked at the subcellular localization of some of them. Several of these molecules were expressed in the lymphatics, but they were also in the blood vasculature: therefore there was no clear candidate. However, we found that some of the molecules, such as PDZ-RGS3 showed different subcellular localization in the wild-type versus mutant lymphatic vessels, suggesting that it may be a candidate molecule involved in ephrinB2 signalling in lymphatic vessels. Ye: I have a question related to the postnatal dependency of VEGFR3. Is anything known about the switch of basement membrane molecular composition around week 2? Mäkinen: Not to my knowledge. Ye: What about pericyte coverage? Mäkinen: The pericytes are recruited into collecting vessels at around that stage. VEGFR3 is more strongly expressed in the lymphatic capillaries than in the collecting vessels even in the adult. Gerhardt: I was wondering about the specificity of the observation that there is a remodelling phase where new sprouts form. In particular, this protrusive activity of the cells seems to be defective in your ephrinB2 mutants. What does this tell us about how the first lymph vessels are forming? Does this mean that they don’t sprout, or that they use different molecules? Mäkinen: The embryonic lymphatic vessels do sprout from the initial lymphatic sacs. I haven’t analysed carefully how the initial sprouting during embryonic development occurs in the ephrinB2 mutants, but since the capillary plexus appears relatively normal at birth, one would not expect any major defects. Weinstein: This gets back to the old models: is it all just sprouting from the initial vein derived from the lateral mesoderm or is there a contribution of mesenchyme? Is it a more vasculogenesis-like or angiogenesis-like process? Perhaps this initial plexus formation is a more vasculogenesis-like process. Mäkinen: If we stain, for example, embryonic day 14 skin we can see active sprouting from the lymphatic sacs. Staining of different development stages suggests that the main process of vasculature expansion is by sprouting from the sacs. Betsholtz: Is there always a continuous connection between the lymphatic vessels and the sac region, or are they ever disconnected? Mäkinen: There are two main sites from which the sprouting starts: the jugular region and the perimesonephric region. Staining of the vasculature during different developmental stages suggests that the vasculature expands by sprouting from these two regions.

LYMPHANGIOGENESIS

101

Betsholtz: Do all the plexuses that you see in the lymphatic system always connect back to these two sites? If so that would argue for sprouting angiogenesis but against vasculogenic formation. Mäkinen: Yes, at least in the skin it looks like they are connected. Wilting: These are mice that you stain with LacZ-VEGFR3 reporter gene. This won’t pick up angioblasts. Mäkinen: I am not saying that there could not be a contribution from angioblasts, but there is clearly sprouting. In the skin we see a continuous network starting from both sides of the sac, extending and finally fusing in the dorsal midline. Wilting: You have shown some LYVE1 staining. I don’t see single cells here. Mäkinen: Around the blood vessels, for example, we do see some single cells. The stainings I showed were from adult skin, and there I don’t see so many single LYVE1 positive cells in the surrounding tissues. However, if I stain embryonic stages, many more single cells are visible. Lammert: I’d like to ask about the basement membrane of the lymph vessels. Some people say that there is no basement membrane around these vessels. Can you comment on where matrix proteins are expressed? Mäkinen: In the literature it is said that lymphatic vessels have discontinuous basement membrane. I don’t know what this means exactly, but I guess it means that there is a much thinner basement membrane as compared with the blood vasculature. I assume that the endothelial cells produce at least some matrix proteins. Shovlin: What can you tell us about the regulation of this system? The reason I ask is because Alk1 null mice have down-regulated ephrinB2 expression, and defects in identity of their arterial and venous endothelial cells. But I don’t think that they (or the human equivalent) have lymphatic abnormalities. Can you explain how we could see these different end results? Mäkinen: I don’t think Alk1 is expressed in the lymphatic endothelium. Shovlin: What would be regulating ephrinB2 expression then? Mäkinen: I don’t know. Yancopoulos: I am not so sure that Ephrin Bs are proximal downstream regulated by Alk. Uv: I have a more general question. What do the vessels look like? Are they blind ended? Are they constructed by a single cell, or several? Mäkinen: It is a single cell layer, but multiple cells are forming the lumen. The capillaries are blind ended. Uv: From what you showed there are single small blind-ended tubes that collected fluid. Mäkinen: Maybe it was confusing in the drawing that I showed: what I marked in the cartoon was anchoring fi laments extending from the endothelial cells. There is a blind-ended capillary and endothelial cells are attached to the tissue

102

MÄKINEN & ALITALO

by fi laments which serve to open the vessels when there is high pressure in the tissue. Uv: If you take a cross-section of the tip of the tube, what does it look like? Is it one cell wrapped around? Mäkinen: In the cross sections the lymphatic capillaries often look collapsed, so I assume that it would be difficult to get a clear picture of the ultrastructure. I am not sure about the very tip of the capillary end, whether it is a single cell, but the lumen is formed by several endothelial cells. Shibuya: Is the ephrinB2 signalling necessary for the cells to differentiate into large-type lymphatic endothelial cells from regular small-type collecting ones? Mäkinen: Yes, it is possible. After the remodelling, ephrinB2 expression becomes restricted to collecting vessels. In wild-type mice when you see the smooth muscle cells around the vessel you don’t see LYVE1 staining, but in the ephrinB2 mutants it doesn’t seem to matter whether the smooth muscle cells are there, the vessels are all positive for LYVE1. This might suggest that in the absence of ephrinB2 signalling the endothelial cells in the collecting lymphatic vessels fail to down-regulate LYVE1 expression, which may be an indication of failure in differentiation. Drake: You have this capillary and then a sprout. There is an identity shift, and this is followed by sprouting. Is this when Prox1 starts to be expressed in a group of the sprout cells that are becoming a lymphatic? Mäkinen: During earlier embryonic development, yes. After the initial differentiation, Prox1 expression stays in all types of lymphatic endothelial cells. Drake: But the blood endothelial cells aren’t expressing Prox1. So now you have a sprout extending into some connective tissue. Where is the signal in the matrix? There are two cells juxtaposed to each other: one of them is becoming a different type of cell. Where is the signal coming from? Mäkinen: The earliest development is that within the vein, on one side there is expression of Prox1. In Prox1 null embryos there is an initial sprout, but the sprouting is arrested in the absence of Prox1. There could be an attraction to something which then gives a signal that turns on Prox1 in these cells. After that the whole differentiation program starts. It is not currently known what the signal that initiates the budding and Prox1 expression is. Drake: Is this when the LYVE1 marker comes on? Mäkinen: LYVE1 is expressed earlier than Prox1, and in a wider area in the vein. Only a subset of LYVE1 positive cells up-regulate Prox1 and differentiate into lymphatic endothelial cells. Subsequently LYVE1 expression becomes restricted to Prox1-positive cells. Yancopoulos: Is VEGFC required for the initial sprouting of the lymphatics? Mäkinen: It isn’t required for Prox1 expression, so I’d assume that the initial small bud is forming, Prox1 is up-regulated, but the further sprouting is blocked and the cells get stuck into the veins and don’t migrate out.

LYMPHANGIOGENESIS

103

Yancopoulos: So VEGFC is required for that primary outgrowth. Mäkinen: I am not sure you can say that. Yancopoulos: In the VEGFC knockout, what don’t you see? Mäkinen: Prox1 expression is induced, but the lymphatic endothelial cells do not sprout out. What I mean by ‘initial budding’ is that in the Prox1-deficient mice it has been characterized that there is an initial bud forming, which is perhaps normally required for induction of Prox1 expression in these cells, but in the absence of Prox1 the budding is arrested. In VEGFC knockout mice Prox1 is induced, so perhaps the very early bud is forming, but the sprouting initiated after lymphatic differentiation fails. Yancopoulos: I am fascinated by your observation that when you treat the adult animals and get regression, it looks like you can get regrowth even in the setting of continued blockade. This suggests that there is a second growth factor system that can induce sprouting, even though both the primary and secondary hierarchical growth both require VEGFC. What is the growth factor? Mäkinen: I don’t know. Sometimes when the vessels regenerate they form a slightly abnormal pattern, suggesting that VEGFC/VEGFR3 signalling may still be required to achieve a completely normal regrowth. However, the data indeed suggest that another signalling system besides VEGFC/VEGFR3 is involved. Yancopoulos: Have you tried the regression regrowth model in the ephrinB2 knockouts? It would be fascinating to see whether that regrowth is ephrinB2 dependent. Mäkinen: The problem is that the ephrinB2 mutants die during the first two weeks after birth. Betsholtz: I’d like to follow up on the possibility of other mechanisms for lymphatic development in the mice in which soluble VEGFR3 is expressed from the K14 promoter. Do these mice produce sufficient amounts of inhibitor to keep VEGF signalling down throughout adulthood? Mäkinen: We measured the serum levels of the VEGFR3 Ig-fusion protein in the transgenic K14 mice, and the levels are higher than neutralizing levels. In addition, the adenovirally infected animals show very high levels, even when the lymphatic regrowth starts. Terhi Karpanen also checked that the soluble VEGFR3 protein precipitated from the serum can still bind the growth factors. Weinstein: Do you get hyper-induction of the ligands with all this stuff floating around, and then you are swamping the inhibitor? Mäkinen: If you have an excess of ligand, you would expect that the receptor precipitated from the serum would no longer be free to bind the ligand. I think Terhi Karpanen also measured the serum levels of the ligands by ELISA, and they were not up-regulated. It is striking that you can completely inhibit tumour lymphangiogenesis in the same conditions, but the normal adult vessels do not regress.

104

MÄKINEN & ALITALO

Kitajewski: Isn’t there a low level of VEGFR2 expression in lymphatic endothelium? Mäkinen: Yes. Kitajewski: Couldn’t there be an up-regulation of VEGFR2 after R3 blockade? Mäkinen: I am not aware of that. In normal conditions VEGFR2 is mostly expressed in the collecting vessels, not in the capillaries, while VEGFR3 levels are higher in the capillaries. Kitajewski: VEGFR3 mutant mice die due to an angiogenic remodelling defect, before lymphatics develop. Is there any further insight about what role VEGFR3 is playing in angiogenesis? Mäkinen: I am not aware of this. Lammert: I have a question about these sprouts. They look like capillary sprouts, but what happens to them in the Dll4 knockouts? Yancopoulos: It’s a bit complicated with the Dll4 knockouts because the homozygotes are embryonic lethal. The heterozygotes are embryonic lethal except on some backgrounds. Very few animals survive, but in the survivors we don’t see lymphatic problems. Now we have pharmacological blockers we could look at this again. When you see regression in adult animals with VEGFR3 treatment, is there lymphoedema? Mäkinen: No, but this is difficult to see in mice even when there are no lymphatic vessels in the skin. Kitajewski: We are about to report that Notch 1 and 4 are expressed in a subset of lymphatics. We haven’t seen lymphatic defects in the Notch mutant animals yet. Uv: Do the lymphatic vessels grow with a single cell leading, like in the capillaries? Mäkinen: It looks more like one cell is leading the sprout, as in blood capillaries. However, the ‘lymphatic tip cell’ has not yet been characterized as in blood vessels. Epstein: What is the embryonic origin of the smooth muscle cells, and does the neural crest contribute? Mäkinen: The origin of lymphatic smooth muscle cells is not known. Epstein: I am interested in the interaction between the endothelium and smooth muscle. This might emerge as a theme in other talks. If neural crest does contribute to a small portion of lymphatics, as it does to the blood vessels, it might be an interesting area to focus on. Augustin: This symposium is on developmental processes, but I would like to raise the issue of invasive lymphangiogenesis in the adult. You have shown beautiful pictures of the sprouting processes in the embryo. In the adult, it is self evident that tumour cells need access to lymphatics to metastasize. Yet, I am increasingly puzzled about the contribution of invasive lymphangiogenesis in the adult. Several

LYMPHANGIOGENESIS

105

years ago the transgenic overexpression of VEGF and VEGFC was first done in the RipTag model. There is dramatic invasion of blood vessels into pancreatic nodules upon VEGF expression. Upon VEGFC expression, there is an intense expansion of lymphatic lacunas around the islands, but there is no invasion of lymph vessels into the neoplastic islands. When we look for lymphatics in human tumours, we find intratumoral lymphatics in all tumours, but we mostly find them in tumours from organs that have a lot of pre-existing lymphatics, such as tumour of the skin or colon. It is therefore my impression that intratumoral lymphatics mostly reflect co-option from pre-existing lymphatics and not invasion. Similarly, blood endothelial cells are strongly invasive in three-dimensional in vitro sprouting angiogenesis assays. Yet, pure lymphatic endothelial cells have very little invasive sprouting angiogenesis capacity in vitro.

Blockade of Dll4 inhibits tumour growth by promoting non-productive angiogenesis1 Irene Noguera-Troise, Christopher Daly, Nicholas J. Papadopoulos, Sandra Coetzee, Pat Boland, Nicholas W. Gale, Hsin Chieh Lin, George D. Yancopoulos2 and Gavin Thurston Regeneron Research Laboratories, 777 Old Saw Mill River Road, Tarrytown, New York 10591, USA

Abstract. Tumour growth requires accompanying expansion of the host vasculature, with tumour progression often correlated with vascular density. Vascular endothelial growth factor (VEGF) is the best-characterized inducer of tumour angiogenesis. We report that VEGF dynamically regulates tumour endothelial expression of Delta-like ligand 4 (Dll4), which was previously shown to be absolutely required for normal embryonic vascular development. To define Dll4 function in tumour angiogenesis, we manipulated this pathway in murine tumour models using several approaches. Here we show that blockade resulted in markedly increased tumour vascularity, associated with enhanced angiogenic sprouting and branching. Paradoxically, this increased vascularity was nonproductive—as shown by poor perfusion and increased hypoxia, and most importantly, by decreased tumour growth—even for tumours resistant to anti-VEGF therapy. Thus, VEGF-induced Dll4 acts as a negative regulator of tumour angiogenesis; its blockade results in a striking uncoupling of tumour growth from vessel density, presenting a novel therapeutic approach even for tumours resistant to anti-VEGF therapies. 2007 Vascular development. Wiley, Chichester (Novartis Foundation Symposium 283) p 106–125

Tumour growth depends on expansion of the host vasculature into the tumour, through the process of tumour angiogenesis (Folkman 1992). The connection between tumour growth and angiogenesis prompted the development of several approaches to limit tumour angiogenesis and thus control tumour growth. The best validated of these approaches involves blockade of the vascular endothelial growth factor (VEGF) pathway. Blockade of VEGF controls tumour growth in 1

This paper is reproduced in modified form from Noguera-Troise et al (2006). Reprinted with permission from Macmillan Publishers Ltd: Nature 444:1032–1037, copyright 2006. 2 This paper was presented at the symposium by George D. Yancopoulos, to whom correspondence should be addressed. 106

DLL4 AND ANGIOGENESIS

107

numerous preclinical models (Ferrara 2004, Rudge et al 2005), and recent results show that potent blockers of VEGF can completely prevent tumour angiogenesis in some models, thereby severely inhibiting tumour growth (Holash et al 2002). The promise of VEGF-blocking approaches has recently been realized in the clinic, as a VEGF-blocking antibody has been shown to have important effects on tumour progression and overall survival in cancer patients (Hurwitz et al 2004, Laskin & Sandler 2005). However, despite the critical role for VEGF in tumour angiogenesis, it is also clear that in some cases tumour growth and angiogenesis can proceed even in the face of potent VEGF blockade (Casanovas et al 2005, Jain et al 2006, Kerbel et al 2001). Thus, additional angiogenesis-targeted therapies are necessary for tumours resistant to VEGF blockade. Mouse genetic studies have demonstrated that, in addition to the VEGF pathway, other signalling pathways are also required for normal embryonic vascular development (for reviews see Yancopoulos et al 2000, Jain 2005a, Carmeliet 2005), raising the possibility that these pathways may also be important during tumour angiogenesis. One signalling pathway implicated in vascular development by gene deletion studies is the Notch pathway (Shawber & Kitajewski 2004). On binding a transmembrane ligand from the Delta/Jagged families, Notch transmembrane receptors generally provide signals to guide cell fate decisions (Artavanis-Tsakonas et al 1999, Gridley 2001). In particular, Delta-like ligand 4 (Dll4) is absolutely required for normal vascular development (Duarte et al 2004, Gale et al 2004, Krebs et al 2004) and is strongly expressed in tumour vessels (Gale et al 2004, Mailhos et al 2001, Patel et al 2005). To determine whether the Dll4/Notch pathway has a role during tumour angiogenesis, we manipulated this pathway in experimental tumour models in mice using a variety of genetic and pharmacological approaches. We report that the Dll4/Notch pathway is a critical negative regulator of tumour angiogenesis, acting to restrain excessive VEGF-induced vascular sprouting and angiogenesis. Increased Dll4/Notch activity resulted in decreased tumour vascular density, whereas blockade of activity resulted in markedly increased vessel density. Paradoxically, this increased vascularity seemed to be non-productive and resulted in decreased tumour growth, even for tumours that are resistant to anti-VEGF therapy. Our findings provide a striking example of an uncoupling of tumour growth from tumour vascular density, and support the model that the Dll4/Notch pathway normally acts as a negative regulator of angiogenic sprouting induced by VEGF or other pathways. VEGF induces high expression of Dll4 in tumour vessels To confirm and extend earlier reports that Dll4 expression was markedly and specifically induced in blood vessels during tumour angiogenesis (Gale et al 2004,

108

NOGUERA-TROISE ET AL

Mailhos et al 2001, Patel et al 2005), we used a combination of Dll4 detection approaches in two different tumour models. First, we exploited ‘Dll4 reporter mice’ in which a β -galactosidase reporter gene was driven by the Dll4 promoter (Gale et al 2004). Lewis lung carcinomas implanted into these mice showed strong reporter-based staining of tumour vessels, apparently at higher levels than of vessels in surrounding normal tissue (Figs 1a, b). At higher resolution, immunostaining for the β -galactosidase reporter protein and comparison with adjacent sections in which all vessels were immunostained with CD-31/PECAM antibodies revealed strong Dll4 reporter expression in tumour blood vessels and relatively weak staining in adjacent subcutaneous and dermal blood vessels (Gale et al 2004). To confirm that the β -galactosidase/Dll4 reporter (Dll4–LacZ ) construct marked sites of Dll4 protein, we generated polyclonal antibodies to murine Dll4; these antibodies also immunostained tumour blood vessels selectively (Figs 1c, d). These findings were further confi rmed by in situ hybridization for Dll4 messenger RNA, which revealed prominent expression in the vessels of C6 glioma tumours (Mailhos et al 2001). Thus, Dll4 is indeed specifically expressed in the tumour vasculature, particularly in the smaller vessels. Moreover, expression of Dll4 was dependent on continuous VEGF signalling, because blockade of VEGF with VEGF Trap, a recombinant soluble receptor that potently blocks VEGFA and placental growth factor (PlGF) (Holash et al 2002), caused a rapid and marked decrease in the expression of Dll4 by tumour vessels (Fig. 1e). Activating and blocking the Dll4/Notch pathway in tumours To manipulate the Dll4/Notch pathway in tumours, we first exploited a retroviral approach to overexpress forms of Dll4, which we reasoned would serve as blockers or activators, in tumour cells. On the basis of previous studies that used soluble versions of Dll to inhibit Notch signalling (Hicks et al 2002), we generated retroviral vectors encoding a soluble dimerized version of Dll4 in which the extracellular region of Dll4 was fused to the human IgG1 Fc constant region (termed Dll4–Fc) as a presumed blocker, as well as full-length membrane-bound Dll4 that presumably would act as an activator; these constructs were transduced into rat C6 tumour cells to produce C6 Dll4–Fc and C6 Dll4 cells. Importantly, C6 tumour cells that overexpressed Dll4–Fc or Dll4 did not have different growth characteristics in vitro than control tumour cells (data not shown). The expected changes on Notch signalling in the host (mouse) stroma of subcutaneously implanted C6 rat tumour cells were confirmed by quantitative messenger RNA analyses using probes specific for the mouse versions (to detect expression in the host stromal cells and not in the rat tumour cells) of three genes that are characteristic targets of Notch signalling (HES1, HEY2 and NRARP) (Taylor et al 2002, Iso et al 2003, Shawber et al 2003, Karsan 2005, Lamar et al

DLL4 AND ANGIOGENESIS

[Image not available in the electronic edition].

109

110

NOGUERA-TROISE ET AL

2001, Krebs et al 2001). That is, in C6 Dll4–Fc tumours, host Notch signalling was consistently reduced as reflected by decreased levels of these target genes, whereas in C6 Dll4 tumours the Notch pathway was activated. Similarly, in coculture studies in which the transduced rat tumour cells were mixed with human umbilical vein endothelial cells, expression levels of the HES1, HEY2 and NRARP genes in the cultured endothelial cells (assayed using human specific probes so as to specifically detect the endothelial versions of these transcripts) were reduced by co-culture with C6 Dll4–Fc cells and induced by C6 Dll4 cells. Finally, treatment of cultured human umbilical vein endothelial cells with purified Dll4–Fc protein rapidly and consistently repressed Notch signalling, as reflected by decreased expression of these target genes. Blockade of Dll4/Notch results in decreased tumour growth To explore the effects of Dll4/Notch pathway manipulation in tumours, we next examined tumours derived from C6 Dll4–Fc and C6 Dll4 cells for their vascular morphologies and tumour growth rates. Strikingly, Dll4–Fc and full-length Dll4 seemed to promote reciprocal changes in the tumour vasculature: the vasculature in C6 Dll4–Fc tumours was much more highly branched and had more fi ne interconnections than that of control tumours (Figs 2a–f). The leading front of the vessels was replete with sprouts and fi lopodia (Figs 2e, h). In contrast, the vasculature in C6 Dll4 tumours was notably straighter and relatively unbranched (Figs 2c, f ), and relatively devoid of sprouts and fi lopodia (Figs 2f, i). These obvious morphological changes were reflected by quantitation of vascular area densities in these tumours, with the C6 Dll4–Fc tumours exhibiting increased vascular density as compared with control tumours, whereas the C6 Dll4 tumours exhibited slightly decreased vascular density (Fig. 2j). Paradoxically, the effects of Dll4–Fc and full-length Dll4 on vascular density were opposite to those that might be expected with respect to their effects on tumour growth. Despite increased vascular density, C6 Dll4–Fc tumours were consistently smaller than control tumours, whereas C6 Dll4 tumours were not substantially different in size from control tumours for small tumours (Fig. 2k) or when C6 Dll4 tumours were harvested at larger sizes. These results have two important implications: first, that the Dll4/Notch pathway normally serves as a negative regulator of sprouting and branching activity during tumour angiogenesis, so that blockade of this pathway results in increased tumour angiogenesis; and second, that the increased angiogenesis resulting from blockade of this pathway is in some sense ‘non-productive’, such that it does not support more robust tumour growth and instead seems to be associated with reduced tumour growth.

DLL4 AND ANGIOGENESIS

[Image not available in the electronic edition].

111

112

NOGUERA-TROISE ET AL

Blockade of Dll4/Notch increases tumour hypoxia To account for the apparent paradox of increased tumour vessel density and decreased tumour growth in C6 Dll4–Fc tumours, we examined the possibility that the increased network of vessels might not be optimally functional. In agreement with this possibility, histological assessment showed more extensive tumour hypoxia in C6 Dll4–Fc tumours than in control tumours (Figs 3a, b; hypoxic regions stained in black). Moreover, whereas the hypoxic region in control tumours was separated from the growing front of tumour vessels by an avascular zone that was itself not hypoxic (corresponding to the oxygen diffusion distance; white asterisks, Fig. 3d), areas of hypoxia were interspersed with the tumour vasculature in C6 Dll4–Fc tumours (arrowheads, Figs 3b, e), indicating that this vasculature was not efficiently delivering oxygen to the surrounding tumour. Quantitative analysis showed that Dll4–Fc tumours contained sevenfold more total hypoxic area within the vascularized tumour (Fig. 3g), and that the hypoxic rim was less separated from the leading front of tumour vessels (Fig. 3h). Corresponding values

DLL4 AND ANGIOGENESIS

[Image not available in the electronic edition].

113

114

NOGUERA-TROISE ET AL

in the C6 Dll4 tumours were not substantially different from controls (Fig. 3c), although there was more variability in the distance from the vascular front to the hypoxic rim (white asterisks, Fig. 3f). The increase in tumour hypoxia in the C6 Dll4–Fc tumours suggested that the dense network of vessels was not fully perfused. We compared the distribution of vessel perfusion (marked by intravascular lectin as a tracer) with the immunohistochemical staining of endothelial cells (stained with CD31/PECAM-1 antibodies) as a marker of total vasculature. Most of the larger vessels of control tumours were perfused, although—as expected—these vessels were associated with some nonperfused sprouts and smaller vessels emanating from the larger vessels (Fig. 3i). In contrast, many of the vascular processes in the C6 Dll4–Fc tumours were not perfused (Figs 3j, m), suggesting that the increased vessel density seen in these tumours is not part of a functional vascular network. Reciprocally, the relatively straight and unbranched vessels seen in the C6 Dll4 tumours were almost completely perfused (Figs 3k, n). Together, the tumour hypoxia and perfusion analyses support the notion that the increased vascular network that results from inhibition of the Dll4/Notch pathway (by Dll4–Fc) is not optimally functional and is instead ‘non-productive’. Systemic Dll4–Fc decreases tumour growth To confirm and extend the finding that locally produced Dll4–Fc promotes excessive angiogenesis that paradoxically blunts tumour growth, we used an adenoviral delivery approach to determine whether systemic Dll4–Fc could also produce this effect. Adenoviruses expressing Dll4–Fc, or human Fc (hFc) as control, were injected intravenously into mice at the time of implanting subcutaneous C6 tumours. The intravenously injected adenovirus infects hepatocytes, and in the case of Dll4–Fc or hFc, produced high serum levels of the encoded protein of 74 ± 20 µg ml−1 (range of 50–100 µg ml−1, n = 5 mice). Circulating Dll4–Fc resulted in an approximately 70% reduction in the size of subcutaneous C6 tumours (Fig. 4a), whereas circulating control hFc had no effect. When examined by histology, circulating Dll4–Fc also caused an increase in the density of the tumour vessels (Fig. 4b). The overall increased vessel density produced by circulating Dll4–Fc was associated with a dense mesh of highly branched and sprouted vessels, particularly at the leading front (Figs 4c–f), similar to that produced by Dll4–Fc overexpression in the tumour cells. Gene expression analysis confirmed that systemic Dll4– Fc suppressed the Notch pathway in the tumour vessel, as indicated by decreased expression of HES1 and NRARP (data not shown). Thus, inhibition of Dll4/ Notch signalling by either local or systemic Dll4–Fc results in smaller C6 tumours and excessive but apparently non-productive tumour angiogenesis. Importantly, the systemic treatment did not seem to have untoward effects on the host animals;

DLL4 AND ANGIOGENESIS

[Image not available in the electronic edition].

115

116

NOGUERA-TROISE ET AL

in addition, preliminary analysis of normal tissues did not reveal obvious changes in tissue vascularity Dll4–Fc or Dll4-blocking antibody act in multiple tumour models To further extend the above findings, we used systemic injection of purified recombinant Dll4–Fc protein as the treatment, and tried additional tumour models. The above studies were carried out using C6 gliomas, which are relatively sensitive to the effects of VEGF blockade (Holash et al 2002), so we assessed the response to Dll4/Notch blockade in other tumours that are more resistant to VEGF blockade. In previous experiments using both bevacizumab (Avastin) and VEGF Trap, we had developed a model of HT1080 tumours (HT1080-resistance model; RM), which is relatively resistant to both Avastin and VEGF Trap (G. Thurston, I. Noguera-Troise and J. Rudge, unpublished results). In contrast to Avastin and VEGF Trap, Dll4–Fc protein was quite effective in reducing the growth of HT1080-RM tumours (Fig. 5a). In a separate experiment (Fig. 5b), we assessed the growth curves of HT1080-RM tumours treated with Dll4–Fc, VEGF Trap or control protein (all at 25 mg kg−1, three times per week; treatment began when tumours were approximately 100 mm3 in size, arrow). Dll4–Fc treatment resulted in a prolonged suppression of tumour growth, whereas VEGF Trap had almost no impact on tumour growth in this resistant tumour model. Dll4–Fc treatment of HT1080-RM tumours also had a marked effect on tumour vessels. The alreadydense vasculature of control HT1080 tumours (Fig. 5c) was further increased by treatment with Dll4–Fc (Fig. 5d), inducing an apparent increase in vascular sprouting and branching, and an apparent disorganization to the network. To verify the specificity of blocking with Dll4–Fc, we also generated polyclonal antibodies to the extracellular portion of Dll4, which inhibited the binding of Dll4

DLL4 AND ANGIOGENESIS

[Image not available in the electronic edition].

117

118

NOGUERA-TROISE ET AL

to Notch1 in in vitro assays. Systemic treatment of mice bearing resistant HT1080RM tumours with blocking Dll4 antibodies also reduced tumour growth (Fig. 5e) and caused marked changes in tumour vessels (data not shown). As for HT1080 tumours, growth of mouse mammary tumours, which are also resistant to VEGF blockade (not shown), was strongly suppressed by systemic treatment with Dll4–Fc. Again, the vascularity of mouse mammary tumours was further increased by treatment with Dll4–Fc. Thus, as with C6 tumours, treatment of other tumours with systemic Dll4–Fc resulted in decreased tumour growth, accompanied by a denser and more highly branched tumour vasculature. Discussion To determine whether the Dll4/Notch pathway has a role during tumour angiogenesis, we manipulated this pathway in tumours using a variety of approaches. Our findings suggest that tumour-derived VEGF induces Dll4 expression in angiogenic endothelial cells as a critical negative regulator of vascular growth, acting to restrain excessive vascular sprouting and branching, and allowing angiogenesis to proceed at a productive rate. Thus, increasing Dll4/Notch activity resulted in decreased vascular density associated with less sprouting and branching of the vascular network. In contrast, Dll4/Notch blockade was associated with enhanced angiogenic sprouting and branching, resulting in a marked increase in tumour vessel density but a decrease in vessel function. Previously, angiogenesisbased treatment of tumours has focused on trying to block angiogenesis; however, our results using Dll4 blockade suggest an alternative approach based on promoting ‘non-functionality’ in the growing tumour vasculature. Although our studies suggest that VEGF blockade may be equally or more effective than Dll4 blockade in many tumour models, certain models that are resistant to VEGF blockade can still be sensitive to Dll4 blockers. We, and others, have shown that Dll4 is specifically expressed in remodelling vessels and is the major Notch ligand in the vasculature. Thus, rather than a general blockade of the Notch pathway, specific blockade of Dll4 may lead to more specific disruption of tumour growth without significant impairment of Notch function in normal host tissues, and thus might be well tolerated in long-term treatments. It seems likely that biological therapeutic agents, which can be specific to a particular ligand or receptor in this complex pathway, may prove more potent and specific than more general pathway blockers, such as the γ -secretase inhibitors that not only block all Notch signalling but also other important γ -secretase-mediated signalling as well. Our findings provide a striking example of an uncoupling of tumour growth from tumour vascular density. Although a large literature supports the notion that tumour growth rate may correlate with tumour vascular density, other studies argue

DLL4 AND ANGIOGENESIS

119

that tumour angiogenesis must be regulated to be productive. For example, a recent study suggests that tumours may have higher vascular densities than is necessary to support their growth, and thus tumour angiogenesis may often exceed an optimally productive rate (Krneta et al 2006). Consistent with the concept of excessive tumour vessel density, some recent studies suggest that pruning of the vasculature might actually improve tumour perfusion and oxygenation (Lee et al 2000, Jain 2002, 2005b). The present studies with blockers of Dll4/Notch seem to provide the other side of the argument; in particular, that Dll4 blockade may further compromise tumour vasculature function by causing excessive non-productive angiogenesis, which can in turn inhibit tumour growth. The overall message seems to be that even tumour vascular networks require a regulated balance of growth factors to generate a hierarchy of well-organized and well-functioning vessels. VEGF clearly has a key angiogenic role in a wide variety of tumours, but Dll4 blockade may present a new therapeutic opportunity in cancer, and one that might be beneficial for patients with tumours that are resistant to anti-VEGF therapies. Acknowledgements We acknowledge the following Regeneron colleagues: Y. Wei for gene expression analysis, A. Adler, A. Rafique, B. Li, H. Huang, E. Pasnikowski, J. McClain, E. Burova, D. Hylton, P. Burfeind and J. Griffiths for technical assistance, S. Staton for assistance with graphics, and S. Wiegand, I. Lobov, T. Daly, S. Davis, E. Ioffe, J. Holash and J. Rudge for scientific input.

References Artavanis-Tsakonas S, Rand MD, Lake RJ 1999 Notch signaling: cell fate control and signal integration in development. Science 284:770–776 Carmeliet P 2005 Angiogenesis in life, disease and medicine. Nature 438:932–936 Casanovas O, Hicklin DJ, Bergers G, Hanahan D 2005 Drug resistance by evasion of antiangiogenic targeting of VEGF signaling in late-stage pancreatic islet tumors. Cancer Cell 8:299–309 Duarte A, Hirashima M, Benedito R et al 2004 Dosage-sensitive requirement for mouse Dll4 in artery development. Genes Dev 18:2474–2478 Ferrara N 2004 Vascular endothelial growth factor as a target for anticancer therapy. Oncologist 9 (Suppl 1):2–10 Folkman J 1992 The role of angiogenesis in tumor growth. Semin Cancer Biol 3:65–71 Gale NW, Dominguez MG, Noguera I et al 2004 Haploinsufficiency of delta-like 4 ligand results in embryonic lethality due to major defects in arterial and vascular development. Proc Natl Acad Sci USA 101:15949–15954 Gridley T 2001 Notch signaling during vascular development. Proc Natl Acad Sci USA 98: 5377–5378 Hicks C, Ladi E, Lindsell C 2002 A secreted Delta1–Fc fusion protein functions both as an activator and inhibitor of Notch1 signaling. J Neurosci Res 68:655–667 Holash J, Davis S, Papadopoulos N et al 2002 VEGF-Trap: a VEGF blocker with potent antitumor effects. Proc Natl Acad Sci USA 99:11393–11398 Hurwitz H, Fehrenbacher L, Novotny W et al 2004 Bevacizumab plus irinotecan, fluorouracil, and leucovorin for metastatic colorectal cancer. N Engl J Med 350:2335–2342

120

NOGUERA-TROISE ET AL

Iso T, Kedes L, Hamamori Y 2003 HES and HERP families: multiple effectors of the Notch signaling pathway. J Cell Physiol 194:237–255 Jain RK 2002 Tumor angiogenesis and accessibility: role of vascular endothelial growth factor. Semin Oncol 29:3–9 Jain RK 2005a Normalization of tumor vasculature: an emerging concept in antiangiogenic therapy. Science 307:58–62 Jain RK 2005b Antiangiogenic therapy for cancer: current and emerging concepts. Oncology 19:7–16 Jain RK, Duda DG, Clark JW, Loeffler JS 2006 Lessons from phase III clinical trials on antiVEGF therapy for cancer. Nat Clin Pract Oncol 3:24–40 Karsan A 2005 The role of notch in modeling and maintaining the vasculature. Can J Physiol Pharmacol 83:14–23 Kerbel RS, Yu J, Tran J et al 2001 Possible mechanisms of acquired resistance to anti-angiogenic drugs: implications for the use of combination therapy approaches. Cancer Metastasis Rev 20:79–86 Krebs LT, Deftos ML, Bevan MJ, Gridley T 2001 The Nrarp gene encodes an ankyrin-repeat protein that is transcriptionally regulated by the Notch signaling pathway. Dev Biol 238:110–119 Krebs LT, Shutter JR, Tanigaki K, Honjo T, Stark KL, Gridley T 2004 Haploinsufficient lethality and formation of arteriovenous malformations in Notch pathway mutants. Genes Dev 18:2469–2473 Krneta J, Kroll J, Alves F 2006 Dissociation of angiogenesis and tumorigenesis in follistatinand activin-expressing tumors. Cancer Res 66:5686–5695 Lamar E, Deblandre G, Wettstein D 2001 Nrarp is a novel intracellular component of the Notch signaling pathway. Genes Dev 15:1885–1899 Laskin JJ, Sandler AB 2005 First-line treatment for advanced non-small-cell lung cancer. Oncology 19:1671–6; discussion 1678–80 Lee CG, Heijn M, di Tomaso E 2000 Anti-vascular endothelial growth factor treatment augments tumor radiation response under normoxic or hypoxic conditions. Cancer Res 60: 5565–5570 Mailhos C, Modlich U, Lewis J, Harris A, Bicknell R, Ish-Horowicz D 2001 Delta4, an endothelial specific notch ligand expressed at sites of physiological and tumor angiogenesis. Differentiation 69:135–144 Noguera-Troise I, Daly C, Papadopoulos NJ et al 2006 Blockade of Dll4 inhibits tumour growth by promoting non-productive angiogenesis. Nature 444:1032–1037 Patel NS, Li JL, Generali D, Poulsom R, Cranston DW, Harris AL 2005 Up-regulation of delta-like 4 ligand in human tumor vasculature and the role of basal expression in endothelial cell function. Cancer Res 65:8690–8697 Rudge JS, Thurston G, Davis S et al 2005 VEGF trap as a novel antiangiogenic treatment currently in clinical trials for cancer and eye diseases, and VelociGene-based discovery of the next generation of angiogenesis targets. Cold Spring Harb Symp Quant Biol 70: 411–418 Shawber CJ, Kitajewski J 2004 Notch function in the vasculature: insights from zebrafish, mouse and man. Bioessays 26:225–234 Shawber CJ, Das I, Francisco E, Kitajewski J 2003 Notch signaling in primary endothelial cells. Ann N Y Acad Sci 995:162–170 Taylor KL, Henderson AM, Hughes CC 2002 Notch activation during endothelial cell network formation in vitro targets the basic HLH transcription factor HESR-1 and downregulates VEGFR-2/KDR expression. Microvasc Res 64:372–383 Yancopoulos GD, Davis S, Gale NW, Rudge JS, Wiegand SJ, Holash J 2000 Vascular-specific growth factors and blood vessel formation. Nature 407:242–248

DLL4 AND ANGIOGENESIS

121

DISCUSSION Augustin: Reduced VEGF levels would lead to vessel normalization, increased intercapillary distance and are believed to provide better access for chemotherapeutic drugs. As you showed, increased VEGF would lead to hypervascularization and a chaotropic network. The straightforward experiment would be to overexpress VEGF. This has been done: it leads to a chaotropic network, but it always induces a pro-tumorigenic effect. Yancopoulos: I agree. We have done experiments in which we have overexpressed VEGF. The effects on tumour growth are rather modest. Among the most markedly induced gene is Dll4. This system is trying to keep itself in check, so it is probably not as bad as when you ablate Dll4, because you are up-regulating the DLR4 system precisely in that setting. Weinstein: Do you see any side effects? One of the concerns about VEGF has been that if it is reduced too much there will be effects on the normal vasculature. There are some hints coming out that this is the case. If you are reducing VEGF even further, perhaps you are going to go too far. Yancopoulos: This is exactly what we were worried about. We had done our own studies and collaborated with Donald MacDonald, using the VEGF Trap. VEGF Trap treatment in the first few weeks of life has dramatic detrimental effects, however, in adult animals the effects are more modest in certain tissues. But we can see profound regression of the thyroid vasculature in adult mice, for example, so there are vasculatures that regress. We can pick up some of these changes by looking at sensitive gene responses that mark hypoxia in these tissues. What we feared was that in tissues where it appears that VEGF is still playing a role, we might see aberrant sprouting. This could cause some problems. So we looked in these tissues with the Dll4 blocker and didn’t see any excess of sprouting or changes in gene markers. We are looking for side effects, but so far the pharmacological blockers have been well tolerated. We are hopeful that if the system is relatively static, we are not going to see over-exuberant effects. Kitajewski: At least for the DLR4 Trap, you could be causing competitive inhibition of other Notch ligands, if in theory it is binding to the Notch receptor. There are a variety of tumour settings where Jagged 1 or Notch are over-expressed. There could be multiple targets. This wouldn’t be true of a neutralizing antibody against Dll4 approach. You might block Dll4 interactions without affecting interactions via other Delta-likes of Jaggeds. Yancopoulos: We haven’t done as extensive work with the antibody, but it looks like it is showing a comparable phenotype. Dll4 is by far the most dramatically induced Notch ligand member in the tumour milieu. Blocking Notch will have all sorts of untoward effects. Our hope is that by coming up with a Dll4-specific blocker we will get maximal efficacy while limiting the inevitable side effects. The

122

NOGUERA-TROISE ET AL

good news is that studies with Dll4 blocking antibody and genetic studies suggest that Dll4 is the right target. Kitajewski: Do you see changes in endothelial proliferation after you block Dll4? It looks like there are more endothelial cells in both the yolk sac and retinal vasculature. In vitro, Notch activation can block endothelial proliferation. Are you seeing some effects on proliferation in addition to the sprouting? Yancopoulos: It would make sense, because there is an increase in total vascular mass. We haven’t looked specifically for proliferation. When we create hypoxia we are also up-regulating VEGF. There will be a complication because of this. In the time frames we would need in order to see proliferation, you might start getting confounded by the hypoxia and up-regulation of the VEGF. It could get complicated. Kitajewski: You might want to look for proliferation early on. Are you stimulating endothelial proliferation in addition to sprouting? I realise the tip cell may not be proliferating, but it does seem that you are ending up with a lot more endothelial cells. Ruhrberg: There is a precedent for a situation in which too much vessel sprouting leads to smaller tumours: Grunstein et al (2000) used different VEGF isoforms to induce tumour growth. They found that all isoforms promoted tumour growth, but VEGF188 produced too many vessel segments and they were too thin, and the tumours were relatively smaller, compared to tumours grown in the presence of VEGF164. When we looked at the brain vasculature of VEGF188 mice, we discovered vessels with too many sprouts, and we saw poor connectivity of vessel segments, i.e. a poorly formed vessel network. Epstein: Is the effect of Dll4 comparable, better or worse than that of the VEGF inhibitors? Yancopoulos: The results broadly look similar. Our hope from the therapeutic perspective is that the biological agents will be more specific and result in fewer side effects. I don’t think we want to block Notch, but rather Dll4. You won’t get this level of specificity without the biological agents. Epstein: It doesn’t seem to me a very attractive approach for treating tumours. It is interesting that you create a poorly developed vasculature, but I’m not sure it will affect metastasis or survival. There are small areas that will be perfused in the tumour. There will be areas of hypoxia and cell death in the areas that are well fed, but I don’t think this will affect survival. Weinstein: It seems to me you are trying to fit these Dll4 results into the VEGF box, but fundamentally it is doing something completely different. It seems that it has more to do with remodelling and creating a functional vasculature, rather than regulating the sprouting. Some of the effects seen in terms of the sprouting may be a secondary effect of a poorly functioning vasculature. Yancopoulos: I think the concept that the overexuberant sprouting and branching are largely driven by VEGF, and the fact that Dll4 is normally a negative

DLL4 AND ANGIOGENESIS

123

regulator of this, is consistent with the fact that at least in the eye model, when VEGF Trap is given in the Dll4 knockout, you almost immediately restrain all the branching. It is not an independent effect. The fact that it is all VEGF driven is consistent with the fact that VEGF is driving it to a certain degree and if Dll14 is removed it drives it more, and then if you give a VEGF blocker you can completely ablate it. Weinstein: I think the idea is more that what you are doing with this DLR4 is taking this network of vessels and making it not work. You are preventing it from remodelling properly to form a functional network. Gerhardt: We have been looking at exactly this aspect. It is important to understand how VEGF and Dll4 work together. The short term inhibition experiments show that if Notch signalling is inhibited with γ -secretase inhibitor for 3 h in the animal, there is an immediate response of new fi lopodia being formed specifically in the area that is exposed to VEGF. These two pathways definitely interact. If at the same time we do VEGF in situ hybridization and look for how these things match, we find the response with new induction of fi lopodia after rapid inhibition of Notch in the area where VEGF is expressed. When the vessel plexus changes over time VEGF starts to be expressed further back in the plexus consistent with the idea that it is not functional if you overdo it in this way. This fits nicely with the tumour studies. VEGF gradients are also interesting. Christiana and I have looked at the number of stalk cells, which seem to be in excess if the VEGF gradients are disrupted. The balance of the migration of the tip cell and the proliferation of the stalk cell is what determines how the initial sprout looks. It will be interesting to try to understand how VEGF gradients and Dll4 signalling interact. The VEGF188 findings are already pointing in one direction. It could be that if VEGF gradients are disrupted, you get excessive numbers of stalk cells and reduced migration of the tip cell. If Notch signalling is inhibited, part of this might be corrected by the generation of more tip cells and sprouts. Drake: If you go back to the knockout embryos and look at why they die, it is because the blood vessels hyperfuse. All of these (genes) are suppressors of the VEGF signalling pathway. You are creating the same scenario in the tumour. The reason I think this might work is because this is what kills the knockout embryos. Yancopoulos: If you think of the embryo as a growing tumour, you are killing it. We must realize that though it looks clear that some of these agents can work on their own, more likely they will be tried in combination. If you can cripple the system and create some hypoxia, there may be opportunities for combination effects. Lammert: What I really like about this blocker is that you can uncouple angiogenesis from lumen formation. You get most of the processes (sprouting, tip cell formation, stalk cell proliferation), but you don’t get a lumen. Interrupting the Notch/Delta signalling leads to over-proliferation and over-production of a

124

NOGUERA-TROISE ET AL

vascular network but without a lumen. What is the balance between proliferation and lumen formation? Yancopoulos: In the Dll4 knockouts, at least some major vessels have lumenal defects. Owens: I think we need to keep in mind that tumour vessels are usually very dysfunctional in terms of their differentiation/maturation. Indeed, among the characteristic features the pathologist uses for assessing tumours is the fact that they have defective vascular networks and circulation, as well as poor pericyte/ SMC investment of tumour blood vessels. These defects tend to be more severe with highly metastatic tumours which have an abundance of ‘giant capillaries’ that lack a SMC-pericyte coat. The systems everyone here studies didn’t evolve to grow ‘good’ tumour vessels. Is it a good or bad thing that the tumour develops a poor circulation? With the Dll4 interruption, now you have an even poorer mural cell investment and an even greater disruption of vascular patterning. An important question is whether Dll4 interruption was associated with increased rates of tumour cell shedding into the circulation. The concern is that by further disrupting vessel maturation you might actually increase tumor cell shedding and possible metastasis. In addition, although controversial, Rakesh Jain and co-workers (Jain 2005) have argued that a concern with combining anti-angiogenic therapies with chemotherapies is that you may exacerbate problems with poor perfusion of tumors and actually decrease the efficacy of conventional chemotherapeutic agents. Yancopoulos: I don’t want to over-sell this, but I can turn that argument around. Because it is a fringe vascular network, it is probably more susceptible than even the embryonic network. It is on the cusp of being functional anyway. There isn’t much evidence in favour of the argument about chemotherapeutic delivery. Shovlin: No one has mentioned the issue that reducing tumour bulk by reducing its vasculature may have knock-on effects as to how susceptible that tumour will be to chemotherapeutic agents used, by reducing the potential for somatic mutations in the tumours. Would there not be a role for getting antiangiogenic strategies in earlier, when you have the immediate bulk reduction time, in order to have a smaller cohort of tumour cells that could mutate in order to bypass the mechanisms needed by chemotherapeutic agents. Yancopoulos: There are a number of potential opportunities like this. Other than VEGF-based therapies there aren’t any other angiogenesis-based therapies that have convincing effects on tumour growth. It is important to find a second pathway. Paradoxically, this second pathway seems to be feeding into the VEGF pathway. Ye: How long does the Dll4 antagonist effect last in the tumour model? Yancopoulos: We haven’t done experiments over the long-term as we have with VEGF. We will be trying to do this soon.

DLL4 AND ANGIOGENESIS

125

References Grunstein J, Masbad JJ, Hickey R, Giordano F, Johnson RS 2000 Isoforms of vascular endothelial growth factor act in a coordinate fashion to recruit and expand tumor vasculature. Mol Cell Biol 20:7282–7291 Jain RK 2005 Normalization of tumor vasculature: an emerging concept in antiangiogenic therapy. Science 307:58–62

HIF in vascular development and tumour angiogenesis Georg Breier, Alexander H. Licht*, Anke Nicolaus, Anne Klotzsche, Ben Wielockx and Zuzana Kirsnerova Division of Endothelial Cell Biolog y, Department of Patholog y, University of Technolog y, Dresden, and *German Cancer Research Centre, Heidelberg, Germany

Abstract. Hypoxia stimulates angiogenesis through the up-regulation of vascular endothelial growth factor and other angiogenic cytokines. Members of the hypoxiainducible factor (HIF) family of transcription factors play a central role in the cellular hypoxia response. To address the function of HIF signalling in physiological and pathological angiogenesis, we used a dominant-negative approach that interferes with the function of both HIF-1 and HIF-2. The expression of a dominant-negative HIF mutant in endothelial cells inhibited endothelial sprouting and disrupted cardiovascular development in mouse embryos, demonstrating that endothelial HIF function is essential for embryogenesis. However, the inhibition of HIF activity in tumour vessels accelerated the growth of experimental fibrosarcoma and osteosarcoma. The over-expression of prolyl hydroxylase domain protein 2 (PHD2), an enzyme that negatively regulates HIF stability, strongly reduced growth of LM8 osteosarcoma cells in vivo. Our results are in line with the complexity of HIF function and indicate that HIF inhibition might not be an ideal anti-tumour strategy. 2007 Vascular development. Wiley, Chichester (Novartis Foundation Symposium 283) p 126–138

All higher organisms possess mechanisms to maintain oxygen homeostasis, which is essential for life. Low oxygen levels, or hypoxia, cause a variety of compensatory reactions at the systemic, local and cellular level (for review, see Pugh & Ratcliffe 2003). For example, hypoxia stimulates erythropoiesis through the induction of the hormone erythropoietin, and induces angiogenesis by up-regulating angiogenic factors such as vascular endothelial growth factor (VEGF). Under hypoxia, cells switch from oxidative phosphorylation to glycolysis, and up-regulate the expression of glucose transporters. Hypoxia also has profound consequences on cell survival: whereas mild hypoxia stimulates signalling pathways that promote cell survival, prolonged periods of hypoxia induce apoptosis. The cellular response to hypoxia is mediated primarily by transcriptional regulators called hypoxia-inducible factors (HIF), which stimulate the expression of a large number of genes (for review, see Pugh & Ratcliffe 2003). Currently, three 126

HIF

127

members of the HIF family are known: HIF-1, HIF-2 and HIF-3. HIF-1 is thought to be the prime regulator of the hypoxia response and is expressed ubiquitously. In contrast, HIF-2 and HIF-3 have a spatially restricted expression pattern, with HIF-2 being expressed predominantly in endothelial cells (Ema et al 1997, Flamme et al 1997, Tian et al 1997). All three HIF family members have a similar structure and are regulated by similar mechanisms. They are heterodimeric proteins consisting of an oxygen-sensitive α subunit and a constitutive β subunit (Fig. 1). Both the α and β subunits possess conserved domains involved in DNA binding and heterodimerization: the basic helix loop helix (bHLH) domain and the Per-Arnt-Sim (PAS) domain. The HIF- α subunit possesses in addition two transactivation domains (TAD) which interact with transcriptional co-activators, as well as two oxygen-dependent degradation domains (ODDD) that are involved in the rapid degradation of the α subunit under hypoxia. HIF activity is regulated by post-translational modification of the HIF- α subunit (for review, see Stolze et al 2006). Under normoxic conditions, HIF- α is hydroxylated on two proline residues located in the ODDD, leading to its rapid degradation. In addition, hydroxylation on an asparagine residue located in the C-terminal TAD of HIF- α prevents its interaction with transcriptional co-activators and thus negatively regulates HIF transcriptional activity. HIF hydroxylation on proline is achieved by a group of enzymes called HIF prolyl hydroxylases (or prolyl hydroxylase domain proteins, PHD) whereas hydroxylation on asparagine is catalysed by a HIF asparaginyl hydroxylase called Factor Inhibiting HIF (FIH). HIF hydroxylases require molecular oxygen, 2-oxoglutarate and Fe2+ for their activity and are thus considered as cellular oxygen sensors. In normoxic cells, PHDs are active and

FIG. 1. Structure of hypoxia-inducible factors (HIF). Members of the HIF family are heterodimeric transcriptional regulators consisting of an oxygen sensitive α subunit and a constitutive β subunit. Details are explained in the text.

128

BREIER ET AL

hydroxylate the HIF- α subunit, which is then bound by the von Hippel-Lindau tumour suppressor protein, the recognition component of the E3 ubiquitin ligase complex, and rapidly degraded by the proteasome. In hypoxic cells, PHDs are inactive, the HIF- α subunit is not hydroxylated. Consequently, HIF accumulates in the nucleus and transactivates HIF target genes following binding to a specific DNA sequence called the hypoxia response element (HRE). It has long been known that hypoxia is a potent stimulator of angiogenesis. The discoveries that VEGF is expressed strongly by hypoxic cells in tumours (Shweiki et al 1992, Plate et al 1992, Vajkoczy et al 2002), and that VEGF up-regulation is mediated by HIF via binding to the HRE (Ikeda et al 1995, Damert et al 1997) provided a molecular explanation for this observation. HIF is therefore thought to promote tumour progression by triggering the angiogenic switch (Pouyssegur et al 2006). Consistently, HIF over-expression appears to correlate with poor prognosis in certain malignancies. On the other hand, it is important to note that hypoxia also has a profound impact on cell survival and apoptosis. HIF is involved in apoptotic pathways, yet hypoxia also selects for cells that have lost the apoptotic potential. Gene targeting experiments in mice have revealed that HIF-1 is essential for embryonic development. HIF knockout mice develop vascular defects, which was originally explained by deregulated VEGF expression, yet the embryos die of mesenchymal cell death (Ryan et al 1998, Kotch et al 1999). HIF-2 expression in endothelial cells was thought to reflect an important function in endothelial cells (Flamme et al 1997). This hypothesis is supported by the observation that HIF-2 regulates the expression of the angiopoietin receptor Tie2. However, most gene targeting experiments have failed to demonstrate a requirement for HIF-2 in vascular development, presumably because HIF-1 can partially compensate for the loss of HIF-2 activity. HIF-2 α knockout mice were reported to display highly variable phenotypes, including metabolic abnormalities, multi-organ pathologies and impaired fetal lung maturation (Compernolle et al 2002, Scortegagna et al 2003), and only one group reported vascular remodelling defects of the embryonic vasculature (Peng et al 2000).

Results and discussion HIF regulates VEGF receptor expression We have previously studied the role of HIF in the regulation of the main signalling VEGF receptor, VEGFR2 (also known as Flk1 or KDR). Using reporter gene studies, we have shown that the mouse Flk1 promoter is strongly stimulated by HIF-2, and to a lesser extent also by HIF-1 (Kappel et al 1999). HIF-2 co-operated with Ets-1 on the Flk1 promoter to stimulate Flk1 transcription. This interaction

HIF

129

is relevant for Flk1 promoter activity in vivo because the mutation of the HIF/Ets binding motifs in a reporter gene construct that normally mediates a uniform endothelial cell-specific expression in transgenic mouse embryos, leads to a complete loss of promoter activity in the developing vasculature (Elvert et al 2003). The apparent discrepancy between these findings and the results of the HIF-2 knockout experiments might be explained by functional redundancy within the HIF family. Therefore we used a dominant-negative (dn) transcription factor mutant that inhibited both HIF-1 and HIF-2. In this dnHIF mutant, the transactivation domains were deleted whereas the bHLH and PAS domains were retained. This mutant can still dimerize with ARNT but lacks the transactivation function, and presumably acts by competing with endogenous HIF α subunits. HIF activity in endothelial cells is essential for cardiovascular development The dnHIF mutant was fused to regulatory sequences of the mouse Flk1 gene in order to direct its expression specifically to the endothelium of developing mouse embryos. The resulting mouse embryos developed severe cardiovascular malformations and died around E12.5 (Licht et al 2006). Specifically, vascular remodelling, angiogenic sprouting, and heart looping were defective, and vessels of the perineural vascular plexus in the head region failed to undergo remodelling (Fig. 2) and to extend sprouts into the neuroectoderm. To confirm that the defective

FIG. 2. Inhibition of HIF activity in embryonic endothelial cells disrupts cardiovascular development. A dominant-negative (dn) HIF mutant was expressed from gene regulatory elements of the mouse Flk1 gene to direct its expression to the developing vasculature. The embryonic vasculature of 10.5-day mouse embryos was visualized by immunohistochemical staining for the endothelial marker PECAM-1. WT, wild-type embryo.

130

BREIER ET AL

sprouting was due to HIF inhibition in endothelial cells (rather than caused by impaired blood circulation), we performed an in vitro sprouting assay. Mouse endothelial cells (MS1) were embedded in collagen, and endothelial sprout formation was monitored. The sprouting activity of MS1 cells that were transduced with retrovirus encoding the dnHIF mutant was significantly lower than of control MS1 cells (Fig. 3), demonstrating that HIF activity is required for angiogenic sprouting of endothelial cells. Heart development in dnHIF mice was also defective. The fetal heart tube failed to loop, and heart trabeculation was defective. Interestingly, these myocardial defects were caused by HIF deficiency in the endocardium, rather than in myocardiac cells, indicating paracrine regulatory interactions. HIF regulates essential endothelial receptor tyrosine kinases The cardiovascular defects observed in dnHIF transgenic mouse embryos are reminiscent of the defects observed in mouse embryos deficient for the receptor tyrosine kinase Tie2. Consistently, Tie2 mRNA and protein levels were considerably down-regulated in dnHIF transgenic embryos (Licht et al 2006). Similarly,

FIG. 3. Inhibition of HIF signalling in endothelial cells results in impaired endothelial sprouting in vitro. Mouse MS1 endothelial cells were transduced with retrovirus encoding a dnHIF mutant, and the length of endothelial sprouts in three-dimensional collagen gels was determined.

HIF

131

VEGFR1 and VEGFR2 transcript levels were also reduced whereas expression of other endothelial receptors (such as VE-cadherin or PECAM-1) were not altered. Taken together, these results demonstrate that HIF activity in endothelial cells has an important function during embryonic cardiovascular development, by regulating the expression of essential endothelial receptor tyrosine kinases, in particular, Tie2. Moreover, HIF members might have partially redundant functions in endothelial cells. It remains unclear however, whether HIF signalling in the embryo is related to hypoxia or to other stimuli. HIF—a target for anti-angiogenic tumour therapy? HIFs regulate various angiogenesis related genes and have been proposed as targets for pro- or anti-angiogenic therapies (Pouyssegur et al 2006). In order to study HIF function in tumour angiogenesis and progression, we used two different approaches that allowed us to dissect HIF function in endothelial cells versus tumor cells. Previous work by others has shown that the conditional inactivation of HIF-1 in endothelial cells of mice affects tumor angiogenesis and growth, by disrupting a VEGF autocrine loop (Tang et al 2004). However, the effects of HIF1 inactivation were relatively modest in comparison to those obtained after VEGF inhibition, which might indicate that HIF-2 expression in tumour endothelial cells contributes to tumour angiogenesis. We therefore inhibited HIF (-1 and -2) signalling in tumour endothelial cells by retrovirus-mediated gene transfer of dnHIF. As a tumour model, we chose a mouse BFS-1 fibrosarcoma whose growth is completely inhibited by retrovirus-mediated gene transfer of dnVEGFR mutants (Heidenreich et al 2004). This growth inhibition is mediated primarily by inhibition of signalling in endothelial cells (and possibly also other host-derived cells), which are transduced with ecotropic retrovirus whereas tumour cells are not. Unexpectedly, dnHIF gene transfer accelerated the growth of BFS-1 tumours in C57BL/6 mice, whereas (as shown previously) dnVEGFR2 completely inhibited tumour growth. The same results were obtained in a second tumour model (an LM-8 osteosarcoma). Thus, inhibition of HIF activity in tumour endothelial cells appears to accelerate tumour growth. The cause of this unexpected result is under investigation. Next, we looked at the consequences of HIF inhibition in tumour cells. BFS-1 fibrosarcoma cells were stably transfected with an expression vector encoding dnHIF, and resulting clones were injected subcutaneously into C57BL/6 mice. Some cell clones analysed grew faster than control cells, although blood vessel density was decreased. A similar observation was made previously by Acker and colleagues in a glioma model (Acker et al 2005). According to these authors, this result was explained by a reduced apoptosis rate of tumour cells that over-express the dnHIF mutant, in line with the known function of HIF in apoptosis.

132

BREIER ET AL

Finally, we tested the effect of PHD2 over-expression on the growth of experimental LM8 osteosarcoma. PHD2 is thought to be the key enzyme involved in HIF regulation (Berra et al 2003). Tumour cells that over-expressed PHD2 grew significantly slower than control cells. Using Affymetrix expression profi ling, we found that several interferon-related or -induced genes were up-regulated in LM8 cells by PHD2 overexpression. Consistently, we observed that numerous inflammatory cells invaded the tumours. The role of these inflammatory cells is under investigation. Conclusions Taken together, our results show that HIF activity in endothelial cells is required for cardiovascular development. However, the inhibition of HIF activity in tumour endothelial cells (and other host-derived cells) does not inhibit tumour growth. The partially contradictory results obtained from HIF inhibition and PHD2 overexpression experiments indicate that HIF and HIF hydroxylase pathways might partially diverge. Future research will clarify this issue, as well as several other outstanding questions. With regard to therapy, it is emerging that HIF inhibition is not an ideal anti-tumour strategy. Whether PHD activation will be more successful in this respect remains to be determined. References Acker T, Diez-Juan A, Aragones J et al 2005 Genetic evidence for a tumor suppressor role of HIF-2alpha. Cancer Cell 8:131–141 Berra E, Benizri E, Ginouves A, Volmat V, Roux D, Pouyssegur J 2003 HIF prolyl-hydroxylase 2 is the key oxygen sensor setting low steady-state levels of HIF-1alpha in normoxia. EMBO J 22:4082–90 Compernolle V, Brusselmans K, Acker T et al 2002 Loss of HIF-2alpha and inhibition of VEGF impair fetal lung maturation, whereas treatment with VEGF prevents fatal respiratory distress in premature mice. Nat Med 8:702–710 Damert A, Machein M, Breier G et al 1997 Up-regulation of vascular endothelial growth factor expression in a rat glioma is conferred by two distinct hypoxia-driven mechanisms. Cancer Res 57:3860–3864 Elvert G, Kappel A, Heidenreich R et al 2003 Cooperative interaction of hypoxia-inducible factor-2a and Ets-1 in the transcriptional activation of Flk-1. J Biol Chem 278:7520–7530 Ema M, Taya S, Yokotani N, Sogawa K, Matsuda Y, Fujii-Kuriyama Y 1997 A novel bHLH-PAS factor with close sequence similarity to hypoxia-inducible factor 1alpha regulates the VEGF expression and is potentially involved in lung and vascular development. Proc Natl Acad Sci USA 94:4273–4278 Flamme I, Frohlich T, von Reutern M, Kappel A, Damert A, Risau W 1997 HRF, a putative basic helix-loop-helix-PAS-domain transcription factor is closely related to hypoxiainducible factor-1 alpha and developmentally expressed in blood vessels. Mech Dev 63: 51–60 Heidenreich R, Machein M, Nicolaus A et al 2004 Inhibition of solid tumor growth by gene transfer of dominant-negative VEGF receptor-1 mutants. Int J Cancer 111:348–357

HIF

133

Ikeda E, Achen MG, Breier G, Risau W 1995 Hypoxia-induced transcriptional activation and increased mRNA stability of vascular endothelial growth factor in C6 glioma cells. J Biol Chem 270:19761–19766 Kappel A, Rönicke V, Damert A, Flamme I, Risau W, Breier G 1999 Identification of VEGF receptor-2 (Flk-1) promoter/enhancer sequences sufficient for angioblast and endothelial cell-specific transcription in transgenic mice. Blood 93:4284–4292 Kotch LE, Iyer NV, Laughner E, Semenza GL 1999 Defective vascularization of HIF1alpha-null embryos is not associated with VEGF deficiency but with mesenchymal cell death. Dev Biol 209:254–267 Licht AH, Muller-Holtkamp F, Flamme I, Breier G 2006 Inhibition of hypoxia-inducible factor activity in endothelial cells disrupts embryonic cardiovascular development. Blood 107: 584–590 Peng J, Zhang L, Drysdale L, Fong GH 2000 The transcription factor EPAS-1/hypoxiainducible factor 2alpha plays an important role in vascular remodeling. Proc Natl Acad Sci USA 97:8386–8391 Plate KH, Breier G, Weich HA, Risau W 1992 Vascular endothelial growth factor is a potential tumour angiogenesis factor in human gliomas in vivo. Nature 359:845–848 Pouyssegur J, Dayan F, Mazure NM 2006 Hypoxia signalling in cancer and approaches to enforce tumour regression. Nature 441:437–443 Pugh CW, Ratcliffe PW 2003 Regulation of angiogenesis by hypoxia: role of the HIF system. Nat Med 9:677–684 Ryan HE, Lo J, Johnson RS 1998 HIF-1 alpha is required for solid tumor formation and embryonic vascularization. EMBO J 17:3005–3015 Scortegagna M, Ding K, Oktay Y et al 2003 Multiple organ pathology, metabolic abnormalities and impaired homeostasis of reactive oxygen species in Epas1–/– mice. Nat Genet 35: 331–340 Shweiki D, Itin A, Soffer D, Keshet E 1992 Vascular endothelial growth factor induced by hypoxia may mediate hypoxia-initiated angiogenesis. Nature 359:843–845 Stolze IP, Mole DR, Ratcliffe PJ 2006 Regulation of HIF: prolyl hydroxylases. In: Signalling pathways in acute oxygen sensing. Wiley, Chichester (Novartis Found Symp 272) p 15–36 Tang N, Wang L, Esko J, et al 2004 Loss of HIF-1alpha in endothelial cells disrupts a hypoxiadriven VEGF autocrine loop necessary for tumorigenesis. Cancer Cell 6:485–495 Tian H, McKnight SL, Russell DW 1997 Endothelial PAS domain protein 1 (EPAS1), a transcription factor selectively expressed in endothelial cells. Genes Dev 11:72–82 Vajkoczy P, Farhadi M, Gaumann A et al 2002 Microtumor growth initiates angiogenic sprouting with simultaneous expression of VEGF, VEGF-receptor-2, and angiopoietin-2. J Clin Invest 109:777–785

DISCUSSION Lammert: This is now the second paper that we have heard at this meeting which suggests that more blood vessels result in less tumour growth. Are there more studies on this? What are the major criteria for discriminating a tumour-promoting vessel versus a non-promoting vessel? Yancopoulos: I was struck by this. It may be an analogous situation, with fewer vessels being more functional. Breier: The experimental BFS-1 fibrosarcoma that we use is highly vascularized; the total vessel area in this tumour is typically more than 10%. I assume that the

134

BREIER ET AL

tumour does not need such a high vessel density, and therefore a reduction in vessel density would not necessarily result in decreased tumour growth. Also, we almost never observe necrosis. This indicates that in this fibrosarcoma model, the vessels perform pretty well. Augustin: We recently published the dissociation of angiogenesis and tumorigenesis in tumours that over-express either follistatin or activin (Krneta et al 2006). Follistatin tumours have more blood vessels but they grow slower. The angiogenesis inhibitor activin results in tumours that grow much faster but have fewer blood vessels. Our explanation for this is that a tumour can grow as long as it is nourished by a minimum microvessel density. We have done tumour experiments in combination with a meta-analysis of the more than 1000 published microvessel counting papers in human tumours. On the basis of these studies we have developed a formula to calculate intercapillary distance from microvessel density. Tumours have different metabolic needs, but we think that if a tumour has 50 microvessels/mm2 , that is enough to sustain its growth. Most human tumours have 80–120 blood vessels per mm2 . This is an anatomical parameter that doesn’t take into account if tumor microvessels are perfused or not, but it suggests that most human tumours have more blood vessels than they need. Yancopoulos: That is interesting and seems to relate to both the stories we have been talking about. The dogma about the relationship between tumours and vessels, and that more is always better, doesn’t seem to be right. This relates to a point that was highlighted by some of the data you showed. This is another example of a situation where the endothelium expresses dominant-negative HIF, there is a profound effect on cardiovascular development that results in lethality. But the same agent is not enough to cripple the tumour growth. The tumour vessel architecture is more abnormal. We have found a lot of pathways that disrupt developmental vascular development and lead to embryonic lethality. It is easier to cripple these developmental processes, but the tumour, which may be more abnormal, can tolerate more types of hit. There are only a few ways to perturb tumour growth. Betsholtz: When comparing blood vessel formation in embryos and tumours we should keep in mind that many of the mouse mutants with defective blood vessel development also have heart defects, which in turn could lead to further secondary perturbation of vascular development. It is often difficult to distinguish primary from secondary effects. Wilting: In chick chorioallantoic membrane it makes a difference whether VEGF is applied as a pure protein or if adenoviruses are used, which produce VEGF. How much of the effects could be inflammatory effects caused by the adenoviruses? Breier: In the tumour experiments I was referring to, we did not apply adenovirus, but transplanted tumour cells that were stably transfected with an expression vector for dominant-negative HIF.

HIF

135

Gerhardt: I’d like to add another component to the issue of whether there is a normalization of the vessel, and whether this contributes to tumour growth. When we discussed this previously, I was relating it to the normalization of the tumour vessel in terms of pericyte coverage. This was in the context of PDGFB being produced by the endothelium to recruit the pericytes. If you interfere with this process, particularly if you do it in a way that disturbs the association of the vessel with the pericyte, this will affect the functionality of the vessel and consequently tumour growth. If you really turn on the sprouting process and increase the density of the vessels, could it be that the normalization of pericyte coverage of these vessels is not really working? Is there a contribution of the pericyte coverage here? Once we normalized tumour blood vessels, incorporating more pericytes, the tumours grew dramatically faster (Xian et al 2006). Have you looked at this? Breier: Not yet. Gerhardt: Donald McDonald presented data at the Sixth European School of Haematology Conference on Angiogenesis, Cannes, France 13–16 May 2006, showing that when he inhibits the VEGF pathway, he sees that pericyte coverage on vessels is also improved. One explanation could be that excessive endothelial sprouting is too quick to allow for the consecutive maturation. Breier: We are addressing this question but we don’t have data yet. Dejana: Isn’t it possible that you are selecting the most resistant or insensitive tumour cells? Therefore they grow more, because they undergo selection, independently of vascular issues. Owens: Donald McDonald presented these data at a recent meeting several of us attended. The pericyte coverage improved but this was because of loss of the VEGF-dependent vessels which had poor pericyte coverage, not by improving the coverage of vessels that were already there. That is, VEGF inhibition was not actually promoting pericyte coverage. These observations fit with why the VEGF blocking works so well: i.e. there is a persistence of VEGF dependence in the tumour vessels. It will be interesting to study the role of HIF in this process. I’d also like to follow up on the comment regarding involvement of CD45 + inflammatory cells. There is evidence from Keshet’s group (Grunewald et al 2006) indicating that bone marrow derived cells play a key paracrine role in regulation of tumour angiogenesis. Moreover, we have seen evidence for this in a model of systemic hypoxia (O’Neill et al 2005). Have people done bone marrow reconstitutions with HIF knockout animals, and then looked at how this affects tumour angiogenesis? Breier: I’m not aware that they have. Shibuya: In the endothelia of HIF-2 knockout mice, you see the effects after vasculogenesis. This means that the expression of VEGFR2 (Flk1) in vasculogenesis is not dependent on endothelial HIF-2. Is it so?

136

BREIER ET AL

Breier: It is important to note that in our mouse model, the Flk1-dnHIF transgene is expressed after the primitive vessels have formed. Therefore, it cannot be expected that the expression of the dominant-negative HIF mutant would affect vasculogenesis. It is not possible to address the question whether HIF might also have a function in earlier stages of vascular development with our transgenic model because of the temporal activity of the Flk1 promoter sequences that drives expression of the transgene. Shibuya: In vasculogenesis Flk1 is already expressed. The dominant negative HIF-2 is also expressed. But this didn’t block vasculogenesis, which means that HIF-2 in endothelial cells isn’t essential for the early stage Flk1 expression. Later it begins to be required. Is this correct? Breier: We observed that the onset of transgene expression is around mid-gestation. Thus, in contrast to the endogenous Flk1 gene, the Flk1 promoter that we used is not active in haemangioblasts, but only later, in defi nitive endothelial cells. The likely cause for this difference is that certain regulatory sequences required for Flk1 expression in the haemangioblast are missing in our Flk1 promoter construct. Lammert: It is hard to imagine that hypoxia plays a role in endothelial cells that are part of vessels. If you look at aorta formation, there is no oxygen. There, HIF might play a more active role in the endothelial cells, than let’s say if you have a sprouting angiogenesis where there is a connection between the endothelial cells and the circulatory system. Perhaps the tip cells are the most hypoxic of all the cells. Adams: Coming back to tumour vascularization, we also have to consider the normal hierarchical organization of the vascular tree. If you expand the microvascular network and blood flow is still supplied by the same artery, oxygen supply to the tumour tissue should be not increased. Drake: Along these lines, we have isolated angioblasts and shown that they express HIF genes and VEGF mRNA. We are starting to believe that there might be an autocrine VEGF signalling pathway running from the beginning of development. Weinstein: In the fish, where we can test the role of oxygen more easily, it doesn’t seem to make much of a difference in the patterning of the major vessels in the embryo. Early on, it is relatively insensitive to oxygen. Breier: There are a few studies on mouse embryos in which hydroxyprobe has been used to detect hypoxic regions in the embryo. Based on these reports, at least certain areas in the embryo are hypoxic. How relevant is this? Drake: Do you need hypoxia? Aren’t there some papers that suggest that the HIF-1 pathway can be regulated under normoxic conditions? Betsholtz: Constitutive expression of VEGF can also occur in well oxygenated tissues, such as the kidney glomeruli.

HIF

137

Breier: I’d like to comment on HIF function with regard to VEGF regulation in mouse development. There is one knockout study that addressed the role of the hypoxia response element in the VEGF gene (Oosthuyse et al 2001). These knockout mice developed motor neuron degeneration. Augustin: Following up on the question of endothelial cell hypoxia sensing, one conceptual view I find attractive is that endothelial cells might respond with higher sensitivity to changes in oxygen than cells in the periphery. Endothelial cells should be the last cells to experience hypoxia because they are closest to the blood vessel. Yet, if they respond with a lower threshold they could be serving a ‘gatekeeper’ function to control what happens in the periphery. It would be interesting to do careful oxygen gradient titration experiments. Wilting: What is hypoxia? Chondrocytes for example do well in 5% oxygen tension, so normoxia (21% pO2 ) would be hyperoxia for these cells. In in vitro studies on hypoxia there is almost no oxygen present in the incubators. In these experiments there is 0.5 or 1% oxygen. From a teleological viewpoint I would suggest that you have to start blood vessel development in the embryo before you run into oxygen deficiencies. Early vessel development would have to be relatively independent from oxygen levels. Shovlin: I want to highlight the fact that endothelial cells are not all equivalent. Pulmonary microvascular endothelial cells are exposed to higher levels of oxygen than cells anywhere else in the body, because these are directly adjacent to the alveoli. It might be worth comparing these with endothelial cells from elsewhere when doing these experiments. Gerhardt: You mentioned VEGF delta mice from Peter Carmeliet (Leuven) which lack the hypoxia response element. If we go to our favourite model, the retina, where there is a graded expression of VEGF in the astrocytes, we would predict that if we were to do hypoxic probe staining the regions of hypoxia in the retina coincide exactly with the expression of VEGF. Marcus Fruttiger (UCL, Ophthalmology) has done a nice study on this (Claxton & Fruttiger 2003, 2005) showing complete overlap. One would predict that the VEGF delta mice would have this pattern disrupted and it eventually would lead to malformations of the retinal vascular plexus. In fact, the primary retinal vascular plexus develops completely normally. Peter Campochiaro has looked at this recently (Vinores et al 2006) This first pattern is not dependent on the HRE. Whether this means it is completely independent of HIF is a separate question. Although there is an overlap between hypoxia and VEGF expression, there seem to be additional things controlling the primary patterning here. If the regulation of VEGF is messed up, there are profound effects, but it isn’t completely a hypoxia-dependent response. Betsholtz: Isn’t the simplest explanation that HIF binds to other sites in addition to the HRE?

138

BREIER ET AL

Gerhardt: Yes, but if you look in the same animals, the downward sprouting into the deeper retinal layers is severely affected, indicating that there may be different requirements in different areas. The total levels of VEGF are different in these two processes (primary plexus and deeper plexus formation), and it could be that the sensitivity is shifting. Betsholtz: In spite of this overlap between the region of hypoxia and the region of VEGF expression, it is only the astrocytes in the developing retina that express VEGF. All the other cells which exist in the same hypoxic area don’t express VEGF. Either there is intracellular hypoxia specifically in certain cell types, or there are other signalling mechanisms that help control VEGF expression in astrocytes. References Claxton S, Fruttiger M 2003 Role of arteries in oxygen induced vaso-obliteration. Exp Eye Res 77:305–311 Claxton S, Fruttiger M 2005 Oxygen modifies artery differentiation and network morphogenesis in the retinal vasculature. Dev Dyn 233:822–828 Grunewald M, Avraham I, Dor Y et al 2006 VEGF-induced adult neovascularization: recruitment, retention, and role of accessory cells. Cell 124:175–189 Krneta J, Kroll J, Alves F et al 2006 Dissociation of angiogenesis and tumorigenesis in follistatin- and activin-expressing tumors. Cancer Res 66:5686–5695 O’Neill TJ, Wamhoff BR, Owens GK, Skalak TC 2005 Mobilization of bone marrow-derived cells enhances the angiogenic response to hypoxia without transdifferentiation into endothelial cells. Circ Res 97:1027–1035 Oosthuyse B, Moons L, Storkebaum E et al 2001 Deletion of the hypoxia-response element in the vascular endothelial growth factor promoter causes motor neuron degeneration. Nat Genet 28:131–138 Vinores SA, Xiao WH, Aslam S et al 2006 Implication of the hypoxia response element of the Vegf promoter in mouse models of retinal and choroidal neovascularization, but not retinal vascular development. J Cell Physiol 206:749–758 Xian X, Hakansson J, Stahlberg A et al 2006 Pericytes limit tumor cell metastasis. J Clin Invest 116:642–651

Imaging the developing lymphatic system using the zebrafish Karina Yaniv*, Sumio Isogai*†, Daniel Castranova*, Louis Dye‡, Jiro Hitomi† and Brant M. Weinstein*1 * Laboratory of Molecular Genetics, National Institute of Child Health and Human Development, National Institutes of Health, 6B/309, 6 Center Drive, Bethesda, MD 20892, USA, † Department of Anatomy, School of Medicine, Iwate Medical University, Morioka, 020-8505, Japan and ‡ Microscopy and Imaging Core, National Institute of Child Health and Human Development, National Institutes of Health, 49/5W-14, 500 Rockville Pike, Bethesda, MD 20892, USA

Abstract. The lymphatic system is essential for immune responses, fluid homeostasis, and fat absorption, and is involved in many pathological processes, including tumour metastasis and lymphoedema. Despite its importance, progress in understanding the origins and early development of this system has been hampered by difficulties in observing lymphatic cells in vivo and performing genetic and experimental manipulation of the lymphatic system. These difficulties stem in part from the lack of a model organism combining these features. The zebrafish is a genetically accessible vertebrate with an optically clear embryo permitting high-resolution in vivo imaging, but the existence of a lymphatic vascular system has not been previously reported in this model organism. Using a series of morphological, molecular and functional studies we have visualized and characterized lymphatic vessels in the developing zebrafish. Our studies show that the zebrafish possesses a lymphatic system that shares many of the characteristic features of lymphatic vessels found in other vertebrates. Using multiphoton time-lapse imaging we have carried out in vivo cell tracking experiments to trace the origins of lymphatic endothelial cells. Our data provide conclusive new evidence supporting a venous origin for primitive lymphatic endothelial cells. 2007 Vascular development. Wiley, Chichester (Novartis Foundation Symposium 283) p 139–151

The lymphatic system, the ‘other vascular system’, is an endothelium-lined network of blind-ended capillaries found in nearly all tissues, draining via collecting vessels into large vascular trunks that eventually empty via an evolutionarily conserved drainage point into the blood circulatory system. This system of vessels, separate and anatomically distinct from the blood vasculature, has been known since antiquity. Hippocrates first described vessels containing ‘white blood’ around 400 1

This paper was presented at the symposium by Brant Weinstein, to whom correspondence should be addressed. 139

140

YANIV ET AL

B.C., and Gasparo Aselli re-identified lymphatic vessels in the 1600s, noting the presence of ‘milky veins’ in the gut of a ‘well-fed’ dog. In the early 20th century, classical vascular anatomists such as Florence Sabin used dye injection methods to characterize the anatomy of the lymphatic system in detail. The lymphatic system has a number of important functions. It transports fluids, plasma macromolecules, and cells extravasated from blood vessels, returning them back into the blood circulation and preventing their build-up in tissues throughout the body. Defects in the lymphatic system, whether congenital (primary lymphoedema, relatively uncommon) or acquired (secondary lymphoedema, a common complication of surgery and certain parasitic infections), can result in severe, disfiguring oedema of affected tissues. The lymphatic system is also a major route for absorption of lipids from the gut (hence the ‘milky’ appearance of the vessels that led to their early identification). Lymphatics are a critical component of the immune system, transporting white blood cells and antigens from distant sites to lymphoid organs. Recent evidence also indicates that the lymphatic system is a major pathway for the dissemination of metastatic cells, making it an important new focus of efforts to develop effective cancer therapies. Despite its importance, the formation of the lymphatic system has remained relatively obscure in comparison to the blood vasculature (Oliver & Alitalo 2005). This is partly due to the intense research focus recently placed on blood vessels, but also due to a number of technical challenges in studying lymphatics. Lymphatic vessels are frequently difficult to visualize in histology or electron micrographs because unlike blood vessels they are often irregular and collapsed. When they are observed, lymphatic vessels are frequently difficult to distinguish definitively from blood vessels. Defects in the lymphatic system often occur relatively late in postnatal life and are slow in onset, making it challenging to assess the proximal aetiology of lymphatic disorders. Molecular identification of lymphatic vessels has only recently become possible. Although most molecular markers of lymphatic vessels are shared with blood vessels, in the past few years a small number of genes have been identified whose expression within lymphatic endothelium is diagnostic for these cells (although none of the genes are expressed exclusively within the lymphatic system). Some of these genes have also been shown to be required for proper formation of the lymphatic system, mostly through murine knockout studies. The origins and assembly of the lymphatic system are still not well understood. Florence Sabin proposed a century ago that the fi rst lymphatic endothelial cells (LECs) arise from embryonic veins, incorporate into primitive lymphatic sacs, and then grow and divide to populate the peripheral lymphatics (Sabin 1902). An alternative view first proposed by Huntington and McClure holds that LECs arise de novo from mesenchyme throughout the animal, forming lymphatic vessels that subsequently connect to large veins (Huntington & McClure 1910). In an effort

ZEBRAFISH LYMPHATIC DEVELOPMENT

141

to better understand the origins of LEC and the assembly of the developing lymphatic system we have turned to the zebrafish, a model system with a number of unique advantages. Zebrafish are genetically tractable and useful for large-scale mutational and small molecule screening. They have externally fertilized, optically clear embryos that are readily accessible to experimental manipulation and highresolution optical imaging of even deep organs and tissues. Although the existence of a lymphatic vascular system had not been previously reported in the zebrafish, as we describe below we have now been able to demonstrate clearly that the zebrafish possesses a bona fide lymphatic system that shares many of the morphological, molecular, and functional characteristics of the mammalian lymphatic system. Using two-photon time-lapse imaging of transgenic zebrafish, we trace the migration and lineage of individual cells incorporating into the lymphatic endothelium, showing that primitive LECs arise from embryonic veins. Confocal images of 4 day-old transgenic Tg( fli1:EGFP) y1 zebrafish (Lawson & Weinstein 2002) reveal an additional vascular tube located between the two axial blood vessels (dorsal aorta and cardinal vein) (Fig. 1). This additional vessel assembles between 3 and 5 days post-fertilization, and can also be seen in histology sections and electron micrographs of both developing larvae and adult zebrafish. In mammals the first lymphatic vessel to form is the thoracic duct, the major trunk lymphatic vessel that similarly runs between the dorsal aorta and cardinal vein. We performed additional studies to explore whether the new vessel we observed in the zebrafish trunk was in fact a bona fide lymphatic vessel and homologue of the mammalian thoracic duct.

FIG. 1. Detection of a previously uncharacterized vascular system in the zebrafish. (a,b) Confocal microangiogram of the 4 dpf zebrafish blood vascular system (a) with a magnification of the boxed region of the trunk (b) showing the dorsal aorta (upper arrow) and posterior cardinal vein (lower arrow), with no intervening vessel (asterisks). (c) Comparable confocal image of a 4 dpf Tg(fl i1:EGFP) yl zebrafish trunk, revealing the presence of a third vascular tube (asterisks) running between the dorsal aorta and posterior cardinal vein. Scale bars = 500 µm (a), and 50 µm (b,c).

142

YANIV ET AL

We began by examining whether this vessel expresses genes identified as markers of lymphatic endothelium in mammals, including the prospero-related homeobox 1 (Prox1) and neuropilin 2 (Nrp2) genes. In mice Prox1 is necessary and sufficient for differentiation of lymphatic endothelial cells (Wigle et al 2002, Wigle & Oliver 1999), while Nrp2 appears to play a role in the development of small lymphatic vessels (Yuan et al 2002). We found that, as in mammals, zebrafish prox1 is expressed in the putative zebrafish thoracic duct, in addition to its conserved expression in other tissues (Fig. 2a,b). Zebrafish Nrp2a is also expressed in this new vessel, like mammalian Nrp2. In addition to showing that this vessel expresses characteristic lymphatic markers, we were also able to demonstrate that its forma-

FIG. 2. Molecular characterization of zebrafish lymphatic vessels. (a) Box on diagram (modified from Kimmel et al 1995) showing approximate location of regions imaged in panels (b–e). (b) In situ hybridization of the zebrafish trunk with prox1 probe at 7 dpf, showing expression in the neural tube (NT) and developing thoracic duct (arrows). (c–e) Two-photon images of the trunks of control morpholino- (c), prox1 morpholino-, (d), or vegfC morpholino- (e) injected 5 dpf Tg(fl i1:EGFP) yl zebrafish. Regions within grey boxes are shown magnified below. The thoracic duct is present in controls between the dorsal aorta (DA) and posterior cardinal vein (PCV) but absent from prox1 or vegfC morpholino-injected animals (arrows). Scale bars = 100 µm.

ZEBRAFISH LYMPHATIC DEVELOPMENT

143

tion requires the function of zebrafish prox1 and veg fC, two genes that have been shown to be important for the formation of lymphatics in mammals (Cao 1998, Jeltsch 1997, Karkkainen 2004, Wigle & Oliver 1999). Injection of antisense oligonucleotides inhibiting the translation of either one of these genes resulted in selective loss of the thoracic duct without significant defects in the formation of trunk blood vessels (Fig. 2c–e). In contrast, the thoracic duct still forms in plcg1y10 mutants that fail to form arteries properly and lack all trunk circulation (Lawson et al 2003), confirming that lymphatic defects in prox1 or veg fC morpholinoinjected animals are not secondary to blood vascular dysfunction. Although our molecular analysis of the thoracic duct supports the view that this is indeed a lymphatic vessel, the functional characteristics of the lymphatic system are perhaps its most definitive identifying features. Lymphatics comprise a distinct, separate network of blind-ended vessels lacking direct connections to the blood vascular system, and do not contain red blood cells. Unlike blood vessels, lymphatics collect and drain fluids and macromolecules from interstitial spaces throughout the animal. We performed additional experiments to determine whether zebrafish lymphatics display these same characteristics (Fig. 3a). Injection of fluorescent microspheres into the blood vascular system (angiography) does not label the thoracic duct (Fig. 3b), or other lymphatic vessels (data not shown). Conversely, fluorescent microspheres injected into the thoracic duct (lymphangiography) do not enter blood vessels at this stage (Fig. 3c). These results demonstrate that blood and lymphatic vessels lack direct, open connections between one another and are distinct vascular systems (see below). Lack of red blood cells is another identifying feature of lymphatic vessels. We did not observe red blood cells in phase contrast micrographs of zebrafish lymphatic vessels (data not shown), but to examine this more comprehensively we used Tg(fli1:EGFP) yl , TG(gata1: dsRed) double transgenic zebrafish expressing DsRed in red blood cells (RBCs) (Long et al 1997) and eGFP in endothelial cells. RBCs are visible in the axial and intersegmental blood vessels of double transgenic animals, but are not detected in the thoracic duct (Fig. 3d). Even with continuous imaging of double transgenic animals for up to 24 hours no DsRed-positive RBCs could be found within the thoracic duct. The ability of lymphatics to collect and drain fluids and macromolecules from surrounding interstitial spaces is perhaps the most defi nitive functional assay for these vessels. Subcutaneously injected dyes are rapidly taken up by and concentrated in lymphatic vessels, highlighting the lymphatic vascular system (Ny et al 2005, Tilney 1971). High molecular weight rhodamine-dextran injected subcutaneously into the tail of developing zebrafish enters the lymphatic thoracic duct within minutes but is not detected within the adjacent vascular dorsal aorta even after extended incubation (Fig. 3e). The dye is highly effective at illuminating even the smallest vessels, revealing a lush and complex system of small and large vessels

144

YANIV ET AL

FIG. 3. Functional characterization of zebrafish lymphatic vessels. (a) Methods used to image blood or lymphatic vessels. The fi rst grey box on the fish drawing (modified from Kimmel et al 1995 Dev Dyn 203:253–310) shows the approximate region of the trunk imaged in panels b, c, d, e, and g. The larger grey box on the drawing shows the approximate region of the trunk imaged in panel f. (b) Angiography of a 14 dpf Tg(fl i1:EGFP) yl zebrafish (lighter grey) injected with fluorescent microspheres (darker grey blobs), labelling dorsal aorta (large arrow) and posterior cardinal vein (asterisk) but not lymphatic thoracic duct (small arrow). (c) Lymphangiography of 3 week Tg(fl i1:EGFP) yl zebrafish injected with fluorescent microspheres, labelling thoracic duct (small arrow) but not dorsal aorta (large arrow). (d) Time-averaged confocal image of a 7 dpf Tg(fl i1:EGFP) yl (green, here depicted as light grey) and TG(gata1:dsRed) (red, here depicted as darker grey) double transgenic animal, showing darker grey (red) fluorescence in the dorsal aorta (large arrow) and cardinal vein (asterisk) but not lymphatic thoracic duct (small arrow). (e–g) Confocal imaging of an 18 dpf Tg(fl i1:EGFP) yl zebrafish (green, here lighter grey) injected subcutaneously with 2 Md rhodamine-dextran (red, here darker grey). (e) Subcutaneously injected rhodamine-dextran drains into the thoracic duct (small arrow) but does not label the adjacent dorsal aorta (large arrow). (f ) Numerous rhodamine-dextran labelled vessels are visible between the blood vessels. (g) Higher magnification image of blind-ended rhodaminedextran labelled vessels. Scale bars = 20 µm (e), 50 µm (b,c,d,g), 100 µm (f ).

ZEBRAFISH LYMPHATIC DEVELOPMENT

145

in the 18 dpf zebrafish (Fig. 3f), a stage when few blood vascular capillaries are yet present. Higher magnification images reveal blind-ended vascular networks (Fig. 3g), as described for the peripheral lymphatic capillaries in other vertebrates (van der Putte 1975a,b). We also compared the anatomy of the developing lymphatics in zebrafish and other vertebrates and were able to demonstrate that zebrafish possess many of the same highly conserved anatomical features found in the developing lymphatics of other developing vertebrates (data not shown, see Yaniv et al 2006). Having established the zebrafish as an effective comparative model for studying lymphatic development, we decided to use time-lapse imaging to probe the cellular origins of this system. As noted above, Sabin proposed that lymphatic endothelial cells emerge from a subpopulation of endothelial cells within early embryonic veins, in particular the cardinal vein, to form the primitive lymphatic sacs, spreading from there to populate the entire peripheral lymphatic system (Sabin 1902). We used two-photon time-lapse imaging of Tg( fli1:EGFP) y1 transgenic animals to capture dynamic images of the sprouting and growth of the thoracic duct. Surprisingly, our images showed that initial thoracic duct sprouts appeared to emerge adjacent to the dorsal aorta (Fig. 4a, see Yaniv et al 2006 for movie), later growing and extending across the trunk to join with other segments and form a continuous thoracic duct (Fig. 4b, see Yaniv et al 2006 for movie). To determine the source of LECs populating the thoracic duct more conclusively, we performed additional two-photon time-lapse imaging using another transgenic line, Tg( fli1:nEGFP) y7, in which the Fli1 promoter drives expression of EGFP fused to a nuclear localization sequence to direct it to endothelial cell nuclei. We performed lineage tracing of LEC by collecting long (3+ day) timelapse sequences and then tracing backwards in time any LEC nuclei found in the thoracic duct at the end of the timelapse sequence. In three separate time-lapse experiments a total of sixteen 4 dpf thoracic duct LECs were traced back to their progenitors at approximately 2 dpf. In each case, the progenitors of thoracic duct LEC were found in the parachordal vessel, a superficial longitudinal vessel running through the horizontal myoseptum dividing the dorsal and ventral halves of the trunk somites (Isogai et al 2003). In one example of such an experiment, two parachordal cells migrated down to positions just ventral to the dorsal aorta and then divided to give a total of five thoracic duct LECs (Fig. 4c–e, see Yaniv et al 2006 for movie). As described in a previous report from our laboratory (Isogai et al 2003), the parachordal vessel is generated from vascular sprouts that emerge directly from the posterior cardinal vein at approximately 1.5 dpf. Thus, the ultimate source of LEC progenitors for the thoracic duct is in fact the posterior cardinal vein. Since every thoracic LEC in our time-lapse experiments could be traced back to the parachordal vessel, mesenchyme does not appear to give rise to LECs in the early thoracic duct. Together, these results support the idea that the primitive lymphatics

146

YANIV ET AL

ZEBRAFISH LYMPHATIC DEVELOPMENT

147

FIG. 4. Assembly and origins of the zebrafish lymphatic thoracic duct. (a,b) Two-photon time-lapse imaging of 2–4 dpf Tg(fl i1:EGFP) yl zebrafish, with a diagram of the area imaged (dark grey box) at time zero shown at top. Selected frames from the time-lapse sequence are shown below, with the lymphatic sprout highlighted in dark grey. (a) Lymphatic sprout emerging ventral to the dorsal aorta (see Yaniv et al 2006 for movie). (b) Lymphatic vessel sprouts growing across the trunk and merging ventral to the dorsal aorta (see Yaniv et al 2006 for movie). (c–e) Two-photon time-lapse imaging of EGFP-positive endothelial cell nuclei in the trunk of a 2–4 dpf Tg(fl i1:nEGFP) y7 zebrafish. (c) Explanatory diagram showing that two cells migrate ventrally from the parachordal vessels to contribute to the lymphatic thoracic duct. The diagram shows the location of the cells and their daughters at different time points. (d,e) Actual images from the time-lapse sequence showing the positions of lymphatic progenitor cells (1, 2) and their daughters (1a, 1b, 2a, 2b) at time zero (d) and 30.5 h after the start of the time lapse sequence (e). See Yaniv et al 2006 for movie. In all panels the time of selected frames is shown as hours:minutes after the start of the time-lapse sequence. Scale bars = 25 µm (a,b), 50 µm (c,d).

have a venous origin, as proposed by Sabin (1902). It remains to be seen whether LEC contributing to later-forming and/or peripheral lymphatics in zebrafish arise predominantly through further proliferation of these initial LECs or via recruitment from mesenchyme as proposed for avians and Xenopus (Ny et al 2005, Wilting et al 2006). In conclusion, we have demonstrated the existence of a well-developed lymphatic system in the zebrafish, and provided direct in vivo evidence for a venous origin for primitive LEC. It seems likely that the easily manipulated, genetic zebrafish model will provide additional novel insights into the formation and function of lymphatic vessels in the future. References Cao Y, Linden P, Farnebo J et al 1998 Vascular endothelial growth factor C induces angiogenesis in vivo. Proc Natl Acad Sci USA 95:14389–14394 Huntington G, McClure C 1910 The anatomy and development of the jugular lymph sac in the domestic cat (Felis domestica). Am J Anat 10:177–311 Isogai S, Lawson ND, Torrealday S, Horiguchi M, Weinstein BM 2003 Angiogenic network formation in the developing vertebrate trunk. Development 130:5281–5290 Jeltsch M, Kaipainen A, Joukov V et al 1997 Hyperplasia of lymphatic vessels in VEGF-C transgenic mice. Science 276:1423–5 (erratum: 1997 Science 277:463) Karkkainen MJ, Haiko P, Sainio K et al 2004 Vascular endothelial growth factor C is required for sprouting of the fi rst lymphatic vessels from embryonic veins. Nat Immunol 5:74–80 Kimmel CB, Ballard WW, Kimmel SR, Ullmann B, Schilling TF 1995 Stages of embryonic development of the zebrafish. Dev Dyn 203:253–310 Lawson ND, Weinstein BM 2002 In vivo imaging of embryonic vascular development using transgenic zebrafish. Dev Biol 248:307–318 Lawson ND, Mugford JW, Diamond BA, Weinstein BM 2003 Phospholipase C gamma-1 is required downstream of vascular endothelial growth factor during arterial development. Genes Dev 17:1346–1351 Long Q, Meng A, Wang H, Jessen JR, Farrell MJ, Lin S 1997 GATA-1 expression pattern can be recapitulated in living transgenic zebrafish using GFP reporter gene. Development 124:4105–4111

148

YANIV ET AL

Ny A, Koch M, Schneider M et al 2005 A genetic Xenopus laevis tadpole model to study lymphangiogenesis. Nat Med 11:998–1004 Oliver G, Alitalo K 2005 The lymphatic vasculature: recent progress and paradigms. Annu Rev Cell Dev Biol 21:457–483 Sabin F 1902 On the origin of the lymphatic system from the veins, and the development of the lymph hearts and thoracic duct in the pig. Am J Anat 1:367–389 Tilney NL 1971 Patterns of lymphatic drainage in the adult laboratory rat. J Anat 109: 369–383 van der Putte SC 1975a The development of the lymphatic system in man. Adv Anat Embryol Cell Biol 51:3–60 van der Putte SC 1975b The early development of the lymphatic system in mouse embryos. Acta Morphol Neerl Scand 13:245–286 Wigle JT, Oliver G 1999 Prox1 function is required for the development of the murine lymphatic system. Cell 98:769–778 Wigle JT, Harvey N, Detmar M et al 2002 An essential role for Prox1 in the induction of the lymphatic endothelial cell phenotype. EMBO J 21:1505–1513 Wilting J, Aref Y, Huang R et al 2006 Dual origin of avian lymphatics. Dev Biol 292:165–73 Yaniv K, Isogai S, Castranova D, Dye L, Hitomi J, Weinstein BM 2006 Live imaging of lymphatic development in the zebrafish. Nat Med 12:711–716 Yuan L, Moyon D, Pardanaud L et al 2002 Abnormal lymphatic vessel development in neuropilin 2 mutant mice. Development 129:4797–4806

DISCUSSION Vargesson: When the cells vacuolate, there are vacuoles of different sizes. How do the cells merge so that they form complete connections that aren’t leaking blood cells? Weinstein: Before these cells transfer their vacuoles into a common space, they must be able to establish an interface that is sealed off from the rest of the environment. This may be tied into the whole idea of how a lumen is made. It may be that the interface is part of a process of establishing an apical domain that corresponds to what will be the inside of the lumenal face of the vessel. Dejana: Why do cells start to make lacunae? We don’t see this vacuolation in the cells that are facing the lumen of the vessels. Could it be that when the cells are in contact with a three dimensional matrix and receive a signal that is all around them they form a lumen while the cells of the large vessels already have a lumen and do not form vacuoles. Weinstein: They didn’t have one initially. Dejana: The cells of the aorta are already polarized while the cells in your model have all their surface in contact with the extracellular matrix. Weinstein: I agree with your premise: the idea that it could be sensing matrix all around versus interface with a cell. I don’t agree though that it is somehow different from the axial vessels. The major axial vessels are initially just an aggregate or a cord of cells. They don’t have a lumen and have to sort it out. One way to think about this is that it is possible that larger vessels are doing it in exactly the analo-

ZEBRAFISH LYMPHATIC DEVELOPMENT

149

gous way, except that instead of pumping vacuoles into a space between just two cells, they are pumping vacuoles into a space bounded by many cells. Lammert: There are slight differences between this system and our in vitro system. The major one is that we see junctions in the stalk of the vessel but in the branching part we see seamless structures without cellular junction. Do you think that the entire inter-somatic vessel is seamless? Or does the vacuole mechanism only play a role in the connection point between the inter-somitic vessel and the aorta? Weinstein: We haven’t just imaged the most ventral cell, we have also imaged some of the cells higher up. They seem to be doing things in a similar way with the vacuoles. But we haven’t done the necessary comparable experiments looking at junctions to assess the role that they are playing. We have done some experiments where we generate mosaics with transgenics. This allows us to half do the experiment: we can see what the junction looks like from the perspective of one cell. We can guess where the interface with the cell below is. From this we can see that the interfaces are quite complex; they are not just discs. It doesn’t look like we have cells next to each other and rearranging to form a tube. Lammert: If you only see seamless lumen in the cells, which connect with the aorta, then you can say that this model of vacuole connection and formation only holds true for a subset of the cells within a vascular tube, namely the ones at the connection between two vessels. Weinstein: It is complicated and will require dynamic imaging with transgenic lines to work out. The junctions are complex: some cells make junctions with themselves. You may think you are looking at two cells but you aren’t. Ye: There is some experimental evidence suggesting that primitive lumens are fi lled with a special matrix. Does blood flow have a role in clearing the matrix? Weinstein: We don’t have any evidence that supports the idea that there is an insoluble matrix inside the nascent lumenal spaces of vessels. When we see this connection happening, we see a very rapid enlargement or inflation of the lumenal space. We don’t know what is causing this rapid enlargement but we see no evidence that there is anything inside that has to be removed. Ye: It would be interesting to use the silent heart mutant. Weinstein: We have done this, but it is complex to interpret results from the silent heart mutant experiments. With no flow or pressure, the vessels are sort of collapsed. If you don’t see a big enlargement of that space, is it because it requires a little pressure to get the thing to open up more, or is it not happening for a more basic mechanistic reason? What we have seen is formation of smaller vacuolar spaces in endothelial cells in silent heart mutants, but they don’t enlarge into larger spaces. Ye: Have you seen the same phenomena in the midline angioblasts that give rise to the axial vessels?

150

YANIV ET AL

Weinstein: We would like to look at the axial vessels, but for technical reasons we started with the segmental vessels because it is easier to do the imaging. Ye: The matrix that surrounds lymphatic vessels is controversial. Some people say it is discontinuous, others say there is no basement membrane. In your EM study, did you see evidence for basement membrane? Weinstein: I’d not want to weigh in too heavily on that. We haven’t seen anything that looks like basement membrane around the vessels, but we’d need to look more carefully to be sure. Drake: You were saying that it would be a bit harder to look at the dorsal aorta. Will you be able to do this? Weinstein: Yes, we hope so. Drake: In the lymphatics, is the cell that is part of the endothelium separating from the endothelium and moving as an individual cell? Weinstein: The question is, is it coming down as individual cells or is it a sprout or segment? It is beginning to look like the latter. At this point I’d be comfortable saying that it is going down the intersegmental boundary. Uv: Is the paracaudal vessel actually a vein as well? Weinstein: It certainly has a venous origin. Uv: As a representative of Drosophila I’d add that we see a very similar mode of tube formation where two branches join each other. It would be conceptually interesting to understand whether the seamless cells seen by electron microscopy are representing only such fusion branches, or whether other vessels form like that. Weinstein: My fantasy experiment would be to simultaneously visualize cellular junctions and cellular vacuoles using a double transgenic system. This would settle a lot of these kinds of issues. Uv: I don’t know how much you read the Drosophila literature, but it appears that lumen fusion between two fusion cells is preceded by establishment of an actinbased cord along which the lumen will form. Do you see any actin cord laid down to mediate lumen fusion? Weinstein: It is some sort of matrix cord. Uv: In the fly, what happens after branch fusion and lumen connection is that an extracellular luminal matrix is needed to push open the lumen. Perhaps flow is taking this role in your system? Weinstein: If there is something there in vessels it would have to be something that could turn over and change with rapid dynamics. Uv: In your system, the lumen looks regular before there is a flow. Then the lumen increases in diameter with the flow. This is why I was asking whether flow helps shape the lumen diameter. Weinstein: It is unlikely to be flow: it is more likely to be pressure. The problem is that there really isn’t any flow in these vessels for some time. All these vessels

ZEBRAFISH LYMPHATIC DEVELOPMENT

151

sprout from the aorta. They grow up and they are almost entirely lumenized before they make any connections with the venous system. There is no way for there to be any flow through the vessels for a while. Shibuya: You showed that the dependency on PLCγ is only for new vessel formation in the arterial system. Since PLCγ -deficient mice showed more severe vascular defects, downstream of the VEGF-receptor might be slightly different between fish and mammals. Also you said that some arterial vessels came from the venous system. When you block vessel formation by knockout of PLCγ at an early stage, do you lose both artery and a part of vein? Weinstein: In the fish we get primarily arterial defects with the knockout of PLCγ. There is much less of a defect in the venous system. Shibuya: I hear that the fish genome contains more than three VEGFR genes. Does the VEGFR3 in fish strictly correspond to that of mammals? Weinstein: One of my former postdocs has been trying to sort out the VEGF receptors and ligands in fish. (Covassin et al 2006). The bottom line is that all of the functions of the VEGF ligands and receptors are covered in the fish, but the precise match-up of which receptor and ligand is doing which part of the function is slightly different. Reference Covassin LD, Villefranc JA, Kacergis MC, Weinstein BM, Lawson ND 2006 Distinct genetic interactions between multiple Vegf receptors are required for development of different blood vessel types in zebrafish. Proc Natl Acad Sci USA 103:6554–6559

Signalling pathways regulating cardiac neural crest migration and differentiation Frances High and Jonathan A. Epstein Department of Cell and Developmental Biolog y and the Cardiovascular Institute, University of Pennsylvania, Philadelphia, PA 19104, USA Abstract. A critical contribution of neural crest to the developing cardiovascular system has been recognized for nearly 25 years. Recently, however, advanced mouse genetic techniques have revealed a series of previously unrecognized molecular pathways that regulate cardiac neural crest migration and differentiation. These involve members of the bone morphogenetic protein (BMP), T-box, myocardin, Gata and Notch families. In addition, molecules previously studied for their role in axon guidance have now been implicated in neural crest and cardiovascular patterning. In particular, members of the semaphorin family of secreted guidance molecules, along with plexin and neuropilin receptors, play critical roles during aortic arch remodelling and are implicated as candidate genes for contribution to congenital heart disease. 2007 Vascular development. Wiley, Chichester (Novartis Foundation Symposium 283) p 152–164

Cardiac neural crest Margaret Kirby first characterized the cardiac neural crest using avian embryos and identified a region of the neural tube between the mid-otic placode and the third somite that gives rise to neural crest cells that migrate through the pharyngeal region and contribute to the great vessels. Ablation of these cells as they emerge from the neural tube results in predictable forms of congenital heart disease including interrupted aortic arch and persistent truncus arteriosus (Kirby et al 1983). Our group and others have studied the migration and fate of cardiac neural crest cells during murine development. Genetic tools available in the mouse, which are not possible in the avian system, have allowed for significant advances in our understanding of the role of neural crest during growth and patterning of the cardiovascular system. Surprisingly, few specific markers of cardiac neural crest are available in the mouse and thus we developed fate-mapping techniques using the Cre-Lox system in order to specifically define the final fate and patterning of neural crest cells. The neural crest markers Pax3 and Wnt1 are expressed in the pre-migratory neural crest beginning at about E8.0 and E8.5, respectively. Both 152

CARDIOVASCULAR DEVELOPMENT

153

genes are inactivated soon after neural crest emergence, and thus cannot be used to follow the migration or fate of these cells. However, Wnt1-Cre transgenic mice and Pax3-Cre transgenic or Cre knockin mice, can be crossed to ‘Cre-reporter’ mice, which activate expression of a reporter gene such as LacZ or GFP under the control of a ubiquitous promoter upon Cre-mediated recombination and removal of a ‘stop’ sequence. This results in expression of the marker gene in all descendents of Wnt1- or Pax3-expressing cells. This type of genetic fate-mapping experiment has demonstrated that cardiac neural crest cells migrate throughout the pharyngeal region in large numbers. They coalesce around the aortic arch arteries, which initially form as endothelial tubes, and subsequently differentiate into vascular smooth muscle (Fig. 1A). In addition, they form the aorto-pulmonary septum, thus dividing the truncus arteriosus into the aorta and pulmonary artery. Smooth muscle in the aortic arch, ductus arteriosus, and carotid arteries are derived from neural crest, but a very sharp demarcation can be demonstrated at the level of the ductus arteriosus, beyond which vascular smooth muscle derives from lateral plate mesoderm (Fig. 1B) (Engleka et al 2005, Jiang et al 2000, Li et al 1999, 2000). Defective formation of neural crest-derived smooth muscle leads to abnormal remodelling of the aortic arch arteries, resulting in characteristic forms of congenital heart disease (Fig. 2). The proximal aorta and pulmonary trunk contain vascular smooth muscle that is not derived from neural crest, but rather derives from the anterior or secondary heart field (Waldo et al 2005). Smooth muscle in the coronary arteries derives from both secondary heart

FIG. 1. Cardiac neural crest cells contribute to smooth muscle of the aortic arches. (A) Depiction of a frontal section through the pharyngeal arches and three major aortic arch arteries of an E10.5 mouse embryo. Neural crest cells migrate through the pharyngeal arches and surround the aortic arch arteries, which initially exist as simple endothelial tubes. They then differentiate to form the vascular smooth muscle layer surrounding these arteries. (B) Fate mapping experiment in which Wnt1-Cre transgenic mice were crossed with a Cre reporter line, resulting in LacZ expression in derivatives of the neural crest. This heart from an adult mouse shows that neural crest cells contribute to the pulmonary artery (pa), the aortic arch (ao), and the proximal branch arteries. Note the sharp boundary between neural crest-derived cells of the aortic arch and non-neural crest-derived cells of the descending aorta (arrow).

154

HIGH & EPSTEIN

field and epicardium (Guadix et al 2006, Tomanek 2005, Waldo et al 2005). It is striking that smooth muscle in various regions of the aorta is derived from vastly different embryonic sources, but appears histologically identical. Nevertheless, it is possible that vascular disorders affecting specific regions of the aorta may be related to these distinct embryologic origins. Signalling pathways involved in neural crest-derived smooth muscle differentiation One of the surprising results that has emerged from tissue-specific knockout experiments and other genetic manipulations in mouse models is that several pathways appear to specifically affect neural crest-derived smooth muscle differentiation, while other aspects of vascular smooth muscle appear unaffected. One example involves the role of the myocardin-related factor MRTF-B. Myocardin, MRTF-A and MRTF-B represent a family of potent transcriptional co-activators that modulate serum response factor (SRF) transcription. Many smooth muscle specific genes are regulated by SRF and its co-factors. Myocardin appears to play a critical role in most, if not all, vascular smooth muscle since knockout mice display severe vascular defects (Li et al 2003). On the other hand, inactivation of MRTF-B results in specific defects in the cardiac outflow tract that mimic common forms of congenital disease including interrupted aortic arch and persistent truncus arteriosus (Li et al 2005, Oh et al 2005). Despite a broad expression domain, neural crest-derived vascular smooth muscle is specifically affected, suggesting potential functional redundancy with other related factors in non-neural crest-derived smooth muscle. The specific role of MRTF-B in neural crest was demonstrated by Cre-mediated rescue of normal cardiac development using a neural crest-specific Cre mouse to restore normal MRTF-B expression in an otherwise null background (Li et al 2005). MRTF-B probably acts as a docking protein to co-ordinate multiple signalling pathways upon SRF-dependent smooth muscle gene transcription. Potential signalling pathways that may be co-ordinated by MRTF-B in neural crest include downstream effectors of the BMP pathway. Tissue-specific inactivation of the Alk2 BMP receptor on neural crest cells results in cardiac outflow tract defects (Kaartinen et al 2004), and it will be of interest to determine if Smads, which mediate intracellular signalling downstream of BMP receptors, can interact directly or indirectly with MRTF-B in the nucleus. A significant body of work from several laboratories implicates the T-box transcription factor Tbx1 in cardiac outflow tract development. TBX1 is located on chromosome 22q11 in the region commonly deleted in humans with DiGeorge syndrome, a disorder characterized by cardiac outflow tract defects and other congenital abnormalities. A significant number of patients with isolated outflow

CARDIOVASCULAR DEVELOPMENT

155

tract defects, but without the other features of DiGeorge syndrome, also have deletions of chromosome 22q11. Some patients with DiGeorge syndrome do not have demonstrable deletions on chromosome 22, and in a few of these cases (though not in the majority) missense mutations in TBX1 have been identified (Yagi et al 2003). Heterozygous mice deficient for Tbx1 also have reproducible cardiac outflow defects, and neural crest-derived smooth muscle differentiation is impaired ( Jerome & Papaionannou 2001, Kochilas et al 2002, Lindsay et al 2001, Merscher et al 2001). This defect may be due to deficient Fgf8 signalling, normally induced by Tbx1 expression in pharyngeal endoderm. Support for this hypothesis comes from the fact that tissue-specific deletion of Fg f8 in pharyngeal endoderm using Tbx1-Cre mice reproduces typical outflow tract defects associated with deficient neural crest-derived smooth muscle differentiation (Brown et al 2004). Interestingly, FGF8 produced by head ectoderm (a location in which Tbx1 is not expressed) also influences cardiovascular patterning (Macatee et al 2003), and Crkl, another gene within the DiGeorge region of chromosome 22q11, may function to influence FGF signalling (Moon et al 2006). Accumulating data from both human genetics and animal models implicates Notch signalling in neural crest-derived smooth muscle differentiation. Human patients with Alagille syndrome suffer from congenital heart disease, including pulmonic stenosis and tetralogy of Fallot (McElhinney et al 2002). This disease has been linked to mutations in JAGGED1, a Notch ligand (Li et al 1997, Oda et al 1997). In zebrafish, gridlock mutants show a phenotype reminiscent of coarctation of the aorta and suffer from defective Notch signalling (Zhong et al 2001). There are four Notch ligands in mouse and man, and these are likely to serve partially redundant functions. Nevertheless, in mice, compound heterozygotes of Notch2 and Jagged1 display pulmonic stenosis strikingly similar to that seen in human Alagille patients (McCright et al 2002). Work from our own laboratory suggests that cell-autonomous Notch signalling in neural crest is required for smooth muscle differentiation and aortic arch remodelling (High et al 2007). Semaphorin–plexin signalling in cardiovascular development Semaphorins represent a family of membrane bound and secreted molecules that mediate attractive or (more commonly) repellent guidance cues for growing axons in the central nervous system. Like the ephrin and netrin families, semaphorins participate in the intricate patterning programmes that allow for the complex and reproducible network of neuronal circuitry. However, semaphorins are widely expressed outside of the central nervous system and play important roles in immune cells and other tissues. Recent work has indicated that several of the axon guidance molecule families also contribute to patterning of the vasculature, and several excellent reviews on this subject have recently appeared (Carmeliet &

156

HIGH & EPSTEIN

Tessier-Lavigne 2005, Eichmann et al 2005, Gu et al 2005, Weinstein 2005). Here, we focus on the data that implicate semaphorin signalling in aortic arch remodelling and neural crest-derived smooth muscle migration. Class 3 semaphorins are secreted molecules that bind to receptors that include plexin-type receptors. Many details of ligand-receptor interactions and subsequent intracellular signalling remain to be elucidated. Plexins bear homology to the c-met tyrosine kinase receptor, but they are unable to mediate tyrosine phosphorylation. In some cases, plexins appear to function as independent receptors, while in other cases they can heterodimerize with either neuropilin 1 or neuropilin 2 (Gu et al 2005, Kolodkin et al 1997). Interestingly, neuropilins can also heterodimerize with Flk1 to form a vascular endothelial growth factor (VEGF) receptor (Soker et al 1998). Until recently, plexins were studied predominantly in the central nervous system where they mediate axon guidance cues. Inactivation of the class 3 semaphorin, SEMA3C, resulted in an unexpected cardiovascular phenotype characterized by interruption of the aortic arch and persistent truncus arteriosus (Feiner et al 2001). This result strongly implicated SEMA3C in neural crest patterning, though SEMA3C itself is expressed both within neural crest cells and in surrounding mesenchyme and myocardium. We identified a putative SEMA3C receptor, PlexinA2, that is expressed more specifically in cardiac neural crest cells and is one of the more specific markers for post-migratory cardiac neural crest that we have utilized (Brown et al 2001). The knockout phenotype for PlexinA2 has not been reported and it remains unclear if PlexinA2 mediates any or all of SEMA3C signalling in the heart and vasculature. While we initially hypothesized that SEMA3C signalled in a paracrine fashion from neighbouring cells to PlexinA2-expressing cardiac neural crest, recent data have suggested an alternative model. In our search for potential SEMA3C receptors in the vascular system, we identified a novel member of the Plexin family (since described by several groups) that we named PlexinD1. Intriguingly, PlexinD1 is expressed primarily by vascular endothelial cells, although expression in the nervous and immune systems has more recently been identified (unpublished observations). In zebrafish, the PlexinD1 orthologue is also expressed by vascular endothelium, and morpholino knockdown results in abnormal patterning of the intersomitic vessels. Mutations in zPlexinD1 account for the previously described out of bounds mutation which is characterized by mispatterned vasculature. Zebrafish offer a significant advantage over murine embryos in that they are transparent, and growing blood vessels can be directly observed in real time. Use of GFP transgenic fish that label endothelial cells facilitates this process. Brant Weinstein’s lab, using time-lapse video microscopy, demonstrated that growth of intersomitic blood vessels proceeds by leading edge endothelial cells producing multiple fi lopodia. These fi lopodia fail to persist in regions of the somitic mesoderm that express class 3 semaphorins. This effect

CARDIOVASCULAR DEVELOPMENT

157

FIG. 2. Examples of congenital heart disease that can result from abnormal remodelling of the aortic arch arteries. Left panel shows a normal aortic arch, with a fully-septated aorta and pulmonary artery, a ductus arteriosus that connects the pulmonary artery to the aortic arch during fetal circulation, and the major aortic arch branches. Abnormal development of the aortic arch arteries can result in a spectrum of aortic arch defects, ranging from mild defects such as an anomalous retroesophageal right subclavian artery (depicted in middle panel), to severe defects such persistent truncus arteriosus with interrupted aortic arch (depicted in right panel). Abbreviations: aorta (ao), ductus arteriosus (da), right subclavian artery (rsa), right carotid artery (rca), left carotid artery (lca), left subclavian artery (lsa), persistent truncus arteriosus (pta).

is mediated by zPlexinD1, and knockdown of either ligand or receptor produces abnormal pathfinding by leading edge endothelium (Torres-Vazquez et al 2004). In zebrafish, aortic arch remodelling and outflow tract development does not adequately model human development because the fish has a two-chambered heart without a pulmonary circulation, and septation of the outflow tract does not occur. Hence, a role for zPlexinD1 in aortic arch remodelling has not been appreciated. However, inactivation of PlexinD1 in the mouse results in neonatal lethality and dramatic, fully penetrant, congenital heart disease. This is characterized by persistent truncus arteriosus and interruption of the aortic arch, a phenotype similar to that seen in Sema3C knockout mice. Peripheral blood vessel patterning defects are also observed (Gitler et al 2004). Controversy persists as to whether PlexinD1 can transduce class 3 semaphorin signals alone, or if heterodimerization with neuropilin subunits is also required. Our group has shown increased affinity of PlexinD1 for SEMA3C and SEMA3A when neuropilin is present, while others have suggested that PlexinD1 may be able to function as an independent SEMA3E receptor in the periphery (Gu et al 2005). Both results may be true, but more detailed analysis of receptor–ligand interaction and subsequent intracellular signalling

158

HIGH & EPSTEIN

events are required. In support of the hypothesis that neuropilin subunits contribute to semaphorin signalling in the outflow tract is the fact that a neuropilin 1 mutation that specifically affects semaphorin signalling (and not VEGF signalling), when expressed on a background of neuropilin 2 deficiency, results in cardiac outflow tract defects almost identical to those seen in PlexinD1 deficient mice (Gu et al 2003). The similarity in phenotype is striking in that both mutants have a single midline outflow vessel and associated identical coronary artery defects. The phenotypic overlap of these neuropilin mutant mice, PlexinD1 mutant mice, and SEMA3C mutant mice suggest that these molecules function in a closely related signalling pathway. Recent results further support the model that SEMA3C functions within neural crest cells to mediate aortic arch remodelling and neural crest migration. GATA6 is a transcription factor expressed in smooth muscle, lung endoderm, and heart. Global gene inactivation results in early embryonic lethality, probably due to an early requirement in endoderm (Morrisey et al 1998). However, tissue-specific inactivation in vascular smooth muscle, using an SM22 α -Cre mouse, results in perinatal lethality and congenital heart disease. Aortic arch derivatives are affected, while other regions of the vasculature are spared. Tissue-specific inactivation of GATA6 in neural crest recapitulates this phenotype, producing interruptions of the aortic arch, persistent truncus arteriosus, and related outflow tract defects. In these embryos, SEMA3C is specifically down-regulated in cardiac neural crest derivatives, while the expression of SEMA3C in myocardium and elsewhere is unaffected. Cardiac neural crest migration and patterning is deficient. Several GATA binding sites are present in the SEMA3C upstream regulatory region, which are conserved across evolution, and GATA6 can directly activate expression of relevant SEMA3C reporter constructs (Lepore et al 2006). These results strongly suggest that GATA6 regulates expression of SEMA3C in cardiac neural crest, and that SEMA3C expression in these cells is required for appropriate patterning and vascular smooth muscle development. Conclusions Congenital cardiac defects involving the outflow tract of the heart and the derivatives of the aortic arch arteries are common birth defects that can be unified by an understanding of normal embryological development and the role of cardiac neural crest in vascular smooth muscle development. GATA6, SEMA3C, PlexinD1 and neuropilins all play important roles in cardiac neural crest migration and patterning (Fig. 3), yet the details of receptor–ligand interactions and subsequent intracellular signalling events remain to be elucidated. Striking similarities between axon guidance in the CNS, and vascular patterning have become apparent, perhaps at least partially explaining the age-old observation that blood vessels and nerves

CARDIOVASCULAR DEVELOPMENT

159

FIG. 3. Schematic representation of cardiac neural crest migration and differentiation into vascular smooth muscle. Some of the genes discussed in this review are indicated.

often travel together. This emerging paradigm offers attractive new targets for angiogenic and anti-angiogenic therapeutic approaches. Many examples of cardiac outflow tract defects in animal models, and perhaps in humans, relate to deficient smooth muscle differentiation of cardiac neural crest, and molecules that play specific roles in neural crest smooth muscle differentiation, but not in other vascular smooth muscle, are emerging (Fig. 3). Thus, the molecular programmes regulating cell fate decision in the vasculature appear to be, at least in some cases, region-specific and related to the cell of origin, despite apparently similar ultimate phenotypes of all vascular smooth muscle cells. These observations provide numerous candidate genes for involvement in congenital heart disease that need to be tested in syndromic and non-syndromic disorders, and may explain why some vascular abnormalities have a predilection for specific regions of the vasculature. The detailed interactions between neural crest, vascular smooth muscle, endothelium and other components of the growing cardiovascular system (including epicardium and primary and secondary heart fields) is an exciting area of active investigation. References Brown CB, Feiner L, Lu MM et al 2001 PlexinA2 and semaphorin signaling during cardiac neural crest development. Development 128:3071–3080 Brown CB, Wenning JM, Lu MM, Epstein DJ, Meyers EN, Epstein JA 2004 Cre-mediated excision of Fgf8 in the Tbx1 expression domain reveals a critical role for Fgf8 in cardiovascular development in the mouse. Dev Biol 267:190–202 Carmeliet P, Tessier-Lavigne M 2005 Common mechanisms of nerve and blood vessel wiring. Nature 436:193–200

160

HIGH & EPSTEIN

Eichmann A, Makinen T, Alitalo K 2005 Neural guidance molecules regulate vascular remodeling and vessel navigation. Genes Dev 19:1013–1021 Engleka KA, Gitler AD, Zhang M, Zhou DD, High FA, Epstein JA 2005 Insertion of Cre into the Pax3 locus creates a new allele of Splotch and identifies unexpected Pax3 derivatives. Dev Biol 280:396 Feiner L, Webber AL, Brown CB et al 2001 Targeted disruption of semaphorin 3C leads to persistent truncus arteriosus and aortic arch interruption. Development 128:3061–3070 Gitler AD, Lu MM, Epstein JA 2004 PlexinD1 and semaphorin signaling are required in endothelial cells for cardiovascular development. Dev Cell 7:107–116 Gu C, Rodriguez ER, Reimer DV et al 2003 Neuropilin-1 conveys semaphorin and VEGF signaling during neural and cardiovascular development. Dev Cell 5:45–57 Gu C, Yoshida Y, Livet J, Reimert DV et al 2005 Semaphorin 3E and Plexin-D1 Control Vascular Pattern Independently of Neuropilins. Science 307:265–268 Guadix JA, Carmona R, Munoz-Chapuli R, Perez-Pomares JM 2006 In vivo and in vitro analysis of the vasculogenic potential of avian proepicardial and epicardial cells. Dev Dyn 235:1014–1026 High FA, Zhang M, Proweller A et al 2007 An essential role for Notch in neural crest during cardiovascular development and smooth muscle differentiation. J Clin Invest 17:353–363 Jerome LA, Papaionannou VE 2001 DiGeorge syndrome phenotype in mice mutant for the T-box gene, Tbx1. Nat Genet 27:286–291 Jiang X, Rowitch DH, Soriano P, McMahon AP, Sucov HM 2000 Fate of the mammalian cardiac neural crest. Development 127:1607–1616 Kaartinen V, Dudas M, Nagy A, Sridurongrit S, Lu MM, Epstein JA 2004 Cardiac outflow tract defects in mice lacking ALK2 in neural crest cells. Development 131:3481–3490 Kirby ML, Gale TF, Stewart DE 1983 Neural crest cells contribute to normal aorticopulmonary septation. Science 220:1059–1061 Kochilas L, Merscher-Gomez S, Lu MM et al 2002 The role of neural crest during cardiac development in a mouse model of DiGeorge syndrome. Dev Biol 251:157–166 Kolodkin AL, Levengood DV, Rowe EG, Tai YT, Giger RJ, Ginty DD 1997 Neuropilin is a semaphorin III receptor. Cell 90:753–762 Lepore JJ, Mericko PA, Cheng L, Lu MM, Morrisey EE, Parmacek MS 2006 GATA-6 regulates semaphorin 3C and is required in cardiac neural crest for cardiovascular morphogenesis. J Clin Invest 116:929–939 Li L, Krantz ID, Deng Y et al 1997 Alagille syndrome is caused by mutations in human Jagged1, which encodes a ligand for Notch1. Nat Genet 16:243–251 Li J, Liu KC, Jin F, Lu MM, Epstein JA 1999 Transgenic rescue of congenital heart disease and spina bifida in Splotch mice. Development 126:2495–2503 Li J, Chen F, Epstein JA 2000 Neural crest expression of Cre recombinase directed by the proximal Pax3 promoter in transgenic mice. Genesis 26:162–164 Li S, Wang D-Z, Wang Z, Richardson JA, Olson EN 2003 The serum response factor coactivator myocardin is required for vascular smooth muscle development. Proc Natl Acad Sci USA 100:9366–9370 Li J, Zhu X, Chen M et al 2005 Myocardin-related transcription factor B is required in cardiac neural crest for smooth muscle differentiation and cardiovascular development. Proc Natl Acad Sci USA 102:8916–8921 Lindsay EA, Vitelli F, Su H et al 2001 Tbx1 haploinsufficiency in the DiGeorge syndrome region causes aortic arch defects in mice. Nature 410:97–101 Macatee TL, Hammond BP, Arenkiel BR, Francis L, Frank DU, Moon AM 2003 Ablation of specific expression domains reveals discrete functions of ectoderm- and endoderm-derived FGF8 during cardiovascular and pharyngeal development. Development 130:6361–6374 McCright B, Lozier J, Gridley T 2002 A mouse model of Alagille syndrome: Notch2 as a genetic modifier of Jag1 haploinsufficiency. Development 129:1075–1082

CARDIOVASCULAR DEVELOPMENT

161

McElhinney DB, Krantz ID, Bason L et al 2002 Analysis of cardiovascular phenotype and genotype-phenotype correlation in individuals with a JAG1 mutation and/or Alagille syndrome. Circulation 106:2567–2574 Merscher S, Funke B, Epstein JA et al 2001 TBX1 is responsible for cardiovascular defects in velo-cardio-facial/DiGeorge Syndrome. Cell 104:619–629 Moon AM, Guris DL, Seo JH et al 2006 Crkl deficiency disrupts Fgf8 signaling in a mouse model of 22q11 deletion syndromes. Dev Cell 10:71–80 Morrisey EE, Tang Z, Sigrist K et al 1998 GATA6 regulates HNF4 and is required for differentiation of visceral endoderm in the mouse embryo. Genes Dev 12:3579–3590 Oda T, Elkahloun AG, Pike BL et al 1997 Mutations in the human Jagged1 gene are responsible for Alagille syndrome. Nat Genet 16:235–242 Oh J, Richardson JA, Olson EN 2005 Requirement of myocardin-related transcription factor-B for remodeling of branchial arch arteries and smooth muscle differentiation. Proc Natl Acad Sci USA 102:15122–15127 Soker S, Takashima S, Miao HQ, Neufeld G, Klagsbrun M 1998 Neuropilin-1 is expressed by endothelial cells and tumor cells as an isoform-specific receptor for vascular endothelial growth factor. Cell 92:735–745 Tomanek RJ 2005 Formation of the coronary vasculature during development. Angiogenesis 8:273–284 Torres-Vazquez J, Gitler AD, Fraser SD et al 2004 Semaphorin-plexin signaling guides patterning of the developing vasculature. Dev Cell 7:117–123 Waldo KL, Hutson MR, Ward CC et al 2005 Secondary heart field contributes myocardium and smooth muscle to the arterial pole of the developing heart. Dev Biol 281:78–90 Weinstein BM 2005 Vessels and nerves: marching to the same tune. Cell 120:299–302 Yagi H, Furutani Y, Hamada H et al 2003 Role of TBX1 in human del22q11.2 syndrome. Lancet 362:1366–1373 Zhong TP, Childs S, Leu JP, Fishman MC 2001 Gridlock signalling pathway fashions the fi rst embryonic artery. Nature 414:216–220

DISCUSSION Ruhrberg: Chenghua Gu and her colleagues made a conditional neuropilin knockout mouse, and deleted neuropilin 1 in endothelial cells with the Tie2-Cre promoter (Gu et al 2003). They suggested that the heart phenotype of these mice was caused by defective VEGF164 signalling through neuropilin 1, but the phenotype seems to be similar to the phenotype of your PlexinD1 knockout mice. Based on what you know now, might the conditional neuropilin 1 knockout phenotype be due to defective Sema3C signalling through a PlexinD1/neuropilin 1 receptor? Epstein: Yes, that’s my belief. The VEGF165 knockout also has outflow tract defects, so it is difficult to work out how much of the neuropilin deficiency is related to VEGF signalling versus SEMA3C signalling. Ruhrberg: Maybe there are two different neuropilin 1-mediated pathways that co-operate during aortic arch remodelling, one involving VEGF165 signalling and one involving SEMA3C signalling. Drake: I was interested in the neural crest population in the cushions. Did the little grouping of neural crest cells that you showed in the cushion ever fan out and integrate with other cushion cells?

162

HIGH & EPSTEIN

Epstein: In the SEMA3C knockout where the patterning of these cells is disrupted, they fan out and disperse into the cushions, and even into the myocardium, but not normally. Drake: Do they show up in the adult valve? Epstein: That is controversial. A recent paper (Nakamura et al 2006) from Jeff Robbins’ lab argues that they do. We see that they mostly die. They persist in the smooth muscle in the septum, not in the valves. Drake: What would be their transient role in the valve? Epstein: I bet that there is an inductive interaction between neural crest and valve mesenchyme that is involved in valve development. Drake: In the PlexinD1 mice you see arch defects, so clearly the mice lived long enough to manifest those defects. You also showed severely disturbed intersomitic vessels. If all the vascular defects were as profound as those observed in the intersomitic vessels, it is unlikely that the mice would have lived long enough to manifest the arch defects. Epstein: The peripheral vasculature is very messed up, as you say. The major vessels other than the arch are normal. The vascular defects aren’t severe enough to kill them before birth. Lammert: So SEMA3C is expressed by the neural crest cells, and its receptors are on neural crest cells as well as endothelial cells. What is the contribution of this signalling at the developmental stages of aortic arch formation? Epstein: We don’t know. Our present model is that neural crest cells signal to endothelium. But we don’t have direct evidence for this. They could be signalling to themselves. Adams: This suggests that there might be the possibility that ligand and receptor are both expressed in the neural crest. Epstein: Cardiac neural crest cells express PlexinA2, which is a potential SEMA3C receptor, but I don’t have direct biochemical evidence that it can function as a semaphorin receptor. Adams: I have a question about the function of the smooth muscle cells on the arch. Are they providing growth factor or matrix that is required to maintain the arch? What happens to the arch in those cases where the smooth muscle cells are missing? Epstein: I don’t know how they are supporting the endothelial tube and preventing it from regressing, other than by providing structural support. This region of the vasculature might be a good model for how supporting cells are working elsewhere because it is such a proscribed area amenable to study. Adams: I have another question about the different sources of smooth muscle cells in the body. Is anything known about the reasons for these different origins? Are there differences in gene expression? Epstein: I am surprised at how different the genetic programme seems to be for smooth muscle differentiation in neural crest cells compared with other cells that

CARDIOVASCULAR DEVELOPMENT

163

form smooth muscle. I would have expected that different cell types would use the same tricks again and again. I am surprised that we can knock out a gene in smooth muscle everywhere and it is expressed in smooth muscle everywhere, like MRTF-B, and yet we get a specific neural crest defect. Perhaps there is functional redundancy in non-neural crest-derived smooth muscle, but it does seem like the programmes have diverged quite dramatically. Owens: This is a theme that is evolving. The MRTF-B knockout results showing selective effects in neural crest-derived smooth muscle cells (SMCs) are one of the clearest examples. However, consistent with this general theme, we have shown that site directed mutagenesis of selected cis elements within the promoters of smooth muscle marker gene promoters will completely abrogate expression in one SMC subtype and have no effect in another (Manabe & Owens 2001). Moreover, in unpublished studies (Qiong and Owens, unpublished results) we have shown that SMCs derived from different embryological origins sometimes employ fundamentally different mechanisms and transcription factors to turn on the same marker genes including smooth muscle α -actin. One of the questions that Eckhard Lammert began to address was the cell autonomous functions of these molecules. You even postulated that the receptor and ligand are in the same cell. Have you done chimeric knockouts combined with lineage tracing methods to selectively eliminate the receptor versus the ligand in ESC and determine whether or not it affects their ability to become smooth muscle? Epstein: We haven’t done this, but we should. I would also be interested in what I suspect is a back and forth conversation with the endothelium in addition to the question of autocrine signalling. Owens: You showed very nicely that Notch impacts the MRTF-B pathway. What is the mechanism for this? Is there direct interaction of Notch with MRTF-B? High: We are not sure, but this is something we are actively pursuing. Epstein: Notch has been implicated both in promoting and preventing smooth muscle differentiation. Smooth muscle cells can act very differently depending on whether they are in a resting state or a contractile state, and their responses to Notch activation may be different. It may depend on how the assays are done. Notch can be difficult. At some points it seems it can repress MRTF-B, and at other times it can activate it. Ye: I have a comment on your response to Chris’ question about how hypervascularization relates to lethality. In your PlexinD1 knockout, the hypervascularization defect seems to be transient. At E10.5 hypervascularization is profound. But later, intercostals hypervascularization is minor, so it seems to resolve itself rapidly. Epstein: There are intercostal abnormalities in the adult. But it is also regional in that model. Ye: It seems to be transient, but if you had a persistent hypervascularization phenotype, it could lead to lethality.

164

HIGH & EPSTEIN

Augustin: I’d like to address the issue of neuropilin-dependent and neuropilinindependent functions of the semaphorins. You showed some overlap in the expression pattern, but mostly in the outflow tract you characterized Sema3C function acting in a paracrine manner. I don’t think you showed the corresponding neuropilin expression pattern? Epstein: It matches with PlexinD1 quite nicely. It is expressed broadly. Augustin: With regard to attractive versus repulsive functions, in the endothelial compartment you primarily stressed repulsive functions. We are working on a different class 3 semaphorin that we think acts in a stimulating manner on endothelial cells. Epstein: In the CNS class 3 semaphorins can be attractive or repulsive. Augustin: In the experiment you showed with SEMA3C, was that full length or processed? Epstein: It was processed. I have a question for the zebrafish people. In the mouse, there are a couple of examples of ectopic vessels forming in the initial sprouting process. This seems to be corrected after a while. What happens in zebrafish? Weinstein: In the plexin mutant fish, we find fish with these defects that make it all the way to adulthood. Perhaps a quarter survive if you are nice to them. We are in the process of looking closely at what is wrong with them. There are many vascular defects present in these adults. We have mainly looked at the superficial vessels. They have a lot of problems, yet they somehow have managed to get together a functioning network. This suggests that there must be some degree of correction, but also that the vasculature is an amazingly plastic system. I make a big point of patterning, but it has an amazing ability to put something together even when you force it to not have pattern. Ruhrberg: Comparing the phenotypes of plexin and semaphorin mutants in zebrafish and mice, it seems to me that the orthologues and homologues of the respective genes may not have been assigned correctly in all cases—the cardiovascular defects in zebrafish and mouse mutants are quite different, even though the expression patterns of the semaphorins and plexins aren’t that different. References Gu C, Rodriguez ER, Reimert DV et al 2003 Neuropilin-1 conveys semaphorin and VEGF signaling during neural and cardiovascular development. Dev Cell 5:45–57 Manabe I, Owens GK 2001 The smooth muscle myosin heavy chain gene exhibits smooth muscle subtype selective modular regulation in vivo. J Biol Chem 276:39076–39087 Nakamura T, Colbert MC, Robbins J 2006 Neural crest cells retain multipotential characteristics in the developing valves and label the cardiac conduction system Circ Res 98: 1547–1554

Investigation of the angiogenic programme with tissue-specific and inducible genetic approaches in mice Ralf H. Adams Vascular Development Laboratory, Cancer Research UK London Research Institute, 44 Lincoln’s Inn Fields, London WC2A 3PX, UK

Abstract. Blood vessels form a highly organized hierarchical network throughout the vertebrate body, integrate functionally into very different tissue environments and remain remarkably adaptable to changing local requirements. The importance of a correctly formed and functional vascular network is highlighted by numerous human pathologies, which involve defects such as compromised blood circulation, destabilization of vessel walls or deregulated angiogenesis. Genetically modified mice are powerful tools for the functional characterization of genes, and can also serve as models of human diseases. However, targeted inactivation of genes essential for blood vessel morphogenesis is usually incompatible with embryonic survival, so studies in adult mice are precluded. Moreover, mutant phenotypes can be a complex combination of defects in different tissues and cell types. Tissue-specific and inducible genetic approaches allow us to overcome many of these limitations. In particular, it is possible to perform manipulations in a spatially (i.e. cell type-specific) and temporally controlled fashion, which makes possible studies throughout development and in adults. We have generated several transgenic Cre and tamoxifen-induble CreERT2 lines that enable the selective targeting of vascular cell populations. Here, I will discuss the characterization of these lines and present examples of their application in the analysis of vascular development. 2007 Vascular development. Wiley, Chichester (Novartis Foundation Symposium 283) p 165–173

The vertebrate body contains an extensive network of blood vessels that mediates the transport of liquids, gases, nutrients, waste products, hormones and cells between tissues and organ systems. Blood circulation is established early during embryonic development and remains indispensable throughout the full lifespan of an organism. Consequently, the extensive growth of the fetal and postnatal body requires a significant expansion of the vascular network without compromising its essential functions. The importance of blood vessels is also highlighted by numerous human pathologies involving malformation or malfunction of the vasculature. The investigation of physiological angiogenesis in the embryo can provide detailed 165

166

ADAMS

insight into the molecular and cellular mechanisms controlling blood vessel growth and, at the same time, lead to the identification of potential therapeutic targets for the treatment of pathological conditions (Carmeliet 2003, Eichmann et al 2005, Gerhardt & Betsholtz 2003, Jain 2003). Important features of the vascular network such as its hierarchical organisation into vessels of larger and smaller diameter, the role of haemodynamic factors, the interactions between different vascular cell types and with surrounding tissues, the role of circulating cells, or the cross-talk between local and systemic signals cannot be sufficiently mimicked by tissue culture systems in the foreseeable future. Thus, angiogenesis is best studied in its physiological context, i.e. in animal models. This approach can also allow us to address the role of genes and potential therapeutic reagents in a range of pathological settings, which has big implications for translational and clinical research. Work over recent years has highlighted the specific advantages of different vertebrate organisms for such studies. Gene expression in fish and frogs can be easily manipulated with antisense RNA technology although it is currently not possible to create clean gene knockouts. The transparency of the zebrafish embryo makes it particularly suitable for the visualization of fine morphological details or dynamic growth processes (Ny et al 2006, Kidd & Weinstein 2003, Weinstein 2002). The chick embryo is very accessible to physical manipulations such as grafting experiments or the alteration of the blood flow pattern (Ribatti et al 2001, le Noble et al 2005). Unfortunately, all these organisms offer a relatively limited experimental time window and are mainly used for studies during certain stages of their embryonic development. In contrast, the mouse, the main mammalian model organism in biological research, allows the investigation of developmental processes and is also a suitable model for homeostasis and pathological conditions in the adult. A key advantage of the mouse is its accessibility to genetic manipulations such as transgenesis, gene knockouts, knock-ins or tissue-specific approaches, which, although expensive and time-consuming, allow very specific changes in a molecular pathway. Thus, it is possible to conclusively address the role of individual genes, unravel the relevance of specific molecular interactions and signalling pathways, or reproduce a genetic defect observed in humans. For the same reasons, the mouse embryo is also widely used for developmental studies although, as outlined above, lower vertebrates and avians are much better suited for certain experimental techniques. Another potential limitation of genetic manipulations is that they can lead to early embryonic lethality, which reflects an essential role of a gene in an early developmental process but also precludes the investigation of processes occurring later in the embryonic or postnatal organism. This applies to all genetic model organisms, even invertebrates, but is particularly relevant for the vascular network in the mouse where the inactivation of key genes will impair blood circulation,

GENETIC APPROACHES FOR STUDYING ANGIOGENESIS

167

cause growth retardation and, ultimately, lead to lethality before midgestation (usually around embryonic day [E] 10.5 to 11). Early embryonic lethality indicates that the gene in question is, in one way or another, essential (and therefore important) but limits all further studies to the narrow time window before effects in the mutant become too catastrophic. It also complicates the identification and characterization of primary defects because of the general growth retardation, which usually sets in from E9.0 onwards, and the interdependence of functional components within the embryonic cardiovascular system, i.e. heart, placenta and blood vessels (Adams et al 2000, Jones et al 2004). These problems can be avoided with the help of inducible genetic approaches, which are usually also cell type-specific, thus offering additional experimental benefits. Numerous methods for inducible overexpression of gene products have been successfully tested in mice. Examples include the bacterial lac operator-repressor system (Cronin et al 2001), transcriptional regulators containing the ligand-binding domain of the receptor for the insect steroid hormone ecdysone (No et al 1996), the cytochrome P450 induction system (Campbell et al 1996) or the tetracycline-responsive (Tet) system (Gossen & Bujard 1992, Kistner et al 1996). So far, only applications of the latter can be frequently found in the published literature. Typically, a binary transgenic approach is used for the Tet system in which animals carrying a target transgene under control of a minimal promoter containing tetracycline-responsive promoter elements (TREs) are bred to a second line expressing tetracycline-controlled regulators in a tissue or cell-type specific fashion. The Tet system comes in two flavours and can be either based on a tetracycline-controlled transactivator protein, tTA (Tet-Off), which is active in the absence of tetracycline or the synthetic derivative doxycycline (Dox), or the reverse tetracycline-controlled transactivator, rtTA (TetOn), which is only active in the presence of the inducing agent. Both systems are very similar and utilize a fusion protein between the Tet repressor DNA binding protein (TetR) from the tetracycline resistance operon of the E. coli transposon Tn10 and the transactivation domain of the Herpes simplex virus VP16 protein. Only four amino acids in the TetR binding region of tTA and rtTA are responsible for the different functional properties of the two proteins (Ryding et al 2001). For the application of the Tet system in the vasculature, several transgenic lines expressing tTA or rtTA under the control of promoters that direct expression to endothelial cells (U. Deutsch, unpublished) or vascular smooth muscle cells (West et al 2004) have been generated. For cell type-specific loss-of-function studies, recombinase-based systems are widely used in mice. The recombinase Cre (causes recombination) of the P1 bacteriophage can release the DNA surrounded by so-called loxP (locus of crossover in P1) sites as a circular fragment. LoxP sites are short sequences of 34 base pairs that can be easily introduced into a gene locus by homologous recombination in cultured mouse embryonic stem cells. The design of the targeting construct should

168

ADAMS

place the loxP sites around an essential area of the gene, for example exons encoding a critical functional domain. Breeding of mice carrying the ‘loxed’ gene to transgenics expressing Cre under the control of a tissue or cell type-specific promoter will lead to the selective excision of the loxP-flanked segment and thereby gene inactivation. Several existing Cre lines allow gene inactivation in the endothelium (Kisanuki et al 2001, Kogata et al 2006, Alva et al 2006, Gustafsson et al 2001, Licht et al 2004) or in pericytes and vascular smooth muscle cells (Foo et al 2006, Lepore et al 2005). As an alternative to Cre, a combination of the Flp integrase of Saccharomyces cerevisiae and FRT (Flp recombination target) sites can be used for the same purpose (Dymecki 1996a, b). However, Flp recombinase is more thermolabile than Cre so that recombination with the first generation of Flp transgenes was usually incomplete and mosaic. Although these limitations have been resolved by the development of a thermostable enhanced FLP recombinase (Rodriguez et al 2000), the Flp-FRT system is today not commonly used for tissue-specific manipulations. Another reason for the predominance of Cre-based approaches in mice is the development of inducible hybrid proteins of Cre and the ligand-binding domain (LBD) of a mutated oestrogen receptor (ER). This CreER fusion is inactively sequestered in the cytoplasm and will only translocate to the nucleus in response to the administration of the oestrogen antagonists tamoxifen or 4-hydroxytamoxifen. Consequently, gene targeting is not only tissue-specific (depending on the promoter controlling expression of CreER) but also temporally restricted (Metzger & Chambon 2001, Ryding et al 2001). This method has obvious benefits for studies in many organs and tissues but is particularly advantageous in the cardiovascular system. Various CreER versions are currently in use, which contain one or two LBDs and carry varying numbers of mutations in the Cre open reading frame to optimize codon usage for mammals. Although Cre-based systems are now reasonably established, reliable endothelial-specific and inducible gene inactivation has been difficult to achieve in the past, perhaps due to choice of gene promoters leading to low and/or heterogeneous expression of the CreER hybrid. We have successfully used homologous recombination in bacteria (Copeland et al 2001) to introduce a special version of CreER, CreERT2 (Feil et al 1997), into large fragments of mouse genomic DNA contained in bacterial or bacteriophage P1-derived artificial chromosomes called BAC and PAC, respectively (Osoegawa et al 2000). The rationale for this strategy is that PAC/BAC-based transgenic mice will presumably contain all the necessary regulatory elements to reproduce the expression pattern of the corresponding endogenous genes. This approach has yielded VE-cadherin(PAC)-CreERT2 and Bmx(PAC)-CreERT2 transgenics, which allow studies in the endothelium of microvessels and arteries, respectively. We found that tamoxifen administration induces efficiently EC-specific Cre activity in a wide range of embryonic and adult

GENETIC APPROACHES FOR STUDYING ANGIOGENESIS

169

tissues. We are now routinely using CreERT2 transgenics for functional studies during late embryonic or postnatal development. For example, global knockout mice lacking the gene for the transmembrane protein ephrin-B2 (Efnb2), a ligand for Eph family receptor tyrosine kinases, die before day 11 of embryonic development due to severe vascular defects (Adams et al 1999, Wang et al 1998). The generation of mice carrying a ‘loxed’ Efnb2 gene (Grunwald et al 2004) and of VE-cadherin(PAC)-CreERT2 transgenics (R.H. Adams, unpublished results) has allowed us to circumvent this lethality and target the gene in the endothelium during late gestation. At the same time, we are also using the Tet system to overexpress ephrin-B2 in the endothelium (R.H. Adams, unpublished results). Together, the two approaches show that the ligand is required for angiogenic remodelling and endothelial sprouting. Moreover, ephrin-B2 is also expressed in the pericytes (PCs) and vascular smooth muscle cells (vSMCs) forming the outer layers of the blood vessel wall. Inactivation of Efnb2 in these mural cells disrupts normal endothelial–mural interactions and leads to the destabilization of microvessels. This work demonstrates ephrin-B2 has indispensable roles in different vascular cell populations and provides an excellent example for the benefits of cell typespecific genetic manipulations. In conclusion, spatially and temporally controlled genetic manipulations using the tetracycline-controlled system or Cre/CreER recombinases have become a key instrument for gain-of-function and loss-of-function studies in the cardiovascular system. It is obvious that inducible systems will have enormous relevance for studies in adult mice and in mouse models of human pathologies. Genetic approaches in the mouse will significantly increase our understanding of gene function in the normal and diseased cardiovascular system and might allow the efficient identification and characterisation of potential therapeutic targets. References Adams RH, Wilkinson GA, Weiss C et al 1999 Roles of ephrinB ligands and EphB receptors in cardiovascular development: demarcation of arterial/venous domains, vascular morphogenesis and sprouting angiogenesis. Genes Dev 13:295–306 Adams RH, Porras A, Alonso G et al 2000 Essential role of p38alpha MAP kinase in placental but not embryonic cardiovascular development. Mol Cell 6:109–116 Alva JA, Zovein AC, Monvoisin A et al 2006 VE-Cadherin-Cre-recombinase transgenic mouse: a tool for lineage analysis and gene deletion in endothelial cells. Dev Dyn 235:759–767 Campbell SJ, Carlotti F, Hall PA, Clark AJ, Wolf CR 1996 Regulation of the CYP1A1 promoter in transgenic mice: an exquisitely sensitive on-off system for cell specific gene regulation. J Cell Sci 109:2619–2625 Carmeliet P 2003 Angiogenesis in health and disease. Nat Med 9:653–660 Copeland NG, Jenkins NA, Court DL 2001 Recombineering: a powerful new tool for mouse functional genomics. Nat Rev Genet 2:769–779 Cronin CA, Gluba W, Scrable H 2001 The lac operator-repressor system is functional in the mouse. Genes Dev 15:1506–1517

170

ADAMS

Dymecki SM 1996a Flp recombinase promotes site-specific DNA recombination in embryonic stem cells and transgenic mice. Proc Natl Acad Sci USA 93:6191–6196 Dymecki SM 1996b A modular set of Flp, FRT and lacZ fusion vectors for manipulating genes by site-specific recombination. Gene 171:197–201 Eichmann A, Makinen T, Alitalo K 2005 Neural guidance molecules regulate vascular remodeling and vessel navigation. Genes Dev 19:1013–1021 Feil R, Wagner J, Metzger D, Chambon P 1997 Regulation of Cre recombinase activity by mutated estrogen receptor ligand-binding domains. Biochem Biophys Res Commun 237:752–757 Foo SS, Turner CJ, Adams S et al 2006 Ephrin-B2 controls cell motility and adhesion during blood-vessel-wall assembly. Cell 124:161–173 Gerhardt H, Betsholtz C 2003 Endothelial-pericyte interactions in angiogenesis. Cell Tissue Res 314:15–23 Gossen M, Bujard H 1992 Tight control of gene expression in mammalian cells by tetracyclineresponsive promoters. Proc Natl Acad Sci USA 89:5547–5551 Grunwald IC, Korte M, Adelmann G et al 2004 Hippocampal plasticity requires postsynaptic ephrinBs. Nat Neurosci 7:33–40 Gustafsson E, Brakebusch C, Hietanen K, Fassler R 2001 Tie-1-directed expression of Cre recombinase in endothelial cells of embryoid bodies and transgenic mice. J Cell Sci 114:671–676 Jain RK 2003 Molecular regulation of vessel maturation. Nat Med 9:685–693 Jones EA, Baron MH, Fraser SE, Dickinson ME 2004 Measuring hemodynamic changes during mammalian development. Am J Physiol Heart Circ Physiol 287:H1561–1569 Kidd KR, Weinstein BM 2003 Fishing for novel angiogenic therapies. Br J Pharmacol 140:585–594 Kisanuki YY, Hammer RE, Miyazaki J et al 2001 Tie2-Cre transgenic mice: a new model for endothelial cell-lineage analysis in vivo. Dev Biol 230:230–242 Kistner A, Gossen M, Zimmermann F et al 1996 Doxycycline-mediated quantitative and tissuespecific control of gene expression in transgenic mice. Proc Natl Acad Sci USA 93: 10933–10938 Kogata N, Arai Y, Pearson JT et al 2006 Cardiac ischemia activates vascular endothelial cadherin promoter in both preexisting vascular cells and bone marrow cells involved in neovascularization. Circ Res 98:897–904 le Noble F, Fleury V, Pries A et al 2005 Control of arterial branching morphogenesis in embryogenesis: go with the flow. Cardiovasc Res 65:619–628 Lepore JJ, Cheng L, Min Lu M et al 2005 High-efficiency somatic mutagenesis in smooth muscle cells and cardiac myocytes in SM22alpha-Cre transgenic mice. Genesis 41:179–184 Licht AH, Raab S, Hofmann U, Breier G 2004 Endothelium-specific Cre recombinase activity in flk-1-Cre transgenic mice. Dev Dyn 229:312–318 Metzger D, Chambon P 2001 Site- and time-specific gene targeting in the mouse. Methods 24:71–80 No D, Yao TP, Evans RM 1996 Ecdysone-inducible gene expression in mammalian cells and transgenic mice. Proc Natl Acad Sci USA 93:3346–3351 Ny A, Autiero M, Carmeliet P 2006 Zebrafish and Xenopus tadpoles: small animal models to study angiogenesis and lymphangiogenesis. Exp Cell Res 312:684–693 Osoegawa K, Tateno M, Woon PY et al 2000 Bacterial artificial chromosome libraries for mouse sequencing and functional analysis. Genome Res 10:116–128 Ribatti D, Nico B, Vacca A et al 2001 Chorioallantoic membrane capillary bed: a useful target for studying angiogenesis and anti-angiogenesis in vivo. Anat Rec 264:317–324 Rodriguez CI, Buchholz F, Galloway J et al 2000 High-efficiency deleter mice show that FLPe is an alternative to Cre-loxP. Nat Genet 25:139–140

GENETIC APPROACHES FOR STUDYING ANGIOGENESIS

171

Ryding AD, Sharp MG, Mullins JJ 2001 Conditional transgenic technologies. J Endocrinol 171:1–14 Wang HU, Chen ZF, Anderson DJ 1998 Molecular distinction and angiogenic interaction between embryonic arteries and veins revealed by ephrin-B2 and its receptor Eph-B4. Cell 93:741–753 Weinstein BM 2002 Plumbing the mysteries of vascular development using the zebrafish. Semin Cell Dev Biol 13:515–522 West J, Fagan K, Steudel W et al 2004 Pulmonary hypertension in transgenic mice expressing a dominant-negative BMPRII gene in smooth muscle. Circ Res 94:1109–1114

DISCUSSION Kitajewski: Ralf, I would like to explore the connection between Notch and Ephrin B in your model. In some ways, your findings are a bit different from the model proposed by Anne and George, where Notch inhibition promotes sprouting and fi lopodia extension. You are reporting Ephrin B2 activation promotes this process, fi lopodia extension, and Ephrin B2 can be regulated positively by Notch. Thus, is there disconnect between Notch’s role in tip-suppression and Notch acting upstream of Ephrin B2 in the tip cell. How do you resolve this? Adams: We have Brant Weinstein here who has done a lot of the work on the Notch connection with Ephrin B2. From what I know, in various models Notch signalling leads to increased Ephrin B2 expression. How this connects with the Dll4 phenotype is difficult to say. We need to do more work. Working on the assumption that Notch signalling induces Ephrin B2, this probably allows adjacent endothelial cells in the vessel wall to modulate their interactions and thereby promote outgrowth. Epstein: I have a technical question. Adult endothelial inducible Cre has been tough for a number of people. Which ones work the best? Adams: Early in our work, it turned out that there was no suitable line but many failed attempts. I am confident that our VE-cadherin-CreERT2 line works very well both in embryos and in adults. There is a second good line, which was generated by Marcus Fruttiger at UCL here in London using the PDGF-B gene to drive expression of inducible Cre. We have also worked with this. Both are probably very useful. Epstein: In non-dividing large vessels, does Cre in the endothelium get activated? Can you delete other genes? Adams: With the VE-cadherin-CreERT2 line in adults, we are limited to microvessels. This includes quiescent vasculature beds, not just actively angiogenic situations. I don’t have the information for the PDGF-B-iCERT line, which, however, works well in embryos. Epstein: What about your Bmx-CreERT2 line? Adams: That works well in adult arteries of all calibre. This allows us to do genetic manipulations in arteries without disrupting microvessels that are

172

ADAMS

particularly sensitive. This line should also be useful for disease modes such as atherosclerosis. Dejana: There are papers (Gory et al 1999) showing that the VE-cadherin promoter is acting in all types of vessels. In addition Luisa Iruela Arispe showed at the Gordon Conference on Angiogenesis and Microcirculation that the VEcadherin–Cre construct is expressed in large and small vessels. Different results likely depend on the type of VE-cadherin promoter construct used. Adams: I remember that Cre activity of that line is rather patchy. Weinstein: How efficient is your Cre? Adams: It is a high percentage, if not complete. Of course, we haven’t looked at all tissues, but if we just follow the blood vessels we don’t have the impression that there is patchy expression. Unfortunately, in the vascular field we have lots of examples of Cre lines with patchy expression. We have tried many lines generated by others, which don’t work as efficiently. Breier: Have you tested your mice in pathological conditions? Adams: Not yet. Mäkinen: When you overexpress Ephrin B2 throughout the vasculature, do you see differences in the arteries versus veins? Adams: That’s an interesting question. I mentioned that we have this massive signalling through Ephrin B4, but the defects that we see are occurring on both the arterial and venous sides. The strongest defects are restricted to really small vessels. The phenotypes we see are not necessarily dependent on Ephrin B4 expression. On the small arteries especially, we would think that even if there is a small amount of Ephrin B4 protein, this should be saturated by the large amount of endogenous Ephrin B2 that is present. This makes me think that there might be cell autonomous functions. There is already good evidence for this from in vitro experiments. Lammert: Would you expect the same results with Ephrin B2-expressing endothelial cells if you were to make a transgenic mouse expressing a truncated form of Ephrin B2? This way you could discriminate between forward and reverse signalling. Adams: A couple of years ago, I made this mouse and it has caused some controversy. The mouse was a direct knock-in and is early embryonic lethal. In vitro we can induce cell responses by overexpressing Ephrin B2 and microinjected Ephrin B2 constructs in endothelial cells. If we use truncated Ephrin B2, this doesn’t happen, which suggests that the cytoplasmic domain is required. In our knock-in mouse and in these experiments we have checked that the truncated protein is on the surface and can be recognized by Ephrin B receptors. The controversy was caused by another publication where the truncated Ephrin B2 is trapped inside for technical reasons. Lammert: You can phenocopy the pericyte/smooth muscle phenotype of the Ephrin B2 knockout with integrin β1 knockout. Is Ephrin B2 upstream of integrin,

GENETIC APPROACHES FOR STUDYING ANGIOGENESIS

173

and is it true that if you have Ephrin B2 signalling, you activate integrin β1 signalling, and this allows for better binding? Adams: I don’t think it’s a phenocopy. But it’s an interesting question because there is clearly a link between ephrin function and adhesion, which I think involves integrins. Betsholtz: It is a quite severe phenotype that you observe both in the endothelialspecific knockouts and in mice overexpressing Ephrin B2. Connecting back to previous discussions on secondary effects of VEGF up-regulation, how much of the phenotypic alterations could be secondary? What is distinctively primary? Adams: That’s a fair question. The inducible methods have the limitation that it is not easy to look at the very early stages of the induction. I only have indirect evidence to answer your question. By comparing the effect of various mutants that affect vascular morphogenesis and tip cell phenotypes, we have found out that the phenotypes are very distinctive suggesting that they are not just caused by generally compromised angiogenesis and poor tissue perfusion. Betsholtz: Are the phenotypes that you observe VEGF independent? Can you test this with VEGF Trap? Adams: In the embryo it would be difficult to do this. Perhaps we could test this in explant systems. Wilting: It was obvious that Ephrin B2 overexpression in the embryo induces massive oedema. The haemorrhage in the skin may have been due to bleeding into the lymphatics. Adams: We have looked at a large number of overexpression and loss-of-function mutants. The extent of the oedema varies a bit. The phenotype is sometimes more severe and sometimes less severe. This may have to do with the exact onset and level of gene overexpression. The defects that we see are not secondary to oedema in the skin, for instance. As far as lymphatic function is concerned, I think Ephrin B2 is playing a cell autonomous function in the lymphatic endothelial cells. Wilting: In the Jagged 1 knockout you see defects in lung morphogenesis. From what you showed, it seemed that the lung was much more mature. Adams: During normal alveogenesis, the air spaces get smaller and smaller, so the surface of the lung increases dramatically. Having larger cavities is actually a developmental defect. Wilting: For the function of the lung, the interalveolar septa are extremely important. In the maturing lung they become thinner and thinner. This was my impression in the knockout. Adams: This isn’t the case. Reference Gory S, Vernet M, Laurent M, Dejana E, Dalmon J, Huber P 1999 The vascular endothelialcadherin promoter directs endothelial-specific expression in transgenic mice. Blood 93: 184–192

Molecular control of vascular smooth muscle cell differentiation and phenotypic plasticity Gary K. Owens University of Virginia School of Medicine, Charlottesville, VA 22908, USA

Abstract. Although the primary role of vascular smooth muscle cells (SMCs) is contraction, they exhibit extensive phenotypic diversity and plasticity during normal development, during repair of vascular injury, and in disease states including arteriosclerosis and tumour angiogenesis. Results of recent studies indicate that there are unique as well as common transcriptional regulatory mechanisms that control expression of various SMC marker genes within vascular SMC subtypes, and that these mechanisms are complex and dynamic even at the single cell level. This chapter will review recent progress in our understanding of the complex processes, environmental cues, and genes that control development of vascular SMCs from embryonic stem cells, as well as mechanisms that contribute to phenotypic switching of SMCs following vascular injury or in disease states. A major focus will be to summarize recent studies in our laboratory and others showing the importance of CArG-SRF-myocardin-dependent mechanisms and epigenetic controls in regulation of vascular SMC lineage. Of major interest, we have shown that SMC precursor cells acquire a unique pattern of epigenetic changes (i.e. chromatype) during early development that distinguish them from other cell lineages, and makes them permissive for activation of cell selective genes required for their specialized function. In addition, we show that phenotypic switching of SMCs in response to PDGF BB in vitro, or vascular injury in vivo is associated with loss of a subset of activating histone modifications at gene loci encoding SMC marker genes, but retention of additional markers such as H3K4 methylation. We postulate that the latter epigenetic changes may provide a mechanism for ‘cell lineage memory’ during reversible phenotypic switching of vascular SMCs. 2007 Vascular development. Wiley, Chichester (Novartis Foundation Symposium 283) p 174–193

Phenotypic modulation/switching of the smooth muscle cell plays a key role in a number of major diseases in human The vascular smooth muscle cell (SMC) in mature animals is a highly specialized cell whose principal function is contraction and regulation of blood vessel tone/ diameter, blood pressure, and blood flow distribution (reviewed in Owens et al 2004). SMCs within adult blood vessels proliferate at an extremely low rate, exhibit 174

VASCULAR SMOOTH MUSCLE CELLS

175

very low synthetic activity, and express a unique repertoire of contractile proteins, ion channels, and signalling molecules required for the cell’s unique contractile function (Owens 1995, Somlyo & Somlyo et al 2003). Unlike either skeletal or cardiac muscle that are terminally differentiated, SMCs within adult animals retain remarkable plasticity and can undergo profound and reversible changes in phenotype in response to changes in local environmental cues that normally regulate phenotype (Owens 1995). Striking examples of SMC plasticity can be seen during vascular development when the SMC plays a key role in morphogenesis of the blood vessel and exhibits high rates of proliferation, migration, and production of extracellular matrix components such as collagen, elastin and proteoglycans that make up a major portion of the blood vessel wall, while at the same time acquiring contractile capabilities. Similarly, in response to vascular injury, the SMC dramatically increases its rate of cell proliferation, migration, and synthetic capacity and plays a critical role in vascular repair. Indeed, the extensive plasticity exhibited by the fully mature SMC is an inherent property of the differentiated SMC that likely evolved in higher organisms because it conferred a survival advantage. That is, mutations that compromised the ability of the SMC to participate in vascular repair were likely detrimental to the organism and did not persist. However, an unfortunate consequence of this high degree of plasticity is that it predisposes the SMC to abnormal environmental cues/signals that can lead to adverse phenotypic switching and acquisition of characteristics that can contribute to development and/or progression of vascular disease. Indeed there is strong evidence that phenotypic switching of the SMC plays a critical role in a large number of major diseases in human including atherosclerosis, asthma, cancer and hypertension (Owens et al 2004). The focus of this chapter will be to summarize our current understanding of molecular mechanisms that control SMC differentiation during vascular development as well as in response to vascular injury or disease. Expression of SMC selective differentiation marker genes is dependent on integration of a multitude of complex local environmental cues The model that has evolved to explain SMC differentiation is that it is highly dependent on the cell integrating complex local environmental cues/signals including mechanical forces, neuronal influences, extracellular matrix components, and various soluble cytokines and growth factors (reviewed in Hungerford & Little 1999, Owen 1995, Owens et al 2004) that influence expression of the repertoire of genes that determine SMC phenotype. That is, SMCs constantly integrate the complex signals present in their local environment, and these in aggregate determine the appropriate patterns of gene expression. This model is similar to that which is believed to control differentiation of virtually all cell types. However, the SMC is somewhat unique in the extent of plasticity that it must exhibit even in

176

OWENS

adult organisms as compared to many other cell types. Key challenges in understanding control of SMC differentiation are: (1) to identify the critical environmental cues/signals that influence differentiation and maturation of the SMC; and (2) to determine the molecular mechanisms and signalling pathways whereby these environmental cues influence expression of genes characteristic of the SMC lineage. An initial step in elucidating molecular controls of SMC differentiation was to identify appropriate marker genes that encode for proteins that are selective or specific for SMCs and that are required for its differentiated functions. Indeed, there has been remarkable progress in this area over the past decade, and we now have a relatively large repertoire of SMC-selective or specific marker genes with which to study SMC differentiation and maturation (reviewed in Owens et al 2004). Appropriate markers include a variety of contractile proteins important for the differentiated function of the SMCs including smooth muscle (SM) α -actin (Gabbiani et al 1981, Owens & Thompson 1986), SM myosin heavy chains (Miano et al 1994, Rovner et al 1986a, b) SM myosin light chains (Hasegawa et al 1992) and SM α -tropomyosin (Hansson et al 1986). In addition, differentiated SMCs also express a number of proteins that are part of the cytoskeleton and/or purported to be involved in regulation of contraction such as h-calponin (Winder et al 1991), SM-22 α (Winder et al 1991), h-caldesmon (Sobue et al 1991), β -vinculin (Geiger et al 1980), metavinculin (Geiger et al 1980), telokin (Herring & Smith 1996), smoothelin (van der Loop et al 1997), LPP (Gorenne et al 2003), and desmin (Mericskay et al 2000) that show at least some degree of SMC specificity/selectivity (reviewed in Hungerford & Little 1999, Owens 1995, Owens et al 2004). The preceding list is not exclusive, and there are certainly other marker genes that are likely to have utility for understanding cell specific transcriptional control mechanisms in SMCs, as well as mechanisms that contribute to co-ordinate gene regulation during SMC differentiation and maturation. It needs to be emphasized that virtually all, if not all, of these SMC differentiation marker genes can also be expressed in non-SMCs under some conditions. Moreover, it is clear that rigorous assessment of the differentiated state of the SMC is dependent on examining multiple marker genes including those that are both increased and decreased with differentiation/ maturation and phenotypic switching. A major advance in this field has been the identification of promoter-enhancer regions of a number of the SMC differentiation marker genes that confer SMCspecific/selective expression in vivo in transgenic mice (Kim et al 1997, Li et al 1996b, Mack & Owens 1999, Madsen et al 1998, Mericskay et al 2000). Studies by Li et al (1996b) and Kim et al (1997) were the first to identify sufficient regions of the SM22 α promoter that conferred expression in transgenic mice. Of interest, they showed that a promoter construct containing 441 bps of the SM22 α 5′ promoter region conferred expression in cardiac, skeletal, and arterial SMC in

VASCULAR SMOOTH MUSCLE CELLS

177

transgenic mouse embryos, but not in gastro-intestinal or other SMC tissues that normally express the endogenous SM22 α gene. These results were of interest in that they provided evidence for differential regulation of SMC marker genes in subsets of SMCs, a paradigm now supported by considerable evidence including studies by our group showing differential activity of specific cis elements within the SM MHC promoter in large versus small arteries (Manabe & Owens 2001c). Our laboratory was the first to identify regions of the SM MHC (Madsen et al 1998), and SM α -actin (Mack & Owens 1999) promoters that could drive expression in vivo in transgenic mice in a manner that recapitulated expression of the endogenous genes. Moreover, we have subsequently characterized multiple regulatory modules and cis elements required for SMC specific/selective expression of these genes in transgenic mice (Adam et al 2000, Kumar et al 2003, Liu et al 2003, Mack & Owens 1999, Madsen et al 1998, Manabe & Owens 2001a), including several that mediate responsiveness of these genes, at least in cultured cell systems, to environmental signals believed to be important in the control of SMC differentiation (Adam et al 2000, Hautmann et al 1999, 1997a, b). Of particular relevance to this presentation, we demonstrated that expression of these genes was dependent on several highly conserved CArG or CArG-like elements (i.e. CC[AT] 6GG) (Mack & Owens 1999, Manabe & Owens 2001a) found in the promoters of most SMC differentiation marker genes identified to date (reviewed in Miano 2003, Owens 1995, Owens et al 2004). Surprisingly, despite the fact that cultured SMC lines are highly modulated, and that phenotypic modulation is a critical process in atherogenesis and vascular injury repair, only two specific factors/pathways have been identified that selectively and directly promote phenotypic modulation of the SMCs, i.e. plateletderived growth factor BB (PDGF BB) (Blank & Owens 1990, Corjay et al 1989, Holycross et al 1992, Li et al 1997) and a CTGF or CTGF-like factor we identified that is produced by cultured endothelial cells (Vernon et al 1996). The reason for the paucity of studies in this area is likely due to two factors: (1) the incorrect belief that SMC phenotypic modulation is simply secondary to growth stimulation (i.e. the old adage that differentiation and proliferation are mutually exclusive processes in all cells); and (2) the assumption that phenotypic modulation of the SMC is a passive rather than active process and due simply to loss of positive SMC differentiation factors. It is now well established that differentiation and proliferation are not mutually exclusive, and that many factors other than the SMC’s proliferation status influences its differentiation state. This topic was extensively reviewed in Owens (1995), and we will only briefly summarize several relevant observations here. First, during late embryogenesis and postnatal development, SMCs are known to have an extremely high rate of proliferation (Cook et al 1994), yet at this time they undergo the most rapid rate of induction of expression of multiple SMC differentiation

178

OWENS

marker genes (Owens & Thompson 1986). Second, SMCs within advanced atherosclerotic lesions show a very low rate of proliferation that approaches that of fully differentiated SMCs, yet are highly phenotypically modulated as evidenced by marked reductions in expression of SMC marker genes (O’Brien et al 1993, Wilcox 1992). These results show that cessation of proliferation alone is not sufficient to promote SMC differentiation and suggest that other SMC differentiation cues are absent and/or that active repressors of SMC differentiation are present. Consistent with the hypothesis that phenotypic modulation of the SMC may be actively controlled and is not simply due to loss of positive differentiation signals, we (Blank & Owens 1990, Corjay et al 1989, Holycross et al 1992) and others (Li et al 1997, Somasundaram et al 1995, Thyberg et al 1983) have shown that treatment of post-confluent cultures of rat aortic SMCs with PDGF BB is associated with rapid down-regulation of expression of multiple differentiation marker genes (Fig. 1). Of particular significance, under the conditions of our experiments, we found that PDGF BB elicited only a transient mitogenic response with cell proliferation returning to control values within 36–48 hours despite repeated daily pulsing with PDGF BB, yet suppression of SMC marker gene expression, including SM α -actin and SM MHC, persisted as long as PDGF BB was present. Indeed, we found that cultured SMCs could be sustained in a highly de-differentiated state

FIG. 1. PDGF BB but not thrombin or bFGF (FGF2) markedly decreased fractional synthesis of SM α -actin in cultured SMC (Owens et al 1996). Confluent cultures of rat aortic SMC were treated with concentrations of each growth factor that gave virtually identical growth responses, and the relative rate of SM α -actin versus non-muscle β -and γ -actin synthesis determined by two dimensional gel electrophoresis following pulsing of cells with 35S-methionine. (Adapted from Owens et al 1996.)

VASCULAR SMOOTH MUSCLE CELLS

179

with virtually no detectable expression of SM α -actin indefinitely by sustained treatment with PDGF BB. Of interest, we also showed that the concentration of PDGF BB required for inducing SMC phenotypic modulation was 10-fold lower than that required to elicit a growth response under these experimental conditions. That is, we could induce down-regulation of SM α -actin expression in the absence of cell cycle entry. In contrast, we found that FGF2 and fetal bovine serum (FBS) had little or no effect on SMC differentiation marker gene expression in postconfluent cultures despite eliciting nearly identical proliferative responses, and thrombin-induced proliferation was associated with increased not decreased expression of SMC marker genes (Blank & Owens 1990, Corjay et al 1990, Owens et al 1996) (Fig. 1). Taken together, these results indicate that PDGF BB is a highly efficacious and selective negative regulator of SMC differentiation, and that its effects on differentiation are not simply a function of growth stimulation. We have also completed extensive studies to determine the mechanisms by which PDGF BB can inhibit expression of SMC differentiation marker genes in cultured SMCs. Of interest, at least part of the effect appears to be mediated through post-transcriptional effects in that we found that PDGF BB induced rapid and selective degradation of mRNAs encoding for SMC marker genes (Corjay et al 1990, Holycross et al 1992). However, we (Dandre & Owens 2004, Liu et al 2005) and others (Somasundaram et al 1995) showed that PDGF BB also markedly inhibits transcription of multiple SMC marker genes in cultured SMCs. Of major interest, we recently found evidence that the effects of PDGF BB on SMC gene expression were mediated, at least in part, by KLF4/GKLF in that PDGF BBinduced decreases in SMC gene expression could be significantly inhibited by KLF4 siRNAs or oligonucleotides, and that expression of KLF4 was markedly increased by PDGF BB treatment (Liu et al 2003, 2005, McDonald et al 2006). In addition, we found that: (1) adenoviral-mediated overexpression of KLF4 profoundly decreased expression of myocardin, as well as multiple SMC marker genes (Fig. 2); (2) KLF4 nearly abolished myocardin-induced activation of SMC marker genes; (3) KLF4 and SRF bound to each other in immunoprecipitation assays; and (4) KLF4 expression was absent from normal vessels but markedly increased following vascular injury in vivo. Taken together, these results support a model in which PDGF BB-induced suppression of SMC genes is mediated, at least in part, through KLF4-induced inhibition of myocardin expression, as well as through inhibition of myocardin-SRF dependent gene activation. SMC-specific transcriptional regulation is dependent on complex combinatorial interactions of multiple cis elements (regulatory modules) and their trans binding factors Tremendous progress has been made in the past decade in identifying mechanisms that contribute to transcriptional regulation of SMC marker genes (see reviews by

OWENS

Relative Gene Copy #

3.00

0.50 0.00

2.00

SM α-actin/GAPDH

1.00 0.00

2.00

1.00

CMV EGFP KLF4

Relative Gene Copy #

Myocardin/GAPDH

KLF4/GAPDH

1.00

CMV

EGFP KLF4

0.00

CMV

EGFP KLF4

SM MHC/GAPDH Relative Gene Copy #

Relative Gene Copy #

180

3.00 2.00 1.00 0.00

CMV EGFP KLF4

FIG. 2. Adenovirus induced overexpression of KLF4 dramatically repressed expression of multiple endogenous SMC marker genes including myocardin, SM α -actin, and SM MHC in cultured SMCs as measured by real time RT-PCR (Adapted from Liu et al 2005, 2003).

Kumar & Owens 2003, Owens et al 1996, 2004, Firulli & Olson 1997, Majesky 2003, Miano 2003). Due to space constraints and the large amount of work in this area we cannot discuss these regulatory pathways in detail. However, suffice to say that the model that has emerged is that regulation of SMC selective gene expression is not dependent on any single factor that is completely specific for SMCs but rather is dependent on unique combinatorial interactions of multiple factors that are either ubiquitously expressed, or that may be selective for SMCs. I will consider one example, CArG-SRF-dependent regulation, to illustrate this general model and to provide important background information of relevance to understanding SMC phenotypic switching associated with vascular injury and disease. Site-directed mutagenesis studies in transgenic mice have shown that expression of most SMC marker genes identified to date are dependent on one or more CArG elements (i.e. a CC(AT) 6GG motif ) found within their promoter and/or intronic sequences (Kim et al 1997, Li et al 1996b, Mack & Owens 1999, Manabe & Owens 2001a, Mericskay et al 2000, Yano et al 1995). For example, we demonstrated that the region of the SM α -actin promoter from −2560 to +2784 completely recapitulated expression patterns of the endogenous SM α -actin gene in vivo in transgenic mice. However, expression of this nearly 5500 base pair promoter enhancer was completely abolished by a 2 or 4 bp mutation of any one of three highly conserved CArG elements contained within it (Mack & Owens 1999). Similarly, mutation of conserved CArG elements within the SM MHC (Manabe & Owens 2001a), SM22 α

VASCULAR SMOOTH MUSCLE CELLS

181

(Kim et al 1997, Li et al 1996b), and desmin (Mericskay et al 2000) promoter enhancers also virtually abolished expression in vivo in transgenic mice. CArG elements bind the transcription factor serum response factor (SRF), a MADS (MCM1, Agamous, Deficiens, SRF) box transcription factor, that was fi rst identified and named because of its ability to confer serum inducibility to the growth responsive gene, c-fos, through binding to a sequence known as the serum response element (SRE, or CArG box). SRF binds CArG boxes as a dimer, with dimerization and DNA binding occurring through the MADS box domain (Shore & Sharrocks 1994). In addition to regulating growth responsive genes such as cfos, and multiple SMC marker genes (Mack et al 2000), SRF binding to CArG boxes also regulates numerous skeletal and cardiac muscle specific genes (Du et al 2003, Sartorelli et al 1993). A long-standing question in the field has been to determine how SRF, a ubiquitously expressed transcription factor, can regulate both growth responsive and cell-specific genes in SMCs and muscle and nonmuscle cell types. Several possible mechanisms have been proposed including a number that appear unique to SMC. Importantly, these mechanisms are not mutually exclusive, and it is likely that SMC selectivity is the result of some combination of the following mechanisms and/or others yet to be discovered (Kumar & Owens 2003). These mechanisms are reviewed in detail in several of our recent reviews (Owens et al 2004, Yoshida & Owens 2005) as well as by Miano (2003), and include: (a) alterations in the level of SRF expression (Belaguli et al 1997, Croissant et al 1996); (b) regulation of SRF binding affinity to the degenerate CArG elements found in many SMC promoters by homeodomain factors such as Prx1/Mhox (Hautmann et al 1997b, Yoshida et al 2004); (c) co-operative interactions between the multiple CArG elements found in most SMC marker genes but not in ubiquitously expressed genes such as c-fos that contain a single CArG (Kim et al 1997, Li et al 1996a, Mack & Owens 1999, Manabe & Owens 2001a, Mericskay et al 2000, Miano et al 2000, Yano et al 1995); (d) post-transcriptional modifications of SRF or SRF co-factors (Manak & Prywes 1993, Rivera et al 2003, Wang et al 2004); (e) co-operative interaction with other cis regulatory elements and their binding factors including GATA6, MEF2, and Nkx3.2 (Nakamura et al 2001, Nishida et al 2002); (f) regulation of the ability of SRF to bind to CArG elements though modulation of chromatin structure (Manabe & Owens 2001b) (Fig. 6); and (g) interaction with SMC-specific/selective SRF co-activators including myocardin or the myocardin-like factors MKL1 or MKL2 (Chen et al 2002, Du et al 2004, Wang et al 2001, 2003, Yoshida et al 2003). One of the most significant and exciting advances for the field of SMC differentiation in recent years was the discovery of myocardin by Olson and co-workers (Wang et al 2001). Myocardin is an extremely potent SRF co-activator that is exclusively expressed in cardiac (Wang et al 2001) and differentiated SMC in vivo (Chen et al 2002, Du et al 2003). Moreover, mouse embryos homozygous for a

182

OWENS

myocardin loss-of-function mutation die by embryonic day 10.5 and show no evidence of vascular SMC differentiation based on a detailed in situ analysis that revealed lack of SMC-selective gene expression (Li et al 2003) although a confounding aspect of this study was that authors also found gross abnormalities in yolk sac vasculogenesis suggesting that effects on SMC development could have been indirect. Indeed, in collaboration with E. Olson, we recently showed that cell autonomous myocardin is not absolutely required for at least early stages of SMC development in that lineage-tagged myocardin null embryonic stem cells (ESC) were capable of forming SMC in the context of chimeric knockout mice (Pipes et al 2005). Nevertheless, myocardin potently and selectively induces expression of all CArG-dependent SMC marker genes tested to date including SM α -actin, SM MHC, SM22 α , and calponin in cultured SMC and embryonic fibroblasts in culture (Chen et al 2002, Du et al 2003, Wang et al 2001, Yoshida et al 2003). Of interest, myocardin appears to be most efficacious in activating those genes that contain multiple CArG elements, and Olson and colleagues have presented evidence for a model whereby the leucine zipper motif of myocardin may bridge adjacent CArG elements and unmask myocardin’s activation domain (Wang et al 2003). This model is intriguing and may provide a mechanism to explain the effects we observed in varying the phasing and spacing of the 5′ SM α -actin CArG elements in our previous studies (Mack et al 2000). We (Yoshida et al 2003) and others (Wang et al 2003) have shown that adenoviral-mediated overexpression of myocardin was sufficient to activate multiple CArG-dependent SMC differentiation markers in ESCs, or embryonic fibroblast systems. Significantly, we showed that administration of a myocardin siRNA (Du et al 2003, Yoshida et la 2003), or adenovirus-mediated overexpression of a myocardin dominant negative construct (Yoshida et al 2003) significantly reduced expression of multiple SMC marker genes including SM MHC, SM22 α , and SM α -actin by up to 80% in cultured SMCs, thus providing direct evidence that endogenous myocardin plays a key role in regulation of expression of multiple SMC marker genes. As yet, little is known regarding role or mechanisms by which myocardin mediates responses of SMC to environmental cues that either activate or repress SMC gene expression in vivo. However, we have evidence showing that it plays a key role in both AII-induced increases (Yoshida et al 2004), as well as PDGF BB-induced decreases (Liu et al 2003, 2005, McDonald et al 2006) in expression of SMC marker genes. In addition, studies by Wang et al (2004) demonstrated that PDGF BB-induced suppression of SMC genes also involved phosphorylation of ELK1, which then competed with myocardin for CArG-SRF binding, findings consistent with results of our studies showing that regulation of SRF binding to degenerate CArG elements is critical for injury-induced suppression of SM α -actin in vivo (Hendrix et al 2005) and suggesting that KLF4 and pELK may act co-operatively to suppress expression of SMC marker genes.

VASCULAR SMOOTH MUSCLE CELLS

183

Evidence that epigenetic controls contribute to SMC lineage determination We found that myocardin was expressed in a unique SMC ‘precursor’ line, designated A404 previously described by our laboratory (Manabe & Owens 2001b), in the absence of detectable expression of any other known SMC marker including the earliest known markers SM α -actin and SM22 α (Yoshida et al 2003). In contrast, myocardin expression was absent from P19 embryonal carcinoma stem cells, the parental line from which A404 cells were derived. Treatment of A404 cells with all-trans retinoic acid (RA), which induced expression of all known SMC marker genes (Manabe & Owens 2001b), was associated with further increases in myocardin expression. In addition, we showed that although SRF was highly expressed in A404 cells, it was unable to bind to the CArG containing regions of SMC genes within intact chromatin (Fig. 3), although it did bind to the constitutively expressed c-fos CArG promoter region. Treatment of A404 cells with RA

FIG. 3. Results of CHIP assays showing increased binding of SRF to CArG-containing regions of SMC promoters in differentiated versus undifferentiated A404 cells within intact chromatin (Manabe et al 2001b).

184

OWENS

resulted in association of SRF with CArG containing regions of SMC promoters, as well as hyperacetylation of histones associated with these regions. Taken together, these results support a model in which myocardin and SRF are expressed in SMC progenitor cells but are unable to bind to CArG containing regions of SMC genes because of spatial restrictions associated with chromatin structure that are selective for SMC promoter regions (Figs. 4–7). However, evidence suggests that upon treatment with RA the chromatin organization within

FIG. 4. Results of quantitative CHIP assays showing marked enrichment of SRF binding to the SM α -actin and SM MHC but not the 5′ c-fos CArG containing regions in cultured SMC versus endothelial cells (EC) within intact chromatin. (Reprinted with permission from McDonald et al 2006.)

EC Silent

3.9 3.4 2.9 2.4 1.9 1.4 0.9 0.4

ESC Silent

SMC Active

EC Silent

25 20 15 10 5 0

ESC Silent

SMC Silent

EC Active

2.9 2.4 1.9 1.4 0.9 0.4

ESC Silent

SMC Silent

EC Active

SMC Silent

EC Silent

25

F o ld E n ric h m e n t

F o ld E n ric h m e n t

ESC Silent

F o ld E n ric h m e n t

SMC Active

EC Active

25 20 15 10 5 0

25 20 15 10 5 0

20 15 10 5 0

ESC Silent

SMC Silent

ESC Active

SMC Active

EC Active

ESC Active

SMC Active

EC Active

ESC Active

SMC Active

EC Active

30 25 20 15 10 5 0

EC Silent F o ld E n ric h m e n t

ESC Silent

SMC Silent

F o ld E n ric h m e n t

25 20 15 10 5 0

ESC Silent

F o ld E n ric h m e n t

EC Silent

F o ld E n ric h m e n t

SMC Active

F o ld E n ric h m e n t

H4 Lys20 Di-methyl

ESC Silent

F o ld E n ric h m e n t

H4 Acetyl

F o ld E n ric h m e n t

H3 Lys4 Di-methyl

c-fos

CM a-MHC

VEC 25 20 15 10 5 0

F o ld E n ric h m e n t

F o ld E n ric h m e n t

SM a-actin 30 24 18 12 6 0

2.9

2.9 2.4

2.4

1.9

1.9

1.4

1.4

0.9

0.9

0.4

ESC Silent

SMC Silent

EC Silent

0.4

FIG. 5. SMCs show unique epigenetic patterning as compared to embryonic stem cells (ESCs) or endothelial cells (ECs). For example, the activating histone modifications of H3K4 dimethylation, and H4 acetylation are enriched at the SM α -actin gene locus in SMCs as compared to ESCs as well as ECs and other differentiated somatic cell types (data not shown). In contrast, the silencing modification H4K20 dimethylation is enriched in ECs as compared to SMCs. (Reprinted with permission from McDonald et al 2006.)

VASCULAR SMOOTH MUSCLE CELLS

mRNA:

Transcript/18S

SM a-actin 1.600 1.200 0.800 0.400

H3 Lys4 DiMethylation:

20.00

0.60

5.000 4.000 3.000 2.000 1.000 0.000 4.50 4.00 3.50 3.00 2.50 2.00 1.50 1.00 0.50 0.00

15.00

0.40 10.00

Absolute Enrichment

Acetylation: H4 Acetylation

0.80

5.00

0.20

0.00

0.00

16.00

12.00

12.00

9.00

8.00

6.00

4.00

3.00

0.00

18.00 15.00 12.00 9.00 6.00 3.00 0.00

c-fos

SM-MHC 0.490 0.420 0.350 0.280 0.210 0.140 0.070 0.000

25.00

Abso lute En richm en t

SRF Binding:

Absolute Enrichment

0.000

185

7.00 6.00 5.00 4.00 3.00 2.00 1.00 0.00

0.00

30.00 24.00 18.00 12.00 6.00 0.00

12.00 9.00 6.00 3.00 0.00

24 hr. vehicle

24 hr. PDGF-BB

72 hr. PDGF-BB

24 hr. PDGF-BB replaced with vehicle for 48 hr.

A Potential Mechanism for Cell Lineage Memory During Reversible Phenotypic Switching FIG. 6. PDGF BB treatment of cultured SMCs was associated with reduced expression of SMC marker genes such as SM α -actin and SM myosin heavy chain (SM MHC), as well as reduced SRF binding and histone acetylation of CArG-containing regions of SMC promoters within intact chromatin. However, there was no change in H3K4 dimethylation at these SMC marker gene loci, a unique epigenetic marker for SMC lineages. (Reprinted with permission from McDonald et al 2006.)

SMC promoter regions is relaxed at least in part by histone acetylation, SRF then binds, recruits myocardin and other possible co-activators, and activates expression of multiple CArG-dependent SMC genes. Consistent with these results Qui & Li (2002) presented evidence that CREB-CArG-dependent expression of the SM22 gene in cultured SMC was dependent on histone acetyltransferase activity (HAT). In brief, they found that treatment of cells with trichostatin A, a histone deacetylase (HDAC) inhibitor, increased whereas overexpression of HDACs decreased SM22 promoter activity. In addition, Cao et al (2005) provided evidence that myocardin can bind both HATs and HDACs and induce acetylation of histones surrounding CArG-containing regions of SMC marker genes. Of major interest, we (McDonald et al 2006) recently presented evidence using novel quantitative CHIP assays developed in our lab, to show that development of SMC lineages from ESCs was dependent on acquisition of SMC-selective changes in chromatin structure that were restricted to gene loci encoding SMC marker genes

186

OWENS SM a-actin promoter Undifferentiated A404 cells

c-fos promoter Undifferentiated A404 cells Ac

SA

M

SRF

M

SRF SRF

SRF

Myocardin

SR

M

SR CArG

Ac

Ac

RA

RA Ac

SA

Ac

SR

M

M

M M

SR CArG

Ac c-fos promoter Differentiated A404 cells

Ac

M

Ac

M

H3K4dM S R

C S Ar R G

Myocardin

SRF

Ac

SRFCArG

SM a -actin promoter Differentiated A404 cells

FIG. 7. Model depicting how alterations in chromatin structure may contribute to activation of CArG/SRF-dependent SMC marker genes upon retinoic acid (RA)-induced differentiation of A404 SMC progenitor cells.

(Figs. 4–5). In addition, we showed that phenotypic switching of SMC in response to PDGF BB treatment in cultured SMC or in response to vascular injury in vivo, was associated with reduced SRF binding to CArG regions of SMC marker genes within intact chromatin, as well as reduced histone acetylation (Fig. 6), an epigenetic change associated with chromatin condensation and transcriptional repression. However, a subset of epigenetic changes, including H3K4 dimethylation appeared to be unchanged by PDGF BB in vitro, or vascular injury in vivo suggesting that this epigenetic mark may contribute to ‘cell lineage memory’. Taken together, the preceding results provide strong evidence that regulation of chromatin structure or ‘epigenetic programming’ (Rice & Allis 2001) plays a key role not only in control of SMC-specific gene expression in fully differentiated SMC but may also contribute to retention of SMC lineage identity during reversible phenotypic switching (Figs. 6–7). In summary, there is extensive evidence showing that vascular SMCs are highly plastic and undergo profound changes in phenotype in response to changes in environmental cues. However, the precise factors and mechanisms that regulate both normal and abnormal differentiation of vascular SMCs in vivo are at present poorly understood. Of particular importance to the vascular development theme of this Novartis Foundation Symposium, further studies are needed to define the

VASCULAR SMOOTH MUSCLE CELLS

187

mechanisms and factors that control investment and subsequent differentiation of SMCs and pericytes during vasculogenesis/arteriogenesis, as well as how this process can be dysregulated in disease states, including defective mural cell investment, arteriogenesis and vascular patterning within solid tumours. Acknowledgements This work was supported by grants R01 HL38854, R37 HL57353, and P01 HL19242 from the National Institutes of Health USA.

References Adam PJ, Regan CR, Hautmann MB, Owens GK 2000 Positive and negative acting krupplelike transcription factors bind a transforming growth factor beta control element required for expression of the smooth muscle differentiation marker SM22alpha in vivo. J Biol Chem 275:37798–37806 Belaguli NS, Schildmeyer LA, Schwartz RJ 1997 Organization and myogenic restricted expression of the murine serum response factor gene. A role for autoregulation. J Biol Chem 272:18222–18231 Blank RS, Owens GK 1990 Platelet-derived growth factor regulates actin isoform expression and growth state in cultured rat aortic smooth muscle cells. J Cell Physiol 142:635–642 Cao D, Wang Z, Zhang CL et al 2005 Modulation of smooth muscle gene expression by association of histone acetyltransferases and deacetylases with myocardin. Mol Cell Biol 25: 364–376 Chen J, Kitchen CM, Streb JW, Miano JM 2002 Myocardin: a component of a molecular switch for smooth muscle differentiation. J Mol Cell Cardiol 34:1345 Cook CL, Weiser MC, Schwartz PE, Jones CL, Majack RA 1994 Developmentally timed expression of an embryonic growth phenotype in vascular smooth muscle cells. Circ Res 74: 189–196 Corjay MH, Thompson MM, Lynch KR, Owens GK 1989 Differential effect of platelet-derived growth factor- versus serum-induced growth on smooth muscle alpha-actin and nonmuscle beta-actin mRNA expression in cultured rat aortic smooth muscle cells. J Biol Chem 264: 10501–10506 Corjay MH, Blank RS, Owens GK 1990 Platelet-derived growth factor-induced destabilization of smooth muscle a-actin mRNA. J Cell Physiol 145:391–397 Croissant JD, Kim JH, Eichele G et al 1996 Avian serum response factor expression restricted primarily to muscle cell lineages is required for [alpha]-actin gene transcription. Dev Biol 177:250–264 Dandre F, Owens GK 2004 Platelet-derived growth factor-BB and Ets-1 transcription factor negatively regulate transcription of multiple smooth muscle cell differentiation marker genes. Am J Physiol Heart Circ Physiol 286:H2042–2051 Du KL, Chen M, Li J, Lepore JJ, Mericko P, Parmacek MS 2004 Megakaryoblastic leukemia factor-1 transduces cytoskeletal signals and induces smooth muscle cell differentiation from undifferentiated embryonic stem cells. J Biol Chem 279:17578–17586 Du KL, Ip HS, Li J et al 2003 Myocardin is a critical serum response factor cofactor in the transcriptional program regulating smooth muscle cell differentiation. Mol Cell Biol 23: 2425–2437 Firulli AB, Olson EN 1997 Modular regulation of muscle gene transcription: a mechanism for muscle cell diversity. Trends Genet 13:364–369

188

OWENS

Gabbiani G, Schmid E, Winter S et al 1981 Vascular smooth muscle cells differ from other smooth muscle cells: predominance of vimentin fi laments and a specific-type actin. Proc Natl Acad Sci USA 78:298–300 Geiger B, Tokuyasu KT, Dutton AH, Singer SJ 1980 Vinculin, an intracellular protein localized at specialized sites where microfi lament bundles terminate at cell membranes. Proc Natl Acad Sci USA 77:4127–4131 Gorenne I, Nakamoto RK, Phelps CP, Beckerle MC, Somlyo AV, Somlyo AP 2003 LPP, a LIM protein highly expressed in smooth muscle. Am J Physiol Cell Physiol 285:C674–685 Hansson GK, Jonasson L, Holm J, Claesson-Welsh L 1986 Class II MHC antigen expression in the atherosclerotic plaque: smooth muscle cells express HLA-DR, HLA-DQ and the invariant gamma chain. Clin Exp Immunol 64:261–268 Hasegawa Y, Ueda Y, Watanabe M, Morita F 1992 Studies on amino acid sequences of two isoforms of 17-kDa essential light chain of smooth muscle myosin from porcine aorta media. J Biochem 111:798–803 Hautmann M, Madsen CS, Owens GK 1997a A transforming growth factor beta (TGF) control element drives TGF-induced stimulation of SM alpha-actin gene expression in concert with two CArG elements. J Biol Chem 272:10948–10956 Hautmann M, Thompson MM, Swartz EA, Olson EN, Owens GK 1997b Angiotensin IIinduced stimulation of smooth muscle alpha-actin expression by serum response factor and the homeodomain transcription factor MHox. Circ Res 81:600–610 Hautmann M, Adam PJ, Owens GK 1999 Similarities and differences in smooth muscle alphaactin induction by transforming growth factor beta in smooth muscle versus non-muscle cells. Arterioscler Thromb Vasc Biol 19:2049–2058 Hendrix J, Wamhoff BR, McDonald T, Sinha S, Yoshida T, Owens GK 2005 5′ CArG degeneracy in smooth muscle alpha-actin is required for injury-induced gene suppression in vivo. J Clin Invest 115:418–427 Herring BP, Smith AF 1996 Telokin expression is mediated by a smooth muscle cell-specific promoter. Am J Physiol 270:C1656–1665 Holycross BJ, Blank RS, Thompson MM, Peach MJ, Owens GK 1992 Platelet-derived growth factor-BB-induced suppression of smooth muscle cell differentiation. Circ Res 71:1525–1532 Hungerford JE, Little CD 1999 Developmental biology of the vascular smooth muscle cell: building a multilayered vessel wall. J Vasc Res 36:2–27 Kim S, Ip HS, Lu MM, Clendenin C, Parmacek MS 1997 A serum response factor-dependent transcriptional regulatory program identifies distinct smooth muscle cell sublineages. Mol Cell Biol 17:2266–2278 Kumar MS, Hendrix J, Johnson AD, Owens GK 2003 The smooth muscle alpha-actin gene requires two Eboxes for proper expression in vivo and is a target of class I basic HLH proteins. Circ Res 92:840–847 Kumar MS, Owens GK 2003 Combinatorial control of smooth muscle specific gene expression. Arterioscler Thromb Vasc Biol 23:737–747 Li L, Miano JM, Cserjesi P, Olson EN 1996a SM22 alpha, a marker of adult smooth muscle, is expressed in multiple myogenic lineages during embryogenesis. Circ Res 78:188–195 Li L, Miano JM, Mercer B, Olson EN 1996b Expression of the SM22alpha promoter in transgenic mice provides evidence for distinct transcriptional regulatory programs in vascular and visceral smooth muscle cells. J Cell Biol 132:849–859 Li S, Wang DZ, Wang Z, Richardson JA, Olson EN 2003 The serum response factor coactivator myocardin is required for vascular smooth muscle development. Proc Natl Acad Sci USA 100:9366–9370 Li X, Van Putten V, Zarinetchi F et al 1997 Suppression of smooth muscle alpha-actin expression by platelet-derived growth factor in vascular smooth-muscle cells involves Ras and cytosolic phospholipase A2. Biochem J 327:709–716

VASCULAR SMOOTH MUSCLE CELLS

189

Liu Y, Sinha S, Owens GK 2003 A transforming growth factor-beta control element required for SM alpha-actin expression in vivo also partially mediates GKLF-dependent transcriptional repression. J Biol Chem 278:48004–48011 Liu Y, Sinha S, McDonald OG, Shang Y, Hoofnagle MH, Owens GK 2005 Kruppel-like factor 4 abrogates myocardin-induced activation of smooth muscle gene expression. J Biol Chem 280:9719–9727 Mack CP, Owens GK 1999 Regulation of SM alpha-actin expression in vivo is dependent upon CArG elements within the 5′ and fi rst intron promoter regions. Circ Res 84:852–861 Mack CP, Thompson MM, Lawrenz-Smith S, Owens GK 2000 Smooth muscle alpha-actin CArG elements coordinate formation of a smooth muscle cell-selective, serum response factor-containing activation complex. Circ Res 86:221–232 Madsen CS, Regan CP, Hungerford JE, White SL, Manabe I, Owens GK 1998 Smooth musclespecific expression of the smooth muscle myosin heavy chain gene in transgenic mice requires 5′-flanking and fi rst intronic DNA sequence. Circ Res 82:908–917 Majesky MW 2003 Decisions, decisions. SRF coactivators and smooth muscle myogenesis. Circ Res 92:824–826 Manabe I, Owens GK 2001a CArG elements control smooth muscle subtype-specific expression of smooth muscle myosin in vivo. J Clin Invest 107:823–834 Manabe I, Owens GK 2001b Recruitment of serum response factor and hyperacetylation of histones at smooth muscle-specific regulatory regions during differentiation of a novel P19derived in vitro smooth muscle differentiation system. Circ Res 88:1127–1134 Manabe I, Owens GK 2001c The smooth muscle myosin heavy chain gene exhibits smooth muscle subtype selective modular regulation in vivo. J Biol Chem 276:39076–39087 Manak JR, Prywes R 1993 Phosphorylation of serum response factor by casein kinase II: evidence against a role in growth factor regulation of fos expression. Oncogene 8:703–711 McDonald OG, Wamhoff BR, Hoofnagle MH, Owens GK 2006 Control of SRF binding to CArG-box chromatin regulates smooth muscle gene expression in vivo. J Clin Invest 116:36–48 Mericskay M, Parlakian A, Porteu A et al 2000 An overlapping CArG/Octamer element is required for regulation of desmin gene transcription in arterial smooth muscle cells. Dev Biol 226:192–208 Miano JM 2003 Serum response factor: toggling between disparate programs of gene expression. Mol Cell Cardiol 35:577–593 Miano JM, Cserjesi P, Ligon K, Perisamy M, Olson EN 1994 Smooth muscle myosin heavy chain marks exclusively the smooth muscle lineage during mouse embryogenesis. Circ Res 75:803–812 Miano JM, Carlson MJ, Spencer JA, Misr RP 2000 Serum response factor-dependent regulation of the smooth muscle calponin gene. J Biol Chem 275:9814–9822 Nakamura M, Nishida W, Mori S, Hiwada K, Hayashi K, Sobue K 2001 Transcriptional activation of beta-tropomyosin mediated by serum response factor and a novel Barx homologue, Barx1b, in smooth muscle cells. J Biol Chem 276:18313–18320 Nishida W, Nakamura M, Mori S et al 2002 A triad of serum response factor and the GATA and NK families governs the transcription of smooth and cardiac muscle genes. J Biol Chem 277:7308–7317 O’Brien ER, Alpers CE, Stewart DK et al 1993 Proliferation in primary and restenotic coronary atherectomy tissue. Implications for antiproliferative therapy. Circ Res 73:223–231 Owens GK 1995 Regulation of differentiation of vascular smooth muscle cells. Physiol Rev 75:487–517 Owens GK, Thompson MM 1986 Developmental changes in isoactin expression in rat aortic smooth muscle cells in vivo. Relationship between growth and cytodifferentiation. J Biol Chem 261:13373–13380

190

OWENS

Owens GK, Vernon SM, Madsen CS 1996 Molecular regulation of smooth muscle cell differentiation. J Hypertens 14:S55–64 Owens GK, Kumar MS, Wamhoff BR 2004 Molecular regulation of vascular smooth muscle cell differentiation in development and disease. Physiol Rev 84:767–801 Pipes GC, Sinha S, Qi X et al 2005 Stem cells and their derivatives can bypass the requirement of myocardin for smooth muscle gene expression. Dev Biol 288:502–513 Qiu P, Li L 2002 Histone acetylation and recruitment of serum responsive factor and CREB-binding protein onto SM22 promoter during SM22 gene expression. Circ Res 90: 858–865 Rice JC, Allis CD 2001 Histone methylation versus histone acetylation: new insights into epigenetic regulation. Curr Opin Cell Biol 13:263–273 Rivera VM, Miranti CK, Misra RP et al 1993 A growth factor-induced kinase phosphorylates the serum response factor at a site that regulates its DNA-binding activity. Mol Cell Biol 13:6260–6273 Rovner AS, Murphy RA, Owens GK 1986a Expression of smooth muscle and nonmuscle myosin heavy chains in cultured vascular smooth muscle cells. J Biol Chem 261:14740– 14745 Rovner AS, Thompson MM, Murphy RA 1986b Two different heavy chains are found in smooth muscle myosin. Am J Physiol 250:c861–870 Sartorelli V, Kurabayashi M, Kedes L 1993 Muscle-specific gene expression. A comparison of cardiac and skeletal muscle transcription strategies. Circ Res 72:925–931 Shore P, Sharrocks AD 1994 The transcription factors Elk-1 and serum response factor interact by direct protein-protein contacts mediated by a short region of Elk-1. Mol Cell Biol 14:3283–3291 Sobue K, Sellers JR, Caldesmon 1991 A novel regulatory protein in smooth muscle and nonmuscle actomyosin systems. J Biol Chem 266:12115–12118 Somasundaram C, Kallmeier RC, Babij P 1995 Regulation of smooth muscle myosin heavy chain gene expression in cultured vascular smooth muscle cells by growth factors and contractile agonists. Basic Appl Myol 6:31–36 Somlyo AP, Somlyo AV 2003 Calcium sensitivity of smooth muscle and non-muscle myosin II: modulation by G proteins, kinases, and myosin phosphatase. Physiol Rev 88:1325–1368 Thyberg J, Palmberg L, Nilsson J, Ksiazek T, Sjolund M 1983 Phenotypic modulation in primary cultures of arterial smooth muscle cells: on the role of platelet-derived growth factor. Differentiation 25:156–167 van der Loop FTL, Gabbiani G, Kohnen G, Ramaekers FCS, van Eys GJJM 1997 Differentiation of smooth muscle cells in human blood vessels as defi ned by smoothelin, a novel marker for the contractile phenotype. Arterioscler Thromb Vasc Biol 17:665–671 Vernon SM, Campos MJ, Haystead TAJ, Thompson MM, Dicorleto PE, Owens GK 1996 Endothelial cell conditioned media downregulates smooth muscle contractile protein expression. Am J Physiol 272:C582–591 Wang D, Chang PS, Wang Z et al 2001 Activation of cardiac gene expression by myocardin, a transcriptional cofactor for serum response factor. Cell 105:851–862 Wang Z, Wang DZ, Pipes GC, Olson EN 2003 Myocardin is a master regulator of smooth muscle gene expression. Proc Natl Acad Sci USA 100:7129–7134 Wang Z, Wang DZ, Hockemeyer D, McAnally J, Nordheim A, Olson EN 2004 Myocardin and ternary complex factors compete for SRF to control smooth muscle gene expression. Nature 428:185–189 Wilcox JN 1992 Analysis of local gene expression in human atherosclerotic plaques. J Vasc Surg 15:913–916 Winder SJ, Sutherland C, Walsh MP 1991 Biochemical and functional characterization of smooth muscle calponin. Adv Exp Med Biol 304:37–51

VASCULAR SMOOTH MUSCLE CELLS

191

Yano H, Hayashi K, Momiyama T, Saga H, Haruna M, Sobue K 1995 Transcriptional regulation of the chicken caldesmon gene. Activation of gizzard-type caldesmon promoter requires a CArG box- like motif. J Biol Chem 270:23661–23666 Yoshida T, Owens GK 2005 Molecular determinants of vascular smooth muscle cell diversity. Circ Res 96:280–291 Yoshida T, Sinha S, Dandre F et al 2003 Myocardin is a key regulator of CArG-dependent transcription of multiple smooth muscle marker genes. Circ Res 92:856–864 Yoshida T, Hoofnagle MH, Owens GK 2004 Myocardin and Prx1 contribute to angiotensin II-induced expression of smooth muscle alpha-actin. Circ Res 494:1075–1082

DISCUSSION Dejana: What are the factors produced by the tumour cells that induce change in the phenotype of the smooth muscle cells? Owens: One obvious candidate was PDGF BB, because we know it can profoundly inhibit differentiation of smooth muscle cells. However, our evidence so far indicates it is not PDGF BB. We did pharmacological inhibition and PDGF BB neutralizing antibody studies and neither inhibited the tumour cell repressor effect. Whereas many tumour cells produce BB, it does not appear to be responsible for the activity we observed in these prostate tumour cell lines. On the basis of cytokine array analyses, we think it may be PDGF-DD. We also have some evidence implicating interleukin (IL)8. Kitajewski: Is KLF4 the same as LKLF? Owens: No, LKLF is KLF2. Jeff Leiden’s lab published a paper on the LKLF (KLF2) knockout (Kuo et al 1997). Based on evidence that LKLF is expressed in leukocytes, they were expecting an immune function phenotype but surprisingly found defective smooth muscle/pericyte investment of blood vessels. They also showed that LKLF was primarily expressed in the endothelium and postulated that loss of KLF2 in endothelial cells somehow disrupts recruitment of mural cells. Of interest, Mike Gimbrone and his colleagues have published results showing that KLF2 appears to be a key downstream effector in mediating an ‘inflammatory phenotype’ in endothelial cells in areas of disturbed blood flow (SenBanerjee et al 2004). It is thus interesting to postulate that the role of KLF2 during development is related to effects of haemodynamic forces, although at present there is no direct evidence for this to my knowledge. Drake: SMCs are of mesodermal origin. There seems to be a temporal component to their differentiation. At the time that we first see differentiated SMCs (i.e. under the dorsal aorta), do you think those mechanisms are opening up in the mesodermal cells that have yet to differentiate? Owens: The dogma used to be that all the chromatin is open and any gene can be activated within ESCs. However, a very elegant recent paper from Eric Lander’s group showed that this is not the case. Rather, they found that the chromatin of

192

OWENS

key developmental genes including Sox, Hox, Pax, and Pou gene family members displayed unique histone modification patterns in ESCs, in that the genes contained large stretches of repressive histone modifications while simultaneously harbouring activating histone modifications. They termed this unusual combination of modifications as ‘bivalent’ histone modifications, and postulated that this mechanism provides a key switching mechanism that makes key lineage regulatory genes within ESCs permissive for either activation or silencing in response to appropriate environmental cues within the developing embryo. Shibuya: I’m interested in the dependency of the SMCs on the PDGF receptor system. SMCs have some heterogeneity, and there has been recent work trying to suppress tumour angiogenesis by using chemicals that inhibit not only the VEGF receptor kinase but also the PDGF receptor kinase. Do you think that inhibition of both receptors suppresses tumour growth better than the single inhibitor? Owens: PDGF BB is very interesting. There is a population of β receptor-positive mural cells that are recruited to the developing endothelium and give rise to pericytes or SMCs. We would love to know what the signals are that regulate this process (i.e. instruct these presumptive SMC/pericytes and ‘tell them’ when and where to invest). Betsholtz, Bowen-Pope, and others have presented very convincing evidence the PDGF β receptor-signalling is critical for this process using both conventional and chimeric β -receptor knockout mice (Lindahl et al 1997, Crosby et al 1998). It was thus paradoxical to us that when we applied PDGF BB to fully differentiated SMCs derived from adult blood vessels they profoundly down-regulated expression of SMC differentiation marker genes. In contrast, there are several papers showing that PDGF BB treatment can enhance formation of SMCs in several in vitro heterogenous cell model systems including embryonic proepicardial organ cells, which give rise to coronary SMCs (Landerholm et al 1999), or in circulating progenitor cells derived from the buffy coat (Simper et al 2002). We believe that in these systems the effects of PDGF BB on SMC differentiation are not direct but rather a function of it promoting selective expansion of PDGF β receptor positive cells within these mixed cell systems, such that there is a larger population of cells that subsequently go onto differentiate into a SMC lineage. The net result is that you get more cells that are expressing SMC markers. References Crosby JR, Seifert RA, Soriano P, Bowen-Pope DF 1998 Chimaeric analysis reveals role of Pdgf receptors in all muscle lineages. Nat Genet 18:385–388 Kuo CT, Veselits ML, Barton KP, Lu MM, Clendenin C, Leiden JM 1997 The LKLF transcription factor is required for normal tunica media formation and blood vessel stabilization during murine embryogenesis. Genes Dev 11:2996–3006 Landerholm TE, Dong X-R, Belaguli N, Schwartz RJ, Majesky MW 1999 A role for serum response factor in coronary smooth muscle differentiation from proepicardial cells. Development 126:2053–2062

VASCULAR SMOOTH MUSCLE CELLS

193

Lindahl P, Johansson BR, Leveen P, Betsholtz C 1997 Pericyte loss and microaneurysm formation in PDGF-B-deficient mice. Science 277:242–245 SenBanerjee S, Lin Z, Atkins GB et al 2004 KLF2 Is a novel transcriptional regulator of endothelial proinflammatory activation. J Exp Med 199:1305–1315 Simper D, Stalboerger PG, Panetta CJ, Wang S, Caplice NM 2002 Smooth muscle progenitor cells in human blood. Circulation 106:1199–1204

Growth factor gradients in vascular patterning Andrea Lundkvist, Sunyoung Lee*, Luisa Iruela-Arispe*, Christer Betsholtz† and Holger Gerhardt1 Vascular Biolog y Laboratory, Cancer Research UK London Research Institute, 44 Lincoln’s Inn Fields, London WC2A 3PX, UK, *Molecular Biolog y Institute, Jonsson Comprehensive Cancer Center and Department of Molecular, Cell and Developmental Biolog y, University of California, Los Angeles, 611 Charles E. Young Drive Ea., Los Angeles, CA 99095, USA and †Laboratory of Vascular Biolog y, Division of Matrix Biolog y, Department of Medical Biochemistry and Biophysics, Scheeles vag 2, Karolinska Institutet, SE-171 77 Stockholm, Sweden

Abstract. Growth factor gradients regulate many developmental processes. VEGF-A is distributed in a graded fashion in growing tissues in order to direct sprouting of new vessels. Growth factor gradients can be formed by regulated production, retention, controlled release and degradation. VEGF-A production is controlled by hypoxia while its retention depends on the C-terminal heparin-binding motifs present in the longer splice-isoforms, VEGF164 and 188. This motif confers binding to the cell surface and the surrounding extracelluar matrix. The short isoform VEGF120 is diffusible and hence fails to direct endothelial tip cell migration. Conditional inactivation of heparan sulfate proteoglycans in the cells that produce VEGF results similarly in misguidance of the tip cells. Studying retinal developmental angiogenesis and pathological neovascularization side-by-side in the mouse retina, we fi nd that endothelial tip cell guidance and stalk cell proliferation control are disrupted in neovascularization due to a loss of VEGFA retention. The cause for this is proteolytic cleavage of VEGF-A by matrix metalloproteases (MMP) derived mostly from macrophages infi ltrating the ischaemic retinal areas. Genetic or pharmacological inhibition of macrophage infi ltration or MMP activity can rescue guided revascularization at the expense of pre-retinal neovascularization. Disruption of VEGF-A gradients provides a novel concept for the mechanism underlying pathological patterning in ocular disease. 2007 Vascular development. Wiley, Chichester (Novartis Foundation Symposium 283) p 194–206

Growth factor and morphogen gradients are versatile tools for patterning previously uniform space into distinct domains, to fine tune cellular responses like differentiation and proliferation and to direct cellular migration (Charron & Tessier-Lavigne 2005). VEGF-A is involved in early differentiation processes of the haemangioblast lineage (Damert et al 2002), but also fulfi ls numerous func1

This paper was presented at the symposium by Holger Gerhardt, to whom correspondence should be addressed. 194

GROWTH FACTOR GRADIENTS

195

tions in subsequent patterning events that lead to the formation of an intricate hierarchical vascular network that is precisely adapted to the various demands of the respective organs (Ferrara et al 2003). The extraordinary dependence of the vascular system on this one growth factor is best illustrated by the fact that both a two-fold reduction (Carmeliet et al 1996, Ferrara et al 1996) and a two-fold increase (Miquerol et al 2000) of VEGF-A levels in mouse embryos is incompatible with survival. Thus VEGF-A levels need to be tightly controlled in the tissue. Additionally, the spatial distribution of VEGF-A appears to be critical for vascular patterning. Our recent studies on the mouse retina and embryonic hindbrain illustrated that a graded distribution of VEGF-A is instrumental in regulating two fundamental processes during sprouting angiogenesis, tip cell migration and stalk cell proliferation (Ruhrberg et al 2002, Gerhardt et al 2003). The tip cell concept The leading endothelial cell of the vascular sprout (the tip cell) differs from the following stalk cells in terms of morphology, gene expression and response to VEGF-A (Gerhardt et al 2003, Gerhardt & Betsholtz 2005). The tip cells are induced by high VEGF-A concentrations, they extend multiple long fi lopodia towards the source of VEGF-A, they migrate towards the VEGF-A source and generate pulling forces, likely through selective adhesion to the provisional matrix by fi lopodial integrins. They also present the VEGF receptor 2 (VEGFR2) on their fi lopodia, suggesting a sensor function for these protrusions. The tip cells further express certain genes related to their function as the leading cell. They express PDGF B in order to recruit pericytes to the nascent sprout, and they express Dll4 potentially to repress protrusive activity in the neighbouring stalk cells (see Anne Eichmann’s and George Yancopoulos’ papers in this book). Mechanisms of tip cell guidance Endothelial tip cell guidance bears similarities to guided migration in other organ systems, like axonal guidance during neural development (Carmeliet & TessierLavigne 2005), or sprouting of the Drosophila tubular airway system, the trachea (Metzger & Krasnow 1999). Here growth factor gradients, matrix scaffolds and specific guideposts all contribute to guidance of the leading tip structures. Whereas long-range gradients may function to determine the gross direction, matrix scaffolds or guideposts may help to determine a specific trajectory of growth. Interestingly, recent studies have shown that a number of neural guidance molecules and receptors are involved in endothelial tip cell guidance. Thus, attractive growth factor gradients may work in concert with repulsive signals to fine-tune directed migration of the tip cells. However, unlike the growing axon, the multicellular angiogenic sprout requires a co-ordinated migration of the tip and proliferation in the stalk.

196

LUNDKVIST ET AL

Balancing migration and proliferation The proliferative response in the growing vasculature is largely restricted to the stalk cells in the vicinity of the sprouting front (Gerhardt et al 2003). Their number appears to be controlled by the local levels of VEGF-A. Altering the concentrations and distribution of VEGF-A in the retina by local administration or genetic manipulation allowed us to study the response of tip and stalk cells in this tissue. Based on our observations we proposed that the graded distribution of VEGF-A regulates the balance of tip cell migration and stalk cell proliferation and thereby controls vascular patterning during sprouting angiogenesis. In the present paper we address the question of how VEGF-A gradients are formed and instruct normal development, and then turn to the model of oxygeninduced retinopathy to ask how pathological neovascularization differs from normal angiogenesis in order to elucidate mechanisms of pathological vascular patterning. We provide the first evidence for a loss of endothelial guidance through disrupted VEGF-A gradients, representing the fundamental cause of retinal vasculopathies (Lundkvist et al 2007, submitted). Graded versus diffuse VEGF-A distribution Summarizing previous work, the morphological appearance of a growing vascular sprout differs considerably in situations of graded versus diffuse VEGF-A distribution. Graded VEGF-A distribution leads to rapid tip cell migration and confined stalk cells proliferation only in the vicinity of the VEGF-A source. The resulting sprout is slender with a homogenous stalk diameter and the vascular network is highly branched. The fi lopodia of the tip cell are polarized, long and confined mostly to the very tip of the sprout. In contrast, diffuse VEGF-A is insufficient to direct tip cell migration but results in widespread stalk cell proliferation. Mechanisms to form a gradient in the tissue Gradient formation will depend on the sum of mechanisms that can contribute to raise the concentration at one place above the concentrations in the surroundings. Gradient formation therefore can be regulated at various levels. First of all, the production of the growth factor needs to be regulated in a spatiotemporal fashion that allows higher concentrations to be spatially restricted. VEGF-A production can be induced by many factors in cultured cells. Interestingly, in vivo, in the retinal system VEGF-A is largely produced by the astrocytes in the area ahead of the growing vascular plexus (Stone et al 1995, Provis et al 1997, Gerhardt et al 2003). Many studies have shown that hypoxia is the main trigger and regulator for VEGF-

GROWTH FACTOR GRADIENTS

197

A expression. Oxygenated tissue rapidly shuts down VEGF-A expression (for details on hypoxia regulation see Georg Breier’s paper in this book). Gradient formation also relies on regulated retention of the produced factor. If concentrations are to be maintained at a certain site, say the astrocytes ahead of the vessels, VEGF-A needs to be bound locally in order to maintain high local concentrations. Splicing of the VEGF-A gene into several different isoforms is a powerful tool to regulate retention by including sequences from exon 6 and 7 that code for a conserved stretch of basic amino acids, termed the retention motif. The longer isoforms, 188 and 164 include this retention motif and thus are retained at the cell surface or in the proximal extracellular matrix (Park et al 1993). The cell surface and extracellular matrix is rich in heparan sulfate proteoglycans (HSPGs), which carry sulfated sugar chains that represent ideal attachment sites for basic peptides through charged interactions (Esko & Lindahl 2001). Although the specificity of these interactions is currently debated, it is clear that HSPG are crucial for the formation of many growth factor and morphogen gradients. A recent study by Robinson and Stringer identified parts of the structural requirements for the sulfated sugar chains to effectively bind the VEGF164 dimer (Robinson et al 2006). Our lab has recently turned towards HS function in vascular patterning and begun to study the effect of conditional HS deletion in the various cell-types that interact with the sprouting front. First data on astrocyte specific deletion of HS by GFAPCre driven recombination at the EXT1 locus in astrocytes, suggest that astrocytic HS is the key to VEGF-A gradient formation in the retinal tissue. We observed a specific defect in fi lopodia guidance and stalk cell proliferation control at the leading vascular front, highly reminiscent of the images seen in the VEGF120/120 mice, in which the growth factor lacks the retention motif. A further mechanism to shape a gradient can be regulated local release and/or degradation. Little is known about these processes in respect to VEGF-A. However, a number of studies have implicated a series of proteases including members of the cathepsin family, matrix metalloproteinases (MMPs), plasmin and others to either alter the binding of VEGF by proteolysis of VEGF itself, or by modifying the binding properties of the matrix (Houck et al 1992, Plouet et al 1997, Lee et al 2005). In fact, it has been suggested that the switch from dormant tumour lesion to highly angiogenic tumour (the so-called ‘angiogenic switch’) is triggered by mesenchymal MMP production, thus releasing VEGF-A to activate neo-angiogenesis (Bergers et al 2000). Whether local release and degradation is involved in VEGF-A gradient formation in normal development is unclear. What is the cause of disturbed vascular patterning in pathological neovascularization? The mouse model of oxygen-induced retinopathy (OIR) closely resembles vascular malformations occurring in situations of ischemia-driven retinal

198

LUNDKVIST ET AL

neovascularization in diabetic patients (Smith et al 1994). Capillary-denuded regions in the retina suffer from tissue hypoxia, triggering a neovascular response. Unlike normal development, this neovascular response fails to re-establish an organized hierarchical pattern, but results in chaotic vascular malformations and hyperproliferation with pre-retinal vessels. These pre-retinal vessels eventually lead to bleeding, fibrosis and detachment. From a therapeutic perspective, two separate aspects require attention: some areas have too few vessels, whereas other areas develop too many and in ectopic positions. Thus the ideal therapeutic approach should work towards guiding effective vessel regrowth into the avascular region while inhibiting formation of pre-retinal vascularization. In the mouse model of OIR, the primary retinal plexus forms normally in the first postnatal week P1–P7. Subsequently, the pups are housed in 75% oxygen, leading to regression of capillaries in the central regions. On day P12, the pups are returned to room air oxygen levels (21%) for another 5 days (Smith et al 1994). During this phase, the neovascularization is triggered leading to pre-retinal vascular tufts, hyperproliferation of the veins and arterial tortuousity. Examining proliferation by BrdU injection and tip cell polarization by lectin immunoflourescence and fi lopodia measurement (Gerhardt et al 2003), we observed that the onset of neovascularization closely resembled organized patterning similar to normal developmental angiogenesis. Endothelial cells only proliferated in the vicinity of the leading sprouting front and tip cells extended long fi lopodia along VEGFproducing astrocytes. This organized pattern of proliferation and fi lopodia extension was disrupted on days 13–17. The resemblance of the vascular phenotype during retinal neovascularization to observations on experimental disruption of VEGF-A gradients, led us to ask whether VEGF-A gradients may be disturbed in OIR. In an attempt to test this, we combined VEGF IHC with VEGF ISH and BrdU labelling to determine the spatial relationship between the site of VEGF-A production, the site of VEGF-A protein localization, and the site of the endothelial response. At day 12, VEGF-A production at the mRNA level was confi ned to astrocytes in the avascular zone and absent around the arteries. VEGF-A protein was also absent around arteries and arterial ECs did not proliferate. However, at day 17, arterial ECs readily proliferated, and VEGF-A protein was also detected in the vicinity of these proliferating cells. Interestingly, in situ hybridization revealed that VEGF-A mRNA was still absent around arteries, suggesting that VEGF-A protein must have relocated (diffused) away from the producing cells. We performed RT-PCR to determine whether this could be explained by up-regulation of the non-heparin binding isoform VEGF120, but found that the heparin binding isoform VEGF164 was predominantly up-regulated at mRNA levels. This suggested that VEGF-A relocalization was not caused by an isoform switch. A recent study showed that a novel truncated form of VEGF-A is produced by proteolytic cleavage at the C-terminal end by members of the MMP family (Lee et al 2005).

GROWTH FACTOR GRADIENTS

199

The resulting proteolytic fragment VEGF113 lacks the heparin binding C-terminus, but contains the receptor binding domains. Using a sandwich ELISA assay with two different antibodies that recognize the N-terminus and the C-terminus of VEGF-A, we determined the ratio of cleaved versus uncleaved VEGF in retinal samples from normal retinal development, OIR at P12 and P14 as well as P17. We also included control samples from mice exposed to continued hyperoxia, in which revascularization of the retina occurs without pathological pre-retinal vessels (Gu et al 2002). The samples from normal developmental stages, P12 and continued hyperoxia controls, showed no detectable levels of cleaved VEGF-A, suggesting that this MMP-dependent processing of VEGF-A does not contribute to normal development. However, in OIR at P14 and increasingly at P17, cleaved VEGF-A constituted up to 80% of the total VEGF-A. Thus, more than half of the VEGF-A present in OIR lacked the heparin-binding C-terminus and consequently would lack retention at the cell surface or in the surrounding extracellular matrix. In comparison, mice that have one VEGF-A allele replaced with the VEGF120 isoform develop similar pre-retinal neovascularization during normal development, arguing that even 50% of diffusible VEGF-A is detrimental for guided vascular growth. We next asked whether MMPs are involved in this cleavage in vivo and therefore tested a broad-range inhibitor of MMPs (GM6001) by i.p. injection during the neovascularization period. Interestingly, we observed restored re-vascularization and normalized patterning concomitant with polarized fi lopodia protrusion and restored proliferation control. Furthermore the VEGF-A protein localization was normalized in these retinas, arguing that indeed MMPs are involved in the VEGFA redistribution in the pathological phase of OIR. In a gain-of-function approach we asked whether local administration of recombinant MMPs to a normal unchallenged retina would mimic a loss of VEGF-A gradients. 24 h after intravitreal delivery of MMP9, MMP3 or MMP12, we observed substantial shortening of tip cell fi lopodia and a hyperplasia of the stalks reminiscent of VEGF-A injection. Injection of pro-MMP9 had no effect. After 48 h, on P7, pre-retinal vascularization occurred, suggesting that these MMPs are sufficient to cause vascular malformations similar to OIR. RT-PCR for all MMPs and their natural inhibitors (TIMPs) surprisingly showed prominent up-regulation of only one member, MMP12. In vitro MMP12 potently cleaved VEGF-A, while addition of GM6001 completely abolished VEGF-A cleavage. MMP12, also known as macrophage elastase, is specifically expressed by macrophages. In OIR, macrophages are most abundant in the capillary-free, hypoxic areas. This recruitment did not occur in CSF-1op/op mice and could also be inhibited by clodronate-liposome administration i.p. and intravitreally (Van Rooijen & Sanders 1994). Inhibition of macrophage recruitment led to a similarly normalized re-vascularization as found after GM6001 treatment. We tested the potential involvement of MMP12 by analysing MMP12-deficient

200

LUNDKVIST ET AL

mice (Shipley et al 1996). MMP12-deficient animals showed no alterations in any aspect of normal retinal angiogenesis and also showed a similar vascular regression following the hyperoxia treatment. However, already heterozygous animals showed significantly improved re-vascularization and reduced epiretinal tuft area following OIR. We conclude that MMPs are necessary and sufficient to disrupt VEGF-A localization, which is otherwise dependent on C-terminal interaction with heparan sulfate. Conclusions Normal vascular patterning depends on VEGF-A gradients formed through localized production and retention of the longer VEGF-A isoforms 164 and 188. In the OIR model of pathological vascular patterning, VEGF164 is cleaved by MMPs resulting in a loss of retention and directed re-vascularization. Inhibition of MMPdependent VEGF-A cleavage is capable of restoring VEGF-A distribution, and therefore restoring guided re-vascularization and functional vascular patterning. Our present results further highlight the involvement of macrophages in OIR. For human complications such as diabetic retinopathy, the involvement of MMPs has been suggested, however, we provide evidence for a novel mechanism by which MMPs may lead to pathological vascular patterning. The wide range of MMPs that have the capability of cleaving VEGF-A however suggests that identification of one definitive member may be difficult and several different MMPs may be involved. Acknowledgements This work was supported by Cancer Research UK. We are grateful to Lieve Moons and Peter Carmeliet for providing MMP-9 and MMP-12 deficient animals for this study. Thanks to Marcus Fruttiger for valuable comments and discussions.

References Bergers G, Brekken R, McMahon G et al 2000 Matrix metalloproteinase-9 triggers the angiogenic switch during carcinogenesis. Nat Cell Biol 2:737–744 Carmeliet P, Tessier-Lavigne M 2005 Common mechanisms of nerve and blood vessel wiring. Nature 436:193–200 Carmeliet P, Ferreira V, Breier G et al 1996 Abnormal blood vessel development and lethality in embryos lacking a single VEGF allele. Nature 380:435–439 Charron F, Tessier-Lavigne M 2005 Novel brain wiring functions for classical morphogens: a role as graded positional cues in axon guidance. Development 132:2251–2262 Damert A, Miquerol L, Gertsenstein M, Risau W, Nagy A 2002 Insufficient VEGFA activity in yolk sac endoderm compromises haematopoietic and endothelial differentiation. Development 129:1881–1892 Esko JD, Lindahl U 2001 Molecular diversity of heparan sulfate. J Clin Invest 108:169–173 Ferrara N, Carver-Moore K, Chen H et al 1996 Heterozygous embryonic lethality induced by targeted inactivation of the VEGF gene. Nature 380:439–442 Ferrara N, Gerber HP, LeCouter J 2003 The biology of VEGF and its receptors. Nat Med 9:669–676

GROWTH FACTOR GRADIENTS

201

Gerhardt H, Betsholtz C 2005 How do endothelial cells orientate? EXS 3–15 Gerhardt H, Golding M, Fruttiger M et al 2003 VEGF guides angiogenic sprouting utilizing endothelial tip cell fi lopodia. J Cell Biol 161:1163–1177 Gu X, Samuel S, El-Shabrawey M et al 2002 Effects of sustained hyperoxia on revascularization in experimental retinopathy of prematurity. Invest Ophthalmol Vis Sci 43:496–502 Houck KA, Leung DW, Rowland AM, Winer J, Ferrara N 1992 Dual regulation of vascular endothelial growth factor bioavailability by genetic and proteolytic mechanisms. J Biol Chem 267:26031–26037 Lee S, Jilani SM, Nikolova GV, Carpizo D, Iruela-Arispe ML 2005 Processing of VEGF-A by matrix metalloproteinases regulates bioavailability and vascular patterning in tumors. J Cell Biol 169:681–691 Lubarsky B, Krasnow MA 2003 Tube morphogenesis: making and shaping biological tubes. Cell 112:19–28 Metzger RJ, Krasnow MA 1999 Genetic control of branching morphogenesis. Science 284:1635–1639 Miquerol L, Langille BL, Nagy A 2000. Embryonic development is disrupted by modest increases in vascular endothelial growth factor gene expression. Development 127: 3941–3946 Park JE, Keller GA, Ferrara N 1993 The vascular endothelial growth factor (VEGF) isoforms: differential deposition into the subepithelial extracellular matrix and bioactivity of extracellular matrix-bound VEGF. Mol Biol Cell 4:1317–1326 Plouet J, Moro F, Bertagnolli S et al 1997 Extracellular cleavage of the vascular endothelial growth factor 189-amino acid form by urokinase is required for its mitogenic effect. J Biol Chem 272:13390–13396 Provis JM, Leech J, Diaz CM, Penfold PL, Stone J, Keshet E 1997 Development of the human retinal vasculature: cellular relations and VEGF expression. Exp Eye Res 65:555–568 Robinson CJ, Mulloy B, Gallagher JT, Stringer SE 2006 VEGF165-binding sites within heparan sulfate encompass two highly sulfated domains and can be liberated by K5 lyase. J Biol Chem 281:1731–1740 Ruhrberg C, Gerhardt H, Golding M et al 2002 Spatially restricted patterning cues provided by heparin-binding VEGF-A control blood vessel branching morphogenesis. Genes Dev 16:2684–2698 Shipley JM, Wesselschmidt RL, Kobayashi DK, Ley TJ, Shapiro SD 1996 Metalloelastase is required for macrophage-mediated proteolysis and matrix invasion in mice. Proc Natl Acad Sci USA 93:3942–3946 Smith LE, Wesolowski E, McLellan A et al 1994 Oxygen-induced retinopathy in the mouse. Invest Ophthalmol Vis Sci 35:101–111 Stone J, Itin A, Alon T et al 1995 Development of retinal vasculature is mediated by hypoxiainduced vascular endothelial growth factor (VEGF) expression by neuroglia. J Neurosci 15:4738–4747 Van Rooijen N, Sanders A 1994 Liposome mediated depletion of macrophages: mechanism of action, preparation of liposomes and applications. J Immunol Methods 174:83–93

DISCUSSION Drake: Your work shows that tip cells send out multiple processes. VEGF appears to be promoting this huge exploratory front, as if the cell is not sure where to go rather than following a tight gradient. When we watch the expanding vascular front in allantois cultures (which look like your retinal network), there are

202

LUNDKVIST ET AL

leading cells and they send out numerous fi lopodia. Based on the predictable vascular pattern that is formed, one can tell which fi lopodial connections will lead to new blood vessels and form the next part of the network. It seems that what VEGF is doing at the tip is promoting the formation of numerous fi lopodia, so the gradient is rather more like a general stimulant. What do you believe mediates the persistence of one fi lopodia versus another? Betsholtz: Are you asking how the fi lopodia that contact those of the other cell are chosen, or how the cell extending the fi lopodia was chosen in the first place? Drake: The tip cell’s business is sending out the fi lopodia, which is the working end of the tip cell. Do you think there is a broad gradient of VEGF? Gerhardt: In the retinal model it is clear that there isn’t a broad gradient. VEGF is definitely stimulating this whole process; if we inhibit VEGF we can show that the protrusive activity is a behaviour dependent on VEGF. The VEGF distribution can be altered in the retinal system by putting animals into hyperoxic or hypoxic conditions. This changes the levels of VEGF that are produced within the plexus (vascularized retina) relative to ahead of the plexus (avascular retinal periphery). Ahead of the plexus where no vessels have yet formed, the level of oxygen that the animal breathes has no effect. It is the balance between VEGF ahead and behind the leading front that is shifted by tweaking the oxygen concentration the animal breathes. By doing this you can fine tune how rapidly the retinal vasculature grows out, and how extensively stalk cells proliferate. In hypoxic conditions there is more VEGF in the already vascularized area, thus reducing the spatial differences in VEGF levels between avascular periphery and vascularized retinal centre. In many cases, these relative differences are tightly controlling the response. If MMPs are inhibited during normal development, nothing happens. We don’t think that MMPs are doing the same job in normal development as in pathological neovascularization. In fact, it is unclear whether they have any role in normal vascular development. The macrophages that I presented as a key to pathological neovascularization seem to have a different role in development. Retinas that lack macrophages have even more extreme directionality of the tip cell response. In these retinas, tip cells show less explorative behaviour, leading to less branching, but the migration of the tip cells is not affected at all. The tip cells run along exactly the direction that you’d predict because of the VEGF gradient. With regard to the question of fusion, or what determines which cells will finally connect, we have made an interesting observation: in the absence of macrophages, endothelial tip cells are not mutually attracting each other. The fusion points coincide with macrophages in almost all tissues in the mouse embryo. In zebrafish, we have been looking at the dynamics of this behaviour. The macrophages whiz around the tip cell. When they do this, the explorative behaviour of the tip cell is dramatically increased. At the fusion points, however, the macrophages appear rather immobile and sit there until the fusion is completed.

GROWTH FACTOR GRADIENTS

203

Drake: Perhaps there is an unappreciated role for macrophages during vasculogenesis. Betsholtz: I am not sure that fi lopodia guidance is the same as cell guidance here. It would be interesting to hear comments from someone more knowledgeable than me about this. I am not convinced that fi lopodia are at all guided towards the VEGF gradient. Instead they might be selectively stabilized through VEGF signalling. The initial response to VEGF might be broad fi lopodial protrusion (which can be modulated by macrophages). Subsequently selective stabilization may occur of those fi lopodia that sense the highest VEGF concentration, or attach to macrophages, or sense soluble macrophage-derived cues. The stabilized fi lopodia may point out the direction of the cell movement. Ruhrberg: I liked your description of potential mechanisms that may operate during endothelial cell guidance, i.e. the gradient, the matrix scaffolds and the guide post cells. These mechanisms may not be mutually exclusive, and, in fact, the retina work shows that they are not mutually excusive. You have the VEGF gradients, but you also have matrix scaffolds, as you showed how the astrocyte network prefigures the patterning of the overlying vessels. In your pictures at high magnifications, this co-patterning is seen already at the level of endothelial cell fi lopodia and astrocyte extensions. So it seems that there is a general growth factor gradient that attracts the vessels and also a scaffold for vessel filopodia to track on. It is probably the intersection of these two different pathways that achieves the final pattern, i.e. there is directionality provided by the gradient and fine patterning due to the shape of the astrocyte scaffolds. Gerhardt: We have been studying these matrix scaffolds in more detail. One student in my lab has done extensive work looking at all the different laminin forms, and fibronectin, in the context of the astrocytes as well as along the growing tip. A couple of laminins are consistently expressed by the astrocytes and fibronectin is also produced by the astrocytes. Interestingly, if we knock out fibronectin specifically in the astrocytes, we don’t see a lack of fi lopodial guidance. It only impacts on the speed of the tip cell migration. The fi lopodia still follow the astrocytes. Drake: During vasculogenesis, we also see fibronectin- as well as integrin α5β1mediated regulation of tip cell behaviour. Gerhardt: It may depend on the tissue context. We are in the process of deleting the other laminins in retinal astrocytes. They produce a complex matrix. We also have data showing that heparan sulfate proteoglycans are produced by the same cells. Our data do suggest that VEGF gradients are indeed involved in guiding fi lopodia. Most importantly, there is a massive misguidance phenotype in VEGF120 mice. It looks like a matrix-bound form of VEGF is involved in this guidance. But you are comparing it to stabilization and we may have a combination of both. Betsholtz: I am trying to reconcile this with Brant Weinstein’s videos of tip cells spearheading the intersegmental vessels in zebrafish embryos. There is apparently

204

LUNDKVIST ET AL

extensive dynamic explorative behaviour, with filopodia extending towards or even into the somites, followed by their retraction. I don’t think the fi lopodia are guided when they go off the track, it is just that they are not becoming stabilized. The net outcome is the guided cell migration, but the fi lopodial protrusions by themselves might be more random and part of a generic machinery triggered by VEGF. Weinstein: That is basically the way that neuronal guidance works. There aren’t selective protrusions. Protrusion is a fairly dynamic and stochastic process, but it is the stabilization of fi lopodia going the right way or repulsion of fi lopodia that go the wrong way that matters. Gerhardt: I agree. When looking at these analogies, I would have thought this is what is happening in our model as well. However, the surprising finding about the macrophages is that when they are gone, the fi lopodia appear not to be extending in the same stochastic way. They appear to be much more confined and directional. What are the macrophages doing? They seem to be binding VEGF very strongly. We are currently studying how they may cause this intensive explorative behaviour in the tip cells. The macrophages appear to bind up VEGF while sitting together with the astrocytes and could be affecting fi lopodia by releasing it again. Localized and matrix bound VEGF could potentially aid in stabilizing the fi lopodia. Weinstein: If you look at a static picture and it looks like there are more filopodia heading in the direction of the macrophage, you could be fooled: you might not be seeing the finest ones but just the largest ones that have become stabilized, and you have to look at the dynamic. How many protrusions per minute are going in a certain direction? Gerhardt: This is exactly why we turned to the zebrafish model. It looks very different, even if you take still pictures of zebrafish intersegmental vessels they look different from what is seen in the retina. This indicates that perhaps there are some aspects that are not quite the same, with more polarization in one system than the other. When we looked at particular issues of the dynamics, it seems stochastic in many areas. In the context of macrophages there are interesting differences, where the protrusive activity goes towards the macrophage and collapses when the macrophage disappears. Lammert: I want to ask about the tube mechanism. When the tip cells fuse to connect the vessels, are they seamless? Are the stalks seamless without junctions? Where is the basal lamina? Gerhardt: We don’t know yet whether the exact fusion point is seamless. However, we have looked at the electron microscopic level on ultrathin sections of the retina and found nice tip cells and fi lopodia in contact with the astrocytes. This allows us to look at the junctions, the lumen and matrix deposition. Interestingly, first of all we didn’t find any vesicles in the retinal system. We didn’t find any lumenal matrix. But to our surprise, we found formations of basal membrane already on the fi lopodia. We also found that the tip cell itself produces specific laminins. It

GROWTH FACTOR GRADIENTS

205

seems that there’s something else to look at in terms of what the tip cell does. I would like to get away from the dogma that the sprout needs first of all to degrade a lot of matrix. For a developmental system it might be different; if you want to invade into a scar tissue degrading matrix might be more important. Lammert: Are the stalks seamless or do they have autocellular junctions? Gerhardt: They have autocellular junctions. But I think there may be situations where one cell or two cells are found. They are definitely not seamless. Ye: I am curious why the new vessel always grows toward the inner limiting membrane and never into the ganglion cell layer. Gerhardt: I think this has something to do with the particular properties of the retina. Ye: What happens if you inject VEGF into the vitreous? Gerhardt: If you inject VEGF, you get these kinds of problems, as you do if you express VEGF from the lens. But also if you just have endogenous cells expressing only diffusible VEGF, then all of a sudden the vessels grow into the vitreous. This indicates that the loss of guided and directional growth into the deeper retinal layers, is what really causes the problem of growing up into the vitreous. This is probably the reason why the retinal vasculature is so sensitive to malformation. The vessels are in very close contact with this inner retinal surface, and once they make it into the vitreous, things go terribly wrong. In normal development, small bursts of fi lopodia are also protruding towards the inner limiting membrane at the very same site where new sprouts are forming to grow into the deeper retinal layers. This indicates that perhaps the initial activation of the vessel to form a new sprout is not a directed response. Then you need the gradient to guide the sprout down. If this doesn’t occur they just make it through to the vitreous. Dejana: As soon as the fi lopodia touch other cells they stabilize. The VEcadherin will be concentrated at that point. There should be a role for these junctional proteins in stabilization of the cells. What happens if some of these junctional proteins are missing? Gerhardt: We have some preliminary data using the VE-cadherin blocking antibody where we find that in the deeper plexus, when we inject into the retina, the tip cells sprout towards each other but they fail to make contact. This would be something that would be much better to examine in a dynamic fashion. If we look at dynamics, it does not look as if the moment a tip cell makes contact with another tip cell, they immediately stabilize contact and start to fuse; there is quite a long interaction. Weinstein: That is not entirely true. In many cases we see behaviour where a vessel has a guided track along which it is supposed to reach and fuse with another vessel, and we see dynamic activity not only in the vessel that is extending and reaching for the point of contact, but also in the vessel that it is going to be contacting. As soon as you make contact all that activity stops.

206

LUNDKVIST ET AL

Gerhardt: We’ve seen the same, but it is not the first fi lopodia that meet by chance that will immediately lead to stabilization. There can be some time. Weinstein: But not much. Once it makes contact it is pretty quick. Drake: In our system we see fi lopodia that make contacts that are clearly not in the right position, based on the pattern of vessels that is generated. These contacts do not appear to lead to the formation of a blood vessel. I don’t know how a particular contact is chosen, but there are many sprouts and connections that do not persist. Dejana: In culture, when two cells meet there is a repeated touching back and forth and then they establish a junction. Owens: It may be that both mechanisms are operative: perhaps a key distinction we need to make is whether it is a ‘genetically programmed’ vascular network versus a searching/probing activity that is more generic for a remodelling network. Brant Weinstein, I believe you suggested that in the former case you have two cells that ‘know’ they have to join, and as soon as they touch they cement that juncture. In contrast, with the tip cell probing we are seeing in these diffuse gradients, they are responding to a lot of gradients and the presence or absence of macrophages. These contacts may be much more transient than fixed depending on the local environmental cues. Weinstein: It could also be a time scale issue. Many cell biologists look at things in time scales of seconds. At that time scale it may look like there is a lot of touching or feeling or exploring. A lot of what we do is looking at minutes or hour time scales, where behaviour seems to be much more rapid and directed. Gerhardt: Our videos were shorter. We are probably talking about different time scales. Shibuya: I am interested in macrophages. In this situation we have up-regulation of VEGF. So, there are two possibilities. One is that inflammatory cytokines such as interleukin (IL)6 recruit macrophages, or another is that up-regulated VEGF stimulates VEGFRs on the macrophage to recruit the cell. Which is the major player? Gerhardt: This is exactly what we need to look at: how much of this macrophage infi ltration is due to VEGF? The data from Dave Shima and Tony Adamis suggest that VEGF164 is a very strong proinflammatory agent, and in this retinopathy model it may actually be VEGF that is the inflammatory agent. Published data show that other factors are also involved. For example, if MCP1 is inhibited, this reduces the recruitment of macrophages but not to full extent. There might be several factors working together, but this is not surprising. Shibuya: Is this kind of macrophage-involvement almost always seen in pathological retinopathies? Gerhardt: There are data suggesting that macrophages contribute to the problem in other retinopathies including choroidal neovascularization in the context of agerelated macular degeneration. It is probably true for most retinal vasculopathies.

Endothelial cell promotion of early liver and pancreas development Deborah A. Freedman, Yasushige Kashima and Kenneth S. Zaret Fox Chase Cancer Center, Cell and Developmental Biolog y Program, 333 Cottman Avenue, Philadelphia, PA 19111, USA

Abstract. Different steps of embryonic pancreas and liver development require inductive signals from endothelial cells. During liver development, interactions between newly specified hepatic endoderm cells and nascent endothelial cells are crucial for the endoderm’s subsequent growth and morphogenesis into a liver bud. Reconstitution of endothelial cell stimulation of hepatic cell growth with embryonic tissue explants demonstrated that endothelial signalling occurs independent of the blood supply. During pancreas development, midgut endoderm interactions with aortic endothelial cells induce Ptf1a, a crucial pancreatic determinant. Endothelial cells also have a later effect on pancreas development, by promoting survival of the dorsal mesenchyme, which in turn produces factors supporting pancreatic endoderm. A major goal of our laboratory is to determine the endothelial-derived molecules involved in these inductive events. Our data show that cultured endothelial cells induce Ptf1a in dorsal endoderm explants lacking an endogenous vasculature. We are purifying endothelial cell line product(s) responsible for this effect. We are also identifying endothelial-responsive regulatory elements in genes such as Ptf1a by genetic mapping and chromatin-based assays. These latter approaches will allow us to track endothelial-responsive signal pathways from DNA targets within progenitor cells. The diversity of organogenic steps dependent upon endothelial cell signalling suggests that cross-regulation of tissue development with its vasculature is a general phenomenon. 2007 Vascular development. Wiley, Chichester (Novartis Foundation Symposium 283) p 207–219

Organ vasculatures have diverse structures and functions. For example, pancreatic islets are rich in capillaries that facilitate the delivery of endocrine signals to the blood stream, while the liver vasculature consists of large sinusoids with fenestrated endothelial cells that function in the massive exchange of metabolic and waste products between the blood and the liver. Diverse developmental processes, including varying methods of lumen formation, likely co-ordinate the formation of these different vascular structures with the differentiation of the native cellular functions of each organ. This co-ordination may be regulated not only in development, but also during adult tissue turnover, regenerative responses to tissue damage, and tumorigenesis. Although much has been studied regarding growth 207

208

FREEDMAN ET AL

factor signalling from organ parenchyma to endothelial cells during the development of the vasculature, it is only recently that studies have investigated communication from developing vessels toward developing organ tissues. Our focus is on this latter communication from endothelial cells during endodermal organogenesis. Specification of endoderm-derived tissues into particular organ fates begins around embryonic day 8 (E8.0) in the mouse, shortly after the endoderm itself is formed in gastrulation. At E8.0, the endoderm is a sheet of cells on the outside of the embryo. From E8.0–E8.5, different cell fates in the endoderm are specified, with cells at different locations along the epithelium being destined to become progenitors of the thyroid, lung, liver, pancreas, and gut tube. By E9.0, the embryo has turned, helping the endoderm to form a closed tube, and organ morphogenesis has begun. In probably all cases, organ morphogenesis begins with thickening of the endodermal epithelium and proliferation and emergence of the cells into the surrounding mesenchyme, creating a tissue bud. While it was thought for many years that generic ‘mesenchymal’ cells provided crucial growth signals to promote the initial steps of organ morphogenesis, these conclusions were drawn from early tissue explant and transplant studies that did not employ markers to carefully define the constituents of the mesenchymal cell population (Fell & Grobstein 1968, Golosow & Grobstein 1962, Le Douarin 1975). Although later studies indeed confirmed that mesenchymal cells have specific signalling roles during endodermal organogenesis, as described below, another cell type crucial for organogenesis was lurking in the explants and transplants: the endothelial cell. In retrospect, the importance of endothelial cells for endodermal organogenesis should have been predicted, considering how critical the vasculature is for the sustenance and endocrine function of endodermal tissues. Endothelial signalling controls hepatic proliferation Historically, it was well established that capillary networks form in emerging endoderm tissue buds, such as in the liver and pancreas. Such capillaries were presumed, but never proven, to sprout from the nearby embryonic blood vessels by angiogenesis. Instead, detailed studies of the expression patterns of CD31 (PE-CAM) and VEGFR2 (Flk1) revealed that a sparse group of angioblast cells surround the thickened hepatic endoderm, within hours of liver specification and prior to the emergence of the endodermal cells to form a liver bud (Matsumoto et al 2001). Notably, these angioblasts intercede between the hepatic endoderm and the surrounding mesenchyme cells. The origin of these angioblasts is in question; they could differentiate from the mesenchyme cells or migrate from elsewhere. During liver bud emergence, the angioblasts coalesce to form vesicles and,

ENDOTHELIUM PROMOTES LIVER & PANCREAS DEVELOPMENT

209

ultimately, capillaries within the emerging liver bud (Matsumoto et al 2001). In summary, hepatic endoderm cells move into an angioblast cell domain prior to entering a mesenchyme-stromal cell domain, and liver bud vascularization appears to occur by vasculogenesis, not angiogenesis. We next asked if the angioblasts surrounding the emergent hepatic endoderm were crucial for organogenesis. Our approach was to employ embryos that were homozygous null for Flk1; such mutants arrest endothelial development at the angioblast stage, so that no mature endothelial cells or blood vessels form (Shalaby et al 1995). We found that in Flk1−/− embryos, hepatic specification occurs normally and the hepatic endoderm thickens (Matsumoto et al 2001). However, in the absence of local endothelial cells, the hepatic endoderm cells do not emerge into the stromal environment, resulting in a failure of liver development. Even though various other embryonic structures developed markedly well in the Flk1−/− embryos, the mutation impairs overall embryo growth and causes embryonic lethality by around E10. We therefore sought to bypass the lethality of the genetic model. Furthermore, we sought to ask whether the failure of liver bud outgrowth in Flk1-null mice was due to deficiencies in oxygen and nutrient supply to the tissue or due to an absence of signalling from endothelial cells. We therefore performed endodermal explant culture experiments, in which the expansion of hepatic tissue with or without endothelial cells could be assessed while keeping oxygen and nutrient levels equivalent between the two conditions. E9.0 liver buds dissected from Flk1−/− mice or control littermates were grown in culture for three days. Explants from Flk1-positive cultures formed vascular networks and the albumin-positive cells increased substantially in number. By contrast, in Flk1−/− explants, no vascular staining was seen, and the albumin-positive cells exhibited a clear defect in proliferation (Matsumoto et al 2001). Because both wild-type and Flk1−/− explants were cultured in the same medium and in the absence of a functioning vasculature, the presence of endothelial cells themselves was necessary to promote hepatic cell outgrowth. Similar results were obtained when we employed a non-genetic perturbation of endothelial development. That is, wild-type explants cultured in the presence of a soluble endothelial cell inhibitor exhibited a defect in the proliferation of albumin-positive cells (Matsumoto et al 2001). These studies showed that endothelial cells send signals to the endoderm, independent of the circulation. A signalling role of endothelial cells in organogenesis was contemporaneously discovered for early pancreatic development (Lammert et al 2001). Furthermore, endothelial cells were subsequently shown to release the hepatic growth factor HGF during adult liver regeneration (LeCouter et al 2003). We expect that endothelial cell signalling outside the cardiovascular system will be pervasive.

210

FREEDMAN ET AL

Endothelial signalling controls pancreatic specification The pancreatic bud emerges from both dorsal and ventral domains of endoderm; later, the dorsal and ventral buds fuse to create the gland. Initial studies from the Melton laboratory (Lammert et al 2001) showed that endothelial signalling from the aorta controls early pancreatic development, but it was unclear which stage(s) were controlled. The dorsal pre-pancreatic endoderm begins to express the transcription factor Pdx1 at the seven somite stage, or approximately E8.25 (Gannon & Wright 1999). This specification factor is also expressed in midgut endoderm that is fated to become stomach and duodenum. Expression of the transcription factor Ptf1a is crucial for the pancreatic specification of a subset of the Pdx1positive midgut endoderm (Kawaguchi et al 2002). At the time of Ptf1a initiation (at the 15–18 somite stage, or approximately E9.0), the dorsal pre-pancreatic endoderm is in direct contact with the dorsal aorta, which is comprised of a single layer of endothelial cells at this time. To ask whether this close proximity of endothelial cells to the endoderm results in signalling that induces pancreatic specification, we again used the Flk1−/− model. We found that the initial induction of Pdx1 occurs normally in Flk1 −/− embryos. Moreover, the ventral pancreatic bud forms normally. However, the dorsal pancreatic bud never forms, and Pdx1 expression is not maintained well in the dorsal endoderm (Yoshitomi & Zaret 2004). Notably, while Ptf1a expression is normal in the ventral endoderm, these embryos fail to initiate expression of Ptf1a dorsally. Thus both specification and morphogenesis of the dorsal pancreas is affected in the Flk1−/− mice. To determine if this deficiency in embryos lacking mature endothelial cells is due to the absence of circulation or to the loss of a direct signal, we again turned to an in vitro explant culture assay. In a tissue recombination experiment, dorsal endoderm explants from E9.5 Flk1−/− embryos were cultured with or without the dorsal aorta explanted from wild-type embryos. The presence of the dorsal aorta induced the expression of Ptf1a in the Flk1−/− endoderm explants (Yoshitomi & Zaret 2004). This result indicates that endothelial cells send signals to neighbouring endodermal tissue independent of the circulation to promote its specification as pancreatic tissue.

Secondary steps of pancreatic development influenced by the vasculature After the Ptf1a induction step, the dorsal aorta and dorsal pancreatic endoderm become separated by the Isl-1-positive dorsal mesenchyme (at ∼20 somites). Sphingosine-1-phosphate (S1P) present in the circulation has been shown to promote the survival of this intervening dorsal mesenchyme (Edsbagge et al 2005), which is required for proliferation of the pancreatic endoderm (Ahlgren et al 1997, Bhushan et al 2001). This study suggests that the endothelium can

ENDOTHELIUM PROMOTES LIVER & PANCREAS DEVELOPMENT

211

deliver signalling molecules important for pancreas development from the circulation that may act directly on the pancreatic bud and/or indirectly by stimulating a signalling event from local endothelium or mesenchyme. Our laboratory has shown that the endothelium plays an additional role in the development of the dorsal pancreas at this stage. That is, we found that endothelial cells release signals independent of the blood supply that promote the survival of the dorsal mesenchyme, which subsequently releases FGF10 ( Jacquemin et al 2006). FGF10, in turn, stimulates the proliferation of the pancreatic endoderm (Ahlgren et al 1997, Bhushan et al 2001). Furthermore, circulating S1P likely helps maintain the local vasculature as well as the local mesenchyme (Edsbagge et al 2005). Our laboratory is currently focusing on the earliest step in this network, namely the signal from the endothelial cells to the Pdx1-positive dorsal endoderm that induces transcription of Ptf1a and thereby completes pancreatic specification. Recent developments and future directions Endothelial cells could transduce their inductive signals to the pre-pancreatic endoderm by cell contact or by insoluble components of the endothelial cell extracellular matrix, such as has been recently described in adult pancreatic islets (Nikolova et al 2006). This latter mode may require cell contact. Alternatively, endothelial cells might release a soluble factor to the endodermal cells. To differentiate between these possibilities, we tested whether cell contact was required for the induction of Ptf1a expression in dorsal endoderm explants by endothelial cells. To accomplish this goal, we first tested whether endothelial cell lines could induce Ptf1a expression in Flk1−/− dorsal endoderm explants as was observed with the dorsal aorta co-cultures (Yoshitomi & Zaret 2004). We plated four endothelial cell lines (mouse eEND.2, mouse bEND.3, human umbilical vein endothelial cells, or human umbilical artery endothelial cells) or control cell lines (mouse 3T3 or human 293T) in either the top or bottom compartment of Transwell culture plates. After 24 hours, Flk1−/− dorsal endoderm explants were placed in the top compartment and allowed to grow for an additional 48–72 hours. Indeed, both configurations led to the induction of Ptf1a in the explants by all four endothelial cells lines, but not by the control cell lines. This result demonstrates that the endothelial cells likely release a soluble factor that can affect gene expression in the endoderm. In addition, it indicates that a functional endothelial cell signal is not specific to a particular endothelial cell type. What signalling pathways are responsible for this endothelial cell-dependent induction of Ptf1a? Our first approach to answering this question was to purify molecules released by endothelial cells that can induce Ptf1a expression. In a separate approach, we are dissecting the endothelial signalling pathway starting from

212

FREEDMAN ET AL

its genetic target. Because the end result of the endothelial signal is a change in gene expression within the endoderm, we predict that there will be an endothelial responsive element in the Ptf1a gene. We have begun to test this hypothesis by two methods. In the first, we are mapping the Ptf1a promoter with reporter constructs (gifts of Raymond MacDonald). The −12.5 kb region of Ptf1a is sufficient to direct expression of a reporter construct in pancreatic buds of E10.5 mouse embryos (Kawaguchi et al 2002). However, no expression of the reporter was seen at the time of endothelial-induced initiation of Ptf1a expression, E9.0–9.5 (Kawaguchi et al 2002). Furthermore, we have found that the −7.5 Kb region of Ptf1a directs expression of a reporter gene throughout the endoderm in transient transgenic mouse embryos at day E9.5, without restriction to the pancreatic buds. These results suggest that the regulation of Ptf1a may be complex, with possible repressive elements and pan-endodermal activating elements, in addition to endothelialresponsive elements. We are currently testing larger segments of the Ptf1a locus to try to find the regions of the gene that recapitulate its normal expression pattern. Our model also predicts that endothelial-dependent changes will be evident within the chromatin of pre-pancreatic endodermal cells. The signal transduction cascade initiated by soluble endothelial cell factors should result in transcription factor binding and chromatin changes at specific regions of the DNA within endodermal cells. We are therefore mapping endothelial-responsive alterations to DNaseI-hypersensitivity profi les in chromatin surrounding Ptf1a and other endothermal genes. We are isolating nuclei from the dorsal pancreatic buds and control tissues of wild-type and Flk1−/− E9.5 embryos. Next, limited DNaseI treatment degrades hypersensitive regions of chromatin. We are then performing ligation-mediated PCR for both region-specific and genome-wide analysis of differential chromatin changes. When we identify regions with endothelial-responsive hypersensitivity, we will evaluate relevant binding factors and track back up the pathway towards the endothelial signal. Perspectives The data from our laboratory and others lead to a model for the different signalling events required for the development of the dorsal pancreas (Fig. 1). Starting at the seven somite stage, Pdx1 expression marks the pre-pancreas, pre-stomach, and pre-duodenum midgut endoderm, under the influence of the transcription factor gene Hnf6 (Jacquemin et al 2003). At approximately 15 somites, the dorsal aorta is in direct contact with the dorsal side of the Pdx1-positive endoderm and sends signals that induce Ptf1a expression in a subset of the Pdx1-positive cells (Lammert et al 2001, Yoshitomi & Zaret 2004). These Ptf1a-expressing cells are now determined as the prospective pancreas. At approximately 20 somites, the

ENDOTHELIUM PROMOTES LIVER & PANCREAS DEVELOPMENT

circulation or S1P endothelium Flk-1 dorsal endoderm Hnf6

survival signals

213

Edsbagge et. al., 2005 Lammert et al. 2001 Yoshitomi & Zaret, 2004 Jacquemin et. al., 2006

Isl-1 mesenchyme Fgf10 Ahlgren et. al., 1997 Bhushan et. al., 2001

Pdx-1

Pdx-1 Ptf1a

7-10S 15S

20S stages

FIG. 1. A signalling network for dorsal pancreas development. Signals from the endothelium that affect pancreas development begin after initial pancreatic specification (as marked by Pdx1 expression at the seven somite stage). The completion of pancreatic specification (Ptf1a induction at 15 somites) is dependent on direct signals from endothelial cells. Survival of the dorsal mesenchyme and its subsequent release of pancreatic growth factor FGF10 are also endothelial cell-dependent, at about 20 somites. Furthermore, sphingosine-1-phosphate present in the circulation also promotes dorsal mesenchyme survival and thus indirectly promotes pancreatic growth.

dorsal aorta and dorsal pancreatic endoderm become separated by the dorsal mesenchyme. The endothelium plays a second role in the development of the dorsal pancreas at this stage, again independent of the blood supply. Endothelial cells release signals that promote the survival of the dorsal mesenchyme, which in turn releases FGF10 (Jacquemin et al 2006). FGF10 then stimulates the proliferation of the pancreatic endoderm (Ahlgren et al 1997, Bhushan et al 2001). Within this signalling network, S1P present in the circulation promotes the survival of the dorsal mesenchyme and possibly the local signalling function of the endothelium (Edsbagge et al 2005). Many classical mesodermal interactions have been known to induce these various stages of dorsal pancreatic development. We can now add endothelial cells to the list (Fig. 2). First, the lateral plate mesoderm induces early patterning in the endoderm by releasing factors such as retinoic acid (RA), bone morphogenic proteins (BMPs), and activins (Kumar et al 2003, Stafford & Prince 2002). Later, the notochord releases activin-βB, resulting in repression of Shh signalling and the expression of Pdx1 in the neighbouring endoderm (Hebrok et al 1998, Kim et al

214

FREEDMAN ET AL Mesoderm– endoderm interaction Lateral plate mesoderm Stafford et al (2002) Kumar et al (2003) Notochord Hebrok et al (1998) Kim et al (2000)

Signalling molecules

Role

RA, BMPs, activins early patterning

activin-bB

repress Shh; enable Pdx1+

Aortic endothelial cells Lammert et al (2001) Yoshitomi & Zaret (2004)

???

Ptf1a induction

Circulating signals Edsbagge et al (2005)

S1P

survival of dorsal mesenchyme and vasculature

Isl1+ dorsal mesenchyme Bhushan et al (2001) Jacquemin et al (2006)

FGF10

maintain Pdx1 & promote bud outgrowth

FIG. 2. Mesodermal interactions that induce dorsal pancreatic endoderm. We now add endothelial cells to the list of known inducers of dorsal pancreatic cell fates. Earlier signals include retinoic acid (RA), bone morphogenic proteins (BMPs), and activins released from lateral plate mesoderm that promote patterning of the pre-pancreatic endoderm (Kumar et al 2003, Stafford & Prince 2002). Next the notochord releases activin- βB, which represses sonic hedgehog expression and enables Pdx1 expression (Hebrok et al 1998, Kim et al 2000). Signals that follow endothelial signals include S1P from the circulation and FGF10 from the dorsal mesenchyme (Edsbagge et al 2005, Jacquemin et al 2006). Notably, the endothelial cell signalling molecule required for Ptf1a induction has not yet been identified.

2000). Endothelial cells then release as yet unidentified signalling molecules to complete specification of the pancreatic endoderm by inducing Ptf1a expression (Lammert et al 2001, Yoshitomi & Zaret 2004). Later, S1P from the circulation and unknown factors released by endothelial cells promote the survival of the dorsal mesenchyme, which in turn releases FGF10 to maintain Pdx1 expression and promote dorsal pancreatic bud outgrowth (Bhushan et al 2001, Edsbagge et al 2005, Jacquemin et al 2006). There is enormous predictive value to our new understanding of endothelial cell signalling during development. For instance, one possible application is to use endothelial cells to help differentiate embryonic stem cells in culture. Some laboratories have already seen some success with this approach on hepatocyte differentiation, and we are eager to assess the effects of endothelial cell products on the differentiation of new insulin-producing β -cells.

ENDOTHELIUM PROMOTES LIVER & PANCREAS DEVELOPMENT

215

Summary of conclusions Endothelial cells send signals to the endoderm during embryonic development. These signals can induce growth and morphogenesis, as is the case with the developing liver. In addition, these endothelial signals can promote tissue differentiation, as with the developing dorsal pancreas. Our goal is to identify the endothelial-derived factors and subsequent signal transduction pathways involved in these inductive events within the endoderm. Acknowledgements We thank Dr Raymond MacDonald for Ptf1a reporter plasmids and Dr Janet Rossant for the Flk1−/− mice. We also thank Xiang Hua of the Fox Chase Cancer Center Transgenic Mouse Facility for generation of transient transgenic mice. Work on this project has been supported by grants from the NIH to KSZ (R01GM36477 and U01DK072503).

References Ahlgren U, Pfaff SL, Jessell TM, Edlund T, Edlund H 1997 Independent requirement for ISL1 in formation of pancreatic mesenchyme and islet cells. Nature 385:257–260 Bhushan A, Itoh N, Kato S et al 2001 Fgf10 is essential for maintaining the proliferative capacity of epithelial progenitor cells during early pancreatic organogenesis. Development 128:5109–5117 Edsbagge J, Johansson JK, Esni F, Luo Y, Radice GL, Semb H 2005 Vascular function and sphingosine-1-phosphate regulate development of the dorsal pancreatic mesenchyme. Development 132:1085–1092 Fell PE, Grobstein C 1968 The influence of extra-epithelial factors on the growth of embryonic mouse pancreatic epithelium. Exp Cell Res 53:301–304 Gannon M, Wright CVE 1999 Endodermal patterning and organogenesis. In: Moody SA (ed) Cell lineage and fate determination. Academic Press, p 583–615 Golosow N, Grobstein C 1962 Epitheliomesenchymal interaction in pancreatic morphogenesis. Dev Biol 4:242–255 Hebrok M, Kim SK, Melton DA 1998 Notochord repression of endodermal Sonic hedgehog permits pancreas development. Genes Dev 12:1705–1713 Jacquemin P, Lemaigre FP, Rousseau GG 2003 The onecut transcription factor HNF-6 (OC-1) is required for timely specification of the pancreas and acts upstream of Pdx-1 in the specification cascade. Dev Biol 258:105–116 Jacquemin P, Yoshitomi H, Kashima Y, Rousseau GG, Lemaigre FP, Zaret KS 2006 An endothelial-mesenchymal relay pathway regulates early phases of pancreas development. Dev Biol 290:189–199 Kawaguchi Y, Cooper B, Gannon M, Ray M, MacDonald RJ, Wright CV 2002 The role of the transcriptional regulator Ptf1a in converting intestinal to pancreatic progenitors. Nat Genet 32:128–134 Kim SK, Hebrok M, Li E et al 2000 Activin receptor patterning of foregut organogenesis. Genes Dev 14:1866–1871 Kumar M, Jordan N, Melton D, Grapin-Botton A 2003 Signals from lateral plate mesoderm instruct endoderm toward a pancreatic fate. Dev Biol 259:109–122 Lammert E, Cleaver O, Melton D 2001 Induction of pancreatic differentiation by signals from blood vessels. Science 294:564–567

216

FREEDMAN ET AL

Le Douarin N 1975 An experimental analysis of liver development. Med Biol 53:427–455 LeCouter J, Moritz DR, Li B et al 2003 Angiogenesis-independent endothelial protection of liver: role of VEGFR-1. Science 299:890–893 Matsumoto K, Yoshitomi H, Rossant J, Zaret KS 2001 Liver organogenesis promoted by endothelial cells prior to vascular function. Science 294:559–563 Nikolova G, Jabs N, Konstantinova I et al 2006 The vascular basement membrane: a niche for insulin gene expression and Beta cell proliferation. Dev Cell 10:397–405 Shalaby F, Rossant J, Yamaguchi TP et al 1995 Failure of blood-island formation and vasculogenesis in Flk-1-deficient mice. Nature 376:62–66 Stafford D, Prince VE 2002 Retinoic acid signaling is required for a critical early step in zebrafish pancreatic development. Curr Biol 12:1215–1220 Yoshitomi H, Zaret KS 2004 Endothelial cell interactions initiate dorsal pancreas development by selectively inducing the transcription factor Ptf1a. Development 131:807–817

DISCUSSION Kitajewski: Is there no smooth muscle cell (SMC) contribution to any of the inductive signals? Freedman: SMCs begin to develop around the dorsal aorta at E9.5. The E9.0 dorsal aorta, which is providing inductive signals to the pre-pancreatic endoderm, is a single sheet of endothelial cells. In the case of liver development, the inductive signals appear to come from angioblasts that are just beginning to form capillaries and are devoid of SMCs. Owens: This is a little bit early for SMCs to be playing a role but we can’t completely rule it out. On the other hand, SMCs and pericyte progenitor cells may contribute in some way. Kitajewski: It might be useful to use cultured SMCs, just to confirm that you don’t get an inductive signal from SMCs, as one of your negative controls. Drake: We looked at a similar inductive interaction in heart development. Another angle you might want to think about is, what is the primordium making that the endothelial cells would like? We found in the heart the myocardial cells are making things like VEGF and angiopoietin. The Flk receptor is clearly present so you might predict that the bud is making VEGF. Wilting: You started your talk by saying that the dorsal pancreas develops close to the aorta and the liver close to the vitelline vein, giving the expectation that you would show specific inductive defects by arterial and venous endothelial cells. However, in a later slide you showed inductive effects on the pancreas by all the endothelial cell types you have been using. Freedman: That is true. I, too, find it surprising that all of the endothelial cell lines are able to induce Ptf1a expression in the dorsal pancreatic explants. In the context of the embryos, however, our model that the signals come from the dorsal

ENDOTHELIUM PROMOTES LIVER & PANCREAS DEVELOPMENT

217

aorta is based on what specific endothelial cells are present. There is no major vein near the dorsal pancreas at E9.0. Only the dorsal aorta is present at this stage, and it is in direct contact with the pre-pancreatic endoderm. In the case of liver development, it is angioblasts and nascent capillaries that are present in the mesenchyme into which the hepatic endoderm buds. We believe that these capillary endothelial cells are sending the morphogenic signals to the endoderm. We do not yet know if, in the embryo, the signalling molecules are specifically made by arterial or venous endothelial cells. Owens: I have a question about your ChIP (chromatin immunoprecipitation) on chip experiments. What are you using in the pull down? Is it a histone H4 acetylation antibody? Freedman: I did not present any ChIP on chip experiments, although our laboratory has looked at histone acetylation, as well as transcription factors in such experiments. The plan that I presented is to do a DNase I hypersensitivity assay on chip. Owens: How much material can you get? Freedman: On a good day, I can get about 1 µg of DNA from five litters’ worth of dorsal pancreatic buds; often I get as low as 400 ng. It is not a lot. This is the main reason that we plan to move the assay to a microarray format. We have been able to look at DNaseI hypersensitivity at individual sites in the Ptf1a locus in E9.5 pancreatic buds. The idea of this approach is that we create a break in the chromatin with DNaseI, where the chromatin is more accessible to the enzyme due to the binding of a transcription factor, for instance. Then we ligate a specific linker to this newly generated end. We then use a primer to this linker and a primer to a region in Ptf1a to PCR a small fragment that we can see on a gel. 400 ng is enough to do one PCR reaction. Although we have had technical success with this technique, we have not found changes that are specific to endothelial signalling in the pancreatic bud by looking at sites one at a time. To move to genome-wide analysis, we will use the method of ligation-mediated PCR to amplify all the DNA hypersensitive sites in the entire cell population. We will then hybridize these products to a microarray that tiles across many kilobases surrounding endodermal genes. We hope to see hypersensitive regions that are specific to the dorsal pancreas region in wild-type mice versus Flk1-deficient mice. Owens: That’s very impressive and a huge amount of work. It would be interesting to compare your data with similar work people are doing with frogs in this area. I know they are attempting to use high throughput proteomic approaches. One of the advantages with frogs is that you can get a lot of synchronized embryos. The major advantage you have is that mouse genetics are far superior. Lammert: You have contradictory results about endothelial cell-derived signals for embryogenesis. These signals are well demonstrated in the mouse field. But in

218

FREEDMAN ET AL

zebrafish, in the cloche mutants, the effects of endothelial cells on organogenesis are extremely mild or do not exist. How do you reconcile the mouse and zebrafish results? Freedman: The mouse pancreas buds in two places: the ventral side and the dorsal side. They later join into one organ. Ptf1a induction in ventral pancreas occurs quite normally in the Flk1-null background. For some reason pancreas development has been split into dorsal and ventral buds during evolution, and the ventral side is more similar to what is seen in zebrafish. Shibuya: I was surprised that most of the endothelial cells in conditioned medium can equally induce Ptf1a. I would like to think that some kinds of endothelial cells may produce more of this inducer. Is it possible to transplant these endothelial cells into different dorsal regions to see which position is more effective in inducing Ptf1a? Freedman: Technically, this is not currently possible in the mouse. We can do some whole-embryo culture, but only if we do not disturb the yolk sac. Gerhardt: You mentioned Eckhard Lammert’s data on the contribution of the matrix. You have shown that endothelial cells produce a soluble factor. Are the two mutually exclusive? Endothelial cells do produce quite a lot of soluble fragments of matrix. Freedman: These two ideas are definitely not mutually exclusive. The ECM could certainly be playing a role. As you say, soluble fragments of ECM could be released by the endothelial cells. Furthermore, we have used E9.5 dorsal pancreatic explants for our endothelial-cell co-culture experiments. There could be indirect effects: the signalling factor from endothelial cells could be inducing matrix changes in other cell types that affect Ptf1a induction in the endoderm. We do not think however that such an effect is through the mesenchyme, because these are Flk1null explants and the dorsal mesenchyme is apparently absent. Lammert: We looked at the pancreatic islets. We asked why every β cell was in direct contact with an endothelial cell, and we then came up with the extracellular matrix that the endothelial cells produce. Actually, VEGF induces capillary formation within pancreatic islets, and these capillaries provide pancreatic β cells with a vascular basement membrane. This is at a late stage of islet development. This mechanism might also play a role when an islet grows to adapt to obesity and insulin resistance. We don’t think that the matrix is an inductive signal. It is creating a permissive environment to allow tissue development. Ye: In the cloche zebrafish mutant, Didier Stanier later found that there are some residual endothelial cells left (presentation at the NAVBO workshop, Asilomar, California, February 2004), therefore, it is possible that there are still some endothelial cells present that could interact with the liver primordium. Weinstein: It is complicated. There is some residual endothelium present, although when I talked to him about it he said it is not in that area where the pancreas is

ENDOTHELIUM PROMOTES LIVER & PANCREAS DEVELOPMENT

219

forming. It is not clear where things are arresting in cloche mutants. The expression of some early markers of the angioblast lineage is common between endothelium and haematopoietic cells, so it is not clear where these cells are going wrong. They may already be expressing some endothelial markers. Ye: In your E9.5 endoderm, do you see a small population of cells that express Ptf1a? Freedman: We never see Ptf1a come on in the Flk1-null dorsal pancreatic explants by RT-PCR. In the wild-type buds, Ptf1a is first weakly detectable at around 15 somites, and at 18 somites we can see a nice signal. It seems to be the initial induction of Ptf1a dorsally that is dependent on the presence of endothelial cells.

Embryonic development and malformation of lymphatic vessels Jörg Wilting, Kerstin Buttler, Jochen Rössler*, Susanne Norgall†, Lothar Schweigerer, Herbert A. Weich† and Maria Papoutsi Department of Pediatrics 1, Georg-August-University, Goettingen, * Department of Pediatrics, AlbertLudwigs-Universität, Freiburg and † Department Gene Regulation and Differentiation, German Research Centre for Biotechnolog y, Braunschweig, Germany

Abstract. In the human, malformations of lymphatic vessels can be observed as lymphangiectasia, lymphangioma and lymphangiomatosis, with a prevalence of 1.2–2.8‰. Their aetiology is unknown and a causal therapy does not exist. We investigated the origin of lymphatic endothelial cells (LECs) in avian and murine embryos, and compared the molecular profi le of LECs from normal and malformed lymphatics of children. In avian embryos, Prox1+ lymphangioblasts are located in the confluence of the cranial and caudal cardinal veins, where the jugular lymph sac ( JLS) forms. Cell lineage studies show that the JLS is of venous origin. In contrast, the lymphatics of the dermis are derived from mesenchymal lymphangioblasts located in the dermatomes, suggesting a dual origin of LECs in avian embryos. The same may hold true for murine embryos, where Lyve1+ LEC precursors are found in the cardinal veins, and in the mesenchyme. The mesenchymal cells express the pan-leukocyte marker CD45, indicating a cell type with lymphendothelial and leukocyte characteristics. In the human, such cells might give rise to Kaposi’s sarcoma. Microarray analyses of LECs from lymphangiomas of children show a large number of regulated genes, such as VEGFR3. Our studies show that lymphvasculogenesis and lymphangiogenesis occur simultaneously in the embryo, and suggest a function for VEGFR3 in lymphangiomas. 2007 Vascular development. Wiley, Chichester (Novartis Foundation Symposium 283) p 220–229

The Novartis Foundation symposium Vascular development has discussed the function of vascular endothelial cells (ECs) during early embryonic development, organogenesis and abnormal tissue development such as tumorigenesis. The vascular system of higher vertebrates consists of two types of vessels: blood vessels and lymph vessels (hemangion vs. lymphangion). In the human, the total blood vessel length is approximately 100 000 km, and the length of the lymph vessels is not much less, since all organs except the CNS, bone marrow and placenta are densely supplied by them. Whereas the main function of the blood vascular system is a nutritive one, lymphatics drain excess fluid from the interstitium (2–4l per day) 220

DEVELOPMENT AND MALFORMATION OF LYMPHATIC VESSELS

221

and serve as highways for circulating leukocytes. Not only leukocytes, but also tumour cells migrate along lymphatic vessels. It has been suggested that about 80% of carcinomas metastasize via lymph vessels (Alitalo et al 2005, Witte et al 2006). Tumour-induced lymphangiogenesis, which has long been denied, obviously takes place from pre-existing lymphatics, whereas the mechanisms of embryonic lymphangiogenesis are still poorly defined. Therefore, the aetiology of lymphangiodysplasias (aplasia, hypoplasia, hyperplasia) has remained unknown and there are no causal therapeutic strategies for children suffering from such dysplasias. Lymphatic malformations can be observed as lymphangiectasia, lymphangioma and lymphangiomatosis, with a prevalence of 1.2–2.8‰ (Filston 1994). The origin of the malformations from lymphatic capillaries, collectors or trunks is not defined, as are the mechanisms of embryonic lymphangiogenesis. For the development of the blood vascular system it is well accepted that various mechanisms take place almost simultaneously. These have been termed vasculogenesis and angiogenesis: the former referring to the development of blood vascular endothelial cells (BECs) from haemangioblastic precursor cells, the latter to the development of vessels from already existing ones by sprouting or intercalating growth. Lymph vessels have mostly been described to develop exclusively by sprouting from specific parts of deep embryonic veins, whereas development of LECs from lymphangioblasts (lymphvasculogenesis) has only been observed in amphibians and birds (Ny et al 2005, Schneider et al 1999, Wilting et al 2006). The mechanisms of embryonic lymphangiogenesis in higher vertebrates and their molecular control are poorly defined. Therefore, basic knowledge is missing that may help to explain the aetiology of lymphatic malformations in the human, as a prerequisite for therapeutic interventions. We have studied the origin of lymphatic endothelial cells (LECs) in avian and murine embryos, and compared the molecular profi les of LECs from normal and malformed lymphatics of children. Our data provide evidence for a dual origin of LECs from specific embryonic veins and from mesenchymal lymphangioblasts. Furthermore, we have identified regulated genes in LECs from human lymphangiomas; among these, up-regulation of vascular endothelial growth factor receptor 3 (VEGFR3) may be a key regulator of lymphangiomas. Experimental procedures Experimental procedures have been described in detail recently (Buttler et al 2006, Wilting et al 2006, Norgall et al 2006). (1) Studies on avian embryos consisted of intravascular injection of 5 µl DiI-conjugated acetylated low-density lipoprotein (DiI-acLDL) into day-4 embryos to label ECs in vivo. Quail–chick grafting experiments were carried out in order to study the origin of the jugular lymph sacs ( JLS) and the dermal lymphatics. In situ hybridization was performed with Prox1 and Tie2 probes. Immunohistological studies were performed with antibodies

222

WILTING ET AL

against BECs and LECs. (2) Studies on murine embryos consisted of immunohistochemisty and immunofluorescence studies of murine embryos from embryonic day (ED) 9.5–13.5. (3) Studies on human endothelial cells consisted of selection and culture of blood vascular and lymphatic endothelial cells, VEGFR3 ELISA, immunocytochemistry and FACS analyses, microarray analysis with 44 k whole genome oligo microarrays (Agilent, Waldbronn, Germany) and immunohistology of lymphangioma tissues. Results Our descriptive and experimental studies on avian embryos strongly suggest a dual origin of lymph vessels. The transcription factor Prox1, which is indispensable for embryonic development of lymphatics in mice (Wigle & Oliver 1999), is expressed in the jugular segment of embryonic veins (Fig. 1A). The jugular region is located at the level of somites 9–11 and is demarcated by the confluence of the cranial and caudal cardinal veins. A considerable time thereafter, at ED 6.5, the JLS can be seen in histological sections, and represent the first morphological evidence of the lymphatic vascular system. In order to study the contribution of venous ECs to the JLS, we injected DiI-acLDL into the vitelline vein of day 4 quail embryos, before lymph sacs are present, and re-incubated the embryos until day 6.5. DiIacLDL is an in vivo marker of ECs, macrophages and hepatoblasts. With the QH1 antibody the region of interest, aortic arch, jugular vein and JLS, were identified

FIG. 1. In situ hybridization of a day 4 quail embryo (stage 20 HH) with a Prox1 probe. (A) Jugular region showing the confluence of the cranial and caudal cardinal veins into the common cardinal vein (arrow). This segment of the venous system is Prox1-positive. (B) Same embryo as in (A), showing inter-limb level. Note scattered Prox1-positive cells in the dermatomes. Spinal and sympathetic ganglia also express Prox1.

DEVELOPMENT AND MALFORMATION OF LYMPHATIC VESSELS

223

and in the same section, the fine granular DiI-acLDL staining could be seen in a number of LECs of the JLS. The data suggest a contribution of early venous ECs to the JLS (Wilting et al 2006). In addition, we observed numerous scattered Prox1-positive cells in the dermatomes all along the body axis (Fig. 1B). These cells co-express QH1, a haemangioblast marker identified some time ago (Pardanaud et al 1987). We grafted dermatomes from inter-limb levels from day 4 quail embryos into the wing region of corresponding chick embryos and re-incubated the hosts until day 9. Double staining with Prox1 (to identify lymph vessels) and QH1 (to identify quail ECs) clearly showed integration of quail LECs into the chick lymph vessels (Fig. 2). This shows lymphangiogenic potential of early dermatomes and suggests that scattered Prox1/QH1-positive cells represent lymphangioblasts. Comparable to the findings in avian embryos we also observed cells expressing LEC markers in two different compartments of murine embryos: in specific veins and in scattered cells in the mesenchyme. We studied with immunofluorescence the expression of a variety of markers in NMRI mice from ED 9.5–13.5: CD31, which detects both BECs and LECs; CD45, which is a marker of leukocytes; Prox1 and Lyve1 (lymphatic vessel endothelial hyaluronan receptor 1), which are highly specific for LECs; CD11b, which is a marker of macrophages; and the proliferation marker Ki67. The earliest marker for developing LECs seems to be Lyve1, which is expressed in the jugular segment of the cardinal veins and in a segment of the vitelline vein in ED 9.5 embryos. Additionally, an increasing number of scattered Lyve1-positive cells can be observed in the mesenchyme during the following days. The central nervous system, which remains free of lymphatics, does not contain such cells. The scattered Lyve1-positive cells are highly proliferative and only rarely co-express the macrophage marker CD11b. All of the cells co-express CD45. This

FIG. 2. Grafting of day 4 quail dermatome into a corresponding chick embryo. Re-incubation 5 days and double staining with Prox1 and QH1 antibodies. Note integration of QH1-positive quail LECs into the chick lymph vessel.

224

WILTING ET AL

seems to be down-regulated when lymph vessels are formed, while CD31 appears to be up-regulated. The LEC marker Prox1 is expressed in a subpopulation of scattered Lyve1-positive cells, and appears to be a later marker of the cell lineage. Our data show the existence of a cell type, which combines leukocyte and lymphendothelial characteristics, and, as discussed below, we assume that such a cell type is also present in the human. In order to identify molecules and mechanisms that regulate lymphangiogenesis in the human, we have isolated LECs from normal and diseased human tissues. We have characterized LECs from normal dermis and from two children suffering from lymphangioma of the upper arm and the axillary region. We compared the LECs with BECs from umbilical vein, aorta and myometrial microvessels. Initially isolated with anti-PECAM-1 antibodies, the LECs were separated by FACS sorting and magnetic beads using anti-podoplanin and/or Lyve1 antibodies. Characterization was performed by FACS analysis, immunofluorescence staining, ELISA and microarray gene analysis. LECs from foreskin and lymphangioma had an almost identical pattern of lymphendothelial markers such as podoplanin, Prox1, reelin, cMaf and integrin α1 and α9; however, Lyve1 was down-regulated and VEGFR3 was up-regulated in lymphangiomas (Table 1). Additionally, we observed high levels of VEGFR2 in lymphangioma LECs. In conclusion, LECs from different sources express slightly variable cell surface markers, but can always be distinguished from BECs by their Prox1 expression. Lymphangioma LECs seem to represent an activated type of LECs, and the high expression levels of VEGFR3 and -2 may contribute to the aetiology of lymphangiomas.

TABLE 1 VEGFR3 quantification in endothelial cell lysates by ELISA Cell type HUVEC HAEC HDMEC HDMEC Lyve1+ LEC patient 1 LEC patient 2

Total protein (mg/ml)

VEGFR3 (ng/ml)

VEGFR3 (ng/mg protein)

0.67 1.09 0.36 0.53

1.77 1.51 2.78 10.25

2.65 ± 0.06 1.38 ± 0.03 7.73 ± 0.61 19.34 ± 0.75

0.60 0.42

17.71 17.15

29.51 ± 2.25 40.84 ± 1.66

VEGFR3 protein in various endothelial cell types was measured by ELISA. Cell lysates were prepared from confluent cells in vitro and measured in triplicate. The mean of VEGFR3 protein concentration in ng/mg total protein (±SD) is shown.

DEVELOPMENT AND MALFORMATION OF LYMPHATIC VESSELS

225

Discussion Recent studies on murine embryos have shown that the first cells committed to the lymphendothelial lineage are located in the endothelial lining of the anterior cardinal vein in the jugular region (Wigle & Oliver 1999, Wigle et al 2002). The ECs of the cardinal vein express VEGFR3, Lyve1, a transmembrane glycoprotein (Banerji et al 1999) and CD31 (PECAM1). Only a subset of them is Prox1-positive. These cells are located in the dorso-lateral part of the vein and seem to sprout into the dorso-lateral mesoderm, where they form the JLS. At ED 14.5 sprouts of the JLS grow toward the periphery and enter different organs (Wigle et al 2002). Mice deficient for VEGF-C, the ligand of VEGFR3, do not form lymph sacs, which can be rescued by the application of VEGF-C (Karkkainen et al 2004). These studies are in line with the traditional view of the development of LECs from specific parts of the venous system. Our studies on avian embryos suggest a dual origin of LECs (Wilting et al 2006). Intravenous application of DiI-LDL, which labels endothelial cells, into day 4 embryos, has revealed labelling of the JLS at day 6.5. This suggests a venous origin of the JLS. Additionally, quail–chick grafting of paraxial mesoderm has shown integration of graft-derived LECs into the superficial parts of the JLS. The lymphatics of the dermis are obviously derived directly from the dermatomes. Lymphangioblasts have recently also been observed in Xenopus tadpoles (Ny et al 2005). In accordance with previous studies, our studies on murine embryos show that the earliest signs of lymphendothelial commitment can be observed in specific parts of the venous system, i.e. the jugular segment of the cardinal veins and the vitelline vein. However, in contrast to previous studies we have observed scattered mesenchymal cells with lymphendothelial characteristics. An increasing number of Lyve1/CD31 and Lyve1/CD45-positive cells can be found in the mesenchyme of the dermis, the mediastinum and the primitive meninx, but not in the CNS, which is free of lymphatics. Only a small number of the cells co-expresses the macrophage marker CD11b. The data are indicative of mesenchymal lymphangioblasts in murine embryos, and they clearly show the existence of cells that coexpress leukocyte and lymphendothelial characteristics. Evidence for the existence of a similar cell type in the human has been published recently by Barozzi et al (2003), who have studied post-transplantation Kaposi’s sarcoma (KS) patients. The cell of origin of KS is not definitely known, but recent studies on the expression of lymphendothelial markers, VEGFR3, podoplanin and Lyve1, and studies showing the mitogenic effect of VEGF-C in KS cells, seem to prove that KS is derived from LECs (reviewed in Cheung & Rockson 2005). In nearly all of the KS lesions investigated, Chang et al (1994) have discovered KS-associated herpes virus (KSHV, HHV8), a member of the gamma herpes virus family, which are often associated with lymphoproliferative disorders. Examination of biopsies from

226

WILTING ET AL

KS skin lesions, which developed 9–40 months after renal transplantation in six HIV-negative female and two male recipients of kidneys from male donors, have shown that a high proportion of KS cells from these patients were of donor origin (Barozzi et al 2003). Furthermore, characteristic KS cells can be isolated from peripheral blood of HIV patients (Browning et al 1994). These findings suggest that KS cells are derived from highly mobile leukocyte-like cells, which express lymphendothelial markers. Malformations of lymphatic vessels, such as lymphangiomas, are associated with a failure in lymph transport and high patient morbidity (Kennedy et al 2001). The aetiology of lymphangiomas is unknown and a causal therapy does not exist. Our microarray gene analyses have identified a large number of genes specifically regulated in lymphangioma LECs; among these were VEGFR2 and -3. With ELISA we have studied VEGFR3 expression at protein level and observed a significant up-regulation of this receptor in lymphangioma LECs as compared to normal LECs. In BECs, VEGFR3 is expressed at very low levels only. High expression levels of VEGFR2 and -3 in lymphangioma LECs show that the cells represent an activated type of lymphatic endothelium, which may be comparable to LECs in the vicinity of solid tumours. Acknowledgements The studies have been supported by grants from the German Research Council (DFG) to JW and HAW (SPP 1069).

References Alitalo K, Tammela T, Petrova TV 2005 Lymphangiogenesis in development and human disease. Nature 438:946–953 Banerji S, Ni J, Wang SX et al 1999 LYVE-1, a new homologue of the CD44 glycoprotein, is a lymph-specific receptor for hyaluronan. J Cell Biol 144:789–801 Barozzi P, Luppi M, Facchetti F et al 2003 Post-transplant Kaposi sarcoma originates from the seeding of donor-derived progenitors. Nat Med 9:554–561 Browning PJ, Sechler JM, Kaplan M et al 1994 Identification and culture of Kaposi’s sarcomalike spindle cells from the peripheral blood of human immunodeficiency virus-1-infected individuals and normal controls. Blood 84:2711–2720 Buttler K, Kreysing A, von Kaisenberg CS et al 2006 Mesenchymal cells with leukocyte and lymphendothelial characteristics in murine embryos. Dev Dyn 235:1554–1562 Chang Y, Cesarman E, Pessin MS et al 1994 Identification of herpesvirus-like DNA sequences in AIDS-associated Kaposi’s sarcoma. Science 266:1865–1869 Cheung L, Rockson SG 2005 The lymphatic biology of Kaposi’s sarcoma. Lymphat Res Biol 3:25–35 Filston HC 1994 Hemangiomas, cystic hygromas, and teratomas of the head and neck. Semin Pediatr Surg 3:147–159 Karkkainen MJ, Haiko P, Sainio K et al 2004 Vascular endothelial growth factor C is required for sprouting of the fi rst lymphatic vessels from embryonic veins. Nat Immunol 5:74–80

DEVELOPMENT AND MALFORMATION OF LYMPHATIC VESSELS

227

Kennedy TL, Whitaker M, Pellitteri P, Wood WE 2001 Cystic hygroma/lymphangioma: a rational approach to managment. Laryngoscope 111:1929–1937 Norgall S, Papoutsi M, Rössler J, Schweigerer L, Wilting J, Weich HA 2007 Elevated expression of VEGFR-3 in lymphatic endothelial cells from lymphangiomas. BMC Cancer, in press Ny A, Koch M, Schneider M et al 2005 A genetic Xenopus laevis tadpole model to study lymphangiogenesis. Nat Med 11:998–1004 Pardanaud L, Altmann C, Kitos P, Dieterlen-Lièvre F 1987 Vasculogenesis in the early quail blastodisc as studied with a monoclonal antibody recognizing endothelial cells. Development 100:339–349 Schneider M, Othman-Hassan K, Christ B, Wilting J 1999 Lymphangioblasts in the avian wing bud. Dev Dyn 216:311–319 Wigle JT, Oliver G 1999 Prox1 function is required for the development of the murine lymphatic system. Cell 98:769–778 Wigle JT, Harvey N, Detmar M et al 2002 An essential role for Prox1 in the induction of the lymphatic endothelial cell phenotype. EMBO J 21:1505–1513 Wilting J, Aref Y, Huang R et al 2006 Dual origin of avian lymphatics. Dev Biol 292: 165–173 Witte MH, Jones K, Wilting J et al 2006 Structure-function relationships in the lymphatic system and implications for cancer biology. Cancer Met Rev 25:159–184

DISCUSSION Lammert: Are the leukocytes Lyve1 positive? Did you also check for Prox1? Wilting: All we can say from these descriptive data is that there is a large population of Lyve1-positive leukocytes. They practically do not appear in the CNS, which is free of lymphatics, and they are numerous in regions such as the skin, which are rich in lymphatics. This is indirect evidence that there might be lymphangioblasts in the mice also. There is indirect evidence in the human that there is at least a cell type that is highly mobile and which expresses lymphendothelial markers and may give rise to post-transplantation Kaposi’s sarcoma (Barozzi et al 2003). Lammert: Are they Prox1 positive? Wilting: A subpopulation of these scattered Lyve1-positive cells is Prox1 positive. Lammert: Why are some not? Wilting: Lyve1 is an earlier marker than Prox1. Lammert: Can you sort these cells and then try in vitro to induce them to form lymph vessels, to make sure that they aren’t just leukocytes that express a lot of Fc receptors? Wilting: They are a form of leukocyte as they have CD45. I don’t know what will happen to this CD45 expression: perhaps this is down-regulated. You are right: we need to isolate this cell and study its behaviour. Gerhardt: I was curious about your grafting experiment. You took a region with scattered Prox1-positive cells. Is it possible to take regions without these cells? Do

228

WILTING ET AL

you think these cells somehow need to be pre-committed, or could you take cells from other regions that don’t express Prox1, graft them and see induction of Prox1 incorporation? Wilting: I think there is a Prox1-negative phase of lymphangiogenesis. When we look at the allantoic bud on day 3, we don’t find Prox1-positive cells. On day 4 it has expanded a bit and there are scattered Prox1-positive cells in the allantois. If you take the allantoic bud from a day 3 embryo and graft it, lymphatics will develop from the Prox1-negative precursors. Also, if an early Prox1-negative somite is taken from an embryo and grafted, lymphatic vessels will develop. Drake: We have observed cells within the somite that we considered to be angioblasts. Do you think angioblasts are induced to a lymphatic fate? If you took angioblasts from somewhere else and put them in the dermamyotome region of the somite, would some of those be selected to express Prox1 and become lymphangioblasts? Wilting: The question is, do these lineages come together? They obviously come together at some part of the body. There are specific parts where the two vascular systems fuse and lympho–venous anastomeses are formed. This is the most complicated part. How do you manage to fuse two vascular systems that are usually not fused? Some of the angioblasts may have an intermediate phenotype and form lymphatic endothelial cells, mainly in the regions where lympho–venous communications are formed. Weinstein: This is a potentially separate issue from what you are saying, which is, is there a lineage relationship between these cell types earlier on. I guess what you are really asking is whether a lymphangioblast is coming from an angioblast. Drake: We could isolate angioblasts from the mouse embryo and then inject them into the somite. By doing this, we could actually determine whether the local milieu of the somite is turning on Prox, as the injected angioblasts do not normally express this. Wilting: There is not ‘the’ lymphangioblast there; there are lymphangioblasts located in the Prox1-positive segments of the early venous system and there may be lympangioblasts located as scattered Prox1-positive cells in the mesenchyme. Drake: Another relevant experiment is one that we have conducted in which a single haematopoietic cell is engrafted into an adult mouse. In the context of endothelial lineage being derived from a common haematopoietic/endothelial cell progenitor, we have done this transplantation and then looked for haematopoieticderived endothelial cells following heart attacks, strokes and tumours. We have never observed haematopoietic-derived endothelial cells. A similar experiment could be done with engrafted mice to evaluate lymphangiogenesis. Are haematopoietic-derived leukocytes incorporated into the lymph system? Wilting: When you are talking about haematopoietic cells you are usually thinking of cells that are residing in the bone marrow. This is one of the rare places

DEVELOPMENT AND MALFORMATION OF LYMPHATIC VESSELS

229

where you do not find lymph vessels. Early in the embryo there are haematopoietic regions in many tissues. I don’t know how many of these cells will reside in the loose connective tissue, but there is evidence that scattered cells integrate into embryonic lymph vessels. Drake: In an adult mouse, is there something you can do to create excessive lymph development? Wilting: You can remove parts of the lymphatic vascular system and look at regeneration. Mäkinen: This has been shown by the induction of excessive tumour lymphangiogenesis. In the study by He et al (2004) they couldn’t see bone marrow-derived cells incorporating into the lymphatic vasculature around tumours. Drake: So it sounds like it is the endothelial lineage. Shibuya: You showed expression of VEGF-C and VEGFR3 in the lymphangioma cells. Is the growth of this lymphangioma depending on interleukin (IL)3 and VEGF-C signalling, or does this receptor remain only as a kind of differentiation maker? Wilting: When we grow the cells in culture we don’t necessarily have to add VEGF-C to the medium. For the development of the pathology it might be relevant that VEGFR3 is so highly expressed. Weinstein: If you were to do similar kinds of somite transfers in the fish you would be carrying along the parachordal vessels with the tissue. This is a potential complication of this sort of experiment. One of the ways we have found to reconcile these results is that when you are doing somite transplant you may be taking along some of these progenitors. Wilting: We have done this experiment with somites and pieces of somites; dorsal half, ventral half. If you isolate the dorsal half of an early epithelial somite, there will be no vessel included, and still you will get lymphatics out of this. References Barozzi P, Luppi M, Facchetti F et al 2003 Post-transplant Kaposi sarcoma originates from the seeding of donor-derived progenitors. Nat Med 9:554–561 He Y, Rajantie I, Ilmonen M et al 2004 Preexisting lymphatic endothelium but not endothelial progenitor cells are essential for tumor lymphangiogenesis and lymphatic metastasis. Cancer Res 64:3737–3740

Role of the neuropilin ligands VEGF164 and SEMA3A in neuronal and vascular patterning in the mouse Joaquim Miguel Vieira, Quenten Schwarz and Christiana Ruhrberg1 Institute of Ophthalmolog y, University College London, 11–43 Bath Street, London EC1V 9EL, UK

Abstract. Blood vessels and neurons use similar guidance cues to control their behaviour during embryogenesis. The semaphorin SEMA3A was originally identified as a repulsive cue for developing axons that acts by signalling through receptor complexes containing NRP1 and A-type plexins. SEMA3A also competes with the VEGF164 isoform of vascular endothelial growth factor for binding to NRP1 to modulate the migration of endothelial cells in vitro. Surprisingly, we have found that SEMA3A and semaphorin signalling through NRP1 were not required for blood vessel development in the mouse. Moreover, we found that there was no genetic interaction between SEMA3A and VEGF164 during vasculogenesis or angiogenesis. Our observations suggest that in vivo vascular NRP1 preferentially confers VEGF164 signals, whilst axonal NRP1 preferentially transmits SEMA3A signals. 2007 Vascular development. Wiley, Chichester (Novartis Foundation Symposium 283) p 230–237

Developing blood vessels and axons branch throughout the vertebrate body and employ similar cell biological mechanisms to invade and navigate within tissues. For example, endothelial tip cells send out fi lopodia to explore their territory for guidance cues, just like axon growth cones (Gerhardt et al 2003). Several recent studies have also explored the idea that axon guidance cues control blood vessel branching. For example, ephrin/EPH ligand/receptor pairs, netrins with their UNC and DCC receptors and neuropilin 1 (NRP1) with its ligands have all been implicated in neuronal and vascular patterning (Eichmann et al 2005). In this chapter, we will discuss recent published data as well as our own unpublished observations to determine the relative contribution of two different NRP1 ligands to neuronal and vascular development, the class 3 semaphorin SEMA3A and an 1

This paper was presented at the symposium by Christiana Ruhrberg, to whom correspondence should be addressed. 230

NEUROPILIN LIGANDS AND VASCULAR PATTERNING

231

FIG. 1. Working models that have been put forward to explain how SEMA3A affects blood vessel growth. According to model (A), SEMA3A binds to an endothelial receptor composed of NRP1 and a plexin. According to model (B), SEMA3A competes with VEGF164 and binds to an endothelial receptor composed of NRP1 and KDR (also known as VEGFR2 or FLK1). (C,D) Visualization of the endomucin-positive cardiovasculature in stage-matched littermate embryos expressing (C) or lacking (D) SEMA3A in a CD1 background at 9.5 dpc, when vessel networks have extended throughout the embryo and have begun to remodel into the large head vessels (open arrowhead); the anterior cardinal vein (arrowhead) and dorsal aorta (arrow) are clearly visible.

isoform of vascular endothelial growth factor termed VEGF164 (Raper 2000, Ruhrberg 2003). Two working models have previously been put forward that implicate SEMA3A in vascular patterning (Fig. 1). According to one model (Fig. 1A), SEMA3A binds to an endothelial receptor composed of NRP1 and a plexin (Serini & Bussolino 2004). Originally, it was thought that the class 3 semaphorin receptor contained an A-type plexin, but recent in vitro work has shown that plexin D1 can also cooperate with NRP1 to form a SEMA3A receptor (Gitler et al 2004). Moreover, in zebrafish plexin D1 controls blood vessel branching, with SEMA3A2, the

232

VIEIRA ET AL

zebrafish homologue of murine SEMA3A, being a possible ligand for plexin D1 (Torres-Vazquez et al 2004). Interestingly, at least in the mouse, NRP1 is not required for semaphorin signalling through plexin D1 (Gu et al 2005). According to an alternative model (Fig. 1B), SEMA3A controls vascular patterning by competing with VEGF164 for NRP1 binding (Miao et al 1999). In this model, VEGF164 acts through a receptor composed of NRP1 and VEGFR2 (alternative names FLK1 and KDR) to stimulate endothelial cell migration and proliferation, and SEMA3A inhibits this pathway. To evaluate the suitability of either model to explain the molecular control of vascular patterning, we examined the genetic requirement for SEMA3A during vascular morphogenesis and blood vessel branching in the mouse, and compared the phenotypes of SEMA3A null mutants to the known defects of mutants lacking VEGF164 alone or both VEGF164 and SEMA3A. We find that neither model is suitable to describe a mechanism operating in vascular development in the mouse, as SEMA3A and semaphorin signalling through neuropilins are not essential for blood vessel formation or angiogenic vessel branching. Firstly, we found that the vasculature of SEMA3A null mutants at E9.5 and E10.5 was indistinguishable from that of stage-matched wild-type littermates in both genetic backgrounds examined, C57BL/6 and CD1 mice (Fig. 1C,D and data not shown; unpublished observations). Specifically, we observed normal vascular remodelling in the head and identified a paired dorsal aorta and paired anterior cardinal veins in all mutants. At both developmental stages, heart development and formation of the pharyngeal arch arteries appeared normal. These observations suggest that SEMA3A is not required for the formation of blood vessel networks. Secondly, we determined if vessel branching in the hindbrain was reduced in the absence of SEMA3A, because (a) brain vessels are known to be affected severely by loss of NRP1, (b) vessel branching is readily quantifiable in this model system, and (c) SEMA3A is expressed in the hindbrain (Chilton & Guthrie 2003, Gerhardt et al 2003, Kawasaki et al 1999, Schwarz et al 2004). We found that hindbrain vessels formed normal networks in the absence of SEMA3A in either the C57BL/6 or CD1 background: At E12.5, a similar number of vessels had branched from the perineural vascular plexus into the hindbrain on its pial side, and a similar number of branch points was present in the vessel plexus that formed in the subventricular zone (C57BL/6 data shown in Fig. 2; CD1 data not shown; unpublished observations). Our findings contrast with those of another group that recently described vessel defects in SEMA3A null mutants in the CD1 background (Serini et al 2003), even though Serini and co-workers reported to have analysed mice carrying the same gene-targeting event we have examined (Taniguchi et al 1997). We therefore confirmed that our SEMA3A null mutants contained the correct gene targeting event by examining their nerve patterning. In agreement with previously published observations, we found that loss of SEMA3A caused

NEUROPILIN LIGANDS AND VASCULAR PATTERNING

233

FIG. 2. Loss of Sema3A does not impair brain vascularization. Visualization (A–D) and quantification (E, F) of vessel branching on the pial side (A, B) and in the subventricular zone (C, D) of PECAM-stained 12.5 dpc littermate hindbrains expressing and lacking SEMA3A in a C57BL/6 background.

extensive axon defasciculation of the cranial and limb nerves in both CD1 and C57BL/6 backgrounds. The alternative model of SEMA3A function in the vasculature suggests that SEMA3A influences blood vessel endothelium by competing with VEGF164 for NRP1 binding (Miao et al 1999). To test this model, we have compared vascular pattering in single mutants lacking either VEG164 or SEMA3A to compound mutants lacking both VEGF164 and SEMA3A, and mutants with an imbalanced expression of these two NRP1 ligands. The VEGF mutants we analysed carry a deletion of exons 6 and 7 of the Veg fa gene; this mutation prevents expression of the VEGF164, but not the VEGF120 isoform and therefore circumvents the embryonic lethality observed in mice lacking all VEGF isoforms (Carmeliet et al 1999). Hindbrains lacking VEGF164 expression contain larger vessels with fewer branch points compared to wild type littermates. This phenotype is caused by an altered distribution of VEGF in the extracellular matrix in the absence of heparinbinding isoforms (Ruhrberg et al 2002), but it is also similar to the phenotype of mutants lacking NRP1 specifically in the vasculature (Gu et al 2003 and our unpublished observations). Importantly, we found that the loss of SEMA3A neither increased nor ameliorated the severity of vascular defects caused by loss of VEGF164 in whole embryos at E9.5 or in the hindbrain model at E12.5 (unpublished observations). Specifically, the vessel branching defect of hindbrains lacking VEGF164 was not affected by abolishing SEMA3A expression, and all compound mutants contained a paired dorsal aorta and paired cardinal vein. These observations suggested that the

234

VIEIRA ET AL

presence of VEGF120 is sufficient to drive angioblast migration and the assembly of the major vessels in the mouse. Moreover, they demonstrated that SEMA3A does not co-operate with VEGF164 to direct either vasculogenesis or angiogenesis in the mouse, even though SEMA3A1 guides angioblasts to the sites where the paired dorsal aorta will form in zebrafish (Shoji et al 2003). To extend our observations on the role of NRP1 ligands in neuronal and vascular patterning to a region of the body where nerves and blood vessels are normally co-patterned, we have examined developing mouse limbs at E12.5. In this organ system, a vascular network develops throughout the limbs prior to axon invasion, but a second vascular network extends alongside growing axons into the digits. We have found that (a) NRP1 mutants display defects in both nerve and vascular patterning in the limb, with extensive axon defasciculation and reduced vessel branching; (b) SEMA3A mutants phenocopy the axonal defect of NRP1 mutants, but show normal vessel patterning; and (c) axonal growth appears at least grossly normal in limbs lacking VEGF164, whilst vessel patterning is disturbed, like in NRP1 mutants. Only the simultaneous deletion of both NRP1 ligands impairs both axon and vessel growth and recreates a full NRP1 defect. Our findings suggest that VEGF164 and SEMA3A control distinct patterning events. In agreement with this idea, we have previously shown that SEMA3A and VEGF164 pattern distinct compartments of the facial nerve in the mouse, with SEMA3A, but not VEGF164 controlling the behaviour of facial nerve axons in the branchial arches, and VEGF164 but not SEMA3A guiding the migration of the facial branchiomotor neuron somata inside the hindbrain (Schwarz et al 2004). Lastly, we would like to draw attention to our observation that, at the time points we have observed, excess axonal branching in SEMA3A mutants does not result in the induction of ectopic vessels, and that loss of vessel branches in VEGF120 mutants does not impair axon branching. It therefore appears that both types of networks initially grow independently of each other, and that the copatterning observed in the adult, exemplified for example by the development of a vasa nervosum, must be set up at later developmental stages. Acknowledgements We thank Drs Masahiko Taniguchi, Hajime Fujisawa, David D. Ginty, Alex L. Kolodkin and David T. Shima for providing mouse strains. C.R. is funded by a Career Development Fellowship from the Medical Research Council and J.M.V. holds a PhD studentship from the Fundação para a Ciência e Tecnologia (SFRH/BD/17812/2004).

References Carmeliet P, Ng YS, Nuyens D et al 1999 Impaired myocardial angiogenesis and ischemic cardiomyopathy in mice lacking the vascular endothelial growth factor isoforms VEGF164 and VEGF188. Nat Med 5:495–502

NEUROPILIN LIGANDS AND VASCULAR PATTERNING

235

Chilton JK, Guthrie S 2003 Cranial expression of class 3 secreted semaphorins and their neuropilin receptors. Dev Dyn 228:726–733 Eichmann A, Makinen T, Alitalo K 2005 Neural guidance molecules regulate vascular remodeling and vessel navigation. Genes Dev 19:1013–1021 Gerhardt H, Golding M, Fruttiger M et al 2003 VEGF guides angiogenic sprouting utilizing endothelial tip cell fi lopodia. J Cell Biol 161:1163–1177 Gitler AD, Lu MM, Epstein JA 2004 PlexinD1 and semaphorin signaling are required in endothelial cells for cardiovascular development. Dev Cell 7:107–116 Gu C, Rodriguez ER, Reimert DV et al 2003 Neuropilin-1 conveys semaphorin and VEGF signaling during neural and cardiovascular development. Dev Cell 5:45–57 Gu C, Yoshida Y, Livet J et al 2005 Semaphorin 3E and plexin-D1 control vascular pattern independently of neuropilins. Science 307:265–268 Kawasaki T, Kitsukawa T, Bekku Y et al 1999 A requirement for neuropilin-1 in embryonic vessel formation. Development 126:4895–4902 Miao HQ, Soker S, Feiner L et al 1999 Neuropilin-1 mediates collapsin-1/semaphorin III inhibition of endothelial cell motility: functional competition of collapsin-1 and vascular endothelial growth factor-165. J Cell Biol 146:233–242 Raper JA 2000 Semaphorins and their receptors in vertebrates and invertebrates. Curr Opin Neurobiol 10:88–94 Ruhrberg C 2003 Growing and shaping the vascular tree: multiple roles for VEGF. Bioessays 25:1052–1060 Ruhrberg C, Gerhardt H, Golding M et al 2002 Spatially restricted patterning cues provided by heparin-binding VEGF-A control blood vessel branching morphogenesis. Genes Dev 16:2684–2698 Schwarz Q, Gu C, Fujisawa H et al 2004 Vascular endothelial growth factor controls neuronal migration and cooperates with Sema3A to pattern distinct compartments of the facial nerve. Genes Dev 18:2822–2834 Serini G, Bussolino F 2004 Common cues in vascular and axon guidance. Physiology (Bethesda) 19:348–354 Serini G, Valdembri D, Zanivan S et al 2003 Class 3 semaphorins control vascular morphogenesis by inhibiting integrin function. Nature 424:391–397 Shoji W, Isogai S, Sato-Maeda M, Obinata M, Kuwada JY 2003 Semaphorin3a1 regulates angioblast migration and vascular development in zebrafish embryos. Development 130:3227–3236 Taniguchi M, Yuasa S, Fujisawa H et al 1997 Disruption of semaphorin III/D gene causes severe abnormality in peripheral nerve projection. Neuron 19:519–530 Torres-Vazquez J, Gitler AD, Fraser SD et al 2004 Semaphorin-plexin signaling guides patterning of the developing vasculature. Dev Cell 7:117–123

DISCUSSION Lammert: Thanks for sharing your results with us that seem to contradict published work. What did the former reports on the role of SEMA3A say? Were there also in vivo reports in mouse? Ruhrberg: Yes, in vivo data were shown for the CD1, but not the BL/6 background by one group (Serini et al 2003). In this paper, the CD1 mutants lacking SEMA3A have a vascular phenotype, which entails lack of an anterior cardinal vein, lack of head vessel remodelling and lack of intersomitic vessel branching. However, we

236

VIEIRA ET AL

see all three processes occurring normally in CD1 mutants lacking SEMA3A. I don’t understand how such a discrepancy can arise, because our lab and theirs have the same mouse mutant (Taniguchi et al 1997) and have backcrossed into the same genetic background, i.e. CD1 mouse obtained from the same company. We presume that this company maintains their CD1 mice as one consistent line of outbred mice. So if there is a modifier that we have no knowledge of, one could suggest that CD1 mice must have evolved in the two different laboratories into two different substrains. However, it does not seem that either my lab nor their lab has backcrossed for long enough to make this happen. Alternatively, it could be that size variation between embryos in the same litter is the root of the problem, because we are dealing with outbred mice. The stage-matching of mutants and controls within a litter should be critical if the sizes of the embryos differed within a litter. We have therefore rigorously stage-matched all embryos analysed. Drake: You have a nice model. What emerged in the limb bud is similar to what we see in the early embryo in which the pattering of the primary vasculature is completely non-nerve dependent. At the tip of the growing limb bud you seemed to have the same phenomenon. In the retina, the patterning is dependent on nerve cells. Thus there seem to be real differences in the molecules that are mediating the pattern in the retina versus the early embryo and limb bud. Ruhrberg: It seems that the cell types are different, but that the molecules are the same. Drake: Perhaps we are just fooled because we see a pattern and we assume that the mechanisms will be the same. Weinstein: It is clear that there isn’t good in vivo evidence for this competitive role that has been looked at quite a bit in vitro. There are a few things about the fish result that I want to make sure are characterized correctly. We never did any work to show that it was specifically SEMA3A that was acting through plexin D1. All the evidence we had suggested that there were a number of different semaphorins that could go through plexin D1. Ruhrberg: At least you had a partial phenotype with SEMA3A disruption—we saw nothing! Weinstein: It was weak. I can easily see how with a modifier or whatever you could go from having a weak phenotype to not really having a phenotype at all with the ligands. With the receptor, both in the mouse and the fish, we get a very dramatic phenotype, with a complete loss of patterning in the trunk. Ruhrberg: Speaking of modifiers, in a way neuropilin 1 may be one itself: In your system it may be playing a different role from its role in Jon Epstein’s system. Weinstein: We have no evidence from our data that neuropilin 1 is required. In the mouse there is some evidence that it may be required. This may explain why you can get more significant defects in the SEMA3A nulls in the nervous system

NEUROPILIN LIGANDS AND VASCULAR PATTERNING

237

than in the vasculature, if there is a capability for dual function with or without neuropilin in the vasculature. Vargesson: It has been known since 1989 that nerves and vessels don’t match in the limbs early on (Martin & Lewis 1989). Could it be because of the simple explanation that the limb is one of the fastest growing tissues and requires rapid changes in the limb vascular tree in order to supply oxygen and nutrients to the proliferating cells to allow the rapid limb outgrowth, thus you don’t want nerves there early on in development as they may interfere with limb outgrowth? Ruhrberg: The early vessel network seems to be growing independently of the nerve, but then there is a second layer of vessels, which seems to track the nerves. We thought originally that this might be an analogous situation to the one described by David Anderson’s lab, in which sensory skin nerves are secreting VEGF to attract or pattern vessels (Mukouyama et al 2002), i.e. we thought that the secondary network may have been induced by the nerves, but it doesn’t look like that now. Weinstein: One thing that has become clear is that almost any possible way you can think of for nerves and vessels to interact will be found in some context. In the trunk, where we were looking, both the vessels and motor axons are using semaphorin as a cue. The motor axons are using it as an attractor signal while the vessels are using it as a repulsive signal. Ye: There is a functional implication of nerves and vessels tracking together. It seems to be occurring when the axons are fasciculating and have a demand for sufficient oxygen supply. References Martin P, Lewis J 1989 Origins of the neurovascular bundle: interactions between developing nerves and blood vessels in embryonic chick skin. Int J Dev Biol 33:379–387 Mukouyama YS, Shin D, Britsch S, Taniguchi M, Anderson DJ 2002 Sensory nerves determine the pattern of arterial differentiation and blood vessel branching in the skin. Cell 109: 693–705 Serini G, Valdembri D, Zanivan S et al 2003 Class 3 semaphorins control vascular morphogenesis by inhibiting integrin function. Nature 424:391–397 Taniguchi M, Yuasa S, Fujisawa H et al 1997 Disruption of semaphoring III/D gene causes severe abnormality in peripheral nerve projection. Neuron 19:519–30

FINAL DISCUSSION Tracheal tube development in Drosophila Uv: I will tell you about a lumenal fi lament that we have found to be involved in shaping tracheal tubes in Drosophila. Tubes are built from a monolayer of epithelial cells and it is extraordinary how these cells can communicate with each other to form a uniform tube with an even diameter. In Drosophila we approach this by forward genetics. Our model system is the embryonic trachea, a network of epithelial tubes that spans the whole embryo. When tubes are first formed they often have a narrow lumen; subsequently they expand to attain their correct diameter. We are focusing on one branch, the main branch, called the dorsal trunk. Over a period of 5 h it expands three- to fivefold in diameter. This is reproducible from embryo to embryo, suggesting that it is highly genetically controlled. The dorsal trunk is built from 10 segments, which fuse to form the continuous tube. The lumen circumference is surrounded by 2–5 cells but, at the branch fusions, which are similar to those of the vasculature, the lumen is surrounded by a pair of donut-shaped fusion cells. Despite this difference in tube architecture, lumen expansion occurs in a uniform manner. We identified three mutants with tube size defects. One was krotzkopf verkehrt (kkv), identified in the 1980s because of a cuticle defect. We discovered that this mutant had a severe tube size defect: the fusion junctions don’t expand but the other parts of the tube expand excessively, and later additional lumen constrictions are seen. Finally, the tube becomes far too elongated. We analysed the apical/basal polarity, microtubule polarity and other lateral junction proteins, in the mutant tracheal cells, but found that these appeared normal. It then turned out that kkv encodes a chitin synthase. Chitin is known as a component of the insect cuticle. In the trachea, towards the end of embryogenesis, an apical cuticle is secreted inside the tube to provide stabilization when liquid is cleared and air enters. Chitin synthases are transmembrane proteins. They polymerize cytosolic GlcNac residues, and the resulting sugar chains are then believed to protrude from the cell through pores formed by the synthases themselves. (They are analogous to hyaluronan synthases in vertebrates.) We wondered whether it was the chitin synthase activity of this protein that was necessary for tube expansion, so we used a specific chitin synthase inhibitor, which we administered by feeding the mothers, and could indeed reproduce the kkv trachael phenotype in their offspring. To confirm that chitin is needed in the lumen for uniform tube expansion, we directed the expression of a chitinase in the developing trachea. Chitinases are 238

FINAL DISCUSSION

239

secreted and cleave chitin into smaller fragments, and this expression also reproduced the tube size phenotype of kkv mutants. We had a difficult time trying to visualize this lumenal chitin. But then we found Congo Red, a fluorescent chemical that intercalates in between chitin chains. Using Congo Red, we were surprised to find that during early embryogenesis chitin forms a fi lament in the lumen rather than lining the apical surface. The fi lament appears just after fusion when the tube begins to expand. After tube expansion it is broken down. The model is therefore that newly born tubes deposit a fi lament in the lumen, which is necessary to co-ordinate cell shape and promote a uniform lumen. We think it forms a kind of scaffold around which the tubular epithelium can rearrange. It is needed both to push out the fusion cells because they remain narrow in the absence of the chitin, and at the same time it seems to have another effect on the epithelium by holding it back. The other two genes we were looking at are called knickkopf (knk) and retroactive, and they have a similar phenotype. We have shown that these work in a chitin-dependent pathway in tube size control. They are both apically secreted proteins. Knickkopf is a GPI-linked protein. In both knickkopf and retroactive mutants chitin is still present but it doesn’t form a fi lament, and fi lls the whole lumen in an amorphous fashion. There is another large group of tube size mutants that affects the septate junctions, which are analogous to vertebrate tight junctions. These also have an amorphous chitin fi lament, and in these mutants the apical localization of Knk is lost; it is more lateral and basal. Thus, it is not sufficient just to have chitin in the lumens; it needs to assemble into fi laments. We also find that if the contact to the apical surface is broken, the chitin fi lament will contract. We think this is why it can hold back the diameter and constrain the length Ye: It would be interesting to visualize integrins in the Drosophila trachea system to see whether the chitin matrix can recruit integrins to the apical surface. Uv: Yes, we haven’t yet found any at the apical surface. Ye: In mouse there is evidence that integrin α5β1 is localized to the lumenal surface of the vascular endothelium (Parsons-Wingerter et al 2005). Ye: Other than integrins, what do you think might be mediating the force? Uv: We are now trying to find other components of this lumenal mesh in addition to chitin. We are also interested in finding what will connect to the apical surface of the cells. Then we are interested in fi nding out what will mediate the change to the cell. The subapical cytoskeleton is disorganized in these cells. Lammert: Do you have an antibody against the chitin synthase 1? Uv: No Lammert: It could be important to see where it is. It could be formed in small sections in the Golgi, and then go to a sorting compartment from where it is delivered towards the lumenal side, where additional steps happen that join these chitin subunits. There could be different steps inside and outside the cell.

240

FINAL DISCUSSION

Uv: Chitin chains are thought to be synthesized as long chains, and antibodylabelling against a chitin synthase in another insect suggests that the chitin synthase localizes to the apical surface. It is however suggested that chitin synthases can assemble on intracellular vesicles, so-called chitosomes, and that chitin synthesis also occurs (perhaps starts) in these vesicles. Gerhardt: If you have a polymer being formed, is the synthesis part of the attachment? If it is transmembrane synthesis with the synthase sitting in the lumenal membrane, it is polarized at the right place, and when you start making the chain. Uv: That is an interesting idea. In longitudinal sections we see ‘waves’ of lumenal material, that we believe are chitin fi laments, and in places these appear connected to the membrane, perhaps to the synthesizing machinery or to something else. Betsholtz: You mentioned that chitin synthase is homologous to hyaluronic acid synthase. What is known about hyaluronic acid synthesis and deposition during angiogenesis? Uv: It apparently is lining the apical surface of the lungs and kidney. Wilting: There are some studies (Feinberg & Beebe 1983) suggesting it is an anti-angiogenic molecule. In embryos, injection of hyaluronic acid will induce an avascular area. Organs that have a lot of hyaluronic acid such as the vitrous body are avascular. Hyaluronic acid is a complicated molecule that appears in many molecular weight forms. As soon as it starts to be degraded it changes from an anti-angiogenic agent into an angiogenic agent. Uv: In situ hybridization has been done with three different hyaluronan synthases. I think it was in frog and there was not anything apparent in the vasculature. Betsholtz: Lyve1 is a receptor for hyaluronic acid. Why would lymphatics need a receptor for this? Wilting: They take up metabolized hyaluronic acid, especially in the skin. There is a constant turnover, and much of this is transported via the lymphatics. Lammert: This is a transient scaffold. It has to be degraded rapidly to allow the air to go through. Is there a mutant that would help you understand what triggers this degradation? Uv: True, the chitin fi laments appear to break down and the lumen is cleared from lumenal components, but we haven’t yet found any mutants in which this is prevented. There is a chitinase expressed in the trachea. There are also proteases expressed after tube expansion. Closing remarks Betsholtz: It is time for me to sum up and close this meeting. I started out by commenting on the many cellular processes that are involved in vascular development,

FINAL DISCUSSION

241

and this meeting has indeed demonstrated how profoundly complex vascular development is, both at the cellular and molecular levels. One way of assessing the molecular complexity of vascular development is to ask how many genes are expressed in a developing vessel. In my own group, we have approached this question by measuring the size of the transcriptome in pure isolates of glomerular capillary fragments. In one experiment we generated and large-scale sequenced glomerular cDNA libraries. From about 15 000 cDNA sequences we annotated more than 6000 different genes. Since this was nowhere near saturation, we also ran Affymetrix chip analyses, which suggested that glomeruli, composed of just three cell types (and mostly endothelial cells and pericytes) express in the order of 12 000 genes, which is about half of the genome. From this result, we may conclude that vascular development is half as complex as life itself! Let’s conclude that it is complex enough to keep us busy for some time yet. Speaking seriously, the angiogenesis research community might even be too small to advance the field rapidly enough. Nevertheless, recent progress is impressive, as demonstrated at this meeting. For example, it has become clear to us that the cell biology has thus far been under-represented in vascular development research, but that this situation is in the process of changing. The complexity of vascular development creates problems in interpreting the information that we receive from genetic and pharmacological manipulations in vivo. There are obvious problems with confounding secondary effects, feedback regulation and other mechanisms that influence the phenotypic outcome in our experiments. Therefore it is becoming increasingly important to study morphological and molecular phenomena at high spatial and temporal resolution. Perhaps it is fair to emphasize that we will need more welldefined pharmacological tools to be able to do proper studies with good temporal resolution. Presently, the genetic tools that we exploit to temporally change gene expression are too slow to allow the temporal resolution that we need. Irrespective of these shortcomings we have seen data in this meeting demonstrating the power of transgenic technologies in the fish and mouse. We have heard about prospects for the speeding up and further refinement of these technologies. In summary, this meeting has been a good representation of the state-of-the-art of vascular development research. The penetrating discussions have illustrated the rapid advancement of the field, but perhaps more importantly also highlighted critical gaps in our knowledge, thereby helping us to focus future efforts on the most critical scientific questions in the area of vascular development. References Feinberg RN, Beebe DC 1983 Hyaluronate in vasculogenesis. Science 220:1177–1179 Parsons-Wingerter P, Kasman IM, Norberg S et al 2005 Uniform overexpression and rapid accessibility of alpha5beta1 integrin on blood vessels in tumors. Am J Pathol 167:193–211

Contributor Index Non-participating co-authors are indicated by asterisks. Entries in bold indicate papers; other entries refer to discussion contributions.

A

E

Adams, R. H. 29, 34, 35, 75, 81, 82, 99, 136, 162, 165, 171, 172, 173 *Alitalo, K. 87 *Argraves, W. S. 61 Augustin, H. G. 15, 29, 42, 43, 44, 59, 75, 104, 121, 134, 137, 164

*Eglinger, J. 46 Eichmann, A. 58, 77, 80, 81, 82, 83 Epstein, J. A. 33, 34, 35, 104, 122, 152, 161, 162, 163, 164, 171 *Ewald, A. J. 37 F

B *Benedito, R. 77 Betsholtz, C. 1, 13, 14, 31, 33, 42, 58, 60, 81, 82, 84, 85, 86, 99, 100, 101, 103, 134, 136, 137, 138, 173, 194, 202, 203, 240 *Boland, P. 106 *Bréant, C. 77 Breier, G. 126, 133, 134, 135, 136, 137, 172 *Buttler, K. 220

*Filvaroff, E. 18 *Fleming, P. A. 61 Freedman, D. A. 207, 216, 217, 218, 219 *Freitas, C. 77 *French, D. 18 G

*Castranova, D. 139 *Coetzee, S. 106

*Gale, N. W. 106 Gerhardt, H. 14, 16, 31, 34, 35, 82, 84, 85, 100, 123, 135, 137, 138, 194, 202, 203, 204, 205, 206, 218, 227, 240 *Gray, A. 18

D

H

*Daly, C. 106 Davies, J. 14, 43, 56, 72, 80 *De Mazière, A. 18 Dejana, E. 4, 13, 14, 15, 16, 28, 29, 35, 59, 99, 100, 135, 148, 172, 191, 205, 206 Drake, C. J. 14, 17, 29, 34, 58, 61, 71, 72, 73, 74, 75, 82, 83, 84, 86, 102, 123, 136, 150, 161, 162, 191, 201, 202, 203, 206, 216, 228, 229, 236 *Duarte, A. 77 *Dye, L. 139

High, F. 152, 163 *Hitomi, J. 139

C

I *Iruela-Arispe, L. 194 *Isogai, S. 139 K *Kashima, Y. 207 *Kirsnerova, Z. 126 242

CONTRIBUTOR INDEX Kitajewski, J. 31, 32, 60, 80, 81, 82, 104, 121, 122, 171, 191, 216 *Klotzsche, A. 126 *Klumperman, J. 18 *Kucˇera, T. 46 L Lammert, E. 13, 33, 43, 44, 46, 56, 57, 58, 59, 72, 74, 81, 85, 98, 101, 104, 123, 133, 136, 149, 162, 172, 204, 205, 217, 218, 227, 235, 239, 240 *Lampugnani, M. G. 4 *le Noble, F. 77 *Lee, S. 194 *Levin, M. 37 *Licht, A. H. 126 *Lin, H. C. 106 *Lundkvist, A. 194 M Mäkinen, T. 87, 98, 99, 100, 101, 102, 103, 104, 172, 229 *McMahon, M. 37 Mostov, K. E. 37, 42, 43, 44, 56, 57, 60 N *Nicolaus, A. 126 *Noguera-Troise, I. 106 *Norgall, S. 220 O Owens, G. K. 15, 16, 32, 73, 74, 75, 86, 124, 135, 163, 174, 191, 192, 206, 216, 217 P *Papadopoulos, N. J. 106 *Papoutsi, M. 220 *Parker, L. 18 R *Rössler, J. 220 Ruhrberg, C. 32, 35, 58, 59, 71, 72, 83, 84, 85, 86, 122, 161, 164, 203, 230, 235, 236, 237

243 *Schweigerer, L. 220 Shibuya, M. 30, 82, 102, 135, 136, 151, 192, 206, 218, 229 Shovlin, C. L. 34, 44, 59, 75, 101, 124, 137 *Smyczek, T. 18 *Strilic´, B. 46 *Suchting, S. 77 T *Thurston, G. 106 U Uv, A. E. 85, 86, 101, 102, 104, 150, 238, 239, 240 V *van Dijk, S. 18 Vargesson, N. A. 34, 56, 148, 237 *Vieira, J. M. 230 W *Weich, H. A. 220 Weinstein, B. M. 100, 103, 121, 122, 123, 136, 139, 148, 149, 150, 151, 164, 172, 204, 205, 206, 218, 228, 229, 236, 237 *Werb, Z. 37 *Wielockx, B. 126 Wilkins, A. 57, 75, 81 Wilting, J. 32, 33, 34, 35, 44, 57, 58, 75, 84, 101, 134, 137, 173, 216, 220, 227, 228, 229, 240 Y Yancopoulos, G. D. 60, 73, 81, 83, 84, 85, 101, 102, 103, 104, 106, 121, 122, 123, 124, 133, 134 *Yaniv, K. 139 Ye, W. 18, 29, 30, 31, 32, 33, 34, 35, 56, 72, 73, 100, 124, 149, 150, 163, 205, 218, 219, 237, 239

S *Schmidt, M. 18 *Schwartz, Q. 230

Z *Zaret, K. S. 207

Subject Index

A

C

α -actin 163, 176, 177, 179, 180, 182, 183 activins 134, 213 adherens junctions control of endothelial cells functions 4–11 VE-cadherin 5 Akt 8 Alagille syndrome 155 Alk1 101 ALK5 75 Ang1 90, 91 Ang2 90 Ang3/4 90 angiogenesis 221 angiogenic growth factors 49 angiogenic switch 197 angiopoietin 216 aplasia 221 Approach for Ranking of Embryonic Vascular Anomalies (AREVA) 62, 63 ARNT 129 Avastin 116

c-fos 181 C-terminal src kinase 9 C6 gliomas 116 h-caldesmon 176 h-calponin 176 cardiac neural crest 152–154 CArG-SRF-dependent regulation 180 α -catenin 5 β -catenin 5, 8, 9, 16, 68 β -catenin/Tcf-Lef-1 binding domain 9 γ -catenin 5, 9, 16 CD11b 223, 225 CD31 (PECAM-1) antibodies 74, 114, 208, 224, 225 CD34 58, 74 CD45 223, 227 CD148/DEP-1 62, 63, 66 Cdc42 49 Cdc42 BTPase 49 ChIP (chromatin immunoprecipitation) 217 chitin 52, 238, 239 chitin synthase 238, 239, 240 chitinases 238 chitosomes 240 cholera toxin B labelling 56 chromatin 183, 186, 212 chromatin immunoprecipitation see ChIP cMaf 224 collagen IV basal polarity 59 deposition during yolk sac vascular development 58 staining in vascular lumen 44, 57 Connexin 45 gene 62, 66, 72, 75 cpr450 gene 62, 65 Cre 167–168, 171 CreER 168

B BAC 168 basic helix loop domain 127 BECs see blood vascular endothelial cells bevacizumab (Avastin) 116 BFS-1 fibrosarcoma 131, 133 blastocyst 57 blood vascular endothelial cells (BECs) 88 BMPs 213 BMP4 74 Bmx-CreERT2 line 171 bone morphogenic proteins see BMPs Brnx(PAC)-CreERT2 transgenics 168 BSA 26

244

SUBJECT INDEX CreERT2 167, 168 cytochrome P450 reductase gene (cpr450) 62, 65 D DAPT 79, 82 Delta 85 Delta/Jagged families 107 Delta/Notch signalling 72, 123 Delta-like ligand 4 see Dll4 Dep-1 (density enhanced phosphatase 1) 9, 11 DEP1/CD148 66, 67, 68, 72 desmin 176 dextran 56, 59 DiGeorge syndrome 154–155 DiI-LDL 225 Dll4 73, 78, 79, 80, 82, 107, 121, 123 lumen formation 60 Dll4 knockout mice 60, 72, 73 Dll4/Notch pathway 81, 107 activation and blocking in tumours 108–110 tumour growth 110–111, 118–119 tumour hypoxia 112–114 Dll4-blocking antibody 116–118 Dll4-Fc 118 decrease in tumour growth 114–116 multiple tumour models 116–118 DLR4 Trap 121 doxycycline (Dox) 167 Drosophila heart 59 tube formation 150, 238–241 E ecdysone 167 EGF pathway 18–27, 58 EGFL7 (VE-statin) 54 binding to endothelial cells 30 death of embryos 34 endothelial cell migration 26–27 extracellular matrix associated protein 24–26 fibronectin 35 fi ne localization 33 formation of vascular lumen 46 positioning of endothelial cells in zebrafish embryos 19–20, 29 regulation of vascular morphogenesis in mouse embryos 21–23, 29

245 tumour angiogenesis in adult mice 23–24, 30–31, 32–33 vascular morphogenesis in mouse embryos 21–23 EGFL8 24, 33 ELISA 103, 224 ELK1 182 embryonic stem cells (ESCs) 182 endoglin 75 endothelial junctions molecular architecture 5 endothelial signalling hepatic proliferation 208–209 pancreatic specification 210 pancreatic development 210–211 EphB4 98, 99, 172 Ephrin 155 Ephrin B 91, 171 Ephrin B receptors 98 ephrinB2 169 function 75 ligation 63 lymphatic vasculature 92 lymphangiogenesis 98–102 Notch connection 171 signalling 172–173 vasculogenesis 62, 66, 67 EphrinB2 mutants 100 Ephrin B4 see EphB4 Ephrin/EPH ligand/receptor pairs 230 epigenetic programming 186 Erk 66 erythropoietin 126 ESCs see embryonic stem cells Ets-1 128 EXT1 locus 197 F FACS analysis 224 Factor Inhibiting HIF (FIH) 127 FAK 63 fetal bovine serum (FBS) 179 fibroblast growth factor (FGF) 16 FGF2 91, 179 FGF10 211, 213 fibronectin 32, 66, 203 gene 62 organization 26, 28–29 Fli1 75 Fli1-positive cells 7

246

SUBJECT INDEX

Flk 72, 74 Flk gene 61, 62, 129 FLK1 209, 232 interaction with α5β1 integrin 29 see also VEGFR2 Flk1 nulls 65 Flk1 promoter 128–129 Flp integrase 167 Flp recombinase 168 Flt 72 Flt1 see VEGFR1 follistatin 134 formation 58 FOXC2 93 G β -galactosidase 108 β -galactosidase (betagal)

reporter gene 78

GATA6 158, 181 GFAP-Cre 197 gliomas, C6 116 GM6001 199 Gpl35 60 gridlock mutants 155 H HDAC inhibitor 185 hemangion 220 heparan sulfate 200 heparan sulfate proteoglycans (HSPGs) 197 hepatocyte growth factor (HGF) 50, 91 hereditary haemorrhagic telangiectasia type 1 (HHT1) 52 Herpes simplex virus VP16 protein 167 HES1 108–110, 114 HEY2 108–110 HGF see hepatocyte growth factor HHT1 see hereditary haemorrhagic telangiectasia type 1 HHV8 225 HIF 126–132, 133–138 anti-angiogenic tumour therapy 131–132 cardiovascular development 129–130 regulation of essential endothelial receptor tyrosine kinases 130–131 regulation of VEGF receptor expression 128–129, 136, 137 HIF-1 126, 127, 128, 129, 131 HIF-2 126, 127, 128, 129, 131

HIF-2 knockout mice 135 HIF-2 α 128 HIF-3 127 HIF asparaginyl hydroxylase 127 HIF deficiency 130 HIF hydroxylase 132 HIF prolyl hydroxylases see PHDs histone acetylation 217 histone acetyltransferase activity 185 histone deacetylase inhibitor see HDAC inhibitor HIV 226 Hnf6 212 Hox 192 HRE 128 HSPGs see heparan sulfate proteoglycans HT1080-RM tumours 116 human umbilical vein endothelial cells (HUVECs) 49 hyaluronan synthase 238 hyaluronic acid 240 4-hydroxytamoxifen 168 hyperplasia 221 hypoplasia 221 hypotrichosis-lymphoedema-telangiectasia 93 hypoxia-inducible factor see HIF hypoxia response element see HRE I ICAM1 59, 75 ICAM2 23 Indian hedgehog 74 inflammatory diseases 93–94 integrins 172 α1 224 α5 integrin gene 62, 66 α5β1 29, 63, 67, 203 α9 224 αvβ3 67 β3 67 interleukins IL1 206 IL3 229 IL6 206 IL8 191 intussusceptive angiogenesis 37–41, 42–44 intussusceptive pillar formation 40 ISH 80 isolectin B4 31

SUBJECT INDEX

247

J

M

Jagged 1 121, 155, 173

macrophase elastase 199 macropinocytic endocytosis 49 MADS (MCM1, Agamous, Deficiens, SRF) box transcription factor 181 MAPK 11, 63 MARTF-A 154 Matrigel 59 matrix metalloproteinase (MMP) inhibitors 40, 199–200 matrix scaffolds 203 MCP1 206 MDCK 59 cells 57 cysts 50–52 MEF2 181 Met receptor 11 metavinculin 176 Milroy disease 92 mitogen-activated protein kinase see MAPK MKL1 181 MKL2 181 MMP family 197, 198, 202 MMP3 199 MMP9 199 MMP12 199 monoclonal antibody BV9 7 BV13 7 E4G10 7 MRTF-B 154, 163 myocardin 154, 181, 182, 183, 185 myocardin-related factor (MRTF-B) 154, 163 myosin light chains 176

K Kaposi’s sarcoma 225, 226, 227 Kaposi’s sarcoma-associated herpes virus (KSHV) 225 KDR see VEGFR2 Ki67 223 KLF2 191 KLF4 182, 191 KLF4/GKLF 179 knickkopf (Knk) 239 KSHV see Kaposi’s sarcoma-associated herpes virus L lac operator-repressor system 167 LacZ 153 LacZ-VEGFR3 reporter gene 101 laminin 203 LKLF 191 loxP 167 LPP 176 Lucifer yellow dye 56 lymphangiectasia 221 lymphangiodysplasia 221 lymphangiogenesis, tumour-induced 221 lymphangioma 221, 226 lymphangiomatosis 221 lymphangion 220 lymphatic endothelial cells (LECs) 140, 145–147, 221 differentiation 88–89 lymphatic metastasis 93 lymphatic vascular system 87 lymphangiogenic growth factors 90–91 regulation of growth 89–91 remodelling 91–92 separation 91 VEGFs and VEGF receptor tyrosine kinases 89–90 zebrafish 139–147, 148–151 lymphoedema 92–93 lymphvasculogenesis 221 Lyve1 (lymphatic vessel endothelial hyaluronan receptor 1) 89, 102, 223, 224, 225, 227, 240

N N-cadherin 14, 16–17 Netrin receptor UNC5B 78 Netrins 75, 155, 230 Netrin 1 78 neural crest-derived smooth muscle differentiation signalling pathways 154–155 neuropilin 86 neuropilin 1 (NRP1) 63, 156, 161, 230, 231, 232, 236 neuropilin 2 (NRP2) 63, 156, 158 genes 142 Nrp2a 142

248 Nkx3.2 181 Notch 85, 121 EphB2 connection 171 Notch/Delta signalling 72, 123 Notch1 63, 79, 104, 118 Notch2 155 Notch4 79, 104 Notch9 79 pathway 60, 107 signalling 79–80, 108–110 see also Dll4/Notch pathway NR ARP 108–110, 114 NRP1 63, 156, 161, 230, 231, 232, 236 NRP2 63, 156, 158 Nrp2 gene 142 Nrp2a gene 142 O osteosarcoma 132 2-oxoglutarate 127 oxygen-dependent degradation domains (ODDD) 127 oxygen-induced retinopathy (OIR) 197–200

SUBJECT INDEX plexin 156, 231 Plexin A2 156, 162 Plexin D1 156–158, 164, 231, 232, 236 PlexinD1 mice 162 podoplanin 224, 225 Pou 192 presenilins 1 and 2 63 prolyl hydroxylase domain proteins see PHDs prospero-related homeobox 1 see Prox1 Prox1 142, 143 expression 102, 103 LEC differentiation 88, 89 lymphatics 222, 223, 224, 227, 228 Prx1/Mhox 181 Ptf1a expression 210, 211–212, 214, 216, 217, 218, 219 induction 218 Q QH1 223 qPCR 80

P

R

p85 7–8 p120 5, 9 PAC 168 PAS see Per-Arnt-Sim Pax 192 Pax3 152 PDGF BB 177, 178–179, 191 PDGF-DD 191 PDGF receptor kinase 192 PDGFB 79, 135 PDGFβ 11 PDGFs 91 Pdx1 210, 214 PECAM 58 PECAM-1 131 PECAM-1 positive cells 7 pELK 182 Per-Arnt-Sim (PAS) 127 PHDs 127 PHD2 126, 132 phorbol ester 49 phosphatidylinositol 3 (PI3) kinase 8 placental growth factor (PlGF) 108 plakoglobin 5 PLCγ 11, 151

RA see retinoic acid Rac, activation of 8 Rac1 GTPase 49 Raf 66 Ras 66 reelin 224 retinoic acid (RA) 65, 213 retroactive 239 Rho, inhibition of 8 RipTag model 105 S S1P see sphingosine-1 phosphate 82 inhibitor 79, 123 SEMA3A 157 murine neuronal and vascular patterning 230–234, 235–237 SEMA3A1 234 SEMA3A2 231 SEMA3C 156, 157, 158, 161, 164 SEMA3E receptor 157 semaphorin class 3 156 semaphorins 75, 156, 164, 236

γ -secretase γ -secretase

SUBJECT INDEX semaphorin–plexin signalling in cardiovascular development 155–158 semaphorin signalling 158 serum response factor (SRF) 154, 181 sFlt1 see soluble VEGF receptor 1 Shc–Grb2 complex 63 Shc isoforms 63 ShcA gene 62, 66 ShcA/Ras/Raf/Mek/Erk signalling cascade 63, 66, 68 Shh signalling 213 SHP2 9 Slits 75 SLP76 91 SM22 gene 185 SM22 α 176, 177, 182, 183 smoothelin 176 soluble VEGF receptor 1 (sFlt1) 79 Sonic hedgehog 74 Sox 192 Sox18 93 sphingosine-1 phosphate (S1P) 210, 211, 213 sprouting angiogenesis 38 Src kinase 8 Src phosphorylation 63 SRF see serum response factor Syk 91 T Tal1 65, 74, 75 Tal1/SCL expression 63 TAD see transactivation domain tamoxifen 168 TβRI 75 Tbx1 154–155 Tek/Tie2 gene 62, 66 telokin 176 telomerase 44 tetracycline-responsive (Tet) system 167 Tiam 8 Tie1 receptor tyrosine kinases 90 Tie2 62, 66, 128, 130 probe 221 receptor tyrosine kinases 90 tight junctions 5 TIMPs 199 Tn10 (E. coli transposon) 167 transactivation domains (TAD) 127 trichostatin A 185 α -tropomyosin 176

249 tumour lymphangiogenesis 93 tumour necrosis factor (TNF) 59 tyrosine kinases 8, 169 tyrosine phosphatases 8 U ubiquitin ligase 128 UNC5B 81 Unc5b gene 78, 79, 83 V vacuolar apical compartment (VAC) 51–52 vascular endothelial growth factor see VEGF vascular hyperfusion 73 vascular lumen formation 46–54, 56–60 cell death and phagocytosis 47, 50 fly tracheae 52–54 in vitro studies 54, 57 role of VEGF 57 vacuole formation and coalescence 47, 48–49, 57 vascular pillars 39, 42, 43 wrapping around extracellular space 47, 49–50, 57 vascular smooth muscle cell (SMC) cell lineage memory 185–187 differentiation, environmental cues and 175–179 epigenetic controls 183–184 phenotypic modulation/switching 174–175 transcriptional regulation 179–182 vascular tube formation 47–50 vasculogenesis critical genes 61, 62, 63–68 angioblast and endothelial cell proliferation and survival 66 angioblast motility and fi lopodial formation/function 66–67 endothelial lineage formation 65 stabilization of nascent blood vessels 67–68 VCAM1 59, 75 VE-cadherin 83, 131, 172, 205 adherens junctions 5 distribution 14 endothelial functions 7 expression at tip cell 14–15 interaction with VEGFR2 9–11, 14 Shc association 63 signalling pathways 7–11

250 stabilization of vascular tubes 46 VE-cadherin gene 62, 67–68 VE-statin see EGFL7 VEGF 71, 85 activation 13 Dll4 induction 60 Dll4 induction in tumour vessels 107–108 lymphatic vessel growth 89–90 production of matrix proteins 29 tip cells 77 tumour angiogenesis 126, 128 upregulation in macrophages 206 VE-cadherin 13 VEGF blocker 123 VEGF gradient 202–204 VEGF receptor see VEGFR VEGF receptor kinase 89–90, 192 VEGF signalling ShcA/Ras/Raf/Mek/Erk signalling cascade 63, 66, 68 vasculogenic genes 62–63 vasculogenesis 61–62 VEGF sink 79 VEGF Trap 108, 116, 121 VEGF113 199 VEGF120 198, 203, 233–234 VEGF120/71 120, 197 VEGF164 122, 161, 197, 198, 200, 206 murine neuronal and vascular patterning 230–234, 235–237 VEGF165 73, 161 VEGF188 35, 71, 122, 123, 200 VEGFA 62, 66, 233 balancing migration and proliferation 196 graded vs diffuse distribution 196 gradient formation 196–200 tip cell 195 vascular patterning 194–195 VEGFC 62, 89, 103, 143, 229 FGF2 91 Kaposi’s sarcoma 225

SUBJECT INDEX lymphangiogenesis 89, 93, 105 therapy 94 VEGFD 89, 91, 93 VEGFR 156 tip cell formation 83 VEGFR1 62, 65, 66, 75, 80, 130 VEGFR2 51, 61, 62, 80, 89, 131, 209, 232 association with α5β1 67 BMP4 74 Dll4 82 growth of lymphatic vessels 90 interaction with integrins 15 interaction with VE-cadherin 9–11, 68, 83 Notch signalling pathway 63, 80 patterns around hepatic endoderm 208 regulation by HIF 128 as regulator of lymphangiomas 224, 226 tip cell concept 14, 195 upregulation 104 vascular lumen formation 58–59 vasculogenesis 135–136 VEGFR3 88, 93, 229 ELISA 222 growth of lymphatic vessels 89–90 lymphangiogenesis 100, 103–104 postnatal dependency 100 regulator of lymphangiomas 221, 224, 225, 226 signalling 94 β -vinculin 176 vitronectin 26 W Wnt1 152 Z zebrafish lymphatic system 139–147, 148–151 zPlexinD1 156, 157

Vascular Development (Novartis Foundation Symposium 283)

Nuclear Organization in Development and Disease (Novartis Foundation Symposium 264)

Cancer and Inflammation (Novartis Foundation Symposium 256)

Anaphylaxis (Novartis Foundation Symposia)

The Biology of Extracellular Molecular Chaperones (Novartis Foundation Symposium 291)

Neuronal and Cognitive Effects of Oestrogens (Novartis Foundation Symposium 230)

Mucus Hypersecretion in Respiratory Disease (Novartis Foundation Symposium 248)

The Limits of Reductionism in Biology (Novartis Foundation Symposium 213)

Mitochondrial Biology: New Perspectives (Novartis Foundation Symposium 287)

The Genetics of Autoimmunity (Novartis Foundation Symposium 267)

Decoding the Genomic Control of Immune Reactions: Novartis Foundation Symposium

Neuronal and Cognitive Effects of Oestrogens (Novartis Foundation Symposium 230)

Mammalian TRP Channels as Molecular Targets (Novartis Foundation Symposium 258)

Tissue Engineering of Cartilage and Bone (Novartis Foundation Symposium 249)

Autism : Neural Basis and Treatment Possibilities (Novartis Foundation Symposium 251)

Growth Factors and Psychiatric Disorders (Novartis Foundation Symposium 289)

Endocrine Facets of Ageing (Novartis Foundation Symposium 242)

Mast Cells and Basophils: Development, Activation and Roles in Allergic Autoimmune Disease (Novartis Foundation Symposium 271)

Cortical Development: Genes and Genetic Abnormalities (Novartis Foundation Symposia)

Tinkering: The Microevolution of Development (Novartis Foundation Symposia)

Empathy and Fairness (Novartis Foundation Symposia)

Osteoarthritic Joint Pain (Novartis Foundation Symposia)

Dietary Supplements and Health (Novartis Foundation Symposia)

2009 (283)

Neuraltransplantation in Neurodegnerative Disease - Current Status and New Directions (Novartis Foundation Symposium 231)

Retinal Dystrophies: Functional Genomics to Gene Therapy (Novartis Foundation Symposium 255)

Signalling Networks in Cell Shape and Motility (Novartis Foundation Symposium 269)

Novel and Re-emerging Respiratory Viral Diseases (Novartis Foundation Symposium 290)

Defining Optimal Immunotherapies for Type 1 Diabetes (Novartis Foundation Symposium 292)

Fatty Acid and Lipotoxicity in Obesity and Diabetes (Novartis Foundation Symposium 286)

Acetaldehyde-Related Pathology: Bridging the Trans-Disciplinary Divide (Novartis Foundation Symposium 285)

Vascular Development (Novartis Foundation Symposium 283)

Nuclear Organization in Development and Disease (Novartis Foundation Symposium 264)

Cancer and Inflammation (Novartis Foundation Symposium 256)

Anaphylaxis (Novartis Foundation Symposia)

The Biology of Extracellular Molecular Chaperones (Novartis Foundation Symposium 291)

Neuronal and Cognitive Effects of Oestrogens (Novartis Foundation Symposium 230)

Mucus Hypersecretion in Respiratory Disease (Novartis Foundation Symposium 248)

The Limits of Reductionism in Biology (Novartis Foundation Symposium 213)

Mitochondrial Biology: New Perspectives (Novartis Foundation Symposium 287)

The Genetics of Autoimmunity (Novartis Foundation Symposium 267)

Decoding the Genomic Control of Immune Reactions: Novartis Foundation Symposium

Neuronal and Cognitive Effects of Oestrogens (Novartis Foundation Symposium 230)

Mammalian TRP Channels as Molecular Targets (Novartis Foundation Symposium 258)

Tissue Engineering of Cartilage and Bone (Novartis Foundation Symposium 249)

Autism : Neural Basis and Treatment Possibilities (Novartis Foundation Symposium 251)

Growth Factors and Psychiatric Disorders (Novartis Foundation Symposium 289)

Endocrine Facets of Ageing (Novartis Foundation Symposium 242)

Mast Cells and Basophils: Development, Activation and Roles in Allergic Autoimmune Disease (Novartis Foundation Symposium 271)

Cortical Development: Genes and Genetic Abnormalities (Novartis Foundation Symposia)

Tinkering: The Microevolution of Development (Novartis Foundation Symposia)

Empathy and Fairness (Novartis Foundation Symposia)

Osteoarthritic Joint Pain (Novartis Foundation Symposia)

Dietary Supplements and Health (Novartis Foundation Symposia)

2009 (283)

Neuraltransplantation in Neurodegnerative Disease - Current Status and New Directions (Novartis Foundation Symposium 231)

Retinal Dystrophies: Functional Genomics to Gene Therapy (Novartis Foundation Symposium 255)

Signalling Networks in Cell Shape and Motility (Novartis Foundation Symposium 269)

Novel and Re-emerging Respiratory Viral Diseases (Novartis Foundation Symposium 290)

Defining Optimal Immunotherapies for Type 1 Diabetes (Novartis Foundation Symposium 292)

Fatty Acid and Lipotoxicity in Obesity and Diabetes (Novartis Foundation Symposium 286)

Acetaldehyde-Related Pathology: Bridging the Trans-Disciplinary Divide (Novartis Foundation Symposium 285)

Recommend Documents