Membrane Protein Structure Determination: Methods and Protocols (Methods in Molecular Biology, Vol. 654)

Methods in Molecular Biology™ Series Editor John M. Walker School of Life Sciences University of Hertfordshire Hatfi...

Author: Jean-Jacques Lacapere

141 downloads 1113 Views 12MB 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

Methods

in

Molecular Biology™

Series Editor John M. Walker School of Life Sciences University of Hertfordshire Hatfield, Hertfordshire, AL10 9AB, UK

For other titles published in this series, go to www.springer.com/series/7651

as

Membrane Protein Structure Determination Methods and Protocols

Edited by

Jean-Jacques Lacapère INSERM U773/CRB3, Université Paris Diderot – Paris 7, Paris, France

Editor Jean-Jacques Lacapère INSERM U773/CRB3 Université Paris Diderot – Paris 7 Paris France [email protected]

ISSN 1064-3745 e-ISSN 1940-6029 ISBN 978-1-60761-761-7 e-ISBN 978-1-60761-762-4 DOI 10.1007/978-1-60761-762-4 Springer New York Dordrecht Heidelberg London Library of Congress Control Number: 2010930685 © Springer Science+Business Media, LLC 2010 All rights reserved. This work may not be translated or copied in whole or in part without the written permission of the publisher (Humana Press, c/o Springer Science+Business Media, LLC, 233 Spring Street, New York, NY 10013, USA), except for brief excerpts in connection with reviews or scholarly analysis. Use in connection with any form of information storage and retrieval, electronic adaptation, computer software, or by similar or Â�dissimilar methodology now known or hereafter developed is forbidden. The use in this publication of trade names, trademarks, service marks, and similar terms, even if they are not identified as such, is not to be taken as an expression of opinion as to whether or not they are subject to proprietary rights. While the advice and information in this book are believed to be true and accurate at the date of going to press, Â�neither the authors nor the editors nor the publisher can accept any legal responsibility for any errors or omissions that may be made. The publisher makes no warranty, express or implied, with respect to the material contained herein. Cover illustration: Membrane protein determination starts from extraction-purification and reaches atomic structure by crystal formation and X-ray diffraction or electron microscopy analysis, or nuclear magnetic resonance studies combined or not with molecular modelling. Printed on acid-free paper Humana Press is part of Springer Science+Business Media (www.springer.com)

Preface Membrane proteins represent almost 40% of all proteins, but only a small number of their structures have been determined. Alone or associated with other proteins, membrane proteins play several roles in the cells. They are involved in signal transduction, ion exchanges, transport of metabolites, molecules or proteins. Cellular communications are controlled or regulated by membrane proteins. Indeed, they are involved in communications between cells, outside/inside cell exchanges, cytosolic traffic among different organelles as well as cytosol/organelles exchanges. Only a few functional classes of membrane proteins have been structurally characterized and mostly are transporters working alone. Membrane proteins are difficult to study mostly because they are often poorly abundant and thus difficult to purify in amounts compatible with structural studies. Heterologous overexpression of recombinant membrane protein is a strategy that has permitted the study of several membrane protein structures at an atomic level. However, membrane proteins are located in an hydrophobic environment such as the cellular bilayers, and their functions often involve hydrophilic contacts with lipids, resulting in the paradox that membrane proteins need lipids to work but they also need detergent addition to be purified. When proteins are associated in complexes in a functional way, their stabilization is often difficult in purification protocols and requires numerous trial and error steps. Determination of a structure is a crucial step but never solves the functional question. Indeed, activation or inactivation of membrane proteins involves numerous factors such as ligand binding, phosphorylation of specific residues, and posttranslational modifications. From a pharmacological point of view, ligands induce or block a functional response that mayÂ€involve either a single protein or a cascade of several proteins mixing membranous and soluble ones. This assembly of proteins can form stable or dynamic interacting complexes. In the cellular environment, these complexes are probably quite easy to form if one considers on one hand the protein concentrations inside the cytosol or the membranes and on the other hand the relative proximity of organelles in these cells. A fundamental aim of structural biology is to move from understanding structure and dynamics to controlling molecular function. This book describes major techniques used in the field of membrane protein structure determination. It is divided into five sections describing different techniques used to solve atomic structure either from purified membrane proteins or in silico. It also describes techniques that permit the capture of atomic scale pictures of membrane proteins in their lipid and protein environment to make “movies” from different instant pictures that will describe membrane protein functioning. It presents techniques scaling up from atomic to molecular that will render protein complexes in membrane of organelles and cells. The first section presents various strategies to purify membrane proteins since getting pure and homogenous material is a significant hurdle. Chapter 1 describes some techniques to characterize membrane protein preparations such as detergent content. Chapter 2 is devoted to the specific case of the adenosine nucleotide transporter (ANT), which a natural abundance in the inner membrane of mitochondria has permitted its three-dimensional structure determination, whereas structure–function relationships have been studied using

v

vi

Preface

mutants over expressed in yeast. Chapter 3 focuses on the importance of overexpression when membrane proteins are not naturally abundant. The specific case of a bacterial expression system used for a small mitochondrial membrane protein, the translocase TSPO is presented. Chapter 4 highlights the difficulties encountered for large membrane proteins overexpression; it describes a different expression system used for the specific case of an ABC transporter. The second section presents various strategies to get three-dimensional crystals and solve the structure by X-ray diffraction. Chapter 5 describes the various steps of Â�membrane protein crystallography in two different approaches that are vapor diffusion and lipidic phases. Chapter 6 discusses the gain for a membrane protein family to solve the atomic structure of one of its members. Chapter 7 analyzes what can be learned about the function of a single protein from its various atomic structures through the example of the sarcoplasmic calcium pump (SERCA-ATPase). Chapter 8 presents recent progress in the study of a membrane protein with high potential as a pharmaceutical target, the G protein-coupled receptor (GPCR) family. The third section presents the various possibilities to gain structural information for a membrane protein using electron microscopy observations. Chapter 9 uses the insect aquaporin AQPcic to go from its characterization in situ to its homotetrameric structure of purified protein reconstituted in membrane. Chapter 10 describes two-dimensional crystal formation and basic electron microscopy image analysis of membrane proteins. Chapter 11 presents a specific combination of cryo-electron tomography and single particle analysis of membrane protein embedded in stacked lipid bilayers. Chapter 12 describes, step-by-step, the process of electron tomography of mitochondria containing numerous membrane proteins. Chapter 13 is devoted to molecular modeling processes that permit to reach atomic structure of membrane protein conformation, combining its electron microscope derived map and atomic structure from a different conformation. The fourth section presents recent advances in nuclear magnetic resonance (NMR) to study membrane proteins and lipids. Chapter 14 goes through the various strategies that are available to solve atomic structure or protein–protein and protein–ligand interactions using different NMR approaches. Chapter 15 is devoted to the analysis of what can be learned from the structure of membrane protein fragments in regard to the overall protein. Chapter 16 used the peculiar example of the phospholamban to show by NMR analysis the structural dynamic of regulation of a membrane protein by a smaller interacting membrane protein. Chapter 17 describes step-by-step detergent solubilized membrane protein structure determination by solution-state NMR. Chapter 18 presents how solid-state NMR is a powerful tool to study lipid structure and dynamics in a membrane environment. The fifth section presents molecular modeling strategies that can be used either to get membrane protein structures or to move from atomic structure to dynamic understanding of a molecular functioning mechanism. Chapter 19 goes through the various possibilities to build and to analyze membrane protein models. Chapter 20 describes step by step how to build a three-dimensional model of a membrane protein. Chapter 21 presents molecular dynamics of membrane peptides and proteins in their lipid environment. Chapter 22 further describes membrane protein dynamics, presenting increasing time scale ranging from femtoseconds to seconds. Chapter 23 shows synergy between experimental data and computational modeling to delineate the ligand binding pocket of a GPCR, a step toward a rational for drug design. Paris, France

Jean-Jacques Lacapère

Contents Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Contributors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

v ix

Part Iâ•… Membrane Protein Purification â•‡ 1 Characterization of Membrane Protein Preparations: Measurement of Detergent Content and Ligand Binding After Proteoliposomes Reconstitution . . . . . . . . . . . . . . . . . å°“. . . . . . . . . . . . . . . . Mariano A. Ostuni, Soria Iatmanen, David Teboul, Jean-Claude Robert, and Jean-Jacques Lacapère â•‡ 2 Native Membrane Proteins vs. Yeast Recombinant: An Example: The Mitochondrial ADP/ATP Carrier . . . . . . . . . . . . . . . . . å°“. . . . . . . . . . . . . . . . Bertrand Arnou, Cécile Dahout-Gonzalez, Ludovic Pelosi, Guy J.-M. Lauquin, Gérard Brandolin, and Véronique Trézéguet â•‡ 3 Bacterial Overexpressed Membrane Proteins: An Example: The TSPO . . . . . . . . . . . . . . . . . å°“. . . . . . . . . . . . . . . . . . å°“. . . . . . . . . . . . . . . . . . å°“. Jean-Claude Robert and Jean-Jacques Lacapère â•‡ 4 Insect Cell Versus Bacterial Overexpressed Membrane Proteins: An Example, the Human ABCG2 Transporter . . . . . . . . . . . . . . . . . å°“. . . . . . . . . . Alexandre Pozza, José M. Pérez-Victoria, and Attilio Di Pietro

3

19

29

47

Part IIâ•… X-Ray Crystallography â•‡ 5 Crystallography of Membrane Proteins: From Crystallization to Structure . . . . . . 79 Aurélien Deniaud, Ekaterina Moiseeva, Valentin Gordeliy, and Eva Pebay-Peyroula â•‡ 6 Structural Approaches of the Mitochondrial Carrier Family . . . . . . . . . . . . . . . . . å°“ 105 Hugues Nury, Iulia Blesneac, Stephanie Ravaud, and Eva Pebay-Peyroula â•‡ 7 What Can Be Learned About the Function of a Single Protein from Its Various X-Ray Structures: The Example of the Sarcoplasmic Calcium Pump . . . . . . . . . . . . . . . . . å°“. . . . . . . . . . . . . . . . . . å°“ 119 Jesper Vuust Møller, Claus Olesen, Anne-Marie Lund Winther, and Poul Nissen â•‡ 8 Recent Progress in the Structure Determination of GPCRs, a Membrane Protein Family with High Potential as Pharmaceutical Targets . . . . . 141 Vadim Cherezov, Enrique Abola, and Raymond C. Stevens

Part IIIâ•…Electron Microscopy â•‡ 9 Observation of Membrane Proteins In Situ: AQPcic, the Insect Aquaporin Example . . . . . . . . . . . . . . . . . å°“. . . . . . . . . . . . . . . . . . å°“. . . . 171 Daniel Thomas and Annie Cavalier

vii

viii

Contents

10 Two-Dimensional Crystallization of Integral Membrane Proteins for Electron Crystallography . . . . . . . . . . . . . . . . . å°“. . . . . . . . . . . . . . . . David L. Stokes, William J. Rice, Minghui Hu, Changki Kim, and Iban Ubarretxena-Belandia 11 Structure Determination of Membrane Protein by Both Cryo-Electron Tomography and Single Particle Analysis . . . . . . . . . . . . . . . . . å°“. . Sylvain Trépout, Jean-Christophe Taveau, and Olivier Lambert 12 Electron Microscope Tomography of Native Membranes . . . . . . . . . . . . . . . . . å°“. . Gabriel Péranzi, Cedric Messaoudi, Leeyah Issop, and Jean-Jacques Lacapère 13 From Electron Microscopy Maps to Atomic Structures Using Normal Mode-Based Fitting . . . . . . . . . . . . . . . . . å°“. . . . . . . . . . . . . . . . . . å°“ Konrad Hinsen, Edward Beaumont, Bertrand Fournier, and Jean-Jacques Lacapère

187

207 221

237

Part IVâ•…Nuclear Magnetic Resonance 14 Determination of Membrane Protein Structures Using Solution and Solid-State NMR . . . . . . . . . . . . . . . . . å°“. . . . . . . . . . . . . . . . . . å°“. . . . . . . . . . Pierre Montaville and Nadège Jamin 15 Membrane Protein Fragments Reveal Both Secondary and Tertiary Structure of Membrane Proteins . . . . . . . . . . . . . . . . . å°“. . . . . . . . . . Philip L. Yeagle and Arlene D. Albert 16 What Can We Learn from a Small Regulatory Membrane Protein? . . . . . . . . . . . . Gianluigi Veglia, Kim N. Ha, Lei Shi, Raffaello Verardi, and Nathaniel J. Traaseth 17 Solution-State NMR Spectroscopy of Membrane Proteins in Detergent Micelles: Structure of the Klebsiella pneumoniae Outer Membrane Protein A, KpOmpA . . . . . . . . . . . . . . . . . å°“. . . . . . . . . . . . . . . Marie Renault, Olivier Saurel, Pascal Demange, Valérie Reat, and Alain Milon 18 NMR Spectroscopy of Lipid Bilayers . . . . . . . . . . . . . . . . . å°“. . . . . . . . . . . . . . . . . Axelle Grélard, Cécile Loudet, Anna Diller, and Erick J. Dufourc

261

283 303

321

341

Part Vâ•… Molecular Modelling 19 Critical Review of General Guidelines for Membrane Proteins Model Building and Analysis . . . . . . . . . . . . . . . . . å°“. . . . . . . . . . . . . . . . . . å°“. . . . . Catherine Etchebest and Gaelle Debret 20 3D Structural Models of Transmembrane Proteins . . . . . . . . . . . . . . . . . å°“. . . . . . . Alexandre G. de Brevern 21 Molecular Dynamics of Membrane Peptides and Proteins: Principles and Comparison to Experimental Data . . . . . . . . . . . . . . . . . å°“. . . . . . . . Patrick F.J. Fuchs 22 Membrane Protein Dynamics from Femtoseconds to Seconds . . . . . . . . . . . . . . . Christian Kandt and Luca Monticelli 23 The Family of G Protein-Coupled Receptors: An Example of Membrane Proteins . . . . . . . . . . . . . . . . . å°“. . . . . . . . . . . . . . . . . . å°“. . . . . . . . . . Irina G. Tikhonova and Daniel Fourmy

363 387

403 423

441

Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 455

Contributors Enrique Abolaâ•‡ •â•‡ Department of Molecular Biology, The Scripps Research Institute, La Jolla CA, USA Arlene D. Albertâ•‡ •â•‡ Department of Molecular & Cell Biology, University of Connecticut, Storrs, CT, USA Bertrand Arnouâ•‡ •â•‡ Laboratoire de Physiologie Moléculaire et Cellulaire, Institut de Biochimie et Génétique Cellulaires-UMR 5095, CNRS-Université Bordeaux 2, Bordeaux, France; CEA, iBiTecS (Institut de Biologie et Technologies de Saclay), CNRS, URA 2096, Gif-sur-Yvette, France Edward Beaumontâ•‡ •â•‡ INSERM U773, Centre de Recherche Biomédicale Bichat-Beaujon (CRB3), Faculté de Médecine X, Bichat, Université Paris 7, Paris, France Iulia Blesneacâ•‡ •â•‡ Institut de Biologie Structurale, CEA-CNRS and Université Joseph Fourier, Grenoble, France Gérard Brandolinâ•‡ •â•‡ Laboratoire de Biochimie et Biophysique des Systèmes Intégrés (BBSI), Institut de Recherches en Technologies et Sciences du Vivant (iRTSV), UMR 5092 CNRS-CEA-Université Joseph Fourier, Grenoble, France Alexandre G. de Brevern â•‡ •â•‡ INSERM UMR-S 665, Dynamique des Structures et Interactions des Macromolécules Biologiques (DSIMB), Institut National de Transfusion Sanguine (INTS), Université Paris Diderot - Paris 7, Paris, France Annie Cavalierâ•‡ •â•‡ CNRS, Interactions Cellulaires et Moléculaires, UMR 6026, Université de Rennes 1, Rennes, France Vadim Cherezovâ•‡ •â•‡ Department of Molecular Biology, The Scripps Research Institute, La Jolla, CA, USA Cécile Dahout-Gonzalezâ•‡ •â•‡ Laboratoire de Biochimie et Biophysique des Systèmes Intégrés (BBSI), Institut de Recherches en Technologies et Sciences du Vivant (iRTSV), UMR 5092 CNRS-CEA-Université Joseph Fourier, Grenoble, France Gaelle Debretâ•‡ •â•‡ Service d’Ingénierie MOléculaire des PROtéines (SIMOPRO), IbiTec-S, DSV, CEA, CE-Saclay, Gif-sur-Yvette, France Pascal Demangeâ•‡ •â•‡ UPS, Institut de Pharmacologie et de Biologie Structurale, Université de Toulouse, CNRS, UMR 5089, Toulouse, France Aurélien Deniaudâ•‡ •â•‡ European Molecular Biology Laboratory, Grenoble Outstation, B.P. 181, Grenoble, France Attilio Di Pietroâ•‡ •â•‡ Institut de Biologie et Chimie des Protéines, UMR5086 CNRS-Université Lyon 1 et IFR128 BioSciences Gerland, Lyon, France Anna Dillerâ•‡ •â•‡ Chimie et Biologie des Membranes et des Nanoobjets (CBMN), UMR5248, CNRS - Université Bordeaux - ENITAB, IECB, Pessac, France Erick J. Dufourcâ•‡ •â•‡ Chimie et Biologie des Membranes et des Nanoobjets (CBMN), UMR5248, CNRS - Université Bordeaux - ENITAB, IECB, Pessac, France

ix

x

Contributors

Catherine Etchebestâ•‡ •â•‡ INSERM UMR-S 665, Equipe Dynamique des Structures et des Interactions des Macromolécules Biologiques (DSIMB), Institut National de Transfusion Sanguine (INTS), Université Paris Diderot - Paris 7, Paris, France Daniel Fourmyâ•‡ •â•‡ INSERM, Institut National de la Santé et de la Recherche Médicale, Université de Toulouse 3, Toulouse, France Bertrand Fournierâ•‡ •â•‡ Laboratoire Léon Brillouin (CEA-CNRS), CEA Saclay, Gif-sur-Yvette, France Patrick F.J. Fuchsâ•‡ •â•‡ Equipe de Bioinformatique Génomique et Moléculaire, INSERM UMR-S726, Institut National de Transfusion Sanguine, Université Paris Diderot – Paris 7, Paris, France Valentin Gordeliyâ•‡ •â•‡ Institut de Biologie Structurale, CEA-CNRS and Université Joseph Fourier, Grenoble, France Axelle Grélardâ•‡ •â•‡ Chimie et Biologie des Membranes et des Nanoobjets (CBMN), UMR5248, CNRS - Université Bordeaux - ENITAB, IECB, Pessac, France Kim N. Haâ•‡ •â•‡ Department of Biochemistry, Molecular Biology, and Biophysics, University of Minnesota, Minneapolis, MN, USA Konrad Hinsenâ•‡ •â•‡ Centre de Biophysique Moléculaire (CNRS), Orléans, France; Division Expériences, Synchrotron SOLEIL, Saint Aubin, Gif-sur-Yvette, France; Laboratoire Léon Brillouin (CEA-CNRS), CEA Saclay, Gif-sur-Yvette, France Minghui Huâ•‡ •â•‡ New York Structural Biology Center, New York, NY, USA Soria Iatmanenâ•‡ •â•‡ INSERM U773, Centre de Recherche Biomédicale Bichat-Beaujon (CRB3), Faculté de Médecine X, Bichat, Université Paris 7, Paris, France Leeyah Issopâ•‡ •â•‡ INSERM U773, Centre de Recherche Biomédicale Bichat-Beaujon (CRB3), Faculté de Médecine X, Bichat, Université Paris 7, Paris, France Nadège Jaminâ•‡ •â•‡ CEA, iBiTecs, URA 2096, SB2SM, Gif-sur-Yvette, France Christian Kandtâ•‡ •â•‡ Computational Structural Biology, Chair of Life Science Informatics B-IT, Life & Medical Sciences (LIMES) Center, Rheinische Friedrich-Wilhelms-University Bonn, Bonn, Germany Changki Kimâ•‡ •â•‡ New York Structural Biology Center, New York, NY, USA Jean-Jacques Lacapèreâ•‡ •â•‡ INSERM U773, Centre de Recherche Biomédicale Bichat-Beaujon (CRB3), Faculté de Médecine X, Bichat, Université Paris 7, Paris, France Olivier Lambert â•‡ •â•‡ CBMN UMR-CNRS 5248, University of Bordeaux, Talence, France Guy J.-M. Lauquinâ•‡ •â•‡ Laboratoire de Physiologie Moléculaire et Cellulaire, Institut de Biochimie et Génétique Cellulaires-UMR 5095, CNRS-Université Bordeaux 2, Bordeaux, France Cécile Loudetâ•‡ •â•‡ IECB, UMS 3033, CNRS - Université Bordeaux - ENITAB, IECB, Pessac, France Cedric Messaoudiâ•‡ •â•‡ INSERM U759, Imagerie intégrative, Orsay, France; Laboratoire Raymond Latarjet, Centre Universitaire d’Orsay, Institut Curie, Centre de Recherche, Orsay, France Alain Milonâ•‡ •â•‡ UPS, Institut de Pharmacologie et de Biologie Structurale, Université de Toulouse, CNRS, UMR 5089, Toulouse, France Ekaterina Moiseevaâ•‡ •â•‡ Institut de Biologie Structurale, CEA-CNRS and Université Joseph Fourier, Grenoble, France

Contributors

xi

Jesper Vuust Møllerâ•‡ •â•‡ Centre for Membrane Pumps in Cells and Disease – PUMPKIN, Danish National Research Foundation, Copenhagen, Denmark; Department of Physiology and Biophysics, University of Aarhus, Aarhus, Denmark Pierre Montavilleâ•‡ •â•‡ CEA, iBiTecs, URA 2096, SB2SM, Gif-sur-Yvette, France Luca Monticelliâ•‡ •â•‡ UMR-S665, DSIMB, INSERM, 6, rue Alexandre Cabanel, 75015, Paris, France Poul Nissenâ•‡ •â•‡ Centre for Membrane Pumps in Cells and Disease – PUMPKIN, Danish National Research Foundation, Copenhagen, Denmark; Department of Molecular Biology, University of Aarhus, Aarhus, Denmark Hugues Nuryâ•‡ •â•‡ Institut Pasteur, Unit if Structural Dynamics of Macromolecules, CNRS, URA 2185, Paris, France Claus Olesenâ•‡ •â•‡ Centre for Membrane Pumps in Cells and Disease – PUMPKIN, Danish National Research Foundation, Copenhagen, Denmark; Department of Physiology and Biophysics, University of Aarhus, Aarhus, Denmark Mariano A. Ostuniâ•‡ •â•‡ INSERM U773, Centre de Recherche Biomédicale Bichat-Beaujon (CRB3), Faculté de Médecine X, Bichat, Université Paris 7, Paris, France Eva Pebay-Peyroulaâ•‡ •â•‡ Institut de Biologie Structurale, CEA-CNRS and Université Joseph Fourier, Grenoble, France Ludovic Pelosiâ•‡ •â•‡ Laboratoire de Biochimie et Biophysique des Systèmes Intégrés (BBSI), Institut de Recherches en Technologies et Sciences du Vivant (iRTSV), UMR 5092 CNRS-CEA-Université Joseph Fourier, Grenoble, France Gabriel Péranziâ•‡ •â•‡ INSERM U773, Centre de Recherche Biomédicale Bichat-Beaujon (CRB3), Faculté de Médecine X, Bichat, Université Paris 7, Paris, France José M. Pérez Victoriaâ•‡ •â•‡ Instituto de Parasitología y Biomedicina “López-Neyra”, CSIC Parque Tecnológico de Ciencias de la Salud, Armilla, Granada, Spain Alexandre Pozzaâ•‡ •â•‡ Institut de Biologie et Chimie des Protéines, UMR5086 CNRS-Université Lyon 1 et IFR128 BioSciences Gerland, Lyon, France Stephanie Ravaudâ•‡ •â•‡ Institut de Biologie Structurale, CEA-CNRS and Université Joseph Fourier, Grenoble, France Valérie Reatâ•‡ •â•‡ UPS, Institut de Pharmacologie et de Biologie Structurale, Université de Toulouse, CNRS, UMR 5089, Toulouse, France Marie Renaultâ•‡ •â•‡ UPS, Institut de Pharmacologie et de Biologie Structurale, Université de Toulouse, CNRS, UMR 5089, Toulouse, France William J. Riceâ•‡ •â•‡ New York Structural Biology Center, New York, NY, USA Jean-Claude Robertâ•‡ •â•‡ INSERM U773, Centre de Recherche Biomédicale Bichat-Beaujon (CRB3), Faculté de Médecine X, Bichat, Université Paris 7, Paris, France Olivier Saurelâ•‡ •â•‡ UPS, Institut de Pharmacologie et de Biologie Structurale, Université de Toulouse, CNRS, UMR 5089, Toulouse, France Lei Shiâ•‡ •â•‡ Department of Chemistry, University of Minnesota, Minneapolis, MN, USA Raymond C. Stevensâ•‡ •â•‡ Department of Molecular Biology, The Scripps Research Institute, La Jolla, CA, USA David L. Stokesâ•‡ •â•‡ Skirball Institute of Biomolecular Medicine, New York University

xii

Contributors

School of Medicine, New York, NY, USA; New York Structural Biology Center, New York, NY, USA Jean-Christophe Taveauâ•‡ •â•‡ CBMN UMR-CNRS 5248, University of Bordeaux, Talence, France David Teboulâ•‡ •â•‡ INSERM U773, Centre de Recherche Biomédicale Bichat-Beaujon (CRB3), Faculté de Médecine X, Bichat, Université Paris 7, Paris, France Daniel Thomasâ•‡ •â•‡ CNRS, Interactions Cellulaires et Moléculaires, UMR 6026, Université de Rennes 1, Rennes, France Irina G. Tikhonovaâ•‡ •â•‡ INSERM, Institut National de la Santé et de la Recherche Médicale, Université de Toulouse 3, Toulouse, France Nathaniel J. Traasethâ•‡ •â•‡ Department of Biochemistry, Molecular Biology, and Biophysics, University of Minnesota, Minneapolis, MN, USA Sylvain Trépoutâ•‡ •â•‡ CBMN UMR-CNRS 5248, |University of Bordeaux, Talence, France Véronique Trézéguetâ•‡ •â•‡ Laboratoire de Physiologie Moléculaire et Cellulaire, Institut de Biochimie et Génétique Cellulaires-UMR 5095, CNRS-Université Bordeaux 2, Bordeaux, France Iban Ubarretxenaâ•‡ •â•‡ Department of Structural and Chemical Biology, Mt. Sinai School of Medicine, New York, NY, USA Gianluigi Vegliaâ•‡ •â•‡ Department of Chemistry and Department of Biochemistry, Molecular Biology, and Biophysics, University of Minnesota, Minneapolis, MN, USA Raffaello Verardiâ•‡ •â•‡ Department of Biochemistry, Molecular Biology, and Biophysics, University of Minnesota, Minneapolis, MN, USA Anne-Marie Lund Wintherâ•‡ •â•‡ Centre for Membrane Pumps in Cells and Disease – PUMPKIN, Danish National Research Foundation, Copenhagen, Denmark; Department of Molecular Biology, University of Aarhus, Aarhus, Denmark Philip L. Yeagleâ•‡ •â•‡ Office of the Dean of Arts & Sciences, Rutgers University, Newark, NJ, USA

as

Part I Membrane Protein Purification

as

Chapter 1 Characterization of Membrane Protein Preparations: Measurement of Detergent Content and Ligand Binding After Proteoliposomes Reconstitution Mariano A. Ostuni, Soria Iatmanen, David Teboul, Jean-Claude Robert, and Jean-Jacques Lacapère Abstract The study of membrane proteins is a difficult task due to their natural embedding in hydrophobic environment made by lipids. Solubilization and purification from native membranes or overexpressed system involves the use of detergent to make them soluble while maintaining their structural and functional properties. The choice of detergent is governed not only by their ability to reach these goals, but also by their compatibility with biochemical and structural studies. A different detergent can be used during purification, and characterization of the detergent amounts present in each purification step is crucial. To address this point, we developed a colorimetric method to measure detergent content in different preparations. We analyzed detergent present in the collected fractions from the purification of the recombinant membrane translocator protein (RecTSPO). We followed detergent removal during the reconstitution of RecTSPO in liposomes and observed by electron microscopy the formation of proteoliposomes. We addressed the RecTSPO functionality by testing its ability to bind high affinity drug ligand [3H]PK 11195. We described the different parameters that should be controlled in order to optimize the measurement of this ligand binding using a filtration procedure. These protocols are useful to characterize functionality and detergent content of membrane protein, both key factors for further structural studies. Key words: Detergent, DPC, Mitochondrial membrane protein, Mitochondrial TSPO, PBR, Purification, Reconstitution, SDS

1. Introduction The quantity and distribution of naturally expressed membrane protein is a limiting step for functional and structural studies, which usually require important amounts of highly purified and concentrated protein (1). Along the past decades, various Jean-Jacques Lacapère (ed.), Membrane Protein Structure Determination: Methods and Protocols, Methods in Molecular Biology, vol. 654, DOI 10.1007/978-1-60761-762-4_1, © Springer Science+Business Media, LLC 2010

3

4

Ostuni et al.

strategies have been developed to either purify or overexpress membrane proteins. These strategies include the isolation and enrichment of naturally abundant proteins from their original tissues or the heterologous overexpression production systems and purification protocols (2). In this latter case, several vectors, constructions, and expression systems, including eukaryotes organism and cells, prokaryotes, and acellular systems (see further chapters), have been developed to optimize both expression and purification yields. Although there are exceptional cases of naturally abundant proteins, which can be highly enriched in a few steps (e.g., sarco– endoplasmic reticulum calcium ATPase (SERCA) from rabbit fast skeletal muscle or adenine nucleotide transporter (ANT) from heart liver mitochondria), membrane protein production protocols often include numerous purification steps mostly in the presence of a detergent (3). The choice of the detergent is complex since, on one hand, the membrane protein should be solubilized but remaining functional, (4) and on the other hand, the detergent should be compatible with biochemical and structural studies. For instance, noncharged detergents are a requisite for ion desorption experiments (MALDI-TOF), and deuterated detergents are needed for nuclear magnetic resonance (NMR) studies. Furthermore, depending on structural studies performed, additional steps might be needed in order to eliminate, diminish, or replace the detergent used to purify the membrane protein. The other major component of a membrane protein preparation is the lipid environment. Lipids could be present from extraction, or added or even exchanged, depending on the purification process. Characterization of the various components present in the final membrane protein preparation is required to assure not only the reproductibleness of experiments, but also the knowledge of the structural and the functional states of the protein. Indeed, any additional step in purification protocols implies the possibility of introducing an unexpected structural modification that may lead to misfolding and/or loss of function. This characterization is important to avoid the loss of time and money used in the analysis of improper samples and it may include several measurements: (1) the assessment of protein purity on gels and the characterization of contaminant proteins as well as the polymeric state of the protein of interest; (2) the analysis of the secondary and tertiary folding not only by spectroscopic (circular dichroism (CD)) techniques, but also by other biophysical techniques (5); (3) the determination of the protein, lipid, and detergent ratios that can be gained from the individual measurement of each component; and (4) the functionality of the purified membrane protein. This can be determined in detergent/protein complexes and/or in reconstituted proteoliposomes.

Characterization of Membrane Protein Preparations: Measurement of Detergent Content

5

Protein concentration determination may be difficult due to the presence of detergent that perturbs classical colorimetric measurements. However, protein concentration could be easily addressed using a nondestructive method based on the absorbance (in the aromatic region at 280Â€nm), in the presence of the detergent, and the calculated extinction coefficient (6). Several protocols of lipid extraction, purification, and characterization have been described from thin layer chromatography (7) to more recent mass spectroscopy (8), and they usually enable good lipid content determination. Detergent content measurements are accurately determined when radioactive forms exist. Some other approaches, such as the use of specific electrodes (9) or thin layer chromatography (10), have been described over the last decades to measure detergent content. We present here methods to address this latter point in different specific cases such as the measurement of detergent present in the membrane protein preparation and the following of detergent removal during membrane protein reconstitution in liposomes. The formation of proteoliposomes is a simple method to assess the protein functionality when coupled to ligand binding for a receptor or transport for a translocator. We illustrate the use of these methods taking advantage of the specific case of recombinant TranSlocator PrOtein (RecTSPO). This membrane protein previously named peripheral-type benzodiazepine receptor (PBR) is a transmembrane protein mostly located in mitochondria (11), whose expression and purification will be discussed in another chapter.

2. Materials 1. Absorption measurements: DO and absorption spectra were measured using an UV-300 Unicam UV-visible spectrometer (ThermoFisher Scientific, Courtaboeuf, France). 2. Phosphate buffer saline (PBS): 50Â€mM H2NaPO4; 150Â€mM NaCl; pH: 7.6. 3. Bio-Rad Protein assay kit and Bio-Beads SM2 (25–50 mesh) were purchased by Bio-Rad (Marne la Coquette, France). 4. Lipid solution: A mixture of dimyristoylphosphatidylcholine (DMPC) and dimyristoyl phosphatidylethanolamine (DMPE) (Avanti Polar Lipids, Alabaster, AL, USA) at a ratio of 9 to 1 weight/weight were mixed in buffer solution and sonicated in a water bath FB 15049 (Fischer Scientific, Illkirch, France) to remove multilamellar vesicles. 5. Detergents used: Sodium dodecyl sulfate (SDS, Sigma, SaintQuentin Fallavier, France) and dodecylphosphocholine (DPC, COGER, Paris, France).

6

Ostuni et al.

6. Protein used: RecTSPO (mouse recombinant TSPO) was expressed in E. Coli Bl21 bacteria (Invitrogen, Paisley, UK), extracted from inclusion bodies with SDS, purified by affinity chromatography on superflow Ni-NTA resin (Qiagen SA, Courtaboeuf, France), and protein concentration was determined from absorption spectra using extinction coefficient of 3.88/(mg/ml)–1 cm–1. 7. Radioactivity: Labeled PK 11195 ([3H]PK 11195) was purchased by New England Nuclear, NEN Life Science Products (Boston, MA, USA). Whatman Filters GF/C and PK 11195 were purchased by Sigma-Aldrich (Saint Quentin Fallavier, France). BSC liquid scintillation cocktail (GE Healthcare Europe, Saclay, France) and Wallac 1409 liquid scintillation counter (Perkin Elmer, Les Ulis, France). 8. Electron microscopy: JEOL 1200EX Transmission Electron microscope equipped with LaB6 filament operated at 120Â€kV and a sample holder (JEOL EM-SQH10). Cupper grids (400 mesh, Delta Microscopies, Labege, France) were covered with carbon film and negatively stained with 2% uranyl acetate (Sigma-Aldrich, Saint Quentin Fallavier, France).

3. Methods One of the key components of membrane protein purification is detergent, since its presence makes water soluble a protein naturally embedded in lipid bilayers. During purification process from either naturally abundant membrane protein as well as from overexpression systems, amount of detergent is usually only characterized from the initial detergent concentration in the used solutions. However, centrifugation, affinity column purification or dialysis steps can differentially concentrate detergent and proteins. Thus, quantification of detergent content of membrane protein preparation before structural or functional analysis is crucial (see Chapter 3 by Robert and Lacapere in this book). 3.1. Detergent Concentration Determination

It has been described for a long time that the presence of detergent modifies or impairs the colorimetric dosage of proteins by Bradford-based protein assay kits (Bio-Rad). We take advantage of the detergent interference to directly measure the effect of detergent addition to this colorimetric dye. 1. Standard calibration of sodium dodecyl sulfate (SDS) and dodecylphosphocholine (DPC) in the absence of protein: the Bio-Rad dye reagent was prepared by fivefold dilution in distilled water. 1Â€ml of diluted reagent was distributed into test tubes, and 0–300Â€ µg of SDS or DPC were added.

Characterization of Membrane Protein Preparations: Measurement of Detergent Content

7

Absorption spectra were recorded (from 250 to 800Â€nm) for each SDS or DPC containing solutions (Figs.Â€1a and 2a). The spectrophotometer saturation level was around OD value as great as 3 (see left side of the spectra). The spectra exhibit a shift of the peak, and the maximal change of absorbance induced by SDS and DPC was observed at 650 and 600Â€nm, respectively. Thus, OD for the various concentrations of SDS and DPC was measured at 650 and 600Â€ nm for SDS and a

b 3

d 2

c b

1

Optical Density

Optical Density

3

d 2

c b m

1

a 0 200 300 400 500

a 0 200 300 400 500

600 700 800

Wavelength, nm

d

2

SDS, mg/mL

c

1,5

e

2,5

25 20 15 10 5 0

1 0,5 0 0

100 200 300

400 500 600

SDS, µg/ ml

RecTSPO, mg/mL

Optical Density at 650 nm

3

600 700 800

Wavelength, nm

20 15 10 5 0 0

5

10

Elution volume, mL

Fig.Â€1. Effect of SDS on the light absorption of the Bradford-based Bio-Rad dye. Panel (a) shows the effect of the addition of different quantities of SDS (traces a, b, c, and d correspond to 0, 75, 125, and 200Â€µg/ml SDS, respectively) on the Bio-Rad dye absorption spectra. Several peaks could be observed, but largest differences in optical density due to SDS were obtained at 650Â€nm. Panel (b) shows the effect of methanol on the dye spectrum in the presence of different quantities of SDS (trace a, dye solution alone; trace m, in the presence of 3% methanol; traces b, c, d, and e, in the presence of methanol and 75, 125, and 200Â€µg/ml SDS, respectively). Panel (c) shows plotted optical density obtained at 650Â€nm without (filled circles) or with 3 and 7.5% added methanol (open triangles and diamonds, respectively). Solid and dashed lines depict the linear portion of the titration curves (100–300 and 200–500Â€µg/ml). Panels (d) and (e) show chromatograms of RecTSPO purification in the presence of SDS. Detergent contents in the eluted fractions (closed triangles in (d)) were measured by precipitating the protein and then mixing an aliquot with Bio-Rad dye to get detergent concentration in the supernatant. Protein contents in the eluted fractions (opened circles in (e)) were measured by recording absorption spectra of each fraction and calculating protein concentration using extinction coefficient of 3.88/(mg/ml)–1 cm–1 at 280Â€nm.

8

Ostuni et al.

a

b 3 Optical Density

Optical Density

3

d

2

c b

1

2

d c

1

m/b

a 0 200 300 400 500 600 700 800

0 200 300 400 500 600 700 800

Wavelenght, nm

Wavelenght, nm

d

3

Optical Density at 450 nm

Optical Density at 600 nm

c

a

2,5 2 1,5 1 0,5 0

1,6 1,4 1,2 1 0,8 0,6

0

100

200

DPC, µg/ml

300

0

100

200

300

DPC, µg/ml

Fig.Â€2. Effect of DPC on the light absorption of the Bradford-based Bio-Rad dye. Panel (a) shows the effect of the addition of different quantities of DPC (traces a, b, c, and d correspond to 0, 75, 125, and 200Â€µg/ml DPC, respectively) on the Bio-Rad dye absorption spectra. Several peaks could be observed but largest differences in optical density were obtained at 600Â€nm. Panel (b) shows the effect of the presence of methanol on the addition of different quantities of DPC (trace a, dye solution alone; trace m, in the presence of methanol; traces b, c, d, and e, in the presence of methanol and 75, 125, and 200Â€µg/ml DPC, respectively). Panel (c) shows plotted optical density obtained at 600Â€nm without (filled circles) or with (open triangles) methanol. Solid and dotted lines depict the linear portion of the titration curve in the absence and in the presence of methanol, respectively. Panel (d) shows plotted OD obtained at 450Â€nm without (filled circles) or with (open triangles) methanol. Solid line depicts the linear portion of the titration curve (100–300Â€µg/ml).

DPC, respectively, and data were plotted versus standard concentrations (Figs.Â€ 1c and 2c). Both curves gained in the presence of SDS and DPC exhibit a linear interval permitting accurate dosages of detergent in the range of 50–200 and 100–250Â€µg/ml for SDS and DPC, respectively (see Note 1). 2. Assay for protein containing samples: Bio-Rad dye reagent is protein sensitive; thus, to avoid cross-reaction with detergent staining, protein can be precipitated in methanol. One volume of protein samples was added to three volumes of methanol in a 1.5-ml tube and strongly mixed. A white precipitate

Characterization of Membrane Protein Preparations: Measurement of Detergent Content

9

is formed and centrifuged for 5Â€min at 200â•›×â•›g. The amounts of detergents present in the supernatant can be determined using standard curves performed in the presence of methanol, since small amount of methanol (3%, accordingly to final concentration in the spectrometer cell) has a small effect on Bio-Rad dye reagent absorption spectrum (Figs.Â€1b and 2b). Absorption spectra were recorded (from 250 to 800Â€nm) for each SDS and DPC containing solutions in the presence of 3% methanol. MeOH shifts the calibration curve of SDS and DPC when looking absorption at 600Â€nm (Figs.Â€1c and 2c), but the effect is smaller for SDS than for DPC. Indeed, a clear effect is observed for 7.5% MeOH on the calibration curve of SDS, whereas only 3% has a significant effect on DPC calibration curve (Figs.Â€1c and 2c). Methanol also induced a small decrease at 450Â€nm of absorption spectrum of Bio-Rad dye reagent. DPC calibration curves performed at this absorption wavelength (Fig.Â€2d) shows that methanol has no effect in the linear part of the curve, for 100–250Â€mg of DPC (see Note 1). 3. Application to membrane protein purification: RecTSPO is overexpressed in E. coli cells (see Chapter 3 by Robert and Lacapere in this book), extracted from inclusion bodies with SDS and purified by immobilized metal ion affinity chromatography in the presence of detergent. SDS-solubilized RecTSPO is loaded into Ni-NTA resin, column is washed and protein eluted with imidazole and SDS containing buffer. Collected fractions were analyzed for detergent and protein content and chromatogram drawn (Fig.Â€1d, e). An aliquot of each fraction was mixed with methanol (see above) and detergent content of the supernatant determined using calibration curves (see Note 2). An absorption spectrum of each fraction (diluted if needed) was recorded and protein contents determined using extinction coefficient of 3.88/(mg/ml)–1 cm–1 at 280Â€nm. 3.2. Detergent Removal Using Bio-Beads

Bio-Beads are macroporous polystyrene beads used in hydrophobic interaction chromatography (12). They have a high surface area for adsorbing organics of molecular weight less than 2,000 from aqueous solutions. The Bio-Beads are useful for the adsorption of nonpolar substances or surface active agents such as detergents. Kinetics of detergent removal by theses Bio-Beads are more or less specific for each detergent (13–15). The above described technique can be used to follow detergent removal from a solution by measuring detergent content of aliquots. 1. Bio-Beads preparation: Dry SM2 Bio-Beads were extensively washed before use, firstly with methanol (three times) to eliminate impurities and polystyrene dust and then with water (four times) to hydrate them.

10

Ostuni et al.

2. Kinetics of detergent removal: 10–15Â€ mg detergents were added to 5Â€ ml solution of PBS. Solution was gently stirred and kept at room temperature (25°C). At various time intervals, weighted amount of wet Bio-Beads were added (see Note 3). Aliquots of 40Â€µl were removed at several time points and their detergent content was determined by mixing in 1Â€ml containing Bio-Rad dye solution (1:5 dilution), which OD was measured at the detergent corresponding wavelength (Fig.Â€3a, b) (see Note 2). 3. Calculation of binding capacity of Bio-Beads: Capacity was defined as the total amount of detergent that was removed by 1Â€g of Bio-Beads. The amount of detergent removed by each

a

b 10 75 mg

14

Total DPC in solution, mg

Total SDS in solution, mg

16 25 mg

12 10 8 6 4 2 0

9

100 mg

8

25 mg

7 6 5 4 3 2 1 0

0

30 60 90 120 150 180 210 240

0

30

60

Time, min

d

8

7

6

6

5 4 3 2 1 0

120

150

180

8

7 DPC removed, mg

SDS removed, mg

c

90

Time, min

5 4 3 2 1

0

25

50

75

Bio-Beads, mg

100

0

0

25

50

75

100

Bio-Beads, mg

Fig.Â€3. Detergent absorption by Bio-Beads. Panel (a) shows the time course of SDS removal by the repeated addition (signaled by arrows) of 25 (closed squares) or 75 (closed triangles) mg of BB in 6Â€ml solution containing ~11–13Â€mg total SDS. Panel (b), kinetics of DPC elimination in the presence of two different amounts of Bio-Beads (closed squares and triangles for 25 and 100Â€mg, respectively). Panels (c) and (d) show the quantity of SDS and DPC absorbed as function of amount of wet Bio-Beads added. A binding capacity of 70Â€ mg SDS or DPC per g of Bio-Beads was calculated from respective slopes.

Characterization of Membrane Protein Preparations: Measurement of Detergent Content

11

addition of Bio-Beads was measured and data plotted (Fig.Â€3c, d for SDS and DPC, respectively). The slope of the curves permitted to calculate a value of 70–80Â€ mg SDS or DPC removed by gram of Bio-Beads (in agreement with previously published results for SDS (15). 3.3. Proteoliposomes Formation

Protein incorporation into liposomes is a powerful tool to investigate both functional and structural aspects of membrane proteins. Several strategies have been developed to achieve functional proteoliposomes including the use of organic solvent, mechanical protocols and the use of detergents (16). The latest is the most frequently chosen strategy as detergents are usually present in isolation and purification protocols. Standard procedures to incorporate detergent solubilized membrane protein in lipid bilayer involves different steps: (1) mixing of membrane protein solubilized in detergent with lipids to form a ternary complex, (2) exchange of detergent surrounding protein by lipid detergent, and (3) bilayers formation, mostly vesicle formation. Detergent removal can be followed by the above described method (Fig.Â€4a, d) and vesicle formation by measuring light scattering changes (Fig.Â€ 4b, e). Finally, proteoliposomes formation as well as the evolution in size and shape of the objects present in the solution can be assessed by electron microscopy (Fig.Â€4c, f). 1. A mixture of DMPC/DMPE (9/1 w/w) lipids at a final concentration of 0.5–1Â€ mg/ml is added to a 6-ml solution of PBS. 2. Membrane protein solubilized in detergent at a final protein concentration of 0.1–0.2Â€ mg/ml is added to the lipid containing solution. 3. OD at 550Â€nm of the solution is measured before and after the addition of protein. The presence of detergent from the purified membrane protein solution is usually enough to solubilize the lipids and is clearly observed by an important decrease in OD. At this wavelength, most proteins do not absorb and the size of the detergent, protein, lipids complexes are too small to induce large light scattering. 4. Detergent content of the starting solution: Aliquots of 40Â€µl was removed and mixed in 1Â€ml containing Bio-Rad dye solution (1:5 dilution). OD was measured at the detergent corresponding wavelength (Fig.Â€ 3a, b) and detergent content calculated from calibration curves (Figs.Â€1c and 2c). 5. For negative staining, 5Â€µl of the reconstitution samples were applied to carbon-coated grids, blotted, and stained with 1% uranyl acetate and the specimens were observed with an electron microscope (Fig.Â€4c, f).

12

Ostuni et al.

a

d 2

1,0

DPC, mg/ml

SDS mg/ml

1,5

0,5

0,0 0

30

60

90

1

0

120 150

0

60

Time, min

0,6 0,5 0,4 0,3

a b

c

d

0,2 0,1 0,0

180

e

0,7

0

60 120 180 240 Over night Time, min

c

0,7

Optical Density at 550 nm

Optical Density at 550 nm

b

120

Time, min

0,6 0,5 0,4 0,3

b a

0,2

d

0,1 0

c

0 30 60 90 120150 Over night Time, min

f a

b

a

b

c

d

c

d

200 nm

200 nm

Fig.Â€4. Formation and characterization of RecTSPO containing proteoliposomes. Reconstitution process was followed by measuring detergent removal and vesicle formations from a solution containing lipids, RecTSPO and detergent (panels (a–c) and (d–f) for RecTSPO purified in SDS or DPC, respectively). Panel (a), time course of SDS removal by a single addition (signaled by arrows) of large amount (closed circles) or repeated addition of small amounts (closed triangles) of Bio-Beads. Panel (b), time course of light scattering changes upon SDS removal by a single (closed circles) or repeated addition of small amounts (closed triangles) of Bio-Beads. Panel (c), electron micrographs of negatively stained samples taken at different time of SDS elimination (0, 30, 90Â€min, and over night, as indicated in panel (b)). Panel (d), time course of DPC removal by a single addition (signaled by arrows) of large (closed squares) or small amount (closed circles) of Bio-Beads. Panel (e), time course of light scattering changes upon DPC removal by a single addition of large (closed squares) or small amounts (closed circles) of Bio-Beads. Panel (f), electron micrographs of negatively stained samples taken at different time of DPC elimination (0, 120 and 30, 90Â€min from large and small Bio-Beads additions, as indicated in panel (e)).

Characterization of Membrane Protein Preparations: Measurement of Detergent Content

13

6. Detergent present in the solution is removed by addition of Bio-Beads. The quantity of added Bio-Beads regulates the rate of vesicle formation by controlling the amount of detergent removed (compare Fig.Â€4a–c for SDS and 4d–f for DPC). Addition of Bio-Beads mass close to the binding capacity can be chosen to get a fast reconstitution, whereas addition of a smaller mass generates a slower reconstitution process. 7. Formation of proteoliposomes is basically correlated to detergent removal, but several remarks should be done. (1) The kinetics of vesicle formation is not exactly parallel with that of the detergent removal as revealed by the difference observed between the measurements of detergent residual (Fig.Â€4a, d for SDS and DPC, respectively) and the changes in OD550 (Fig.Â€ 4b, e for SDS and DPC, respectively). (2) The final OD550 reached after detergent removal correlates the size of the vesicles formed, but the size of the vesicles seems to vary with the type of detergent used to solubilize the protein (see Note 4). Comparison of Fig.Â€4c, f shows that vesicles are smaller when SDS was the solubilizing detergent. (3) In some cases, vesicles tend to fuse leading to an increase of OD550 well correlated with the observation of vesicle aggregation. 3.4. Ligand Binding

Ligand binding is a useful method to characterize functional state of a membrane protein. Several techniques can be used among which incubation of proteoliposomes in the presence of radioactive ligands is a very sensitive one. Separation of bound and free ligand can be obtained by filtration of proteoliposomes. In this case, the choice of the filter is crucial. In the specific case of TSPO, whose high affinity drug ligands are hydrophobic compounds, preliminary studies have to be performed to optimize the ligand binding experiments. 1. Choice of the filter: They should retain the proteoliposomes and minimize the radioactive background (mostly due to filter dead volume and nonspecific binding on the filter). Different composition and pore size filters are commercially available, and TableÂ€ 1 shows the characteristics obtained with TSPO proteoliposomes and its high affinity drug ligand, PK 11195. Proteoliposome retention was performed by filtering a solution and measuring the TSPO intrinsic fluorescence of the filtrate compared to the proteoliposmes solution. The radioactive background was calculated by measuring retained radioactivity in the filters in the absence of proteoliposomes. TableÂ€ 1 shows that the best filters to measure [3H]PK 11195 binding to RecTSPO are the Whatman GF/C and that background can be reduced by washing the filters with buffer.

14

Ostuni et al.

TableÂ€1 Different filters tested for measuring radioactive ligand binding to TSPO Filter

Proteoliposomes retention (% of total)

Background radioactivity (% of total) Nonwashed

Washed

Whatman GF/B

74â•›±â•›13

29.4

7.7

Whatman GF/C

62â•›±â•›12

24.1

2.6

Whatman GF/F

60â•›±â•›15

42.9

14.3

Millipore

86â•›±â•›10

94.3

70.5

For proteoliposome retention by the various filters, 1Â€ml of a solution containing 0.5–10Â€µg of RecTSPO was filtered and the protein content of the filtrate (i.e., nonretained on the filter) was measured by intrinsic fluorescence of the RecTSPO using a calibration curve. Results are expressed as percentage of total protein filtered. For ligand retention by the various filters, 0.3Â€ml of a solution containing cold and [3H] PK 11195 was filtered under the same condition of binding experiments, filters were washed or not two times with 4Â€ml cold PBS, and radioactivity trapped in the filters was measured in a liquid scintillation counter. Retained radioactivity was expressed as a percentage of total sample radioactivity. Based on the results showed in this table, the GF/C grade glass microfiber filters were chosen to perform RecTSPO binding studies

2. Experimental conditions: The signal-to-background ratio can be optimized adjusting protein concentration and incubation time. FigureÂ€ 5a shows that signal-to-background ratio increases linearly with RecTSPO amounts. For drug ligands, such as PK 11195, the affinity for RecTSPO is in the nanomolar range; thus, the optimal protein concentration used in the experimental condition of ligand binding should be lower to measure affinity constant. FigureÂ€5b shows the incubation time is also important for such hydrophobic ligand such as PK 11195. Indeed, the binding of PK 11195 to the RecTSPO is fast, but the PK 11195 bound and free decreases with time. The optimal signal-to-background ratio is observed for 5–15-min incubation (see Note 5). 3. Direct ligand binding experiments: Proteoliposomes were incubated in the presence of increasing concentration of [3H] PK 11195 and filtered. For each ligand concentration, the specific binding was calculated as the difference between total binding and nonspecific binding (measured in the presence of a large excess of cold PK 11195). FigureÂ€5c shows that [3H] PK 11195 bound increases with the total PK 11195 concentration and reaches a plateau corresponding to the saturation

Characterization of Membrane Protein Preparations: Measurement of Detergent Content

a

b

5000

3000 2000

4000 2000

1000 0

0 0

1 2 TSPO, (µg)

c

3

0

10

20 30 40 Time, (min)

50

60

d 120 [3H]PK 11195 bound, (%)

[3H]PK 11195 bound, (nmol/mg)

8000 6000

C.P.M.

C.P.M.

4000

15

40 30 20 10 0

0

10 20 PK 11195, (nM)

100 80 60 40 20 0 0,01

1 100 PK 11195, (nM)

10000

Fig.Â€5. Control of protein functionality by ligand-binding experiments. Panel (a) shows the ligand binding as function of reconstituted RecTSPO quantities incubated for 15Â€ min at 25°C. Circles indicate the radioactivity due to total [3H]PK 11195 binding, whereas triangles show the nonspecific binding to filters and membranes in the presence of a large excess of nonradioactive PK 11195. Panel (b) shows the time course of ligand binding using 2Â€ ng of reconstituted RecTSPO incubated at 25°C. Circles and triangles indicate the radioactivity due to total PK 11195 bound and nonspecific binding, respectively. Panel (c) shows the ligand-binding saturation curve obtained incubating 2Â€ ng of reconstituted RecTSPO for 15Â€min at 25°C. Binding of [3H]PK 11195 is expressed as nmol of bound ligand per mg of RecTSPO. Panel (d) shows [3H]PK 11195 binding of 2Â€µg of RecTSPO incubated for 30Â€min at 25°C in the presence of raising concentration of cold PK 11195.

of the RecTSPO binding sites. The binding constants (affinity constant, Kd, and maximal binding value, Bmax) can be obtained by fitting the saturation curve with the following equation Yâ•›=â•›(Bmaxâ•›×â•›S)/(Kdâ•›+â•›S), giving a Kd of 6.5â•›±â•›0.5Â€ nM and a Bmax of 43â•›±â•›8Â€nmol/mg protein. 4. Competition binding experiments: Proteoliposomes were incubated in the presence of a constant concentration of [3H] PK 11195 and rising concentration of cold PK 11195. The various solutions were filtered and the specific binding was calculated as the difference between total binding and nonspecific binding (the asymptotic value observed in the presence of a 1,000-fold excess of cold PK 11195). FigureÂ€ 5d

16

Ostuni et al.

shows that [3H]PK 11195 bound decreases with increasing concentration of cold PK 11195. The inhibition constant (IC50) can be obtained by fitting the curve with the following equation Yâ•›=â•›Max−(Maxâ•›×â•›S)/(Kdâ•›+â•›S), giving an IC50 of 10â•›±â•›2Â€nM. 3.5. Conclusions

The method presented here permits to reproducibly measure the amount of detergent present in various conditions during the processes leading from the purification, the functional characterization and the structure determination of a membrane protein. It has been characterized for detergent used to purify RecTSPO in SDS or DPC for the reconstitution of this membrane protein in liposomes and the further measurements of ligand binding. The results gained from this method permit to know the detergent at the different steps of sample preparations for functional and structural studies. It permits not only to follow complete detergent removal to form proteoliposomes used for ligand binding experiments but also to control the protein–detergent ratio, which is an important factor in the crystallization process.

4. Notes 1. It should be mentioned that calibration curves adjust to sigmoid function with threshold and saturation levels. The linear part corresponds to a narrow range of detergent concentrations but with an important change in OD giving a very sensitive method in this range of concentration (Figs.Â€1 and 2). The effect of methanol is small, but calibration curve has to be done taking into account the methanol content in the aliquot from the precipitate membrane protein. Moreover, volume added has to be as small as possible (10–40Â€µl in 1Â€ml) to reduce dilution effect. 2. When detergent content was too low to be in the linear part of the calibration curve, a first addition of well-characterized detergent solution was performed to reach the linear region and then an aliquot of the sample is added. The detergent concentration in the sample is finally calculated by difference. 3. Bio-Beads are prepared in aqueous solution, and thus, weighting needs water removal. However, beads also need to be maintained wet to keep their detergent absorption capacity. Real amounts of weighted Bio-Beads might vary from one operator to another since water present with the “wet” Bio-Beads might vary. The best is that each operator performs its own calibration curves.

Characterization of Membrane Protein Preparations: Measurement of Detergent Content

17

4. We have no clear explanation for the reasons leading to the formation of different sizes of vesicles when starting from RecTSPO purified in SDS or in DPC. It might be due to different types of interactions between detergent and proteins that either favor or delay lipid–proteins interactions. Different compositions of ternary complexes (protein–lipid–detergent) might be generated as function of the type of used detergent. This might lead to larger or smaller vesicles related to the lipid amounts around the protein. 5. Incubation time should not be extended too long since free PK 11195 diminishes with time. This might be due to nonspecific binding of hydrophobic PK 11195 to the glass tube surface.

Acknowledgment The authors would like to thank Professor V. Papadopoulos for the generous gift of TSPO plasmids, G. Péranzi and A. Letort for their contribution in preliminary experiments. This work was supported by CNRS (Centre National de la Recherche Scientifique) and ANR (Agence National pour la Recherche) Grant 06-Blan0190-01 to JJL. References 1. Lacapere J-J, Pebay-Peyroula E, Neumann J-M, Etchebest C (2007) Determining membrane protein structures: still a challenge! Trends Biochem Sci 32:259–270 2. Grisshammer R, Tate C (2003) Preface: overÂ� expression of integral membrane proteins. Biochim Biophys Acta 1610: 1 3. Garavito RM, Ferguson-Miller S (2001) Detergents as tools in membrane biochemistry. J Biol Chem 276:32403–32406 4. LeMaire M, Champeil P, Møller JV (2000) Interaction of membrane proteins and lipids with solubilizing detergents. Biochim Biophys Acta 1508:86–111 5. Pebay-Peyroula E (ed) (2008) Biophysical analysis of membrane protein. Investigating structure and function. Wiley-VCH, Weinheim 6. Gasteiger E, Gattiker A, Hoogland C, Ivanyi I, Appel RD, Bairoch A (2003) ExPASy: the proteomics server for in-depth protein knowledge and analysis. Nucleic Acids Res 31:3784–3788 7. Fleischer S, Rouser G, Fleischer B, Casu A, Kritchevski G (1967) Lipid composition of

8.

9.

10.

11.

mitochondria from bovin heart, liver, and kidney. J Lipid Res 8:170–180 Schiller J, Arnhold J, Benard S, Müller M, Reichl S, Arnold K (1999) Lipid analysis by matrix-assisted laser desorption and ionization mass spectroscopy: a methodological approach. Anal Biochem 267:46–56 Mokus M, Kragh-Hansen U, Letellier P, le Maire M, Møller JV (1998) Construction and use of a detergent-sensitive electrode to measure dodecyl sulfate activity and binding. Anal Biochem 264:34–40 Eriks LR, Mayor JA, Kaplan RS (2003) A strategy for identification and quantification of detergents frequently used in the purification of membrane proteins. Anal Biochem 323:234–241 Papadopoulos V, Baraldi M, Guilarte TR, Knudsen TB, Lacapere J-J, Lindemann P, Noremberg MD, Nutt D, Weizman A, Zhang M-R, Gavish M (2006) Translocator protein (18Â€kDa): new nomenclature for the peripheral-type benzodiazepine receptor based on

18

Ostuni et al.

its structure and molecular function. Trends Pharmacol Sci 27:402–409 12. Holloway PW (1973) A simple procedure for removal of triton X-100 from proteins samples. Anal Biochem 53:304–340 13. Rigaud J-L, Mosser G, Lacapere J-J, Olofson A, Levy D, Ranck J-L (1998) Bio-Beads: an efficient strategy for two-dimensional crystallization of membrane proteins. J Struct Biol 118:226–235 14. Lacapere J-J, Stokes DL, Olofsson A, Rigaud J-L (1998) Two-dimensional crystallization

of Ca-ATPase by detergent removal. Biophys J 75:1319–1329 15. Lacapere J-J, Delavoie F, Li H, Péranzi G, Maccario J, Papadopulos V, Vidic B (2001) Structural and functional study of reconstituted peripheral benzodiazepine receptor. Biochem Biophys Res Commun 284:536–541 16. Rigaud J-L, Pitard B, Levy D (1995) Reconstitution of membrane proteins into liposomes: application to energy-transducing membrane proteins. Biochim Biophys Acta 1231:223–246

Chapter 2 Native Membrane Proteins vs. Yeast Recombinant: An Example: The Mitochondrial ADP/ATP Carrier Bertrand Arnou, Cécile Dahout-Gonzalez, Ludovic Pelosi, Guy J.-M. Lauquin, Gérard Brandolin, and Véronique Trézéguet Abstract The mitochondrial ADP/ATP carrier (Ancp) has long been a paradigm for studies of the mitochondrial carrier family due to, among other properties, its natural abundance and the existence of specific inhibitors, namely, carboxyatractyloside (CATR) and bongkrekic acid (BA), which lock the carrier under distinct and stable conformations. Bovine Anc1p isolated in complex with CATR in the presence of an aminoxyde detergent (LAPAO) was crystallized and its 3D structure determined. It is the first mitochondrial carrier structure resolved at high resolution (2.2â•›Å, as reported by Pebay-Peyroula etÂ€al. (Nature 426:39–44, 2003)). Analyses revealed a monomer while most of the biochemical studies led to hypothesize Ancp functions as a dimer. To address the structural organization issue, we engineered a mutant of the yeast Ancp that corresponds to a covalent homodimer in view of 3D structure determination. We compare in this chapter the purification yield and quality of the chimera tagged either with six histidines at its C-ter end or nine histidines at its N-ter. We show that, as expected, length and position of the tag are important criteria for qualitative purification. We also discuss the advantages and drawbacks of purifying Ancp either from a natural source or from engineered yeast cells. Key words: Mitochondrial ADP/ATP carrier, Mitochondrial carrier, High yield purification

1. Introduction Major metabolic pathways occur within mitochondria, which is the primary site of ATP synthesis. It is isolated from the cytosol by two membranes (outer and inner) that are therefore important sites for the regulation of metabolic functions. The mitochondrial carrier family (MCF) members are integral membrane proteins that transport various metabolites through the mitochondrial inner membrane. Among them, the mitochondrial ADP/ATP carrier (Ancp) was the first to be identified around 40Â€years ago. Jean-Jacques Lacapère (ed.), Membrane Protein Structure Determination: Methods and Protocols, Methods in Molecular Biology, vol. 654, DOI 10.1007/978-1-60761-762-4_2, © Springer Science+Business Media, LLC 2010

19

20

Arnou et al.

It plays a key role in the energetic cell metabolism because it exchanges ATP and ADP, respectively product and substrate of the mitochondrial ATP synthase. Ancp is the most abundant among MCF members (up to 10% of mitochondrial proteins in beef heart mitochondria) and has to cope with high nucleotide amounts in order to fulfill cell energetic requirements. It can be purified pretty easily in one or two steps from beef mitochondria in high amount and in crystallization compatible quality. However, the high-resolution structure of the beef Anc1p in complex with its inhibitor carboxyatractyloside (CATR) evidenced a monomeric organization (1). This was contradictory to many previously published results about different MCF members, of which Ancp shares functional and structural features and was shown to be organized as dimers (for a review see (2)). We hypothesized that protein preparation, and more precisely, the protein concentration step necessary to crystallization trials was responsible for Ancp dimer dissociation. Therefore, we used an engineered Ancp mutant (3) that forces a dimeric organization of the carrier throughout the purification and crystallization processes; we named it AA in this chapter. However, we previously showed that it was necessary to tag the yeast Ancp to get it highly purified (4). Consequently, the chimera was tagged with either six histidines at its C-ter (AAH6) or nine histidines at its N-ter (H9AA).

2. Materials 2.1. Chemicals

1. n-dodecyl-b-d-maltoside (DDM) was purchased from Anatrace. [14C]DDM was kindly provided by Marc Le Maire (CEA, CNRS, Université Paris-Sud 11, France). 2. Atractyloside (ATR) and carboxyatractyloside (CATR) were obtained from Sigma. 3. The nickel-nitrilotriacetic acid agarose matrix (Ni-NTA) for metal affinity chromatography is from Qiagen. It is provided as a 1:1 (vol/vol) suspension in 50% ethanol (vol/vol). 4. 3-Laurylamido-N,N¢-dimethylpropylaminoxide (LAPAO) was synthesized as described in (5). It can now be purchased from Anatrace. 5. Hydroxyapatite Bio-Gel® was purchased from Bio-Rad. 6. Thrombin was from Sigma and used as recommended by the supplier.

2.2. Buffer Compositions

1. Buffer A: 150Â€mM NaPi pH 7.3, 10% (w/v) glycerol, 0.1% (w/v) DDM.

Native Membrane Proteins vs. Yeast Recombinant

21

2. Buffer B: 100Â€ mM Na2SO4, 10Â€ mM Tris–HCl pH 7.3 and 1Â€mM Na2-EDTA pH 8.0. 3. Buffer C: 33% (v/v) glycerol, 6Â€mM MgSO4, 150Â€mM NaPi pH 7.3. 4. Buffer D: 150Â€mM NaPi pH 7.3, 10% (w/v) glycerol, 0.05% (w/v) DDM, 40Â€mM imidazole pH 8.0. 5. Buffer E: 10Â€ mM MOPS-NaOH pH 6.8, 150Â€ mM NaCl, 10% (w/v) glycerol, 0.05% (w/v) DDM. 6. Buffer F: 500Â€mM imidazole pH 8.0, 10Â€mM MOPS-NaOH pH 6.8, 150Â€mM NaCl, 10% glycerol, 0.05% (w/v) DDM. 7. Buffer G: 10Â€ mM MOPS-NaOH pH 6.8, 150Â€ mM NaCl, 10% glycerol, 0.05% (w/v) DDM. 8. Buffer H: 100Â€mM NaCl, 10Â€mM Tris–HCl pH 7.4, 1Â€mM Na2-EDTA, 0.05% LAPAO (w/v). 9. Buffer I: 500Â€mM NaCl, 10Â€mM Tris–HCl pH 7.4, 1Â€mM Na2-EDTA. 10. Buffer J: 10Â€mM Tris–HCl pH 7.4, 1Â€mM Na2-EDTA, 0.05% LAPAO.

3. Methods 3.1. The 6-Histidine Tag: Mutagenesis, Plasmids, and Strains

1. Construction of the gene encoding Anc2p tagged with six histidines at its C-ter (AH6) is described in (4).

3.2. The 9-Histidine Tag: Mutagenesis, Plasmids, and Strains

1. The His9(ANC2)2 gene was obtained by site directed mutagenesis using the Transformer™ Site-directed Mutagenesis Kit (CLONTECH Laboratories). The protein produced from

2. To construct the gene coding for the covalent tandem dimer of Anc2p tagged with six histidines at its C-ter (AAH6), the wild type ScANC2 gene was amplified with the following primers: 5¢-GAATTCGGATCCATGTCTTCCAACGCCC AAGTCAAAA-3¢ and 5¢-CTGGATCCTTTGAACTTCTTA CCAAACAAGATC-3¢ to introduce BamHI sites (underlined) on each side of the ORF and remove the stop codon. The fragment was then subcloned into the unique BamHI site of the pAH6 plasmid (4), which contains the gene encoding for AH6. After control of the orientation of the fragment, the 3¢ ScANC2 terminator region (amplified with two primers containing each an XbaI site) was introduced. The resulting (PstI-NotI) fragment containing the promoter, the AAH6 encoding gene and the terminator was used for its integration into the JL1-3∆2 strain (6) by homologous recombination at the ANC2 locus. The resulting strain was named JL-AAH6.

22

Arnou et al.

this gene corresponds to a covalent tandem homodimer of Anc2p (3) tagged with nine histidines at its N-ter. 2. The mutagenesis target plasmid was obtained from KSDIM5¢3¢ (3) that was digested first by SalI and NotI to shorten the ScANC2 3¢ noncoding region. The resulting plasmid was named KSDIM5¢3¢∆SN. 3. The mutagenic primer 5¢catacatataagcaaatacaattgccATG GGT CAC CAT CAC CAC CAT CAC CAT CAC CAC TCT TCA GGT TTA GTT CCT AGA GGT TCT TCC AAC GCC CAA G3¢ was designed (1) to replace the single EcoRI site of KSDIM5¢3¢∆SN with a MfeI site (lower case, underlined) to select for mutagenesis events; (2) to introduce a cluster of nine histidines (capital, bold) at the N-terminus of the covalent tandem dimer of Anc2p (capital, italic); and (3) to introduce the recognition and digestion sites of thrombin (capital, underlined). Successful mutagenesis was assessed by DNA sequencing. The resulting plasmid was named KSH9DIM. 4. The His9ANC2 gene was obtained by digestion of KSH9DIM by HindIII and re-ligation of the biggest fragment to obtain a single copy of the ScANC2 gene. The resulting plasmid was named KSH9MON. 5. The KpnI-SacI fragment containing either the His9ANC2 or the His9(ANC2)2 gene was used for its integration into the JL1-3∆2 strain (6) by homologous recombination at the ANC2 locus. The resulting strain was named JL-H9A or JL-H9AA, and the produced protein is H9A or H9AA. 3.3. Isolation of Mitochondria from Yeast

1. Protocol and materials used to perform mitochondria isolation are described in (6). They are frozen in small beads in liquid nitrogen and stored at −70°C prior to protein purification (see Note 1). 2. In the case of H9AA and AAH6, a cocktail of protease inhibitors (1Â€ µg/mL pepstatin A, 1Â€ µg/mL leupeptin, 1Â€ µg/mL antipain, 5Â€ µg/mL aprotinin, and 1Â€ mM Na2-EDTA, final concentrations) is added to the buffers during mitochondria isolation to prevent protein degradation (7).

3.4. Isolation of Mitochondria from Bovine Heart 3.5. AAH6 Purification

Bovine mitochondria were isolated from heart muscle by differential centrifugation as described by (8). They were suspended in 0.27Â€ M sucrose, 2Â€ mM Tris–HCl pH 7.4 and stored in liquid nitrogen (see Note 1). 1. The protocol used to purify AH6 and described in (4) is applied to purify AAH6. It consists of three steps: hydroxyapatite Bio-Gel® chromatography, immobilized metal ion chromatography, and removal of imidazole by dialysis or AcA202 chromatography.

Native Membrane Proteins vs. Yeast Recombinant

3.6. H9A and H9AA Purifications

23

1. The Ni-NTA resin (Qiagen) (1Â€ ml resin for 5Â€ mg total protein) is washed twice with 5Â€ volÂ€ H2O then equilibrated with Buffer A (2â•›×â•›3 resin volumes). 2. When necessary, isolated mitochondria (100Â€ mg) are incubated in the presence of 600Â€nmol CATR for 15Â€min at 4°C. 3. Membrane proteins are solubilized with 1% (w/v) or 3% (w/v) DDM in Buffer B for 15Â€ min at 4°C supplemented with an antiprotease cocktail as described in (7) in the case of H9AA. During this step the final protein concentration is 10Â€mg/mL. 4. The lysate is centrifuged at 24,000â•›×â•›g for 10Â€min at 4°C and the supernatant is supplemented with Buffer C (0,5 vol for 1 vol of supernatant) and 20Â€mM imidazole. It is loaded onto the equilibrated Ni-NTA resin using a batch procedure and incubated for 1Â€h at 4°C in a rotating apparatus. 5. The flow-through fraction is removed by a 10-min centrifugation at 4,000â•›×â•›g and 4°C. 6. The resin is then resuspended with two volumes of Buffer D and poured into a column. It is thereafter washed with two volumes of the same buffer. The two eluted fractions (four resin volumes) are pooled for further analyses. 7. The resin is washed with five volumes of Buffer E. The eluate is usually collected in two fractions: one corresponds to the first 4.5â•›×â•›resin volumes and the other to the last 0.5â•›×â•›resin volume. 8. The protein is eluted with three resin volumes of Buffer F. Elution of the protein is followed by absorbance measurement at 280Â€nm. 9. Imidazole is removed by an Ultrogel® AcA 202 size exclusion chromatography (four resin volumes for one sample volume). The eluant is Buffer G. Elution of the protein is followed by absorbancy measurement at 280Â€nm. Both H9A and H9AA proteins are obtained highly purified as shown in Fig.Â€1.

3.7. Protein Concentration

1. The protein concentration after Ultrogel® AcA 202 chromatography is far from being sufficient for crystallization trials. The fraction of interest is loaded on Amicon® Centriprep® YM30 units (Millipore). The molecular weight cutoff is 30,000Â€Da. The units are centrifuged at 4°C and 900â•›×â•›g to reduce ten times the fraction volume. 2. Detergent is concentrated at the same time and this may lead to protein denaturation. Thus, detergent is removed using by Bio-Beads® SM2. 3. Bio-Beads are activated as described in (10).

24

Arnou et al.

Fig.Â€ 1. Analyses of purified H9A and H9AA. Ten microliters of the imidazole elution fraction (lanes 1 and 2â•›: H9A; lanes 3 and 4: H9AA) is analyzed by SDS-PAGE (12.5%). The gels are divided in two parts, one is Coomassie blue stained (lanes 1 and 3â•›) and the other is transferred onto a nitrocellulose membrane (lanes 2 and 4â•›) and immunostained with an antibody directed against a peptide corresponding to the last 14 amino acids of ScAnc2p as described in (17).

4. The amount of Bio-Beads necessary to remove 90% of the detergent is calculated considering that 100Â€mg (dry weight) of Bio-Beads after activation can bind 1.8Â€mg of DDM after 3Â€h at 4°C, as determined using [14C]DDM. Therefore, the sample should contain at the end of this step about 0.05% DDM corresponding approximately to the initial DDM concentration. 5. The concentrated protein fraction is incubated with the appropriate Bio-Beads amount for 3Â€ h at 4°C under mild shaking (see Note 2). Bio-Beads are removed by three successive centrifugations (3Â€min, 3,000â•›×â•›g). 6. Steps 4 and 5 are repeated until the Ancp concentration reaches around 10Â€mg/mL. Protein concentration is determined by absorbance measurement at 280Â€ nm. The molar extinction coefficients are calculated from the H9A and H9AA tryptophanyl and tyrosyl residue content (11): 35,870/MÂ€cm for H9A and 71,740/MÂ€cm for H9AA. 7. Concentrated H9A and H9AA can be stored at 4°C but preferably are immediately submitted to crystallization trials or to tag removal process. 3.8. Removal of the 9-Histidine Tag

1. Enzyme units (0–20) of thrombin are added to 1Â€mg of tagged H9AA in 1Â€ml and left over for 20Â€h at 4°C with mild shaking. Under such conditions, the histidine tag is almost completely removed for a thrombin to protein ratio of 20Â€U/mg (Fig.Â€2).

3.9. Purification of bAncp

1. Hydroxyapatite is suspended in an ice-cold Buffer H and washed with the same buffer according to the supplier’s instructions. 2. Mitochondria (50Â€mg of mitochondrial proteins) are unfrozen and incubated in Buffer I (4Â€ ml final volume) with 25Â€ µM CATR for 10–15Â€min at 0°C.

Native Membrane Proteins vs. Yeast Recombinant

25

Fig.Â€2. Polyhistidine tag removal. 0, 5, 10, or 20 enzymatic units of thrombin are added to 1Â€mg of tagged H9AA in 1Â€ml and left over for 20Â€h at 4°C. The samples are analyzed by SDS-PAGE (12.5%), and the protein is revealed by Coomassie blue staining.

3. Membrane proteins are solubilized by addition of 1Â€mL 10% (w/v) LAPAO and stirring of the mixture. After standing on ice for 10Â€ min, the mitochondrial lysate is centrifuged at 20,000â•›×â•›g for 10Â€min. 4. The supernatant is layered on a hydroxyapatite Bio-Gel® column (2.5Â€cm diameter, 25Â€mL settled gel) and the elution is performed using Buffer H and monitored online by UV absorbance at 280Â€nm. 5. The pass-through fraction is collected and concentrated to approximately 10Â€ mL by pressure dialysis on an Amicon YM30 membrane. It is then subjected to a gel-exclusion chromatography on Ultrogel AcA202 resin (BioSepra) to remove small solutes such as Ca2+ ions, phosphate and nucleotides from the bAncp preparation. In addition, this step allows the protein preparation to be placed under appropriate conditions with respect to salt concentration and pH; whereas equivalent to dialysis, it is considerably shorter and in this way is probably less damaging. In most cases, the chromatography is carried out in a column (2.5Â€cm diameter) containing 40Â€ml of settled gel equilibrated in Buffer J, supplemented with either 5 or 100Â€mM NaCl. 6. The protein fraction containing the purified BAncp∙CATR complex is treated with moist activated (see above) Bio-Beads (Bio-Rad, 100Â€mg per mg protein) for 2Â€h at 4°C to remove excess detergent (see Note 2). It is then filtered on a 0.45-mm nitrocellulose filter and concentrated to approximately 10Â€mg/mL on a Centricon YM 30 device (Amicon). In this preparation, the protein is exclusively present as the Anc1p isoform as assessed from in-gel proteolysis combined to mass spectrometry analysis. 3.10. Comments on Use of Native bAncp vs. Recombinant ScAncp 3.10.1. Practical Aspects

Experiments carried out with the bovine ADP/ATP carrier obviously benefit from the abundance and easiness to obtain the amount of biological material required for biochemical investigations. This is first illustrated by the fact that 5–10Â€g of mitochondrial proteins can be routinely extracted within a few hours from one beef heart. This amount corresponds to almost 0.5–1Â€g of bAncp. In addition, purification of bAncp is undoubtedly facilitated by the easiness of mitochondria solubilization with detergents

26

Arnou et al.

and by the high efficiency of adsorption chromatography on hydroxyapatite (12). This allows recovery of pure carrier in a single step with a 90–100% yield. Yeast mitochondria are isolated to a lower yield and operation extends over a longer time if considering 1–2Â€days required for cell growth. Approximately 10–20Â€mg of mitochondrial proteins, corresponding to 0.5–1Â€mg of ScAnc2p depending on the nature of strains, are extracted from 1Â€l of culture. Attempts to isolate ScAnc2p by chromatography on hydroxyapatite led to preparations containing other mitochondrial proteins such as the phosphate carrier and VDAC in various amounts, depending on the detergent nature. Therefore, it was necessary to engineer polyhistidine-tagged forms of ScAnc2p that we purified by IMAC. 3.10.2. The Genetic Approach

Yet, yeasts offer the undeniable advantage of genetic approaches to investigate structure–function relationships of the ADP/ATP carrier essentially because (1) they contain endogenous Ancps; (2) their genetics is well known; and (3) they are well suited for the rapid screening of the functional state of carrier mutants due to their ability to grow under either fermentative or respiratory conditions. Site-directed mutagenesis has been used to locate strategic amino acid residues expected to play a role in the transport mechanism. Mutated forms of Anc2p able to sustain the growth of yeast on nonfermentable carbon sources were characterized with respect to ADP/ATP transport in isolated mitochondria or in proteoliposomes. Other approaches consisted in removing/introducing appropriate residues in ScAncp without impairing its full transport activity for probing ligand-induced conformational changes of the isolated carrier or for assessing the topography of the membrane embedded carrier. This is illustrated, for example by the use of tryptophanyl mutants using fluorometric approaches (13,14) or that of cysteinyl mutants for chemical labeling experiments (reviewed in (2)). Genetic handling of yeasts was used with the purpose of functional and/or structural explorations to engineer chimeras in which ScAncp was linked to itself (3,15), to the phosphate carrier (7) or to cytochrome c (16), and also for the heterologous expression of the human Ancps to understand the role of pathogenic point mutations (6). Yeast will undoubtedly afford the means to isolate an Anc2p mutant stabilized in the BA conformation which so far could not be crystallized from isolated beef Ancp due to the difficulties to handle a stable bAncp∙BA complex. BA refers to bongkrekic acid, which is the other specific Ancp inhibitor. It is recognized that CATR and BA, the binding of which to Ancp is mutually exclusive, stabilize the ADP/ADP carrier in two distinct conformations. This led to the conclusion that Ancp adopts at least two

Native Membrane Proteins vs. Yeast Recombinant

27

different conformations in the membrane, which are probably involved in nucleotide transport. Therefore understanding this process at the molecular level involves deciphering the 3D structure of the ScAnc2p BA complex in comparison with the 3D structure of the Ancp CATR complex.

4. Notes 1. It is recommended to store the frozen mitochondria as 50–100Â€ µL beads. This facilitates the withdrawing of the wanted amount of material prior to each experiment. Beads are made immediately after isolation of mitochondria by dropping the suspension from a pipette into a small volume of liquid nitrogen. They are then handled with a spatula or with forceps. 2. The mixture is gently stirred in a tube rotating horizontally to prevent extensive bead break up. References 1. Pebay-Peyroula E, Dahout-Gonzalez C, Kahn R, Trézéguet V, Lauquin GJ-M, Brandolin G (2003) Structure of mitochondrial ADP/ATP carrier in complex with carboxyatractyloside. Nature 426:39–44 2. Nury H, Dahout-Gonzalez C, Trézéguet V, Lauquin GJ-M, Brandolin G, Pebay-Peyroula E (2006) Relations between structure and function of the mitochondrial ADP/ATP carrier. Annu Rev Biochem 75:713–741 3. Trézéguet V, Le Saux A, David C, Gourdet C, Fiore C, Dianoux A, Brandolin G, Lauquin GJ-M (2000) A covalent tandem dimer of the mitochondrial ADP/ATP carrier is functional inÂ€vivo. Biochim Biophys Acta 1757:81–93 4. Fiore C, Trézéguet V, Roux P, Le Saux A, Noël F, Schwimmer C, Arlot D, Dianoux A-C, Lauquin GJ-M, Brandolin G (2000) Purification of histidine-tagged mitochondrial ADP/ATP carrier: influence of the conformational states of the C-terminal region. Protein Expr Purif 19:57–65 5. Brandolin G, Doussiere J, Gulik A, GulikKrywicki T, Lauquin GJM, Vignais PV (1980) Kinetic, binding and ultrastructural properties of the beef heart adenine nucleotide carrier protein after incorporation into phospholipid vesicles. Biochim Biophys Acta 592:592–614 6. De Marcos Lousa C, Trézéguet V, Dianoux A-C, Brandolin G, Lauquin GJ-M (2002)

7.

8.

9.

10. 11.

12.

The human mitochondrial ADP/ATP carriers: kinetic properties and biogenesis of wild type and mutant proteins in the yeast S. cerevisiae. Biochemistry 41:14412–14420 Postis V, De Marcos Lousa C, Arnou B, Lauquin GJ-M, Trézéguet V (2005) Subunits of the yeast mitochondrial ADP/ATP carrier: cooperation within the dimer. Biochemistry 44:14732–14740 Smith AL (1967) Preparation, properties and conditions for assay of mitochondria, slaughterhouse material, small-scale. Methods Enzymol 10:81–86 Fiore C, Trézéguet V, Le Saux A, Roux P, Schwimmer C, Dianoux A-C, Noël F, Lauquin GJ-M, Brandolin G, Vignais PV (1998) The mitochondrial ADP/ATP carrier: structural, physiological and pathological aspects. Biochimie 80:137–150 Holloway PW (1973) A simple procedure for removal of Triton X-100 from protein samples. Anal Biochem 53:304–308 Pace CN, Vajdos F, Fee L, Grimsley G, Gray T (1995) How to measure and predict the molar absorption coefficient of a protein. Protein Sci 4:2411–2423 Riccio P, Aquila H, Klingenberg M (1975) Purification of the carboxyatractylate binding protein from mitochondria. FEBS Lett 56:133–138

28

Arnou et al.

13. Le Saux A, Roux P, Trézéguet V, Fiore C, Schwimmer C, Dianoux A-C, Vignais PV, Brandolin G, Lauquin GJ-M (1996) Conformational changes of the yeast mitochondrial adenosine diphosphate/adenosine triphosphate carrier studied through its intrinsic fluorescence. 1. Tryptophanyl residues of the carrier can be mutated without impairing protein activity. Biochemistry 35:16116–16124 14. Roux P, Le Saux A, Trézéguet V, Fiore C, Schwimmer C, Dianoux AC, Vignais PV, Lauquin GJ-M, Brandolin G (1996) Conformational changes of the yeast mitochondrial adenosine diphosphate/adenosine triphosphate carrier studied through its intrinsic fluorescence. 2. Assignment of tryptophanyl residues of the carrier to the responses

to specific ligands. Biochemistry 35: 16125–16131 15. Hatanaka T, Hashimoto M, Majima E, Shinohara Y, Terada H (1999) Functional expression of the tandem-repeated homodimer of the mitochondrial ADP/ATP carrier in Saccharomyces cerevisiae. Biochem Biophys Res Commun 262:726–730 16. Dassa EP, Dahout-Gonzalez C, Dianoux AC, Brandolin G (2005) Functional characterization and purification of a Saccharomyces cerevisiae ADP/ATP carrier-iso 1 cytochrome c fusion protein. Protein Expr Purif 40:358–369 17. Marty I, Brandolin G, Gagnon J, Brasseur R, Vignais PV (1992) Topography of the membrane-bound ADP/ATP carrier assessed by enzymatic proteolysis. Biochemistry 31:4058–4065

Chapter 3 Bacterial Overexpressed Membrane Proteins: An Example: The TSPO Jean-Claude Robert and Jean-Jacques Lacapère Abstract The mitochondrial membrane TranSlocator PrOtein (TSPO) is a 18-kDa transmembrane protein involved in various mitochondrial functions, among which the best characterised is cholesterol transport and steroid formation. Determination of its structure would be an important step to understand the mechanism of transport and its regulation. Purification from native membranes is difficult in respect with amounts of homogeneous purified proteins needed for biophysical, structural, and functional studies. Efficient heterologous overexpression in bacterial system, purification on affinity column, and biochemical characterisation has been successfully developed. Large-scale production of detergent-solubilized TSPO has been obtained with fermentation coupled to fast protein liquid chromatography procedure. Small-scale production at lower cost for isotopically labelled recombinant TSPO and/or detergent is also presented. Key words: Peripheral-type benzodiazepine receptor (PBR), Expression vector, E. coli bacteria, Inclusion body, Ni-NTA resin, NMR

1. Introduction Structural studies of membrane proteins require large amounts of purified and concentrated proteins (1). However, most membrane proteins are not naturally abundant and overexpression is one way that has been developed to overcome such difficulty. Bacteria, such as Escherichia coli (E. coli) cells, are the preferred host for recombinant protein expression because they are rather easy to genetically manipulate and expression is fast, typically producing protein in a single day or less than 24Â€h. A large number of vectors with different fusion tags for purification have been developed over the last decades (2). Histidine fusion tag is one of

Jean-Jacques Lacapère (ed.), Membrane Protein Structure Determination: Methods and Protocols, Methods in Molecular Biology, vol. 654, DOI 10.1007/978-1-60761-762-4_3, © Springer Science+Business Media, LLC 2010

29

30

Robert and Lacapère

the most commonly used since it often enables easy purification on NTA-Ni affinity columns. TranSlocator PrOtein (TSPO), previously named peripheraltype benzodiazepines receptor (PBR) is a transmembrane protein mostly located in mitochondria and initially discovered as a class of binding site for benzodiazepine distinct from the GABAa receptors from the central nervous system (3). TSPO expression is too low to permit easy purification from native membranes. cDNA from mouse was cloned, inserted into vectors, and E. coli cells were transformed with these vectors (4, 5). A small fraction of TSPO was found in bacterial membrane, and the major fraction was detected in inclusion bodies. Different fusion tags for purification were tested (4, 5), good results were obtained with six histidines fusion tag added in N-terminal position of the TSPO. Theoretically, the fusion tag placed in C-terminal position permits to remove non-fully expressed proteins, but in the case of TSPO, the C-terminal domain plays an important role in cholesterol binding (6) and transport (5) and thus is affected by the presence of a tag. The present chapter describes production and purification of recombinant TSPO (RecTSPO). The first part is a comparison of RecTSPO production in an incubator with Luria–Bertani (LB) broth or minimum medium (M9) complemented with isotopes for nuclear magnetic resonance (NMR) studies and in a fermentor with LB broth. The second part describes purification on affinity column. We developed two protocols with Ni-NTA resin, large-scale purification with fast protein liquid chromatography (FPLC) procedure and small-scale purification with “manual chromatography”. Advantages and disadvantages are discussed in the light of the target, that is, the structural approach used. Exchange of detergent on Ni-NTA resin is presented, since deuterated detergents are needed for NMR studies, whereas nonionic detergents are requested for mass spectroscopy analysis with MALDITOF. The last part presents protocols for characterisation of detergent-purified RecTSPO, for instance, to find optimal conditions of detergent/protein ratios, which appeared very useful for structural studies and reconstitution of membrane proteins into liposomes for functional studies.

2. Materials 2.1. Expression

1. cDNA from mouse and human TSPO was inserted in pET15b vector and E. coli bacteria strain BL21(DE3) were transformed with the TSPO cDNA containing plasmid (Novagen, VWR, Fontenay sous bois, France).

Bacterial Overexpressed Membrane Proteins: An Example: The TSPO

31

2. Bacteria were cultured either in LB broth medium (Sigma, Saint-Quentin Fallavier, France) or minimum medium M9 (7). Basic M9 medium: 2Â€g KH2PO4, 8Â€g Na2HPO4, 0.5Â€g NaCl, and 0.5Â€g MgCl2∙6H2O in a final volume of 1Â€L and adjusted to pH 7.2. Enriched M9 medium: basic M9 medium supplemented with 10-mL oligoelement stock solution and 1-mL solution A. Oligoelement stock solution: 0.5Â€g EDTA, 83Â€mg FeCl3, 8.4Â€mg ZnCl2, 1.3Â€mg CuCl2 2H2O, 1Â€mg CoCl2 6H2O, and 1Â€mg H3BO3 in a final volume of 100Â€mL. Solution A: 50Â€mg CaCl2, 1Â€mg biotine, 1Â€mg thiamine, and 50Â€mg ampicillin in 1Â€mL water. M9 “plus” H2O or M9 “plus” D2O: Enriched M9 medium supplemented with isotopes (d-Glucose 13C6, d-Glucose 13C6 2 H7, (15NH4)2SO4, D2O, Euriso-Tope, Gif sur Yvette, France). 3. Incubation was performed in 0.5–2.0-L flasks and 1.8–2.0-L Fernbachs (D. Dutscher, Brumath, France) in a swiss minitron INFORS HT incubator (INFORS sarl, Massy, France). 4. Fermentation was performed in a 7.5-L bioreactor BioFlo 110 (New Brunswick Scientific, Paris, France). Temperature, pH, dissolved O2, and stirring were controlled and set at 37°C, 7.0â•›±â•›0.1, 20% and 300–1,200Â€rpm, respectively. 2.2. Purification

1. Bacteria lysis was performed at 4°C in a sonicator (PG 1509) equipped with a 3/16″ tip (MSE, Pocklington, UK). 2. Inclusion bodies from bacteria were collected in 0.5-L bottle with by centrifugation at 5,000â•›×â•›g, 4°C for 15Â€min (Beckman J21, Gagny, France). 3. Elimination of long DNA fragments from solubilized inclusion bodies was performed by addition of benzonase (Novagen, VWR, Fontenay sous bois, France). 4. “Manual” purification was performed on 1.2-mL superflow Ni-NTA resin (Qiagen SA, Courtaboeuf, France) loaded in 12-mL Poly-Prep columns (Bio-Rad, Marne la Coquette, France). 5. Large-scale purification was performed with 1-mL His Trap columns (GE Healthcare SA, Orsay, France) and fast protein liquid chromatography (FPLC) equipment (GE Healthcare SA, Orsay, France). 6. Buffer A: 150Â€mM NaCl and 50Â€mM Hepes-Na pH 7.8. 7. Detergents used during solubilization and purification processes: Sodium dodecyl sulphate (SDS, Sigma, Saint-Quentin Fallavier, France), N-lauroylsarcosine sodium salt (Sarkosyl), dodecylmaltopyranoside (DDM), dodecylphosphocholine

32

Robert and Lacapère

(DPC), 1,2-dihexanoyl-sn-glycero-3-phosphocholine (DHPC C6), and 1,2 diheptanoyl-sn-glycero-3-phosphocholine (DHPC C7) (COGER, Paris, France). Isotopically labelled dodecylphosphocholine D38 (Euriso-Top, Gif sur Yvette, France). 8. Buffer B: Buffer A supplemented with 1% (w/v) SDS. 9. Buffer C: Buffer B supplemented with 5Â€mM imidazole. 10. Buffer D: Buffer B supplemented with 250Â€mM imidazole. 11. Buffer E: Buffer A supplemented with 2.5% (w/v) DPC and 5Â€mM imidazole. 12. Buffer F: Buffer A supplemented with 1.0% (w/v) DPC and 5Â€mM imidazole. 13. Buffer G: Buffer A supplemented with 1.0% (w/v) DPC and 250Â€mM imidazole. 2.3. Characterization

1. Absorption spectra were recorded on UNICAM UV 300 (Fisher Scientific BIOBLOK, Illkirch, France). Extinction coefficient for RecTSPO was calculated using full sequence composition (tag sequence and mRecTSPO or hRcTSPO amino acids) and the ProtParamTools of ExPASy (8). Calculated extinction coefficients are 3.88 and 4.1â•›(mg/mL)−1â•›cm−1 for mRecTSPO and hRcTSPO, respectively. 2. Protein concentration was determined using Bio-Rad protein assay and DC protein assay in the presence of detergents (BioRad, Marne la Coquette, France). Bovin serum albumin (Fraction V, Sigma, Saint-Quentin Fallavier, France) was used as a standard. 3. Protein composition was analysed with SDS-polyacrylamide gel electrophoresis (SDS-PAGE) using Miniprotean II electrophoretic materials and power supply (Bio-Rad, Marne la Coquette, France). Stock solution (Sigma, Saint-Quentin Fallavier, France) of 30% aqueous solution of acrylamide– bisacrylamide (ratio 37.5/1) was used to prepare 12.5% gels. Reducing buffer was prepared as described in the Bio-Rad instruction manual. Protein electrophoresis SDS-PAGE standards were broad range (Bio-Rad, Marne la Coquette, France) and Seablue Plus 2 pre-stained standards (Invitrogen, CergyPontoise, France). Gels were stained with EZBlue staining solution (Sigma, Saint-Quentin Fallavier, France). Gels were dried on gel dryer model 583 (Bio-Rad, Saint-Quentin Fallavier France). 4. Detergent concentration was determined using modified colorimetric assays (Bio-Rad, Marne la Coquette, France) as described in Chapter 1 (this volume).

Bacterial Overexpressed Membrane Proteins: An Example: The TSPO

33

3. Methods 3.1. Transfection of Bacteria with TSPO Containing Plasmid

1. Plasmid containing mTSPO and hTSPO cDNA was sequenced to verify insertion into pET15b vector between Nde1 and BamH1 restriction sites, just following His tag and Thrombin cleavage site structurally constituent of pET15b vector. T7 promotor was used as initiator for sequencing. 2. Competent E. coli bacteria strain BL21(DE3) were transformed by plasmid as previously described (7). 3. Transformed E. coli BL21(DE3) cells were inoculated into 10-mL sterile LB broth (25Â€g/L) supplemented with ampicillin (50Â€mg/L) to select bacteria containing plasmid. 4. Clones were selected on LB/agarose plates containing ampicillin. 5. We observed that the level of RecTSPO expression decreases with time, thus transformation of E. coli bacteria has to be repeated to maintain a high level of protein expression.

3.2. Expression of TSPO

3.2.1. Small-Scale Production in a 200Â€rpm Shaking Incubator

Bacterial cells were grown at 37°C in a medium supplemented with ampicillin in an incubator for small-scale production or in a bioreactor for large-scale production. 1. Pre-culture: Transformed E. coli BL21(DE3) cells were inoculated into 10Â€mL sterile LB broth (25Â€g/L) supplemented with 50Â€µg/â•›mL ampicillin. After 8Â€h incubation, pre-culture was repeated twice by inoculating 0.5Â€ mL of bacteria into 10Â€mL fresh LB supplemented with ampicillin. We observed that these pre-cultures increased RecTSPO productions and particularly in the case of mutants. 2. Culture was initiated by transferring the 10Â€mL pre-culture into 500Â€mL LB supplemented with ampicillin either in a 2-L flask or 1.8-L Fernbach. Incubation was continued until optical density (OD at 600Â€nm) reaching a value of 0.7. At this point, 1Â€ mM Isopropyl-1-Thio Beta-d-galactopyranoside (IPTG) was added to induce recombinant protein production (see Note 1). 3. Bacterial cells were harvested by centrifugation usually after 5Â€h shaking.

3.2.2. Large-Scale Production in a 5-L Bioreactor

1. Three successive pre-cultures were performed. 2. 100Â€mL of transformed E. coli BL21(DE3) cells were transferred into a bioreactor filled with 5Â€L LB supplemented with ampicillin.

34

Robert and Lacapère

3. Growing parameters were controlled by computer acting on bioreactor probes and pumps. Temperature was maintained at 37°C by a mix of heating and freezing by circulating water. Stirring was set at 800Â€rpm. pH was stabilised at 7.1â•›±â•›0.1 by two pumps adding either 1Â€ N NaOH or 1Â€ M phosphoric acid. 4. At 0.7 OD, 1Â€ mM IPTG was added to produce proteins. After 3Â€h , a mixture of 200Â€mL glucose 6.25% and ammonium sulphate 1.5% was infused overnight via a peristaltic pump at 0.25Â€mL per minute of the E. coli BL21(DE3) cells (see Note 2) 5. At the end of the production, when OD at 600Â€nm reached a value of 8.5, bacterial cells were harvested by centrifugation. Clearly it appeared that fermentation with addition of d-Glucose and ammonium sulphate increased production of mTSPO (see TableÂ€1 and Note 3). 3.2.3. Production of Labelled Protein for NMR Studies

For structural studies by NMR, efficient productions of uniformly 15 N, 13C, and 2H-labelled proteins are needed. This is obtained by adding labelled nutriments to bacteria cultured in a minimal medium (M9) (see Note 4). In order to reduce the cost of such production, we used an efficient method for labelling of recombinant proteins previously described (9) with our medium composition (see Note 5).

TableÂ€1 Yields of mRecTSPO and hRecTSPO produced per litre of broth using the different protocols Method

Broth

Expression mRecTSPO hRecTSPO Volume (L) Final OD time (H) (mg/L) (mg/L)

Incubation

LB only

0.5

1.8

5

22

–

Incubation

LB only

0.13

2.1

4

–

17

Fermentation LB only

2.5

2.5

18

82

–

Fermentation LB complemented with glucose and ammonium sulphate

5.3

8.5

18

195

–

Fermentation LB complemented with glucose and ammonium sulphate

2

8.3

18

–

22

Bacterial cells were cultured either by incubation in Fernbach in a thermostated chamber or by fermentation with regulated pH, temperature, and pO2

Bacterial Overexpressed Membrane Proteins: An Example: The TSPO

35

1. Preparation of M9 media (7). M9 minimum medium was prepared and sterilised. Stock solution of oligoelement, solution A and nutriments were prepared with autoclaved H2O and injected through sterile filters (0.22Â€µm) into sterilised M9 medium. 2. Production of isotopic labelled RecTSPO. After three successive pre-cultures of transformed E. coli BL21(DE3) cells in 20Â€mL (each) LB/ampicillin medium, cells were transferred into two Fernbachs (1.8Â€L) and grown at 37°C in 0.5Â€L LB/ampicillin per Fernbach, shaken at 200Â€rpm in the incubator. When OD (at 600Â€nm) reaches a value of 0.7, cells were centrifuged for 20Â€min at 3,300â•›×â•›g and 4°C. Cells in the pellet were washed with 120Â€mL enriched M9 H2O solution and recentrifuged. Cells were resuspended in 250Â€mL of the appropriate medium (enriched M9 H2O supplemented with glucose and ammonium sulphate or M9 “plus” H2O or M9 “plus” D2O) and incubated at 37°C in a small Fernbach (450Â€mL) shaken at 200Â€rpm to allow the recovery of growth and clearance of unlabelled metabolites. Proteins expression was induced after 1Â€ h by addition of 1Â€mM IPTG. After 18Â€h, bacterial cells were harvested by centrifugation. 3.3. Purification of Mouse and Human RecTSPO by Immobilised Metal Ion Affinity Chromatography (IMAC)

3.3.1. Solubilization of Inclusion Bodies

Most of RecTSPO was produced in the inclusion bodies of E. coli BL21(DE3) cells. Generally, recombinant membrane proteins can be removed from inclusion bodies either with denaturing agents and the released proteins are then refolded by gradual removal of the denaturing reagents by dilution or dialysis in the presence of detergents, or inclusion bodies can be directly solubilized in SDS or Sarkosyl detergent (7, 10–12). TSPO is a membrane protein that needs detergent to be maintained solubilized in buffers; thus, SDS was used to solubilize TSPO containing inclusion bodies (5, 13) (see Note 6). 1. The bacterial cells were harvested, at the end of each TSPO’s production, by centrifugation at 5,000â•›×â•›g and 4°C for 15Â€min, the supernatant was discarded into sodium hypochloride and the pellets resuspended in buffer A (without protease inhibitors). Washed bacterial cells were centrifuged (same conditions as above), the supernatant discarded and the pellets kept at −30°C. The content of one pellet depends on the production conditions and corresponds to roughly 50–100Â€mL LB biomass (except for specific production for NMR samples). 2. Bacterial lysates were obtained from one (or several) pellet(s) resuspended by adding 20Â€mL of buffer A, chilled at 0–1°C (for one pellet) and sonicated at 14Â€µm for 1Â€min in a water iced bath.

36

Robert and Lacapère

3. Inclusion bodies were collected by centrifugation (5,000â•›×â•›g, 4°C, 20Â€min). 4. Inclusion bodies were solubilized at room temperature by adding 20Â€mL of buffer B to the pellet. The tube was vigorously vortexed (2Â€ min), agitated on a roller (20Â€ min), and centrifuged at 15,000â•›×â•›g for 1Â€h at 20°C to avoid precipitation of SDS. Supernatant containing solubilized inclusion bodies was reserved at room temperature before purification. In order to fluidify the solution, 2Â€µL of benzonase (25 units per µL) was added 10Â€min before purification. 3.3.2. Large-Scale Purification, Fast Protein Liquid Chromatography (FPLC) Procedure

1. 200Â€ mL of (1%SDS) solubilized inclusion bodies from mRecTSPO production gained by fermentation were injected on 1-mL HisTrap (Ni-NTA) column pre-equilibrated with buffer B. 2. Loading and purification on ÄKTA optimised purification protein system was analysed following changes in OD and protein content of collected fractions (Fig.Â€ 1). The first upward step observed in the chromatogram corresponds to the pass through of unbound proteins. The second upward step corresponds to the saturation of the column (corresponding to the binding capacity of 40Â€mg protein per mL of resin for such column). The third downward step corresponds to the washing of the column. The fourth peak step was induced by the linear imidazole gradient and corresponds to elution of mRecTSPO.

3.3.3. Small-Scale Purification, Manual Chromatographic Procedure

1. Preparation of the column: Empty Poly-Prep column was filled with 1.2Â€mL (named now one column volume, 1Â€V) of Ni-NTA resin. Resin was successively washed (by gravity) with 5Â€V of water, 5Â€V of 0.1Â€M ethylenediaminetetraacetic acid solution adjusted to pH 8.0, 5Â€V of water, 5Â€V of 0.1Â€M nickel sulphate solution, 5Â€V of water, 5Â€V of buffer A, and 5Â€V of buffer B. We noticed that the columns can be prepared “in advance” since the resin did not dry when no effluent flew away. 2. Standard purification: 10–60Â€mL of solubilized proteins were layered on the column with a peristaltic pump (worked at 0.3Â€mL per minute). Column was washed with 4–5Â€V of buffer B and collected fractions were named Wn. Protein was eluted with imidazole containing buffer C and collected fractions were named En. Fraction volume of 1Â€V or 0.25Â€V was collected in Eppendorf tubes of 1.5Â€mL for washing and elution fractions, respectively. 3. Purification with exchange of detergent (see Fig.Â€2): column was loaded as described above, but detergent was exchanged

Bacterial Overexpressed Membrane Proteins: An Example: The TSPO

37 100

3000

OD at 300nm (m AU)

80 70

2000

60 50

1500

40 1000

30 20

500

Imidazole gradient

90 2500

10 0

0 0

2

4

6

8

10

12

14

16

25

35

Fraction Number

116 kDa 97.4 kDa 66 kDa 45 kDa 31 kDa

mRecTSPO

21.5 kDa 14.5 kDa

L fractions

4

W MW FT E 14 17 27 28 29

Fig.Â€1. Purification of mRecTSPO in SDS. FPLC profile (top panelâ•›) and SDS-PAGE of fractions (bottom panel). Bacterial inclusion bodies were solubilized in SDS (10Â€mg/mL SDS), loaded on 1Â€mL His Trap column at a flow rate of 1Â€mL/min. The chromatogram shows OD at 300Â€nm since OD at 280Â€nm is higher and goes rapidly out of scale. It can be divided in four parts (arrows correspond to fractions analysed by SDS-PAGE). Part 1 (fractions 1–6) shows an increase of OD and corresponds to proteins that are not fixed on the column. Part 2 (fractions 4–15) is characterised by a further increase of OD, attributed to the saturation of the column. Part 3 (fraction 16) shows a reduction of the OD and is correlated to the washing of the column (buffer A with 50Â€mg/mL SDS). Part 4 (fractions 16–35) corresponding to the elution step induced by flow of the imidazole gradient (0–250Â€mM). A narrow but concentrated peak of mRecTSPO is observed clearly observed. Volume of fractions is 10, 4 and 2Â€ mL for fractions 1–15, 16 and 17–36, respectively. The silver stained SDS-Page (12.5% polyacrylamide) shows that loaded material (lane 1) contains mRecTSPO (arrow on the leftâ•›) and numerous other proteins. Flow through fractions (4 and 14 in lanes 2 and 3, respectively) show that no mRecTSPO was present in the first part of the load, whereas significant amount is observed in the second part suggesting an over load of the column. Almost no proteins are detected during the washing step (fraction 17, lane 4). Purified mRecTSPO is clearly observed within the elution peak (fractions 27–29 in lanes 5–7, respectively). Lane 8 shows molecular weight (MW) of broad range from Bio-Rad. The major spot in the gel (lane 5 corresponding to fraction 27) contains 200Â€ng pure mRecTSPO. The major peak in the chromatogram (fraction 27) contains 20Â€mg/mL mRecTSPO.

Robert and Lacapère

mRecTSPO ( ) and DPC ( ), mg/mL

38

W1

30

W2

D1

D2

E

25 20 15 10 5 0 0

5 10 Elution volume, mL

15

Fig.Â€2. Purification of mRecTSPO in DPC. Bacterial inclusion bodies solubilized in SDS (10Â€mg/mL) were loaded on manually prepared 1.2Â€mL of Ni-NTA (Qiagen) in 12Â€mL Poly-Prep column. The chromatogram presents three main steps, washing of the column (W1 and W2), detergent exchange on the column (D1 and D2) and elution of protein (E). Volume of fraction during the washing and detergent exchange steps is 1.4Â€ mL, whereas collected fraction during the elution step is 0.3Â€mL. The chromatogram was drawn by recording absorption spectra of each fraction, measuring the OD at 280Â€nm and calculating protein concentration (black diamonds) using extinction coefficient of 3.88â•›(mg/mL)−1â•›cm−1. The washing step is initially done (W1) with flowing through loading buffer (buffer B containing SDS) followed by (W2) low imidazole buffer (buffer C) to remove impurities such as protein weakly attached to the resin. The detergent exchange step is performed on protein attached on the column by flowing through high DPC (D1) containing buffer (buffer E) followed by low DPC (D2) containing buffer (buffer F). mRecTSPO is eluted (E) in the presence of high concentration of imidazole (buffer G). Detergent content (opened square) in the eluted fraction was measured using colorimetric method (see Chapter 1, this volume). Detergent concentration profile shows a peak similar to that observed for mRecTSPO, suggesting that the protein is surrounded by detergent molecules. For instance the maximum protein concentration peak is 14Â€mg/ mL with a DPC concentration of 25Â€mg/mL, giving detergent protein ratio of 1.7 w/w (i.e. 106Â€mol DPC per mol mRecTSPO).

within the column during the washing steps. Resin was washed and eluted successively by 2Â€ V buffer B (W1), 2Â€ V buffer C (W2), 3Â€V buffer E (D1), and 3Â€V buffer F (D2). Protein was eluted with 2Â€V of buffer G (E). 4. Purification of samples for NMR studies: The above described protocol allowed changing nature of detergents and their concentrations during washing step with small volume addition. Deuterated DPC is therefore added just before elution (see Note 7). 3.3.4. General Considerations on IMAC Purification

1. The IMAC instruction manual (from Qiagen) gives some specific considerations on its use, in particular for compatibility of reagents and detergent with Ni-NTA matrices. In our

Bacterial Overexpressed Membrane Proteins: An Example: The TSPO

39

TableÂ€2 Compatible and incompatible detergents for recovery of mRecTSPO sPDC

SDS

DDM

Sarkosyl

DPC

insPDC

C12E8

NOG

Triton X-100

DHPC C6 or C7

CHAPS

Detergents forming soluble and insoluble protein detergent complexes (sPDC and insPDC) were characterised by their ability to exchange or not with SDS in the affinity column loaded with mRecTSPO. Sodium dodecyl sulphate (SDS), dodecylmaltopyranoside (DDM), N-Lauroylsarcosine sodium salt (sarkosyl), dodecylphosphocholine (DPC), octaethylene glycol monododecylether (C12E8), N-octyl b d-glucopyranoside (NOG), Triton X-100, 1,2 dihexanoylsn-glycero-3-phosphocholine (DHPC C6), 1,2 diheptanoyl-sn-glycero-3-phosphocholine (DHPC C7), and 3-[(-cholamidopropyl)dimethylammonio]-1-propane sulphonate (CHAPS) were tested at a concentration of 50Â€mg/mL

hands, no troubleshooting with SDS and sarkosyl up to 5% appeared when used with Ni-NTA resin. 2. The choice of detergent to purify a membrane protein remains an empirical “decision” or a trials and errors procedure. We successfully solubilized mRecTSPO in SDS, exchanged it with DPC and other detergents (see TableÂ€2). However, some other detergents were unable to exchange SDS (see TableÂ€2). Baneres etÂ€ al. (11) tested different detergents to purify G-protein coupled receptor (GPCR) overexpressed in E. coli. Their succeeding protocol involves GPCR removing from inclusion bodies with urea and refolding within the column in the presence of lauryldimethylamine oxide (LDAO) detergent. Bane etÂ€al. (12) attempt to overexpress different human GPCR in E. coli, and NK1 receptor was the only one giving good expression level. Extraction of proteins, mostly localised in inclusion bodies, was only possible with ionic detergent and purification was obtained with DPC. Columbus etÂ€ al. (14) overexpressed several a-helical membrane proteins from Thermotoga maritima in E. coli. Proteins were extracted from membrane with DDM, fixed on affinity column and detergent exchanged. They screened a broad range of detergent with different physical and chemical characteristics, evaluated various biophysical properties of the solubilized proteins, and emphasised on the need to characterise protein–detergent complex in order to pursue structural studies. 3. The recovery yields after purification depends strongly on the quantity of protein immobilised in the column. We observed a linear relationship between amounts of total proteins deposited on the column and RecTSPO recovered. FigureÂ€3 shows a summary of the various purification of mRecTSPO performed in the presence of different detergents using manual chromatography procedure.

40

Robert and Lacapère

Recovered mRecTSPO, mg

16 14 12 10 8 6 4 2 0 0

25

50

75

100

125

150

Layered Proteins, mg

Fig.Â€ 3. Linear relationship between recovered purified mRecTSPO from the 1.2Â€ mL Ni-NTA resin and total proteins deposited on the column. mRecTSPO was isolated in SDS (closed squares), DPC (opened triangles), Sarkosyl (opened circles) or DDM (closed diamonds).

3.4. Characterisation of Purified Membrane Protein 3.4.1. Protein Composition and Oligomeric State

Analysis of samples along the various purification steps (Fig.Â€1) as well as at the end of the purification process often reveals not only the purity of the sample but also the possible oligomeric state of the protein (Fig.Â€ 4). For such purpose, different gel coloration (Coomassie-based and silver stain) and reducing conditions can be used. 1. Prepare a 0.75-mm thick, 12.5% gel following the manufacturer’s instructions based on Laemli method (15). Make the resolving (lower) 12.5% gels the day before of use and let it stand at 4°C to obtain reproducible results. Prepare the staking gel 2Â€h before sample migration. 2. Prepare samples in reductive or non-reductive conditions with or without 650Â€mM mercaptoethanol and in mild reductive conditions with variable concentrations of dithiothreitol (DTT) from 60Â€µM to 180Â€mM. To avoid protein aggregation, samples should not be heated. 3. All wells were filled with identical volumes of the buffer in which samples were diluted. The recommended voltage conditions for optimal resolution with minimal thermal band distortion were 200Â€V at constant voltage setting. Separations ran over less than 50Â€min. 4. Two methods of staining were used, a “single” step Coomassie blue-based one with the EZBlue gel staining reagent and the second one with silver nitrate staining (16). The gels were fixed in 50% ethanol/10% acetic acid for 30Â€min and they were slowly rehydrated by incubation in 5% ethanol/1% acetic acid for 15Â€ min, followed by three times 5Â€ min wash in distilled water.

Bacterial Overexpressed Membrane Proteins: An Example: The TSPO

41

Fig.Â€4. SDS-PAGE profile of hRecTSPO in various reducing conditions. Each lane was loaded with 7.5Â€µL of a solution made of 2.5Â€µL SDS non-reducing buffer and 5Â€µL of SDS solubilized hRecTSPO (4.5Â€µg total amounts) in the presence of variable concentrations of DTT. None (lane 2), 60Â€µM (lane 3), 180Â€µM (lane 4), 600Â€µM (lane 5), 1.8Â€mM (lane 6), 6Â€mM (lane 7), 18Â€mM (lane 8), 60Â€mM (lane 9) and 180Â€mM (lane 10). hRecTSPO is at a final concentration of 30Â€µM and DTT over protein molar ratio ranged thus from 2 to 6,000. Lane 11 contained 650Â€mM mercaptoethanol, concentration used in Laemli reducing conditions. Lanes 1 and 12 correspond to Invitrogen protein standards in non-reductive and reductive conditions, respectively. Gel was made with 12.5% polyacrylamide and stained with Ezblue.

For EZBlue method, staining intensity generally reached a maximum within 45Â€ min to 1-h reagent incubation, after gels were rinsed with distilled water to minimise the background. For silver staining method, fixed and rehydrated gels were incubated in freshly prepared sodium thiosulphate (20%) for 1Â€min and then rinsed in distilled water three times for 30Â€s. Then, gels were stained by 20-min incubation in freshly prepared silver nitrate solution (100Â€ mg silver nitrate and 37.5Â€ µL saturated formaldehyde for 50-mL water solution) and then washed in distilled water two times for 30Â€s. Staining was revealed by incubation in developer solution (3% sodium carbonate, 0.4% sodium thiosulphate and 25-µL saturated formaldehyde for 50-mL water solution) as long as needed (1–2Â€ min). Finally, reaction was stopped in 5% acetic acid for at least 5Â€min and gels were kept in distilled water. 3.4.2. Optimal Detergent Concentration

Detergent can be exchanged during the purification process using manual chromatography procedure (see SubheadingÂ€ 3, Fig.Â€ 2). The concentration needed to get full protein solubilization can be determined experimentally by repeating purification in the presence of various detergent concentrations in the washing and elution buffers. FigureÂ€5 shows the total recovery of mRecTSPO as a function of DPC concentration. At low DPC concentrations, some proteins remained stacked on the Ni-NTA resin and were collected by elution in the presence of high SDS concentration.

Robert and Lacapère

mRecTSPO recovered, mg

42

10 8 6 4 2 0 0

1

2

3

4

5

6

7

8

9

10 11

DPC, mg/ mL

Fig.Â€5. Retrieval of purified mRecTSPO from affinity column as a function of added DPC concentrations during washing steps (D2). Protocol was identical to that presented in Fig.Â€3.2. mRecTSPO eluted in DPC (round black symbols). Resting mRecTSPO stacked in the column is eluted with buffer A supplemented with 50Â€mg/mL SDS (square white symbols). Total amount of mRecTSPO eluted in both DPC and SDS (triangular white symbols) is almost constant and close to 9Â€mg.

We suggested that a minimal concentration of 5Â€ mg/mL DPC was necessary to elute a stable mRecTSPO detergent complex, which did not precipitate. It suggests that when DPC/mRecTSPO ratio was too low, protein tends to precipitate. Using the method described in a previous chapter, we measured the detergent content of each fraction during the purification process (Fig.Â€2). The DPC concentration started at a value of 10Â€mg/mL corresponding to the content of the elution buffer G. Then DPC concentrations increases similarly to the mRecTSPO elution profile. This shows that mRecTSPO forms a complex with DPC and the calculated DPC/mRecTSPO ratio was 1.6 (w/w) for the elution peak. This ratio was higher for the following fractions.

4. Conclusions Optimising bacterial E. coli cultures permit us to overexpress and purify high amounts of both mRecTSPO and hRecTSPO. Both proteins are pure and appear as a single band in SDS-PAGE in reducing conditions. However, in non-reductive conditions mRecTSPO, which has no cystein in its amino acid sequence, still appeared as a 20-kDa single band (Fig.Â€1), whereas polymers are observed for hRecTSPO (Fig.Â€4). This is consistent with the presence of two cysteins in the hTSPO amino acid sequence that could form disulphide bridges inducing the formation of polymers. Increasing the concentration of reducing agent induced

Bacterial Overexpressed Membrane Proteins: An Example: The TSPO

43

progressive polymers disappearance (Fig.Â€4). Furthermore, careful analysis of the gel profile reveals that apparent molecular weight of hRecTSPO monomers increases with DTT concentrations (compare lanes 1–7, in Fig.Â€4). This suggests that cysteins of hRecTSPO can form an intramolecular disulfide bridge. The initial monomer with S–S bond has a more compact structure run as a low molecular weight protein, whereas final monomer without S–S bridge has a more opened structure running as higher molecular weight protein. Along with the change in reduction conditions, the presence of the two forms of the monomers can coexist (see lane 5 in Fig.Â€4). This demonstrates that electrophoresis in reducing, non-reducing and partial reductive conditions could give information on protein conformations.

5. Notes 1. Final optical density (OD at 600Â€nm) reached for cultures in Fernbachs was higher than in flasks, probably because surface contact between LB medium and air was larger in Fernbachs than in flasks. 2. Many media and feeding solutions are described in the literature (17), but we choose to add only essential nutriments (glucose and ammonium sulphate) to LB. 3. For unknown reasons, we observed that amount and concentration of hTSPO were always lower than mTSPO whatever the production process used (see TableÂ€1). We noticed that hTSPO was produced in polymeric forms in the inclusion bodies. 4. Production of labelled protein for NMR studies was performed step by step. A first enriched M9 medium solution was made in H2O with non-labelled ammonium sulphate and glucose. Then, labelled nutriments 15N and 13C were substituted sequentially to the medium in order to follow any changes in bacterial growth and recombinant protein production (see TableÂ€3). Finally, a medium made with inclusion of labelled nutriments in D2O (M9D2O) was used to get fully labelled protein. 5. Marley etÂ€al. (9) described a protocol where cells were first grown in LB medium (since it favours protein expression yields), then concentrated and grown in a smaller volume of minimal M9 medium supplemented with labelled isotopes. They studied the effect of M9 over LB ratio (from 0.125 to 1), and suggested that the best conditions taking into account cost and protein production was the 0.25 ratio that we used in the present work (see TableÂ€ 3). Using our enriched M9 media, the results show that averaged amount of recovered

44

Robert and Lacapère

TableÂ€3 Yields of mRecTSPO produced per litre of broth in different isotopic conditions Isotope

None

15

Ratio M9/LB

â•‡ 0.25

â•‡ 0.25

â•‡ 0.25

0.25

Number of productions

â•‡ 1

â•‡ 3

â•‡ 2

6

17â•›±â•›1

14â•›±â•›1

6.3â•›±â•›0.4

mRecTSPO per litre 14.2 of LB (mg/L)

N

15

N, 13C, deuterated H2O

15

N, 13C, 2H, deuterated H2O

Columns show non-isotopic M9 medium (second column), 15N-enriched M9 medium (third column), 15N, 13C and deuterated H2O-enriched M9 medium (third column), and 15N, 13C, 2H and deuterated H2O-enriched M9 medium (fourth column). 13C, 2H were gained from labelled d-Glucose and 15N from ammonium sulphate

mRecTSPO by LB litre (15Â€mg/L) is close to that obtained in sole LB broth (22Â€mg/L) using a different protocol (see TableÂ€ 1). We do not have explanation for the reduction observed using deuterated d-Glucose. 6. Solubilization of integral membrane proteins (such as TSPO) from inclusion bodies in SDS seems to be less drastic than high dilution processes used for renaturation after urea and guanidine hydrochloride treatments. Such treatment greatly diluted proteins and thus needs further concentration steps often difficult with detergent-solubilized membrane proteins. 7. Purification procedure has to be adapted to needs of functional and structural studies. Batch procedures are often used by molecular biologist and permit to get easily small amounts of diluted purified protein (use of large washing, elution volumes compared to resin volume). Manual chromatography might be “old fashion” but reduces strongly void volumes of tubes, connections and pumps present in FPLC or HPLC apparatus. It permits collection of small concentrated elution volumes. If the advantages of FPLC are numerous (automatisation, control of parameters, use of pre-packed columns), it needs important volumes of mobile phases (from 50 to 500Â€ mL), very expensive when using deuterated detergent and water.

Acknowledgment The authors would like to thank Professor V. Papadopoulos for the generous gift of TSPO plasmids. They would gratefully thank M.A. Ostuni for his help in the critical reading of this chapter. This work was supported by CNRS (Centre National de la Recherche Scientifique) and ANR (Agence National pour la Recherche) Grant 06-Blan-0190-01 to JJL.

Bacterial Overexpressed Membrane Proteins: An Example: The TSPO

45

References 1. Lacapere J-J, Pebay-Peyroula E, Neumann J-M, Etchebest C (2007) Determining membrane protein structures: still a challenge! Trends Biochem Sci 32:259–270 2. Peti W, Page R (2007) Strategies to maximise heterologous protein expression in Escherichia coli with minimal cost. Protein Expr Purif 51:1–10 3. Papadopoulos V, Baraldi M, Guilarte TR, Knudsen TB, Lacapere J-J, Lindemann P, Noremberg MD, Nutt D, Weizman A, Zhang M-R, Gavish M (2006) Translocator protein (18Â€kDa): New nomenclature for the peripheral-type benzodiazepine receptor based on its structure and molecular function. Trends Pharmacol Sci 27:402–409 4. Garnier M, Dimchev AB, Boujrad N, Price JM, Musto NA, Papadopoulos V (1994) In vitro reconstitution of a functional peripheral-type benzodiazepine receptor from mouse Leydig tumor cells. Mol Pharmacol 5:201–211 5. Li H, Papadopoulos V (1998) Peripheric-type benzodiazepine receptor function in cholesterol transport. Identification of a putative cholesterol recognition/interaction aminoacid sequence and consensus pattern. Endocrynology 139:4991–4997 6. Jamin N, Neumann J-M, Ostuni MA, Kim NV, Yao ZX, Murail S, Robert J-C, Giatzakis C, Papadopoulos V, Lacapere J-J (2005) Characterization of the cholesterol recognition amino acid consensus sequence of the peripheral-type benzodiazepine receptor. Mol Endocrinol 19:588–594 7. Sambrook J, Fritsch EF, Maniatis M (1989) Molecular cloning: a laboratory manual, 2nd edn. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY 8. Gasteiger E, Gattiker A, Hoogland C, Ivanyi I, Appel RD, Bairoch A (2003) ExPASy: The proteomics server for in-depth protein knowledge and analysis. Nucleic Acids Res 31:3784–3788

9. Marley J, Lu M, Bracken C (2001) A method for efficient isotopic labeling of recombinant proteins. J Biomol NMR 20:71–75 10. Charbonnier F, Köhler T, Pechère JC, Ducruix A (2001) Overexpression, refolding, and purification of the Histidine-tagged outer membrane efflux protein OprM of Pseudomonas aeruginosa. Protein Expr Purif 23:121–127 11. Baneres J-L, Martin A, Hullot P, Girard J-P, Rossi J-C, Parello J (2003) Structure-based analysis of GPCR function: conformational adaptation of both agonist and receptor upon leukotrienes B4 binding to recombinant BLT1. J Mol Biol 329:801–814 12. Bane SE, Velasquez JE, Robinson AK (2007) Expression and purification of milligram level of inactive G-protein coupled receptors in E. Coli. Protein Expr Purif 52:348–355 13. Lacapere J-J, Delavoie F, Li H, Péranzi G, Maccario J, Papadopulos V, Vidic B (2001) Structural and fonctional study of reconstituted peripheral benzodiazepine receptor. Biochem Biophys Res Commun 284: 536–541 14. Columbus L, Lipfert J, Klock H, Millett I, Doniach S, Lesley SA (2006) Expression, puriÂ� fication, and characterization of Thermotoga maritima membrane proteins for structure determination. Protein Sci 15:1–15 15. Laemli UK (1970) Cleavage of structural proteins during the assembly of the head of acteriophage T4. Nature 227:680–685 16. Yan JX, Wait R, Berkelman T, Harry RA, Westbrook JA, Wheeler CH, Dunn MJ (2000) A modified silver staining protocol for visualisation of proteins compatible with matrixassisted laser desorption/ionizatioin and electrospray ionization-mass spectroscopy. Electrophoresis 21:3666–3672 17. Riesenberg D, Gunthke R (1999) High cell density cultivation of microorganisms. Appl Microbiol Biotechnol 51:422–430

as

Chapter 4 Insect Cell Versus Bacterial Overexpressed Membrane Proteins: An Example, the Human ABCG2 Transporter Alexandre Pozza, José M. Pérez-Victoria, and Attilio Di Pietro Abstract The multidrug resistance phenotype of cancer cells has been often related to overexpression of plasma membrane ATP-binding cassette transporters, which are able to efflux many types of drug by using the energy of ATP hydrolysis. ABCG2 is a half-transporter recently involved. Its purification would help to understand the mechanism of both transport and its inhibition. Biophysical, structural, and functional studies are consuming great amounts of homogeneous purified proteins and require efficient overexpression systems. Heterologous overexpression of human membrane proteins is actually a challenge because these proteins are toxic for the host, and both translation and chaperone systems of the host are not well adapted to the biosynthesis of human proteins. Overexpression of ABCG2 has been assayed in both bacterial and insect cell/baculovirus systems. Although it was highly overexpressed in bacterial system, neither transport nor ATPase activity was found within inverted membrane vesicles. By contrast, insect cells/baculovirus system produces a low amount of protein, a part of which is active. Key words: ABCG2, ATPase activity, Baculovirus/insect cells, Biosynthesis and activity inhibitors, Drug transport, Heterologous membrane protein overexpression, Selected bacteria strain

1. Introduction ABCG2 is a human ABC (“ATP-binding cassette”) half-transporter involved in the multidrug resistance phenotype of cancer cells (1–3). This membrane protein has been heterologously overexpressed in Lactoccocus lactis bacteria (4), Pichia pastoris yeast (5), and both Sf9 and High-Five baculovirus-infected insect cells (6–9) as well as in Xenopus laevis (10). The insect cells/baculovirus system has been used despite significant criticisms. Indeed, the insect cell membranes contain a very low amount of cholesterol, which has been recently shown to modulate ABCG2 activity (11, 12). For instance, purified ABCG2 has been obtained only Jean-Jacques Lacapère (ed.), Membrane Protein Structure Determination: Methods and Protocols, Methods in Molecular Biology, vol. 654, DOI 10.1007/978-1-60761-762-4_4, © Springer Science+Business Media, LLC 2010

47

48

Pozza et al.

from High-Five insect cells (8, 9). Other human ABC transporters, also involved in the multidrug resistance phenotype, have been overexpressed in heterologous systems. P-glycoprotein has been overexpressed in Escherichia coli bacteria (13–15), P. pastoris yeast (16), baculovirus/insect cell system (17), and also in X. laevis (18). Multidrug resistance protein 1 has also been obtained in P. pastoris yeast (19) and baculovirus/insect cells system (20). The heterologous overexpression of human membrane proteins is quite difficult for several reasons. First, the lipid composition of host membranes is different from human cells (21), and specific protein–lipid interactions have a significant impact on the correct insertion, folding, structural integrity, and optimal functionality (22). Second, transcription, translation, and chaperone machinery are not fully adapted to human protein biosynthesis, and the quality control system of integral membrane proteins is less efficient than in human cells; for example, an inactive ABCG2 mutant was addressed to the cell surface membrane in insect Sf9 cells but not in human HEK-293 cells where it was hardly expressed, underglycosyled and localized inside the ER (23). Finally, some motifs, such as ATPase domains and transmembrane domains, may be toxic for the host (24) and induce growth suppression. It is worthwhile mentioning that functional overexpression of some membrane proteins in either bacteria, yeast, or insect cells is not possible as described by Tate etÂ€ al. for the rat serotonin transporter (25) where only mammalian cells could produce a functional protein. In our case, ABCG2 overexpression has been tried in both bacterial and baculovirus/insect cell systems. In E. coli, different combinations of bacteria strains, expression plasmid, and growth conditions were assayed. Under the best conditions, a high amount of protein was produced, but the growth was stopped upon induction, thus limiting the bacteria biomass. To overcome this problem, we isolated a toxicity mutant from BL21 (DE3) as described by Miroux and Walker (26). Although the selected strain grew upon induction and produced ABCG2 as well as parental strain, neither ATPase activity nor transport has been detected within inverted membrane vesicles. This result was contradictory because an ABCG2-mediated transport of ethidium bromide has been detected by fluorescence on intact bacteria. Moreover, bacteria-overexpressed ABCG2 could only be solubilized by SDS. The insect cell/baculovirus system allowed obtaining functional ABCG2 within both whole cells and inverted membrane vesicles. Transport activity was monitored on whole cells by flow cytometry with rhodamine 123 as a substrate and on inverted membrane vesicles by fluorescence with Hoechst33342. Insect cell-overexpressed ABCG2 was partially solubilized by mild detergents, allowing its purification through nickel chromatography.

Insect Cell Versus Bacterial Overexpressed Membrane Proteins

49

Purified ABCG2 exhibited a high vanadate-sensitive ATPase activity and was able to bind a number of substrates and inhibitors with high affinity. However, additional studies indicated that the insect cell/baculovirus system was not ideally adapted since ABCG2 was produced as two forms with different migration in SDS–PAGE; a number of evidences suggested that only the upper form was active. The homogeneity of any heterologously overexpressed protein should be checked prior to biochemical, biophysical, and especially structural studies. The transport activity within either inverted membrane vesicles or whole cells is not sufficient to guarantee that all expressed protein is indeed fully functional.

2. Materials 2.1. Host Strain and Culture Medium

1. Bacterial strains: E. coli strains BL21(DE3), BL21 (DE3) pLysS, Origami B(DE3), and Rosetta(DE3) (Novagen). The C41(DE3) and C43(DE3) strains were kindly given by Dr. J.E. Walker (MRC Cambridge, UK). 2. Bacterial medium: LB medium (10Â€ g tryptone, 5Â€ g yeast extract, 10Â€g NaCl, qsp 1Â€l) or 2TY medium (16Â€g tryptone, 10Â€ g yeast extract, 5Â€ g NaCl, qsp 1Â€ l), both available from Bio101. 3. Insect cell strains: Sf9 and High-Five strains were kindly given by Dr. D. Hulmes from our Institute. 4. Insect cells medium: BacVector (Novagen) medium where 10% FCS were added, and High-Five cells in EXPRESS FIVE SFM supplemented with l-glutamine. Drug and chemical chaperone treatments were performed in 2× Grace medium upon twofold dilution with versole water (Aggettant, Lyon).

2.2. Molecular Biology

1. Expression plasmids: pET21b(+) and pTriEx-4-Neo were purchased from Novagen. The pcDNA3 plasmid containing R482T-ABCG2 cDNA was kindly provided by Dr. D. D. Ross (University of Baltimore, MD). 2. Site-directed mutagenesis: The R482T-ABCG2 cDNA was mutated to obtain R482-ABCG2 cDNA by site-directed mutagenesis using a “Quick-Change Site-directed mutagenesis” kit (Stratagene, La Jolla, CA). The primers used were: 5¢-CTGTTATCTGATTTATTACCCATGAGGATGTTA CCAAGTATTATATTTACC-3¢ and 5 ¢ - G G TA A ATATA ATA C T T G G TA A C AT C C T C AT GGGTAATAAATCAGATAACAG-3¢.

50

Pozza et al.

3. Baculovirus generation: Baculovirus were generated using BacVector Triple Cut-3000 transfection kit according to the manufacturer’s instructions (Novagen, VWR, Fontenay-sousBois, France). PtriEx-4-Neo was used as recombinant transfer plasmid and resuspended in 10Â€ mM Tris–HCl, 0.1Â€ mM EDTA, pH 8. All molecular biology experiments were performed with aerosol-barrier pipette tips to avoid contaminations. 2.3. Biochemistry

1. Chemical compounds: Most chemical were purchased from Sigma-Aldrich (Saint-Quentin Fallavier, France). Detergents: Fos-choline 16, Anatrace, Maumee, USA; Tris(2-carboxyethyl) phosphine (TCEP), Perbio Brebières, France; 3-((3-cholamidopropyl) dimethylammonio]-1-propanesulfonate (CHAPS), Euromedex, Souffelweyersheim, France; n-Dodecyl b dmaltopyranoside, Alexis Biochemicals, Axxora, Villeurbanne, France. Other compounds: Pheophorbide a (Frontier Scientific, Logan, USA), Ko 143 (kindly provided by Dr. A. H. Schinkel, The Netherlands Cancer Institute of Amsterdam, The Netherlands), 6-prenylchrysin, and other flavonoids (synthesized by Dr. A. Boumendjel, University of Grenoble, France), GF120918 kindly given by GlaxoSmithKline (Madrid, Spain), and perfluoro-octanoic acid (PFO) from Fluorochem, Derbyshire, UK. 2. Protein purification and concentration: Ni–NTA agarose (Qiagen, Courtaboeuf, France). Purified protein was concentrated using a centrifugal concentrator with a molecular weight cut-off of 30Â€ kDa (Millipore, UK). Imidazole was removed using desalting column (Econo-Pac 10 DG, Bio-Rad). 3. Bacterial solubilization buffer: 0.5% SDS, 0.1Â€M KPO4, 15% glycerol, 0.1Â€M NaCl, 1Â€mM DTT. 4. Cell solubilization buffer: 50Â€ mM HEPES/NaOH, pH 8, 18Â€ mM CHAPS, 0.5Â€ M NaCl, 10Â€ mM imidazole, 20% glycerol. 5. Baculovirus/insect cell solubilization buffer A: 50Â€mM HEPES, pH 8, 20% glycerol, 0.3Â€M NaCl (with CHAPS or fos-choline 16), 5Â€mM TCEP, 10Â€µl/ml protease inhibitors. 6. Baculovirus/insect cell solubilization buffer B: 2% SDS, 50Â€mM HEPES, pH 8, 5Â€ mM TCEP, and 10Â€ µg/ml protease inhibitors. 7. Washing buffer A: 0.05% SDS, 0.1Â€ M KPO4, 15% glycerol, 0.1Â€M NaCl, 1Â€mM DTT. 8. Washing buffer B: 50Â€ mM HEPES/NaOH, pH 8, 18Â€ mM CHAPS, 0.5Â€M NaCl, 30Â€mM imidazole, and 20% glycerol.

Insect Cell Versus Bacterial Overexpressed Membrane Proteins

51

9. Elution buffer A: 0.05% SDS, 0.1Â€ M KPO4, 15% glycerol, 0.1Â€M NaCl, 1Â€mM DTT, 0.2Â€M imidazole. 10. Elution buffer B: 50Â€ mM HEPES/NaOH, pH 8, 18Â€ mM CHAPS, 0.5Â€ M NaCl, 250Â€ mM imidazole, and 20% glycerol. 11. Dialysis buffer: 0.05% SDS, 0.1Â€M KPO4, 15% glycerol, 0.1Â€M NaCl, 1Â€mM DTT. 12. Fluorescence: Photon Technology International Quanta Master I spectrofluorimeter. Samples were loaded into a Suprasil quartz cuvette with an optical path of 5Â€ mm (Hellma). 13. Circular dichroism: CD6 Jobin-Yvon dichrograph using quartz Suprasil cuvettes (Hellma) with an optical path of 2Â€mm. Protein concentration was determined with a visible/ UV DU640 Beckman spectrophotometer. 14. Dynamic light scattering: Zetaiser Nano S from Malvern Instruments (Worcestershire, UK). Buffer viscosity was measured with an Uddelohde viscosimeter from Schott-Geräte, and the refractometric coefficient with an ABBE refractometer. A Suprasil quartz cuvette with a 5-mm optical path was used to monitor particle size. 15. Cell lysis: SLM AMINCO French press. Insect cells were broken by successive passages through a 25Â€Gâ•›×â•›5/8″ needle. 16. Cell lysis buffer: 1% SDS, 100Â€ mM NaCl, 20% glycerol, 100Â€mM TCEP, 10Â€µl/ml protease inhibitors. 17. SDS–Polyacrylamide gel electrophoresis (SDS–PAGE) and western blotting: Mini-protean III from Bio-Rad. 18. Separating gel: 8% acrylamide, 0.21% bisacrylamide, 0.320Â€M Tris–HCl, pH 8.8, 0.1% SDS, 0.1% ammonium persulfate, 0.1% TEMED. 19. Staking gel: 5% acrylamide, 0.13% bisacrylamide, 0.130Â€ M Tris–HCl, pH 6.8, 0.1% SDS, 0.1% ammonium persulfate, 0.1% TEMED. 20. Running buffer: 25Â€ mM Tris–HCl, 192Â€ mM glycine, 0.1% SDS. 21. Sample buffer: 62.5Â€mM Tris–HCl, pH 6.8, 10% glycerol, 2% SDS, 0.00625% bromophenol blue, and 0.1Â€M b-mercaptoethanol (added extemporaneously). 22. SDS–PAGE migration: Followed using prestained molecular weight markers (broad range, from Bio-Rad). 23. SDS–PAGE gels stain: 0.1% coomassie blue R 250 (w/v), 40% methanol (v/v), 10% acetic acid from 15 to 30Â€ min. The SDS–PAGE gels were destained using 40% methanol, 10% acetic acid.

52

Pozza et al.

24. Electrotransfer membranes: Nitrocellulose (Bio-Rad). 25. Electrotransfer: Bio-Rad Trans-Blot SD electroblotter and a PowerPac 200 generator (for 60Â€min at 90Â€mA). 26. Transfer buffer: 25Â€mM Tris–HCl, pH 7.4, glycine 0.7Â€M. 27. Blocking buffer: 1% nonfat-dry milk (w/v), 0.1% Tween 20 (v/v), 10Â€ mM Tris–Borate, 150Â€ mM NaCl, pH 7.4). The blocking was performed for 1Â€h at room temperature. 28. Primary antibody buffer: Same composition that blocking buffer supplemented with monoclonal antibody anti-ABCG2 BXP21 (Alexis Biochemicals) diluted at 1:250. Primary antibody binding was performed for 1Â€h at room temperature. 29. Secondary antibody: Goat anti-mouse IgG (Hâ•›+â•›L)-alkaline phosphatase (AP) conjugate (Bio-Rad), used at 1:2,300 dilution for 1Â€h at room temperature. 30. Immunorevelation: “AP color Development Kit” from BioRad. Between each step, three washings for 5Â€min were performed with 10Â€ml of blocking buffer without dry milk. 31. Buffer A: 50Â€ mM Tris–HCl, pH 8, 5Â€ mM MgCl2, 1Â€ mM DTT, 1Â€µg/ml DNase/RNase. 32. Buffer B: 20Â€mM Tris–HCl, pH 8, 1.5Â€mM EDTA. 33. Buffer C: 50Â€ mM Tris–HCl, pH 8, 0.3Â€ M sucrose, 1Â€ mM EDTA. 34. Buffer D: 10Â€mM HEPES, pH 8, 10Â€mM NaCl. 35. Buffer E: 10Â€ mM HEPES, pH 8, 50Â€ mM NaCl, 0.3Â€ M mannitol. 36. Transport buffer: 50Â€ mM HEPES, pH 8, 2Â€ mM MgCl2, 8.5Â€mM NaCl, 20Â€µg/ml pyruvate kinase, 4Â€mM phosphoenolpyruvate. 37. Reaction buffer (for ATPase activity measurements): 167Â€mM HEPES, pH 8, 5Â€ mM DTT, 13.3Â€ mM phosphoenol pyruvate, 200Â€ µg/ml pyruvate kinase, 33.3Â€ mM sodium azide, 1.7Â€mM EGTA, 3.3Â€mM ouabaine. 38. Stopping solution: 0.5% ammonium molybdate, 6% SDS, 3% ascorbic acid, and 0.5% HCl. 39. Revelation solution: 2% sodium citrate, 2% sodium arsenate, 2% acetic acid. 40. Circular dichroism detergent buffer: 10Â€ mM NaPO4, pH 8, 2% glycerol, 1Â€ mM DTT, 0.05% detergent (either SDS for ABCG2 R482T from bacteria or dodecylmaltoside for the protein from insect cells). 41. Fluorescence detergent buffer: 50Â€mM HEPES/NaOH, pH 8, 18Â€mM CHAPS, 0.5Â€M NaCl, and 20% glycerol.

Insect Cell Versus Bacterial Overexpressed Membrane Proteins

53

3. Methods 3.1. cDNA Integration in Host Cells and Protein Overexpression

1. E. coli strains were transformed with pET21b(+)/ABCG2 R482T using a thermal-shock procedure. The transformed bacteria were plated onto an LB plate containing 100Â€µg/ml ampicillin and incubated overnight at 37°C (see Note 1).

3.1.1. Bacteria Transformation and ABCG2 Expression

2. A single colony was picked and inoculated into a 500-ml erlen containing 100Â€ ml of LB supplemented with 100Â€ µg/ml ampicillin. The preculture was incubated overnight at 37°C with shaking at 140Â€rpm.

3.1.1.1. Bacterial Transformation and Culture

3. Next morning (see Note 2), the preculture was diluted at OD600Â€ nmâ•›=â•›0.1 into a 5-l flask containing 1Â€ l of prewarmed 2YT, and the culture was performed at 37°C with shaking at 140Â€rpm. 4. When OD600Â€nm reached about 0.6, IPTG was added to a final concentration of 1Â€ mM and bacteria were incubated with shaking. After induction, the growth stopped with BL21(DE3), Origami B(DE3) and Rosetta(DE3) strains, and no expression was observed with BL21(DE3)pLysS. Under these conditions, the biomass was low, and other parameters were therefore tested. First, several induction temperatures (20, 25, 30, and 37°C) and inducer concentrations (0.025, 0.05, 0.1, and 1Â€ mM) were checked but without success. Second, induction was tried at high OD600Â€nm (1.5 to 3): a high biomass was obtained but the growth was stopped. Poor-culture medium, such as LB, has been used but growth arrest was still observed. Growth arrest was also observed with the Rosetta (DE3) strain, excluding codon bias as a possible cause. Since ABCG2 is composed of two toxic motifs known to induce growth suppression upon induction (25), a mutant of the BL21(DE3) strain that was still able to grow after induction has been selected. Using pTriEx in bacteria gave even worse results than pET21, and protein expression was assayed with either N-terminal or C-terminal hexahistidine tag, or without tag.

3.1.1.2. Obtaining Mutant Strains by Host Selection

The mutant strain was selected as described by Miroux and Walker (26) with minor modifications. In our case, bacteria should be taken just before the recovery growth. 1. A single colony was inoculated into 5Â€ ml of 2YT medium supplemented with 100Â€ µg/ml ampicillin. The culture was incubated at 30°C with shaking at 140Â€rpm. 2. When OD600Â€nmâ•›=â•›0.8, the culture was inducted by 1Â€mM IPTG.

54

Pozza et al.

3. Cell growth was monitored by measuring absorbance at 600Â€nm every hour. 4. Upon induction cell growth stopped, and just before the recovery growth (4Â€h in our case) the cells were collected. 5. The cells were immediately diluted with 2YT medium according to serial dilutions between 1:10 and 1:10,000. 6. The different dilutions were then plated onto 2YT agar plates containing 1Â€mM IPTG and 100Â€µg/ml ampicillin. 7. After overnight incubation at 37°C, two cell populations were present: large colonies represented mutants having lost the ability to express ABCG2 but keeping ampicillin resistance, whereas small colonies corresponded to mutants still able to express ABCG2 and ampicillin resistant. These last mutants should be isolated. 8. Protein expression and growth kinetics were monitored on several mutants by western blot. The mutant with highest expression level, and without toxic effects, was selected and named PV6. 9. The selected PV6 mutant strain was cured (see Note 3) of the plasmid using the following procedure. They were maintained in exponential phase in LB medium containing 1Â€mM IPTG, and every day, serial dilutions were plated onto LB agar plates containing either 1Â€ mM IPTG or 100Â€ µg/ml ampicillin or both. The plasmid was lost, when the absence of colony on LB agar plate containing ampicillin was concomitant with a lot of colonies observed on LB agar plate containing IPTG taking the size of the colony into account. This procedure took about 1 month and was checked monitoring the absence of plasmidic DNA. 10. Finally, cured PV6 mutants were transformed with pET21b(+)/ABCG2 R482T, and protein expression and growth kinetics were monitored. This step was very important to check that the mutation is indeed in the bacteria and not in the plasmid. This procedure allowed obtaining a selected strain producing ABCG2 and growing as fast as noninduced selected and parental strains (Fig.Â€1a, b). In contrast to PV6, C41(DE3) and C43(DE3) failed to express ABCG2, showing that the mutant host must be adapted to the protein to be expressed. On the contrary, PV6 was able to produce high level of active BmrA, under similar conditions as with C41. Electron microscopy analysis showed that ABCG2 was not located in intracellular membranes of PV6, as described by Arechaga etÂ€al. (27) with C41(DE3) and C43(DE3) strains (see Note 4).

Insect Cell Versus Bacterial Overexpressed Membrane Proteins

55

Fig.Â€1. ABCG2 overexpression in bacterial and insect cells/baculovirus systems. (a) Time course of ABCG2 expression in the parental and selected PV6 strain. After induction by 1Â€mM IPTG, when OD600Â€nm reached 0.5, the ABCG2 expression was assessed at the times indicated by SDS–PAGE gel, stained with coomassie blue and revealed by western blot with the BXP-21 antibody. (b) Kinetics of growth of parental or selected strain upon induction. When OD600Â€nm reached 0.5, the parental and selected strains were either induced or not, then OD600Â€ nm was measured every hour during 5Â€ h. (■), BL21(DE3)/pET21a(+) ABCG2 R482T; (§), BL21(DE3)/pET21a(+) ABCG2 R482T induced; (▲), PV6/pET21a(+) ABCG2 R482T; ( ), PV6/pET21a(+) ABCG2 R482T induced. (c) Time course of ABCG2 expression in Sf9 and High-Five insect cells. Indicated times represent the number of hours after cell infection. Insect cells were infected with an M.O.I of 20 for 2Â€h and cultured in T25 flasks. 10Â€µg of protein were loaded on each lane. (d) Comparative overexpression of ABCG2 in High-Five using different culture methods. Lane 1, ABCG2 overexpressed in roller bottle; lane 2, ABCG2 overexpressed in spinner; lane 3, ABCG2 overexpressed in flask. The insect cells were infected with an M.O.I of 20 for 2Â€h then cultured in the different culture systems. The roller bottle and spinner cultures were conducted at 20Â€rpm. The insect cells were harvested 72-h postinfection, and 10Â€µg of protein were loaded on each lane.

3.1.1.3. ABCG2 Overexpression Using a Selected Strain

1. A single PV6 colony was inoculated into a 500-ml erlen containing 100Â€ml of LB supplemented with 100Â€µg/ml ampicillin. The preculture was incubated overnight at 37°C with shaking at 140Â€rpm. 2. Next morning, the preculture was diluted at OD600Â€ nmâ•›=â•›0.1 into a 5-l flask containing 1Â€l of prewarmed 2YT and the culture was performed at 30°C with shaking at 140Â€rpm. 3. When OD600Â€nmâ•›=â•›0.8, IPTG was added to a final concentration of 1Â€ mM and the bacteria were incubated under shaking. 4. After 5-h postinduction (2â•›<â•›ODâ•›<â•›2.5), the bacteria were harvested. Functional assays have been tried on intact bacteria by transport measurement of fluorescent compounds, but unfortunately without success because Gram-negative bacteria are

56

Pozza et al.

poorly permeable. The TolC and TolR genes have therefore been inactivated (see Note 5). Mutant bacteria were more permeable and an ethidium bromide transport inhibitable by specific inhibitors was observed, but no ABCG2-dependent multidrug resistance has ever been observed. We tried to select a host mutant by spreading over drug-containing plates but also without success. ABCG2 overexpression using autoinduction method (28) has not been tried, although it constitutes an effective method to overproduce toxic proteins. 3.1.2. Baculovirus Generation and ABCG2 Overexpression

1. Seed 0.5â•›×â•›106 exponentially growing Sf9 insect cells in each well of a 6-well plate, then incubate at 27°C for 30Â€min. Use only the cultures that contain nearly 95% healthy cells.

3.1.2.1. Insect Cell Transfection

2. In a sterile 6-ml polystyrene tube, combine 0.1Â€ µg of the BacVector-3000 Triple Cut Virus DNA, 0.5Â€ µg pTriEx-4Neo/ABCG2, and BacVector Insect cell Medium qsp 25Â€µl. 3. In a separate sterile 6-ml polystyrene tube, add 20Â€µl of nuclease-free water and 5Â€µl of insect juice transfection reagent. 4. Immediately, add the entire DNA/medium mixture into the latter tube, mix gently, and incubate at room temperature for 15Â€min. 5. The DNA/medium/transfection reagent mixture was dilued at 1:10, 1:50, and 1:250 with BacVector medium. 6. The cell monolayer was washed twice, with 2Â€ ml of the BacVector medium. 7. Immediately, 100Â€µl of each dilution was slowly loaded onto the center of the monolayer. 8. Incubate the plates at room temperature for 1Â€h. To keep the cells moist and to increase the contact of the DNA/liposome complex with the cells, the plates should be submitted to gentle rocking every 20–30Â€min. 9. During this period, the agarose overlay was prepared. Sterile BacPlaque agarose at 3% was completely melted using a microwave oven. Next, place the bottle in a 37°C water bath and add two volumes of prewarmed BacVector medium (at 37°C) containing 5% FCS. The agarose was kept at 37°C until use. 10. Two milliliters of the agarose was added into each well. The agarose was slowly poured at the side of the plate. Keep the plates at room temperature for 20Â€min, until the agarose has solidified. 11. One milliliter of BacVector medium containing 5% FCS was then added by gently pipetting the medium onto the middle of the well. 12. The plates were incubated at 27°C for 4–5 days.

Insect Cell Versus Bacterial Overexpressed Membrane Proteins 3.1.2.2. Plaque Identification and Isolation

57

1. The plaques appeared cloudy against a more transparent background. Several well-isolated plaques were picked up using a sterile glass pasteur pipette (see Note 6). 2. The plug was transferred into 1Â€ml of BacVector Medium in the sterile screw-cap vial. The plug was incubated overnight at 4°C to allow virus to diffuse into the medium. 3. The isolated virus were diluted at 1:10, 1:100, and 1:1,000, and 200Â€µl of each dilution was then used to infect cell monolayer as described in step 7 in Subheading “Insect Cell Transfection.” 4. Repeat the steps 8–12 described in Subheading “Insect Cell Transfection.” This procedure should be performed twice to purify each virus.

3.1.2.3. Baculovirus Titration

1. Make several dilutions of purified virus from 1:10 to 1:107 in BacVector medium. 2. Seed 0.5â•›×â•›106 exponentially growing Sf9 insect cells in each well of the 6-well plate, then incubate at 27°C for 30Â€min. Use only the cultures that contains nearly 95% healthy cells. 3. The medium was carefully aspirated, just before the addition of 200Â€µl of each virus dilution to the center of the well. 4. The plates were incubated for 1Â€h at room temperature and submitted to gentle rocking every 20Â€min. 5. During this period, the agarose overlay was prepared. See step 9 in Subheading “Insect Cell Transfection.” 6. Repeat the steps 10–12 in Subheading “Insect Cell Transfection.” 7. The plaques were visualized and counted.

3.1.2.4. Baculovirus Amplification

1. Seed one T25 flask with 2.5â•›×â•›106 exponentially growing Sfâ•›9 cells. Incubate at 27°C for 30Â€min. 2. Remove the supernatant and add 0.3Â€ml of virus suspension to the cell. Then, the flasks were incubated for 1Â€h at room temperature and submitted to gentle rocking every 20–30Â€min. 3. After infection, 5Â€ml of medium was added and the flasks were incubated at 27°C until a cytopathic effect was observed. 4. The supernatant was harvested, constituting the seed-stock virus. 5. Seed one T75 flask with 20â•›×â•›106 exponentially growing Sfâ•›9 cells. Incubate at 27°C for 30Â€min. 6. Remove the supernatant and add 0.5Â€ml of seed-stock virus suspension to the cells. Then, the flasks were incubated for 1Â€h at room temperature, with two gentle rocking. 7. After infection, 10Â€ ml of medium was added and the flasks were incubated at 27°C until a cytopathic effect was observed.

58

Pozza et al.

8. The supernatant was harvested, constituting the master-stock virus. 9. Titrate the master-stock virus: see Subheading “Baculovirus Titration.” 10. Seed one T225 flask with 60â•›×â•›106 exponentially growing Sfâ•›9 cells. Incubate at 27°C for 30Â€min. 11. Remove the supernatant and infect the cells with master-stock virus at a multiplicity of infection equal to 0.1. After incubation for 1Â€ h at room temperature, the flasks were gently rocked, twofolds. 12. After infection, 30Â€ml of medium were added, and the flasks were incubated at 27°C until a cytopathic effect was observed (4–5 days). 13. The supernatant was harvested after centrifugation at 1,000â•›×â•›g for 10Â€min, constituting the high-stock virus. The viral stocks should be stored in the dark at 4°C (see Note 7). 3.1.2.5. ABCG2 Overexpression in Insect Cells

The highest ABCG2 overexpression in flasks is detailed here. Roller bottles and spinners (see Note 8) have also been used, but the expression level was low (Fig.Â€ 1d). Moreover, two insect cell strains have been tried: Sf9 and High-Five. The ABCG2 expression level was greater with the High-Five strain (Fig.Â€1c). 1. Seed on T225 flask with 30â•›×â•›106 exponentially growing Sf9 or High-Five cells. Incubate at 37°C for 24Â€h. 2. Just before infection, the medium was removed. The insect cells were infected at a multiplicity of 20 for 1Â€h at room temperature with high-stock virus. 3. After infection, 30Â€ml of medium was added, and the flasks were incubated at 27°C. 4. The infected insect cells were harvested 72-h postinfection. ABCG2 functionality in intact cells has been checked by flow cytometry with rhodamine 6G as the substrate. ABCG2 expressed in insect cell was indeed able to efflux rhodamine 6G, an activity that could be abolished by specific inhibitors. 1. The bacteria was harvested by centrifugation at 7,500â•›×â•›g for 10Â€min at 4°C, and then washed twice with PBS.

3.2. Preparation of Inverted Membrane Vesicles and Functional Assays

2. Next, the pellets were resuspended in buffer A at a final concentration of 1Â€mg protein/ml.

3 .2.1. Inverted Membrane Vesicles from the Selected Bacteria Strain

3. The suspension was passed twofold through the French press at 18,000Â€ psi, and 10Â€ mM EDTA was immediately added.

Insect Cell Versus Bacterial Overexpressed Membrane Proteins

59

4. Unlysed bacteria and cellular debris were removed by centrifugation at 15,000â•›×â•›g for 30Â€min at 4°C. 5. Next, the supernatant was centrifuged at 100,000â•›×â•›g for 1Â€h. 6. The pellets were washed by buffer B and centrifuged at 100,000â•›×â•›g for 1Â€h at 4°C. 7. The pellets were resuspended in buffer C at a final protein concentration of 20Â€mg protein/ml. 8. The inverted membrane vesicles were homogenized with an Elvehjem potter. 9. Inverted membrane vesicles were quickly frozen and stored in liquid nitrogen. 10. The overexpression level was checked using an SDS–PAGE gel (Fig.Â€3a). ABCG2 expressed in bacteria seemed to be located within inverted membrane vesicles, although we have not checked if the overexpressed protein might form “proteo-lipidic aggregates” as described by Montigny etÂ€al. (29). Indeed, these aggregates were present in the cytosolic fraction and co-sedimented with the membrane fraction. 3.2.2. Inverted Membrane Vesicles from Insect Cells

1. After centrifugation at 500â•›×â•›g for 15Â€ min, the cells were washed twice with PBS without Ca2+ and Mg2+. 2. The cell contents of T225 flasks were suspended into 1Â€ ml of lysis buffer D for 1Â€h at 4°C, next lysed with a syringe by 20 passages through a 25Gâ•›×â•›5/8″ needle. 3. Cellular debris and unlysed cells were pelleted by centrifugation at 500â•›×â•›g for 15Â€min, and discarded. 4. The supernatant was diluted into buffer E, prior to ultracentrifugation at 100,000â•›×â•›g for 1Â€h. 5. The pellets were then suspended into the same buffer including 10% glycerol, by passing through the 25Gâ•›×â•›5/8″ needle. 6. Finally, membrane fractions were quickly frozen and stored in liquid nitrogen. 7. The expression level was checked using western blot (Fig.Â€3d, lane 2).

3.2.3. Functional Assays on Inverted Membrane Vesicles 3.2.3.1. FluorescentSubstrates Transport Measurements

Hoechst33342 is a fluorescent compound, the quantum yield of which is increased in hydrophobic surroundings as plasmatic membranes. When ATP/Mg2+ is added, ABCG2 within inverted membrane vesicles translocates the Hoechst33342 toward the aqueous compartment, and the Hoechst33342 fluorescence decreases. The rate of Hoechst33342 transport must be greater

60

Pozza et al.

than its rate of diffusion to make the transport visible. These conditions are hardly obtained. 1. Four hundred microgram of membrane protein were diluted into transport buffer in either presence or absence of the BCRP-specific inhibitor Ko143. 2. The mixture was incubated for 1Â€min at 37°C under magnetic stirring. 3. Then, 0.1Â€ µM Hoechst33342 was added, and the fluorescence followed until equilibrium. Fluorescence was read at 457Â€nm upon excitation at 355Â€nm, and the bandwidths were set at 4Â€nm. 4. Finally, 5Â€mM ATP was added, and the fluorescence was measured during several minutes (Fig.Â€2a). No Ko143-inhibitable transport activity was detectable on inverted membrane vesicles from bacteria, whereas a transport activity was indeed observed from insect cells. 3.2.3.2. ATPase Activity Measurements

The ATPase assay was based on a colorimetric ascorbic acid/ ammonium molybdate assay for measuring the release of inorganic phosphate as described by Doige etÂ€al. (30) with minor modifications. 1. The ABCG2 effectors were prepared as tenfold concentrated solutions in the presence of 10% DMSO. Inverted membrane vesicles were diluted at 1Â€µg/µl. All solutions were prepared before using and prewarmed at 37°C. 2. The reaction buffer (30Â€ µl), ABCG2 effector (10Â€ µl), and inverted membrane vesicles (10Â€µl) were added together and homogenized. 3. The reaction was started by adding 50Â€ µl of 10Â€ mM ATP/20Â€mM MgCl2. 4. After 40Â€min at 37°C, the reaction was stopped by 100Â€µl of stopping solution. 5. Five minutes later, 100Â€µl of revelation solution was added. 6. After 20Â€ min, absorbance was read at 750Â€ nm. The basal ATPase activity was determined on inverted membrane vesicles from control insect cells expressing b-galactosidase or inactive K86M ABCG2. The amount of released inorganic phosphate was expressed in nanomoles of phosphate per minute (per milligram of total protein). The basal ATPase activity of ABCG2 was modulated by drug substrates or inhibitors, and it depended on a mutation located at position 482 (Fig.Â€2b). No significant ABCG2-mediated ATPase activity was measured on inverted membrane vesicles from bacteria.

Insect Cell Versus Bacterial Overexpressed Membrane Proteins

61

Fig.Â€ 2. Functional assay on inverted membrane vesicles from bacteria, and insect cells infected with baculovirus. (a) Hoechst33342 transport on inverted membrane vesicles from bacteria and insect cells containing ABCG2. 1,000Â€µg of inverted membrane vesicles from bacteria or 400Â€µg of inverted membrane vesicles from insect cells were diluted in 2Â€ml of 50Â€mM HEPES, pH 8, 2Â€mM MgCl2, 8.5Â€mM NaCl, 20Â€µg/ml pyruvate kinase, and 4Â€mM phosphoenolpyruvate. After 1Â€min, Hoechst33342 was added at 0.5Â€µM, then when fluorescence was stable, 5Â€mM ATP was added and the fluorescence was measured during 1,000Â€s. —, inverted membrane vesicles containing ABCG2 from bacteria; –â•›–, inverted membrane vesicles containing ABCG2 from bacteria + 1Â€µMÂ€Ko 143; -â•›-â•›-, inverted membrane vesicles containing ABCG2 from insect cells; …, inverted membrane vesicles containing ABCG2 from insect cells + 1Â€µMÂ€Ko 143. (b) Effects of substrates and inhibitors on the ATPase activity of ABCG2 within the inverted membrane vesicles from insect cells. The ATPase activity of R482 (black columns) or R482T (gray columns) ABCG2 overexpressed in Sf9 was measured with 2.5Â€µg of protein, 5Â€mM ATP, and 20Â€mM MgCl2 at 37°C for 40Â€min.

62

Pozza et al.

3.3. Solubilization and Purification of ABCG2 3.3.1. ABCG2 Solubilization, See SubheadingÂ€3.5.1 3.3.2. Purification of ABCG2 Overexpressed in Selected or BL21 (DE3) Strain

ABCG2 expressed in bacteria was not solubilized by zwiterionic and mild detergents, in contrast to a little fraction of the protein expressed in insect cells. The reasons for such a difference will be discussed below. 1. Inverted membrane vesicles from bacteria were treated with bacterial solubilization buffer at a final protein concentration of 2Â€mg/ml for 60Â€min with gentle shaking at 30°C. 2. The sample was centrifuged at 100,000â•›×â•›g for 1Â€h at 30°C. 3. The supernatant containing solubilized proteins was immediately diluted tenfold (see Note 9) and applied onto a Ni–NTA agarose resin, which was pre-equilibrated with the bacterial solubilization buffer containing only 0.05% SDS for 16Â€h at room temperature with gentle shaking. 4. The Ni–NTA agarose resin (478Â€µl/mg of total protein) with bound protein was washed with two-column volumes of washing buffer A. The protein was eluted with three-column volumes of elution buffer A. 5. Imidazole was removed by two successive dialyses (against 100 volumes of the dialysis buffer). 6. Protein purity was assessed on SDS–PAGE gel (Fig.Â€ 3b). Further studies by electronic microscopy indicate that ABCG2 was oligomerized after purification. Some optimizations, such as changing the detergent to DDM, have been tried, but no improvement has been obtained. Purified ABCG2 appeared as soluble aggregates. Neither ATPase activity nor drug binding has ever been observed, except for 6-prenylchrysin.

3.3.3. Purification of ABCG2 Overexpressed in High-Five Cells

1. Inverted membrane vesicles from High-Five cells infected by a baculovirus vector encoding the R482 or R482T transporter were treated with cell solubilization buffer at a final protein concentration of 2Â€ mg/ml, for 30Â€ min with gente shaking at 4°C. 2. The sample was then centrifuged at 15,000â•›×â•›g for 30Â€min at 4°C. 3. The supernatant containing solubilized proteins was immediately applied onto a Ni–NTA agarose resin, which was preequilibrated in the cells solubilization buffer, for 2Â€h at 4°C with gente shaking. 4. The Ni–NTA agarose resin (80Â€µl/mg of total protein) (see Note 10) with bound protein was washed with ten-column volumes of washing buffer B. 5. The protein was eluted with two-column volumes of elution buffer B.

Insect Cell Versus Bacterial Overexpressed Membrane Proteins

63

Fig.Â€ 3. ABCG2 purification from inverted membrane vesicles. (a) SDS–PAGE gel of inverted membrane vesicles from bacteria. The bacteria were lysed using French press, and 10Â€µg of protein was deposited by lane. (b) Purification of ABCG2 from inverted membrane vesicles of bacteria. ABCG2 was purified here from PV6. (c, d) SDS–PAGE gel and western blot, respectively, with BXP-21 antibody of the different steps of ABCG2 purification from inverted membrane vesicles of High-Five insect cells. Lane 1: prestained molecular weight markers; lane 2â•›: inverted membrane vesicles of High-Five cells infected with the baculovirus vector encoding either R482 or R482T ABCG2; lane 3â•›: supernatant after solubilization; lane 4: supernatant after centrifugation (15,000â•›×â•›g, 30Â€ min); lane 5â•›: detergent-insoluble pellet after centrifugation; lane 6â•›: supernatant after binding for 2Â€h; lane 7: Ni–NTA agarose gel after binding for 2Â€h; lane 8â•›: washing; lane 9â•›: Ni–NTA agarose gel after washing; lane 10â•›: elution; lane 11: Ni–NTA agarose gel after elution; lane 12â•›: after imidazole removal by gel filtration.

6. Imidazole was removed by gel filtration (Econo-Pac 10DG Columns) (see Note 11), and the purified protein was quickly frozen and stored in liquid nitrogen. 7. Protein concentration was determined before freezing with Bio-Rad protein assay using bovine serum albumin as a standard. 8. The absence of oligomerization of purified ABCG2 in detergent solution was monitored by dynamic light scattering, either before freezing or after freezing and thawing. All steps were performed in the presence of 1,4-dithio-DL-threitol (5Â€mM) (see Note 12) and protease inhibitors (10Â€µl cocktail/ml buffer). 9. Each step during the purification was controlled by western blot and SDS–PAGE (Fig.Â€3c, d) (see Note 13). Some detergents have been tried to solubilize and purify ABCG2, but none was satisfactory, except for CHAPS. Although ABCG2

64

Pozza et al.

purified in CHAPS exhibited a high vanadate-sensitive ATPase activity and was able to bind many substrates and inhibitors with high affinity, its stability was limited once purified; it showed a great propensity to form high molecular weight oligomers displaying vanadate-sensitive ATPase activity. An exhaustive study of the protein stability by dynamic light scattering has indicated that, in order to get ABCG2 stable at 0.1Â€mg/ml, the buffer must contain at least 0.3Â€M NaCl, 20% glycerol, and 15Â€ mM CHAPS. The presence of these adjuvants is a great problem for additional steps of purification (gel filtration, ionic chromatography) and functional tests such as reconstitution into liposomes. Indeed, glycerol may induce membrane interdigitation, resulting in loss of protein activity (31). ABCG2 reconstitutions into liposomes have been tried, but without success (see Note 14). 3.4. Functional Assays on Purified ABCG2 3.4.1. Quantification of Secondary Structure

1. The protein concentration was precisely determined by absorbance at 280Â€ nm using the molecular extinction coefficient calculated by the ProtParam tool software from protein sequence. 2. The measurements were performed with the circular dichroism detergent buffer solution and the spectra was scanned from 185–190 to 250Â€ nm, with 0.2-nm step and 1-s data acquisition time, at 25°C under constant nitrogen flushing. The baseline was recorded under the same conditions with the detergent buffer solution. Each spectrum was the average of ten measurements. 3. Spectra were then subtracted from the contribution of detergent buffer solution, smoothed and normalized to 0 at 250Â€nm.

Fig.Â€4. Functional assay on purified ABCG2. (a) and (b) Circular dichroism spectra of purified ABCG2 from bacteria and insect cells, respectively. Measurements were performed under the following conditions: 10Â€mM NaPi, pH 8, 2% glycerol, 1Â€mM DTT, 0.05% detergent (SDS for 0.5Â€µM ABCG2 R482T from bacteria or dodecylmaltoside for 0.24Â€µM ABCG2 R482T from insect cells), and spectra was scanned from 185–190 to 250Â€nm with 0.2-nm step and 1-s data acquisition time at 25°C. (c) Binding of 6-prenylchrysin on purified ABCG2 R482T from bacteria or insect cells. The measurements were performed under the following conditions: 0.5Â€µM ABCG2 R482T, 0.05% SDS, 0.1Â€M KPi, 15% glycerol, 0.1Â€M NaCl, 1Â€mM DTT at 25°C for ABCG2 from bacteria (■), and 0.6Â€µM ABCG2 R482T, 50Â€mM HEPES/NaOH, pH 8, 18Â€mM CHAPS, 0.5Â€M NaCl, 20% glycerol ABCG2 for insect cell-expressed transporter (§). The fluorescence emission spectra were recorded from 310 to 370Â€nm at 25°C upon excitation at 295Â€nm. Curves were fitted with SigmaPlot with an on-site model. For bacteria ABCG2, apparent Kdâ•›=â•›0.96â•›±â•›0.10Â€µM, Bmaxâ•›=â•›86.40â•›±â•›4.0; for insect cell ABCG2 apparent Kdâ•›=â•›0.67â•›±â•›0.010Â€µM, Bmaxâ•›=â•›33.84â•›±â•›1.9. (d) Inhibition by vanadate of the ATPase activity of purified ABCG2 from insect cells. The inhibition of ATPase activity of R482 (squares) and R482T (triangles) transporter was measured in the presence of 5Â€mM MgATP, 0.5Â€µg protein, 50Â€µg azolectin, and a range of orthovanade concentrations from 0.001 to 10Â€mM at 37°C for 40Â€min. (e) Lineweaver–Burk plots of ATP hydrolysis by the purified transporter from insect cells. The ATPase activity of either R482 (squares) or R482T (triangles) purified ABCG2 was measured with 0.5Â€µg protein in the presence of 50Â€µg azolectin at 37°C for 40Â€min. (f) Summary of the apparent KD1 values obtained by drug-induced quenching of ABCG2 intrinsic fluorescence. Measurements were performed under the following conditions: 0.045Â€µg/µl ABCG2 (either R482 or R482T), 50Â€mM HEPES/NaOH, pH 8, 18Â€mM CHAPS, 0.5Â€M NaCl, 20% glycerol. The fluorescence emission spectra were recorded from 310 to 370Â€nm at 25°C upon excitation at 295Â€nm. Experimental data were fitted according to a two-site model, as described by Doppenschmitt etÂ€al. (32).

Insect Cell Versus Bacterial Overexpressed Membrane Proteins

65

4. Finally, the arbitrary CD units were converted to values of mean residue ellipticity (degree/cm/dmol) (Fig.Â€4a, b, see Note 15), and these values were used for prediction of secondary structure content using different deconvolution programs available on Dichrowed site. The K2d deconvolution was used to compare ABCG2 expressed in the two expression systems.

66

Pozza et al.

Bacterial ABCG2 was constituted of 37% alpha helix, 18% beta sheet, and 45% random coil, whereas ABCG2 expressed in insect cell were composed of 64% alpha helix, 11% beta sheet, and 24% random coil. It clearly appears that purified ABCG2 from bacteria displays a greater content of random coil and a lower content of alpha helix in comparison to the same protein from insect cells. 3.4.2. Ligand-Induced Quenching of Tryptophan Intrinsic Fluorescence

The fluorescence spectroscopy is a good technique to monitor ligands binding on a purified protein. Some controls, however, need to be performed before the experiments: the ligands must not be fluorescent and not have a high absorbance around the excitation wavelength. 1. The purified protein was quickly thawed at 37°C. 2. Next, the purified protein was diluted at 0.045Â€µg/µl in the fluorescence detergent buffer. The same buffer was used for both the blank and the inner-filter control with N-acetyltryptophanamide (NATA). The inner-filter control is indeed necessary to correct the fluorescence attenuation produced by ligand absorption. The concentration of NATA was chosen to give a similar area as the emission fluorescence spectrum area of the purified protein. 3. A magnetic stirrer bar was added into the three cuvettes containing Protein, Blank, and NATA, which were left under magnetic stirring for 1Â€h at room temperature. This step was necessary to obtain a stable fluorescence spectrum. 4. Then, the fluorescence was read from 310 to 370Â€nm upon excitation at 295Â€nm four times for each cuvette to take into account any potential photo-induced fluorescence quenching. 5. The effector additions were done alternatively: after each fluorescence reading, the effector was added to the cuvette so that the time interval between all measurements was constant. 6. The blank spectra were subtracted from both the protein and NATA spectra. The area under spectra was used to calculate the percentage of fluorescence in comparison to the fluorescence value without drug addition for both protein and NATA. Then, the corrected fluorescence was obtained by dividing the protein fluorescence percentage by the NATA fluorescence percentage. Finally, the corrected values were used to determine the Kd values using Grafit program or to determine apparent Kd values using SigmaPlot program after curve fitting (Fig.Â€4c). Drug substrates and inhibitors showed biphasic quenching curves that were fitted with a two-site model as described by Doppenschmitt etÂ€al. (32). The apparent Kd values were not modified by the R482T point mutation (Fig.Â€4f), suggesting that the R482T point mutation does not act on binding but on subsequent step to transport.

Insect Cell Versus Bacterial Overexpressed Membrane Proteins 3.4.3. ATPase Assay on Purified ABCG2

67

The measurement was performed as described in Subheading “Fluorescent-Substrates Transport Measurements,” with minor modifications. The inhibitors (sodium azide, EGTA, ouabain) of contaminant ATPase was removed in the reaction buffer. The purified protein was prepared as follows: 1. Purified protein was quickly thawed at 37°C. 2. Immediately, azolectine multilamellar vesicles (see Note 16) were added to give a final lipid-to-protein ratio of 100 (w/w), and the mixture was incubated at 25°C for 30Â€min. 3. Then, the mixture was incubated at 4°C for 30Â€min before using. 4. 0.5Â€ µg of purified protein was used for ATPase assay. The basal amount of phosphate was determined with the lipid– detergent buffer. The amount of released inorganic phosphate was expressed in nmoles of Pi per minute (per milligram of ABCG2). The ATPase activity of purified ABCG2 was modified by the R482T point mutation, which both increased Vmax and decreased Km (ATP/Mg2+) (Fig.Â€ 4e). Moreover, the ATPase activity of purified ABCG2 was inhibited by vanadate but less efficiently for the single-point mutant (Fig.Â€4d).

3.5. Limitations of the Baculovirus/ Insect Cell System to Overexpress ABCG2

Two forms of ABCG2 were clearly observed on western blot revealed with either the specific BCRP monoclonal antibody BXP-21 or the anti-histidine monoclonal antibody (see Note 17). These two forms were characterized by an apparent difference in molecular weight of about 2.8Â€kDa in SDS–PAGE but concomitant biosynthesis. Some experiments have been investigated to understand the origin of such a difference.

3.5.1. Differential Solubilization of the Two ABCG2 Forms by Various Detergents

The compared ability of detergents to solubilize ABCG2 was checked with 3-[(3-cholamidopropyl)-dimethylammonio]-1propanesulfonic acid (CHAPS), PFO, fos-choline 16 (FC-16), and sodium dodecylsulfate (SDS). 1. Membranes were solubilized at 2Â€ mg/ml in Baculovirus/ insect cell solubilization buffer A for 30Â€ min at either 4°C (with CHAPS or fos-choline 16) or room temperature (with SDS or PFO). 2. Then, soluble and insoluble fractions were separated by ultracentrifugation at 100,000â•›×â•›g for 60Â€min, at either 4°C (for CHAPS and fos-choline 16) or 25°C (for SDS and PFO). 3. The insoluble fractions were solubilized baculovirus/insect cell solubilization buffer B and the suspension volume used was the same as the centrifuged one. 4. The solubilization was monitored by western blot as revealed with the specific BCRP monoclonal antibody BXP-21 (Fig.Â€5a).

68

Pozza et al.

Fig.Â€5. Limitations of the baculovirus/insect cell system to overexpress ABCG2. (a) Differential solubilization of the two ABCG2 forms by various detergents. Solubilization was performed at a constant protein concentration of 2Â€mg/ml and variable detergent concentrations, corresponding to the different detergent-to-protein ratio (DPR) values, as indicated, in a final volume of 500Â€µl. After 30Â€min with gentle shaking, soluble (supernatant) and insoluble (pellet) fractions were separated by ultracentrifugation at 10,000â•›×â•›g for 60Â€min. Insoluble fractions were solubilized with a volume equal to the centrifuged volume. After addition of Laemmli sample buffer and heating for 30Â€min at 37°C, 10Â€µl of sample were loaded on 8% polyacrylamide gels. S soluble fraction; I insoluble fraction. Western blots were revealed with BXP-21. (b) Conversion of lower to upper form of ABCG2 during SDS–PAGE in presence of 9Â€M urea. Lane 1, upper form purified at low DPR with fos-choline 16; lanes 2–7, increasing amounts, as indicated, of purified ABCG2 with SDS. The 8% polyacrylamide gel was used 4Â€ h after polymerization, run at 75Â€V, and stained with coomassie blue. (c) Effects of translation and transcription inhibitors on ABCG2 expression in High-Five cells. After 13-h postinfection, inhibitors were added to the medium at the final indicated concentration. The cell contents of T25 flasks were harvested 12Â€h later, washed twice with PBS, and lysed with 500Â€µl of lysis buffer; 10Â€µl of cell total extracts were loaded on 8% polyacrylamide gel, and the western blot was revealed with BXP-21.

Insect Cell Versus Bacterial Overexpressed Membrane Proteins

69

The upper form was partially solubilized by CHAPS and other mild detergents, whereas the lower form was solubilized only by SDS, PFO, and fos-choline 16 at a high detergent-toprotein ratio. The reasons for such differences in solubility were not known. Two hypotheses are then possible, based on either a local difference in membrane lipid composition or, more likely, a protein posttranslational modification or conformational change. Three evidences indicate that the upper form has undergone a posttraductional modification. First, ABCG2 glycosylation was monitored using periodic acid-Schiff specific staining for polyacrylamide gels according to Kapitany’s procedure (33), and no mature glycosylation was experimentally observed on both upper and lower forms although a glycosylation core could not be excluded. Second, the tunicamycin treatment known to both inhibit protein glycosylation and induce protein misfolding favored the lower form. The two last arguments suggest that ABCG2 indeed has a glycosylation core. Finally, the lower form was converted into the upper form during SDS–PAGE in the presence of 9Â€M urea. 3.5.2. Conversion of Lower into Upper Form of ABCG2 During SDS–PAGE in Presence of 9Â€M Urea

Urea at 9Â€ M produced two types of effect on purified ABCG2 during electrophoresis: first, it was able to disturb SDS binding to protein in a protein-dependent way (34); second, it altered ABCG2 structure since the maximal wavelength of tryptophanintrinsic fluorescence emission was slightly red-shifted from 335 to 336Â€nm. Others chaotropic agents are known to be more efficient in altering ABCG2 structure but their use is not compatible with SDS–PAGE. In the presence of 8Â€M guanidinium chloride or 6Â€ M guanidinium thiocyanate, the maximal wavelength of tryptophan-intrinsic fluorescence emission was red-shifted by 2 or 11Â€nm, respectively. 1. The SDS gels, either with or without 9Â€M urea, were used 4Â€h after polymerization. 2. A range of SDS-purified ABCG2 (see Note 18), from 1 to 14Â€µg, was loaded on SDS polyacrylamide gels with or without 9Â€M urea. A control was performed with the upper band (6Â€µg). 3. Electrophoresis was run at 75Â€ V until the 50-kDa marker nearly went out of the gel. 4. Then, the SDS gels were stained with coomassie blue (Fig.Â€5b). The results showed that the lower form was converted into the upper one, indicating that the lower and upper forms might indeed be two conformations of the same protein. However, the difference in conformation seemed to be very subtle.

70

Pozza et al.

Since urea is known to produce limited denaturing effects on hydrophobic cores, the differential effects produced here on ABCG2 conformation might likely concern the cytosolic nucleotide-binding domain. 3.5.3. Effects of Translation and Transcription Inhibitors on ABCG2 Expression in High-Five cells

The addition of inhibitors of either translation or transcription, by decreasing the rate of protein synthesis, might allow a better maturation by the molecular chaperone system. 1. After infection by baculovirus encoding ABCG2, at a multiplicity of 20 for 2Â€h, the virus suspension was removed, and 5Â€ml of Grace medium was immediately added. The cells were incubated at 27°C. 2. After 13-h postinfection (see Note 19), the drugs were added into the medium at various concentrations: cycloheximide 1–20Â€µg/ml, mitoxantrone 2Â€µg/ml, hygromycin D 400Â€µg/ ml, emitine 5Â€µg/ml, anisomycin 2Â€µg/ml, and actinomycin D 0.1–1Â€µg/ml. 3. The cells were harvested 12Â€ h later and washed twice with phosphate-buffered saline. 4. The cells from T25 flasks were lysed with 500Â€µl of cells lysis buffer. 5. Cell lysate was heated at 37°C for 30Â€min. 6. The drug effects on ABCG2 overexpression were monitored by western blot using the specific BCRP monoclonal antibody BXP-21 (Fig.Â€5c). The addition of inhibitors of either translation or transcription, favored the upper form; it also seemed to diminish the higher molecular weight smear, possibly due to SDS-resistant ABCG2 aggregates or oxidized forms. Similar results have been observed when the insect cells were cultured in the presence of chemical chaperones (see Note 20) (e.g., glycerol, mannitol, trehalose). Pharmacological chaperones, such as mitoxantrone, did not improve the expression pattern of ABCG2. Chemical chaperones and inhibitors of either translation or transcription favored the upper form, which could indicate that the formation of the lower form is caused by a too high rate of protein synthesis during insect cell infection, compared with the normal rate of maturation fully controlled by molecular chaperones. The comparison of ABCG2 ATPase activity within inverted membrane vesicles obtained after treatment with or without cycloheximide indicates that the ATPase activity was essentially associated to the upper form and confirms that the upper form is probably active and mature while the lower form appears inactive and immature. These last studies indicate that the insect cells/baculovirus system is not ideally adapted to

Insect Cell Versus Bacterial Overexpressed Membrane Proteins

71

overexpress ABCG2, although part of the protein is active and a significant improvement can be provided by addition of agents reducing the rate of protein synthesis.

4. Notes 1. It is not recommended to use glycerol stock because BL21 strains are able to transform plasmids containing genes encoding toxic proteins. Use fresh colonies after transformation to avoid such a problem. Moreover, ABCG2 overexpression was better with fresh transformed bacteria. 2. The preculture is very critical because the bacteria in stationary phase are able to lose plasmids. It is recommended to check the gene integrity by sequencing for validating this step the first time. 3. Do not use chemical compounds such as acridine orange, which is highly mutagenic. 4. The intracellular membranes have a lipid composition and a lipid-to-protein ratio that differs from that of the cytoplasmic membranes. This difference in density modifies the sedimentation pattern. 5. TolC is a porine that brings together both internal and external membranes. TolR is a protein that contributes to external membrane stability. These mutations were performed by Dr. J.-C. Lazzaroni from UMR5122 CNRS/INSA/UCBL1 at Villeurbanne (France). They are known to increase cell permeability and to sensitize E. coli to many different drugs. 6. Avoid neutral red coloration because this colorant is mutagenic. 7. The medium may be supplemented with 10% CFS, which allows a better conservation of the viral titer. 8. The culture conditions in both spinners and roller bottles might stress the insect cells, which are subdued to shear and gravity stress. 9. ABCG2 expressed in bacteria contains an N-terminal hexahistidine tag located. This chimeric protein has low affinity for Ni–NTA agarose, and a concentration of SDS higher than 0.15% inhibits the protein from binding to Ni–NTA and releases the bound protein during washing. 10. The ratio between Ni–NTA and total protein should be optimized; if it is too low, the bound protein will aggregate on the gel.

72

Pozza et al.

11. Desalting column allows removing other undesirable small molecules, such as divalent cations and residual DNA, that may induce protein aggregation or interfere in further functional tests. Purified ABCG2 can be concentrated with centrifugal concentrator, but about half of the purified protein was lost by unspecific binding to the concentrator surface, in spite of many trials to avoid this problem. Moreover, the ABCG2-specific activity after concentration was drastically decreased. 12. Dithiothreitol could be replaced by TCEP that is more stable and does not reduce Ni2+ ions. 13. The purification of overexpressed His-tagged-protein in insect cells is very difficult for many reasons. Some insect cell proteins are rich in histidine residues, which increase the presence of contaminants. The protein overproduction induces no specific protein–protein interaction, which increased the co-elued contaminants. Two steps are critical to avoid this problem: the protein solubilization and the washing. These phenomena explain the low reproducibility of protein purity among different preparations of inverted membrane vesicles from different ABCG2 productions. 14. Solubilized ABCG2 exhibits a lower vanadate-sensitive ATPase activity than ABCG2 within membrane and is also not subject to drug stimulation. The ABCG2 reconstitution to recover these enzymologic properties failed, probably, because the buffer composition was not adapted to both maintain ABCG2 dispersed and allow a good interaction between lipid/detergent micelles and lipid/detergent/protein complex. The research of the adapted buffer requires the determination of the second osmotic virial coefficient. This work is very time- and material consuming. 15. The CD spectra of ABCG2 purified from bacteria have a great background noise from 185 to 200Â€nm, in spite of the corrections and smoothing, which explains why only K2d deconvolution was used to determine the percentage of secondary structure. 16. Azolectine multilamellar vesicles were prepared using the following procedure: first, azolectine was dissolved at 1Â€mg/ml in chloroform/methanol (90/10, v/v) in the presence of 1% d,l-a-tocopherol into a round-bottomed flask. Then, the solvent was removed by rotary evaporation. The residual traces of solvent were removed with a vacuum pump for 4Â€ h. Azolectine was rehydrated at a final concentration of 20Â€mg/ ml for 1Â€h at 37°C with the following buffer: 50Â€mM HEPES, pH 8, 10Â€mM NaCl. 17. The hexahistidine tag is difficult to reveal by western blot in cell total extracts or after solubilization by different detergents. To overcome this problem, protein extracts were solubilized

Insect Cell Versus Bacterial Overexpressed Membrane Proteins

73

at 1 or 0.5Â€ µg/ml in sample buffer (SDS-to-protein ratio between 40,000 and 20,000) for 30Â€min at 37°C. Then, the sample was run at low voltage (about 75Â€V). 18. Insect cell ABCG2 purified in SDS was used to avoid detergent effects and detergent exchange, which would disturb the electrophoretic migration. ABCG2 was purified using the protocol described in SubheadingÂ€3.3.2 with minor modifications. The buffer is composed of 20% glycerol, 0.3% SDS, 50Â€mM HEPES, pH 8, and 5Â€mM TCEP. Protein purification was achieved at room temperature. 19. The inhibitors either of translation or transcription were added 13-h postinfection to reduce their toxicity. Moreover, ABCG2 synthesis started 13-h postinfection with our baculovirus. When the inhibitors were added either 24-h postinfection or later, then the two forms were produced. 20. When chemical chaperones were used, they were added immediately after infection.

Acknowledgments We warmly thank Drs. A.H. Schinkel (Amsterdam, The Netherlands) for providing Ko143, J.-C. Lazzaroni (University of Lyon 1) for constructing the TolR- and TolC- mutants, A. Boumendjel (Grenoble, France) for synthesizing 6-prenylchrysin and other flavonoid derivatives, O. Coulombel and R. Haser (Lyon, France) for giving access to refractometer and viscosimeter, and to dynamic light scattering apparatus, respectively, and J.-L. Rigaud and D. Levy for advices on membrane protein reconstitution into liposomes and TEM analysis of purified protein by negative staining. CD experiments were performed on the platform “Production et Analyse des Proteines” of the IFR 128 BioSciences Gerland-Lyon Sud. We also thank F. Penin for advice on membrane protein biochemistry, E. Vaganay on PAS staining, R. Monserret on circular dichroïsm, and A. Chaboud and I. Grosjean on cell culture. This work was supported by grants from the Association pour la Recherche sur le Cancer (ARC 3519 and 3942), the Ligue Nationale contre le Cancer (comités du Rhône, Drôme et Savoie), the Région Rhône-Alpes (thématique prioritaire Cancer), French–Spanish PAI-Picasso interministerial agreements (SAF-2003-04200-CO2-01), and the Agence Nationale de la Recherche (ANR-06-BLAN-0420). A.P. is a recipient of doctoral fellowships from the Ligue Nationale contre le Cancer (Comité du Rhône) and the Association pour la Recherche sur le Cancer, J.M.P.-V. of a Marie-Curie postdoctoral fellowship from the European Commission (HMPF-CF-2001-01244).

74

Pozza et al.

References 1. Allikmets R, Schriml LM, Hutchinson A, Romano-Spica V, Dean M (1998) A human placenta-specific ATP-binding cassette gene (ABCP) on chromosome 4q22 that is involved in multidrug resistance. Cancer Res 58:5337–5339 2. Doyle LA, Yang W, Abruzzo LV, Krogmann T, Gao Y, Rishi AK, Ross DD (1998) A multidrug resistance transporter from human MCF-7 breast cancer cells. Proc Natl Acad Sci U S A 95:15665–15670 3. Miyake K, Mickley L, Litman T, Zhan Z, Robey R, Cristensen B, Brangi M, Greenberger L, Dean M, Fojo T, Bates SE (1999) Molecular cloning of cDNAs which are highly overexpressed in mitoxantrone-resistant cells: demonstration of homology to ABC transport genes. Cancer Res 59:8–13 4. Janvilisri T, Venter H, Shahi S, Reuter G, Balakrishnan L, van Veen HW (2003) Sterol transport by the human breast cancer resistance protein (ABCG2) expressed in Lactococcus lactis. J Biol Chem 278:20645–20651 5. Mao Q, Conseil G, Gupta A, Cole SP, Unadkat JD (2004) Functional expression of the human breast cancer resistance protein in Pichia pastoris. Biochem Biophys Res Commun 320:730–737 6. Ozvegy C, Litman T, Szakacs G, Nagy Z, Bates S, Varadi A, Sarkadi B (2001) Functional characterization of the human multidrug transporter, ABCG2, expressed in insect cells. Biochem Biophys Res Commun 285:111–117 7. Ahmed-Belkacem A, Pozza A, MunozMartinez F, Bates SE, Castanys S, Gamarro F, Perez-Victoria JM, Di Pietro A (2005) Flavonoid structure activity studies identify 6-prenylchrysin and tectochrysin as potent and specific inhibitors of breast cancer resistance protein ABCG2. Cancer Res 65:4852–4860 8. Pozza A, Perez-Victoria JM, Sardo A, AhmedBelkacem A, Di Pietro A (2006) Purification of breast cancer resistance protein ABCG2 and role of arginine-482. Cell Mol Life Sci 63:1912–1922 9. McDevitt CA, Collins RF, Conway M, Modok S, Storm J, Kerr ID, Ford RC, Callaghan R (2006) Purification and 3D structural analysis of oligomeric human multidrug transporters ABCG2. StructureÂ€14:1623–1632 10. Nakanishi T, Doyle LA, Hassel B, Wei Y, Bauer KS, Wu S, Pumplin DW, Fang HD, Ross DD (2003) Functional characterisation of human breast cancer resistance protein (BCRP, ABCG2) expressed in the oocytes of Xenopus laevis. Mol Pharmacol 64:1452–1462

11. Telbisz A, Muller M, Ozvegy-Laczka C, Homolya T, Szente L, Varadi A, Sarkadi B (2007) Membrane cholesterol selectively modulates the activity of the human ABCG2 multidrug transporter. Biochim Biophys Acta 1768:2698–2713 12. Pal A, Méhn D, Molnar E, Gedey S, Meszaros P, Nagy T, Glavinas H, Janaky T, von Richter O, Bathori G, Szente L, Krajcsi P (2007) Cholesterol potentiates ABCG2 activity in a heterologous expression system: improved inÂ€vitro model to study function of human ABCG2. J Pharmacol Exp Ther 321:1085–1094 13. Achard-Joris M, Bourdineaud JP (2006) Heterologous expression of bacterial and human multidrug resistance proteins protect Escherichia coli against mercury and zinc contamination. Biometals 19:695–704 14. El-Masry EM, Abou-Donia MB (2003) Reversal of P-glycoprotein expressed in Escherichia coli leaky mutant by ascorbic acid. Life Sci 73:981–991 15. George AM, Davey MW, Mir AA (1996) Functional expression of the human MDR1 gene in Escherichia coli. Arch Biochem Biophys 333:66–74 16. Beaudet L, Urbatsch IL, Gros P (1998) Mutations in the nucleotide-binding sites of P-glycoprotein that affect substrate specificity modulate substrate-induced adenosine triphosphatase activity. Biochemistry 37:9073–9082 17. Sarkadi B, Price EM, Boucher RC, Germann UA, Scarborough G (1992) Expression of the human multidrug resistance cDNA in insect cells generates a high activity drug-stimulated membrane ATPase. J Biol Chem 267:4854–4858 18. Nelson JA, Dutt A, Allen LH, Wright DA (1995) Functional expression of the renal organic cation transporter and P-glycoprotein in Xenopus laevis oocytes. Cancer Chemother Pharmacol 37:187–189 19. Cai J, Daoud R, Georges E, Gros P (2001) Functional expression of multidrug resistance protein 1 in Pichia pastoris. Biochemistry 40:8307–8316 20. Ren XQ, Furukawa T, Chen ZH, Okumura H, Aoki S, Sumizawa T, Tani A, Komatsu M, Mei XD, Akiyama SI (2000) Functional comparison between YCF1 and MRP1 expressed in Sf21 insect cells. Biochem Biophys Res Commun 270:608–615 21. Opekarova M, Tanner W (2003) Specific lipid requirement of membrane proteins – a putative bottleneck in heterologous expression. Biochim Biophys Acta 1610:11–22

Insect Cell Versus Bacterial Overexpressed Membrane Proteins 22. Hunt C (2005) Specific protein–lipid interactions in membrane proteins. Biochem Soc Trans 33:938–942 23. Polgar O, Robey RW, Bates SE (2008) ABCG2: structure, function and role in drug response. Expert Opin Drug Metab Toxicol 4:1–15 24. Inoue S, Sano H, Ohta M (2000) Growth suppression of Escherichia coli by induction of expression of mammalian genes with transmembrane or ATPase domains. Biochem Biophys Res Commun 268:553–561 25. Tate CG, Haase J, Baker C, Boorsma M, Francesca Magnani F, Vallis Y, Williams DC (2003) Comparison of seven different heterologous protein expression systems for the production of the serotonin transporter. Biochim Biophys Acta 1610:141–153 26. Miroux B, Walker E (1996) Over-production of proteins in Escherichia coli: mutant hosts that allow synthesis of some membrane proteins and globular proteins at high levels. J Mol Biol 260:289–298 27. Arechaga I, Miroux B, Karrasch S, Huijbregts R, de Kruijff B, Runswick MJ, Walker JE (2000) Characterisation of new intracellular membranes in Escherichia coli accompanying large scale over-production of the b subunit

28. 29.

30.

31. 32.

33. 34.

75

of F(1)F(0) ATP synthase. FEBS Lett 482: 215–219 Studier FW (2005) Protein production by auto-induction in high-density shaking cultures. Protein Expr Purif 41:207–234 Montigny C, Penin F, Lethias C, Falson P (2004) Overcoming the toxicity of membrane peptide expression in bacteria by upstream insertion of Asp-Pro sequence. Biochim Biophys Acta 1660:53–65 Doige CA, Yu X, Sharom FJ (1992) ATPase activity of partially purified P-glycoprotein from multidrug-resistant Chinese hamster ovary cells. Biochim Biophys Acta 1109:149–160 Lu JZ, Huang F, Chen JW (1999) The behaviors of Ca2+-ATPase embedded in interdigitated bilayer. J Biochem 126:302–306 Doppenschmitt S, Spahn-Langguth H, Regardh CG, Langguth P (1998) Radioligandbinding assay employing P glycoprotein-overexpressing cells: testing drug affinities to the secretory intestinal multidrug transporter. Pharm Res 15:1001–1006 Kapitanay RA, Zebrowski EJ (1973) A high resolution PAS stain for polyacrylamide gel electrophoresis. Anal Biochem 56:361–369 Schägger H (2006) Tricine–SDS-PAGE. Nat Protoc 1:16–23

as

Part II X-Ray Crystallography

as

Chapter 5 Crystallography of Membrane Proteins: From Crystallization to Structure Aurélien Deniaud, Ekaterina Moiseeva, Valentin Gordeliy, and Eva Pebay-Peyroula Abstract Although crystallographic studies of membrane proteins have progressed in the last 5 years, the field still remains challenging with several severe bottlenecks. The chapter focuses on the crystallization and describes two approaches, the classical vapor diffusion method and the more recent use of lipidic phases. General aspects on the crystallization principles as well as more practical details are given. In a more synthetic way, the chapter also addresses how structures are solved by X-ray crystallography, and highlights aspects that are specific to membrane proteins. Key words: Crystallization, Crystallography, Detergent, Diffraction, Lipid

1. Introduction Membrane proteins (MP) are the main functional units of membranes and represent roughly one-third of the proteins encoded in the genome. It has become clear in recent years that the study of membranes at the molecular level is of great importance not only in the deciphering of all cellular processes, but also in the understanding of the alterations leading to abnormal transformed cells and the action of the drugs. Indeed, 70% of drugs target membrane proteins. Although the number of membrane protein structures deposited since 1985, date of the first membrane protein structure (1), increases drastically it does not yet reach the rate achieved for soluble proteins (2). Currently, the RCSB Protein Data Bank contains more than 50,000 structures, among which less than 500 are Jean-Jacques Lacapère (ed.), Membrane Protein Structure Determination: Methods and Protocols, Methods in Molecular Biology, vol. 654, DOI 10.1007/978-1-60761-762-4_5, © Springer Science+Business Media, LLC 2010

79

80

Deniaud et al.

structures of membrane proteins (3). Major efforts undertaken in several laboratories in the last years as well as high-throughput crystallography offer a hope of correcting this imbalance. Nevertheless for the large-scale membrane protein structural biology to realize its full promise, significant challenges must be overcome. Two major bottlenecks await the experimentalist along the process toward the structure, the production of pure, stable and functional protein solubilized in amphiphiles, and the growth of well diffracting crystals. Most of the structural studies need to extract the proteins from the membrane using detergents. The matter is that the natural environment of these proteins is a lipid bilayer, and although detergents shield the hydrophobic surface of the protein from the solvent, they do not fully mimic membranes. Crystallization of a membrane protein should be undertaken only if sufficient biochemical and biophysical characterizations have been made. Solving the structure of a membrane protein by X-ray crystallography can encounter specific difficulties due to the detergent present in the crystals. This step is not defined as a major bottleneck but can prove to be difficult. This chapter describes the general strategy for solving membrane protein structures from the crystallization to the structure. In addition to the “classical” crystallization, it presents also alternate ways exploiting lipidic phases. Most of the crystallography is not specific to MPs, and only the main lines are indicated with references to other textbooks (4). The chapter focuses on aspects that are specific to MPs. All the steps described herein in a linear way, will not necessarily be followed sequentially. MPs very often crystallize with poor diffraction quality. In such cases, the procedure of getting good diffracting crystals and solving the structure will be an iterative process in which one has to cycle between the initial step of preparing the protein in a pure and soluble form for the crystallization, the various crystallization approaches, and the diffraction experiments.

2. Materials 1. Detergents: The chemical purity of detergents that are necessary for the crystallization is more important than for purification. In particular, it has a strong influence on the phase diagram. Most companies have several purity levels; the highest purity level is advisable (for example, high-purity ANAGRADE from ANATRACE®). 2. Lipids: The requirement for the purity of lipids is different when lipids are added as additives to the crystallization and when used for the crystallization in lipidic phases. In the latter case, high purity is needed in order to obtain the appropriate phases.

Crystallography of Membrane Proteins

81

As an example, monoolein (MO) for the cubic phase is purchased from NU-Chek Prep, Elysian, MN. 3. Storage of detergents and lipids: Detergents and lipids have to be stored frozen and dry. When large quantities are bought, they should be aliquoted. Flush away the air in each aliquot with a nitrogen flux before freezing the sample at −20°C. Repeat this each time the sample will be frozen again. When warming lipids or detergent at room temperature, keep the samples sealed in order to avoid water absorption. Store prepared detergent solutions at 4°C in the dark. 4. Proteins: Most membrane proteins are very unstable once they are solubilized in the presence of detergent. Each protein is different. Some protein solutions can be frozen. In general, set up crystallization straight after purification.

3. Methods 3.1. Crystallization by the Vapor Diffusion Method

After extracting the protein from the lipidic membrane and after purification, the membrane protein remains in solution surrounded by detergent molecules. The crystallization can then be achieved by adding a precipitant agent in a similar way as for soluble proteins (5). A series of precipitants and additives have then to be screened in order to identify the appropriate conditions. Although the process is quite similar to soluble proteins, several major differences make it more difficult. First, the solution is intrinsically heterogeneous due to the fact that it will contain protein–detergent complexes (PDC), detergent micelles, and individual detergent molecules (Fig.Â€1). If lipids are copurified with the protein, the solution is even more heterogeneous and the size of PDCs will be difficult to control. Extensive efforts to select the solubilization conditions (choice of detergent, concentration, ligands) are required to enhance the success rate of the crystallization using the vapor diffusion methods (Fig.Â€2). This initial step should not be overlooked and has to be monitored experimentally. Several methods used to characterize the protein–detergent complexes prior to crystallization exist and have to be chosen accordingly. Each membrane protein has a specific behavior also dependent on the detergent used for its purification. The steps listed below are indicative. 1. Start the crystallization experiments with well-defined pure proteins (SDS-gel, mass spectrometry, if possible N-terminal sequencing) solubilized in neutral or zwitter-ionic detergents known to be useable for crystallization with appropriate purity (see Note 1). Perform functional assays when possible (can also be ligand binding) (see Note 2).

82

Deniaud et al.

Fig.Â€1. From the membrane to the crystal. The solubilization and crystallization steps are represented schematically and highlight the role of the detergents.

Fig.Â€2. The vapor diffusion method. From the initial condition, vapor diffuses from the drop in order to equilibrate the precipitant concentrations in the reservoir and in the drop. In the final condition, the protein and precipitant concentrations of the drop are increased.

2. Concentrate the protein to 5–10Â€mg/mL by using molecular mass concentrators adapted to the protein (see Note 3). 3. Characterize the protein–detergent complexes before and after concentration, in size (size-exclusion chromatography, light scattering, analytical ultracentrifugation) and composition (content in detergent: radioactive labeled detergent or ATR-IR, content in lipids: Thin-layer chromatography, phosphate dosage, or mass spectrometry) (see Note 4). 4. Test the stability of the protein solubilized in detergent, i.e., by following its behavior in a size-exclusion chromatography

Crystallography of Membrane Proteins

83

after purification, and one or several days after. Identify and analyze proteolytic fragments as a function of time. Compare the stability of the protein with and without inhibitors or ligands (see Note 5). 5. Set up crystallization with commercial screens or PEG-screens with the vapor diffusion method (see Note 6). When a nanodrop robot is available, make 100 or 200Â€nL drops composed of ½ volume of protein solution and ½ volume of reservoir; the reservoir has a volume of 100Â€ mL. Hanging drops are preferable to sitting drops (easier to observe). 6. From 1 or 2 days to 1 month after their setup, the drops should be observed regularly (see Note 7). 7. Conditions leading to crystalline precipitates, or to phase separation with precipitates have to be repeated and refined (see Note 8). Depending on the number of conditions, the drops will be done with the robot or manually. In the latter case, drops of typically 2Â€ mL (1Â€ mL protein solutionâ•›+â•›1Â€ mL reservoir) are equilibrated against reservoirs of 500Â€mL. The scale-up from 200Â€nL to 2Â€mL modifies the kinetics driving to the equilibrium, and therefore, the crystallization conditions are not always directly transposable from the small volume to the larger one. 8. Explore additives including lipids and detergents (see Note 9). 9. Exchange the detergent to another detergent during purification, and explore crystallization conditions starting with large screens (see Note 10). For this, bind the protein to an affinity column and wash with several column volumes of buffer containing the replacement detergent. 10. Modify the protein by cleaving floppy N- or C-termini (i.e., additional Tags), or increasing the protein surface accessible to protein–protein interactions by crystallizing the protein in complex with an antibody fragment (see Note 11). 11. Test the crystals for diffraction quality as soon as possible. Crystals can be flash-frozen or tested directly from the crystallization setups (see Note 12). 3.2. Practical Approach to lipidic cubic phase (LCP) Crystallization

In order to understand the mechanisms of function of membrane proteins, their structures have to be determined up to high resolution. A formidable obstacle in this way is a lack of well-ordered three-dimensional crystals, which is largely the result of the difficulties in handling membrane proteins. In addition, even if crystals are obtained, they are not necessarily of sufficient quality for X-ray crystallography means. From the low number of successes despite the efforts, “standard” crystallization from detergent solutions may not be an optimal way to crystallize membrane proteins.

84

Deniaud et al.

Fig.Â€3. Schematic representation of in meso crystallization process.

In 1996, a new alternative crystallization method, employing lipidic cubic phases (see Note 13), was developed by Rosenbusch and Landau to improve this situation (6). A fundamental difference between methods of standard crystallization and crystallization in the lipidic cubic phase is that in the latter, the solubilized protein is incorporated back in the native lipid bilayer and the crystallization is induced by addition of dry salt as precipitant agent (Fig.Â€3). Liquid crystalline systems formed by lipids in aqueous media can exist as infinite periodic minimal surfaces, which have a zero mean curvature and a periodicity in all three dimensions characterized by a cubic lattice (7). Luzzati first described the bicontinuous liquid crystal cubic phase (8) and later its geometric structure was supplied by (9). The system consists of two compartments: a continuous curved lipid bilayer forming a three-dimensional well-ordered structure, interwoven with continuous aqueous channels. The phase is very viscous, isotropic, and optically transparent. Bicontinuous cubic phases are ubiquitous in lipid–water mixtures. Cubic membranes are found in cell life (10), and they are used in food industry (11) as well as for drug delivery (12). 3.3. Example of Bacteriorhodopsin (BR)

Practical aspects of crystallization in the lipidic cubic phase look very simple and an example – crystallization of BR (see Note 14 for other examples) – can be described as the following procedure (13): 1. Weight into the PCR tube (200Â€mL) approximately 5Â€mg of dry MO, incubate tubes with MO at 40°C, and spin the lipid down for 10Â€min at 13,000â•›×â•›g at room temperature.

Crystallography of Membrane Proteins

85

2. To begin with the crystallization processes, keep MO at 40°C during an additional 20Â€min to gain the isotropic fluid lipidic phase and then let the lipid phase cool to room temperature. If the MO phase becomes nontransparent at room temperature, repeat the melting procedure. Mix 1Â€ mL of prepared 10Â€ mg/mL BR solution comprising about 1.2 w/w% of n-octyl-b-d-glucopyranoside (OG) with 1Â€ mg of MO. To homogenize lipid/protein mixture as well as to gain the cubic phase, centrifuge the PCR tubes with the sample at 9,200Â€RCF for at least 1Â€h at 22°C (rotating tubes within the rotor every 15Â€min by 90°). Incubate the samples during the day in the dark at 22°C. 3. To induce the crystallization, add a precipitant – a ground powder of KH2PO4 mixed with Na2HPO4 (95/5 w/w) with a final concentration of the salt mixture 1–2.5Â€M (pH 5.6). Repeat homogenizing centrifugation of samples as described in the previous item. Leave the crystallization batch in the dark at 22°C. The first BR microcrystals (10–20Â€mm in diameter) usually appear within 1 weak after induction of crystallization (Fig.Â€ 4) (see Note 15 for a more detailed crystallization mechanism). The described protocol of crystallization is quite close to the original one provided by Rosenbusch and Landau (6). 4. Harvest crystals directly from LCP by mechanical manipulation using microtools during observation with high resolution light microscopy, and flash-freeze a single crystal surrounded by some remaining lipidic phase for cryoprotection. Alternatively, add lipase to the lipidic phase to digest the monoolein at room temperature during several hours or days (14), and fish the crystal from the fluid phase (see Note 16).

Fig.Â€4. A single BR crystal and BR crystallization probe which is placed in a PCR tube.

86

Deniaud et al.

3.4. Perspectives for the Crystallization in Lipidic Phases 3.4.1. Crystallization in the Sponge Phase

The sponge phase (L3-phase) is the liquid analogue of the lipidic cubic phase with the reduced bending rigidity of membranes and without long-range order. The ordered cubic phase structure is perturbed by thermally excited shape fluctuations of membranes. The transformation of the cubic to the sponge phase may be induced by adding a solvent such as dimethyl sulfoxide, propylene glycol, polyethyleneglycol (Mwâ•›»â•›400), 2-methyl-2,4-pentanediol (MPD), or Jeffamine M600 to a lipid/water system (15). The diameter of aqueous pores in the monoolein cubic phase is relatively narrow (ca. 5Â€nm) compared to that of the sponge phase (10–15Â€ nm) (16). The size of the pores of the L3-phase allows one to incorporate membrane proteins with large hydrophilic parts and let them diffuse freely within the plane of the membrane surface (15). A schematic representation of L3-phase is shown in Fig.Â€5. Among the literature regarding membrane proteins crystallization in the sponge phase, there was a report on the appearance of the sponge phase proved by “visual inspection, small-angle X-ray scattering and NMR spectroscopy” (17). Crystals of the reaction center from Rhodobacter sphaeroides were grown in the L3-phase by a conventional hanging-drop experiment, and were harvested directly without the addition of lipase or cryoprotectant, and the structure was refined to 2.2â•›Å resolution. In contrast to the earlier lipidic cubic phase reaction center structure (18), the mobile ubiquinone could be built and refined. In these experiments, the components similar to those of cubic phase crystallization (structural resolution 2.35â•›Å) (18) were used: the MO/membrane protein/detergent/buffer. The only additional component was a small amphiphilic molecule 1,2,3-heptanetriol or Jeffamine M600. In another work (19), crystals of the light harvesting II complex

Fig.Â€5. A schematic representation of the sponge phase.

Crystallography of Membrane Proteins

87

with structural 2.45â•›Å resolution were obtained by the sponge phase approach. In this study, the additives used were KSCN, butanediol, pentaerythritol propoxylate (PPO), t-butanol, Jeffamine, and 2-methyl-2,4-pentanediol (MPD). A 2.0â•›Å structure was available for the light harvesting II complex. It was obtained using vapor diffusion-grown crystals of the detergentsolubilized complex (20). The liquid properties of the sponge phase at room temperature can be used directly in hanging- or sitting-drop vapor-diffusion crystallization by commercially available robots. Recently, a sponge phase sparse matrix crystallization screen consisting of 48 different conditions was designed, and crystals for 8 proteins out of 11 tested were obtained (21). Is the “sponge phase” approach better than crystallization in the cubic phase? Unlike the former method, this one has not led to a breakthrough in structural biology of membrane protein. There was no structure of a new membrane protein or a principal improvement in structural resolution achieved by this method. In addition, the use of small additives to obtain a sponge phase can sometimes be harmful for a membrane protein. We have tested the additives described in (19) for BR crystallization. The results of this experiment presented in TableÂ€1 show that indeed, most of the additives are harmful for BR, and the approach does not allow one to obtain BR crystals of the same quality as in the LCP (Moiseeva, personal communication). Does it mean that the sponge phase approach does not have the same (or higher)

Table 1 Crystallization of BR in swollen lipidic L3 mesophases. Probes with the stable protein are shown in dark boxes. In the other cases, the proteins were denaturated or were considerably stressed

88

Deniaud et al.

power as the LCP method? No, for some of the proteins it works well and a further development of this new method is possible. As in the case of the previous approach, unfortunately, there is lack of information about the behavior of in meso system in the course of crystallization. Again, such information could be obtained by a complimentary use of neutron and X-ray scattering. 3.4.2. Crystallization from Vesicles

As is known, bacteriorhodopsin of Halobacterium salinarum forms a hexagonal two-dimensional lattice within the purple membrane. Another crystallization approach (22, 23) exposed the observation that purple membranes treated with the neutral detergent under certain conditions lead to the creation of spherical protein clusters (~50Â€nm in diameter). Using a standard vapor diffusion method for crystallization from BR vesicles with a high protein/lipid ratio, birefringent hexagonal crystals diffracting X-rays beyond 2.5â•›Å resolution were obtained (22–24). This new crystal belongs to the space group P622 with unit cell dimensions of aâ•›=â•›bâ•›=â•›104.7â•›Å and câ•›=â•›114.1â•›Å. The highest announced structural resolution achieved by this method is 2.3â•›Å. The scheme of such an experiment is illustrated in Fig.Â€6. A number of experiments have been carried out in our laboratories to optimize the

Fig.Â€6. Crystallization from vesicles. (a) Schematic representation of crystallization process from vesicles. (b) The BR crystals grown from vesicles.

Crystallography of Membrane Proteins

89

method (Golubev etÂ€ al., personal communication). It has been shown that one can obtain different types of BR crystals (Fig.Â€6); however, at present, there is no other membrane protein crystallized by this method. Nevertheless, it is not yet clear whether this approach is limited to some specific cases, like BR, or has a more general application. Unfortunately, as in two other cases mentioned above, this system of crystallization has not been sufficiently characterized. 3.4.3. Crystallization from Bicelles

Another approach is the crystallization from bicelles, which was first applied to obtain high quality BR crystals (25, 26). Bicelles are a liquid crystal phase consisting of disc-shaped lipid-rich bilayer particles formed from mixtures of dimyristoylphosphatidylcholine (DMPC) with certain detergents. The most employed detergents usually are either dihexanoylphosphatidylcholine (DHPC) or zwitterionic bile salt derivative, CHAPSO. The bicelles sizes at a 1:3 DMPC/DHPC molar ratio are: the bilayer thickness – 40â•›Å and the diameter – 400â•›Å. The 1ipid:detergent ratios present in the bicellar systems are quite high compared to other micellar systems, which have previously been used as model membranes in biophysical or biochemical studies (27, 28). Moreover, the bicellar systems are unusual in that bicelles can be magnetically oriented to yield oriented solid state NMR spectra of high quality (29). It is interesting to mention that investigation of bicelle formation is of great biological importance as well. It may help to understand better one of the fundamental processes in the cell: formation and transformation of biomembranes. The procedure of crystallization of membrane proteins from bicelles can be described as follows. The first step is preparation of bicelles. Then, solubilized protein is mixed with bicelles. It is believed that at this stage, the proteins are reconstituted into bicelles. These steps are schematically illustrated in Fig.Â€ 7. The last step is crystallization of the protein by a standard vapor diffusion method. BR crystals grown at room temperature are essentially identical to previously obtained at 37°C twinned crystals: space group P21 (2.0â•›Å resolution) with unit cell dimensions of aâ•›=â•›44.7â•›Å, bâ•›=â•›108.7â•›Å, câ•›=â•›55.8â•›Å, ßâ•›=â•›113.6°. The other room-temperature crystals are untwinned and belong to space group C2221 (2.2â•›Å resolution) with the following unit cell dimensions: aâ•›=â•›44.7â•›Å, bâ•›=â•›102.5â•›Å, câ•›=â•›128.2â•›Å. By this method, crystals of the human b2-adrenergic G-protein-coupled receptor were obtained (30). The structure was solved to 3.5/3.7â•›Å resolution, which is considerably lower than what was obtained by protein crystallization in the cubic phase (31). However, taking into account the long standing attempts to crystallize a GPCR, one has to accept that such a limited resolution

90

Deniaud et al.

Fig.Â€7. Crystallization from bicelles. (a) The bicelle components: DMPC, DHPC, CHAPSO. (b) Schematic representation of the bicelles: CHAPSO–DMPC, DHPC–DMPC and DHPC–DMPC bicelle with reconstituted integral membrane protein. (c) The BR crystals grown by this method.

is still a considerable success of the method under discussion. Very recently, the murine voltage dependent anion channel (mVDAC) was also crystallized in lipidic bicelles (32). The 2.3â•›Å resolution structure reveals a high-resolution presentation of membrane protein architecture. Also, the position and structural features of the voltage-sensing N-terminal segment were clearly assigned unlike the recent published NMR structure of human VDAC1 solubilized in detergent micelles (33). 3.4.4. On the Way to a General Method

Almost half of the small amount of known MP structures is determined at relatively low resolutions (with resolution 3â•›Å and lower), and only a small number of high-resolution structures are available (34). In contrast, for the RCSB (3) as a whole, ~4% of the entries are at 3â•›Å resolution or worse. The majority of membrane proteins with determined structures were crystallized in a nonmembrane environment by a standard method of vapor diffusion in the presence of detergents (Fig.Â€ 8) (35). Nevertheless, new approaches – crystallization from lipid bilayers (better to say “from membranes”) – have led to a breakthrough in the field of structural biology of membrane proteins. The important role of native lipidic environment in the stability and function of membrane proteins is known. Recent progress in crystallization, employing the most promising biomimetic environments: bicelle and lipidic cubic phases, has given hope to gain the three-dimensional structure of many other membrane proteins. Also a rational design of the crystallization experiment is not possible without deep knowledge of the used systems. A complimentary approach to the studies of complex fluids used for crystallization is of great importance for the development of more general and efficient methods of membrane protein crystallization. Physics with such powerful experimental methods as neutron and X-ray scattering as well as theoretical tools will play a key role in further

Crystallography of Membrane Proteins

91

Fig.Â€8. Statistics of membrane protein crystal structures determined by X-ray diffraction versus crystallization method used to obtain the crystals. Note: Count is high because methods, as listed, are not mutually exclusive (35).

development of crystallization methods. Success would have a high impact on our understanding of molecular mechanisms of the function of the living matter. 3.5. From Crystals to Diffraction Patterns 3.5.1. Heavy Atom Derivatives

3.5.2. Crystallography

Most of the membrane proteins solved are new structures and cannot be solved by molecular replacement. When expressed in Escherichia coli, the proteins can be labeled with Selenomethionines, and the structure is then solved using anomalous dispersion. Still several membrane proteins will need heavy atom derivatives to be solved. Binding heavy atom derivatives to MPs at specific locations may encounter two problems. First, polar protein surfaces are by nature limited, and therefore polar compounds have a more restricted number of putative binding sites. Second, hydrophobic compounds will be trapped within hydrophobic tails of detergent molecules. The search for heavy atom derivatives can be therefore more tedious than for a soluble protein. As soon as high quality diffraction data as well as phase information are available, solving the structure of a membrane protein is very similar to a soluble protein. It has to be noted that because of the difficulty in MP crystallization, whenever micro-crystals are available, these should be tested for their diffraction if possible on a micro-focus synchrotron radiation beam-line (or at least on a

92

Deniaud et al.

beam-line that allows to collimate the beam down to the size of the crystal). In some cases, structures can be solved from microcrystals. Several developments go along this direction (36). Finally, once the protein model is built, extra densities corresponding to lipids or detergents are detected in the electron density maps. They can be modeled from 2Fo-Fc or Fo-Fc electron density maps. Refining their positions and temperature factors will depend on the resolution.

4. Notes 1. Most of the crystallizations achieved so far imply only a limited number of detergent families (37): alkyl-dimethylamineoxides (CnDAO), with nâ•›~â•›12, alkyl-glucosides (CnG) with nâ•›~â•›8, alkyl-maltosides (CnM) with nâ•›~â•›12 or alkyl-oligoethylene glycol-monoethers (CnEm) i.e., C12E8, with small variations on the length of the alkyl chains. Detergents with short chains are not well suitable for membrane protein stability. However, the use of longer detergents can lead to steric hindrance during the crystal formation. It is admitted that ionic detergents would prevent protein–protein interactions by repelling the PDCs, therefore only nonionic detergents are appropriate for crystallization. The best detergent for the solubilization is not necessarily the best for crystallization and vice-versa. In that case, the detergent has to be exchanged during or after the purification procedure, during affinity chromatography, size exclusion chromatography, or by dialysis. However, depending on the critical micellar concentrations, the time necessary for the dialysis might be incompatible with protein stability. 2. The detergent has to maintain the protein in solution in a functional state or more precisely, in a conformation that is relevant to the function. Functional assays are not always possible in PDCs, in particular for transport processes. If ligands are known (substrates or analogs, cofactors, inhibitors), then binding activities can be assayed; they will provide partial indications on the “quality” of the protein. The success of this initial step also depends upon the knowledge of the physicalchemistry of the detergent used, for review (38). 3. The choice and the concentration of the detergent during the purification steps prior crystallization is well discussed in (39). During protein concentration, the detergent is usually concentrated. Removing excess detergent can be essential (40). 4. Several biophysical methods will help to characterize the folding. Thermal fluorescence scans adapted to membrane

Crystallography of Membrane Proteins

93

protein will provide information on the stability (41). Circular dichroism will provide information on the secondary structure. Infrared spectroscopy using ATR-IR will show the secondary structure as well as the quantity of detergent (42). All these methods can be particularly efficient for relative measurements when comparing the properties of PDCs with different detergents. Same measurements can also be performed in the presence of lipids or any other ligand to identify conditions under which the protein has a proper folding and is more stable. It might also be useful to identify and quantify endogenous lipids bound to the protein. The size, the shape of the PDCs, their content in protein and the oligomerization state of the protein, amounts of detergent and lipid molecules and their identification, the stability of the PDCs, have to be characterized as best as possible and can be studied by several biochemical and biophysical methods described in the chapters of (43). Size exclusion chromatography coupled to UV absorption, refractive index, static and dynamic light scattering is well adapted to such measurements (44) and can be complemented by analytical ultracentrifugation (45). 5. The dynamics of proteins in PDCs might be different from the native membrane, and the proteins will therefore be able to adopt a large variety of conformations (not necessarily relevant for their functions), making precise molecular contacts for crystal growth difficult. Stabilizing one single conformation in the solution will therefore enhance the success rate in crystal growth. This was achieved by the mutation of a single residue for LacY (46), or by co-crystallization with an inhibitor for the ADP/ATP carrier (47). The detergent concentration can also be critical. 6. Several precipitants will favor the phase separation and protein precipitation. If enough protein is available, all the commercial screens can be tested using a nanodrop robot. Even with 100Â€nL drops, 80Â€mL are sufficient to screen more than 600 conditions. However, with membrane proteins, it is not expected to have protein crystals from the first screens. They will only indicate conditions where phase separations and interesting precipitates occur, which have then to be refined. Alternately, instead of screening blindly any available conditions, screens can be restricted to fewer precipitants favorable to many membrane proteins such as polyethylene glycols (PEGs) in the presence of various salts and other additives. For this strategy, grid screens with PEGs of different molecular masses (i.e., PEG400, PEG4000, PEG8000) or similar molecules (Jeffamine or PEGMME) and various salts at 0.1 or 0.2Â€M can be set up for different pH values. FigureÂ€9 shows a few examples of initial hits obtained with AcrB.

94

Deniaud et al.

Fig.Â€9. Initial hits obtained for AcrB in the presence of different PEGs. Small crystals are identified in (a) and (c). Higher PEG concentrations lead to phase separations as seen in (b) and (d).

7. A large part of the crystallization experiments consists in observing the crystallization setups. In contrast to many soluble proteins, even with a large number of conditions, protein crystals should not be expected from initial screens. The difficulty is then to recognize crystalline precipitates or small angular objects as possible early events in the crystallization process. Large amounts of detergent combined with salts will lead to detergent crystals. Dyes coloring protein crystals don’t easily diffuse in membrane protein crystals which therefore remain uncolored as salt crystals do. Observing crystals under UV light is a more efficient way of detecting the molecular nature of small promising objects. 8. FiguresÂ€ 10a, b show two typical phase diagrams of nonionic detergents. They highlight the existence of the consolution boundary, which separates the micellar phase from a domain where two immiscible phases coexist. One is detergent enriched, the other is detergent depleted. Micelle–micelle interactions will increase while approaching the consolution boundary from the micellar phase. In the presence of proteins, the same effect will occur and increasing PDC interactions will thus favor protein–protein interactions. When reaching the phase separation,

Crystallography of Membrane Proteins

95

Fig.Â€10. Typical nonionic detergent phase diagrams. The detergent concentration versus temperature is shown in (a) and (b) for two different nonionic detergents. The diagrams highlight different phases separated by the critical micellar concentration (CMC) and the consolution boundary. (c) represents the detergent concentration as a function of the precipitant showing the influence of the precipitant on the CMC and the consolution boundary.

the protein will concentrate in detergent-enriched phase as illustrated with colored proteins in Fig.Â€ 11b. Very often, crystallization is obtained across or near the consolution boundary (Fig.Â€11c–e). This boundary will also be influenced by the precipitant itself (Fig.Â€10c). The phase diagram of the protein versus the precipitant (Fig.Â€ 12) indicates that the

96

Deniaud et al.

Fig.Â€11. A bacterial photosynthetic complex crystallized in the presence of PEG. (a)–(c) show the effect of three PEG concentrations. At 8% (a), the protein solution remains clear. At 15% (b), a phase separation occurs, and the protein is concentrated in one of the phase (as seen by the dark droplets containing the colored protein). At 10% (c), the conditions are close to the consolution boundary and crystals have appeared. (d) is a close-up of a condition similar to (b), small crystals have formed using all the protein in their surroundings. (e) is a close-up view of (c).

Fig.Â€12. Phase diagram of protein concentration versus precipitant. The diagram is the standard diagram valid for all types of proteins. For membrane proteins, this diagram superimposes with the detergent diagrams.

Crystallography of Membrane Proteins

97

protein will go from a soluble forward to a metastable zone above the solubility curve, in which increasing protein–protein interactions might lead to the nucleation of a crystal (5). Closer to the solubility curve, nucleation will be weaker and the growth of crystals will be favored. The protein phase diagram has to be combined with the phase diagram of the detergent. The general strategy for the crystallization setups is to search for precipitants that induce phase separations with crystalline precipitates in one of the phases. In a second step, the precipitant should be reduced to approach the consolution boundary without crossing it. This can be a smoother way and produces often fewer and larger crystals (Fig.Â€11c, e). It has to be noticed that phase diagrams of a detergent might be perturbed by various parameters such as its chemical purity, or the ratio of different isomers. Crystallization will thus be influenced by these parameters (48). 9. When refining the conditions, additives have to be considered. In particular, some additives are known to modify the micelle sizes and therefore will influence the size of the detergent belt in the PDC. This can be important in the crystal packing. Indeed, heptane-triol, known to shrink the radius of detergent micelles, was successfully used as an additive to improve the diffraction quality of the reaction center crystals probably by adapting the size of the detergent belt to the cavities in the crystal left over by protein–protein contacts (49). With the same idea, detergents can also be screened (48, 50). Detergents that are successful as additives might then be tested as main detergent in the PDC. The temperature will drastically influence the phase diagrams of the protein and of the detergent and is therefore an important parameter to explore. 10. Because of the high degree of flexibility of detergent molecules surrounding the protein in the PDCs, crystal contacts have to be made through protein–protein interactions to ensure crystals of high diffraction quality. These contacts are in most cases restricted to extramembrane domains of the protein, and the number of possible contacts is thus much lower than for a soluble protein. The organization of the detergent in membrane protein crystals was highlighted by low-resolution neutron diffraction for the reaction center (49) and for the OmpF porin (51). These studies have shown that the bulk volume occupied by the detergent around the proteins has to fit perfectly in the cavities of the crystal left over by protein–protein contacts. To improve the crystal contacts, the size and shapes of PDCs can be modified.

98

Deniaud et al.

11. Crystallizing the protein in complex with an antibody fragment increases the size of the extramembrane domain of the protein and favors the crystal contacts. The method was first illustrated for a cytochrome c oxidase, which was crystallized with Fv fragments (52). Fab fragments were also used as shown for the KcsA K+ channel (53). Several other examples were reviewed by Hunte etÂ€al. (54). More recently, a similar approach based on ankyrin repeat proteins (DARPins) was proposed (55). Modifications on the protein itself can also improve the crystallization. Truncations of N- or C-termini or the presence or absence of Tags can influence the quality of crystal packing (50). 12. As any protein crystal, the crystal is in equilibrium with the drop in which it grew. The so-called “mother liquor” is difficult to reproduce and becomes even more difficult to substitute for MPs because it contains also detergent and sometimes lipids in amounts that are difficult to determine precisely. Therefore, transferring the crystals to a solution containing cryoprotectant or soaking the crystals in solution containing heavy-atom derivatives can be difficult. In most cases, crystals have to be flash-frozen at 100â•›K for the diffraction experiments. It is usual to test several cryoprotectants with various procedures (time, concentration, intermediate transfer solutions). If in all cases the diffraction is poor, it is worthwhile testing the crystal quality prior to flash-cooling and even prior to any handling. It is now feasible to test directly the diffraction quality of crystals in the crystallization drop by setting the whole crystallization plate in the X-ray beam (56). Such setups are now available on a few synchrotron beam lines (i.e., BM30AFIP at ESRF, Grenoble). 13. The matrix most widely used for crystallization is composed of unsaturated monoglyceride monoolein and water. The temperature–composition phase diagram of the monoolein in water is shown in Fig.Â€ 13. At lower water content, two lamellar (Lc and La) phases exist (57). At hydration, over 20% MO spontaneously forms bicontinuous cubic phases of different symmetries (Ia3d and Pn3m), and the latter are stable in the presence of excess of water. Bicontinuous lipidic cubic phases are quite “resistant” to changes in temperature. The cubic phase is a gel-like material with high viscosity and stability. 14. The crystallization in the lipidic cubic phase environment has already enabled one to obtain crystal structures for a number of proteins such as bacteriorhodopsin (58–60), halorhodopsin (61), sensory rhodopsins (62–65), and the first complex of two membrane proteins: sensory rhodopsin II with transducer protein HtrII (66, 67), bacterial photosynthetic

Crystallography of Membrane Proteins

99

Fig.Â€13. A phase diagram of monooleoyl/water system (redrawn from (19)).

reaction centers (18, 68, 69), the cobalamin transporter BtuB (70), and the human b2-adrenergic G-protein-coupled receptor (31, 71, 72), which were the first high resolution structures of the first human “ligand” binding GPCRs. 15. Over the past several years, the behavior of lipid/water phases has been investigated under different conditions, in particular, the phase diagram of monoglyceride/detergent/water system in water (73). Further, there is a theoretical attempt to understand the mechanism of membrane protein crystal growth in the lipidic cubic phase (74). Theoretical calculations showed that the elastic energy of the deformation of the curved bilayer due to embedded proteins may be the driving force of crystallization. Unfortunately, this theory is not complete, and in addition, it is not yet known whether the model of crystallization system used for the calculation is correct. There are several possible reasons for efficiency of the method, and they have been discussed in the literature. However, the first question is whether the system is really in this phase. In fact, the probe for crystallization is different from the pure MO/water system. Indeed, it contains salts in the buffer and, what is most important, a certain amount of the detergent. A priori it was not evident that the phase diagram shown in Fig.Â€ 13 somehow describes the lipidic/membrane protein/ detergent/salt/water system as well. Therefore, the detailed kinetics of a proteolipidic cubic phase was studied by neutron scattering in the course of crystallization of the membrane protein bacteriorhodopsin. It was shown, that the initiation of crystallization process by salt addition leads to a dramatic decrease of the lattice constant, but no phase transition takes place. The cubic phase of Pn3m symmetry is observed during the entire crystallization process. No other phases are present in

100

Deniaud et al.

a macroscopic amount (75). Unfortunately, this does not exclude the presence of a small amount of other than the cubic phase (75). There is evidence of the presence of a lamellar phase around the growing protein crystals, but there is no definite proof of that (76). Further experimental and theoretical studies are required to elucidate the mechanism of crystallization. First of all, a very detailed experimental characterization of the system for crystallization via neutron and X-ray scattering techniques is necessary. 16. For X-ray data collection and also for other applications, single crystals grown in lipidic matrix need to be released from the highly viscous and sticky environment, placed onto nylon cryo-loops, and dumped into liquid nitrogen. There are several techniques for crystal mounting. Crystals could be directly removed from LCP by mechanical manipulation. In this case, the lipidic surrounding is mechanically withdrawn using microtools during observation with high resolution light microscopy. A single crystal with the rest of lipidic phase may be directly cryocooled as lipid served as cryoprotectant. Alternative fishing procedure is based on lipolysis of cubic liquid crystal dispersions of monoolein. The protein could be freed from the surrounding lipidic matrix by digesting the monoolein with lipase at room temperature during several hours or days (14). Lipase breaks ester bonds between glycerol and fatty acids of monoacylglycerols thereby embedded crystals are spontaneously released from the lipidic cubic phase and can easily be mounted from the aqueous phase or in oleic acid droplets by cryoloops. In the third technique, the cubic phase, which itself is membranous, may be solubilized by detergent (59). In the case of BR, the lipidic matrix with embedded crystals is transferred into 3Â€M Na/Na–Pi buffer (pH 5.6) containing 0.1% OG by microspatula and incubated during several days before crystal harvesting. It should be mentioned that in all these three cases, all manipulations with the crystals should be gentle since the crystals are fragile and very sensitive to environmental changes. For instance, dehydration quickly leads to crystal quality degradation. References 1. Deisenhofer J, Epp O, Miki K, Huber R, Michel H (1985) Structure of the protein subunits in the photosynthetic reaction centre of Rhodopseudomonas viridis at 3â•›Å resolution. Nature 318:618–624 2. White SH (2004) The progress of membrane protein structure determination. Protein Sci 13:1948–1949

3. http://www.rcsb.org/pdb/ 4. Drenth J (1994) Principles of protein X-ray crystallography. Springer, New York 5. Reiss-Husson F, Picot D (1999) Crystallization of membrane proteins. In: Ducruix A, Giégé R (eds) Crystallization of nucleic acids and proteins. A practical approach. Oxford University Press, Oxford

Crystallography of Membrane Proteins 6. Landau EM, Rosenbusch JP (1996) Lipidic cubic phases: a novel concept for the crystallization of membrane proteins. Proc Natl Acad Sci U S A 93:14532–14535 7. Mariani P, Luzzati V, Delacroix H (1988) Cubic phases of lipid-containing systems. Structure analysis and biological implications. J Mol Biol 204:165–189 8. Luzzati V, Tardieu A, Gulik-Krzywicki T, Rivas E, Reiss-Husson F (1968) Structure of the cubic phases of lipid-water systems. Nature 220:485–488 9. Scriven L (1976) Equilibrium bicontinuous structure. Nature 263:123–125 10. Landh T (1995) From entangled membranes to eclectic morphologies: cubic membranes as subcellular space organizers. FEBS Lett 369:13–17 11. Fontell K (1990) Cubic phases in surfactant and surfactant-like lipid systems. Colloid Polym Sci 268:264–285 12. Ericsson B. Eriksson P, Löfroth J, Engström S (1991) Cubic phases as delivery systems for peptide drugs. ACS symposium series 469:251–265 13. Gordeliy VI, Schlesinger R, Efremov R, Buldt G, Heberle J (2003) Crystallization in lipidic cubic phases: a case study with bacteriorhodops, vol 228, Methods in molecular biology. Humana Press, Clifton, NJ, pp 305–316 14. Nollert P, Landau EM (1998) Enzymic release of crystals from lipidic cubic phases. Biochem Soc Trans 26:709–713 15. Engström S, Alfons K, Rasmusson M, LjusbergWahren H (1998) Solvent-induced sponge (L3) phases in the solvent–monoolein–water system. Prog Colloid Polym Sci 108:93–98 16. Ridell A, Ekelund K, Evertsson H, Engström S (2003) On the water content of the solvent/monoolein/water sponge (L3) phase. Colloids Surf A Physicochem Eng Asp 228:17–24 17. Wadsten P, Wohri AB, Snijder A, Katona G, Gardiner AT, Cogdell RJ, Neutze R, Engstrom S (2006) Lipidic sponge phase crystallization of membrane proteins. J Mol Biol 364:44–53 18. Katona G, Andreasson U, Landau EM, Andreasson LE, Neutze R (2003) Lipidic cubic phase crystal structure of the photosynthetic reaction centre from Rhodobacter sphaeroides at 2.35Â€ A resolution. J Mol Biol 331:681–692 19. Cherezov V, Clogston J, Papiz MZ, Caffrey M (2006) Room to move: crystallizing membrane proteins in swollen lipidic mesophases. J Mol Biol 357:1605–1618

101

20. Papiz MZ, Prince SM, Howard T, Cogdell RJ, Isaacs NW (2003) The structure and thermal motion of the B800-850 LH2 complex from Rps. acidophila at 2.0A resolution and 100K: new structural features and functionally relevant motions. J Mol Biol 326:1523–1538 21. Wohri AB, Johansson LC, WadstenHindrichsen P, Wahlgren WY, Fischer G, Horsefield R, Katona G, Nyblom M, Oberg F, Young G, Cogdell RJ, Fraser NJ, Engstrom S, Neutze R (2008) A lipidic-sponge phase screen for membrane protein crystallization. StructureÂ€16:1003–1009 22. Kouyama T, Yamamoto M, Kamiya N, Iwasaki H, Ueki T, Sakurai I (1994) Polyhedral assembly of a membrane protein in its three-dimensional crystal. J Mol Biol 236:990–994 23. Takeda K, Sato H, Hino T, Kono M, Fukuda K, Sakurai I, Okada T, Kouyama T (1998) A novel three-dimensional crystal of bacteriorhodopsin obtained by successive fusion of the vesicular assemblies. J Mol Biol 283:463–474 24. Denkov ND, Yoshimura H, Kouyama T, Walz J, Nagayama K (1998) Electron cryomicroscopy of bacteriorhodopsin vesicles: mechanism of vesicle formation. Biophys J 74:1409–1420 25. Faham S, Bowie JU (2002) Bicelle crystallization: a new method for crystallizing membrane proteins yields a monomeric bacteriorhodopsin structure. J Mol Biol 316:1–6 26. Faham S, Boulting GL, Massey EA, Yohannan S, Yang D, Bowie JU (2005) Crystallization of bacteriorhodopsin from bicelle formulations at room temperature. Protein Sci 14: 836–840 27. Sanders CR, Schwonek JP (1992) Characterization of magnetically orientable bilayers in mixtures of dihexanoylphosphatidylcholine and dimyristoylphosphatidylcholine by solid-state NMR. Biochemistry 31:8898–8905 28. Sanders CR, Prestegard JH (1990) Magnetically orientable phospholipid bilayers containing small amounts of a bile salt analogue, CHAPSO. Biophys J 58:447–460 29. Sanders CR 2nd, Landis GC (1995) Reconstitution of membrane proteins into lipid-rich bilayered mixed micelles for NMR studies. Biochemistry 34:4030–4040 30. Rasmussen SG, Choi HJ, Rosenbaum DM, Kobilka TS, Thian FS, Edwards PC, Burghammer M, Ratnala VR, Sanishvili R, Fischetti RF, Schertler GF, Weis WI, Kobilka BK (2007) Crystal structure of the human beta2 adrenergic G-protein-coupled receptor. Nature 450:383–387

102

Deniaud et al.

31. Cherezov V, Rosenbaum DM, Hanson MA, Rasmussen SG, Thian FS, Kobilka TS, Choi HJ, Kuhn P, Weis WI, Kobilka BK, Stevens RC (2007) High-resolution crystal structure of an engineered human beta2adrenergic G protein-coupled receptor. Science 318:1258–1265 32. Ujwal R, Cascio D, Colletier JP, Faham S, Zhang J, Toro L, Ping P, Abramson J (2008) The crystal structure of mouse VDAC1 at 2.3Â€ A resolution reveals mechanistic insights into metabolite gating. Proc Natl Acad Sci U S A 105:17742–17747 33. Hiller S, Garces RG, Malia TJ, Orekhov VY, Colombini M, Wagner G (2008) Solution structure of the integral human membrane protein VDAC-1 in detergent micelles. Science 321:1206–1210 34. Ivetac A, Sansom MS (2008) Molecular dynamics simulations and membrane protein structure quality. Eur Biophys J 37:403–409 35. http://www.mpdb.ul.ie/ 36. Riekel C, Burghammer M, Schertler G (2005) Protein crystallography microdiffraction. Curr Opin Struct Biol 15:556–562 37. http://www.mpibp-frankfur t.mpg.de/ michel/public/memprotstruct.html 38. Garavito RM, Ferguson-Miller S (2001) Detergents as tools in membrane biochemistry. J Biol Chem 276:32403–32406 39. Wiener MC (2004) A pedestrian guide to membrane protein crystallization. Methods 34:364–372 40. Dahout-Gonzalez C, Brandolin G, PebayPeyroula E (2003) Crystallization of the bovine ADP/ATP carrier is critically dependent upon the detergent-to-protein ratio. Acta Crystallogr A 59:2353–2355 41. Alexandrov AI, Mileni M, Chien EY, Hanson MA, Stevens RC (2008) Microscale fluorescent thermal stability assay for membrane proteins. StructureÂ€16:351–359 42. daCosta CJ, Baenziger JE (2003) A rapid method for assessing lipid:protein and detergent:protein ratios in membrane-protein crystallization. Acta Crystallogr A 59:77–83 43. Pebay-Peyroula E (2007) Biophysical analysis of membrane proteins. Investigating structure and function. Wiley, United States 44. Slotboom DJ, Duurkens RH, Olieman K, Erkens GB (2008) Static light scattering to characterize membrane proteins in detergent solution. Methods 46:73–82 45. le Maire M, Arnou B, Olesen C, Georgin D, Ebel C, Moller JV (2008) Gel chromatography and analytical ultracentrifugation to determine

the extent of detergent binding and aggregation, and Stokes radius of membrane proteins using sarcoplasmic reticulum Ca2+-ATPase as an example. Nat Protoc 3:1782–1795 46. Abramson J, Smirnova I, Kasho V, Verner G, Kaback HR, Iwata S (2003) Structure and mechanism of the lactose permease of Escherichia coli. Science 301:610–615 47. Pebay-Peyroula E, Dahout-Gonzalez C, Kahn R, Trezeguet V, Lauquin GJ, Brandolin G (2003) Structure of mitochondrial ADP/ATP carrier in complex with carboxyatractyloside. Nature 426:39–44 48. Screpanti E, Padan E, Rimon A, Michel H, Hunte C (2006) Crucial steps in the structure determination of the Na+/H+ antiporter NhaA in its native conformation. J Mol Biol 362:192–202 49. Roth M, Lewit-Bentley A, Michel H, Deisenhofer J, Huber R, Oesterhelt D (1989) Detergent structure in crystals of a bacterial photosynthetic reaction center. Nature 340:659–662 50. Lemieux MJ, Reithmeier RA, Wang DN (2002) Importance of detergent and phospholipid in the crystallization of the human erythrocyte anion-exchanger membrane domain. J Struct Biol 137:322–332 51. Pebay-Peyroula E, Garavito RM, Rosenbusch JP, Zulauf M, Timmins PA (1995) Detergent structure in tetragonal crystals of OmpF porin. StructureÂ€3:1051–1059 52. Iwata S, Ostermeier C, Ludwig B, Michel H (1995) Structure at 2.8Â€A resolution of cytochrome c oxidase from Paracoccus denitrificans. Nature 376:660–669 53. Zhou Y, Morais-Cabral JH, Kaufman A, MacKinnon R (2001) Chemistry of ion coordination and hydration revealed by a K+ channel-Fab complex at 2.0Â€ A resolution. Nature 414:43–48 54. Hunte C, Michel H (2002) Crystallisation of membrane proteins mediated by antibody fragments. Curr Opin Struct Biol 12:503–508 55. Steiner D, Forrer P, Pluckthun A (2008) Efficient selection of DARPins with subnanomolar affinities using SRP phage display. J Mol Biol 382:1211–1227 56. Jacquamet L, Ohana J, Joly J, Borel F, Pirocchi M, Charrault P, Bertoni A, Israel-Gouy P, Carpentier P, Kozielski F, Blot D, Ferrer JL (2004) Automated analysis of vapor diffusion crystallization drops with an X-ray beam. StructureÂ€12:1219–1225 57. Briggs J, Ching H, Caffrey M (1996) The temperature–composition phase diagram and mesophase structure characterization of

Crystallography of Membrane Proteins

58.

59.

60.

61.

62.

63.

64.

65.

66.

67.

the monoolein/water system. J Phys II 6:723–751 Belrhali H, Nollert P, Royant A, Menzel C, Rosenbusch JP, Landau EM, Pebay-Peyroula E (1999) Protein, lipid and water organization in bacteriorhodopsin crystals: a molecular view of the purple membrane at 1.9Â€A resolution. StructureÂ€7:909–917 Luecke H, Schobert B, Richter HT, Cartailler JP, Lanyi JK (1999) Structure of bacteriorhodopsin at 1.55Â€ A resolution. J Mol Biol 291:899–911 Pebay-Peyroula E, Rummel G, Rosenbusch JP, Landau EM (1997) X-ray structure of bacteriorhodopsin at 2.5 angstroms from microcrystals grown in lipidic cubic phases. Science 277:1676–1681 Kolbe M, Besir H, Essen LO, Oesterhelt D (2000) Structure of the light-driven chloride pump halorhodopsin at 1.8Â€ A resolution. Science 288:1390–1396 Luecke H, Schobert B, Lanyi JK, Spudich EN, Spudich JL (2001) Crystal structure of sensory rhodopsin II at 2.4 angstroms: insights into color tuning and transducer interaction. Science 293:1499–1503 Royant A, Nollert P, Edman K, Neutze R, Landau EM, Pebay-Peyroula E, Navarro J (2001) X-ray structure of sensory rhodopsin II at 2.1-A resolution. Proc Natl Acad Sci U S A 98:10131–10136 Vogeley L, Sineshchekov OA, Trivedi VD, Sasaki J, Spudich JL, Luecke H (2004) Anabaena sensory rhodopsin: a photoÂ� chromic color sensor at 2.0Â€ A. Science 306:1390–1393 Reshetnyak AA, Borshchevskiy VI, Klare J, Moiseeva ES, Engelhardt M, Buldt G, Gordeliy VI (2008) Comparative analysis of Sensory rhodopsin II structures in complex with a transducer and without it. J Surf Invest 2:894–899 Gordeliy VI, Labahn J, Moukhametzianov R, Efremov R, Granzin J, Schlesinger R, Buldt G, Savopol T, Scheidig AJ, Klare JP, Engelhard M (2002) Molecular basis of transmembrane signalling by sensory rhodopsin II-transducer complex. Nature 419:484–487 Moukhametzianov R, Klare JP, Efremov R, Baeken C, Goppner A, Labahn J, Engelhard M,

68.

69.

70.

71.

72.

73.

74. 75.

76.

103

Buldt G, Gordeliy VI (2006) Development of the signal in sensory rhodopsin and its transfer to the cognate transducer. Nature 440:115–119 Chiu ML, Nollert P, Loewen MC, Belrhali H, Pebay-Peyroula E, Rosenbusch JP, Landau EM (2000) Crystallization in cubo: general applicability to membrane proteins. Acta Crystallogr A 56:781–784 Katona G, Snijder A, Gourdon P, Andreasson U, Hansson O, Andreasson LE, Neutze R (2005) Conformational regulation of charge recombination reactions in a photosynthetic bacterial reaction center. Nat Struct Mol Biol 12:630–631 Misquitta LV, Misquitta Y, Cherezov V, Slattery O, Mohan JM, Hart D, Zhalnina M, Cramer WA, Caffrey M (2004) Membrane protein crystallization in lipidic mesoÂ�phases with tailored bilayers. StructureÂ€12:2113–2124 Rosenbaum DM, Cherezov V, Hanson MA, Rasmussen SG, Thian FS, Kobilka TS, Choi HJ, Yao XJ, Weis WI, Stevens RC, Kobilka BK (2007) GPCR engineering yields high-resolution structural insights into beta2-adrenergic receptor function. Science 318:1266–1273 Jaakola VP, Griffith MT, Hanson MA, Cherezov V, Chien EY, Lane JR, Ijzerman AP, Stevens RC (2008) The 2.6 angstrom crystal structure of a human A2A adenosine receptor bound to an antagonist. Science 322:1211–1217 Persson G, Edlund H, Lindblom G (2004) Phase behaviour of the 1-monooleoylracglycerol/n-octyl-b-d-glucoside/water system. Prog Colloid Polym Sci 123:36–39 Grabe M, Neu J, Oster G, Nollert P (2003) Protein interactions and membrane geometry. Biophys J 84:854–868 Efremov R, Shiryaeva GN, Islamov A, Kuklin A, Yaguzhinsky L, Fragneto-Cusani G, Bueldt G, Gordeliy VI (2005) SANS investigations of the lipidic cubic phase behaviour in course of bacteriorhodopsin crystallization. J Cryst Growth 275:1453–1459 Cherezov V, Caffrey M (2007) Membrane protein crystallization in lipidic mesophases. A mechanism study using X-ray microdiffraction. Faraday Discuss 136:195–212 (discussion 3–29)

as

Chapter 6 Structural Approaches of the Mitochondrial Carrier Family Hugues Nury, Iulia Blesneac, Stephanie Ravaud, and Eva Pebay-Peyroula Abstract The transport of solutes across the inner mitochondrial membrane is highly selective and necessitates membrane proteins mainly from the mitochondrial carrier family (MCF). These carriers are required for the transport of a variety of metabolites implicated in all the important processes occurring within the mitochondrial matrix. Due to its high abundance, the ADP/ATP carrier (AAC) is the member of the family that was studied most. It is the first mitochondrial carrier for which a high-resolution X-ray structure is known. The carrier was crystallized in the presence of a strong inhibitor, the carboxyatractyloside (CATR). The structure gives an insight not only into the overall fold of mitochondrial carriers in general but also into atomic details of the AAC in a conformation that is open toward the intermembrane space (IMS). Molecular dynamics simulations indicate the first events occurring to the carrier after the binding of ADP. A careful analysis of the primary sequences of all the carriers in light with the structure highlights properties of the protein that are related to the substrate. Key words: Membrane protein, Transport, Inner mitochondrial membrane, ADP and ATP transport

1. Introduction Mitochondria are organelles in which major metabolic cycles, Krebs cycle, degradation of fatty acids, and others take place. They also provide the environment to produce ATP, the main fuel for cellular processes. The molecular complexes of the respiratory chain located in the inner membrane create a proton gradient that is further used by the ATP-synthase, a membrane machinery, which catalyzes the production of ATP from ADP and inorganic phosphate. The structures of several components of the respiratory chain, including domains of the ATP-synthase (1, 2), are now known at quasi-atomic resolution and have shed light Jean-Jacques Lacapère (ed.), Membrane Protein Structure Determination: Methods and Protocols, Methods in Molecular Biology, vol. 654, DOI 10.1007/978-1-60761-762-4_6, © Springer Science+Business Media, LLC 2010

105

106

Nury et al.

Fig.Â€1. The MCF motif. The characteristic MCF motif is found within each repeat of 100 amino acids present three-times in MCF carriers. The limits of the helices determined in the AAC structure are superimposed on the consensus MCF sequence. It spans over the second part of an odd-numbered helix (Hn) to the first part of the following evennumbered helix (Hn+1). “a” represents an aromatic amino acid.

into major mechanisms, such as the rotary mechanism of ATP-synthase. The production of ATP as well as the function of all the major cycles is strongly connected to the import and export of metabolites between the cytosol and the mitochondrial matrix. The metabolites are small molecules with a large variety of size and chemical characteristics. They are able to cross the outer membrane relatively easily via a porin-like membrane protein, VDAC, which structure has been solved in 2008 by NMR (3) and X-ray crystallography (4). The transport through VDAC is rather unspecific and mainly limited by the size of the substrate with a slight preference for anionic molecules. On the contrary, the inner membrane is impermeable, and molecules are only transported through very specific membrane carriers. Many of these carriers belong to the MCF characterized by a specific motif of amino acids (Fig.Â€1) first identified by Walker etÂ€al. (5) and further described in (6). More than 40 carriers were described in human. They are encoded by the SLC25 gene family, and several are related to diseases as reviewed in (7). An overview of plant mitochondrial carriers is presented in (8). The ADP/ATP carrier (AAC) is the most abundant – it represents 10% of the proteins in bovine heart mitochondria – and was therefore studied extensively in the last 30 years (9). It serves as a model for other MCF carriers. This chapter first details the structural studies of the bovine AAC, discusses possible implications for the transport mechanism, and in particular, the import of ADP, and finally discusses general implications for the whole MCF.

2. Methods to Solve the Structure of Bovine AAC 2.1. General Strategy

The 3D structure of the bovine AAC was solved by X-ray crystallography (10). The same year, Kunji etÂ€al. published a 2D projection map obtained by electron microscopy from 2D crystals (11). Transporters often undergo large conformational changes in order to shuttle molecules from one side of the membrane to the other. As a consequence, in the absence of ligands, they display large conformational heterogeneities in solution. In addition, detergent molecules that surround the protein after solubilization

Structural Approaches of the Mitochondrial Carrier Family

107

do not provide the same environment as the native lipidic membrane, and large conformational changes are not necessarily well controlled and might favor the loss of the tertiary structure. It is therefore of interest to find the appropriate conditions to stabilize a single conformation and block the transport process. This can be achieved by several ways: binding a ligand that is not transported, crystallizing a single point mutant that has no transport activities as done for the lactose permease (12), or crystallizing at a pH in which the transport is blocked as exemplified with the Na/H antiporter (13). All these strategies necessitate a good knowledge of the biochemical and functional properties of the transporter. Inhibitors of AAC were well described since the 1980s. Two families of molecules are known, atractylosides (ATR) and carboxyatractyloside (CATR) on one hand, and bongkrekic acid (BA) and iso-bongkrekic acid on the other. The first is known to block the entrance of ADP from the intermembrane space (IMS) (14), the second prevents the export of ATP from the matrix (15). They were described to induce the extreme conformations adopted by the carrier during the transport process. These inhibitors are therefore adequate tools for structural characterization of AAC. 2.2. 3D Crystallization

AAC was extracted and purified from bovine heart mitochondria with a protocol derived from (16). Importantly, mitochondria have to be incubated in the presence of CATR in order to stabilize the carrier from the first step of extraction (17). Because of the constraints related to the purification of a nontagged protein, many lipid and detergent molecules are copurified with the carrier and cannot be washed away easily. As a consequence, the protein solution after concentration to 10Â€ mg/mL for the crystallization contains about 20Â€ g of detergent per gram of protein, a much larger amount than usual. In particular, the very large amount of detergent was incompatible with the lipid cubic phase approach (18) and prevented crystallization in the standard vapor diffusion method. Excess detergent was therefore removed using polystyrene beads (moist activated SM2 Biobeads, Bio-Rad) known to adsorb detergent molecules at their surface. Biobeads are commonly used for 2D crystallization (19). In the presence of lipids, they are added to proteins solubilized in detergent. The adsorption of detergents drives the incorporation of proteins into the liposomes. For AAC, the amount of removed detergent as a function of time could be followed by radioactive labeled detergent (17). Although the crystallization was achieved by vapor diffusion, the presence of endogenous lipids probably favored the crystallization in a similar way as for 2D crystallization. Partially removing the detergent forces the protein to interact with lipids, and the presence of precipitants favors protein–protein interactions. As a result, the crystals can be described as ordered stacks of 2D crystals.

108

Nury et al.

2.3. Structure Determination

Two crystal forms were necessary to solve the structure. The first crystal form had a Pâ•›â•›21212 symmetry. Most of these crystals contained a defect, that is, a few protein layers (perpendicular to the c-axis) were displaced along the b-axis by approximately 2â•›Å (17). However, a diffraction data set to 2.2â•›Å could be collected from a perfect crystal and was used for the refinement of the molecular model. The high rate of defects prevented the search for heavy atom derivatives necessary for phase determination. Therefore, another crystal form was needed. Because the crystallization trials had highlighted the strong influence of various salts, the salt concentration in the buffer used for the final purification step was reduced (5Â€mM NaCl instead of 100Â€mM). New crystallization set-ups were made, and a Câ•›2221 crystal form was identified. Despite anisotropic diffraction, this form could be used for finding heavy atom derivatives that allowed the phases to be determined to a resolution of 3â•›Å. Three mercury derivatives were used, and they interacted with the four cysteins of AAC with different occupancies. A first model was constructed from the experimental electron density and transposed by molecular replacement to the 2.2â•›Å data of the primitive crystal form. The model was progressively completed and refined on the basis of 2Fo-Fc electron density maps combined with the experimental MIR maps (translated and rotated so that they would match with the position of the carrier in the Pâ•›â•›21212 cell). The refinement was then conducted by alternate cycles of manual modifications using the program O (20) and energy minimization or simulated annealing using CNS (21). After complete reconstruction of the molecule with the exception of a little number of disordered amino acids in the N- and C-termini, the location of the inhibitor was clearly visible from 2Fo-Fc, and Fo-Fc electron density maps in the center of the large cavity formed by the six tilted and partially kinked transmembrane (TM) helices. In the final electron density, a few lipid molecules could be modeled. By preparing the protein without any salt in the final purification step, a slightly different unit cell in the C2221 symmetry was obtained. The refinement of this structure showed three cardiolipins organized around the TM helices on the matrix-side leaflet of the membrane (22).

2.4. Description of the Structure

As predicted from the primary sequence analysis, the carrier has six helices that span over the membrane, connected by short loops on the side of the IMS. The overall fold is shown in Fig.Â€2. On the matrix side, the amino acids connecting the TM helices do not form long unstructured loops as depicted in previous published articles but are partially structured in small amphipathic helices as predicted in (23). The TM helices are all six tilted with respect to the axis that is perpendicular to the membrane, and the three odd-numbered ones are kinked by more than 30° at the level of the conserved prolines of the MCF motifs. A cavity largely open

Structural Approaches of the Mitochondrial Carrier Family

109

Fig.Â€2. Ribbon representation of AAC. Panels (a), (b) and (c) view the bovine ADP/ATP carrier from the intermembrane space, the matrix and the side, respectively. The N- and C-termini are labeled in panel (c). Panels (a) and (b) indicate the transmembrane and the short amphipathic helices, respectively.

toward the IMS and penetrating deeply toward the matrix side is formed by the tilts and kinks of the TM helices. The three oddnumbered helices are longer than initially predicted. It can also be noted that all the TM helices are largely surrounded by solvent in the central cavity that penetrates deeply into the carrier (Fig.Â€3). This is particularly the case for H2, which has a low hydrophobic score and was not easily predictable as a TM helix from the primary sequence. The structure solved to 2.2â•›Å revealed several interesting aspects (10). The inhibitor, CATR, cocrystallized with the carrier, is located in the cavity with its diterpene moiety oriented toward the matrix and the two sulfate groups pointing toward the IMS. Almost all the chemical groups of CATR interact with AAC via hydrogen bonds or van der Waals contacts, except the primary alcohol group located on the sugar moiety that points toward the cavity. The tight interaction between AAC and CATR is consistent with the nanomolar affinity of the inhibitor for AAC. Both carboxylates of CATR interact with the carrier, one through a direct electrostatic interaction, the other via a water molecule,

110

Nury et al.

Fig.Â€3. The cavity and the binding site of CATR. The figure shows a longitudinal section through the cavity, which is largely open toward the intermembrane space. The surface of the protein is represented. The inhibitor, CATR, is depicted in dots filling the van der Waals volumes of CATR atoms.

thus explaining the slightly lower affinity for ATR, which has only one carboxylate group. The most interesting findings are the locations of the MCF residues. The first halves of the MCF motifs are located on the odd-numbered helices, the two first charged amino acids following the proline interact via salt bridges and connect the helices two by two as shown in Fig.Â€ 4. These salt bridges were predicted from mutant studies (24). The kinks at the level of the signature prolines bring the charged residues in close vicinity and allow the formation of salt bridges, thus forcing the closure of AAC toward the matrix. 2.5. Oligomerization, Organization in the Membrane, 2D Crystals

The structure solved by X-ray crystallography demonstrated that a monomeric form can exist as a stable entity, and that it contains a cavity, the size of which is compatible with nucleotide binding. With a few structural rearrangements, it is plausible that the transport occurs through this single monomer. The Câ•›2221 crystal form highlights protein–protein contacts mediated by cardiolipins (22). However, the contacts are rather weak and could be induced by the crystallization. At the same time, a 2D projection map of the yeast AAC3 was determined by electron microscopy (11). The carriers located in the projection maps are compatible with the X-ray structure and are organized in the 2D layer without obvious dimerization. The carrier solubilized in detergent was analyzed by several functional, biochemical, and biophysical approaches, most of them showing a monomer (25–27). All these findings seem to

Structural Approaches of the Mitochondrial Carrier Family

111

Fig.Â€4. Salt bridges formed by MCF residues. The residues of the first part of the MCF are located on helices H1, H3 and H5. Prolines are responsible for the sharp kinks of the odd-numbered helices, acidic and basic residues form salt bridges that connect these three helices two by two.

contradict previous evidences for dimerization. Whether the carrier forms dimers in the membrane and whether its oligomerization is necessary for the function are still open questions. It is highly probable that local high concentration of AAC in the mitochondrial membrane favors close packing that could account for an efficient transport. The 2D crystallization was done in the presence of ATR and the carrier has therefore the same conformation as in the 3D crystals. Even though the resolution of 2D crystals is much lower than 3D crystals, 2D crystals have the advantage of being formed by a monolayer of proteins that are accessible from both sides making the exchange of inhibitors by soaking possible.

3. The Transport Mechanism 3.1. What Is Suggested by the Structure?

The current structure is that of AAC in a single conformation. It represents most probably the conformation able to attract and bind ADP from the IMS or at least is close to this ADP conformation.

112

Nury et al.

Although the structure is not sufficient to fully understand the transport mechanism, it highlights the location of residues that are known to be important for the function of AAC. It also serves as a basis for molecular dynamic simulations. It was known from the amino acid sequence analysis that AACs from all species contain a highly conserved sequence in the third MCF motif, RRRMMM, where the first and last arginines belong to the MCF motif. In addition, this sequence is highly specific and is not found in any other carriers (and even any other protein). Therefore, the RRRMMM signature has to be related to the function. The structure revealed that the sequence spans over the 10â•›Å from the bottom of the cavity to almost the matrix side. R234 and R235 both point toward the cavity and interact with the inhibitor. R236 is oriented toward the matrix, not far from the surface, and its side chain is implicated in a salt bridge with E264. In the current known conformation, R234 and R235 are positioned to interact with an ADP that would enter from IMS. R236 is buried, but even a small conformational change would allow this residue to be accessible from the matrix so that it could possibly interact with ATP. Because the carrier transports negatively charged molecules, and before the structure was known, many of the conserved basic residues were mutated in order to test their functional relevance (28–30). The structure reveals that several of them are located in the cavity and form basic patches that line the cavity from the entrance to the bottom. In parallel, three tyrosines form an intriguing “ladder” from the entrance of the cavity to the bottom and were postulated to guide the nucleotide ring (31). The mutation of two of these tyrosines is known to prevent AAC to function (32). The location of the MCF residues suggests that the kinks could be modulated by the existence, or nonexistence of the salt bridges, while the second part of the motif would act as a hinge that could follow the movement induced by a kink modification of the odd-numbered helices. 3.2. Molecular Dynamics Simulations on AAC

Since the structure of AAC was published, several groups have performed molecular dynamics simulations on the AAC. The first approach aimed at following conformational changes occurring to the carrier in the absence of CATR (33). These authors compared the trajectories in the presence and in the absence of CATR and found that a number of salt bridges close to the matrix side are rather unstable in the absence of CATR. Their studies also highlighted that odd-numbered helices can undergo wobble movements around the conserved prolines, whereas even-numbered helices can undergo a face rotation. The rotation of these helices is partially consistent with experimental data that evidenced a difference in their accessibility measured between the states blocked by CATR and BA, respectively (34). Although these simulations

Structural Approaches of the Mitochondrial Carrier Family

113

were performed on very short time scales (10Â€ns), such observations could indicate the first events of conformational changes that occur during the transport. Similar calculations but with longer time scales (4â•›×â•›20Â€ns) also hinted to a possible hinge role of the conserved prolines (35). Longer MD simulations up to 0.3Â€ ms (36) and 0.53Â€ms (37) in the presence of ADP highlighted a strong electrostatic funnel that drives ADP3− very fast into the cavity. Both studies showed the binding site of ADP at the bottom of the cavity with its phosphates pointing toward the matrix side and interacting with the residues of the conserved salt bridges. The binding of ADP could therefore perturb the salt bridges formed by the conserved MCF residues and induce the first events leading to the opening toward the matrix side. The three salt bridges are not equally involved. In particular MCF residues from helix 3 are not involved in ADP binding, compared to MCF residues in helices 1 and 5 (37). It has to be noted that MCF residues from helix 3 are the less conserved among all the AACs. 3.3. Comparison and Differences with Other Transporters

The structures of several membrane transporters are available in the Protein Data Base and listed in (38). Interestingly, many transporters have more than ten transmembrane helices. The transport pathway for the substrates is usually located in the middle of the protein surrounded by an inner shell of TM helices, whereas the other helices constitute an outer shell. These transporters adopt a more compact form, and conformational changes during the transport process mainly occur within the central part of the protein. The outer layer of TMs could dampen the movements. In AAC, within its half facing the IMS, the six TM helices form a thin protein-layer, which separates the polar cavity from the lipidic membrane. Therefore, each single helix is in contact with the solvent on one side and with lipids on the other. The helical bundle is quite dynamic on the side of the IMS as shown by normal mode calculations and by the experimental temperature factors of the N- and C-termini and cytoplasmic loops. The structure and dynamics of AAC could explain the extreme fragility of the carrier outside the lipidic bilayer. An excess of detergent can interfere with helix–helix interactions and destroy the 3D structure, unless it is stabilized by a strongly bound inhibitor. Individual cardiolipin molecules were shown to be still bound to AAC after purification from the mitochondrial membrane. These lipids formed by four acyl chains (Fig.Â€ 5a) are abundant in the mitochondrial membrane. Three of them were identified in the crystal structure (Fig.Â€5b, c). Their tight interaction with AAC residues belonging to the second half of the MCF motif is noticeable. If a hinge movement around these residues occurs, then the lipids could control the movement, and follow the helices in order to avoid hydrophobic mismatch provoked by twists or kink modifications of the helices.

114

Nury et al.

Fig.Â€5. Interactions of cardiolipins and AAC. (a) Stick representation of a cardiolipin molecule. (b, c) The three cardiolipins interacting with AAC are represented as van der Waals spheres in black, and residues from the second half of the MCF motif in grey, the backbone of AAC is depicted in white. (b) is viewed from the side and highlights the large surface of AAC covered by the cardiolipins. (c) is viewed from the matrix and represents the interactions between cardiolipins and the residues from the second half of the MCF motif.

4. The Mitochondrial Carrier Family

The structure of AAC is still the only structure of a MCF carrier. Despite the relative low sequence identities (maximum of 20%), the presence of the three MCF motifs most probably constrains the structures so that the overall fold of each carrier should be similar. Indeed, Fig.Â€6 shows the volume occupied by the MCF motif residues. The first half of the three MCF motifs including the prolines forming the kinks and the salt bridges that stabilize the kinked helices, form a compact core of the carrier (represented in dark gray in Fig.Â€6). The second part of the three MCF motifs (in light grey) surrounds this core volume. The conservation of the MCF motifs and their location in the core part of the structure as seen from AAC, most probably shapes all of these carriers in a similar way. However, the substrates transported differ drastically by their size and chemical nature. Several homology models of other MCF members were built on the basis of AAC. Because the sequence identities are less than 20%, their significance as such is questionable but the models guide the choice of mutants in functional studies as shown for the ornithine/citrulline carrier (39), or for the oxoglutarate carrier (40). Residues that are highly conserved among AACs are concentrated within the cavity (41). This is consistent with the high selectivity of the carriers for their substrates. Although the overall

Structural Approaches of the Mitochondrial Carrier Family

115

Fig.Â€6. Location of the MCF motif residues. The residues that are characteristic of the MCF motif are represented as van der Waals spheres. The first part of the motif is depicted in dark grey, the second half in light grey, the surface of the carrier is in white. Panel (a) is viewed from the side, panel (b) from the matrix.

backbone of AAC (and most probably of all other MCF carriers) displays a threefold pseudo-symmetry, the deviation from the symmetry by the side chains present in the cavity, is certainly related to the binding properties of non-symmetrical substrates. This idea was exploited by the group of Kunji. After a first sequence analysis coupled to homology modeling from which they postulated a common substrate binding site for MCFs (42), Kunji etÂ€al. compared the three repeats within each MCF member and scored the deviations from a threefold symmetry within these repeats. The representation of the deviations on the 3D structures highlights clusters of high deviation within the cavity postulated to be related to substrate binding (43).

5. Conclusion The knowledge of the biochemical behavior of AAC purified from bovine heart mitochondria gained over two decades, together with the development and expertise in crystallization of membrane proteins, made it possible to obtain crystals of high diffracting quality. The X-ray structure of AAC determined in the presence of the inhibitor CATR gave a first insight in ADP/ATP transport. It opened the way to molecular dynamics and new experiments. However, a complete understanding of the transport mechanism and its selectivity awaits further experimental results, that is, structures of other conformations and other MCF carriers. In particular the AAC structure in the presence of BA would be a major step forward.

116

Nury et al.

References 1. Abrahams JP, Leslie AG, Lutter R, Walker JE (1994) Structure at 2.8 A resolution of F1-ATPase from bovine heart mitochondria. Nature 370:621–628 2. Stock D, Leslie AG, Walker JE (1999) Molecular architecture of the rotary motor in ATP synthase. Science 286:1700–1705 3. Hiller S, Garces RG, Malia TJ, Orekhov VY, Colombini M, Wagner G (2008) Solution structure of the integral human membrane protein VDAC-1 in detergent micelles. Science 321:1206–1210 4. Ujwal R, Cascio D, Colletier JP, Faham S, Zhang J, Toro L, Ping P, Abramson J (2008) The crystal structure of mouse VDAC1 at 2.3 A resolution reveals mechanistic insights into metabolite gating. Proc Natl Acad Sci U S A 105:17742–17747 5. Walker JE, Runswick MJ (1993) The mitochondrial transport protein superfamily. J Bioenerg Biomembr 25:435–446 6. Jezek P, Jezek J (2003) Sequence anatomy of mitochondrial anion carriers. FEBS Lett 534:15–25 7. Palmieri F (2008) Diseases caused by defects of mitochondrial carriers: a review. Biochim Biophys Acta 1777:564–578 8. Laloi M (1999) Plant mitochondrial carriers: an overview. Cell Mol Life Sci 56:918–944 9. Klingenberg M (2008) The ADP and ATP transport in mitochondria and its carrier. Biochim Biophys Acta 1778:1978–2021 10. Pebay-Peyroula E, Dahout-Gonzalez C, Kahn R, Trezeguet V, Lauquin GJ, Brandolin G (2003) Structure of mitochondrial ADP/ATP carrier in complex with carboxyatractyloside. Nature 426:39–44 11. Kunji ER, Harding M (2003) Projection structure of the atractyloside-inhibited mitochondrial ADP/ATP carrier of Saccharomyces cerevisiae. J Biol Chem 278:36985–36988 12. Abramson J, Smirnova I, Kasho V, Verner G, Kaback HR, Iwata S (2003) Structure and mechanism of the lactose permease of Escherichia coli. Science 301:610–615 13. Hunte C, Screpanti E, Venturi M, Rimon A, Padan E, Michel H (2005) Structure of a Na/H antiporter and insights into mechanism of action and regulation by pH. Nature 435:1197–1202 14. Block MR, Lauguin GJ, Vignais PV (1981) Atractyloside and bongkrekic acid sites in the mitochondrial ADP/ATP carrier protein. An appraisal of their unicity by chemical modifications. FEBS Lett 131:213–218

15. Henderson PJ, Lardy HA (1970) Bongkrekic acid. An inhibitor of the adenine nucleotide translocase of mitochondria. J Biol Chem 245:1319–1326 16. Kramer R, Aquila H, Klingenberg M (1977) Isolation of the unliganded adenosine 5’-diphosphate, adenosine 5’-triphosphate carrier-linked binding protein and incorporation into the membranes of liposomes. Biochemistry 16:4949–4953 17. Dahout-Gonzalez C, Brandolin G, PebayPeyroula E (2003) Crystallization of the bovine ADP/ATP carrier is critically dependent upon the detergent-to-protein ratio. Acta Crystallogr 59:2353–2355 18. Landau EM, Rosenbusch JP (1996) Lipidic cubic phases: a novel concept for the crystallization of membrane proteins. Proc Natl Acad Sci U S A 93:14532–14535 19. Lacapere JJ, Stokes DL, Olofsson A, Rigaud JL (1998) Two-dimensional crystallization of Ca-ATPase by detergent removal. Biophys J 75:1319–1329 20. Jones TA, Zou JY, Cowan SW, Kjeldgaard M (1991) Improved methods for building protein models in electron density maps and the location of errors in these models. Acta Crystallogr A 47:110–119 21. Brunger AT, Adams PD, Clore GM, DeLano WL, Gros P, Grosse-Kunstleve RW, Jiang JS, Kuszewski J, Nilges M, Pannu NS, Read RJ, Rice LM, Simonson T, Warren GL (1998) Crystallography & NMR system: a new software suite for macromolecular structure determination. Acta Crystallogr 54:905–921 22. Nury H, Dahout-Gonzalez C, Trezeguet V, Lauquin G, Brandolin G, Pebay-Peyroula E (2005) Structural basis for lipid-mediated interactions between mitochondrial ADP/ ATP carrier monomers. FEBS Lett 579:6031–6036 23. Panneels V, Schussler U, Costagliola S, Sinning I (2003) Choline head groups stabilize the matrix loop regions of the ATP/ADP carrier ScAAC2. Biochem Biophy Res Commun 300:65–74 24. Nelson DR, Felix CM, Swanson JM (1998) Highly conserved charge-pair networks in the mitochondrial carrier family. J Mol Biol 277:285–308 25. Bamber L, Harding M, Monne M, Slotboom DJ, Kunji ER (2007) The yeast mitochondrial ADP/ATP carrier functions as a monomer in mitochondrial membranes. Proc Natl Acad Sci U S A 104:10830–10834

Structural Approaches of the Mitochondrial Carrier Family 26. Bamber L, Slotboom DJ, Kunji ER (2007) Yeast mitochondrial ADP/ATP carriers are monomeric in detergents as demonstrated by differential affinity purification. J Mol Biol 371:388–395 27. Nury H, Manon F, Arnou B, le Maire M, Pebay-Peyroula E, Ebel C (2008) Mitochondrial bovine ADP/ATP carrier in detergent is predominantly monomeric but also forms multimeric species. Biochemistry 47:12319–12331 28. Nelson DR, Lawson JE, Klingenberg M, Douglas MG (1993) Site-directed mutagenesis of the yeast mitochondrial ADP/ATP translocator. Six arginines and one lysine are essential. J Mol Biol 230:1159–1170 29. Muller V, Heidkamper D, Nelson DR, Klingenberg M (1997) Mutagenesis of some positive and negative residues occurring in repeat triad residues in the ADP/ATP carrier from yeast. Biochemistry 36:16008–16018 30. Heimpel S, Basset G, Odoy S, Klingenberg M (2001) Expression of the mitochondrial ADP/ATP carrier in Escherichia coli. Renaturation, reconstitution, and the effect of mutations on 10 positive residues. J Biol Chem 276:11499–11506 31. Pebay-Peyroula E, Brandolin G (2004) Nucleotide exchange in mitochondria: insight at a molecular level. Curr Opin Struct Biol 14:420–425 32. David C, Arnou B, Sanchez JF, Pelosi L, Brandolin G, Lauquin GJ, Trezeguet V (2008) Two residues of a conserved aromatic ladder of the mitochondrial ADP/ATP carrier are crucial to nucleotide transport. Biochemistry 47:13223–13231 33. Falconi M, Chillemi G, Di Marino D, D’Annessa I, Morozzo della Rocca B, Palmieri L, Desideri A (2006) Structural dynamics of the mitochondrial ADP/ATP carrier revealed by molecular dynamics simulation studies. Proteins 65:681–691 34. Kihira Y, Iwahashi A, Majima E, Terada H, Shinohara Y (2004) Twisting of the second transmembrane alpha-helix of the mitochon-

35.

36.

37.

38. 39.

40.

41.

42.

43.

117

drial ADP/ATP carrier during the transition between two carrier conformational states. Biochemistry 43:15204–15209 Johnston JM, Khalid S, Sansom MS (2008) Conformational dynamics of the mitochondrial ADP/ATP carrier: a simulation study. Mol Membr Biol 25:506–517 Wang Y, Tajkhorshid E (2008) Electrostatic funneling of substrate in mitochondrial inner membrane carriers. Proc Natl Acad Sci U S A 105:9598–9603 Dehez F, Pebay-Peyroula E, Chipot C (2008) Binding of ADP in the mitochondrial ADP/ ATP carrier is driven by an electrostatic funnel. J Am Chem Soc 130:12725–12733 http://blanco.biomol.uci.edu/Membrane_ Proteins_xtal.html Tonazzi A, Giangregorio N, Palmieri F, Indiveri C (2005) Relationships of Cysteine and Lysine residues with the substrate binding site of the mitochondrial ornithine/citrulline carrier: an inhibition kinetic approach combined with the analysis of the homology structural model. Biochim Biophys Acta 1718:53–60 Cappello AR, Miniero DV, Curcio R, Ludovico A, Daddabbo L, Stipani I, Robinson AJ, Kunji ER, Palmieri F (2007) Functional and structural role of amino acid residues in the odd-numbered transmembrane alphahelices of the bovine mitochondrial oxoglutarate carrier. J Mol Biol 369:400–412 Nury H, Dahout-Gonzalez C, Trezeguet V, Lauquin GJ, Brandolin G, Pebay-Peyroula E (2006) Relations between structure and function of the mitochondrial ADP/ATP carrier. Annu Rev Biochem 75:713–741 Robinson AJ, Kunji ER (2006) Mitochondrial carriers in the cytoplasmic state have a common substrate binding site. Proc Natl Acad Sci U S A 103:2617–2622 Robinson AJ, Overy C, Kunji ER (2008) The mechanism of transport by mitochondrial carriers based on analysis of symmetry. Proc Natl Acad Sci U S A 105:17766–17771

as

Chapter 7 What Can Be Learned About the Function of a Single Protein from Its Various X-Ray Structures: The Example of the Sarcoplasmic Calcium Pump Jesper Vuust Møller, Claus Olesen, Anne-Marie Lund Winther, and Poul Nissen Abstract Improvements in the handling of membrane proteins for crystallization, combined with better synchrotron sources for X-ray diffraction analysis, are leading to clarification of the structural details of an ever increasing number of membrane transporters and receptors. Here we describe how this development has resulted in the elucidation at atomic resolution of a large number of structures of the sarcoplasmic Ca2+ATPase (SERCA1a) present in skeletal muscle. The structures corresponding to the various intermediary states have been obtained after stabilization with structural analogues of ATP and of metal fluorides as mimicks of inorganic phosphate. From these results it is possible, in accordance with previous biochemical and molecular biology data, to give a detailed structural description of both ATP hydrolysis and Ca2+ transport through the membrane, to serve as the starting point for a fuller understanding of the pump mechanism and, in future studies, on the regulatory role of this ubiquitous intracellular Ca2+-ATPase in cellular Ca2+ metabolism in normal and pathological conditions. Key words: X-ray diffraction, Sarcoplasmic reticulum, Ca2+-ATPase, Ca2+ transport, Metal fluorides, P-type ATPases

1. Introduction Studies on the calcium pump present in skeletal muscle (SERCA1a) can be traced back to the demonstration by Marsh (1) of a socalled relaxation factor, remaining as a particulate fraction in the supernatant of a muscle homogenate after removal of the fibrils by low-speed centrifugation. The fraction had the remarkable property of temporarily inducing relaxation by a process that could be attributed to the removal of Ca2+ from the suspension medium Jean-Jacques Lacapère (ed.), Membrane Protein Structure Determination: Methods and Protocols, Methods in Molecular Biology, vol. 654, DOI 10.1007/978-1-60761-762-4_7, © Springer Science+Business Media, LLC 2010

119

120

Møller et al.

both when added to a contractile actomyosin preparation and to an ion permeabilized muscle fibre (2, 3). Further studies carried out by a number of investigators in the 1960s identified the factor as the sarcoplasmic reticulum (SR), an organelle central to the removal and release of Ca2+ during the relaxation/contraction cycle. The characteristics of the Ca2+ uptake process and its ATP dependence were described by Ebashi and Lipmann (4). Rather than being due to binding it was demonstrated by Hasselbach and Makinose that Ca2+ is accumulated inside the SR by an ATP requiring transport in isolated membrane preparations (5). The relation of the Ca2+-transporting ATPase to that of Na+,K+ATPase became clear with the realization that both pumps utilize a common acid-stable aspartylphosphorylated intermediate to perform active transport (6). The important physiological function of the sarcoplasmic 2+ Ca -ATPase in the relaxation/contraction cycle and easy access to purification in large yield, soon made this enzyme a favourite for fundamental studies on the nature of active transport mechanisms. In the following decades (1970–2000), hundreds, even thousands of papers were devoted to functional and structural studies of the enzyme. Concurrent with developments in the study of other P-type ATPases, in particular of Na+,K+-ATPase this has led to the formulation of different reaction schemes. Among these, the scheme originally proposed for Ca2+-ATPase by de Meis and Vianna (7) has gained general acceptance. Figure 1 shows a present day version of the scheme on which (8) the functional interpretation of our structural data are based. In the initial reaction a Ca2+-dependent phosphorylation by ATP leads to the formation of a high energy Ca2E1~P intermediate,

Fig. 1. Reaction scheme of the sarcoplasmic Ca2+-ATPase, showing binding of Ca2+ to the E1 conformations on the cytosolic side (upper part of the scheme), and release of Ca2+, in exchange with protons, to the E2P conformation on the lower part of the scheme. , transition states, with occluded Ca2+ and H+ shown in brackets; (K+) refers to the steps accelerated by binding of K+ to a modulatory binding site near the P-domain.

What Can Be Learned About the Function of a Single Protein

121

phosphorylated at Asp 351. This is followed by a conformational change of the protein to an E2P state during which two Ca2+ ions, bound inside the membrane, are released towards the SR lumen. This occurs in exchange for 2–3 H+ to form an (Hn)E2-P intermediate (9–11) in which form protons in a backward transport reaction are transferred from the luminal to the cytoplasmic side of the SR membrane. This translocation is accompanied by dephosphorylation of the protein to form an E2 conformation with the intramembranous binding sites oriented towards the cytoplasmic side. In between the E1/E1P and E2P/E2 forms the protein is present in transition states, marked by stars with occluded Ca2+ and H+, shown in brackets, barred from contact with both the cytosolic and intralumenal medium. For the kinetic analysis more elaborate and branched reaction schemes than that shown in Fig. 1 have been formulated (e.g., refs. 8, 12, 13) that take into account other factors such as the effect of stepwise association of Ca2+ with the Ca2+-ATPase, the modulatory effect of ATP, the effect of protein–protein interactions (8) and of K+ on the E2 reactions (14, 15), and a number of other environmental variables like pH, inhibition by intra cellularly accumulated Ca2+ and Ca2+/Ca2+ exchange (16, 17). However, with respect to ATP the intracellular concentrations in the physiological situation are sufficiently high that, as indicated in the scheme, all intermediates (with the exception of the product of the ATP phosphorylation step, ([Ca2]E1~P-ADP) can be regarded as having ATP bound in a modulatory mode, irrespective of nucleotide level. Thus although a high intracellular ATP level undoubtedly is important for transport efficiency, even severe physiological changes in ATP concentration can be expected to occur without regulatory consequences. A similar statement can be made concerning the effect of intracellular K+ on the potassium-sensitive E2 reactions, since the intracellular level of K+ by far exceeds the dissociation constant for the binding of this cation to the ATPase. Regardless of the inclusion of these and other kinetic aspects, it has to be stated that reaction schemes are by themselves not of much help to explain how transport across a biological membrane occurs. In the present case, it is clear that efficient structurally based mechanisms must exist for the enzyme to be capable of maintaining a precise 2:1 Ca2+/ATP transport stoichiometry over a wide range of cytosolic Ca2+ concentrations and to be able to accumulate Ca2+ against very high (up to 104:1) concentration gradients (18). The Ca2+-ATPase was found early on by electron microscopy of SR vesicles to contain both a large globular and cytosolic exposed head as well as an intramembranous docked part (19–21). Structural aspects such as studied by analytical ultracentrifugation and SAXS on detergent-solubilized Ca2+-ATPase (22, 23), neutron scattering of stacked membranes (24), and structural analysis

122

Møller et al.

of vanadate-stabilized 2D crystals (25–27) confirmed that the 112 kDa Ca2+-ATPase protein has an elongated shape, with a large cytoplasmic head and a narrow membrane inserted tail. Continued studies with 2D-crystals by cryo EM, combined with a theoretical analysis of the amino acid sequence (28, 29) led to a more detailed, but tentative description of the ATPase in terms of three cytoplasmic domains and ten transmembrane helices. Interestingly, studies with fluorescent probes suggested that the phosphorylation site, located in one of the cytoplasmic domains, is at least 50 Å removed from the intramembranous Ca2+ binding sites (30–32). Evidently, chemical energy derived from hydrolysis of ATP had to be mediated from the phosphorylation domain to the intramembranous cation-binding sites via the intervening stalk segment to energize Ca2+ translocation (33). But how this occurred remained a matter of speculation, although at the same time, following the elucidation of the DNA-based amino acid sequence (34, 35), great strides were made by site directed mutagenesis to pinpoint the role of particular residues, resulting in, for example identification of the intramembranous Ca2+ liganding residues (36) and of residues critical for the conformational changes accompanying transport (37). At the turn of the twentieth century a new era was introduced with the publication of the 3D structure of Ca2+-ATPase in the Ca2E1 conformation at atomic (2.6 Å) resolution by Toyoshima et al. (38). The relative ease with which this membrane protein has since then been found to form crystals, has resulted in a multitude of structures at atomic resolution. These, in effect, provide a library of structures of the ATPase in various conformational states with which the transport mechanism can be analyzed in sufficient detail to provide plausible models for the transport mechanism.

2. Use of Structural Analogues to Stabilize Intermediary Transport Steps

Since the essence of protein X-ray crystallography is to deduce 3D structure from the diffraction pattern provided by the ordered arrangement of many unit cells in the crystal, the technique leaves little room for studying the dynamic aspects, apart from those associated with localized vibrational modes of small amplitude (as indicated to some extent by the refined atomic B-factor). But if a series of well-defined and functionally relevant forms of the same protein is available, it becomes possible from these “snapshots” to piece together the main structural changes that the protein undergoes during its functional cycle. The SERCA1a protein is a good case in point. At present (end of year 2010) 29 structures have been deposited in the Brookhaven Protein Data Bank, representative of

What Can Be Learned About the Function of a Single Protein

123

structures of the ATPase during the various stages of transport and in combination with specific ligands and inhibitors. However, it should be noted that to obtain structures of intermediary states, requires these to be stabilized before crystallization. In principle, this can be achieved in one of two ways: (1) by the use of structural analogues, in particular of transition state analogues, that stabilize the protein in the desired intermediary state, and (2) stabilization by the use of mutated protein. For SERCA1a both approaches have been used, but by far most structures have been obtained from nonmutated Ca2+-ATPase prepared from SR of skeletal muscle. Classical P-type ATPase transition state analogues are Cr ATP (39–42) and vanadate (43) that stabilize the Ca2+-ATPase in conformations corresponding to the(Ca2) E1~P and (Hn)E2-P and transition states, with occluded Ca2+ and protons, respectively. However, the use of these compounds has never led to the formation of diffracting crystals. Instead it has turned out that fluoride-metal complexes of Al3+, Be2+, and Mg2+ can be regarded as structural analogues of phosphate (44–48) and used for crystallization of Ca2+-ATPase in a similar way as they have been used for crystallization of other proteins, for example transducin (49, 50), myosin ATPase (51), G-proteins (52), and phosphotransferases (53). Figure 2 shows how tetrahedral BeF3−, ligated to oxygen-containing ligands, mimicks phosphate covalently bound to the carboxyl group of Asp 351 in the E2P (ground state) of Ca2+-ATPase. While tetragonal-bipyrimidal AlF4− with a planar configuration of the four F− surrounding the central Al3+ functions, and two apical oxygen-liganding groups, can mimick the phosphate of the [Ca2]E1~P-ADP and the (Hn) E2-P transition states. Finally, that MgF42− can be used to mimick the inorganic phosphate bound to the enzyme in the E2·Pi state. a

b

c

Fig. 2. (a) Tetrahedral conformation of MgF42−, forming with ATPase a noncovalent intermediate, mimicking the E2·Pi product state. (b) Tetragonal bipyrimidal conformation of AlF4− forming a transition state structure by interaction with Asp 351. (c) Tetrahedral conformation of E2BeF3−, forming with ATPase a covalent intermediate, similar to the E2P ground state.

124

Møller et al.

We prepare crystals of these forms by the hanging drop technique after solubilization by C12E8 of Ca2+-ATPase, isolated from rabbit skeletal muscle, in the presence of metallofluoride. For crystallization we use polyethylene glycol as a precipitant, together with sucrose or glycerol and additives as conformation stabilizing agents (54). Optimization of crystallization is performed with screens where we vary the salt and buffer composition. To obtain E1 conformations of Ca2+-ATPase usually require the presence of a rather high (10 mM) Ca2+ concentrations, while stabilization of E2 usually is performed in the presence of a specific inhibitor, such as thapsigargin. There is also an obligatory requirement for phospholipids which in the present set-up is derived from the endogenous phospholipids present in the C12E8 solubilized Ca2+-ATPase membranes. The crystals obtained are usually plate like, formed by the stacking of planar (bilayer type) layers of protein–phospholipid–detergent complex. Mutant forms of rabbit SERCA1a, with a biotinylation acceptor domain (BAD) tag fused to the C-terminal end of Ca2+-ATPase through a thrombin-sensitive linker, are prepared by heterologous expression in yeast according to established procedures (55, 56). Purification from yeast membranes is performed by solubilization with dodecylmaltoside (DDM) and affinity chromatography, from which the Ca2+-ATPase is released from the column without the tag by proteolytic cleavage with thrombin. This is followed by an HPLC size exclusion chromatography in which DDM is exchanged by C12E8, and finally relipidated with DOPC. Crystallization by the hanging drop technique then proceeds in the same way as for the Ca2+-ATPase prepared from sarcoplasmic-derived Ca2+-ATPase and similar crystal forms are obtained (55).

3. Phosphorylation of Ca2+-ATPase by ATP

The overall structure of SR Ca 2+-ATPase, as illustrated in Fig. 3 of the Ca2E1-AMPPCP structure, is constituted by (1) an intramembranous domain with ten transmembrane helices (M1– M10) that harbour the two Ca2+-binding sites, (2) three cytoplasmic domains, and (3) a connecting stalk region. The cytoplasmic domains comprise the nucleotide binding N-domain, the phosphorylation P-domain with the Asp 351 phosphorylation site, and the actuator (A) domain, which plays a key role in the E1/E2 conformational switches, leading to Ca2+/H+ exchange. The connecting stalk region is formed by the cytosolic extensions of M1–M6 that together with the A-domain is involved in the conformational changes by which bound Ca2+ induces phosphorylation by ATP. Phosphorylation in turn gives rise to the changes that result in Ca2+ translocation and H+ exchange.

What Can Be Learned About the Function of a Single Protein

125

Fig. 3. The structure of Ca2+-ATPase in the Ca2E1~P-AMPPCP conformation, shown in a graphical representation with the N-domain in red, P-domain in blue, and A-domain in yellow. The transmembrane helices and their cytosolic stalk extensions are coloured as follows: M1–M2 purple, M3–M4 green, and M5–M10 in orange. Ligands and central amino acid residues are shown in black. Notice the binding of nucleotide by the N-domain, with the g-phosphate close to Asp 351in the P-domain, while the 181TGES184 dephosphorylation motif is drawn away from the phosphorylation region, together with the A-domain that is located in a notch on the backside between the N- and P-domain.

Structures relevant for different stages of the ATP phosphorylation reaction are depicted in Fig. 4. We start in Fig. 4a by showing the interaction of the protonated E2 intermediate with modulatory-bound ATP (represented by the (H n) E2-AMPPCP structure). This ATPase form via the transitional states depicted in Fig. 4b, c is converted to the phosphorylated [Ca2+]E1~P ADP intermediate (Fig. 4d). During these transitions the adenosyl moiety of ATP remains bound to the N-domain in a preformed notch, stabilized by p–p stacking with Phe 487 and by interactions with Lys 515, Glu 482, Glu 442, and a number of other, mainly hydrophobic, amino acid residues. The ribose-a,bphosphate moiety of the bound nucleotide is localized at the mouth of the binding site, where it is stabilized in a characteristic bent conformation by interaction with Arg 560 and a Mg2+ ion that in the E2 conformation, via a coordinated water molecule, interacts with Glu 439. The g-phosphate protruding from the N-domain interacts with a conserved TGD motif, located at the periphery of the P-domain (57). In this position, the g-phosphate is about 9 Å displaced from the phosphorylation site by the A-domain, out of reach for reaction with the carboxylate group of Asp 351, which is located in a notch in the central

126

Møller et al.

Fig. 4. Structural events taking place at the phosphorylation site during the Ca2+dependent phosphorylation of Ca2+-ATPase by ATP. (a) The structure of E2-AMPPCP(TG), corresponding to the E2·ATP conformation, showing binding of nucleotide to the N-domain (in red) and interaction with the Gly residue of the 625TGD627 motif of the P-domain (in blue), 9 Å removed from Asp 351 (the phosphorylation site). (b) The structure of the Ca2E1-AMPPCP, corresponding to the Ca2E1·ATP intermediate. The figure shows the approach of the g-phosphate of the nucleotide to Asp 351 that occurs as the result of Ca2+ binding in the transmembrane domain and conversion of the ATPase to the E1 conformation. (c) The structure of [Ca2]E1-ADP-AlF4−, representative of the [Ca2] E1~P-ADP transition state, with AlF4− as a mimick of g-phosphate. (d) Structure of [Ca2] E1~P·AMPPN, representative of the [Ca2]E1~P·ADP conformation. Nucleotide is coloured in CPK, green spacefilling refer to bound divalent cation, with 2 Mg2+ bound in c, and one Ca2+ in b replacing the physiologically bound Mg2+ (see text).

part of the P-domain. In the next step, following binding of 2 Ca2+ inside the membraneous domain, there is an E2 → E1 transition which primarily is the result of displacement of the A-domain from its interactions with the phosphorylation site as will be discussed later. This paves the way for the N-domain with bound ATP, by bending of the long S5 helix and the hinge region attaching the N-domain to the P-domain, to move in close contact with the P-domain to form the Ca2E1-ATP and [Ca2]E1-PADP transition state intermediates. These states are in Fig. 4b, c represented by the Ca2E1-AMPPCP and [Ca2]E1-ADP-AlF4− crystalline structures. Considering first the latter structure where

What Can Be Learned About the Function of a Single Protein

127

AlF4− mimicks g-phosphate, the central Al3+ is found to be linearly and equidistantly (~2 Å) interposed between the b-phosphate of ADP and the carboxyl group of Asp 351. This is in accordance with AlF4− mimicking the intermediate formed in the transfer of g-phosphate from ATP to the accepting Asp 351 carboxylate group, according to an SN-2 associative nucleophilic reaction scheme. From this structure the transition state, with unfavourable electrostatic interactions between the negatively charged nucleotide and Asp 351 carboxylate, is seen to be stabilized by the positive charge imparted by the closely positioned Lys 684 residue, and by Mg2+ that coordinates with both the Asp 351 carboxylate and the g-phosphate group, as well as with the carbonyl group of main chain Thr 353 and the Asp 703 carboxylate group. Furthermore, the side chains of Thr 353 and Thr 625 stabilize the b- and g-phosphate as they arrange for phosphorylation of Asp 351.The structure of the environment for the phosphate transfer reaction is in good agreement with deductions previously drawn from studies involving mutation of residues critical for the ATP phosphorylation reaction (58, 59). But in addition to the Mg2+ coordinated with the phosphorylation site in the [Ca2] E1-AlF4−ADP structure, a second Mg2+ ion is also present, bound to the a,b-phosphate of ADP. It can be noted that the second bound Mg2+ ion interacts with the same part of AMPPCP as in the (Hn)E2AMPPCP form, but with different ATPase residues due to movements of the cytosolic domains associated with the E2 → E1 transition. The functional role of the second bound Mg2+ can be seen as an aid to additionally lower the activation energy required for the transfer of the phosphoanhydride bond in ATP to form the aspartylphosphate bond. with Asp 351 and, possibly, also as an aid to the subsequent product removal of nucleotide from the ATPase as MgADP (60). The structure of Ca2E1 with bound AMPPCP (Fig. 4b) is strikingly similar to that of the transition state. However, the g-phosphate of the nucleotide with a slightly longer distance from carboxylate group (3 Å) is in a position where it has not entered into covalent interactions with Asp 351. Furthermore, the Mg2+ bound at the second site is missing and, in addition, Site I is in fact occupied by a Ca2+ ion instead of Mg2+ (61). This is suggested by the fact that the density remains in the absence of Mg2+ in the solvent and on the basis of anomalous difference density at this site. Furthermore, a better correspondence of the B-factor was obtained when calculated on the assumption of Ca2+ binding (61). The replacement of Mg2+ for Ca2+ can be rationalized on the basis of the high (10 mM) Ca2+ concentration present in the crystallization medium Ca2+ that is required for producing crystals with good diffracting properties.

128

Møller et al.

Despite the similarities of the ATPase with bound nucleotide and the transition state structure there are distinct differences in the biochemical properties of the complexes in the native − membraneous states: while the [Ca2]E1-ADP-AlF4 form stably 2+ occludes Ca , this is not the case for ATPase with bound − AMPPCP which also, in contrast to the ADP-AlF4 form, is susceptible to proteolytic degradation and sulfhydryl modification (60). Thus, while the ATPase with bound MgAMPPCP appears to be present in the native membranes in a dynamically fluctuat− ing state, this is not the case for the [Ca2]E1-ADP-AlF4 intermediate. However, the difference between the two forms is much less pronounced when examined in the membraneous state at the same high concentration of Ca2+ as required for crystallization of the E1 forms of Ca2+-ATPase (62). Thus, it appears likely that the lack of difference between the two crystalline forms, at least partly, is attributable to the presence of Ca2+ at the phosphorylation site which we consider to stabilize the ATPase with AMPPCP in a more stable structure, reflecting perhaps the kinetic evidence for the formation of a presteady state complex of ATPase with MgATP, before phosphorylation (63). Of further interest for understanding of the phosphorylation reaction we found in studies on the D351A mutant (which does not hydrolyze ATP) that ATP is bound with the same characteristic bend of the triphosphate moiety as AMPPCP. The presence of a carbon atom bridging the b- and g-phosphate in AMPPCP thus does not affect the conformation of bound ATP. The binding region was indistinguishable from that of the wild type, except for the replacement of Asp 351 with alanine. Furthermore, in agreement with previous studies (64, 65) we confirmed that the mutant binds ATPase with a considerably higher affinity than the wild type, consistent with the abolition in the mutant of opposing negative electrostatic interactions imparted by the negatively charged nucleotide phosphate and phosphorylation site in the wild type ATPase (56). Finally, Fig. 4d shows the structure of the fully phosphorylated Ca2+-ATPase (66). This was obtained by the use of the slowly hydrolyzed nucleotide analogue, AMPPNP, which stabilized the Ca2+-ATPase in the Ca2E1~P-AMPPN nucleotide bound form. The formation of a covalent phosphoaspartyl bond was documented by mass spectrometry, consistent with the electron density shifting towards the Asp 351 side chain in this structure. The remaining part of the nucleotide (AMPPN) is retracted from the aspartylphosphate, but remains enclosed by the P- and N-domain, with only slight changes in the topology of the amino acid residues at the phosphorylation site, indicating that the structure corresponds to that of the ADP-sensitive phosphorylated form of the ATPase ([Ca2]E1~P-ADP) in the reaction scheme in Fig. 1.

What Can Be Learned About the Function of a Single Protein

129

4. Membrane Binding of Ca2+ Concomitant with the E1P transitions described above the two Ca2+ ions, serving as the substrate for transport, are bound inside the membrane domain at sites, commonly designated as Sites I and II, where they are surrounded by a number of intramembranous amino acid residues that form a cagelike network, involving many hydrogen bonds around them (67). Within this network, as can be seen from Fig. 5 the Ca2+ binding is stabilized in tetragonal-bipyrimidal form by coordination with oxygen-containing groups. At site I the equatorial coordination comprises the side chains of Asn 768 (M5), Glu 908 (M8), Thr 799 (M6), and a water molecule, and these groups are adjoined by the carboxylate groups of Glu 771 (M5) and Asp 800 (M6) in the apical positions to complete the bipyrimidal form (67). At Site II Ca2+ is nested inside M4, interacting with main chain Val 304, Ala 305, Ile 307, and with carboxyl Glu 309 by bidentate linkage. Furthermore, Site II is stabilized by the side chains of Asp 800 and Asn 796 in M6 that together with Val 304 and Ile 307 are assembled as a tetragon around the central Ca2+ ion. As shown by the figure the two Ca2+-binding sites are located side by side, linked by the central Asp 800 residue, and with their tetragonal planes tilted with respect to each other and to the plane of the membrane. With both Ca2+-binding sites occupied and the presence of four ligating carboxylate groups an electroneutral cavity is formed. It is important to realize that the cage is not a preformed structure,

Fig. 5. Coordination of the membrane-bound Ca2+ by carboxylic groups and amino acid – side chain and backbone carbonyl groups in the [Ca2]E1-ADP-AlF4 (1T5T) conformation. 2+ At Site I, Ca is hexa-coordinated by N768, E771, E908, and a H2O molecule in equatorial position, and by D800 at proximal- and E771 distal apical positions. At Site II, Ca2+ is hepta-coordinated at V304, D800, N796, and I307, with A305 at distal- and E309 at proximal apical positions, and in the latter case with bidentate coordination.

130

Møller et al.

but the result of the interaction of the ATPase with both Ca2+ ions. It thus differs significantly from the structure of Ca2+-ATPase in the E2 state, from which it is formed by a number of both translational and rotational changes of the Ca2+-associated transmembrane helices to fix the liganding groups in accurate position for optimal interaction with the bound Ca2+. The formation of the Ca2+ cage is the background for the above-described events in the cytosolic domains, leading to phosphorylation of the ATPase by ATP. These are initiated by the displacement of the A-domain which instead of being locked to the P-domain close to the phosphorylation site, by an upward and counterclockwise rotation, becomes bound in a crevice formed between the upper part of the P- and N-domain (60, 68). The altered location of the A-domain exerts a traction and elevation of the M1/M2 helices that blocks a putative cytosolic N-terminal entrance for Ca2+ which presumably is an important background for the occlusion of the bound Ca2+ ions associated with the formation of [Ca2]E1~P-ADP (60). In conclusion, the phosphoryltransfer from ATP to Asp 351 takes place in an environment that is formed between the N- and P-domain in the E1 state. In the absence of nucleotide there is no interaction between the N- and P-domain, as indicated by the open conformation of the Ca2E1 form. Thus, the close association between the N- and P-domain is critically dependent on the binding of nucleotide that glues the N- and P-domain together, and once this has occurred the stage is set for the catalytic transfer of energy-rich phosphate from ATP to Asp 351 and the ensuing reactions leading to active transport.

5. The Translocation Step ((Ca2)E1P to E2P Transition)

In all structures published until recently, the distal (luminally) oriented part of the ten transmembrane segments in the structures were seen to present a formidable hydrophobic barrier which in our view made it difficult to argue for any specific mechanism by which bound Ca2+ is translocated through the membrane (18). On this background, it was of interest when we recently succeeded in solving a structure which qualifies as a candidate for the conformation that the Ca2+-ATPase assumes immediately after the two Ca2+ ions occluded inside the membrane in the [Ca2]E1~P-ADP state have been delivered into the SR lumen (66). It is a structure stabilized by BeF3− which, according to biochemical and spectroscopic data (48), after covalent interaction with the phosphorylation site functions as a structural analogue of phosphate in the E2P ground state. The preparation of E2–BeF3− used to obtain this structure included the presence of EGTA and

What Can Be Learned About the Function of a Single Protein

131

Fig. 6. The mechanism of Ca2+ translocation through the membrane. (a) E2P ground – state as represented by the E2–BeF 3 structure which suggests that release of 2+ the membrane bound Ca occurs through a wide water-filled channel arising from the spreading out of M1–M2 (green) and M3–M4 (purple) membrane segments from the remainder of the membraneous domain (M5–M10, in orange ). (b) Changes in the disposition of the Ca2+ coordinating residues, resulting in exposure of E771, N796, and E309 towards the luminal channel during the E1P to E2P transition. In the E1P and E2P state, Ca2+ coordinating residues are shown as yellow and purple sticks, respectively, based on superimposition of the C-terminal membrane domain with the Ca2+ occluded E1P (1T5T) structure. For further explanation, see text.

a high Mg2+ concentration (50 mM) during crystallization, but absence of any inhibitor like thapsigargin that would be expected to destabilize the formation of the E2P ground state complex (48, 69). Surprisingly, the structure of the E2–BeF3− complex, solved at 2.65 Å resolution, showed that the compact organization of the membranous domain had become changed into a trilobed structure (Fig. 6a) as a result of lateral and rotational movement of the M1/M2 and M3/M4 pairs, relative to the M5–M10 complex which remained intact (66). This resulted in the formation of a funnel-shaped luminal opening, approx 15 Å broad at the luminal base and tapering into approx 4 Å at the apex where it is lined by three of the Ca2+ ligating residues in the E1 structures (Glu 309, Asn 796, and Glu 771). As shown in Fig. 6b this configuration has become possible, mainly because M4 with Glu 309 has moved towards Glu 771, and because the carboxamide group of Asn 796 has flipped sidewards by counterclockwise rotation of M6. The positions of Glu 908 and Asn 768 are essentially unaltered, while Thr 799 and Asp 800 as a result of the M6 rotation have moved away from their Ca2+ ligating position, leaving room for the bound Ca2+ to move to the distal position where the new constellation of Glu 309, Asn 796, and Glu 771 can be conceived of as constituting the basis for binding of Ca2+ with low affinity. In this connection, it is of interest that exactly the same residues on the basis of mutation studies were pointed out as candidates for the luminal proton exchange with two to three

132

Møller et al.

protons during the reaction cycle (70, 71). Thus luminal exposure of these residues can be considered to result in the release of Ca2+ in exchange for luminal protons. To what extent protons have been bound in our structure remains an open question, since one Mg2+ is present, coordinated with the carboxylate groups of Glu 309 and Glu 59 in M1. Two somewhat different structures of E2–BeF3− were concomitantly published by Toyoshima et al. (72), one with thapsigargin bound (at 2.4 Å resolution) and one in the absence of thapsigargin (3.8 Å resolution). Overall, these two structures are much alike and both are without the evidence for the luminal opening, created by dissociation of the intramembranous segments that is a characteristic of our structure. Instead, on the basis of a detailed analysis of the movements of the intramembranous N-terminal segments (M1–M4), Toyoshima et al. propose that in the absence of thapsigargin a narrow slit between these transmembrane helices is formed through which the bound Ca2+ can escape to the lumen. We may characterize this as a gating model for Ca2+ translocation, similar to what has earlier been proposed by the Japanese group as an extrapolation from the compact organizations of the E2Pi product state (73). Whereas in our model luminal Ca2+ release is the result of more extensive conformational changes of the intramembranous domain, leading to direct exposure of Ca2+ to the luminal space. What is the background for these different models? Fundamentally, the two E2–BeF3− structures without bound thapsigargin, E2–BeF3−Olesen and E2–BeF3−Toyoshima, differ with respect to the movements that the A-domain performs during the E1P to E2P transition. While the rotation in E2–BeF3−Toyoshiama is incomplete (90°), compared to the E2 → Ca2E1 transition (108°), it is 120° in E2–BeF3−Olesen, as the result of a movement where the A domain wedges into the N-domain, providing a tightly fitting seal on top of the phosphorylation site. This is a hyperextended state compared with that observed in the Ca2E1 → E2 transition which readily accounts for the luminal opening as being the result of a drag exerted by this movement on the N-terminal transmembrane segments via the A-M2 and A-M3 linkers. Accordingly, the luminally open state in our model represents a strained conformation of the ATPase where the A-domain via the A-M1, A-M2, and A-M3 linkers with the membrane exerts a pull on the M1/M2 and M3/M4 membrane segments decisive for their moving apart from the C-terminal transmembrane segments. In accordance with this view proteolytic cleavage of the A-M3 (74) and A-M2 (75) linkers block the processing of [Ca2]E1~P intermediate. Furthermore, in mutagenesis experiments the length of A-M1 has been found to be very critical, shortening by one or two residues preventing the [Ca2]E1~P → E2P transition (76), whereas lengthening the linker by two glycine residues stabilizes a Ca2+ occluded E2P form of the ATPase (77).

What Can Be Learned About the Function of a Single Protein

6. The Mechanism of E2-P Dephosphorylation

133

Ultimately, Ca2+ transport must be powered by events taking place in the P-domain as the result of the formation of the “energyrich” and thermodynamically unstable aspartylphosphate intermediate. In all E2 conformations published so far the P-domain is seen to have an increased tilt towards the membrane, compared to the E1 conformations. This, together with the liberation of ADP at the [Ca2]E1~P → Ca2E2P transition presumably are decisive factors for permitting the A-domain to move from its elevated position on top of the P-domain in the [Ca2]E1~P-ADP structure to engage in tight contact with the phosphorylation site. The movement of the A-domain seems to occur in two steps, the first one as discussed above leading to opening of the intramembraneous domain to allow passage of Ca2+ from the binding sites to the luminal space. In the second stage the exposed TGES loop of the A-domain is turned from an outward orientation and becomes docked into the phoshorylation site (Fig. 7). This is accompanied by closure of the luminal gate, with protons from the luminal space that have entered into contact with the membrane to neutralize the surplus of negative charge created by the carboxylate groups after the release of Ca2+ (78). The structure of the ATPase in this transition state is best portrayed by the structures of the E2–AlF4− complex that have been published both in the presence (72, 78) and absence (66, 72) of thapsigargin. In these structures, the

Fig. 7. Interaction of the A-domain with the phosphorylation sites in the E2P states. − (a) The E2P ground state, represented by the E2–BeF3 structure. Here the TGES motif is close to the phosphorylation site, but the catalytically active Glu 183 and Thr 181 are turned away from Asp 351. (b) In the (Hn)E2P conformation, as represented by the (Hn) E2–AlF4− structure the carboxyl group of E183 and the hydroxyl group of T181, via a bound water molecule, are lined up for hydrolysis of the aspartylphospho anhydride by an SN-2 associative mechanism.

134

Møller et al.

carbonyl group of Thr 181 and carboxyl group of Glu 183 are seen via an intervening water molecule to form a linear complex with the AlF4− that is coordinated with the carboxyl group of Asp 351. The complex very much resembles that formed with ADP and AlF4− mimicking the ATP phosphorylation reaction, except that now it is the TGES loop of the A domain that via the coordinated water molecule reacts with Asp 351 in lieu of ADP (Fig. 7). From this arrangement we conclude that in the physiological situation dephosphorylation of E2-P occurs by base catalysis as the result of proton extraction from the bound water molecule, in complete accordance with mutational data on the TGES loop related to the dephosphorylation in the E2-P state (79, 80), while luminal protons are required at the ion binding sites in the membrane for the dephosphorylation site to become activated (78). Other mutational data of amino acid residues related to the A-domain that have been shown to lead to enzymatic inactivation suggest that the necessary E1–E2 rotation and interaction between the A- and P-domain, required for docking of the TGES loop is dependent on the formation of two hydrophobic clusters where one is formed from five residues centred around Tyr 122 at the lower end of the A-domain (81). The other hydrophobic cluster is centred on Val 200, located on the top of the A-domain, which interacts with Val 705 and Val 726 of the P-domain (82). The docking of the TGES loop into the phosphorylation site is probably also associated with the formation of a helical bundle by the adjoinment of three of the C-terminal helices of the P-domain and two helical segments, linking the A-domain to the M2 and M3 transmembrane helices (the A-M2 and A-M3 linkers). At the base of this bundle there is a monovalent cation (K+) binding site that stabilizes the association of A-M3 with the P-domain and probably facilitates the dephosphorylation of E2P by the TGES loop, thereby providing a structural explanation for the accelerating effect of K+ on E2-P dephosphorylation known for a number of years (14). Furthermore, the transition state is probably stabilized by hydrogen bond formation between Ser 186 and Glu 439 and by other interactions between the A- and N-domain. After E2-P hydrolysis these interactions weaken, leaving room for the liberated phosphate to escape through a tunnel formed between the A- and N-domain and for modulatory bound nucleotide to interact with the TGD motif at the periphery of the P-domain as illustrated in Fig. 4a (57).

7. The Dephosphorylated E2 and E1 States

After E2-P dephosphorylation, we have almost completed our tour through the Ca2+ transport cycle. It only remains for the Ca2+-ATPase to unload the bound protons towards the cytosolic

What Can Be Learned About the Function of a Single Protein

135

side in exchange for Ca2+. Crystal structures representative of these E2 forms are available, stabilized by thapsigargin (83) or other inhibitors of Ca2+ membrane binding (67, 84). Overall, these structures are very similar to those representing the E2-P transition or E2·Pi product states, except that the A-domain appears to be in a more relaxed state with the TGES loop retracted from its docking interaction with the phosphorylation site (78). However, there is the caveat associated with these crystal structures that according to extensive functional evidence thapsigargin stabilizes the ATPase in an occluded and functionally slightly different (Hn)E2 state (85, 86), and thus the representation of the membrane region probably differs somewhat from that of the genuine E2 form which readily reacts with Ca2+. Despite that the affinity of the thapsigargin stabilized ATPase for nucleotide is reduced (57), there is spectroscopic evidence that the cytosolic domains of Ca2+-ATPase, as opposed to the membrane domain, are only minimally affected by thapsigargin binding (85). This is the reason why we have felt justified to use the (Hn)E2-AMPPCP (Tg) form as the starting point in the description of the ATP phosphorylation of ATPase from (Hn)E2-P·ATP in Fig 7.4a. With respect to the membraneous domain, the internal position of the carboxylic groups in the E2 structures is indeed a good indication of their proton occluded state, since their removal would result in the formation of charged carboxylate groups inside the membrane, an event that would be highly energy requiring. On the other hand, the lability of the “naked” E2 conformations, in the absence of ligand stabilizers, both in the membraneous and detergent-solubilized (87) state would seem to be in accord with the view that E2 forms cover a rather wide dynamic range of slightly different conformations (86). There is thus good reason to consider that E2 represents a labile state in the membrane where some of the carboxylic groups could be spending part of the time close to the cytosolic space to enable the cytosolic release of bound protons in exchange for Ca2+. The intramembranous binding of cytoplasmic Ca2+ to the dephosphorylated enzyme in exchange for protons evokes the profound changes in the structure of the Ca2+-ATPase molecule that we ordinarily refer to as the E2 → Ca2E1 transition (83). How are they brought about? Ultimately, the formation of the Ca2E1 and [Ca2]E1~P states must be the result of changes in the position of the intramembraneous cation liganding residues to fix them in a position optimal for coordination with Ca2+. This involves a reversal of the changes that we saw was associated with the [Ca2]E1~P to E2P transition (Fig. 6b), i.e. there is a clockwise rotation of M6 to bring the three liganding residues of M6 (Asp 800, Asn 796, and Thr 799) in position for binding of Ca2+ at sites I and II. In addition, to adapt the ATPase for interaction with Ca2+ at site II there is an upward displacement of Glu 309,

136

Møller et al.

accompanied by a ~20° inclination of the cytosolic extensions of M4 (57). The latter changes are decisive, because they probably lead to disruption of the interactions between the A-domain and the C-terminal part of the P-domain as well as that they are required for maintaining the integrity of the hydrophobic cluster centred on Tyr 122 involved in holding the A- and P-domain firmly together in the E2 state.1 As a result, the A-domain then becomes free to move away from the position it occupies close to the phosphorylation site in the E2 conformation. Other and presumably weaker interactions between the N- and top of the A-domain are also disrupted, allowing the P-domain and N-domain to move towards an upright position, resulting in the well-known open structure of the Ca2E1 where there is minimal interaction between the cytosolic domains (38). In the physiological situation these events are modified by the binding of ATP which induces the N-domain, via its hinge attachment to the P-domain and by bending of the S5 helix, to close down like a lid on the P-domain to form the pre-steady complex leading to phosphorylation as explained in the beginning of our review (Fig. 4b).

8. Concluding Comments The recent explosion of X-ray structures of the SERCA Ca 2+-ATPase at atomic resolutions, corresponding to the various functional intermediates, has made it possible to give a description of the whole transport cycle at the structural level. The new structures by and large confirm and extend conclusions reached from previous biochemical and biophysical studies as well as investigations on Ca2+-ATPase mutants. Although some aspects are still missing such as the detailed mechanism for the passage of the transported Ca2+ ions and protons through the membrane, the role of modulatory bound ATP etc., this is, of course, a pleasing state of affairs. However, at the more fundamental level we still need to understand not only how, but why transport occurs – what are the forces behind the complementary movements of the Ca2+ATPase domains and polypeptide chains required for the transport processes, and how is ATP hydrolysis in the cytosolic part of the molecule coupled to transport in the membraneous domain?

1 Note that in the originally published structures of E2–AlF4 − (Tg) (78) and E2·MgF42− (73) with thapsigargin bound the carboxyl group of Glu 309 was modelled towards the cytosolic space rather than towards the interior of the membranous domain as found through later and improved refinement. However, this does not exclude the possibility that an outward orientation of carboxyl 309 could be a feature of the genuine E2 structure, in agreement with previous suggestions (83, 88) that Glu 309 serves as a gating residue in the release of protons and binding of Ca2+.

What Can Be Learned About the Function of a Single Protein

137

Single-molecule studies will be of significant importance in this regard. In addition, there are still many challenges ahead concerning the regulation and structures of various Ca2+-ATPase isoforms, and the use of Ca2+-ATPase as a drug target in connection with the treatment of human diseases.

References 1. Marsh BB (1951) A factor modifying muscle fibre synaeresis. Nature 167:1065–1066 2. Bendall JR (1952) Effect of the Marsh factor on the shortening of muscle fibre models in the presence of adenosine triphosphate. Nature 170:1058–1060 3. Bendall JR (1953) Further observations on a factor (The ‘Marsh’ factor) effecting relaxation of ATP-shortened muscle-fibre models and the effect of Ca and Mg ions upon it. J Physiol 121:232–254 4. Ebashi S, Lipmann F (1962) Adenosine triphosphate-linked concentration of calcium ions in a particulate fraction of rabbit muscle. J Cell Biol 14:389–400 5. Hasselbach W, Makinose M (1961) Die Calciumpumpe der “Erschlaffungsgrana” des Muskels und ihre Abhängigheit von der ATPspaltung. Biochem Z 333:518–528 6. Tonomura Y (1972) Muscle proteins, muscle contraction, and cation transport. Univ Tokyo Press, Tokyo 7. de Meis L, Vianna AL (1979) Energy interconversion by the Ca2+-dependent ATPase of the sarcoplasmic reticulum. Annu Rev Biochem 48:275–292 8. Mahaney JE, Thomas DD, Froehlich JP (2004) The time-dependent distribution of phosphorylated intermediates in native sarcoplasmic reticulum Ca2+-ATPase from skeletal muscle is not compatible with a linear kinetic model. Biochemistry 43:4400–4416 9. Levy D, Seigneuret M, Bluzat A, Rigaud JL (1990) Evidence for proton countertransport by the sarcoplasmic reticulum Ca2+-ATPase during calcium transport in reconstituted proteoliposomes with low ionic permeability. J Biol Chem 265:19524–19534 10. Cornelius F, Moller JV (1991) Electrogenic pump current of sarcoplasmic reticulum Ca2+ATPase reconstituted at high lipid/protein ratio. FEBS Lett 284:46–50 11. Yu X, Carroll S, Rigaud JL, Inesi G (1993) H+ countertransport and electrogenicity of the sarcoplasmic reticulum Ca2+ pump in reconstituted proteoliposomes. Biophys J 64:1232–1242

12. Lee AG, East JM (2001) What the structure of a calcium pump tells us about its mechanism. Biochem J 356:665–683 13. Mintz E, Guillain F (1997) Ca2+ transport by the sarcoplasmic reticulum ATPase. Biochim Biophys Acta 1318:52–70 14. Shigekawa M, Pearl LJ (1976) Activation of calcium transport in skeletal muscle sarcoplasmic reticulum by monovalent cations. J Biol Chem 251:6947–6952 15. Shigekawa M, Dougherty JP (1978) Reaction mechanism of Ca2+-dependent ATP hydrolysis by skeletal muscle sarcoplasmic reticulum in the absence of added alkali metal salts. II. Kinetic properties of the phosphoenzyme formed at the steady state in high Mg2+ and low Ca2+ concentrations. J Biol Chem 253:1451–1457 16. Gerdes U, Moller JV (1983) The Ca2+ permeability of sarcoplasmic reticulum vesicles. II. Ca2+ efflux in the energized state of the calcium pump. Biochim Biophys Acta 734:191–200 17. Smith WS, Broadbridge R, East JM, Lee AG (2002) Sarcolipin uncouples hydrolysis of ATP from accumulation of Ca2+ by the Ca2+ATPase of skeletal-muscle sarcoplasmic reticulum. Biochem J 361:277–286 18. Møller JV, Nissen P, Sørensen TL-M (2006) X-ray crystallographic structures of sarcoplasmic reticulm Ca2+-ATPases at the atomic level. In: Grisshammer R, Buchanan SK (eds) Structural biology of membrane proteins. RSC Publishing, Cambridge, UK, 288–306 19. Deamer DW, Baskin RJ (1969) Ultrastructure of sarcoplasmic reticulum preparations. J Cell Biol 42:296–307 20. Hardwicke PMD, Green NM (1974) The effect of delipidation on the adenosine triphosphatase of sarcoplasmic reticulum. Electron microscopy and physical properties. Eur J Biochem 42:183–193 21. Hasselbach W, Elfvin LG (1967) Structural and chemical asymmetry of the calciumtransporting membranes of the sarcotubular system as revealed by electron microscopy. J Ultrastruct Res 17:598–622

138

Møller et al.

22. le Maire M, Jorgensen KE, Roigaard-Petersen H, Moller JV (1976) Properties of deoxycholate solubilized sarcoplasmic reticulum Ca2+-ATPase. Biochemistry 15:5805–5812 23. le Maire M, Moller JV, Tardieu A (1981) Shape and thermodynamic parameters of a Ca2+-dependent ATPase. A solution x-ray scattering and sedimentation equilibrium study. J Mol Biol 150:273–296 24. Herbette L, DeFoor P, Fleischer S, Pascolini D, Scarpa A, Blasie JK (1985) The separate profile structures of the functional calcium pump protein and the phospholipid bilayer within isolated sarcoplasmic reticulum membranes determined by X-ray and neutron diffraction. Biochim Biophys Acta 817:103–122 25. Taylor KA, Dux L, Martonosi A (1986) Three-dimensional reconstruction of negatively stained crystals of the Ca2+-ATPase from muscle sarcoplasmic reticulum. J Mol Biol 187:417–427 26. Stokes DL, Green NM (1990) Threedimensional crystals of CaATPase from sarcoplasmic reticulum. Symmetry and molecular packing. Biophys J 57:1–14 27. Stokes DL, Green NM (1990) Structure of CaATPase: electron microscopy of frozenhydrated crystals at 6 A resolution in projection. J Mol Biol 213:529–538 28. Toyoshima C, Sasabe H, Stokes DL (1993) Three-dimensional cryo-electron microscopy of the calcium ion pump in the sarcoplasmic reticulum membrane. Nature 362:467–471 29. Zhang P, Toyoshima C, Yonekura K, Green NM, Stokes DL (1998) Structure of the calcium pump from sarcoplasmic reticulum at 8-A resolution. Nature 392:835–839 30. Gutierrez-Merino C, Munkonge F, Mata AM, East JM, Levinson BL, Napier RM, Lee AG (1987) The position of the ATP binding site on the (Ca2+ + Mg2+)-ATPase. Biochim Biophys Acta 897:207–216 31. Bigelow DJ, Inesi G (1992) Contributions of chemical derivatization and spectroscopic studies to the characterization of the Ca2+ transport ATPase of sarcoplasmic reticulum. Biochim Biophys Acta 1113:323–338 32. Baker KJ, East JM, Lee AG (1994) Localization of the hinge region of the Ca2+-ATPase of sarcoplasmic reticulum using resonance energy transfer. Biochim Biophys Acta 1192: 53–60 33. Moller JV, Juul B, le Maire M (1996) Structural organization, ion transport, and energy transduction of P-type ATPases. Biochim Biophys Acta 1286:1–51 34. MacLennan DH, Brandl CJ, Korczak B, Green NM (1985) Amino-acid sequence of a

35.

36.

37.

38.

39.

40.

41.

42.

43. 44.

45.

46.

47.

Ca2+ + Mg2+-dependent ATPase from rabbit muscle sarcoplasmic reticulum, deduced from its complementary DNA sequence. Nature 316:696–700 Brandl CJ, Green NM, Korczak B, MacLennan DH (1986) Two Ca2+ ATPase genes: homologies and mechanistic implications of deduced amino acid sequences. Cell 44:597–607 Clarke DM, Loo TW, Inesi G, MacLennan DH (1989) Location of high affinity Ca2+binding sites within the predicted transmembrane domain of the sarcoplasmic reticulum Ca2+-ATPase. Nature 339:476–478 Andersen JP (1995) Dissection of the functional domains of the sarcoplasmic reticulum Ca2+-ATPase by site-directed mutagenesis. Biosci Rep 15:243–261 Toyoshima C, Nakasako M, Nomura H, Ogawa H (2000) Crystal structure of the calcium pump of sarcoplasmic reticulum at 2.6 A resolution. Nature 405:647–655 Serpersu EH, Kirch U, Schoner W (1982) Demonstration of a stable occluded form of Ca2+ by the use of the chromium complex of ATP in the Ca2+-ATPase of sarcoplasmic reticulum. Eur J Biochem 122:347–354 Coan C, Scales DJ, Murphy AJ (1986) Oligovanadate binding to sarcoplasmic reticulum ATPase. Evidence for substrate analogue behavior. J Biol Chem 261:10394–10403 Vilsen B, Andersen JP (1992) Interdependence of Ca2+ occlusion sites in the unphosphorylated sarcoplasmic reticulum Ca2+-ATPase complex with CrATP. J Biol Chem 267:3539–3550 Cantley LC Jr, Cantley LG, Josephson L (1978) A characterization of vanadate interactions with the (Na, K)-ATPase. Mechanistic and regulatory implications. J Biol Chem 253:7361–7368 Pick U (1982) The interaction of vanadate ions with the Ca-ATPase from sarcoplasmic reticulum. J Biol Chem 257:6111–6119 Murphy AJ, Coll RJ (1992) Fluoride is a slow, tight-binding inhibitor of the calcium ATPase of sarcoplasmic reticulum. J Biol Chem 267:5229–5235 Murphy AJ, Coll RJ (1992) Fluoride binding to the calcium ATPase of sarcoplasmic reti culum converts its transport sites to a low affinity, lumen-facing form. J Biol Chem 267:16990–16994 Murphy AJ, Coll RJ (1993) Formation of a stable inactive complex of the sarcoplasmic reticulum calcium ATPase with magnesium, beryllium, and fluoride. J Biol Chem 268:23307–23310 Troullier A, Girardet JL, Dupont Y (1992) Fluoroaluminate complexes are bifunctional

What Can Be Learned About the Function of a Single Protein analogues of phosphate in sarcoplasmic reticulum Ca2+-ATPase. J Biol Chem 267: 22821–22829 48. Danko S, Yamasaki K, Daiho T, Suzuki H (2004) Distinct natures of beryllium fluoridebound, aluminum fluoride-bound, and magnesium fluoride-bound stable analogues of an ADP-insensitive phosphoenzyme intermediate of sarcoplasmic reticulum Ca2+-ATPase: changes in catalytic and transport sites during phosphoenzyme hydrolysis. J Biol Chem 279:14991–14998 49. Chabre M (1990) Aluminofluoride and beryllofluoride complexes: a new phosphate analogs in enzymology. Trends Biochem Sci 15:6–10 50. Antonny B, Chabre M (1992) Characterization of the aluminum and beryllium fluoride species which activate transducin. Analysis of the binding and dissociation kinetics. J Biol Chem 267:6710–6718 51. Park S, Ajtai K, Burghardt TP (1999) Inhibition of myosin ATPase by metal fluoride complexes. Biochim Biophys Acta 1430: 127–140 52. Sondek J, Lambright DG, Noel JP, Hamm HE, Sigler PB (1994) GTPase mechanism of Gproteins from the 1.7-A crystal structure of transducin alpha-GDP-AlF4-. Nature 372: 276–279 53. Wang W, Cho HS, Kim R, Jancarik J, Yokota H, Nguyen HH, Grigoriev IV, Wemmer DE, Kim SH (2002) Structural characterization of the reaction pathway in phosphoserine phosphatase: crystallographic “snapshots” of intermediate states. J Mol Biol 319:421–431 54. Sorensen TL, Olesen C, Jensen AM, Moller JV, Nissen P (2006) Crystals of sarcoplasmic reticulum Ca2+-ATPase. J Biotechnol 124:704–716 55. Jidenko M, Nielsen RC, Sorensen TL, Moller JV, le Maire M, Nissen P, Jaxel C (2005) Crystallization of a mammalian membrane protein overexpressed in Saccharomyces cerevisiae. Proc Natl Acad Sci USA 102: 11687–11691 56. Marchand A, Lund Winther AM, Holm PJ, Olesen C, Montigny C, Arnou B, Champeil P, Clausen JD, Vilsen B, Andersen JP, Nissen P, Jaxel C, Moller JV, le Maire M (2008) Crystal structure of D351A and P312A mutant forms of the mammalian sarcoplasmic reticulum Ca2+-ATPase reveals key events in phosphorylation and Ca2+ release. J Biol Chem 283:14867–14882 57. Jensen AM, Sorensen TL, Olesen C, Moller JV, Nissen P (2006) Modulatory and catalytic modes of ATP binding by the calcium pump. EMBO J 25:2305–2314

139

58. Clausen JD, McIntosh DB, Woolley DG, Andersen JP (2001) Importance of Thr-353 of the conserved phosphorylation loop of the sarcoplasmic reticulum Ca2+-ATPase in MgATP binding and catalytic activity. J Biol Chem 276:35741–35750 59. McIntosh DB, Clausen JD, Woolley DG, MacLennan DH, Vilsen B, Andersen JP (2004) Roles of conserved P domain residues and Mg2+ in ATP binding in the ground and Ca2+-activated states of sarcoplasmic reticulum Ca2+-ATPase. J Biol Chem 279:32515–32523 60. Sorensen TL, Moller JV, Nissen P (2004) Phosphoryl transfer and calcium ion occlusion in the calcium pump. Science 304:1672–1675 61. Picard M, Jensen AM, Sorensen TL, Champeil P, Moller JV, Nissen P (2007) Ca2+versus Mg2+ coordination at the nucleotide-binding site of the sarcoplasmic reticulum Ca2+-ATPase. J Mol Biol 368:1–7 62. Picard M, Toyoshima C, Champeil P (2005) The average conformation at micromolar (Ca2+) of Ca2+-Atpase with bound nucleotide differs from that adopted with the transition state analog ADP.AlFx or with AMPPCP under crystallization conditions at millimolar (Ca2+). J Biol Chem 280:18745–18754 63. Pickart CM, Jencks WP (1982) Slow dissociation of ATP from the calcium ATPase. J Biol Chem 257:5319–5322 64. Pedersen PA, Rasmussen JH, Jorgensen PL (1996) Consequences of mutations to the phosphorylation site of the alpha-subunit of Na, K-ATPase for ATP binding and E1-E2 conformational equilibrium. Biochemistry 35:16085–16093 65. McIntosh DB, Woolley DG, MacLennan DH, Vilsen B, Andersen JP (1999) Interaction of nucleotides with Asp(351) and the conserved phosphorylation loop of sarcoplasmic reticulum Ca2+-ATPase. J Biol Chem 274:25227–25236 66. Olesen C, Picard M, Winther AM, Gyrup C, Morth JP, Oxvig C, Moller JV, Nissen P (2007) The structural basis of calcium transport by the calcium pump. Nature 450:1036–1042 67. Obara K, Miyashita N, Xu C, Toyoshima I, Sugita Y, Inesi G, Toyoshima C (2005) Structural role of countertransport revealed in Ca2+ pump crystal structure in the absence of Ca2+. Proc Natl Acad Sci USA 102:14489–14496 68. Toyoshima C, Mizutani T (2004) Crystal structure of the calcium pump with a bound ATP analogue. Nature 430:529–535 69. Seekoe T, Peall S, McIntosh DB (2001) Thapsigargin and dimethyl sulfoxide activate medium P(i)<–>HOH oxygen exchange catalyzed by sarcoplasmic reticulum Ca2+-ATPase. J Biol Chem 276:46737–46744

140

Møller et al.

70. Andersen JP, Vilsen B (1995) Structure-function relationships of cation translocation by Ca2+and Na+, K+-ATPases studied by site-directed mutagenesis. FEBS Lett 359:101–106 71. Vilsen B, Andersen JP (1998) Mutation to the glutamate in the fourth membrane segment of Na+, K+-ATPase and Ca2+-ATPase affects cation binding from both sides of the membrane and destabilizes the occluded enzyme forms. Biochemistry 37:10961–10971 72. Toyoshima C, Norimatsu Y, Iwasawa S, Tsuda T, Ogawa H (2007) How processing of aspartylphosphate is coupled to lumenal gating of the ion pathway in the calcium pump. Proc Natl Acad Sci USA 104:19831–19836 73. Toyoshima C, Nomura H, Tsuda T (2004) Lumenal gating mechanism revealed in calcium pump crystal structures with phosphate analogues. Nature 432:361–368 74. Moller JV, Lenoir G, Marchand C, Montigny C, le Maire M, Toyoshima C, Juul BS, Champeil P (2002) Calcium transport by sarcoplasmic reti culum Ca2+-ATPase. Role of the A domain and its C-terminal link with the transmembrane region. J Biol Chem 277:38647–38659 75. Lenoir G, Picard M, Gauron C, Montigny C, Le Marechal P, Falson P, Le Maire M, Moller JV, Champeil P (2004) Functional properties of sarcoplasmic reticulum Ca2+-ATPase after proteolytic cleavage at Leu119-Lys120, close to the A-domain. J Biol Chem 279:9156–9166 76. Daiho T, Yamasaki K, Wang G, Danko S, Iizuka H, Suzuki H (2003) Deletions of any single residues in Glu40-Ser48 loop connecting a domain and the first transmembrane helix of sarcoplasmic reticulum Ca2+-ATPase result in almost complete inhibition of conformational transition and hydrolysis of phosphoenzyme intermediate. J Biol Chem 278:39197–39204 77. Daiho T, Yamasaki K, Danko S, Suzuki H (2007) Critical role of Glu40-Ser48 loop linking actuator domain and 1st transmembrane helix of Ca2+-ATPase in Ca2+ deocclusion and release from ADP-insensitive phosphoenzyme. J Biol Chem 282:34429–34447 78. Olesen C, Sorensen TL, Nielsen RC, Moller JV, Nissen P (2004) Dephosphorylation of the calcium pump coupled to counterion occlusion. Science 306:2251–2255 79. Clausen JD, Vilsen B, McIntosh DB, Einholm AP, Andersen JP (2004) Glutamate-183 in

80.

81.

82.

83. 84.

85.

86.

87.

88.

the conserved TGES motif of domain A of sarcoplasmic reticulum Ca2+-ATPase assists in catalysis of E2/E2P partial reactions. Proc Natl Acad Sci USA 101:2776–2781 Anthonisen AN, Clausen JD, Andersen JP (2006) Mutational analysis of the conserved TGES loop of sarcoplasmic reticulum Ca2+ATPase. J Biol Chem 281:31572–31582 Yamasaki K, Wang G, Daiho T, Danko S, Suzuki H (2008) Roles of Tyr122-hydrophobic cluster and K+ binding in Ca2+ -releasing process of ADP-insensitive phosphoenzyme of sarcoplasmic reticulum Ca2+ -ATPase. J Biol Chem 283:29144–29155 Wang G, Yamasaki K, Daiho T, Suzuki H (2005) Critical hydrophobic interactions between phosphorylation and actuator domains of Ca2+-ATPase for hydrolysis of phosphorylated intermediate. J Biol Chem 280:26508–26516 Toyoshima C, Nomura H (2002) Structural changes in the calcium pump accompanying the dissociation of calcium. Nature 418:605–611 Takahashi M, Kondou Y, Toyoshima C (2007) Interdomain communication in calcium pump as revealed in the crystal structures with transmembrane inhibitors. Proc Natl Acad Sci USA 104:5800–5805 Montigny C, Picard M, Lenoir G, Gauron C, Toyoshima C, Champeil P (2007) Inhibitors bound to Ca2+-free sarcoplasmic reticulum Ca2+-ATPase lock its transmembrane region but not necessarily its cytosolic region, revealing the flexibility of the loops connecting transmembrane and cytosolic domains. Biochemistry 46:15162–15174 Inesi G, Lewis D, Toyoshima C, Hirata A, de Meis L (2008) Conformational fluctuations of the Ca2+-ATPase in the native membrane environment. Effects of pH, temperature, catalytic substrates, and thapsigargin. J Biol Chem 283:1189–1196 Møller JV, le Maire M, Andersen JP (1986) Uses of non-ionic bile salt detergent in the study of membrane proteins. In: Watts A, de Pont JJHHM (eds) Progress in protein-lipid interactions. Elsevier, Amsterdam Inesi G, Ma H, Lewis D, Xu C (2004) Ca2+ occlusion and gating function of Glu309 in the ADP-fluoroaluminate analog of the Ca2+ATPase phosphoenzyme intermediate. J Biol Chem 279:31629–31637

Chapter 8 Recent Progress in the Structure Determination of GPCRs, a Membrane Protein Family with High Potential as Pharmaceutical Targets Vadim Cherezov, Enrique Abola, and Raymond C. Stevens Abstract G protein-coupled receptors (GPCRs) constitute a highly diverse and ubiquitous family of integral membrane proteins, transmitting signals inside the cells in response to an assortment of disparate extracellular stimuli. Their strategic location on the cell surface and their involvement in crucial cellular and physiological processes turn these receptors into highly important pharmaceutical targets. Recent technological developments aimed at stabilization and crystallization of these receptors have led to significant breakthroughs in GPCR structure determination efforts. One of the successful approaches involved receptor stabilization with the help of a fusion partner combined with crystallization in lipidic cubic phase (LCP). The success of using an LCP matrix for crystallization is generally attributed to the creation of a more native, membrane-like stabilizing environment for GPCRs just prior to nucleation and to the formation of type I crystal lattices, thus generating highly ordered and strongly diffracting crystals. Here we describe protocols for reconstituting purified GPCRs in LCP, performing pre-crystallization assays, setting up crystallization trials in manual mode, detecting crystallization hits, optimizing crystallization conditions, harvesting, and collecting crystallographic data. The protocols provide a sensible framework for approaching crystallization of stabilized GPCRs in LCP, however, as in any crystallization experiment, extensive screening and optimization of crystallization conditions as well as optimization of protein construct and purification steps are required. The process remains risky and these protocols do not necessarily guarantee success. Key words: GPCR, Lipidic cubic phase, Membrane proteins, Fluorescence recovery after photobleaching, Crystallization, Crystallography, Minibeam

1. Introduction G protein-coupled receptors (GPCRs) comprise the largest family of integral membrane proteins in the human genome, with ~800 GPCR family members identified (1). These receptors communicate Jean-Jacques Lacapère (ed.), Membrane Protein Structure Determination: Methods and Protocols, Methods in Molecular Biology, vol. 654, DOI 10.1007/978-1-60761-762-4_8, © Springer Science+Business Media, LLC 2010

141

142

Cherezov, Abola, and Stevens

signals across the plasma membrane in response to a variety of extracellular stimuli, ranging from photons, ions, and small molecules, to peptides and proteins. The signals are amplified and transmitted to downstream effectors inside the cells primarily through coupling to heterotrimeric guanidine nucleotide binding proteins (G proteins). Upon activation, receptors trigger complex cascades of reactions controlling crucial physiological and cellular processes. As such, GPCRs are implicated in multitude of diseases, making them important pharmaceutical targets. In fact, about 40% of drugs on the market act on the G protein-coupled receptors (2). GPCR family members share common architecture of a seven transmembrane a-helical bundle (3). It remains intriguing how such a relatively simple scaffold has evolved to selectively bind to thousands of diverse ligands transmitting signals to dozens of different effectors. The primary questions in the GPCR field include understanding mechanisms of signal transduction and aspects of ligand selectivity and specificity. These and other essential questions can only be answered when a variety of biophysical and biochemical data are combined with high-resolution structures of family members stabilized in different functional states. Despite the great importance of GPCRs and the immense worldwide efforts aimed at obtaining their structures, very little progress has been achieved until recently. Indeed, until late 2007, a high resolution structure of only one highly atypical GPCR member, visual rhodopsin, was known (4). Rhodopsin is unusual in that it is highly abundant in the eye’s retina and contains a covalently bound ligand, 11-cis retinal, keeping the receptor conformationally stable in a fully inactive state. However, all other receptors are expressed in miniscule amounts in native tissues and activated by diffusible ligands. For structural studies, it has therefore become necessary that these proteins be heterologously overexpressed, solubilized, and purified. Human integral membrane proteins, including GPCRs, are notoriously difficult to express and they are highly unstable when solubilized by detergents. What makes GPCRs even more challenging is their observed conformational heterogeneity highlighting dynamic equilibrium between multiple functional states (5), obviously vital for signal transduction; however, a formidable roadblock for structural studies. For example, most GPCRs possess a certain level of basal activity in the apo, or unliganded state. Ligands that increase activity are called agonists, those that bind to the receptor and block it without changing its activity are called antagonists, and ligands that decrease the activity below basal are called inverse agonists. Typically, there is a whole range of intermediate states between completely inactive and fully active receptors, and ligands that induce or stabilize them are called partial

Recent Progress in the Structure Determination of GPCRs

143

inverse agonists and partial agonists. In general, receptors are more stable and less conformationally heterogeneous in the inactive states and become progressively less stable and more heterogeneous as they are converted into more active states. Thus, it is now generally accepted that in order to crystallize a GPCR, it has to be stabilized in a single conformational state. An explosion of technological breakthroughs in the past few years have made it possible to stabilize the conformation and determine the high-resolution structures of several members of the GPCR family. One of the successful approaches involved developing a monoclonal antibody toward a structural epitope on the third intracellular loop of the human b2-adrenergic receptor (b2AR) (6). The antibody/receptor complex was crystallized in bicelles, and the structure was determined at a moderate resolution of 3.4–3.7â•›Å (7). In the second approach, a highly flexible intracellular loop between helices 5 and 6 of the human b2AR was replaced by a compact and stable protein, lysozyme from T4 bacteriophage (T4L) (8). The chimeric receptor (b2AR-T4L) was crystallized in lipidic cubic phase (LCP) and the structure solved to a resolution of 2.4â•›Å (9). The same approach was recently used to obtain the structure of the human adenosine A2A receptor at 2.6â•›Å resolution (10). Finally, the third approach toward receptor stabilization involved a systematic mutagenesis strategy (11, 12). This strategy helped to identify six mutations in turkey b1-adrenergic receptor that increased its melting temperature in detergent solution by 21â•›°C. The thermally stabilized receptor was crystallized by a traditional vapor diffusion technique and the structure was solved to 2.8â•›Å (13). All of these studies were done on receptors with bound ligands, typically inverse agonists or antagonists, locking them in a stable conformation. Additional stabilization was achieved by crystallizing b2AR/Fab complex in bicelles (7), and both T4L fusion receptors, the b2AR-T4L and A2AR-T4L, in the lipidic cubic phase (9, 10, 14). Comparison of these new structures highlights important details related to the plasticity of the GPCR 7TM core and the involvement of the extracellular loops in the ligand binding (15, 16). In addition, the new structures provided superior templates for homology-based modeling of other GPCRs and for performing virtual ligand screening and structure based drug design (17). While new approaches have enabled GPCR structure determination, more remains to be done to address crucial questions centering on mechanisms of signal transduction upon receptor activation. To gain insights into these mechanisms, it is essential to obtain a high-resolution structure of an activated receptor with an agonist bound. Recently, a crystal structure of the bovine opsin bound to a transducin fragment shed some light on the receptor activation mechanism (18). However, the atypical mode of the

144

Cherezov, Abola, and Stevens

rhodopsin activation and the lack of an agonist in the structure prevent generalizations applicable to other GPCRs responding to diffusible ligands. It is conceivable that a similar approach of stabilizing an activated receptor by a G protein, Ga subunit, or by a fragment of Ga will yield a high resolution structure in the near future. Two of the best resolution GPCR structures to date, not including rhodopsin, have been obtained using crystallization in LCP, also referred to as in meso crystallization (19). The success of in meso crystallization can be attributed primarily to two factors: stabilization of highly flexible receptors in a membrane-like LCP environment and formation of type I crystal lattice leading to highly ordered and strongly diffracting crystals. Here we describe methods related to reconstitution of purified receptors in LCP, performing pre-crystallization assays, setting up crystallization trials, harvesting crystals, and acquiring crystallographic data. We recommend the reader to get familiar with more general and detailed protocols for in meso crystallization recently published (20). In this chapter, the emphasis is on working with highly unstable GPCRs, which require pre-crystallization analysis of multiple constructs, extensive crystallization optimization, and crystallographic data collection on multiple microcrystals using a highly collimated synchrotron beam.

2. Materials 2.1. Reconstitution of Protein in LCP

1. Monoolein (1-oleoyl-rac-glycerol) (Sigma or Nu-Chek Prep).

2.1.1. Lipid Mixing

3. Chloroform (HPLC grade, Sigma).

2.1.2. Mixing Lipid with Protein

1. Purified protein solution (see Note 1).

2. Additive lipids (Sigma or Avanti Polar Lipids).

2. Two 100Â€mL gas-tight removable needle syringes without a needle (Hamilton, cat.# 81065). 3. 25 or 50Â€mL gas-tight syringe with a flat tip (point style 3) removable needle (gauge 26s) (Hamilton, cat.# 80265). 4. Syringe coupler made of two gauge 26 removable needles (Hamilton, cat.# 7768-02) and two removable needle (RN) nuts (Hamilton, cat.# 30902) as shown in Fig.Â€1 and described in (21).

2.2. LCP-FRAP Crystallization Pre-screening Assays 2.2.1. Protein Labeling

1. Purified protein solution at 1–5Â€mg/mL concentration. 2. Stock solution of 5Â€ mg/mL Cy3 monofunctional N-hydroxysuccinimide ester (Cy3 NHS ester) (GE Healthcare)

Recent Progress in the Structure Determination of GPCRs

145

Fig.1. (a) Attaching the syringe coupler to a gas-tight syringe. After the coupler is attached, the syringe is filled with lipid. Then a second syringe, filled with protein solution, is attached to the other end of the coupler. (b) Section through the syringe coupler.

in dimethylformamide (DMF) (Sigma). Can be stored at −20â•›°C for up to 2Â€weeks (see Note 2). 3. Labeling buffer: 50Â€ mM Hepes pH 7.2, 150Â€ mM NaCl, 0.05%w/v n-dodecyl-b-d-maltopyranoside (DDM), 0.01%w/v cholesterol hemisuccinate (CHS), ligand. 4. Wash buffer: 50Â€mM Hepes pH 7.5, 150Â€mM NaCl, 0.05%w/v DDM, 0.01%w/v CHS, 20Â€mM imidazole, ligand. 5. Elution buffer: 50Â€mM Hepes pH 7.5, 150Â€mM NaCl, 0.05%w/v DDM, 0.01%w/v CHS, 200Â€mM imidazole, ligand. 6. Desalting column PD-10 (GE Healthcare). 7. Ni Sepharose resin (GE Healthcare). 8. Empty 2Â€mL gravity-flow columns (Thermo Scientific). 2.2.2. Sample Setup

1. Glass slides, 127.8â•›×â•›85.5Â€ mm, 1Â€ mm thick (Erie Scientific) (see Note 3). 2. Cover slips, 112â•›×â•›77Â€mm, 0.2Â€mm thick (Erie Scientific) (see Note 3). 3. Perforated double stick spacer (3M 9492MP double stick tape, 60Â€ mm thick, cut to 112â•›×â•›77Â€ mm, with 96 punched 7Â€mm in diameter holes making 12 columns and 8 rows with 9Â€ mm distance between the centers of the adjacent holes) (Saunders Corp.) (see Note 3). 4. Screening solutions (see Note 4).

146

Cherezov, Abola, and Stevens

2.2.3. FRAP Data Collection

1. FRAP station consisting of a Zeiss AxioImager A1 fluorescent microscope with an EC-Plan 10× objective (NAâ•›= â•›0 .3), an HBO 100 epi-illumination, and a Cy3 fluorescence filter set (excitation at 543Â€ nm with 22Â€ nm bandwidth, emission at 605Â€nm with 60Â€nm bandwidth); a tunable dye cell (set at 551Â€ nm) MicroPoint laser system (Photonic Instruments); a CoolSnap HQ2, 14 Bit, cooled (−30â•›°C) CCD (1,392â•›×â•›1,040Â€ pixels, 6.45Â€ µm/pixel) monochrome FireWire camera (Photometrics); an XYZ automated microscope stage MS-2000 and an automated shutter with controller (Applied Scientific Instrumentation). All the FRAP station hardware is controlled by an ImagePro Advance Microscopy program suite (Media Cybernetics) (see Note 5).

2.3. Crystallization in Lipidic Mesophases

1. Protein reconstituted in LCP at 10–30Â€mg/mL final concentration prepared as described in SubheadingÂ€3.1. 2. 10Â€mL gas-tight removable needle syringes without a needle (Hamilton, cat.# 80065). 3. Repetitive syringe dispenser (Hamilton, cat.# 83700) (see Note 6). 4. Short (0.375-in.), flat-tipped needle (point style 3, gauge 26, Hamilton, cat.# 7804-03). 5. Glass slides, 127.8â•›×â•›85.5Â€ mm, 1Â€ mm thick (Erie Scientific) (see Note 3). 6. Cover slips, 112â•›×â•›77Â€ mm, 0.2Â€ mm thick (Erie Scientific) (see Note 3). 7. Perforated double stick spacer (3M 9500PC double stick tape, 140Â€ mm thick, cut to 112â•›×â•›77Â€ mm, with 96 punched 5Â€mm in diameter holes making 12 columns and 8 rows with 9Â€ mm distance between the centers of the adjacent holes) (Saunders Corp.) (see Note 3). 8. Crystallization screens (see Notes 7 and 8). 9. MiteGen MicroMounts (see Note 9).

3. Methods 3.1. Reconstitution of Protein in LCP

Protocols for pre-crystallization assays and crystallization trials described in this chapter begin with reconstitution of membrane proteins in the lipid bilayer of LCP. The reconstitution is achieved spontaneously upon mechanical mixing of a purified protein in detergent solution with special lipids or lipid mixtures prepared as described in SubheadingÂ€3.1.1.

Recent Progress in the Structure Determination of GPCRs 3.1.1. Lipid Mixing

147

The procedure described here can be used to supplement monoolein or any other LCP host lipid with lipophilic additives. Such additives can modulate properties of LCP or can preferentially interact with proteins reconstituted in LCP and therefore are useful for optimization of crystallization conditions. Cholesterol was found to be the best lipid additive significantly improving the b2AR-T4L crystal size (9, 14). Cholesterol was also used for crystallization of adenosine A2A receptor (10). Compatibility of monoolein-based LCP with several lipid-like molecules was reported in (22). 1. Weigh a small amount (few mg) of an additive lipid in a small (1–2Â€mL) amber glass vial with a Teflon-lined cap. 2. Add an appropriate amount of monoolein to obtain required concentration of the additive lipid (see Note 10). 3. Dissolve the lipids in ~200–400Â€mL of chloroform. 4. Evaporate bulk of the solvent using a gentle stream of dry, filtered nitrogen, keeping the vial warm (~37â•›°C) to prevent the lipid from freezing. 5. Remove the last traces of chloroform under a vacuum (<100Â€mTorr) for at least 4Â€h (preferably overnight). 6. Flush the vial with argon gas, close the cap, and store at −20â•›°C or lower temperature until used.

3.1.2. Mixing Lipid with Protein

1. Transfer a necessary amount (15–50Â€ mg) of a host cubic phase lipid (e.g., monoolein) or a lipid mixture prepared in SubheadingÂ€3.1.1 into a small 0.5Â€mL plastic vial. 2. Melt lipid at ~40â•›°C (see Note 11). 3. Weigh a 100Â€ mL gas-tight syringe with attached coupler (Fig.Â€1a). 4. Remove the plunger and transfer molten lipid into the syringe barrel through the plunger end using an adjustable volume pipette (see Note 12) (Fig.Â€2a). 5. Insert the plunger back and slowly move the lipid up in the syringe to remove any trapped air bubbles. Stop when the lipid reaches the end of the coupler needle (Fig.Â€2b). 6. Weigh the syringe to determine the total mass of the lipid in the syringe. 7. Use a 25 or 50Â€mL syringe with a flat tipped 26s gauge needle to transfer an appropriate amount of the protein solution in the second 100Â€ mL gas-tight syringe to achieve 40Â€ wt% of protein solution in the final mixture (for example, for 30Â€mg of lipid use 20Â€ mL of protein solution) (Fig.Â€ 2c). Avoid trapping air bubbles during the syringe loading.

148

Cherezov, Abola, and Stevens

Fig.Â€2. Sequence of steps during manual set up of in meso crystallization trials. Reconstitution of protein in LCP (a–e) is described in SubheadingÂ€3.1.2. Setting up trials in a glass sandwich plate (f–i) is described in SubheadingÂ€3.3.2.

8. Move the protein solution with the plunger up as far as possible to minimize the air trapping after assembling the syringe mixer (Fig.Â€2d). 9. Screw the syringe containing protein solution to the open end of the coupler attached to the lipid syringe. 10. Move the protein solution and the lipid through the coupler inner needle back and forth from one syringe to another by pushing alternatively on the corresponding plungers until the lipid mesophase in the syringe mixer become homogeneous and transparent (Fig.Â€ 2e). This sequence of motions will mechanically mix the lipid with the protein solution through the action of shearing forces forming inside the narrow coupler needle. Upon mixing, a lipidic cubic phase will form spontaneously and the protein will become inserted into the lipid bilayer of the LCP. Complete mixing requires a few hundred passages and typically takes less than 5Â€min (see Note 13). 3.2. LCP-FRAP Crystallization Pre-screening Assay

LCP-FRAP assay is designed to measure diffusion properties of membrane proteins reconstituted in LCP at a variety of different screening conditions (23). The long-range diffusion of membrane

Recent Progress in the Structure Determination of GPCRs

149

proteins in LCP is essential for successful crystallization; however, the microstructure of LCP imposes spatial constraints on diffusion of large proteins or oligomeric protein aggregates. Our data indicate that one of the primary reasons for failure of the in meso crystallization trials with GPCRs is due to a fast nonspecific protein aggregation. While this event is equivalent to massive precipitation, and is easily recognizable in crystallization trials in solution, it does not produce any visual feedback in LCP, since the size of the non-diffusing stuck protein oligomeric aggregates is well below the optical resolution. It has been found that the aggregation behavior of a protein depends on the particular protein construct, the host lipid, and the additives employed in crystallization trials. The LCP-FRAP protocol, described below, involves labeling the protein of interest with a fluorescent dye, reconstituting the labeled protein into LCP, loading a 96-well glass sandwich plate with LCP and screening solutions, performing an FRAP data acquisition, and analyzing the data. Work on automation of data acquisition and data analysis is in progress. The LCP-FRAP assay is useful for prescreening multiple protein constructs, for assessing the role of ligands and lipid additives, and for identifying precipitants that are non-conducive to protein diffusion. The latter are then excluded from the subsequent crystallization screens. 3.2.1. Protein Labeling

1. Bind the protein to a Ni Sepharose resin in a batch mode (see Note 14). 2. Exchange the protein buffer to labeling buffer (see Note 15). 3. Add Cy3 NHS ester stock solution in DMF to the resin to achieve Dye/Protein molar ratio of 2–5. Mix the resin with the dye by pipetting the slurry up and down repetitively with a 1Â€mL pipette. 4. Incubate at 4â•›°C in the dark for 2–3Â€ h using a gentle rocking. 5. Wash out the bulk of unreacted dye and labeled lipids with five column volumes (CVs) of wash buffer. 6. Incubate the resin overnight in the dark with 5Â€ CVs of wash buffer at 4â•›°C on a gentle rocker to dissociate unreacted dye and labeled lipids that are tightly bound to the protein. 7. Wash with 5–10Â€CVs of wash buffer until the flow-through does not contain any detectable dye. 8. Elute the protein with elution buffer. 9. Concentrate to 1–5Â€ mg/mL using a Vivaspin concentrator with 100Â€kDa cutoff.

150

Cherezov, Abola, and Stevens

10. Measure absorbance at 280Â€nm (A280) and 552Â€nm (A552) and determine the protein labeling percentage using the following equation: e 280 [dye] A552 = , [prot] e 552 ( A280 - k ´ A552 )

where e552â•›=â•›150,000/MÂ€cm is the molar extinction coefficient of Cy3 at 552Â€ nm; e280 is the molar extinction coefficient of the protein at 280Â€nm; and kâ•›=â•›0.08 is the correction coefficient due to absorption of Cy3 at 280Â€nm. The expected labeling percentage is between 2 and 10%. 3.2.2. Sample Setup

1. Mix the labeled protein with a lipid to form an LCP using a syringe mixer as described in SubheadingÂ€3.1.2. 2. Set up samples in a glass sandwich plate with a 60Â€mm thick spacer, either manually (SubheadingÂ€ 3.3.2) or robotically, (24) using FRAP screening solutions instead of crystallization screens. 3. Incubate the plate at 20â•›°C for at least 12Â€h to allow for equilibration of screening solutions with LCP before starting the FRAP measurements.

3.2.3. FRAP Data Collection

The protocol describes data collection using a custom build FRAP station controlled by an ImagePro Advance Microscopy program suite (Fig.Â€3a). Most of the steps are automated using the scripting language of the ImagePro. Similar protocol can be implemented on any FRAP capable microscope system. 1. Place the sample on a microscope stage and focus on a homogeneous area of the sample through a 10× objective. 2. Take 3–5 fluorescence images to capture the pre-bleached state of the sample using a CoolSnap HQ2 cooled to −30â•›°C CCD camera. Reduce the acquisition area to a 501â•›×â•›501 pixels central sensor area (corresponding to the sample area of 325â•›×â•›325Â€mm) to minimize the read-out and image transfer times. 3. Trigger the MicroPoint laser firing 10–20 short 5Â€ns pulses at 20Â€MHz repetition rate (see Note 16). 4. Immediately after triggering the laser, start recording a fast post-bleached sequence of 200 images (100–500Â€ ms exposure time per image) streaming the images as fast as possible into the computer memory. 5. Follow with a slow post-bleached sequence of 50 images, selecting the delay time between images (1–20Â€s) depending on the diffusion rate of the protein. Close the shutter during

Recent Progress in the Structure Determination of GPCRs

151

Fig.Â€3. (a) The LCP-FRAP station. (b) Fluorescence recovery curves obtained for b2AR-T4L reconstituted in monoolein based LCP after 12Â€h incubation in the presence of 0.1Â€M Bis tris propane pH 7.0, 25%(v/v) PEG 400, 5%(v/v) 1,4-butanediol and different concentration of sodium sulfate: 0Â€M (triangles), 0.05Â€M (squares), 0.1Â€M (diamonds), and 0.4Â€M (circles).

the pauses between images to minimize sample bleaching by the incident light. 6. Save all recorded images into a single multiframe 16-bit tiff file for data analysis.

152

Cherezov, Abola, and Stevens

3.2.4. Data Analysis

FRAP data analysis steps described here are implemented using ImagePro (Media Cybernetics), Microsoft Excel, and Prizm (GraphPad); however, other image analysis and curve fitting packages can also be employed to perform similar tasks. 1. Open a multiframe FRAP recovery image sequence tiff file with ImagePro. 2. Locate the frame with the darkest bleached spot. 3. Select a circular region of interest (ROI) around the bleached spot. 4. Select four square ROIs in homogeneous regions of LCP at a distance from the bleached spot to serve as reference signals to compensate for decrease of fluorescence intensity due to bleaching during the image acquisition sequence. 5. Integrate total fluorescence intensities inside the selected ROIs in all frames and transfer all the data into an Excel spreadsheet. 6. Correct for the bleaching and any light intensity fluctuations during the acquisition by dividing the intensity inside the bleached spot by the averaged intensity of the reference squares. 7. Normalize the signal to make the pre-bleached intensity equal to 1 and the intensity of the bleached spot equal to 0. 8. Fit the normalized intensity vs. time (extract exact time of acquiring each frame from the original multiframe tiff file), F(t), with GraphPad Prizm using the following equation (25):

æ 2T ö æ æ 2T ö æ 2T ö ö F (t ) = M ´ exp ç - ÷ ´ ç I 0 ç ÷ + I1 ç ÷ ÷ , è t ø è è t ø è t øø

(1)

where M is the mobile fraction of diffusing molecules, T is the characteristic diffusion time, t is the real time of each recorded frame, I0 and I1 are 0th and 1st orders modified Bessel functions (see Note 17) (Fig.Â€3b). 9. Determine the size of the bleached spot by radially integrating the spot intensity with the ImagePro and fitting the integrated intensity profile with the GraphPad Prizm using a Gaussian shape. Use the half width at half maximum (HWHM) of the Gaussian as a measure of the spot radius, R. 10. Calculate the diffusion coefficient, D, as:

R2 D = 4T

(2)

11. Compare the mobile fractions and diffusion coefficients obtained at different screening solutions.

Recent Progress in the Structure Determination of GPCRs

153

12. Design new crystallization screens based on components facilitating protein diffusion. Exclude conditions for which protein diffusion were not observed from subsequent crystallization trials. If the protein did not diffuse in any of the screened conditions, consider broadening the screening space or trying a new protein construct. 3.2.5. High-Throughput LCP-FRAP

Recording a complete fluorescence recovery curve is time consuming, taking 10–30Â€min per sample depending on the protein diffusion coefficient. This limits the number of samples that can be processed in a reasonable time. We noticed that in the case of GPCRs, the majority of screens result in no measurable recovery of signal. Thus, when a 96-well plate is used to set up FRAP samples, it is possible to sequentially bleach all the samples first and then measure the total recovered intensity for each sample after 30–60Â€ min incubation. This procedure provides protein mobile fractions for all 96 samples within ~2Â€h. Full recovery curves are then recorded only for the samples with significant protein mobility.

3.3. Crystallization in Lipidic Mesophases

In meso crystallization, trials can be performed manually or robotiÂ� cally. Protocols for the manual setup of crystallization are provided in this section; the use of robotics for in meso crystallization is extensively discussed in (24, 26). It is preferable to perform in meso crystallization trials in glass sandwich plates, as described in (24, 27). These plates have excellent optical properties for the detection of very small colorless protein crystals growing in an LCP matrix. At the time of writing, glass sandwich plates were not available commercially and must be assembled from separately ordered glass slides and perforated spacers (SubheadingÂ€3.3.1). Alternatively, most commercial micro-batch, sitting, or hanging drop plates could be used for the manual crystallization setup with the caveat that one may obtain less than optimal conditions for detection of small colorless crystals, primarily due to the scattering of light from a rough boundary between the LCP bolus and precipitant solution (Fig.Â€4a–c). The problem can be circumvented by sandwiching the LCP bolus with a 5Â€mm diameter glass coverslip (Warner Instruments, cat.# W2 64-0700) as shown in Fig.Â€4d, e and used in (28). Crystallization setup starts with mixing protein solution with lipid as described in SubheadingÂ€ 3.1.2. It is advisable to have plates and screening solutions ready before starting protein reconstitution in LCP, and to proceed to crystallization setup (SubheadingÂ€ 3.3.2) immediately after forming protein-laden LCP, because some proteins may not be stable in LCP without added precipitant solution. The whole process of setting up a 96-well plate manually including mixing of protein and lipid takes about 1Â€h. In addition to the description of the manual in meso crystallization setup, this section contains protocols for crystal detection, optimization, and harvesting.

154

Cherezov, Abola, and Stevens

Fig.4. Different ways of setting up crystallization trials in LCP: (a) Microbatch, (b) Sitting drop, (c) Hanging drop, (d) Modified hanging drop, (e) Modified microbatch. In (d and e) LCP, bolus (black) is sandwiched using a small 5Â€mm in diameter glass coverslip in order to improve optical properties and facilitate detection of crystals growing in LCP. The best conditions for detecting small colorless crystals are achieved when crystallization trials are performed in glass sandwich plates. (f–h) show top views of the glass sandwich plates used in our laboratory. Plates (f and g) are based on a standard microscope slide, which are convenient to use for setting up crystallization trials manually. Plate (h) has the SBS footprint and is suitable for both robotic and manual setups.

3.3.1. Assembling Glass Sandwich Plates

1. Silanize both base glass slide and coverslip with AquaSil according to manufacturer’s instructions. 2. Peel off protective liner from one side of a perforated spacer sheet. 3. Attach the spacer to a base glass slide, aligning them along their top and left sides. 4. Wrap the plate into a clean aluminum foil. 5. Use a brayer to apply a uniform pressure on the spacer for ensuring better adhesion.

Recent Progress in the Structure Determination of GPCRs

155

6. Wrap the coverslip in a separate aluminum foil to protect it from dust. 7. The plate is now ready for crystallization setup. Plates can be stored at 20â•›°C for up to several months. 3.3.2. Manual Crystallization Setup

1. Attach a 10Â€ mL gas-tight syringe to a repetitive syringe dispenser. 2. Transfer protein-laden LCP, prepared in SubheadingÂ€ 3.1.2, into the 10Â€ mL syringe affixed to the repetitive dispenser (Fig.Â€2f). 3. Attach a short removable needle to the 10Â€mL syringe. 4. Push the plunger until the cubic phase starts coming out from the needle. Fix the plunger with a gripping nut of the repetitive dispenser. 5. Unwrap a glass sandwich plate (prepared in SubheadingÂ€3.3.1), and place it on a flat surface. 6. Position the syringe needle at the center of the first well ~200–300Â€mm above the surface, and press on the button of the dispenser to deliver 200Â€ nL of the protein laden cubic phase bolus (see Note 18) (Fig.Â€2g). 7. Repeat step 5 to dispense cubic phase into three more wells, forming a 2â•›×â•›2 square. 8. Add 1Â€mL of precipitant solutions on top of the cubic phase boluses in each of the four wells (Fig.Â€2h). 9. Cap the four loaded wells with a 18Â€mm square glass coverslip (Fig.Â€2i). Use a wooden toothpick to press on the coverslip around the wells to properly seal them. 10. Repeat steps 5–8 until the whole plate is filled up. 11. Incubate the plate at 20â•›°C (see Note 19).

3.3.3. Crystal Detection

Typically crystals start appearing in the time frame of between a few hours and a few weeks after setup. Majority of GPCR crystals are first detected between 12Â€ h and 7Â€ days of incubation. Crystallization wells should be inspected periodically either under a microscope or with an automated imager. We use the RockImager line of incubator/imagers from Formulatrix, which is compatible with imaging the glass sandwich plates. In this section, we describe inspection of wells and detection of crystals using a microscope equipped with a polarizer and a rotating analyzer. An automatic imager adds a convenience of taking images of wells at scheduled time points and storing them in a database for later viewing. In difficult cases, however, it is useful to take the plate out of the imager and to inspect it under a microscope using a higher magnification than what is normally achievable with the imager.

156

Cherezov, Abola, and Stevens

Fig.Â€5. Needle-like crystals of the engineered adenosine A2a receptor grown in LCP. (a) Bright-field illumination. (b) With cross-polarizers. (c) Fluorescence. The protein was trace-labeled with Cy 3 NHS. Fluorescence picture in (c) was recorded using excitation at 543Â€nm (22Â€nm bandwidth) and emission at 605 (60Â€nm bandwidth).

1. Inspect each well under a microscope with a 10â•›×â•›objective using a bright-filed illumination and cross-polarizers. Under cross-polarizers, the cubic phase is dark as it is isotropic while most crystals appear as bright objects. Make a note of any such birefringent object or any crystal-like object to detect if they grow up with time (Fig.Â€ 5a, b). Use 40× or higher magnification objective to better see the shape of small objects. Sometimes the whole lipid phase bolus can turn birefringent meaning that it has transformed into a lamellar or hexagonal phase, both of which are unlikely to support growth of membrane protein crystals. Reference (20) provides extensive examples of images one can encounter when following up in meso crystallization trials. 2. When a crystalline object is observed during the initial screening, there are several options to establish whether this object contains protein or not: (a) Set up replicate trials using exactly the same components lacking the protein (see Note 20). (b) Label a small fraction of the protein (typically less than 1%) with a fluorescent dye to see if the generated crystals are fluorescent (29) (see Note 21) (Fig.Â€5c). (c) Use a UV fluorescence microscope to distinguish between fluorescent protein crystals and non-fluorescent salt crystals (see Note 22). (d) If the crystals are relatively large, harvest them and check for diffraction using a minibeam at APS or at other synchrotron sources. 3.3.4. Crystal Optimization

Initial crystal hits in LCP often appear as showers of very small crystals, and crystallization conditions have to be optimized. In general, in meso crystal optimization strategies are similar to optimization strategies for crystallization of macromolecules in solution. One difference is that along with tweaking the precipitant solutions composition and pH, one has the ability to

Recent Progress in the Structure Determination of GPCRs

157

optimize the LCP host lipid and lipid additives. When planning for optimization of in meso trials, it is important to understand how LCP responds to changes in concentration of a given component. For example, salts at high concentration can transform LCP into a hexagonal phase, which is not conducive to crystallization, while some small organic molecules, like 2-methyl-2,4-pentanediol (MPD), can completely dissolve the lipids, usually inducing massive protein precipitation. Effects of soluble and lipid-like components on LCP are described in (22, 30, 31). Other parameters that we found to be critical for optimization of crystal growth in meso are the size and the shape of LCP bolus and the protein concentration. In this section, we describe steps that have proven to be useful for optimization of b2AR-T4L crystals (9). 1. After initial crystal hit is detected and verified to be the protein, initiate the first round of optimization by applying coarse concentration–pH grid screens. As an example, vary PEG 400 concentration between 10 and 40%(v/v) with 5%(v/v) steps, salt concentration between 0 and 500Â€mM with 100Â€mM steps, and buffer pH between 6 and 8 with 0.5 steps. 2. Using the best conditions from Step 1 as a guide, screen for the salt identity using a broad selection of salts as, for example, available in the Salt StockOptions kit from Hampton Research (see Note 23). 3. Optimize the buffer identity and concentration. Try several different buffers with the same pH. Try different buffer concentrations between 50 and 200Â€mM. (see Note 23). 4. Set up fine grid-screens around the best conditions obtained during the first three steps. Useful increment for PEG 400 concentration is 1–2%(v/v); for salt: 20–50Â€mM; for pH: 0.1–0.2. 5. Screen for lipid additives using the best fine grid screens from Step 4. Lipid additives should be mixed with the host LCP lipid as described in SubheadingÂ€3.7.2. A good starting concentration of lipid additives is 5Â€ wt%. For most promising additives, concentration should be optimized (see Note 23). 6. Screen for soluble additive: Commercial additive screens (for example, from Hampton Research or Emerald BioSystems) can be conveniently used for this purpose. Use ten times dilution for initial screening. Optimize concentrations of promising additives (see Note 23). 7. Create fine pH–concentration screens including all components that improve crystallization. 8. Optimize protein concentration by setting up trials with protein concentration varied between 10Â€mg/mL and maximum achievable and using the screens prepared in Step 7 (see Note 23). 9. Optimize the volume of lipidic cubic phase bolus using the screens prepared in Step 7 (see Notes 23 and 24).

158

Cherezov, Abola, and Stevens

3.3.5. Crystal Harvesting

Due to a strong adhesion of the double sticky spacer to glass slides, it is impossible to peel off the cover slip to expose a single well for crystal harvesting. Therefore, a piece of the cover glass should be cut and removed first. When crystals are obtained in commercial plastic trays, in which well opening is a trivial procedure, skip Subheading “Opening a Well in the Glass Sandwich Plate” and proceed directly to crystal harvesting Subheading “Harvesting”.

3.3.5.1. Opening a Well in the Glass Sandwich Plate

1. Place a glass sandwich plate with crystals on a harvesting stereo microscope with a variable zoom. 2. Focus the microscope on a well with crystals, using a low zoom so that the whole well is visible in the field of view (Fig.Â€6a). 3. Score the cover glass in four strokes forming a square inside the well boundaries using a sharp corner of a Hampton Research ceramic capillary cutting stone (Fig.Â€6b). 4. Using a strong sharp point tweezers, press around the scored perimeter to propagate the scratches through the thickness of the cover glass (Fig.Â€6c). 5. Punch two small holes in the cover glass at opposite corners of the square. 6. Inject 2–3Â€mL of precipitant solution into the well through one of the holes to reduce dehydration during the subsequent well opening steps (see Note 25) (Fig.Â€6d). 7. Use an angled sharp needle probe to free one or two edges the glass square (Fig.Â€ 6e) and carefully lift it up (Fig.Â€ 6f). The cubic phase bolus can remain on the bottom of the well or can be lifted up with the glass square piece. In the latter case, flip the glass piece over and place it inside the well. 8. Add an extra 5Â€mL of precipitant solution, supplemented with a cryo-protectant, if necessary, on top of the exposed cubic phase bolus (see Note 26) (Fig.Â€6g, h).

3.3.5.2. Harvesting

1. Increase magnification of the microscope to ~100×, focus on a crystal, and adjust the polarizer and analyzer angles so that the birefringent crystal has a good contrast with the background, while making sure that there is still enough light to see the harvesting loop. 2. Harvest the crystal by scooping it directly from the lipidic cubic phase using a MiteGen MicroMount with a diameter matching the size of the crystal (see Note 27). 3. Immediately plunge the MicroMount with the harvested crystal in a dewar with liquid nitrogen to flash freeze it and then transfer it into a storage or shipping dewar.

Recent Progress in the Structure Determination of GPCRs

159

Fig.Â€6. Sequence of steps illustrating opening of a well in the glass sandwich plate for crystal harvesting (see Subheading “Opening a Well in the Glass Sandwich Plate” for details).

3.4. Data Collection from Microcrystals Using a 10Â€mm Minibeam

The protocol describes data collection strategies from small LCP grown GPCR crystals using a 10Â€mm minibeam on the GM/CA CAT beamlines at the Advance Photon Source (Argonne, IL). In most parts, the protocol can be directly translated to data collection on any other microcrystallography beamline with beam size of 10Â€mm or less.

3.4.1. Alignment of Invisible Crystals with the Minibeam

This part of the protocol was originally designed to align crystals of b2AR-T4L, invisible in a frozen opaque lipidic mesophase, with a 10Â€mm minibeam at the GM/CA CAT beamlines. Recently, an automatic rastering was implemented at the GM/CA CAT to facilitate crystal centering (44); however, the original procedure described here can be useful at other microfocus beamlines. 1. Mount a cryo-loop with a crystal on the beamline goniometer. 2. Center the loop on the rotation axis. 3. Attenuate the beam 20 times (see Note 28). 4. If the area of the loop is smaller than 60â•›×â•›30Â€ mm, then go directly to Step 5. Otherwise, use 50â•›×â•›25 mm slits with a 300Â€mm collimator to scan the whole area of the loop taking 1Â€s exposures with 0.5° oscillation at each point. 5. When diffraction from a protein crystal is observed, switch to a 10Â€ mm minibeam and start scanning within the area of 60â•›×â•›30Â€mm using 10Â€mm steps. 6. After detecting diffraction with the 10Â€mm minibeam, select a strong spot and use the total intensity in that spot to search

160

Cherezov, Abola, and Stevens

for the best diffracting position by moving the crystal in the direction of increasing the intensity initially with 5Â€mm and then with 2Â€mm steps. 7. Rotate the crystals 90°. 8. Scan across the loop in the direction perpendicular to the rotation axis using 10Â€mm steps. 9. When diffraction is detected, fine tune the crystal position as described in Step 6. 3.4.2. Crystallographic Data Collection

Radiation damage limits high resolution data collection from small crystals. Collecting and merging data from multiple crystals is necessary to acquire a full dataset. It is essential to optimize data collection strategy to obtain the best possible results. Here we describe the protocol used to collect high quality data fromâ•›~â•›30â•›×â•›15â•›×â•›5Â€ mm3 b2AR-T4L crystals andâ•›~â•›60â•›×â•›10â•›×â•›5Â€ mm3 crystals of A2AR-T4L. 1. Collect five frames at 0 and 90° from a single crystal to determine the space group, lattice parameters, mosaicity, and resolution. Adjust the sample to detector distance and the oscillation width per frame to achieve the optimal data collection. 2. Estimate the susceptibility of crystals to radiation damage by collecting a sequence of exposures at the same crystal orientation. Determine the absorbed dose at which the total diffracting intensity drops two times and use this number as a guide of maximum dose for planning data collection in Step 3. 3. With a strongly diffracting crystal, collect a full low resolution dataset using an attenuated 10Â€mm minibeam. After collecting the first five frames, run a Strategy as implemented, for example, in HKL2000 (32) or XDS (33) to optimize the starting phi angle and the range of rotation. Choose the beam attenuation and the exposure time taking into account the number of needed exposures and the maximum X-ray dose tolerated by the crystals determined in Step 2 (see Note 29). 4. Collect high-resolution wedges of data from several crystals. Choose the beam attenuation and exposure time to be able to collect at least five to ten frames per crystal with the highest achievable resolution. Typically, we used 1–2Â€s exposure and 1° oscillation per frame with unattenuated beam. 5. After collecting each new wedge, merge it with previously collected data using the low resolution set as a reference for scaling. Discard the wedge of data if it does not scale well and significantly increases the Rmerge factor (see Note 30). 6. Repeat Steps 4 and 5 until a complete dataset at the desired resolution is assembled. Remove the low resolution data from the final scaling step (see Note 31).

Recent Progress in the Structure Determination of GPCRs

161

4. Conclusion The in meso crystallization method was introduced more than a decade ago (34), and since that time, it has proven successful in obtaining high resolution structures of difficult membrane proteins, such as human GPCRs (9, 10, 14). The method, however, has been in limited use mostly due to the difficulties involved in the handling of sticky and viscous, gel-like LCP material. The protocols in this chapter provide directions for approaching in meso crystallization of GPCRs or any other membrane proteins. As with any crystallization experiment, the in meso crystallization approach requires substantial screening and optimization efforts, with no guarantee of success. Additionally, for many membrane proteins, and especially in the case of GPCRs, extensive protein engineering aimed at protein stabilization and elimination of highly flexible and disordered parts is an absolute pre-requisite before entering into crystallization trials. Lastly, although crystallography provides atomic resolution 3-d structures, these static snapshots of preferred stable conformations preclude a full understanding of the dynamic nature of these molecules. With increases in the number of GPCR crystal structures, other complementary approaches, such as NMR and computer modeling, also covered in this volume, are becoming essential for determining the structural basis for ligand specificity and for deciphering the mechanisms of signal transduction.

5. Notes 1. Protein prepared for crystallization trials or pre-crystallization assays should run on an analytical size exclusion chromatography column as a single sharp peak without significant contribution from oligomeric aggregates. Protein solution should not contain excessive amounts of detergent as high concentration of detergents could destroy the lipidic cubic phase (35, 36). If the high detergent concentration creates a problem, try reducing the detergent content as low as possible during purification steps. For example, for His-tagged proteins, use saturated binding to a Ni sepharose resin and elution with a minimal amount of buffer to concentrate the protein without increasing the detergent content, and finally use the largest possible cutoff centrifugal concentrators. 2. Protein can be labeled with a variety of fluorescent dyes, readily available from Invitrogen, GE Healthcare, and other companies. We have tried Fluorescein, Rhodamine 6G, Tetramethylrhodamine, and Cy3. Our preferable choice is Cy3

162

Cherezov, Abola, and Stevens

due to its good solubility, low environmental sensitivity, and suitable bleaching properties. There are two common conjugation chemistries for protein labeling: thiol-reactive and amino-reactive. The advantage of using amino-reactive dyes is that virtually any protein can be labeled and that these dyes impose a minimal disturbance to the protein when attached predominantly to the N-terminus. The drawback of the amino-labeling is that co-purified with the protein free amino group-containing lipids, such as phosphatidylethanolamines, are also labeled. This introduces unwanted background signal, which is difficult to separate from the protein signal. The thiol-reactive dyes do not label lipids; however, they rely on the availability of free cysteins exposed on the protein surface. Additionally, we have found that cystein labeling often destabilizes GPCRs. In this protocol, we describe universal protein labeling at its N-terminus with succinimidyl ester derivatives of fluorescent dyes. We also recommend trying cystein labeling if it does not disturb the properties of the protein. Labeling with thiol-reactive dyes can be performed according to the manufacturer’s instructions. 3. The 96-well plate is designed primarily for compatibility with robotic crystallization setup and imaging (Fig.Â€ 4h). In case of manual operations, it is more convenient to make glass sandwich plates of standard 25â•›×â•›75Â€mm2 microscope slides, spacers with punched holes arranged in 8 rows and 2 columns (Fig.Â€4f ) or 9 rows and 3 columns (Fig.Â€4g), and 18â•›×â•›18Â€mm2 (Fig.Â€4f, h) or 25â•›×â•›25Â€mm2 (Fig.Â€4g) glass cover slips, depending on the wells configuration. 4. Screening solutions for LCP-FRAP should be selected to represent a range of common precipitants. We found that PEG-ion-pH and ion-pH screens with 30%v/v PEG 400 or 20%w/v PEG 4000, variety of salts and pH ranging from 5 to 8 are good starting points in screening for diffusion of GPCRs with and without fused lysozyme. 5. To perform FRAP measurements, we have used a simple and affordable system consisting of a fluorescent microscope with an attached laser. Alternatively, any dedicated commercial FRAP system and most confocal microscopes can be used to run the LCP-FRAP assays that are described in this protocol. 6. The Hamilton syringe dispenser can be modified as described in (37) to decrease the dispensing volume ~three times, from 200 to 70Â€nL. 7. It is convenient to use commercial sparse matrix screens for initial screening. On average, however, about 30% of commercial screen conditions are not compatible with LCP crystallization, since they transform LCP into lamellar or hexagonal phase or completely dissolve the lipid (30).

Recent Progress in the Structure Determination of GPCRs

163

Such conditions can be diluted two times with water to increase the compatibility percentage. Additionally, more specific grid or sparse matrix screens can be prepared following results of the LCP-FRAP assays described in SubheadingÂ€3.2. 8. During in meso crystallization the soluble fraction of LCP is diluted ~50 times due to the vast excess of precipitant (1Â€mL) over the LCP bolus (50Â€nL), thus significantly depleting the ligand in the vicinity of the receptor. Therefore, depending on the ligand binding affinity, solubility, and off-rate, it may be necessary to supplement crystallization screens with the ligand. 9. We prefer MiTeGen MicroMounts over standard nylon loops for harvesting crystals from LCP due to their rigidity and thinner profile, which allow for easier penetration into LCP and for picking minimal amount of lipids along with the crystal. 10. Do not exceed 100–200Â€ mg of total lipid mixture per vial, since a complete removal of solvent will be difficult. Use several vials or larger volume vial for making larger volume stocks of mixed lipids. 11. Melting temperature of monoolein is 37â•›°C. Different lipids can melt at higher or lower temperatures. Adjust the incubation temperature to a few degrees above the lipid melting temperature. Incubation time should not be longer than that necessary to melt the lipid in order to avoid possible degradation. 12. After the molten lipid is taken out from the incubator, it remains liquid at room temperature for a few minutes before it solidifies, allowing for transfers with a pipette. 13. The syringe mixer can warm up due to friction between the lipidic mesophase and the syringe coupler needle. To avoid heating up the sample, limit the mixing rate to ~1Â€stroke/s. A useful trick is to cool down the mixer slightly by putting it for a few seconds on ice or in a refrigerator, which will speed up the process of achieving a homogeneous cubic phase. Do not overcool it, however, as below 18â•›°C, monoolein-based LCP is unstable and can convert into a lamellar crystalline phase, damaging the protein. 14. We provide the protocol for labeling and cleaning proteins with C-terminal His-tag. Procedure for labeling untagged proteins should be modified accordingly to include either ion-exchange or size exclusion chromatography for removing the unreacted dye. 15. The pH of the buffer should be between 7 and 7.5 to label predominately the free N-terminus of the protein. The buffer should not contain free amino groups. 16. The number of laser pulses depends on the dye, filters, laser power, sample thickness, etc. Adjust the number of pulses to achieve ~30–50% bleaching.

164

Cherezov, Abola, and Stevens

17. If the experimental recovery data do not follow Eq.Â€1, it is likely that more than one population of diffusing molecules is present in the sample. The extra signals can originate from different protein oligomeric states or from labeled lipids. If two populations of molecules have substantially (more than an order of magnitude) different diffusion rates, then it is possible to fit the experimental recovery curve with a two component equation and extract both contributions. In practice, when amino-labeling is used, even after an extensive wash in an attempt to remove labeled lipids, there is a residual recovery signal of 5–15% coming from the labeled lipids. The protein diffusion signal can in most cases be extracted using a two component diffusion equation with a fixed characteristic diffusion time for lipid molecules determined from a separate FRAP measurement on a sample containing only labeled lipids (23). 18. If the needle is too far from the glass surface, the cubic phase bolus coming out of the needle as a tube curls back and sticks to the needle. If the needle is too close to the glass surface, the cubic phase bolus balls up and sticks to the needle. Both of these cases result in no or incomplete delivery. The right distance and feeling for accurate dispensing come with practice and is relatively easy to achieve. 19. Crystallization plates can be incubated at different temperatures, however, monoolein-based cubic phase is stable only at temperatures above 18â•›°C (38). Recently, several new lipids were developed specifically for low temperature in meso crystallization (39, 40). It is important to avoid temperature fluctuations during plate incubation and imaging, because fluctuations by just few degrees can induce formation of liquid droplets inside the cubic phase. These droplets scatter light, making it difficult to detect crystals. 20. In such control trials, the appearance of crystals similar to the original can confirm that something else apart from the protein is crystallizing. Control trials resulting in no crystals, however, do not prove that the original crystal was from the protein, since it is difficult to exactly reproduce protein-free conditions. 21. Be aware that sometimes even low percentage labeling can prevent protein from crystallizing. Try different dyes and conjugation chemistries as discussed in Note 2. 22. Before attempting UV imaging, check that the material from which crystallization plates are made is sufficiently transparent to UV light. For example, for glass sandwich plates, we use 1Â€ mm slides made of electroverre glass (Erie Scientific) with transmittance of 40% at 280Â€nm, allowing for efficient excitation of protein’s tryptophans. The coverslips used in

Recent Progress in the Structure Determination of GPCRs

165

these plates are made of 0.2Â€mm borosilicate glass, which is essentially non-transparent to UV light. 23. When the described optimization procedure was applied to initial crystal hits of b2AR-T4L, improvements were achieved almost at each step. Salt in the initial hit, lithium sulfate, was replaced by sodium sulfate as a result of Step 2. Initial Hepes buffer was replaced by bis tris propane, which consistently gave better crystal size and shape, on Step 3. Additions of cholesterol and dioleoylphosphatidylethanolamine have improved crystal size on Step 5. The optimum concentration of cholesterol was found to be 10%(wt/wt). The best soluble additive was identified as 5% of 1,4-butanediol on Step 6. We found that increasing concentration of b2AR-T4L from 20 to 30Â€mg/mL, which was used in the initial screening to 50Â€mg/ mL resulted in larger crystal size on Step 8. Increasing protein concentration even further to 60Â€ mg/mL destabilized the lipidic cubic phase and abolished crystal growth. Finally, we observed that decreasing the volume of cubic phase from 50 to 20Â€nL on Step 9 improved the crystal size even further. 24. The rationale behind adjusting the LCP bolus volumes is that the larger the volume the longer it takes for the precipitants to diffuse in the LCP and establish an equilibrium. The transient gradients forming during this process may affect the nucleation and crystal growth rates. An in meso crystallization robot (24) can be used to vary lipidic cubic phase volume between 20 and ~500Â€ nL. With the manual setup, the cubic phase volume range is limited to 70–700Â€nL when using the modified syringe dispenser coupled to 10 and 100Â€mL syringes (37). 25. Certain precipitants may transform LCP into a sponge phase (41). It is difficult to harvest crystals grown in the sponge phase using the described technique. The sponge phase has liquid-like properties and is drawn away from the well by the surface tension when the well is opened up. To overcome this problem, the sponge phase can be transformed back into the cubic phase by lowering the concentration of the precipitant. The cubic phase has a gel-like consistency and stays in place when well is opened. For example, the common precipitant, PEG 400, at concentrations above 35%(v/v) can induce sponge phase formation. Lowering PEG 400 concentration to 30%(v/v) on this step will bring LCP back within few minutes. When lowering PEG 400 concentration, keep concentrations of other ingredients unchanged. 26. At this step, it is possible to dissolve the lipid and free the crystals using enzymatic lipid digestion (42), detergent solutions (43), sponge-inducing compounds, (41) or mineral oils. We have observed that any of these treatments were

166

Cherezov, Abola, and Stevens

detrimental to the diffraction quality of harvested crystals of b2AR-T4L and A2AR-T4L. These crystals diffracted the best when harvested directly from the cubic phase. 27. When harvesting a crystal from the cubic phase, try picking up as little lipid as possible to reduce scattering background from lipids during data collection. If the crystal is located deep inside the cubic phase bolus, use a micro-tool or an empty MiTeGen MicroMount to remove excess of the cubic phase from the top and expose the crystal to the surface. Then use a fresh MicroMount to harvest the crystal. 28. Beam attenuation will depend on the flux. Attenuate as much as possible to reduce the radiation damage while still having the ability to observe low resolution diffraction from the crystals. 29. At this step, it is important to collect a full dataset with minimal damage to the crystal. Resolution is not critical as long as the dataset is complete and scales well, thus giving reasonable Rsym. Typical datasets that we collected at this step had resolution ~5–6â•›Å and Rsym <10%. 30. This method relies on a high isomorphicity between crystals. The in meso grown crystals of b2AR-T4L and A2AR-T4L did have this quality. The data rejection rate due to poor merging was less than 20%. 31. To obtain a full dataset for b2AR-T4L in complex with carazolol, we merged data from 27 crystals (9). For b2AR-T4L/ timolol (14) and A2AR-T4L/ZM241385 (10), 11 and 13 crystals were used, respectively. Fewer crystals were required in the latter two cases because of the higher symmetry space group and slightly lower resolution.

Acknowledgements This work was supported in part by the NIH Roadmap Initiative grant P50 GM073197 (JCIMPT), the Protein Structure Initiative grant U54 GM074961 (ATCG3D) and R21 RR025326. The authors acknowledge contributions from colleagues Michael A. Hanson, Wei Liu, Jeffrey Liu, Mark Griffith, Ellen Chien, Veli-Pekka Jaakola, Chris Roth and Peter Kuhn. The authors acknowledge the support of Janet Smith, Robert Fischetti, and the GM/CA-CAT team at the Advanced Photon Source, for assistance in development and use of the minibeam and beamtime. The GM/CA-CAT beamline (23-ID) is supported by the National Cancer Institute (Y1-CO-1020) and the National Institute of General Medical Sciences (Y1-GM-1104).

Recent Progress in the Structure Determination of GPCRs

167

References 1. Fredriksson R, Lagerstrom MC, Lundin LG, Schioth HB (2003) The G-protein-coupled receptors in the human genome form five main families. Phylogenetic analysis, paralogon groups, and fingerprints. Mol Pharmacol 63:1256–1272 2. Rubenstein K (2008) GPCRs: dawn of a new era? Cambridge Healthtech Institute, Needham 3. Jacoby E, Bouhelal R, Gerspacher M, Seuwen K (2006) The 7 TM G-protein-coupled receptor target family. ChemMedChem 1:761–782 4. Palczewski K, Kumasaka T, Hori T, Behnke CA, Motoshima H, Fox BA, Le Trong I, Teller DC, Okada T, Stenkamp RE, Yamamoto M, Miyano M (2000) Crystal structure of rhodopsin: A G protein-coupled receptor. Science 289:739–745 5. Bosier B, Hermans E (2007) Versatility of GPCR recognition by drugs: from biological implications to therapeutic relevance. Trends Pharmacol Sci 228:438–446 6. Day PW, Rasmussen SG, Parnot C, Fung JJ, Masood A, Kobilka TS, Yao XJ, Choi HJ, Weis WI, Rohrer DK, Kobilka BK (2007) A monoclonal antibody for G protein-coupled receptor crystallography. Nat Methods 4: 927–929 7. Rasmussen SG, Choi HJ, Rosenbaum DM, Kobilka TS, Thian FS, Edwards PC, Burghammer M, Ratnala VR, Sanishvili R, Fischetti RF, Schertler GF, Weis WI, Kobilka BK (2007) Crystal structure of the human beta2 adrenergic G-protein-coupled receptor. Nature 450:383–387 8. Rosenbaum DM, Cherezov V, Hanson MA, Rasmussen SG, Thian FS, Kobilka TS, Choi HJ, Yao XJ, Weis WI, Stevens RC, Kobilka BK (2007) GPCR engineering yields high-resolution structural insights into beta2-adrenergic receptor function. Science 318:1266–1273 9. Cherezov V, Rosenbaum DM, Hanson MA, Rasmussen SG, Thian FS, Kobilka TS, Choi HJ, Kuhn P, Weis WI, Kobilka BK, Stevens RC (2007) High-resolution crystal structure of an engineered human beta2-adrenergic G proteincoupled receptor. Science 318:1258–1265 10. Jaakola VP, Griffith MT, Hanson MA, Cherezov V, Chien EY, Lane JR, Ijzerman AP, Stevens RC (2008) The 2.6Â€ Angstrom crystal structure of a human A2A adenosine receptor bound to an antagonist. Science 322:1211–1217 11. Serrano-Vega MJ, Magnani F, Shibata Y, Tate CG (2008) Conformational thermostabilization of the beta1-adrenergic receptor in a

detergent-resistant form. Proc Natl Acad Sci U S A 105:877–882 12. Magnani F, Shibata Y, Serrano-Vega MJ, Tate CG (2008) Co-evolving stability and conformational homogeneity of the human adenosine A2a receptor. Proc Natl Acad Sci U S A 105:10744–10749 13. Warne T, Serrano-Vega MJ, Baker JG, Moukhametzianov R, Edwards PC, Henderson R, Leslie AG, Tate CG, Schertler GF (2008) Structure of a beta1-adrenergic G-protein-coupled receptor. Nature 454: 486–491 14. Hanson MA, Cherezov V, Griffith MT, Roth CB, Jaakola VP, Chien EY, Velasquez J, Kuhn P, Stevens RC (2008) A specific cholesterol binding site is established by the 2.8Â€A structure of the human beta(2)-adrenergic receptor. StructureÂ€16:897–905 15. Katritch V, Cherezov V, Hanson MA, Roth RB, Abagyan R (2009) Analysis of the b2AR structure provides insight into agonist binding and role of the TM5 helix in the activation mechanism. J Mol Recognit 22:307–319 16. Mustafi D, Palczewski K (2009) Topology of class A G protein-coupled receptors: insights gained from crystal structures of rhodopsins, adrenergic and adenosine receptors. Mol Pharmacol 75:1–12 17. Reynolds KA, Katritch V, Abagyan R (2009) Identifying conformational changes of the beta(2) adrenoceptor that enable accurate prediction of ligand/receptor interactions and screening for GPCR modulators. J Comput Aided Mol Des 23:273–288 18. Scheerer P, Park JH, Hildebrand PW, Kim YJ, Krauss N, Choe HW, Hofmann KP, Ernst OP (2009) Crystal structure of opsin in its G-protein-interacting conformation. Nature 455:497–502 19. Caffrey M (2009) Crystallizing membrane proteins for structure determination: use of lipidic mesophases. Annu Rev Biophys 38: 29–51 20. Caffrey M, Cherezov V (2009) Crystallizing membrane proteins using lipidic mesophases. Nat Protocols 4:706–731 21. Cheng A, Hummel B, Qiu H, Caffrey M (1998) A simple mechanical mixer for small viscous lipid-containing samples. Chem Phys Lipids 95:11–21 22. Cherezov V, Clogston J, Misquitta Y, Abdel-Gawad W, Caffrey M (2002) Membrane protein crystallization in meso: lipid type-tailoring of the cubic phase. Biophys J 83:3393–3407

168

Cherezov, Abola, and Stevens

23. Cherezov V, Liu J, Hanson MA, Griffith MT, Stevens RC (2008) LCP-FRAP assay for pre-screening membrane proteins for in meso crystallization. J Cryst Growth Design 8:4307–4315 24. Cherezov V, Peddi A, Muthusubramaniam L, Zheng YF, Caffrey M (2004) A robotic system for crystallizing membrane and soluble proteins in lipidic mesophases. Acta Crystallogr D Biol Crystallogr 60:1795–1807 25. Soumpasis DM (1983) Theoretical analysis of fluorescence photobleaching recovery experiments. Biophys J 41:95–97 26. Cherezov V, Caffrey M (2007) Miniaturization and automation for high-throughput membrane protein crystallization in lipidic mesophases. In: Chayen NE (ed) Protein crystallization strategies for structural genomics. San Diego, International University Line 27. Cherezov V, Caffrey M (2003) Nano-volume plates with excellent optical properties for fast, inexpensive crystallization screening of membrane proteins. J Appl Crystallogr 36: 1372–1377 28. Lunde CS, Rouhani S, Facciotti MT, Glaeser RM (2006) Membrane-protein stability in a phospholipid-based crystallization medium. J Struct Biol 154:223–231 29. Forsythe E, Achari A, Pusey ML (2006) Trace fluorescent labeling for high-throughput crystallography. Acta Crystallogr D Biol Crystallogr 62:339–346 30. Cherezov V, Fersi H, Caffrey M (2001) Crystallization screens: compatibility with the lipidic cubic phase for in meso crystallization of membrane proteins. Biophys J 81:225–242 31. Vargas R, Mateu L, Romero R (2003) The effect of increasing concentrations of precipitating salts used to crystallize proteins on the structure of the lipidic Q224 cubic phase. Chem Phys Lipids 127:103–111 32. Otwinowski Z, Minor W (1997) Processing of X-ray diffraction data collected in oscillation mode. Methods Enzymol 276:307–326 33. Kabsch W (1993) Automatic processing of rotation diffraction data from crystals of initially unknown symmetry and cell constants. J Appl Crystallogr 26:795–800 34. Landau EM, Rosenbusch JP (1996) Lipidic cubic phases: a novel concept for the crystal-

lization of membrane proteins. Proc Natl Acad Sci U S A 93:14532–14535 35. Misquitta Y, Caffrey M (2003) Detergents destabilize the cubic phase of monoolein: implications for membrane protein crystallization. Biophys J 85:3084–3096 36. Ai X, Caffrey M (2000) Membrane protein crystallization in lipidic mesophases: detergent effects. Biophys J 79:394–405 37. Cherezov V, Caffrey M (2005) A simple and inexpensive nanoliter-volume dispenser for highly viscous materials used in membrane protein crystallization. J Appl Crystallogr 38: 398–400 38. Qiu H, Caffrey M (2000) The phase diagram of the monoolein/water system: metastabiÂ� lity and equilibrium aspects. Biomaterials 21:223–234 39. Misquitta Y, Cherezov V, Havas F, Patterson S, Mohan JM, Wells AJ, Hart DJ, Caffrey M (2004) Rational design of lipid for membrane protein crystallization. J Struct Biol 148:169–175 40. Yamashita J, Shiono M, Hato M (2008) New lipid family that forms inverted cubic phases in equilibrium with excess water: molecular structure–aqueous phase structure relationship for lipids with 5, 9, 13, 17-tetramethyloctadecyl and 5, 9, 13, 17-tetramethyloctadecanoyl chains. J Phys Chem B 112: 12286–12296 41. Cherezov V, Clogston J, Papiz MZ, Caffrey M (2006) Room to move: crystallizing membrane proteins in swollen lipidic mesophases. J Mol Biol 357:1605–1618 42. Nollert P, Landau EM (1998) Enzymic release of crystals from lipidic cubic phases. Biochem Soc Trans 26:709–713 43. Luecke H, Schobert B, Richter HT, Cartailler JP, Lanyi JK (1999) Structure of bacteriorhodopsin at 1.55Â€ A resolution. J Mol Biol 291:899–911 44. Cherezov V, Hanson MA, Griffith MT, Hilgart MC, Sanishvili R, Nagarajan, V, Stepanov, S, Fischetti RF, Kuhn P, Stevens R (2009) Rastering strategy for screening and centering of microcrystal samples of human membrane proteins with a sub-10 micron size X-ray synchrotoron beam. J R Soc Interface 6 (Suppl 5): S587–S597

as

Part III Electron Microscopy

Chapter 9 Observation of Membrane Proteins In Situ: AQPcic, the Insect Aquaporin Example Daniel Thomas and Annie Cavalier Abstract Aquaporins are water-selective channels widely distributed in prokaryotes, plants, and animals. Looking for the presence of a water channel in the filter chamber (FC) of a homopteran insect (Cicadella viridis), we conducted an electron microscopic study. On thin sections, FC displays thin epithelia with developed basal membrane folds (BMFs). Freeze fracture performed on FC shows an amazing network of intramembrane particles. Epithelial cell membranes were purified and observed by negative staining for control purity. Membrane solubilisation followed by PAGE showed that a 25-kDa polypeptide (P25) is the major protein constituent. Using a specific antibody, we located P25 on thin sections on the microvilli and on BMFs of the epithelial cells. Immunogold localisation of P25 on negatively stained membranes and examination of Pt/C shadowed membranes demonstrated that P25 has an asymmetric insertion within the membrane. cDNA cloning and heterologous expression confirmed that P25 is an aquaporin; thus, we called it AQPcic. The native state of crystallisation of this aquaporin in the membrane appeared to be unique and favourable for a structural investigation by negative staining, cryo-electron microscopy, and image processing. We demonstrated that, in the native membrane, AQPcic is a homotetramer forming a regular two-dimensional array. Key words: Aquaporin, Membrane proteins, Insect, Electron microscopy, Thin sections, Freeze fracture, Metal shadowing, Negative staining, Immunogold labelling, Cryo-electron microscopy, Image processing

1. Introduction Water is the key solvent for the chemical process of life and its movement across the plasma membrane accompanies essential physiological functions. Water crosses the plasma membranes of most cells by diffusion through the lipid bilayer; however, some cells exhibit high water permeability due to water-selective membrane proteins now identified as aquaporins. These membrane proteins form transmembrane channels and are found Jean-Jacques Lacapère (ed.), Membrane Protein Structure Determination: Methods and Protocols, Methods in Molecular Biology, vol. 654, DOI 10.1007/978-1-60761-762-4_9, © Springer Science+Business Media, LLC 2010

171

172

Thomas and Cavalier

Fig.Â€1. Thin section on the filter chamber of Cicadella viridis. The epithelium of the filter chamber is extremely thin. The membrane surfaces are amplified by the presence of apical microvilli (Mi) and basal membrane folds (BMF), which reach the apical region of the cell. Scale bar represents 4Â€µm.

in bacteria, plants, and animals. In an early investigation on an epithelial complex found in the digestive tract of some homopteran insects feeding on plant sap, Gouranton (1968) made the assumption that, in this complex called the “filter chamber” (FC), a significant water transfer occurs down a transepithelial osmotic gradient (1). Looking for the presence of a water channel, we conducted a biochemical and structural study on the FC membranes. At the EM level, the FC displays very thin epithelia with extraordinarily developed basal membrane folds (BMFs) (Fig.Â€ 1). We performed freeze fracture on the FC of Cicadella viridis and observed that epithelial cell membranes exhibit an amazing network of intramembrane particles that covers the whole surface of the membrane (Fig.Â€2) (2). To characterise such particles, we purified the epithelial cell membranes and checked their purity by negative staining (2); some of them, favourably flattened, display a faint regular square array of particles (Fig.Â€3). After membrane solubilisation by detergents and PAGE, we showed that their major constituent is a 25-kDa hydrophobic polypeptide (P25) (2). As a result of its high representation in the membranes, it appeared likely that this polypeptide takes an important part in the constitution of the regular array and should be involved in the water transport function. We hypothesised that P25 could be a water channel and thus belongs to the aquaporin family. Using a specific antibody, we located this putative aquaporin on thin sections, on the microvilli, and on BMFs of the epithelial cells (Fig.Â€4). Immunogold localisation of P25 on

Observation of Membrane Proteins In Situ: AQPcic, the Insect Aquaporin Example

173

Fig.Â€2. Freeze-fracture preparation carried out on the filter chamber of Cicadella viridis. The folded cell membrane reveals a regular arrangement of intramembrane particles on the whole surface. Each particle has a size of about 9Â€nm. Scale bar represents 0.2Â€µm.

Fig.Â€3. Membrane isolated and purified from the filter chamber observed after negative staining. The membrane displays a regular two-dimensional array of particles. Inset shows the power spectrum calculated from this area. The unit cell dimensions are 9.6â•›×â•›9.6Â€nm; 90°. Scale bar represents 0.15Â€µm.

purified negatively stained membranes (Fig.Â€5) and examination of Pt/C shadowed membranes (Fig.Â€ 6) demonstrated that P25 has an asymmetric insertion within the membrane. An experimental approach, using cDNA cloning, overexpression, and functional study in Xenopus oocyte led us to confirm that the

174

Thomas and Cavalier

Fig.Â€4. Immunogold localisation of AQPcic on ultra-thin sections of the filter chamber. Gold particles are abundant on microvilli (Mi) and on basal membrane folds (BMF). Scale bar represents 1Â€µm.

Fig.Â€5. Immunogold localisation of AQPcic on purified negatively stained membranes. Numerous gold particles are located only in clearly delineated zones corresponding to one side of the membrane, suggesting an asymmetrical structure of the membrane. Scale bar represents 0.25Â€µm.

P25 protein is an insect member of the aquaporin family (3). We consequently called it AQPcic (for AQuaPorin of Cicadella). The native state of crystallisation of this aquaporin in the FC

Observation of Membrane Proteins In Situ: AQPcic, the Insect Aquaporin Example

175

Fig.Â€ 6. Freeze-dried, shadowed membrane preparation from the filter chamber. This folded membrane displays both sides: a smooth and a rough surface. Therefore, the membrane has an asymmetrical structure as suggested by the immunolabelling experiment. Scale bar represents 0.3Â€µm.

Fig.Â€7. Cryo-electron microscopy on purified membrane embedded in vitreous ice. Within ice, isolated membranes appear mostly as tubes, vesicles, and folded sheets (inset). In some favourable cases, patches of membranes display a faint regular array. Scale bar represents 0.1 and 0.2Â€µm for inset.

membrane was very favourable for a structural investigation since native membranes can be used directly for negative staining (Fig.Â€3) and cryo-electron microscopy (Fig.Â€7) followed by image

176

Thomas and Cavalier

Fig.Â€8. Two-dimensional projection map of a membrane protein tetramer. The resulting image corresponds to the average of 680 units cells from a single membrane crystal embedded in ice. Fourfold symmetry was imposed. Scale bar represents 2.5Â€nm.

processing (Fig.Â€ 8). By combining this structural information with mass data provided by scanning transmission electron microscopy (STEM) in dark-field mode (4), we were able to conclude that AQPcic exists in the native membrane as a homotetramer forming a regular two-dimensional array (5).

2. Materials 2.1. Animals

Cicadella viridis L. (Homoptera, Jassidae) was harvested from wheat meadows in Brittany from summer to autumn. After dissection, freshly collected FCs were fixed for cytology or homogenised for membrane purification. Membranes were purified over a discontinuous sucrose gradient and then washed in an alkaline buffer (2).

2.2. Products

1. Fixative preparation for immunolabelling: Dissolve 4.0Â€ g of paraformaldehyde powder in 45Â€ mL H2O for 2Â€ h at 65°C, then add 20–80Â€µL of NaOH 1.0Â€M until paraformaldehyde is dissolved (repeat the NaOH step if necessary). Adjust to 50Â€ mL with H2O and then add 50Â€ mL 0.2Â€ M phosphate buffered saline (PBS). Check the final pH, correct with HCl if necessary. Store frozen in aliquots.

Observation of Membrane Proteins In Situ: AQPcic, the Insect Aquaporin Example

177

2. Blocking buffer for immunolabelling: 1% (w/v) Fraction V, bovine serum albumin (BSA) in a standard buffer PBS, pH 7.4. 3. Primary antibody and protein A-gold dilution buffer for immunolabelling: PBS supplemented with 0.5% (w/v) BSA. 4. Protein A conjugated to 10Â€nm colloidal gold. 5. Secondary antibody (10-nm goat anti-rabbit (GAR)-gold). 6. Aqueous uranyl acetate solution: 2% (w/v) in distilled water. Stored at 4°C in darkness. 7. Lead citrate solution: 0.04Â€ g lead citrate in 10Â€mL distilled water containing 100Â€µL of 10Â€mM NaOH. Stored at 4°C. 2.3. Sectioning

2.4. Freezing

Ultra-thin sections were obtained with a diamond knife on an ultramicrotome LKB 2088 Ultrotome V. 1. Freeze fracture apparatus: Cryofract 190 from Reichert-Jung (now Leica Microsystems, Vienna, Austria) JFD-9000 FreezeEtching Equipment from JEOL, Tachikawa, Tokyo, Japan. 2. Fast freezing devices: Homemade “guillotine”, Vitrobot (FEI).

2.5. Observation

Conventional electron microscopy was carried out on a Philips CM12 microscope operating at 100Â€kV.

2.5.1. Electron Microscopes

Cryo-electron microscopy was carried out on a Philips CM12 microscope operating at 100Â€kV and on a Tecnai G2 Sphera (FEI) operating at 200Â€ kV. During observation, frozen samples were maintained below −160°C using a Gatan cryoholder.

3. Methods 3.1. Ultrastructural Cytology

1. Chemical fixation: FCs were prefixed by immersion in 2.5% glutaraldehyde in 0.1Â€M PBS, pH 7.4 for 30Â€min to 1Â€h at 4°C (see Note 1). After washing in PBS, the specimens were post-fixed in 2% (w/v) osmium tetroxide solution in PBS for 1Â€h at 4°C (see Note 2). 2. Resin infiltration and embedding: Fixed samples were washed three times with the same buffer, dehydrated in a graded series of acetone (70, 90, and 100%) (v/v), 20Â€min each and then embedded in an epon–araldite mixture. 3. Polymerisation: The specimens were placed in the adequate holder filled with epon–araldite mixture and placed into the polymerisation chamber at 60°C. 4. Sectioning: Ultra-thin sections, (grey gold) of about £100Â€nm thickness, were cut on an ultramicrotome using a diamond knife oriented at 45° inclination.

178

Thomas and Cavalier

5. Staining: Resin epoxy sections were stained with 5% aqueous uranyl acetate followed by 10Â€ min lead citrate (6) at room temperature in a dust-free atmosphere. Example of a thin section on the FC of C. viridis is shown in Fig.Â€ 1. Typical FC cells display a reduced cytoplasm, but have developed microvilli and BMFs. 3.2. Freeze Fracture

Freeze fracture permits to investigate the fine structure of cell surfaces but also inside fractured membranes, thus revealing intramembrane particles. 1. Fixation: FCs were fixed in cold 2.5% glutaraldehyde, pH 7.2, 0.1Â€M cacodylate buffer for 1Â€h. After washing, the tissue was cryoprotected in 30% glycerol in buffer overnight. 2. Freezing: The bulk specimen was mounted between two copper platelets and frozen by quickly plunging them into liquid propane. The frozen specimen was quickly introduced into the high vacuum chamber of the freeze-fracture device and mounted on the pre-cooled specimen stage at −150°C (see Note 3). 3. Fracture and metal evaporation: The sandwich was broken into two pieces and immediately shadowed by evaporating 2Â€nm of platinum carbon at an elevation angle of 45°. The metal thickness was monitored with the quartz crystal thickness monitor. Replicas were reinforced by evaporation of a layer of 20Â€nm of carbon under an elevation angle of 90° (see Note 4). 4. Cleaning: Cleaning of the replica is performed after the transfer of the shadowed specimen out of the preparation unit and specimen thawing. Replicas were cleaned on sodium hypochlorite (household bleach) for 2–12Â€ h and rinsed with several bath of distilled water until the bleach is replaced with distilled water. 5. Replicas were collected on uncoated 100–300 mesh copper grids. FigureÂ€2 shows a freeze-fracture preparation carried out on the FC of C. viridis. The cell membrane exhibits a regular network of intramembrane particles, each particle having a size of 9Â€ nm in diameter.

3.3. Negative Staining

To characterise the intramembrane particles, we purified the epithelial cell membrane and control their purity by negative staining. Negative staining is a simple and quick method that should be applied in first instance to control the purity of a given fraction; moreover, it provides fast medium resolution information on the structure under study, avoiding fixation, dehydration, and resin-embedding artifacts.

Observation of Membrane Proteins In Situ: AQPcic, the Insect Aquaporin Example

179

1. Aliquots of membranes (5Â€µL) were applied 1Â€min to 400mesh collodion/carbon-coated copper grids previously glow discharged and allowed to stand for 1Â€min. 2. Grids were extensively washed with distilled water. 3. Grids were then stained with a drop of 2% aqueous uranyl acetate for 1Â€min. 4. Excess stain was removed by blotting with filter paper, and grids were air-dried. FigureÂ€ 3 shows a purified membrane observed after negative staining. The membrane displays a regular two-dimensional array of particles. 3.4. Embedding in London Resin White

Ultrastructural localisation of macromolecules needs specimen preparations (fixation, embedding) that give a better preservation of the fine cellular structure than the conventional methods used in cytology. 1. Chemical fixation: FCs were chemically fixed for 4Â€h with a solution of 4% (w/v) paraformaldehyde with less than 0.05% (v/v) glutaraldehyde in 0.1Â€M PBS, pH 7.4 at 4°C (see Note 5). The samples were washed overnight in the same buffer at 4°C. 2. Acrylic resin embedding (7, 8) (see Note 6): After buffer rinses, the samples were dehydrated using ethanol at different concentrations (55, 70, and 90%) (v/v), 30Â€ min each, at room temperature. After partial dehydration (see Note 7), the samples were infiltrated with increasing concentrations of London Resin (LR) White resin medium in solvent (usually use 2:1, 1:1, 1:2, ethanol:resin). Finally, the samples were impregnated in a 100% resin, 2–4Â€ h at room temperature, overnight at 4°C and transferred to fresh resin for 2Â€ h. Afterwards, 4–5Â€ µL of LR White catalyst was mixed thoroughly into the resin. To minimise the effect of heating, the adequate holder was maintained in an ice bath and filled with LR White mixture. Immediately, the samples were placed into holder for embedding (see Note 8). 3. Polymerisation: The polymerisation of the resin was triggered at 4°C for at least 48Â€h. 4. Sectioning: Ultra-thin sections (silver–gold to gold sections) were obtained with diamond knives at room temperature and picked up on collodion carbon-coated nickel grids.

3.5. Gold Immunolabelling on Ultra-thin London Resin White Sections

Isolated membranes contain a highly hydrophobic polypeptide P25. To locate this polypeptide on the epithelial cells, immunogold localisation was performed using a specific antibody against P25 and the protein A-gold technique (9, 10). All of the following steps were done in a moist chamber at room temperature.

180

Thomas and Cavalier

1. Blocking non-specific sites: Ultra-thin resin sections were floated on a droplet ³30Â€µL of blocking buffer (1% BSA in PBS) on a piece of Para-Film and stored overnight at 4°C (see Note 9). 2. Formation of the antigen–antibody complex: The sections were incubated with the primary polyclonal rabbit anti-AQPcic diluted 300-fold in PBS–BSA for 2Â€h. Then, they were rinsed three times with the PBS–BSA (see Note 10). 3. Detection of the complex: The preparations were incubated 1Â€ h on a droplet of protein A-gold (10-nm colloidal gold) diluted 40-fold in 0.5% (w/v) BSA in PBS. After three washings with 0.5% BSA in PBS, the sections were rinsed two times with PBS. 4. Complex fixation: For stabilisation of the immunocytochemical reaction and the preservation of cellular structure, sections were briefly floated on 1% glutaraldehyde in PBS. After washing two times with PBS and three times with distilled water, the sections were finally rinsed with a jet of distilled water from a wash bottle. 5. Contrast by conventional method (11): The gold-labelled sections were stained 10Â€min with 5% aqueous uranyl acetate in a dark box, washed three times on a drop of distilled water, and rinsed with a jet of distilled water. Immunogold localisation of P25 (AQPcic) on ultra-thin sections is shown in Fig.Â€4. Gold particles are abundantly present in membranes of the apical microvilli (Mi). A less intense labelling is observed on the BMFs. 6. Control was performed by using a non-immune serum. 3.6. Gold Immunolabelling on Isolated Membranes

Immunogold labelling of isolated membranes was performed as described (3). The different steps of immunodetection were carried out on drops £40–50Â€µL placed on a piece of Para-Film at room temperature. 1. Freshly isolated membranes were deposited on collodion carbon-coated nickel grids. 2. Blocking non-specific sites in blocking buffer: MemÂ� branes were incubated in 1% BSA in PBS for 2Â€ h at room temperature. 3. Formation of the antigen–antibody complex. Membranes were then incubated for 2Â€h at room temperature with primary polyclonal rabbit anti-AQPcic antibody diluted (1:300 dilution) in 0.5% BSA in PBS. 4. Washing: Rinsed three times with the same buffer.

Observation of Membrane Proteins In Situ: AQPcic, the Insect Aquaporin Example

181

5. Detection of the complex: The complex was revealed by a secondary antibody coupled to 10-nm gold particles. The membranes were incubated for 1Â€h at room temperature with a 10-nm GAR-gold diluted 40-fold in 0.5% PBS–BSA, washed three times with 0.5% BSA in PBS and two times with PBS. 6. Staining: After immunogold labelling, membranes were stained with 2% (w/v) aqueous uranyl acetate and viewed using a Philips CM12 microscope operating at 100Â€kV. In Fig.Â€5, gold particles are located only in restricted zones that correspond to a given side of the membrane. 3.7. Freeze Drying and Metal Shadowing

This technique enables to observe the surface topography of a biological object. For a membrane, the contrast is limited to the surface exposed to the evaporated metal. 1. Aliquots of membranes (5Â€µL) were applied 1Â€min to nickel or copper grids, 300–400 mesh, covered with a carbon film or a carbon collodion film freshly glow discharged, and allowed to stand for 1Â€min. 2. Grids were washed 3–5 times with distilled water or with a volatile solvent-like ammonium acetate. Excess of liquid was removed with filter paper. 3. Freezing: Grids were plunged quickly into liquid nitrogen and mounted on the specimen holder maintained at liquid nitrogen temperature. 4. Introduce the specimen on the pre-cooled specimen stage (−150°C) into the high vacuum chamber of the freeze-fracture device. 5. Freeze drying: The temperature of the controlled specimen stage was raised to −80°C and the specimen was maintained at this temperature for about 2Â€h. 6. Shadowing: The specimen stage was tilted to 45°, then 1Â€nm Pt/C was evaporated. Replicas were reinforced by evaporation of a layer of 20Â€ nm of carbon under an elevation angle of 90°. FigureÂ€ 6 shows a Pt/C shadowed membrane from the FC. Folding of the membrane evidences that the membrane is asymmetric as a result of the membrane protein insertion.

3.8. Cryo-Electron Microscopy

The native state of crystallisation of the aquaporin AQPcic within the membrane was a unique favourable case to conduct a structural study on the quaternary structure of this aquaporin. Cryoelectron microscopy gives images of such fully hydrated membranes without the need of staining.

182

Thomas and Cavalier

1. 5Â€ µL of membrane suspension was placed on copper grid coated with a holey carbon grid freshly glow discharge. 2. The grid was then quickly blotted with filter paper and immediately plunged into liquid ethane, using a fast freezing device (see Note 11). 3. The frozen grid was transferred to a Gatan cryoholder maintained at liquid nitrogen temperature and introduced into the electron microscope. 4. The specimen temperature was maintained below −160°C throughout the observations. 5. Images were recorded under low-dose conditions at a magnification of 28,000. FigureÂ€ 7 represents isolated membranes embedded in vitreous ice (inset). At high magnification, a regular array of membrane particles is clearly observed. 3.9. Image Processing

The goal of image processing is to increase the signal/noise (S/N) ratio in an image to reveal fine structural details that are already present in the original image but hidden with random noise. Averaging many images is a very efficient technique since it increases the S/N ratio by a factor of √n, where n is the number of averaged images. Processing of the digitised images was achieved using the SPIDER software system (12). 1. Micrographs were checked by optical diffraction and selected images were digitised on a Nikon coolscan 9000 ED with a step size of 10Â€µm (2,540Â€dpi). 2. Compute a reference image: To calculate an initial reference image, the Fourier transform and the power spectrum were computed from a suitable area and the diffraction pattern indexed. The indexing was used to calculate a Fourier filter mask that was applied to the Fourier transform to produce a filtered image. A subarea of the filtered image was used as a reference. 3. Cross correlation mapping: The raw image was cross-correlated with the reference in order to map similar area in the raw image. 4. Averaging: Areas centred on the peaks in the cross-correlation map were extracted from the raw image and averaged. Rotational coefficients were calculated for quantitative assessment of the symmetry. FigureÂ€8 represents a projection map of one membrane particle. Each particle is formed by four subunits; one subunit corresponds to one monomer of aquaporin. The resolution associated to this projection has been estimated at 2.4Â€ nm. The structural

Observation of Membrane Proteins In Situ: AQPcic, the Insect Aquaporin Example

183

information presented here is limited to the tetrameric organisation as there is no detail about the monomer. The main reason for this lack of information is the limited number and size of the membrane patches available for analysis.

4. Conclusions To overcome this problem and thus get high-resolution structures, two-dimensional crystallisation of proteins followed by electron crystallography has been a powerful alternative. Medium-resolution maps of AQP1 obtained by electron crystallography first revealed the structure of the monomer (13, 14). The monomer is formed by a right-handed bundle of six highly tilted a helices that allows the formation of an hourglass-shaped pore. Later, the first atomic model of a human membrane channel at 3.8â•›Å resolution (Fig.Â€9) derived from electron crystallography data gave the first insight into water specificity (15). Finally, the structure of bovine AQP1 at 2.2-Å resolution has been obtained by X-ray crystallography (16). It reveals the site of water molecule and confirmed the refined structure of human AQP1 derived from electron crystallography. By combining electron crystallographic data at 3-Å resolution and molecular replacement, the structure of the AQP0 membrane junction was resolved (17). It revealed that the water pore is closed in AQP0 junctions and that AQP0 also contains an additional pore constriction not seen in other known aquaporin.

Fig.Â€9. Atomic model at 3.8-Å resolution from electron crystallographic data ((15), PDB accession N° 1FQY). (a) Ribbon representation of the AQP1 tetramer viewed from the periplasmic side. (b) End-on view of the monomer from the extracellular surface (top) and side view (side) depicting the six membrane-spanning helices, two pore helices, and the connecting loops.

184

Thomas and Cavalier

5. Notes 1. To obtain a good fixation, samples smaller than 1Â€mm3 size are considered optimal. 2. Osmium tetroxide preserves and stains structures particularly membranes. Caution: Osmium tetroxide is highly toxic, and the preparation of 2% osmium tetroxide solution should be carried out under chemical hood. 3. Liquid ethane and liquid propane could be used; they have better cryogenic properties than liquid nitrogen. 4. Freeze-fracture refers to experiments where ice removal is negligible; it is performed at low temperature (−120 to −196°C), and Pt/C evaporation is started before breaking the specimen. 5. The least crosslinking aldehyde fixative (i.e. paraformaldehyde) in buffer pH 7.4 is recommended for the preservation of maximum biological material antigenicity. However, it is possible to add ³0.05% (v/v) aqueous glutaraldehyde in the paraformaldehyde solution for a better preservation of the cellular ultrastructure. Osmium tetroxide is avoided since it destroys antigenic sites. 6. Carlemalm etÂ€al. (7) introduced acrylic resin embedding (i.e. Lowicryls, LR White, Bioacryl). These resins were specially developed to be very hydrophilic embedding medium and thus compatible with the localisation of antigens on ultra-thin sections. Caution: LR White resin, a mixture of methacrylate (i.e. polyhydryaromatic acrylic resin) is toxic; refer to the safety information provided with the product. 7. During dehydration, the water sample was progressively replaced by the organic solvent. LR White resin tolerates partial dehydration, accepting tissue from 70 to 90% ethanol. Reducing the time of dehydration and the concentration of ethanol used in dehydration will often improve antigenic sites preservation. 8. LR White may be polymerised by two different procedures: one is the addition of a chemical accelerator benzoyl peroxide and the other is the heat polymerisation without catalyst (50°C). It is important to prepare the resin mixture very quickly since the catalyst–resin mixture will polymerise in £5Â€min. The reaction is highly exothermic and should be controlled in the cold. Embedding in LR White medium at 4°C for immunolabelling is a rapid method but the least sensitive immunodetection method, not appropriated if the protein to be localised is poorly expressed.

Observation of Membrane Proteins In Situ: AQPcic, the Insect Aquaporin Example

185

9. To prevent all non-specific binding, a blocking protein was dissolved in a standard buffer; we routinely use 1% (w/v) BSA, Fraction V in PBS, pH 7.4. 10. Fractions containing 1–3Â€ µg of pure AQPcic, isolated from SDS-PAGE and mixed with Freud’s adjuvant, were injected to rabbits at 3-week intervals until immunisation. 11. Fast freezing is used to vitrify biological specimens contained within thin films. Such specimens include purified suspensions of macromolecules or suspensions of bacteria. The main advantage is that the specimen is fully embedded in its native environment, thus preserving its three-dimensional structure. References 1. Gouranton J (1968) Ultrastructures en rapport avec un transit d’eau. Etude de la “chambre filtrante” de Cicadella viridis L. (Homoptera, Jassidae). J Microsc (Paris) 7:559–574 2. Hubert JF, Thomas D, Cavalier A, Gouranton J (1989) Structural and biochemical observations on specialized membranes of the “filter chamber”, a water-shunting complex in sapsucking homopteran insects. Biol Cell 66:155–163 3. Le Cahérec F, Deschamps S, Delamarche C, Pellerin I, Bonnec G, Guillam MT, Thomas D, Gouranton J, Hubert JF (1996) Molecular cloning and characterization of an insect aquaporin. Functional comparison with aquaporin 1. Eur J Biochem 241:707–715 4. Thomas D, Schultz P, Steven AC, Wall JS (1994) Mass analysis of biological macromolecular complexes by STEM. Biol Cell 80:181–192 5. Beuron F, Le Caherec F, Guillam MT, Cavalier A, Garret A, Tassan JP, Delamarche C, Schultz P, Mallouh V, Rolland JP, Hubert JF, Gouranton J, Thomas D (1995) Structural analysis of a MIP family protein from the digestive tract of Cicadella viridis. J Biol Chem 270:17414–17422 6. Venable JH, Coggeshall R (1965) A simplified lead citrate stain for use in electron microscopy. J Cell Biol 25:407–408 7. Carlemalm E, Garavito RM, Villiger W (1982) Resin development for electron microscopy of embedding at low temperature. J Microsc 126:123–143 8. Hobot JA (1989) Lowicryls and low temperature embedding for colloidal gold methods. In: Hayat MA (ed) Colloidal gold: principles,

9.

10.

11. 12. 13.

14.

15.

16.

17.

methods and applications, vol 2. Academic Press, San Diego, pp 75–115 Bendayan M (1984) Protein A-Gold electron microscopy immunocytochemistry: methods, applications, and limitations. J Electron Microsc Tech 1:243–270 Roth J, Bendayan M, Orci L (1978) Ultrastructural localization of intracellular antigens by the use of protein A-gold complex. J Histochem Cytochem 26:1074–1088 Roth J, Taatjes DJ, Tokuyasu KT (1990) Contrasting of Lowicryl K4M thin sections. Histochemistry 95:123–136 Frank J, Shimkin B, Dowse H (1981) SPIDER-A modular software system for electron image processing. Ultramicroscopy 6:343–358 Walz T, Hirai T, Murata K, Heymann JB, Mitsuoka K, Fujiyoshi Y, Smith BL, Agre P, Engel A (1997) The three-dimensional structure of aquaporin-1. Nature 387:624–627 Cheng A, van Hoek AN, Yeager M, Verkman AS, Mitra AK (1997) Three-dimensional organization of a human water channel. Nature 387:627–630 Murata K, Mitsuoka K, Hirai T, Walz T, Agre P, Heymann JB, Engel A, Fujiyoshi Y (2000) Structural determinants of water permeation through aquaporin-1. Nature 407:599–605 Sui H, Han BG, Walian P, Jap BK (2001) Structural basis of water-specific transport through the AQP1 water channel. Nature 414:872–878 Gonen T, Sliz P, Kistler J, Cheng Y, Walz T (2004) Aquaporin-0 membrane junctions reveal the structure of a closed water pore. Nature 429:193–197

as

Chapter 10 Two-Dimensional Crystallization of Integral Membrane Proteins for Electron Crystallography David L. Stokes, William J. Rice, Minghui Hu, Changki Kim, and Iban Ubarretxena-Belandia Abstract Although membrane proteins make up 30% of the proteome and are a common target for therapeutic drugs, determination of their atomic structure remains a technical challenge. Electron crystallography represents an alternative to the conventional methods of X-ray diffraction and NMR and relies on the formation of two-dimensional crystals. These crystals are produced by reconstituting purified, detergentsolubilized membrane proteins back into the native environment of a lipid bilayer. This chapter reviews methods for producing two-dimensional crystals and for screening them by negative stain electron microscopy. In addition, we show examples of the different morphologies that are commonly obtained and describe basic image analysis procedures that can be used to evaluate their promise for structure determination by cryoelectron microscopy. Key words: Membrane protein structure, Electron crystallography, Two-dimensional crystals, Electron microscopy, Crystal screening

1. Introduction Electron crystallography is a method of membrane protein structure determination that involves imaging two-dimensional (2D) crystals by cryoelectron microscopy (cryo-EM). Such 2D crystals are produced when membrane proteins adopt a regular array within the plane of a lipid bilayer. Methods for electron crystallography were originally pioneered in the 1970s by Henderson and Unwin in their studies of bacteriorhodopsin (1). Intense efforts over the next two decades led to developments both in electron microscope design and in software for analyzing the resulting images, thus producing the first atomic resolution structure of a membrane protein in its native membranous Jean-Jacques Lacapère (ed.), Membrane Protein Structure Determination: Methods and Protocols, Methods in Molecular Biology, vol. 654, DOI 10.1007/978-1-60761-762-4_10, © Springer Science+Business Media, LLC 2010

187

188

Stokes et al.

environment (2). Since then, electron crystallography has developed into a powerful tool to elucidate the 3D structure of a wide range of membrane proteins (3–6), thus offering a plausible alternative to X-ray crystallography. Similar to X-ray crystallography, the bottleneck preventing a more generalized use of electron crystallography is the preparation of well-ordered crystals. In the case of membrane proteins, there are three predominant morphologies adopted by 2D crystals: (1) flattened lipid vesicles with two, overlapping 2D lattices; (2) tubular lipid vesicles which retain a cylindrical shape and comprise a helical array of membrane proteins; and (3) a planar lipid bilayer with a single, coherent 2D array of proteins. Historically, 2D crystallization of membrane proteins has involved either in situ crystallization or inÂ€vitro reconstitution (7, 8). The advantage of in situ crystallization is that the membrane protein is never removed from its native membrane. However, this approach requires a high concentration of the relevant protein in the native membrane and is not therefore generally applicable. More generally, 2D crystals have been grown by reconstitution of purified, detergent-solubilized membrane proteins into lipid bilayers under defined conditions (7, 9). Reconstitution involves the controlled removal of detergent, either by dialysis or by adsorption to a hydrophobic resin, in the presence of exogenous lipid. Under favorable conditions, the membrane proteins assemble into an ordered 2D array within the resulting lipid bilayers (7). In this chapter, we will describe protocols and procedures to obtain 2D crystals of membrane proteins and to evaluate their quality and potential for structure determination by electron crystallography.

2. Materials 2.1. Analysis of Lipid

1. Solid phase: 20â•›×â•›20Â€ cm thin layer chromatography (TLC) plates with silica gel 60 F254 coating (Merck & Co., Whitehouse Station, NJ). 2. Glass cutter. 3. Glass chromatography tank (25Â€cmâ•›×â•›27Â€cmâ•›×â•›10Â€cm). 4. Graduated disposable 5Â€ml glass capillaries (CAMAG, Muttenz, Switzerland). 5. Mobile phase: mixture of chloroform/methanol/25% aqueous ammonia (65:25:5, by volume). 6. Compressed N2 gas. 7. Kontes Chromatography TLC Reagent Sprayer with Standard Ground Joint Sprayer (Fisher Scientific, Pittsburgh, PA).

Two-Dimensional Crystallization of Integral Membrane Proteins

189

8. A 0.1% aqueous solution in water of 8-anilino-1-naphthalene sulfonic acid (ANSA) (Sigma, St. Louis, MO). 9. E. coli polar phospholipid extract (Avanti Polar Lipids, Alabaster, AL) solubilized at 1Â€mg/ml in chloroform as a standard. 2.2. Biobead Reconstitution

1. SM2 Biobeads (BioRad, Hercules, CA), washed with ethanol and then with water and stored under water at 4°C with 1Â€mM NaN3. Care should be taken to prevent the Biobeads from drying out. Prior to use, bulk liquid should be removed by aspiration with a pipette and they should be kept covered and cold until water can be added back. 2. Compressed N2 gas. 3. Lyophilizer or vacuum desiccator. 4. 50Â€mM Tris pH 7.3. 5. Lipid stock solutions purchased in chloroform (Avanti Polar Lipids), e.g., egg yolk phosphatidyl choline, egg yolk phosphatidyl ethanolamine, egg yolk phosphatidic acid, or dioleoylphosphatidylcholine, dioleoylphosphatidyl ethanolamine, palmitoyl-oleoylphosphatidic acid. 6. 50% sucrose stock solution in 50Â€mM Tris–HCl pH 7.3. 7. Detergent stock solution 10Â€ mg/ml (e.g., C12E8, Anatrace, Maumee, OH) and stored at 4°C.

2.3. Detergent Removal by Dialysis

1. 50–200Â€ µl microdialysis buttons and o-rings (Hampton Research, Aliso Viejo, CA). 2. 40Â€mm dialysis membrane 12,000–14,000 MW cutoff (Spectra/ Pore, VWR International, West Chester, PA). Prepare 30â•›mm × 30Â€mm square pieces by cutting all four sides of the strip and soak in a beaker of water. Beaker may be stored in the refrigerator.

2.4. Solid Carbon Films

1. Carbon rod, either 1/8″ or 1/4″ depending on the source provided with your vacuum evaporator. Alternatively, a carbon thread can also be used (Electron Microscopy Sciences, Hatfield, PA). 2. Mica sheets (Electron Microscopy Sciences). 3. Crystallizing dish (190â•›×â•›100Â€mm) (VWR International). 4. Small test tube rack, tygon tubing, 5–10Â€ ml glass pipette, hemostats, plastic tray, 2 small coins. 5. 90Â€ mm plastic petri dishes, 90Â€ mm Grade 4 Whatman filter paper. 6. 300 mesh Cu EM grids (Ted Pella, Redding, CA). 7. Stainless steel forceps (e.g., #5 Dumont). 8. Vacuum evaporator (e.g., Edwards Auto306, Edwards Vacuum, Wilmington, MA).

190

Stokes et al.

2.5. Holey Carbon Films

1. Glass slides: standard 1â•›×â•›4″ slides for light microscopy cleaned by soaking in acetone overnight in glass staining dish (20–30 slide capacity, Electron Microscopy Sciences or Ted Pella). 2. 2–3 Coplin staining jars (Electron Microscopy Sciences or Ted Pella). 3. Stock solution of 0.4% Formvar in chloroform (Electron Microscopy Sciences). Store in glass-stoppered flask at room temperature. 4. Stock solution of 50% glycerol in water. 5. Plastic boxes for storing glass slides (Electron Microscopy Sciences). 6. Crystallizing dish (190â•›×â•›100Â€mm) (VWR International). 7. EM grids, carbon rod/thread, and vacuum evaporator as above. 8. Methanol, ethylene dichloride, 50% hydrofluoric acid.

2.6. Negative Staining

1. Solid carbon films (prepared in SubheadingÂ€3.4). 2. Stainless steel forceps, either reverse grip or with o-ring to hold them closed. 3. Glow discharge apparatus (e.g., Edwards Auto306, Edwards Vacuum). 4. Stock solution of 1% uranyl acetate in water. 5. Grade 1 Whatman filter paper, torn into strips. 6. Plastic boxes for storage of EM grids (Electron Microscopy Sciences).

2.7. Screening of Samples by Electron Microscopy

1. Electron microscope: 80Â€ kV or higher accelerating voltage (e.g., JEOL, USA, Peabody MA). 2. Stainless steel forceps (e.g., #5 Dumont, Ted Pella). 3. ImageJ computer software for calculating Fourier transform of images (10).

3. Methods The following methods focus on the reconstitution, crystallization, and screening of membrane proteins. Prior to applying these methods, we assume that the membrane protein of interest has been purified in a detergent-solubilized state, either from natural sources or from a cell expression system. In either case, it is generally useful to know the lipid composition of the purified protein; so the first set of methods describes a suitable analysis using TLC. After that, we describe two alternative methods

Two-Dimensional Crystallization of Integral Membrane Proteins

191

for 2D crystallization. Conceptually, crystallization involves two steps: reconstitution of the membrane protein into a lipid bilayer and ordering of molecules within that bilayer to form a 2D array. When dialysis is used, detergent is removed slowly (up to 2Â€ weeks) and these two processes occur simultaneously. Detergent removal by hydrophobic resins is two orders of magnitude faster (3–4Â€ h) and in this case, crystallization is often carried out as a second, subsequent step. The conditions for 2D crystallization generally require a screen involving a number of different factors – such as pH, salt concentration, lipid composition, detergent composition – and the resulting samples must be assessed by EM. In particular, EM is used to evaluate the morphology of the resulting proteoliposomes and to look for the presence of a regular lattice. Therefore, we describe methods for preparing two different kinds of carbon-coated EM grids and for preparing negatively stained samples. Furthermore, examples are given for different morphologies produced by reconstitution and for two different lattice assemblies that can be produced. For structure determination by electron crystallography, cryoEM is required to image the 2D crystals. Although these methods are beyond the scope of this chapter, we also include a protocol for producing holey carbon films, which are useful for preparing frozen-hydrated samples of 2D crystals. 3.1. Analysis of Lipid Species Copurifying with Membrane Proteins

1. Using a graduated glass cylinder, add 100Â€ ml chloroform/ methanol/25% aqueous ammonia (65:25:5, by volume) to a glass chromatography tank that is lined with filter paper. Cover tightly and allow the vapors in the tank to equilibrate for 10Â€min. 2. Use a glasscutter to divide the TLC plates into 5â•›×â•›10Â€cm pieces. Use gloves to avoid contamination of the Si-coated surface. 3. Using a pencil, draw a very light line across the plateÂ€1.5Â€cm from the bottom. Be careful not to scratch the Si coating. 4. Using glass capillaries, spot 2–4Â€ml of sample at 5Â€mm intervals along the pencil line. Standard samples (e.g., E. coli polar lipid extract) should be included so that lipid species associated with the protein can be identified. Amount of lipid added in each spot should be 1–5Â€µg. 5. Dry sample using a stream of N2 gas. 6. Using a pair of long forceps, grasp the top of the plate and place it into the tank such that the spots of sample lie just above the level of the solvent. Allow the top edge of the plate to lean against the back wall of the developing chamber. 7. Cover the tank tightly and allow the solvent to progress up the plate. Leave undisturbed until the ascending solvent front reaches the top of the plate (approx. 30–45Â€min).

192

Stokes et al.

8. Remove the TLC plate with forceps, and use a stream of N2 gas in chemical fume hood to dry the plate thoroughly. 9. Once dry, spray the plate briefly with 0.1% ANSA solution in the chemical fume hood. 10. Illuminate plate with an UV lamp either from above or from below and use a digital camera to record a picture. If the staining is weak, step 9 may be repeated. 3.2. 2D Crystallization Using Biobeads

1. Prepare a thin film of dried lipid by pipetting 500Â€µg of lipid stock solution into a glass test tube (7Â€ cmâ•›×â•›1Â€ cm) or small round bottom flask (25Â€ml) (see Notes 1 and 2). Rotate the flask rapidly to swirl the solution around to the bottom while blowing N2 gas across the solution to evaporate the chloroform. This procedure should produce a thin, even film of dry lipid on the bottom of the tube. The film should appear white and not contain any thick, clear areas. If the film appears too thick, dissolve it with more chloroform and try again. Once satisfied, place tube into a lyophilizer or vacuum desiccator for 1Â€h to remove residual chloroform. Then add 100Â€µl of detergent stock solution (1Â€mg) and vortex to create a suspension of lipid. This represents the lipid stock solution to be used the same day for reconstitution. 2. Prepare solutions with 1Â€ mg/ml protein concentration and with lipid-to-protein ratios ranging from 0.25 to 2.0 by weight. The final detergent concentration should be at least double the lipid concentration to ensure complete solubilization. The buffer conditions should be tailored to the particular protein under investigation, initially to prevent aggregation and eventually to induce 2D crystallization. 3. Add 100Â€ µl of these protein–detergent–lipid solutions to a test tube. Add a small stir bar and place on a magnetic stirrer at room temperature. Several test tubes can be placed in a small beaker and stirred simultaneously. 4. Over a 3–4Â€h period, add Biobeads at a Biobead-to-detergent mass ratio of 16g:1mg for C12E8 (see Note 3). This includes detergent that accompanies both the protein and lipid stock solutions. There are two strategies for adding Biobeads, which often produce different results. Fast reconstitution involves simply adding all the Biobeads at the beginning of the incubation period (3Â€ h). Slow reconstitution involves four consecutive additions at 30Â€min intervals of a 1Â€g:1Â€mg Biobeads:detergent mass ratio, a 4:1 ratio for a 1Â€h incubation, followed by a final addition of the remaining 8:1 ratio followed by 1Â€ h incubation, all at room temperature (see Note 4).

Two-Dimensional Crystallization of Integral Membrane Proteins

193

5. Remove the reconstituted protein solution and layer on top of a discontinuous sucrose gradient composed of three layers (10, 25, and 50%) of 0.7Â€ml each in a 2Â€ml centrifuge tube. Place the centrifuge tube in a swinging bucket rotor, and centrifuge at 100,000 gâ•› for at least 3Â€h. 6. Collect the protein band from the sucrose gradient, and dilute 10- to 20-fold in buffer (see Note 5). Centrifuge either in the same swinging bucket rotor or in a refrigerated benchtop centrifuge at 20,000–30,000 g for 30Â€min. Repeat this wash step at least once to remove sucrose. 7. Freeze–thaw cycles are often effective in increasing the size of proteoliposomes and in promoting crystallization. For this procedure, do not resuspend the final pellet from the previous step. Instead, remove the supernatant and add the buffer that you wish to use for crystallization to produce a final protein concentration of ~1Â€mg/ml. Without resuspending the pellet, immerse the tube into liquid N2. Then thaw the pellet between thumb and forefinger. Repeat 3–4 times. Finally, resuspend any remaining pellet with a pipette. Crystallization often requires incubation for 1–7Â€days. This final solution can also be dialyzed against a crystallization buffer if desired (see below). 3.3. 2D Crystallization by Detergent Removal

1. Prepare lipid stock solutions and protein solutions as described in steps 1 and 2 of SubheadingÂ€3.2. 2. Prepare dialysis solutions with a range of pH, salts and other additives that could be effective in promoting 2D crystal formation and aliquot into 100Â€ml beakers. 3. Add lipid/protein/detergent solutions to microdialysis buttons with a volume that exceeds their stated capacity by 5–10% such that the meniscus extends above the surface of the button. 4. Place o-ring on top of a 3â•›×â•›3Â€cm2 piece of dialysis membrane. In one motion, place dialysis membrane on top of the button and press o-ring into groove. The goal is to prevent the introduction of air bubbles into the dialysis button (Fig.Â€1a). 5. Place 2–4 dialysis buttons on the bottom of each 100Â€ml beaker with dialysis membrane oriented upwards (Fig.Â€1b). Only the detergent and buffer components will exchange with the dialysis buffer; so buttons with different lipid-to-protein ratios or different lipid species can be mixed within one beaker. In this case, a permanent marker can be used to label the bottom of individual buttons. 6. Allow dialysis to continue for 2 weeks, changing the buffer daily. Solutions can be kept refrigerated or at room temperature or cycled between different temperatures as part of the crystallization strategy. Stirring is not generally required.

194

Stokes et al.

Fig.Â€1. Dialysis buttons for removal of detergent. (a) Individual dialysis button is shown with dialysis membrane held in place by a black o-ring. After securing the dialysis membrane in this way, the excess can be cut away with scissors. (b) Dialysis buttons in the bottom of a 100Â€ml beaker filled with dialysis solution. Stirring is not necessary, though the solution should be changed at least once per day to promote efficient detergent removal.

7. To harvest samples, remove buttons from the beaker and place on a piece of parafilm. Use a scalpel to cut an “X” in the dialysis membrane. Then use a micropipette to remove the solution, and store in a microfuge tube. 3.4. Preparation of Solid Carbon Film for Electron Microscopy

1. Sharpen carbon rod, blow off dust remnants, and mount in vacuum evaporator. Handle the carbon rod with gloves to prevent contamination with body oil. 2. Cut 4â•›×â•›2Â€cm2 piece of mica with scissors. Cleave in half longitudinally using your fingernail, and place on filter paper in a petri dish with freshly cleaved surface facing upwards. Cut triangular piece of filter paper, fold in half such that one vertex overlaps the opposite side of the triangle, and place in petri dish alongside the mica (Fig.Â€2). Place petri dish into vacuum evaporator, and obtain a vacuum of at least 1â•›×â•›10−6â•›Torr. 3. If available, use retractable shield to cover mica sheets. Increase current through carbon rod until vacuum begins to degrade. Reduce current until vacuum recovers. Repeat until degradation is minimal. This step is designed to burn contamination off the carbon rod prior to evaporation. 4. Retract shield to expose mica sheets. Gradually increase current until carbon begins to evaporate. Maintain lowest possible current during evaporation to prevent sparking. After 15–30Â€s, turn down current and inspect shadow produced by depositing carbon on the folded triangle of filter paper. Repeat evaporation as required to obtain a light, but distinct gray shadow (see Note 6 and Fig.Â€2).

Two-Dimensional Crystallization of Integral Membrane Proteins

195

Fig.Â€2. Carbon coating of EM grids. A group of EM grids lie on a piece of white filter paper in the bottom of a petri dish. Two triangular pieces of filter paper have been folded over to cast a shadow during carbon evaporation. Four coins were placed on the filter paper to hold it down during the evaporation process, but have been removed to reveal their shadow. An array of EM grids have been placed in the center of the Petri dish; eight of these grids have been removed from the right side of the array. The contrast of these shadows is a good measure of the thickness of the carbon that has been deposited either, as in this case, on the plastic coated grids, or on the freshly cleaved mica.

5. Vent vacuum evaporator, remove petri dish, and store it covered at room temperature (see Note 7) until ready to coat EM grids. 6. Fill crystallization dish with water, and place on a tray. Place test tube rack under water. Cut 3â•›×â•›5Â€cm2 piece of filter paper, and place on top of the test tube rack. Weigh down the filter paper by placing coins on the edges. 7. Individually place 300 mesh EM grids on top of the filter paper with forceps in a close packed array that matches the size of the mica (see Notes 8 and 9). 8. Carefully add water until crystallization dish is almost overflowing. Wipe surface with 10Â€ml glass pipette to remove dust from the surface. Insert 40–50Â€ cm length of tygon tubing into bottom of crystallization dish, and suck on end to create siphon. Clamp tygon tubing to stop siphon until later. 9. Pick up mica with forceps. “Huff” on the mica to humidify the carbon-coated surface (don’t blow with pursed lips; say “hot” without pronouncing the “t”). Hold mica at 10–20° relative to the water surface, and gently lower it into the water. The carbon should float off of the mica and remain at the surface. Once the carbon is completely separated, drop the mica into the bottom of the crystallization dish.

196

Stokes et al.

10. Use the glass pipette and/or the forceps to maneuver the carbon over the filter paper. Release the hemostats to allow the siphon to remove water gradually from the crystallization tray. The carbon should land on top of the grids as the water recedes below the filter paper. The siphon can be paused just before the carbon contacts the filter paper in order to perfect the positioning of the carbon over the grids. 11. Lift the filter paper together with the carbon-coated grids, and place on a piece of dry filter paper in a plastic petri dish. Store at room temperature until ready for use. 3.5. Preparation of Holey Carbon Films

1. Mix 50Â€ml of the Formvar stock solution with 10–50 drops of the glycerol solution. Sonicate with a probe sonicator for 2–5Â€ min, and wait for 0–10Â€ min after the sonication. Conditions for sonication should be adjusted to obtain the desired size and distribution of holes in the plastic film (see Note 10 and Fig.Â€3). 2. Using forceps, hold a glass slide vertically, dip it into the sonicated solution, and slowly pull it out. Allow the slide to air dry and examine it in the light microscope using phase contrast optics. Once suitable plastic films have been obtained, they can be stored in a plastic box for years. It is helpful to make notes with a permanent marker on the top of the slides for later reference. 3. To transfer the holey plastic film to EM grids, it must be floated onto the surface of water as follows. Use a razor blade to scrape the edges of the glass slide and also to scrape away the upper and lower margins of the plastic film to provide a distinct edge.

Fig.Â€3. Holey carbon films. The density and size of holey carbon films can be controlled by the concentration and sonication of the glycerol/chloroform mixture (see text). (a) Closely packed holes are useful for cryoelectron microscopy of helical crystals, where one wishes to image the crystal as it lays across the hole suspended in vitreous ice. (b) Widely spaced holes are useful for cryoelectron microscopy of planar 2D crystals that require support by the carbon. These holes allow the grids to be blotted from behind, thus concentrating crystals on the carbon surface. This is a useful strategy for 2D crystals that fail to adsorb efficiently to solid carbon supports. The size of holes in (a) is ~2Â€µm.

Two-Dimensional Crystallization of Integral Membrane Proteins

197

Fill the crystallizing dish with water. In order to facilitate release of the plastic film released from the glass slide, wave the slide (plastic film down) over an open bottle of hydrofluoric acid to loosen the association. Holding the glass slide with hemostats at a 10–20° angle relative to the surface of the water, lower it gradually such that the plastic film floats off. 4. Place EM grids directly on top of the plastic film with the polished side down. Gently lay either a piece of parafilm or its backing paper on top of the grids and plastic film. Allow water to saturate the region between the plastic and the parafilm. Gently lift the parafilm/backing paper, which should also lift the plastic film and associated grids off the water surface. Place this sandwich onto filter paper in a plastic petri dish and allow it to dry. 5. Etch the plastic films to remove “pseudo-holes” as follows. Place a piece of filter paper in a glass petri dish. Add enough methanol to saturate the filter paper as well as a residual amount of liquid in the bottom of the dish. Individually place the grids onto the filter paper plastic-side down. Cover and allow to incubate for 10–15Â€min. Remove the grids and place them on a dry piece of filter paper. 6. Select a group of grids by examining them individually in a light microscope using phase contrast optics and a 40× objective. Place the grids on a glass slide plastic-side up into the vacuum evaporator. Include a folded triangular piece of filter paper as described above for carbon evaporation. Prepare carbon rod/thread and evaporate carbon as described above (Fig.Â€2). The carbon coated grids can be stored at room temperature though they should be used within a few days/weeks of the evaporation. 7. Remove the plastic films using ethylene dichloride. Similar to step 5 above, place 2–3 pieces of filter paper into a small glass petri dish and saturate them with ethylene dichloride. Place the grids on the filter paper carbon side facing upwards. Cover and incubate for 10–30Â€min. Examine the grids with a light microscope using phase contrast optics and the 40× objective to ensure that the carbon film remains intact. 3.6. Preparation of Negatively Stained Samples

1. Place solid carbon grids on a clean glass slide with the carbon facing upwards. Place slide inside glow discharge apparatus (see Note 11), and evacuate to ~150Â€mTorr. Turn on electrical discharge, and observe a faint blue/purple glow throughout chamber (more visible with room light turned down). Allow discharge to proceed for ~30Â€s. If sparking is observed or grids are jumping, the discharge is too strong and the carbon film will break. If in doubt, examine one or two grids by phase-contrast light microscopy to ensure that the carbon film was not broken.

198

Stokes et al.

2. Grasp grid by the edge using reverse-grip forceps with the carbon coat facing up. Pipette 5Â€µl of reconstituted protein sample onto the surface of the grid. If glow discharge was effective, sample should spread evenly across grid. Pooling of sample indicates that grid is too hydrophobic. 3. Incubate ~30Â€s, then blot excess sample from grid by touching torn edge of filter paper to edge of the grid. Do not scrape filter paper across top of grid. 4. Immediately pipette 5Â€ µl of 1% uranyl acetate solution onto grid. Do not allow grid to dry before adding this solution. It is easiest if a pipette is preloaded with stain prior to the blotting step. Incubate 15–30Â€ s and blot stain from side as in step 3. Depending on the sample, this staining step may be repeated 2–3 times. After the final application, thoroughly remove all the stain from the grid and forceps by touching a torn filter paper to the side of the grid and to the side of the forceps where the stain may have wicked. Try to remove as much liquid as possible. 5. Allow the grid to air-dry and store in a plastic grid box. Uranyl acetate stained grids should last for several years. Other stains may be more volatile (see Note 12). 3.7. Screening of 2D Crystallization Trials by Electron Microscopy

1. Insert stained grid into electron microscopy according to the manufacturer’s instructions. 2. Start with the microscope at low magnification (150–300×). Be sure that the objective aperture has been removed as it will limit the field of view in this mode. Observe the overall appearance of the grid. Most squares should be light grey, possibly with some dark flecks, whereas broken squares will appear bright (Fig.Â€ 4). Too much sample and/or stain will result in very dark, possibly black squares. If too many grid squares are broken or black, see Note 13. 3. Choose an area of the grid with several good squares. Increase magnification to ~3,000×, insert objective aperture to increase contrast, and adjust sample height to the eucentric position. Adjust focus to achieve 5–10Â€µm underfocus. 4. Scan grid square noting features and recording images. Look for the following basic morphologies (Fig.Â€4): Vesicles with poor contrast and irregular shape, which are likely to be pure lipid. Large aggregates of protein. Round vesicles with good contrast, which are likely to be proteoliposomes. Rod-shaped vesicles, which are possibly a sign of helical crystals. Large sheets, which are possibly a sign of 2D crystals.

Two-Dimensional Crystallization of Integral Membrane Proteins

199

Fig.Â€4. Screening of negatively stained crystal trials. (a) Low magnification image showing multiple grid squares coated with continuous solid film. Although a few grid squares are broken on the lower right, the film is mostly intact. (b) Low magnification image showing considerable precipitate on this grid and many broken grid squares on the right hand side. This grid is less likely to produce nice images. (c) Pure lipid vesicles generally form extended, amorphous structures. (d) Proteoliposomes are generally spherical, with a tendency to collapse and form wrinkles when dried in negative stain. (e) Elongated shapes can indicate that the protein is forming an ordered lattice within the bilayer, which after further screening could produce helical crystals. (f) Single-layered membrane sheets lie flat on the carbon support and offer the best opportunity for high resolution structure determination. Unlike the vesicles in (d), these sheets do not show wrinkles that result from the collapse of the top bilayer onto the lower bilayer. Scale bars are 500Â€µm in (a, b) and 100Â€nm in (c–f).

5. When rod-shaped vesicles or large sheets are found, increase the magnification to ~50,000× and defocus to 1–2Â€ mm underfocus (see Note 14). At this magnification, you may be

200

Stokes et al.

able to see a lattice if sample is crystalline, depending on how much of the protein is outside of the membrane where negative stain is most likely to produce contrast. FigureÂ€ 5 shows examples of helical and 2D crystals. 6. After recording an image (preferably on a CCD camera, see Note 15), calculate a Fourier transform. A 2D crystal will produce discrete diffraction spots, and a helical crystal will produce layer lines (Fig.Â€5). In addition, Thon rings may be visible in the background and can be used to judge defocus and astigmatism. 7. Continue screening other grid squares in areas all over grid. Stain is inherently variable, and some areas will be better stained than others. Signs of suitable staining include good

Fig.Â€5. Negatively stained 2D crystals. (a) Planar 2D crystal shows a regular lattice. (b) A helical crystal also has a regular lattice, though a moire pattern is produced by overlap of lattices from the front and back sides of the cylindrical tube. (c) Fourier transform of a 2D crystal is characterized by discrete spots at regular intervals. A black circle surrounds the innermost spots, which is produced by the defocus used to record the image. Careful inspection will reveal additional spots appearing beyond this circle near the edge of the panel. (d) After rotating a helical crystal to align the long axis vertically, its Fourier transform is characterized by a regular series of horizontal “layer lines.” The pattern possesses mirror symmetry about both the horizontal and vertical axes. Scale bars in (a, b) correspond to 60Â€nm.

Two-Dimensional Crystallization of Integral Membrane Proteins

201

contrast, a halo of stain around the object, and a low, featureless background. 8. Computer analysis of images requires considerable expertise, but preliminary evaluation of ordered arrays can be done as follows: ●●

●●

●●

●●

●●

●●

Open 50,000 × magnification image in ImageJ software (10). Box out an area of interest. Under “Process” menu, choose to calculate Fourier transform. Adjust brightness/contrast of Fourier transform until diffraction spots or layer lines are visible. Record resolution of diffraction spots closest to the origin as well as maximum resolution of spots furthest from the origin by positioning mouse over individual diffraction spots. ImageJ will report resolution of any given location in pixels (edge of the Fourier transform corresponds to the Nyquist limit, which is 2 pixel resolution). Multiply values reported by ImageJ by the calibrated pixel size to calculate the resolution in Angstroms. Example: ImageJ reports resolution of 33 pixels and the image was recorded at a magnification producing 2â•›Å/pixel, so resolution of spot is 33â•›×â•›2â•›=â•›66â•›Å. Calculate a filtered image by using the circle tool to encircle each of the visible spots in the Fourier transform. Choose to clear each spot, and then calculate an inverse Fourier transform (under the “Process” menu).

9. To consult with a potential collaborator about prospects for structure determination, assemble the following data: ●●

●●

●●

●●

Images. Sample Fourier transform showing diffraction spots or layer lines. Estimate of highest resolution visible in the Fourier transform. Estimate of lattice size for the 2D crystal (i.e., resolution of the diffraction spots closest to the origin of the Fourier transform).

4. Conclusion Although the methods for growing 2D crystals are relatively straightforward, it is unfortunately not possible to predict which conditions to use for a particular membrane protein. As with X-ray crystallography, a wide range of conditions must therefore be

202

Stokes et al.

screened in order to identify a general set of parameters that maintain protein stability and that produce suitable membrane morphologies. Promising conditions are then subjected to subsequent screening designed to optimize size of membrane assemblies and order of any resulting 2D crystals. Unlike soluble proteins, membrane proteins generally require detergent solubilization, which complicates the physical chemistry of the crystallization solution and represents an additional parameter that must be included in the screen. For 2D crystallization, the situation is even more complex, as the composition of added lipid and the procedures used for reconstitution are also crucial to successful crystallization. As a result, a high-throughput approach is necessary for sampling a sufficient number of crystallization conditions. Although liquid-handling robots, multi-well crystallization plates, and automated imaging are standard for X-ray crystallography, 2D crystallization has, until recently, been done largely by hand. This is due to the lack of suitable apparatus for multi-well dialysis and to the cumbersome nature of sample preparation and imaging by EM. Recently, some progress has been made on both these fronts. Strategies for automated imaging of samples in the electron microscope have been reported, using either a special 100-grid holder (11) or a sample-loading robot (12,19). Also, devices for microdialysis and negative staining have been developed on a 96-well format (13). By integrating all of these pieces, it is now possible to construct a pipeline for high-throughput screening of 2D crystallization trials (20). Thus, electron crystallography is becoming a plausible strategy not only for structure determination of membrane proteins, but also for participation in the genome-wide effort being taken by the NIH sponsored Protein Structure Initiative.

5. Notes 1. 2Â€µg of butylatedhydroxytoluene (BHT) can be added to the lipid solution prior to drying to minimize oxidation. 2. Lipid composition has a significant influence on the results of reconstitution and crystallization. To start, try a mixture of three lipids in the following ratio: 80% phosphatidylcholine as the bulk lipid, 10% phosphatidylethanolamine which tends to destabilize bilayer structures and thus induce fusion into larger sheets, 10% phosphatidic acid or phosphatidylserine which carries a negative charge and minimizes aggregation of vesicles. 3. Biobeads have different capacities for different detergents. According to Rigaud and colleagues (14), the following absorptive capacities can be expected in mg detergent/g Biobeads: Triton X100 – 185, C12E8 – 190, dodecylmaltoside – 105,

Two-Dimensional Crystallization of Integral Membrane Proteins

203

Cholate – 80, CHAPS – 85, octylglucoside – 117. An excess of Biobeads relative to these values is always used due to the slow kinetics of binding. 4. Fast reconstitution often produces a more homogeneous preparation of proteoliposomes with an even distribution of randomly inserted protein molecules. Slow reconstitution can produce a heterogeneous preparation of vesicles with varying lipid-to-protein ratios and an asymmetric insertion of proteins into the bilayer (15). Presumably, this heterogeneity and preferential sidedness is a result of protein–protein interactions during bilayer formation, which depends on the kinetics of reconstitution. 5. The protein should be visible as a white band at the interface of the 25 and 50% sucrose layers. However, there may be multiple bands if a heterogeneous population of proteoliposomes were produced. At first, it may therefore be useful to collect samples systematically throughout the sucrose gradient and assay them for lipid and protein content to characterize the results of the reconstitution (15). 6. Some report that multiple evaporations of carbon produce a stronger, flatter, and more conductive carbon film. Some prefer graphite rods rather than amorphous carbon, but in each case, the rods should be of the highest possible purity. 7. Storage conditions affect the ability of the carbon film to separate from the mica in subsequent step 9. Generally, the carbon will not separate immediately after evaporation and should be stored for at least a couple of days. Separation may be helped by equilibrating the mica in a humid atmosphere immediately before floating. 8. Grids can be cleaned if desired by bath sonication in a small beaker filled with acetone. This removes traces of machine oil and other contaminants that may be present. After sonication, decant bulk acetone, then turn the beaker upside down on a piece of filter paper. As the grids dry, they tend to drop onto the filter paper, especially after tapping the bottom of the beaker. 9. Grids should be placed on the filter paper in a consistent way, usually with the polished (shiny) side up. This removes ambiguity about which side the carbon is on. Grids can be purchased with a Rh coating on one side, which makes it easy to distinguish the two different sides. 10. Two parameters should be explored to obtain the desired size and distribution of holes: amount of glycerol added to the Formvar solution, and incubation time between sonication and dipping of the glass slide. More glycerol solution and longer incubation times produce larger holes. If very few holes separated by a large distance is desired, use lower amounts of

204

Stokes et al.

glycerol and longer incubation times (e.g., Fig.Â€3b). This method was previously described by Baumeister and Seredynski (16). An alternative method for producing holey films is described by Fukami and Adachi (17). 11. Glow discharge apparatuses vary greatly. You should check with local EM staff for detailed operating instructions. For negatively charged proteins, it may be helpful to have a small amount of amylamine vapor inside the glow discharge apparatus. This will tend to make the carbon film positively charged. Glow discharged grids are good for about 1 day, after which they will become hydrophobic again. 12. There are many possible stains, and a book by Robin Harris provides an excellent reference (18): ●●

●●

●●

●●

Uranyl acetate: Most commonly used stain. pH, generally not adjusted, is 4.5. Stock solution lasts for months. Uranyl formate: A finer grain stain than uranyl acetate, but must be freshly prepared. Sodium phosphotungstate: Can be titrated to neutral or basic pH. Ammonium molybdate: pH 6.0–7.0. Much less contrast than heavy metal salts.

13. If all or most carbon squares are broken, then glow discharge may have been too aggressive. Check grids in light microscope before and after glow discharge to see if this was a problem. If not, poor sample handling, application of too much sample or of a sample composed of very coarse precipitate may result in broken carbon. If most squares are nearly or completely black, sample may be too concentrated and should be diluted. Alternatively, a wash step can be added before the first stain step. After removing sample, pipette 5Â€ml of buffer onto grid, remove with filter paper, and proceed with staining. 14. A low-dose software kit on the electron microscope is very helpful at this point. This kit will allow you to jump between screening magnification and imaging magnification. It also facilitates focusing on an area adjacent to the feature of interest, thus minimizing electron radiation and helping to preserve high-resolution features. 15. CCD imaging is preferred for screening samples because (1) an unlimited number of images can be taken in any given session and (2) Fourier transforms can be calculated immediately, thus allowing rapid evaluation of image quality and crystalline order. A large format, high-sensitivity CCD is not needed for screening; a 1 mega pixel format is generally sufficient, allows for faster image readout, and reduces data storage requirements.

Two-Dimensional Crystallization of Integral Membrane Proteins

205

Acknowledgement This work was supported by NIH grant R01 GM81817 and NSF grant MCB 0546087. References 1. Henderson R, Unwin PNT (1975) Threedimensional model of purple membrane obtained from electron microscopy. Nature 257:28–32 2. Henderson R, Baldwin JM, Ceska TA, Zemlin F, Beckmann E, Downing KH (1990) Model for the structure of bacteriorhodopsin based on high-resolution electron cryo-microscopy. J Mol Biol 213:899–929 3. Subramaniam S, Hirai T, Henderson R (2002) From structure to mechanism: electron crystallographic studies of bacteriorhodopsin. Philos Transact A Math Phys Eng Sci 360:859–874 4. Stahlberg H, Fotiadis D, Scheuring S, Remigy H, Braun T, Mitsuoka K, Fujiyoshi Y, Engel A (2001) Two-dimensional crystals: a powerful approach to assess structure, function and dynamics of membrane proteins. FEBS Lett 504:166–172 5. Unger VM (2001) Electron cryomicroscopy methods. Curr Opin Struct Biol 11:548–554 6. Gonen T, Cheng Y, Sliz P, Hiroaki Y, Fujiyoshi Y, Harrison SC, Walz T (2005) Lipid-protein interactions in double-layered two-dimensional AQP0 crystals. Nature 438:633–638 7. Kühlbrandt W (1992) Two-dimensional crystallization of membrane proteins. Q Rev Biophys 25:1–49 8. Yeager M, Unger VM, Mitra AK (1999) Threedimensional structure of membrane proteins determined by two-dimensional crystallization, electron cryomicroscopy, and image analysis. Methods Enzymol 294:135–180 9. Jap BK, Zulauf M, Scheybani T, Hefti A, Baumeister W, Aebi U, Engel A (1992) 2D crystallization: from art to science. Ultramicroscopy 46:45–84 10. Rasband W (1997–2008) ImageJ. U.S. National Institutes of Health, http://rsb.info. nih.gov/ij/ 11. Lefman J, Morrison R, Subramaniam S (2007) Automated 100-position specimen loader

and image acquisition system for transmission electron microscopy. J Struct Biol 158: 318–326 12. Cheng A, Leung A, Fellmann D, Quispe J, Suloway C, Pulokas J, Abeyrathne PD, Lam JS, Carragher B, Potter CS (2007) Towards automated screening of two-dimensional crystals. J Struct Biol 160:324–331 13. Vink M, Derr K, Love J, Stokes DL, Ubarretxena-Belandia I (2007) A highthroughput strategy to screen 2D crystallization trials of membrane proteins. J Struct Biol 160:295–304 14. Rigaud J-L, Levy D, Mosser G, Lambert O (1998) Detergent removal by non-polar polystyrene beads: applications to membrane protein reconstitution and two-dimensional crystallization. Eur Biophys J 27:305–319 15. Young HS, Rigaud J-L, Lacapere J-J, Reddy LG, Stokes DL (1997) How to make tubular crystals by reconstitution of detergent-solubilized Ca2+-ATPase. Biophys J 72:2545–2558 16. Baumeister W, Seredynski J (1976) Preparation of perforated films with predeterminable hole size distributions. Micron 7:49–54 17. Fukami A, Adachi K (1965) A new method of preparation of a self-perforated micro plastic grid and its application (1). J Electron Microsc 14:112–118 18. Harris JR (1997) Negative staining and cryoelectron microscopy. BIOS Scientific, Oxford, UK 19. Hu M, Vink M, Kim C, Derr K, Koss J, D’Amico K, Cheng A, Pulokas J, UbarretxenaBelandia I, Stokes D (2010) Automated electron microscopy for evaluating twodimensional crystallization of membrane proteins. J Struct Biol in press 20. Kim C, Vink M, Hu M, Love J, Stokes DL, Ubarretxena-Belandia I (2010) An automated pipeline to screen membrane protein 2D crystallization. J Struct Funct Genomics in press

as

Chapter 11 Structure Determination of Membrane Protein by Both Cryo-Electron Tomography and Single Particle Analysis Sylvain Trépout, Jean-Christophe Taveau, and Olivier Lambert Abstract The structure determination of membrane protein in lipid environment can be carried out using cryo-electron microscopy combined with the recent development of data collection and image processing. We describe a protocol to study assemblies or stacks of membrane protein reconstituted into a lipid membrane using both cryo-electron tomography and single particle analysis, which is an alternative approach to electron crystallography for solving 3D structure. We show the organization of the successive layers of OprM molecules revealing the protein–protein interactions between OprM molecules of two successive lipid bilayers. Key words: Cryo-electron tomography, Membrane protein, Multidrug efflux pump, Single particle analysis

1. Introduction Electron microscopy (EM) and cryo-electron microscopy (cryo-EM) are used to image ultrastructure of tissues and cells as well as to provide information on the structure of purified molecules including membrane proteins. Cryo-EM enables the visualization of the specimen embedded in amorphous ice that represents a state close to the native state. To study membrane protein in a lipid environment, structural analysis is performed on membrane protein assembled in twodimensional (2D) arrays either occurring naturally or induced after reconstitution into a lipid membrane. Unilamellar 2D arrays of membrane protein are amenable to electron crystallography studies. Instead, stacks of 2D arrays protein or multilamellar crystal are less suitable mainly because of complications to characterize Jean-Jacques Lacapère (ed.), Membrane Protein Structure Determination: Methods and Protocols, Methods in Molecular Biology, vol. 654, DOI 10.1007/978-1-60761-762-4_11, © Springer Science+Business Media, LLC 2010

207

208

Trépout, Taveau, and Lambert

the unit cell and process the data, the number of layers and their geometry being difficult to estimate despite few attempts (1, 2). Both single particle analysis and cryo-electron tomography (cryo-ET) are also useful approaches for the structural analysis of biological components by transmission EM. Classically, single particle analysis enables to determine the three-dimensional (3D) structures of relatively large proteins and macromolecular complexes from a large set of images, assuming that all the samples have the same architecture. Single particle analysis needs to determine the Euler angles of each particle randomly oriented in ice. Then the final 3D density map is computed not from a single particle but from thousands of particles. With this method, the specimen is illuminated only once, but the determination of the Euler angles is a key issue (random conical tilt method, common line approach, etc.). Purified membrane proteins or assemblies into small regular structures have been studied by 3D negative-stain EM and cryo-EM using single particle method (3 see references inside; 4â•›–7). There is no notable difference between membrane protein and soluble protein methodologies except the fact that the detergent associated to the membrane protein and the detergent in solution also contribute to the images that require careful attention for their interpretation especially in the membrane region. Electron tomography (ET) is often used to visualize 3D subcellular structures and supramolecular assemblies (8) at molecular resolution after collecting a tilt series of a unique object. In ET, the specimen is illuminated multiple times from various different view angles to compute a 3D structure. Examples of recent applications of ET to negatively stained supramolecular assemblies include studies of fibrillin-rich microfibrils and the corneal collagen fibril structure (9, 10) or frozen-hydrated proteoliposome– bacteriophage complexes (11) leading to the overall description of the assembly. To get more precise structural details, ET has been combined with image analysis tools. Over the last decade, tomograms computed from freeze-substituted sections (12, 13), from frozen-hydrated specimens (14–18), and from cryo-sections (19) were then subjected to a classification procedure identifying local variances, followed by class averaging and model building. Thus, combining both cryo-ET and single particle analysis appears as a new trend in membrane protein structure at nanometer resolution. In this chapter, OprM represents a typical case to illustrate this methodology because this transmembrane protein – component of the tripartite multidrug efflux pump MexAB-OprM of Pseudomo nas aeruginosa (20) – forms small stacks of 2D arrays after incorporation into lipid membrane that is a limitation for structural studies using electron crystallography (21). Thus, we aim to combine cryo-ET and single particle analysis to study the overall

Structure Determination of Membrane Protein by Both Cryo-Electron Tomography

209

Fig.Â€1. General scheme of the procedure for studying membrane protein structure arranged in a complex assembly (i.e., proteoliposome stacks) using cryo-electron tomography and single particle analysis. (1) Reconstitution of purified membrane protein into a lipid bilayer after detergent removal leads to the formation of flattened proteoliposomes and induces stacks due to protein–protein interactions (unsuitable for electron crystallography study). (2) These membrane protein assemblies are amenable for 3D cryo-electron tomography studies, provided their sizes are comprised in range of few hundreds of nanometer. A tilt series is automatically collected. A tomogram of the overall 3D architecture is then computed using a weighted backprojection method. (3) To improve the signal-to-noise ratio and the resolution, subtomograms, each containing a single membrane protein is 3D aligned (in translation and in rotation) to calculate an average of the membrane protein 3D structure.

organization of OprM membrane protein. The purpose of the present methodological work, summarized in Fig.Â€1, is to widen the scope of cryo-ET/single particle analysis to the specimen made of membrane protein complexes forming self-assemblies with irregular size and geometry unsuitable for electron crystallography.

2. Materials 1. Detergent: Octyl b d glucopyranoside (bOG; Sigma-Aldrich, Saint-Quentin Fallavier, France). 2. Proteins extraction buffer A: 20Â€mM Tris–HCl, pH 8, 10% glycerol (v/v), 15Â€mM imidazole, 2.5% bOG (w/v). 3. Lipid: Dioleoyl phosphadidyl choline (DOPC; Avanti Polar Lipids, Alabaster, AL, USA).

210

Trépout, Taveau, and Lambert

4. Detergent removal with SM2 BioBeads (Bio-Rad, Marne la Coquette, France). 5. Specimen preparation for cryo-TEM: Holey carbon copper grid (300 mesh, Ted Pella, Redding, CA) and EM CPC workstation for specimen vitrification (bare grid method, Leica Vienna, Austria). 6. Transmission electron microscope: FEI Polara (equipped with Field Emission Gun, FEG) operated at 300Â€kV, goniometer allowing rotation from −70° to +70° and a camera 4Â€kâ•›×â•›4Â€k (Gatan, Inc., Pleasanton, CA). 7. Image acquisition program: UCSF tomography software freely available (see ref 22). 8. Computer for image analysis: Standard workstation or PC-based computer. Just be aware that calculation of tomograms, such as 2,048â•›×â•›2,048â•›×â•›1,024, can be usual and then generate large files. It is recommended to have for convenience 1Â€ GB RAM, 1Â€ TB hard disk, and modern graphics card.

3. Methods 3.1. Protein Purification

As a standard scheme, membrane protein purification starts with the total solubilization of protein-enriched membranes by detergent. Then, after several purification steps, the protein is purified at a high concentration and maintained soluble in the presence of detergent. The transmembrane protein OprM (MM 50Â€ kDa) expression and purification follows this standard method (21). Briefly, the membrane envelopes from broken Escherichia coli cells were solubilized in buffer A overnight at 20°C. The solubilized membrane proteins were loaded onto a Ni-NTA resin column and then were eluted with a linear gradient of imidazole (60–500Â€mM). The fractions containing the OprM protein were pooled and concentrated to 5Â€ mg/ml. Finally, OprM was exchanged for suitable buffer by dialysis in the presence of buffer A without imidazole. OprM is trimeric in solution and forms a cylindrical structure composed of a transmembrane b barrel and 12 long a helices protruding to the solvent.

3.2. Membrane Protein Reconstitution

Various strategies have been proposed to incorporate purified membrane proteins into lipid bilayer such as organic solventmediated reconstitutions, mechanical means, spontaneous incorporation in preformed liposomes, and detergent-mediated reconstitutions (for a review (23)). The latter is a soft procedure suitable for fragile proteins and consists in the detergent removal

Structure Determination of Membrane Protein by Both Cryo-Electron Tomography

211

from a micellar solution composed of detergent, lipid and protein. Four main techniques for detergent removal are available: dilution, dialysis, gel-filtration and the use of polystyrene beads. OprM reconstitution in proteoliposomes was carried out after detergent removal using polystyrene beads. OprM proteins (4Â€µg/ml) were mixed with DOPC lipids (10Â€ µg/ml) and 30Â€ mM bOG under gentle stirring for 30Â€min at 4°C. The sample was then treated with three successive additions of 50Â€ mg/mg SM2 BioBeads according to the batch procedure described by Rigaud etÂ€al. (23). After 3Â€ h of incubation, the reconstitution material was then pipetted off. Freshly prepared samples were used subsequently for cryo-ET experiments and were kept at 4°C for further studies. 3.3. Cryo-EM and Tilt Series Acquisition

Samples for cryo-EM and cryo-ET are prepared according to Dubochet etÂ€al. (24) (Fig.Â€2). 1. A small droplet of the reconstituted protein solution (3Â€µl) is deposited on a holey carbon microscopy grid of copper. For cryo-ET, the grid is pretreated with the deposition of few µl of a 10-nm gold bead solution (fiducial markers) and let it dry (Fig.Â€2e). 2. The grid is then mounted on a “guillotine” device hung by tweezers. 3. The excess of the solution is absorbed on a filter paper (Whatman Filter Paper), and subsequently the grid is plunged into a bath of liquid ethane, cooled down at liquid nitrogen temperature. To prevent damages in the preserved structures of the sample, the microscopy grid is maintained at liquid nitrogen temperature during the whole transfer process to the electron microscope. The grid is inserted into the microscope using a single tilt cryo holder. 4. A tilt series of images is automatically acquired via software controlling the electron microscope (various proprietary and academic packages are available for this task). The classical collection scheme is linear with an increment step with a range of 1–3° from −70° to +70°. Each image is collected in low dose mode, and a total electron dose should be lesser than 200Â€e−/Å2 for limiting the beam damage.

3.4. Tomogram Generation from Cryo-Tilt Series

The observations of membrane proteins in situ are only available with cryo-ET. However, the main drawbacks of ice-embedded specimen are the high sensitivity to the radiation damage and a very poor contrast (to minimize the beam-induced damage, the illumination must be at extraordinary low dose, i.e., ~â•›0.2Â€e−/Å2). Likewise due to mechanical constraints (limited tilt angle range) and poor signal-to-noise ratio of tomograms, the averages of membrane proteins are required to substantially increase the

212

Trépout, Taveau, and Lambert

Fig.Â€2. Cryo-EM of OprM-containing proteoliposomes stacks. (a) Side view of typical small stacks of proteoliposomes that possess a preferential orientation within the ice layer. (b) Enlarged view showing OprM molecules sandwiched between lipid membranes separated by a distance of 13Â€nm (four horizontal lipid bilayers are visible). (c) Top view of large OprM stacks laying flat on the carbon grid. The moiré pattern associated to protein densities reveals a regular assembly of OprM molecules, but their arrangement is only unveiled from the calculation of the 3D architecture. (d) Other image of small stacks of OprM proteoliposomes. (e) For cryo-electron tomography, stacks with an oblique orientation were selected to give access to the top view. The white square delineates the area used for tomogram alignment in Fig.Â€3. Scale bars 100Â€nm (a), 20Â€nm (b, c) and 50Â€nm (d, e).

Structure Determination of Membrane Protein by Both Cryo-Electron Tomography

213

resolution. Thus, the procedure is composed of two parts: (1) the tomogram generation and (2) the averaging of subtomograms containing the objects of interest (e.g., the membrane protein). Several tomography software packages are available and the choice of one of them depends on many criteria such as your computer environment (operating system, parallel computer/ cluster, etc.). For the sake of clarity in the rest of this section, the description of various steps of tomogram generation will be completed with the commands/tools of IMOD software (25). The use of other software should require few adaptations. 1. Setup of tomography software (in IMOD, use etomo a graphical user-friendly interface) by entering name of your image series, gold particle diameter, and location of the tilt angles file (see Note 1). 2. A preprocessing step is required to remove aberrant pixel values (due to X-ray) in the image series. 3. A coarse alignment is carried out by cross-correlation and allows the visualization of roughly aligned image series by using the “Movie” feature of IMOD::3dmod (visualization tool). At this stage, you are able to appreciate the quality of your series (Fig.Â€3a, b) (see Note 2). 4. The crucial step in tomogram generation is the semiautomatic procedure consisting of alignment refinement by using gold particles as fiducial markers. 5. In a first manual step – in the 0° tilt image (median image in the series) – select precisely (by clicking the particle center with the middle button of mouse pointer and press the “N” key) as many as possible gold particles while checking that the marker will not be hidden/overlapped/disappeared in the other images of the series. In the IMOD philosophy, this initialization step defines the seeds for the automatic tracking. 6. Then, the software automatically searches the positions of the seeds in the rest of the image series (IMOD::beadtracker tool) (Fig.Â€3c). 7. Finally, the positions of the fiducials (IMOD::beadfixer tool) are adjusted to minimize the variations in the whole series. Do as many cycles as the software proposes and check that the refined positions are well centered (Fig.Â€3d) (see Note 3). 8. To optimize the overall size of the final tomogram, the software requires bounding box positions to compute various offsets and angles. (In IMOD, this step is done by first calculating three small tomograms allowing the drawing of the upper and lower bounding box) (see Note 4). 9. Compute the final aligned image series and visualize the result using a “Movie” mode to check that the animation is fluent

214

Trépout, Taveau, and Lambert

Fig.Â€3. Tomogram generated from one tilt series of an OprM proteoliposomes stack. Tilt axis alignment is required after acquisition due to various drift effects (e.g., thermal drift, mechanical stage movements) that is performed in two steps. (a) Ninety-two raw images with an increment step of 1.5° are collected with UCSF tomography acquisition software (22). For the sake of clarity, only a series of 10 superimposed images from −30° to +30° is presented. Black dots correspond to fiducial markers (10Â€ nm gold beads). (b) Coarse alignment using cross-correlation function provides a reliable alignment of the images along the tilt axis (long white double arrow). Fiducial markers create linear traces perpendicular to the tilt axis. (c) Automatic tracking for a more accurate alignment. Black arrows indicate the seed positions of fiducial markers. Black circles point out the results of automatic tracking. (d) Final alignment with tilt axis oriented vertically. (e) Stereo views of calculated tomogram depicted by an oblique slice revealing the regular assembly of OprM molecules. Isosurfaces of gold beads show fiducial markers.

showing no shift between images (adjust the animation speed for a better impression). 10. Compute the 3D reconstruction (Fig.Â€3e) (see Note 5). 3.5. Single Particle Analysis

The single particle analysis consists in (1) extracting subtomograms, (2) aligning them (rotation and translation applied in 3D) and (3) calculating averages. As these procedures are intensively used in single particle 3D reconstruction, we recommend the use of software usually devoted for these tasks. Here, examples of commands of SPIDER (26) and of Xmipp (27) are presented.

Structure Determination of Membrane Protein by Both Cryo-Electron Tomography

215

1. Using a 3D visualization tool displaying simultaneously the three orthogonal planes (IMOD::3dmod using concomitantly “XYZ” and “ZaP” views ), accurately select the particles (see Note 6) and save the coordinates’ file (in IMOD, this file is called a “model” and convert it in ASCII format with IMOD::imodinfo -a) (Fig.Â€4). 2. Extraction of the subtomograms from the coordinate files: For the sake of convenience during the 3D alignment process, the extracted subtomograms must be cubic and thus, if your proteins have an elongated shape (like OprM), you must choose a larger bounding box including several membrane proteins (herein, one centralâ•›+â•›six neighboring OprM molecules were extracted per subtomograms) (Fig.Â€5). The script must be organized as follows: for each coordinate triplet, calculate the upper-left corner and extract (SPIDER::WI command) the subtomogram. 3. Preprocessing of the subtomograms: Most single particle software require volumes with inverted contrast and positive

Fig.Â€4. Extraction of subtomograms from the tomogram. For the subtomogram picking, the protein layers are oriented in X–Y plane. Then, slices along Z axis reveal specific and recognizable patterns allowing to discriminate between protein and lipid layers. Five equidistant 0.4-nm-thick slices clearly show densities typically corresponding to welldefined cylindrical OprM molecules (1, 5) and to moiré pattern ascribed to the protein– protein interactions located in the lipid bilayer (2, 4) and in extramembrane moiety (4). Each of the 347 subtomograms contains a central OprM with its six neighbors to fit into a cubic bounding box more convenient for 3D alignment operations as schemed with the black box.

216

Trépout, Taveau, and Lambert

Fig.Â€5. Exploration of the average volume of OprM molecules. (a) Electron density map calculated from OprM X-ray atomic structure (PDB ID: 1WP1; (20)) (XMIPP::xmipp_ convert_pdb2descr and XMIPP::xmipp_phantom_create) and low-pass filtered up to 30â•›Å (XMIPP::xmipp_fourier_filter). The molecule is a trimer composed of a b barrel transmembrane domain and a 12-helice domain. The density map is then fitted within the subtomogram average (resolution limit estimated to 30â•›Å using FSC) allowing the construction of a 3D model. (b) A series of 0.4-nm-thick Z-slices of the average volume spaced 1.6Â€nm apart. (c–f) OprM models are represented as isosurfaces and fitted into the subtomogram average whose densities are drawn as isocontour lines. (c) Bottom layer of three OprM (central densities of slice #1 in (b)) with their transmembrane domains facing up. (d) Central layer of OprM with their transmembrane domains facing down. The isolines in 3D views (c and d) correspond to the location of one lipid bilayer (slice #2 in (b)) with a characteristic moiré pattern resulting from densities of the transmembrane domains of the bottom and central layers of OprM molecules. (e) Top layer of

Structure Determination of Membrane Protein by Both Cryo-Electron Tomography

217

pixel values (light protein densities on dark background) (SPIDER::AR) and corrected ramp effect (SPIDER::RA). 4. Choice of a 3D reference subtomogram: Various strategies are possible concerning the 3D reference to limit its influence along the alignment process: (1) several runs of 3D alignments using different reference volumes are carried out until the result appears stable and (2) you assume that the picking is rather good and the particles are approximately centered; therefore, run a 3D rotational alignment to compute an average used as reference for next cycles of 3D alignment. 5. To align the whole set of subtomograms, several cycles of successive rotation and translation operations executed in 3D update the orientation (three rotations) and position (three translations) of the subtomograms. At each end of a 3D alignment cycle, compute the average, which will be used as a new reference for the next cycle. –â•ﬁ

3D Rotational alignment requires the determination of three Euler angles (f, q, y). A brute force approach consists in scanning the whole Euler space to find the best correlation between the reference and the subtomogram. In certain circumstances (a priori knowledge of protein orientation), this search may be limited to a part of Euler space. A faster approach is to use programs searching function minimization (SPIDER::OR 3Q based on quasi-Newtonian method). For the latter approach, try several initial guesses to find the best correlation.

–â•ﬁ

3D Translational alignment is carried out by 3D crosscorrelation (SPIDER::CC 3 and SPIDER::PK 3D) (see Note 7).

6. Merge the geometric transformations (rotations and translations) to those obtained in the previous cycle (4â•›×â•›4-matrix calculation; SPIDER::SA P) and apply them to the original subtomograms. Finally, compute an average of the aligned subtomograms (SPIDER::AS R) (Fig.Â€5b). This average will be used as reference for a new cycle. 7. Resolution determination. Split the aligned set of tomograms in two parts and compute two average volumes. Then compare these two averages using a Fourier Shell Correlation (SPIDER::RF 3).

Fig.Â€ 5. (continued) OprM interacting with the central proteins through their helical domains leading to a different moiré pattern (slice #7 in (b)). (f) The top isocontour lines represent the second lipid bilayer (slice #11 in (b)). This view summarizes the complete assembly of OprM layer between two lipid bilayers and is in good agreement with our previous electron crystallography results (21).

218

Trépout, Taveau, and Lambert

4. Notes 1. Depending on your acquisition of the software, you need to check the value of Y-tilt axis angle either given in clockwise or counter-clockwise. If you do not know the right sign of your angle, the software will be unable to process a correct alignment. In this case, restart the session (remove all the script files) and change the sign of tilt angle. 2. If the images of high tilt are not well aligned or are of poor quality (drifted, too thick, etc.), it is reasonable to remove them for the image series knowing that high tilt images are crucial in the quality of the final tomogram. For the sake of convenience, we suggest to create a new truncated image series (IMOD::newstack), to delete the script files and to rerun a new session. 3. During the semiautomatic procedure of position refinement, it can occur that one seed is not well tracked (likely because the gold particle has moved during the acquisition) do not hesitate to remove it, choose another one if possible and entirely restart step 4. 4. In the specific case of membrane protein reconstitution, it is often difficult to observe the boundaries of the tomogram compared to cell sections and thus, by increasing the thickness of the sample tomograms should help in defining them. 5. The implementation of weighted backprojection algorithm (the most classical algorithm for this kind of 3D reconstruction) in various tomography softwares may differ in the Point Spread Function definition and thus the resulting tomogram may appear more or less satisfying. Thus, compute 3D reconstructions with different software and choose the most appropriate. 6. Due to the “missing wedge” of the whole tomogram (missing data in Fourier space due to the limited tilt angle range of ±70° of data collection), the objects of interest are not well defined along the Z axis (perpendicular to the X Y-plane) and therefore, for specific orientations, the picking in the particle center may be difficult. For membrane proteins, a way to circumvent this problem is to use the membrane bilayer as a guide or to choose an oblique view revealing the proteins end-on orientation (Fig.Â€4). 7. In case of erroneous 3D alignment, constrained cross-correlation should be tried because this approach (28, 29) takes into account the “missing wedge” of subtomograms during computation.

Structure Determination of Membrane Protein by Both Cryo-Electron Tomography

219

Acknowledgments We would like to thank Dr. Achilleas Frangakis for fruitful discussions and for giving us access to the EM facilities of EMBL (Heidelberg) and Pr. A. Ducruix for providing purified membrane protein samples. S. Trépout is recipient of PhD fellowship from French Ministry of Education and Research and Technology (MENRT). This work has been supported in part by Agence Nationale de la Recherche (ANR-06-PCV1-001) and by Conseil Régional d’Aquitaine (20071302007) grants. References 1. Cheong GW, Young HS, Ogawa H, Toyoshima C, Stokes DL (1996) Lamellar stacking in three-dimensional crystals of Ca(2+)-ATPase from sarcoplasmic reticulum. Biophys J 70:1689–1699 2. Shi D, Lewis MR, Young HS, Stokes DL (1998) Three-dimensional crystals of Ca2+-ATPase from sarcoplasmic reticulum: merging electron diffraction tilt series and imaging the (h, k, 0) projection. J Mol Biol 284:1547–1564 3. Rubinstein JL (2007) Structural analysis of membrane protein complexes by single particle electron microscopy. Methods 41:409–416 4. Gregorini M, Wang J, Xie XS, Milligan RA, Engel A (2007) Three-dimensional reconstruction of bovine brain V-ATPase by cryoelectron microscopy and single particle analysis. J Struct Biol 158:445–454 5. Nakagawa T, Cheng Y, Ramm E, Sheng M, Walz T (2005) Structure and different conformational states of native AMPA receptor complexes. Nature 433:545–549 6. McDevitt CA, Collins RF, Conway M, Modok S, Storm J, Kerr ID, Ford RC, Callaghan R (2006) Purification and 3D structural analysis of oligomeric human multidrug transporter ABCG2. StructureÂ€14:1623–1632 7. Chami M, Steinfels E, Orelle C, Jault JM, Di Pietro A, Rigaud J-L, Marco S (2002) Threedimensional structure by cryo-electron microscopy of YvcC, an homodimeric ATPbinding cassette transporter from Bacillus subtilis. J Mol Biol 315:1075–1085 8. Baumeister W (2002) Electron tomography: towards visualizing the molecular organization of the cytoplasm. Curr Opin Struct Biol 12:679–684 9. Shuttleworth CA, Kielty CM (2001) The supramolecular organization of fibrillin-rich microfibrils. J Cell Biol 152:1045–1056

10. Holmes DF, Gilpin CJ, Baldock C, Ziese U, Koster AJ, Kadler KE (2001) Corneal collagen fibril structure in three dimensions: structural insights into fibril assembly, mechanical properties, and tissue organization. Proc Natl Acad Sci 98:7307–7312 11. Bohm J, Lambert O, Frangakis AS, Letellier L, Baumeister W, Rigaud JL (2001) FhuAmediated phage genome transfer into liposomes: a cryo-electron tomography study. Curr Biol 11:1168–1175 12. Winkler H, Taylor KA (1999) Multivariate statistical analysis of three-dimensional crossbridge motifs in insect flight muscle. Ultramicroscopy 77:141–152 13. Chen LF, Blanc E, Chapman MS, Taylor KA (2001) Real space refinement of acto-myosin structures from sectioned muscle. J Struct Biol 133:221–232 14. Förster F, Medalia O, Zauberman N, Baumeister W, Fass D (2005) Retrovirus envelope protein complex structure in situ studied by cryo-electron tomography. Proc Natl Acad Sci USA 102:4729–4734 15. Bostina M, Bubeck D, Schwartz C, Nicastro D, Filman DJ, Hogle JM (2007) Single particle cryoelectron tomography characterization of the structure and structural variability of poliovirus-receptor-membrane complex at 30â•›Å resolution. J Struct Biol 160:200–210 16. Zanetti G, Briggs JA, Grünewald K, Sattentau QJ, Fuller SD (2006) Cryo-electron tomographic structure of an immunodeficiency virus envelope complex in situ. PLoS Pathog 2:e83 17. Zhu P, Liu J, Bess J Jr, Chertova E, Lifson JD, Grisé H, Ofek GA, Taylor KA, Roux KH (2006) Distribution and three-dimensional structure of AIDS virus envelope spikes. Nature 441:847–852

220

Trépout, Taveau, and Lambert

18. Strauss M, Hofhaus G, Schröder RR, Kühlbrandt W (2008) Dimer ribbons of ATP synthase shape the inner mitochondrial membrane. EMBO J 7:1154–1160 19. Al-Amoudi A, Díez DC, Betts MJ, Frangakis AS (2007) The molecular architecture of cadherins in native epidermal desmosomes. Nature 450:832–837 20. Akama H, Kanemaki M, Yoshimura M, Tsukihara T, Kashiwagi T, Yoneyama H, Narita S, Nakagawa A, Nakae T (2004) Crystal structure of the drug discharge outer membrane protein, OprM, of Pseudomonas aeruginosa: dual modes of membrane anchoring and occluded cavity end. J Biol Chem 279:52816–52819 21. Lambert O, Benabdelhak H, Chami M, Jouan L, Nouaille E, Ducruix A, Brisson A (2005) Trimeric structure of OprN and OprM efflux proteins from Pseudomonas aeruginosa, by 2D electron crystallography. J Struct Biol 150: 50–57 22. Zheng QS, Braunfeld MB, Sedat JW, Agard DA (2004) An improved strategy for automated electron microscopic tomography. J Struct Biol 147:91–101 23. Rigaud J-L, Mosser G, Lacapere J-J, Olofsson A, Levy D, Ranck J-L (2005) Bio-Beads: an

24.

25. 26.

27.

28.

29.

efficient strategy for two-dimensional crystallization of membrane proteins. J Struct Biol 118:226–235 Dubochet J, Adrian M, Chang JJ, Homo JC, Lepault J, Mc Dowall AW, Schultz P (1988) Cryo-electron microscopy of vitrified specimens. Q Rev Biophys 21:129–228 Mastronarde DN (1997) Dual-axis tomography: an approach with alignment methods that preserve resolution. J Struct Biol 120:343–352 Frank J, Radermacher M, Penczek P, Zhu J, Li Y, Ladjadj M, Leith A (1996) SPIDER and WEB: processing and visualization of images in 3D electron microscopy and related fields. J Struct Biol 116:190–199 Sorzano CO, Marabini R, Velázquez-Muriel J, Bilbao-Castro JR, Scheres SH, Carazo JM, Pascual-Montano A (2004) XMIPP: a new generation of an open-source image processing package for electron microscopy. J Struct Biol 148:194–204 Förster F, Pruggnaller S, Seybert A, Frangakis AS (2008) Classification of cryo-electron subtomograms using constrained correlation. J Struct Biol 161:276–286 Schmid MF, Booth CR (2008) Methods for aligning and for averaging 3D volumes with missing data. J Struct Biol 161:243–248

Chapter 12 Electron Microscope Tomography of Native Membranes Gabriel Péranzi, Cedric Messaoudi, Leeyah Issop, and Jean-Jacques Lacapère Abstract Membrane proteins are often present in low amounts in cells. Their function can be modulated by interactions with other proteins. Moreover, these complexes can be transiently formed, thus making them difficult to be isolated and to be purified. One way to overcome these difficulties is to visualize these complexes in situ in the cells. For such purpose, electron microscopy coupled to tomography is a promising approach that has been developed over the last decades. Mitochondria are a good example of organelles where many membrane proteins form different functional complexes within the outer and the inner membranes. The latter is either close to the former or projects within the matrix to form cristae. Structure of these cristae involves different proteins and can vary from lamellar to tubular forms in normal mitochondria. In pathological conditions, other mitochondrial morphologies have been described, for instance, vesicular structures for inner boundary membrane have been observed. Key words: Mitochondrial ultrastructure, Resin embedded material, HT-29 cells, Inflammation, ImageJ, TomoJ, Chimera, Three-dimensional reconstruction

1. Introduction Many membrane proteins function in situ as protein complexes. These complexes may be formed by homo or hetero-functional polymers and be stable over long time periods, but they can also be due to transiently associated proteins. In this latter case, their purification is often difficult. Moreover, their amount in cells can be so low that it increases the difficulty to get enough amounts to determine their structures (1). Confocal microscopy permits cells imaging and membrane protein complexes characterization using fluorescence probes. However, resolution is limited to micron scale (2). Determination Jean-Jacques Lacapère (ed.), Membrane Protein Structure Determination: Methods and Protocols, Methods in Molecular Biology, vol. 654, DOI 10.1007/978-1-60761-762-4_12, © Springer Science+Business Media, LLC 2010

221

222

Péranzi et al.

of membrane protein atomic structure is mostly obtained with purified proteins and does not give information of their natural partners in their membrane environment. Electron tomography of tissues section, cells and organelles makes the link between the two types of observation (3, 4). In particular, protein–protein interactions from different membranes, such as those involved in mitochondria have gained from electron tomography (5–8). Details of the organization of internal membrane structure of mitochondria have been revealed to be more complicated than previously popularized in textbooks. This chapter describes protocols to get three-dimensional structures of mitochondria within cells by electron tomography using freely accessible programs. It compares structures observed in normal and pathological conditions.

2. Materials 2.1. Cells

1. Human colonic adenocarcinoma cell line HT-29 (passages 179–186) was kindly provided by Dr. C.L. Laboisse (Nantes, France). 2. Cells were grown in Dulbecco’s minimal essential medium (Invitrogen, Cergy Pontoiseand France) and supplemented with 10% fetal calf serum (Invitrogen, Cergy Pontoise, France) as previously described (9). 3. Cells were cultured for 24Â€h in the absence or in the presence of, 7.5Â€ng/ml TNF (Tumor Necrosis Factor, RD Systems Europe, Lille, France), for inflammatory stress conditions (10).

2.2. Sample Preparation for Electron Microscopy

1. Cell isolation buffer A: 50Â€mM cacodylate pH 7.4 incubated at 37.5°C. 2. Fixation in buffer B: Buffer A supplemented with 12.5Â€mM lysine, 5Â€mM EGTA and 50Â€mM glutaraldehyde. 3. Washing buffer C: 50Â€mM barbital pH 7.6. 4. Postfixation in buffer D: Buffer C supplemented with 1% v/v osmium. 5. Staining medium: 0.05Â€M uranyl acetate in water. 6. Solutions of increasing alcohol were 50, 70, 90 and 100% ethanol in water. 7. Epon mixture: 24Â€g LX112 resin solution, 16Â€g of dodenyl succinic anhydryde, 10Â€ g nadic methyl anhydride and 1.5 DMP-30 (Euromedex, Mundolsheim, France). 8. Preembedding medium: 50/50 1,2-epoxypropane and epon mixture.

Electron Microscope Tomography of Native Membranes

223

9. Sections (200Â€ nm, purple-red color): Ultramicrotome Ultracut E (Reichert-Leica, Rueil-Malmaison, France). 2.3. Transmission Electron Microscope 2.4. Image Acquisition and Processing

JEOL 1200EX equipped with LaB6 filament operated at 120Â€kV, goniometer allowing rotation from −60° to +60° and sample holder (JEOL EM-SQH10). 1. Lhesa camera (LH4036-B, Lheritier SA, StOuen l’aumone, France) with scintillator screen (low level of signal). Camera placed the JEOL screen. Averaging of images (DSP 1000 box, Lhesa). 2. Dell Precision 390 computer (Core duo E-6420 (2.66Â€GHz)), 8 Go RAM, Graphic card OPEN GL, Windows XP (64 bits) to acquire and to process images. 3. ImageJ, TomoJ and Chimera programs: ImageJ is a public domain, Java-based image processing program developed at the National Institutes of Health (11). TomoJ is a freeware processing program developed at the Curie Institute and University Pierre and Marie Curie (12). Chimera is a molecular graphics program, free of charges for academics (13) that can visualize three-dimensional volumes generated by TomoJ coupled ImageJ programs.

3. Methods 3.1. Sample Preparation

1. HT-29 cells in culture medium were washed briefly with buffer A stocked at 37°C. 2. Cells were fixed in buffer B (14) 1Â€h at room temperature. 3. Cells were washed twice with buffer C. 4. Cells were postfixed with buffer D for 15Â€min at 4°C. 5. Cells were washed three times with distilled water at 4°C. 6. Cells were stained with uranyl acetate solution overnight at 4°C. 7. Cells were washed with distilled water (to remove unbound uranyl acetate). 8. Cells were scrapped and dehydrated by incubation for 10Â€min in increasing alcohol solutions (50–90%). Cells were collected after, each incubation by 5-min centrifugation at 2,200â•›×â•›g. In a final step of dehydration, cells were incubated for 1Â€ h in 100% alcohol and then centrifuged 5Â€min at 3,000â•›×â•›g. 9. Cells were incubated in a solution of 1,2-epoxypropane for 15Â€min and then centrifuged 10Â€min at 3,000â•›×â•›g.

224

Péranzi et al.

10. Cells were incubated overnight in preembedding medium at 4°C and centrifuged 20Â€min at 3,600â•›×â•›g to collect cells. 11. Cells were incubated 4Â€h in pure epon at room temperature and polymerization was performed by incubation at 60°C overnight. 12. Blocks were cut in slices of 200Â€ nm that were collected on copper grids of 200 mesh (see Note 1). 3.2. Images Acquisition

1. Grids were observed with electron microscope, image recorded and saved as files in the computer. 2. A first selection of interesting cells with mitochondria has to be done. FigureÂ€ 1 shows HT-29 cells at low magnification and different types of mitochondria at higher magnification. The size and shape of these mitochondria vary and permit a classification in three types of morphologies: type 1 (normal, Fig.Â€1b), type 2 (normal–vesicular, Fig.Â€1c) and type 3 (vesicular, Fig.Â€1d). Further characterization of the internal structure of these mitochondria required third spatial dimension provided by electron microscope tomography. This can be

Fig.Â€1. Transmission electron images of slices of HT-29 cells. Panel (a), view of sectioned HT-29 cells at low magnification. Numerous cells are observed, their have a large nucleus (N) with a dense nucleolus (n) and mitochondria (m with arrows). Panel (b, c, and d), views of different types of mitochondria at high magnification.

Electron Microscope Tomography of Native Membranes

e−

225

e− 60° T nom. 200nm

T eff. 400nm

T nom Tilt angle (°)

0

20

40

50

60

70

80

Thickness (nm)

200

213

261

311

400

585

1152

Fig.Â€2. Schematic representation of electron beam crossing nontilted slice and highly tilted slice (60°). Table describes the increase of slice effective thickness (Teff) crossed by electron beam as a function of tilt angle (a). The variation of the effective thickness is linked to the nominal thickness (Tnom) by the following equation: Teffâ•›=â•›Tnom/cos a.

achieved by collecting images of the slices at different tilted angles. Most recent electron microscopes are equipped with sample holders that permit image acquisition at high tilt angles but several difficulties are raised in such conditions. 3. The thickness of the slice is an important factor since at high tilt angle the electron beam will have to cross a higher thickness (see Fig.Â€2.). The intensity of the electron beam is governed by the acceleration voltage of the microscope and thus the thicker section will require a stronger electron beam to get images. However, stronger is the electron beam, higher are the irradiation damages. Finally, the choice of the thickness of a section is a compromise. It takes into account the size of the object, the electron microscope and the maximal dose of irradiation that can be used for a section. This latter point can be minimized by using low-dose system that equipped most of the recent microscope (see Note 2). 4. The number of images acquired, the tilt angle and the angular increments have to be adjusted to optimize data acquisition for tomogram reconstruction and to reduce sample irradiation damage. For such purpose, higher the tilt angle of the sample, greater will be the information acquired. Previous work has analyzed the different strategies (15) for tilt increments. Briefly, this increment can be linear or even not linear. In this latter case, it uses cosine increment (also known as Saxton scheme) or an approximation of this increment presented in Fig.Â€3 with an angular increment of 5°, 2°, and 1° from 0° to 20°, 20° to 44°, and 44° to 60°, respectively.

226

Péranzi et al.

Fig.Â€3. Schematic representation of the tilt image series that have to be acquired to start image processing. The angular increment is not the same for the overall series. View of the same mitochondria acquired at each change of angular increment. Values in boxes indicate the different increments used for a tilt series from −60° to +60°.

Fig.Â€4. ImageJ and TOMOJ tool bar viewed by user.

3.3. Data Processing (ImageJ, TomoJ and Chimera Programs)

This process has been performed with the TomoJ program (see Fig.Â€4). This is the most important step to get good reconstruction (tomogram).

3.3.1. Image Alignment

1. Image acquired should be named and numbered such that the program will be able to be read and be made into a series (see Note 3).

Electron Microscope Tomography of Native Membranes

227

2. Import the image sequence in the ImageJ program (File/ Import/Image Sequence). 3. Open the TomoJ Interface 1.0).

program

(Plugins/TOMOJ/Tomoj

4. When asked, specify the tilt angle increment between the higher tilt image (starting) and the following one. Another possibility is the use of a separate file “home made” or gained from acquisition software. It is an ASCII file containing one line per image with the value of the corresponding tilt angle. You may also modify later the angles using the TOMOJ function (File/assign angles). In all cases, it is useful to save the rename series (IMAGEJ: File/save as/name.tif). 5. Contrast and brightness of each image can be improved to facilitate their alignment (IMAGEJ: Image/Adjust/Brightness/ Contrast). 6. A first alignment of the image series can be obtained using the Correct Shift function (specific button in TOMOJ panel). It has to be mentioned that this can be applied to nonsquared images, but the calculations will be faster with images containing power of 2Â€pixels in each dimensions. 7. A square box (with each dimension power of 2Â€ pixels) is applied to the series and around the chosen object (IMAGEJ: rectangular selections button) to reduce the data to be treated and increase the calculation rate. After checking that selected box contains the chosen object all along the image series, a crop is applied to extract the data (IMAGEJ: Image/Crop) and the new series saved (IMAGEJ: File/save as/name.tif). 8. A band-pass filter followed by a correct shift is applied to the crop series (specific buttons in TOMOJ panel). This first operation (filtering) largely improves the shift correction and thus the alignment of the images. 9. The parameters used for the shift of each image have to be saved in a separate file (specific button “save transforms” in TOMOJ panel, generating a file name.txt). 10. The quality of the alignment can be checked by loading the nonfiltered crop series (IMAGEJ: File/open/name.tif), applying the shifts (TOMOJ Panel: Load Transforms/name.txt) and visually testing the entire image sequence (for instance from −60° to +60°). If the alignment is correct, the tomogram reconstruction can start (see SubheadingÂ€3.3.2); otherwise, several methods can be applied to improve the alignment of the images. 11. The correct shift function can be applied (button in TOMOJ panel). To reduce the time of processing, a square box (power of 2Â€pixels) can be made in the center of the images.

228

Péranzi et al.

The center is reached using the cross (first, switch toolbar with the button in the TOMOJ panel, second, select the rectangular selections with button in IMAGEJ panel and third, place the cross in the center of the image with the mouse and check coordinates in the IMAGEJ panel). The square box is made by simultaneously pressing the ctrl button of the desktop and moving the mouse. The size of the box appears in the IMAGEJ panel. The new series is saved (IMAGEJ: File/save as/name.tif). If the alignment is improved, the tomogram reconstruction can start (see SubheadingÂ€3.3.2) otherwise, a new correction can be applied to improve the alignment of the images. 12. The correction rotation function can be applied (button in TOMOJ panel). Select the method for rotation determination in the option panel (TOMOJ Panel: Files/Options), usually the “classic in real space.” The corrected series are saved (IMAGEJ: File/save as/name.tif). If the alignment is improved, the tomogram reconstruction can start (see following section) otherwise; all the previous steps of correction can be re-applied to get an improved alignment of the images. One might also test other filters accessible in ImageJ. At the end of the alignment process, if one image cannot be aligned it might be easier to remove it, but it has to be noted that it will affect the resolution of the tomogram. 3.3.2. Tomograms Reconstructions

Different processes can be applied such as backprojection, weighted backprojection, ART or SIRT (16) and all present in the right part of the TOMOJ Panel. Whatever the process used, two values have to be determined (1) the tilt axis (see Note 4) and (2) the thickness of the reconstruction (see Note 5). 1. The tilt axis is a specific characteristic of the electron microscope and linked to the magnification used. It can be experimentally determined in a Fourier Transform (FT) of a vertically stacked aligned series (Fig.Â€5). Indeed, if the tilt axis is not aligned with the X or Y axis, a line of densities passing in the center of the FT can be observed in between the X and Y axes. Measuring the angle between this line and the vertical axis gives a value of the tilt axis. It has to be mentioned that no line will be observed if the tilt axis is along the X or Y axis. Briefly, a series is loaded (IMAGEJ: File/ open/name.tif), the images are vertically stacked (IMAGEJ: Image/Stack/Z-project and in the projection panel, choose the mean intensity) and the FT is processed (IMAGEJ: Process/FFT/FFT). 2. The thickness (Z in pixels) of the reconstruction can be calculated knowing the nominal thickness of the section (Tnom in nm),

Electron Microscope Tomography of Native Membranes

229

Fig.Â€5. Determination of tilt axis. Panels (a and c), stacks of two aligned tilted series acquired at different magnifications. Panels (b and d), Fourier Transform (FT) of the two stacked series. Dotted line shows tilt axis. Solid black line with arrows in Panels (a and c) shows hand draft 90° axis respective to tilt axis. White solid line represents the horizontal of the image. White arrows represent the angle needed in the TOMOJ program, i.e., between the tilt axis and the vertical of FT.

the magnification and the size (in pixel) of the acquired images. The Z value is calculated as follows: Zâ•›=â•›Tnom/cal, where cal (in nm per pixels) is the specific value of the acquired image. A calibration of the couple electron microscope and camera has to be performed to determine this value. 3. Backprojection (chosen in the rolling window in the right part of the TOMOJ panel), this reconstruction process is relatively faster than the others (in our case around 5–10Â€ min for a series of 65 images of 512â•›×â•›512 pixels). The projections are simply backprojected into the volume. This technique is known to produce quite bad reconstruction because of inhomogeneity of sampling in Fourier space. After computing the tomogram, the reconstruction can be saved for further visualization (IMAGEJ: File/save as/name.tif). It enables to check that (1) the images alignment is correct and (2) the tilt angle has correctly been determined. For instance, the image in the center of the reconstructed volume (Fig.Â€6b) should look as close as possible to the 0° tilt image (Fig.Â€6a). A fussy image would mean that the alignment has to be improved. An effect

230

Péranzi et al.

Fig.Â€6. Process of treatment of a tilt series of mitochondria. Panel (a), zero tilt image of the aligned series. Panels (b, c, d, and e), central image of the tomograms reconstructed by backprojection, weighted backprojection, ART and SIRT, respectively. Panels (f and g), error graph from ART and SIRT reconstructions, respectively.

of shearing in the image sequence is observed if the tilt axis is not correct. 4. Weighted backprojection (chosen in the rolling window in the right part of the TOMOJ panel), this reconstruction process is a little slower than backprojection as there is a weighting function applied to each pixel in each projection. This gives better results than backprojection (Fig.Â€ 6c) by limiting the overrepresentation of some information (low frequencies in Fourier space). This is the classical reconstruction algorithm in tomography. The drawback of this algorithm is that the weighting does not consider the presence of noise and so tends to amplify it. 5. ART (chosen in the rolling window in the right part of the TOMOJ panel), this reconstruction is performed in an iterative manner and is slow. The principle is to have a reconstruction

Electron Microscope Tomography of Native Membranes

231

volume (empty at the beginning) and project it, compare this projection to the true TEM images and modify the volume to have a better correlation. The number of iterations and the relaxation coefficient has to be chosen (17). In our case, 17 iterations and a coefficient of 0.1 have given good results (Fig.Â€6d). One iteration runs over 40Â€min, the improvement of the quality of the reconstruction (calculated with the mean square error) as a function of the iterations is shown on a graph (Fig.Â€6f). After computing the tomogram, the reconstruction can be saved for further visualization as slices using ImageJ program (IMAGEJ: File/save as/name.tif). 6. SIRT (chosen in the rolling window in the right part of the TOMOJ panel), this reconstruction is performed in an iterative manner and is like ART slow. The principle of SIRT is very similar to ART. The main difference is that the modification of the volume is done only after all comparisons between current volume and TEM images are finished. This results in an average of modifications and a smoother reconstruction at the expense of computing time. The number of iterations (around 30) and the relaxation coefficient (initially 1) have to be set (17). In our case, 30 iterations and a coefficient of 1 have given good results Fig.Â€6e. One iteration is made of two steps and runs over 30Â€min, the improvement of the quality of the reconstruction (calculated with the mean square error) as a function of the iterations is shown on a graph (Fig.Â€6g). 3.3.3. A 3D Visualization

A 3D visualization of the tomogram can be obtained using the Chimera program. 1. The reconstruction saved as a Tif extension file has to be transformed inverting black and white (IMAGEJ: Edit/ Invert), then the result is saved in a format, which can be used by Chimera program (IMAGEJ: File/save as/spider writer/name.spi). 2. When chimera is opened, the tomogram is loaded (File/ open/name.spi). Contrasts are adjusted within the volume viewer window, after selection of the solid option (Fig.Â€7a). 3. Membrane “individualization” can be performed using a segmentation selection within the tomogram in IMAGEJ program and be saved as individual files for each membrane (see Note 6). Different colors can be used for the different elements of interest (Fig.Â€7b).

3.4. Conclusions

1. In noninflammatory and inflammatory conditions of cell growth, cristae morphology of mitochondria changes from linear to circular shapes with intermediate states (Fig.Â€ 1). These morphologies are observed in two dimensions and

232

Péranzi et al.

Fig.Â€7. Three-dimensional view of the tomogram. Panel a, nontreated tomogram showing not only the mitochondria but also external component such as the nuclear membrane (Nmb), for instance. Panel b, mitochondria has been isolated and external double membrane colored in blue, whereas internal cristae has been colored in red.

Fig.Â€8. Three-dimensional view of the tomogram of the three types of mitochondria presented in Fig.Â€1. Panels a, b, and c, mitochondria that have been isolated from tomograms and where internal component specific of each three types of mitochondria have been highlighted in pink.

could be interpreted in different ways such as the formation of closed vesicular cristae. The main progress over the last years is the possibility to get three-dimensional views of these mitochondria from electron tomograms (Fig.Â€8). It permits to characterize that linear cristae are formed by lamellar staking that may interact at certain levels (Fig.Â€8a). It also permits to observe that circular cristae are the vertical superposition of two clockwise and anticlockwise curved lamellae and not closed vesicles (Fig.Â€ 8b). In inflammatory conditions, the third type of mitochondria exhibits lamellae that do not cross the matrix but rather fold on the same place, forming closed

Electron Microscope Tomography of Native Membranes

233

domains (Fig.Â€8c). Further work is needed to see what is happening on top and bottom of these domains, for instance, are they closed? The inner membrane remodeling has also been observed in pathology and apoptosis conditions (8, 18). 2. Electron tomography is a new approach that attempts to observe protein complexes in situ. For example, formation and dynamic of cristae depend on several factors such as the distribution of key proteins and/or lipids that would be important to observe and characterize. Observation of these protein complexes close to atomic level will need an improvement of resolution and a good selection of samples. Recent developments from the last years permit the correlation of light microscopy to select areas containing elements to be observed, with electron tomography to characterize molecular objects (8, 19). 3. Perspectives: Since its early development (20), electron tomography has been largely developed; further progress in high resolution toward membrane protein characterization might be gained by conjunction of different factors. First, sample preparation, selection and preservation can be improved using cryo or high pressure freezing followed by cryo-substitution to preserve better the structures (21). Second, the image acquisition can be improved using cryoholder that permits to reduce shrinkage of the resin-embedded samples (22). Image acquisition is also improved by the use of energy filter to remove electron having inelastic interactions with sample and so increase signal-to-noise ratio. Third, volume reconstruction might be improved either by increasing high tilt possibility (up to 80° using specialized holder), or dual-axis image acquisition (15) to minimize missing information.

4. Notes 1. Despite the loss of contrast, cell sections were not overstained, which has often been done by floating grid on lead or uranyl acetate solutions since this produces granular artifacts at the level of membranes. These artifacts are observed by transmission electron microscopy at high magnification, but more clearly evidenced on tomograms after three-dimensional reconstruction. 2. Thickness of the section has to be adapted to the dimension of the studied object and to electron microscope. Indeed, in this latter case, electron generated by high acceleration voltage will be able to cross thicker sections than lower acceleration voltage.

234

Péranzi et al.

However, they will produce higher damages. Usually, maximal thickness of the section is close to 200Â€ nm using electron microscope at 120Â€kV. This allows observations of organelles whereas membrane can be studied with thinner section of 100Â€nm. Another important factor is the irradiation dose (i.e., the number of electron per surface unit) needed to get an image. It increases with the magnification and the acquisition time. The choice of the lowest acceptable magnification to observe the details of interest is thus important to reduce the irradiation of the sample. In this case, it has to be stressed that the use of a camera positioned below the negative chamber in some microscopes might help since it will permit to work at lower nominal magnification to acquire the same details. 3. The same number of digits in the file name, as well as the absence of signs such as plus or minus (which are not considered as digit), is an absolute request to load data an ordered true series. 4. Accuracy of the tilt axis is important for the reconstruction process and to generate high-resolution structure. However, it has no effect on the process itself since it does not affect iteration process performed with ART and SIRT. At low resolution, an error of 10° will generate a shirring effect along the various plans clearly observable on the reconstructed volume. 5. If the chosen value of thickness for the reconstruction is too small compared to the true value, it will not only reduce the information but also an artifact might appear as saturated lines parallel to the tilt axis. This is due to data from the images having no place in the reconstruction (not enough thickness) that the program tries to insert somewhere. If the value of thickness of reconstruction is too large (by much), it might generate “ghost object” on the top and bottom of the reconstruction. Therefore, if the true thickness is impossible to estimate accurately, the best is to take a thickness larger by a small amount, this would only induce a larger computation time. 6. Segmentation might involve the selection of slices containing the information of interest. A mask surrounding the chosen details permit to focus on them and to exclude the others. Filters and threshold can be applied to segmentation selections to reinforce the information. Binarization of data can help to present data or “interpretation” of data such as membrane continuity or discontinuity. Another possibility is drawing of details from data to focalize on specific membrane contacts for instance. These different steps are accessible by the various commands included in IMAGEJ or can be added as plugging.

Electron Microscope Tomography of Native Membranes

235

Acknowledgments The authors would like to thank M. A. Ostuni for his help in the critical reading of this chapter. This work was supported by CNRS (Centre National de la Recherche Scientifique) and ANR (Agence National pour la Recherche) Grant 06-Blan-0190-01 to JJL. References 1. Lacapère J-J, Peybay-Peyroula E, Neumann J-M, Etchebest C (2007) Determining membrane proteins structures: still a challenge! Trends Biochem Sci 32(6):259–270 2. Jasiwal JK, Simon SM (2007) Imaging singles events at the cell membrane. Nat Chem Biol 3:92–98 3. Lucic V, Leis A, Baumeister W (2008) Cryoelectron tomography of cells: connecting structure and function. Histochem Cell Biol 130:185–196 4. Sartori A, Gatz R, Beck F, Rigort A, Baumeister W, Plitzko JM (2007) Correlative microscopy: bridging the gap between fluorescence light microscopy and cryo-electron tomography. J Struct Biol 160:135–145 5. Frey TG, Mannella CA (2000) The internal structure of mitochondria. Trends Biochem Sci 25:319–324 6. Frey TG, Renken CW, Perkins GA (2002) Insight into mitochondrial structure and function from electron tomography. Biochim Biophys Acta 1555:196–203 7. Frey TG, Perkins GA, Ellisman MH (2006) Electron tomography of membrane-bound cellular organelles. Annu Rev Biophys Biomol Struct 35:199–224 8. Frey TG, Sun MG (2008) Correlated light and electron microscopy illuminates the role of mitochondrial inner membrane remodelling during apoptosis. Biochim Biophys Acta 1777:847–852 9. Ostuni M, Ducroc R, Peranzi G, Tonon M-C, Papadopoulos V, Lacapère J-J (2007) Translocator protein (18 kDa) ligand PK11195 induces transient mitochondrial Ca2+ release in HT-29 human colon cancer cells. Biol Cell 99:639–647 10. Valette G, Jarry A, Lemarre P, Branka J-E, Laboisse CL (1997) NO-dependent and NO-independent IL-1 production by a human colonic epithelial cell line under inflammatory stress. Br J Pharmacol 121:187–192 11. ImageJ program. http://rsbweb.nih.gov/ij/ 12. TomoJ program. http://u759.curie.u-psud. fr/softwaresu759.html

13. Chimera program. http://current.cs.ucsb. edu/projects/chimera/index.html 14. Boyles J, Fox JE, Phillips DR, Stenberg PE (1985) Organization of the cytoskeleton in resting, discoid platelets: preservation of actin filaments by a modified fixation that prevents osmium damage. J Cell Biol 101:1463–1472 15. Mastronarde DN (1997) Dual-axis tomography: an approach with alignment methods that preserve resolution. J Struct Biol 120:343–352 16. Messaoudi C, de Loubresse NG, Boudier T, Dupuis-Williams P, Marco S (2006) Multipleaxis tomography: applications to basal bodies from Paramecium tetraurelia. Biol Cell 98:415–425 17. Messaoudii C, Boudier T, Sanchez Sorzano CO, Marco S (2007) TomoJ: tomography software for three-dimensional reconstruction in transmission electron microscopy. BMC Bioinformatics 8:288 18. Acehan D, Xu Y, Stokes DL, Schlame M (2007) Comparison of lymphoblast mitochondria from normal subjects and patients with Barth syndrome using electron microscopic tomography. Lab Invest 87:40–48 19. Sun MG, Williams J, Munoz-Pinedo C, Perkins GA, Brown JM, Ellisman MH, Green DR, Frey TG (2007) Correlated three-dimensional light and electron microscopy reveals transformation of mitochondria during apoptosis. Nat Cell Biol 9:1057–1065 20. Frank J (1992) Electron tomography. Threedimensional imaging with the transmission electron microscope. Plenum, New York 21. Al-Amoudi A, Chang JJ, Leforestier A, McDowall A, Salamin LM, Norlén LP, Richter K, Blanc NS, Studer D, Dubochet J (2004) Cryo-electron microscopy of vitreous sections. EMBO J 23:3583–3588 22. Boudier T, Lechaire J-P, Frébourg G, Messaoudi C, Mory C, Colliex C, Gaill F, Marco S (2005) A public software for energy filtering transmission electron tomography (EFTET-J): application to the study of granular inclusions in bacteria from Riftia pachyptila. J Struct Biol 151:151–159

as

Chapter 13 From Electron Microscopy Maps to Atomic Structures Using Normal Mode-Based Fitting Konrad Hinsen, Edward Beaumont, Bertrand Fournier, and Jean-Jacques Lacapère Abstract Electron microscopy (EM) has made possible to solve the structure of many proteins. However, the resolution of some of the EM maps is too low for interpretation at the atomic level, which is particularly important to describe function. We describe methods that combine low-resolution EM data with atomic structures for different conformations of the same protein in order to produce atomic models compatible with the EM map. We illustrate these methods with EM data from decavanadate-induced tubular crystals of a pseudophosphorylated intermediate of Ca-ATPase and the various atomic structures of other intermediates available in the Protein Data Bank (PDB). Determination of atomic structure permits not only to analyse protein–protein interactions in the crystals, but also to localize residues in the proximity of the crystallizing agent both within Ca-ATPase and between Ca-ATPase molecules. Key words: Ca-ATPase, Ion pump, Decavanadate, Tubular crystals

1. Introduction In the course of the last decade, important advances in the field of cryo-electron microscopy (EM) have lead to the solution of many low-resolution structures of proteins and biomolecular complexes (1). Since the elucidation of the function of these molecules and complexes often requires structural information at the atomic level, various modelling techniques have been developed that create atomic-resolution models compatible with the low-resolution structures by combining the cryo-EM data with atomic-level information from other sources. The first techniques of this kind took atomic-level structures for subdomains of a protein or for Jean-Jacques Lacapère (ed.), Membrane Protein Structure Determination: Methods and Protocols, Methods in Molecular Biology, vol. 654, DOI 10.1007/978-1-60761-762-4_13, © Springer Science+Business Media, LLC 2010

237

238

Hinsen et al.

individual molecules of a complex and placed them into the low-resolution structure of the assembly by applying rigid-body motions (2–8). However, the available atomic structures often correspond to different conformations than the cryo-EM structures. This observation leads to the development of flexible docking and fitting techniques. Most of these methods use the normal modes of an elastic network model to describe the flexibility of the molecules (9–12), but simulation techniques based on geometric constraints have also been used (13). In this chapter, we present a series of methods that make it possible to obtain an atomic structure from an EM map, given an atomic structure for a different conformation of the same protein. We show not only how to obtain an atomic structure, but also how to analyse the conformational change from the initial to the fitted atomic structure, and how to assess the quality of the fitted model. We demonstrate these methods using the sarco/endoplasmic reticulum Ca-ATPase (SERCA) as an example. The sarco/endoplasmic reticulum Ca-ATPase (SERCA) is a transmembrane protein that belongs to a large family of P-type ATPases (14). This 100-kDa protein uses ATP as energy source to transport calcium and forms a stable aspartyl-phosphoryl intermediate that distinguishes it from F- and V-type ATPases, which are ATP synthases. The catalytic cycle consists of several intermediates that alternate steps of ion and nucleotide binding/release (Fig.Â€1). The calcium binding step required for enzyme activation permits the ATP terminal phosphate to be transferred to the enzyme. Then, calcium becomes occluded and vectorial transport occurs followed by hydrolytic cleavage of covalent phospho-intermediate. Proton countertransport from the lumen of reticulum is associated with calcium transport from the cytosol (15, 16) with a stoïchiometry of 1–1.5 leading to electrogenic transport (17). The structures of several intermediates have been defined by X-ray crystallography and EM (18, 19). Calcium binding sites reside within the membrane-inserted region formed by ten transmembrane helices. The ATP and catalytic sites are in the cytoplasmic headpiece formed by three domains (called N, P, and A for nucleotide, phosphorylation, and actuator, respectively). Coupling of ATP hydrolysis and calcium transport is effectuated by conformational changes that involve long-range intramolecular rearrangements (20). Atomic models of various conformational states (Fig.Â€1) have been derived from crystals obtained in the presence or in the absence of calcium. In the former case, calcium was present alone (21), with a nonhydrolyzable ATP analogue (22, 23), or with ADP and fluoroaluminate (22, 24). In the latter case, the enzyme was stabilized by the presence of a Ca-ATPase inhibitor, thapsigargin, without (25–27) or with a nonhydrolyzable ATP analogue (27), or with fluoride (24) or fluoroaluminate (28). Historically, the first structures for Ca-ATPase were obtained by

From Electron Microscopy Maps to Atomic Structures Using Normal Mode-Based Fitting

239

Fig.Â€1. The functional cycle of Ca-ATPase with intermediates that have been biochemically and/or biophysically characterized. The conformational states that have been identified by crystallography are shown in boxes; the PDB IDs are indicated in italics. Steps that are subjected to modulation by non-covalently bound ATP are shown by dotted lines. The upper part corresponds to calcium-bound intermediates, whereas the lower part corresponds to calcium-free/protonated intermediates.

cryo-EM from tubular crystals (29, 30), but their resolution was not sufficient to yield atomic coordinates. Stabilization with thapsigargin increased the resolution, but atomic structures were only obtained by docking and manual modification of atomic models within the EM map (31, 32). The use of fluorescein(FITC)labelled Ca-ATPases made it possible for the first time to crystallize a pseudo-phosphorylated intermediate (32, 33). High density corresponding to decavanadate is clearly observed, but interacting amino acids cannot be determined from the EM data alone. Moreover, the location of FITC has not been clearly determined. It is important because the presence of this FITC molecule blocks calcium-dependent phosphorylation by ATP.

2. Materials 1. The EM maps used in this chapter were obtained from tubular crystals of sarcoplasmic reticulum (SR) Ca-ATPase in the presence of decavanadate (32) and stored in a CCP4 format density map file.

240

Hinsen et al.

2. The atomic structures of SR Ca-ATPase were obtained from the PDB. 3. The program DensityFit, an Open Source program for fitting atomic structures into EM density maps, is available for download at http://dirac.cnrs-orleans.fr/plone/software/ densityfit/. This program uses the Molecular Modelling Toolkit, an Open Source library for molecular simulations, which is available for download at http://dirac.cnrs-orleans. fr/MMTK/ (34). 4. The program DomainFinder, an Open Source program for identifying dynamical domains in proteins, is available for download at http://dirac.cnrs-orleans.fr/DomainFinder/. While we do not use DomainFinder itself, we use some of the utility scripts for studying conformational transitions that come with DomainFinder. 5. The molecular visualization program PyMol is available for download at http://pymol.sourceforce.net. 6. Plotting software (such as Gnuplot or Excel) is recommended to visualize and analyse the results. 7. Any desktop computer with a common operating system (Linux, MacOS X, Windows XP) can be used to perform the calculations.

3. Methods In the following, we describe the methods that we used to obtain and analyse atomic-resolution structures corresponding to the EM map for Ca-ATPase. The central method concerns fitting a Ca-model for a different conformation into the EM map (see Note 1 for a discussion of the use of all-atom models). This method can be transferred to other proteins rather easily. The subsequent analysis methods are more specific to the protein under study and may have to be adapted. 3.1. Flexible Fitting of Atomic Structures into an EM Map of an Isolated Protein

The program density_fit applies the normal mode-based fitting technique described in (10). It consists of iteratively deforming an initial structure under the influence of an external force derived from the cryo-EM map. The force is calculated as the gradient of a potential energy that is given by the square of the difference of the experimental map and a theoretical map calculated from the model. The mechanical response of the protein to this external force is calculated using the large-amplitude normal modes of an elastic network model for the Ca atoms (35). The structural integrity of the protein is preserved by keeping the nearest-neighbour Ca distances constant during the deformations.

From Electron Microscopy Maps to Atomic Structures Using Normal Mode-Based Fitting

241

In the following, we detail the steps that need to be executed. The second line for each step is the command line to be typed into the computer. The starting point is an EM map file (in CCP4 or EZD format) and a starting structure in a PDB-format file, containing at least the Ca atom of each residue. In our example, these files are called “caatpase.ccp4” and “1SU4.pdb”, respectively. 1. Preprocess the EM map: density_fit prepare caatpase.ccp4 0.5 0.9 This step creates a pre-processed EM map and stores it in the file “caatpase.map”. The last two parameters indicate which parts of the map are modified. Map values between 0.5 times the largest map value and 0.9 times the largest map value are left unchanged. Map values above 0.9 times the largest map value are set to 0. Map values below 0.5 times the largest map value are considered susceptible to noise, which is removed using an algorithm described in (10). This noise elimination step is likely to be required with all EM maps. The upper limit of 0.9 times the largest map value is used for Ca-ATPase to remove the contribution of the decavanadate, which because of the high number of electrons in the vanadium atoms creates the highest map values. Since the structure that is fitted does not contain decavanadate, the corresponding map data must be removed as well. 2. Find the right orientation of the atomic structure relative to the EM map: density_fit orientation 1SU4.pdb caatpase.map This command produces several PDB files, named “1SU4_ caatpase_orientation_0.pdb”, “1SU4_caatpase_ orientation_1. pdb”, etc. The number of files can vary from one to eight. Each file contains the initial structure (reduced to its Ca atoms) rotated to a different orientation. In most cases, the first file (“1SU4_ caatpase_ orientation_0.pdb”) contains the most suitable initial orientation of the protein with respect to the map. However, for proteins that have a near-symmetry, the optimal orientation can be contained in one of the other files. In such a situation, it is recommended to repeat the following steps for each of the candidate orientations. 3. Launch the fitting process: density_fit fit 1SU4_caatpase_ orientation_0.pdb caatpase.mapÂ€11.5 This command performs the main fit algorithm and can take a long time to execute. It produces a protocol file called “fit_1SU4_ caatpase_m42_fiterror.txt” that contains the remaining fit error at each step and a PDB file called “fit_1SU4_caatpase_m42.pdb” containing the final fitted structure. The number 42 in these file names is the number of modes that is used in the fit. It is determined from the last parameter on the command line (11.5 in the example). This parameter specifies the factor between the energy

242

Hinsen et al.

of the last mode and the energy of the first non-trivial mode (the first six normal modes have zero energy and describe global translation and rotation; they are never used in the fit). The advantage of specifying an energy factor rather than a number of modes is that the energy factor has a similar meaning for all proteins, whereas the number of modes corresponding to it increases with the size of the protein. Note 2 explains how this choice affects the outcome of the fit. 4. Extend the fitted Ca structure to an all-atom structure (optional): density_fit atomic_coordinates fit_1SU4_caatpase_m42.pdb 1SU4.pdb This command takes the peptide plane and side chain conformations from the second input file and applies a rotation and translation to their atoms that makes them follow the displacements that the Ca atoms have undergone during the fit. After this reconstruction, the all-atom structure is energy minimized to remove bad contacts. The final structure is written to the file “fit_1SU4_ caatpase_m42_all_atom.pdb”. It is important to note that the EM data is not used at all in this step. As a consequence, the side chain conformations should not be interpreted as information from experiment. The PDB contains 12 atomic structures for Ca-ATPase, corresponding to different reaction intermediates of the transport cycle (Fig.Â€ 1). TableÂ€ 1 shows the calculated root mean square deviation (RMSD) between all these atomic structures. A cluster analysis using the ScientificPython library (36) reveals four groups of structures that can be distinguished by the range of RMSD variation. This classification was expected since they correspond to four different reaction intermediates crystallized either with a different compound or by different research groups. We started the fit procedure from these 12 atomic structures and also varied the number of modes used in the procedure. In this way, we obtained a range of final structures. FigureÂ€2 shows the fit error of the final structure as a function of the number of modes for each of the 12 input structures; the value for 0 modes corresponds to the initial structure. The four groups already identified by the RMSD are clearly recognizable. The fit error of the final structure decreases with an increasing number of modes up to about 65 modes. When the number of modes is increased further, the fit error of the final structure behaves erratically. We explain this phenomenon in Note 2. In the following, we consider only the “best fit” for each initial structure, i.e. the final structure that leads to the smallest fit error with a number of modes sufficiently small to avoid overfitting. Among these best fits, the one with the absolute lowest fit error is the one obtained from the input structureÂ€ 1WPG. We discuss the analysis of the fitted structures in SubheadingÂ€3.3.

0

1.40

1.81

1.84

1.82

4.23

4.73

14.37

10.51

10.44

10.44

10.43

1IWO

2AGV

2C8L

2C88

2C8K

1XP5

1WPG

1SU4

1VFP

1T5T

1WPE

1T5S

10.48

10.49

10.49

10.54

14.40

4.69

4.35

1.86

1.90

1.85

0

0.78

2AGV

10.50

10.49

10.51

10.57

14.31

4.47

4.38

0.71

0.68

0

1.22

1.17

2C8L

10.52

10.51

10.53

10.57

14.31

4.47

4.36

0.60

0

0.34

1.22

1.17

2C88

10.50

10.49

10.51

10.55

14.32

4.48

4.38

0

0.29

0.35

1.21

1.17

2C8K

11.86

11.85

11.87

11.94

14.13

2.25

0

4.00

3.97

3.99

4.07

3.99

1XP5

11.88

11.82

11.89

11.93

13.78

0

1.62

4.15

4.12

4.13

4.37

4.35

1WPG

13.95

13.96

13.97

13.86

0

13.49

13.81

13.97

13.97

13.97

14.06

14.01

1SU4

2.00

1.11

1.99

0

13.86

11.60

11.59

10.24

10.26

10.24

10.23

10.18

1VFP

0.33

2.03

0

1.31

13.68

11.55

11.53

10.21

10.22

10.20

10.19

10.12

1T5T

2.03

0

1.21

0.63

13.68

11.48

11.47

10.17

10.18

10.16

10.16

10.10

1WPE

0

1.22

0.18

1.31

13.65

11.54

11.52

10.20

10.22

10.19

10.18

10.12

1T5S

Upper part Ca only, lower part all-atom. A cluster analysis reveals four groups corresponding to (a) Ca-free protein with or without nucleotide (1WO, 2AGV, 2C8L, 2C88, 2C8K), (b) phosphate bound to Ca-free protein (1XP5, 1WPG), (c) Ca-bound protein (1SU4), and (d) Ca-bound protein with nucleotide (1VFP, 1T5T, 1WPE, 1T5S)

1IWO

Model

TableÂ€1 RMSD between the atomic structures for Ca-ATPase in the PDB

From Electron Microscopy Maps to Atomic Structures Using Normal Mode-Based Fitting 243

244

Hinsen et al. 2 1SU4

Fit error (x 10−5)

1.5 1T5S, 1T5T, 1VFP, 1WPE

1IWO, 2AGV, 2C88, 2C8K, 2C8L

1

1WPG, 1XP5

0.5 0

20

40

60

80

100

Number of modes

Fig.Â€2. The fit error of the final structures (after convergence of the fitting procedure) for the 12 initial atomic structures shown as a function of the number of modes used in the fit. See Note 2 for an explanation of the behaviour beyond approximately 60 modes.

3.2. Analysis of the Fitting Procedure as a Conformational Transition

The fitting procedure can be considered as a simulation of a conformational transition from the initial to the final conformation. An analysis of this conformational transition is instructive both for understanding the fitting procedure and for illustrating the conformational differences between the two states of the protein. For a large protein, such as Ca-ATPase, it is convenient to analyze the conformational transition in terms of dynamical domains, that is, large regions of the proteins that can, to a good approximation, be considered rigid bodies for the purpose of describing the collective motions of the protein. The program DomainFinder (37, 38) is a useful tool for identifying such dynamical domains, and it has been applied to Ca-ATPase before (39). In the following, we use the definitions of the domains A (activator), N (nucleotide), and P (phosphorylation) from this article. We also look at the rigid-body displacements of the ten transmembrane helices although they cannot be considered dynamical domains because they are too small. A schematical view of the dynamical domains and the transmembrane helices in Ca-ATPase is shown in Fig.Â€3a. As a first analysis, we look at how the rigid-body motions of the three domains and of the transmembrane helix group are decomposed into normal modes. This analysis helps to understand the behavior of the fitting procedure described in the preceding subheading. The calculations are performed by the scripts

From Electron Microscopy Maps to Atomic Structures Using Normal Mode-Based Fitting

a

b

245

1

Cumulative projection

N

0.8 0.6

0.8

N A

0.6

1VFP 1T5S 1T5T 1WPE

P

0.4 0.2 0 0

Cumulative projection

Cumulative projection

d

1

TH 20

40

60

Number of modes

80

100

P

1SU4

0.4 0.2 0

c

A

TH 0

20

40 60 Number of modes

80

100

1 0.8

N A

0.6

P

0.4 0.2 0 0

1IWO 1WPG 1XP5 2AGV 2C88 2C8K 2C8L

TH 20

40

60

80

100

Number of modes

Fig.Â€3. The contributions of the first 100 normal modes to the rigid-body displacements of the three dynamical domains (A, N, P) and of the transmembrane helices (TH) during the conformational transition from the initial to the final fitted structures. A schematic drawing of the domains is shown in (a). The 12 structures are shown in three plots for visual clarity: (b) 1SU4; (c) 1T5S, 1T5T, 1VFP, and 1WPE; (d) 1IWO, 1WPG, 1XP5, 2AGV, 2C88, 2C8K, and 2C8L.

DomainFinderModes and RigidBodyMotionsByMode that are distributed with the program DomainFinder (38): DomainFinderModes 1SU4.pdb 1SU4.modes 2982 RigidBodyMotionsByMode 1SU4.modes 1-40+124-243 360-604 330-359+605-737 49-78,89-121,248-274,289-329, 740-779,789-808,831-852,896-914,931-949,965-985 The first line calculates the normal modes of Ca-ATPase (input structureÂ€ 1SU4.pdb) and stores them in the file 1SU4. modes. The last argument is the number of normal modes; we choose to calculate all 2,982 modes (3â•›×â•›994, since Ca-ATPase has 994 residues). The second line calculates the decomposition into normal modes for the three domains (arguments 2–4) and the group of ten transmembrane helices (last argument). Each domain is described by a combination of several chain segments which are specified by their residue number ranges. FigureÂ€ 3 shows the contributions of the first 100 normal modes to the rigid-body displacements of the three dynamical

246

Hinsen et al.

domains and of the transmembrane helices. The value for n modes can be interpreted as the percentage of the rigid-body conformational change that is described by the first n normal modes of the elastic network model for the protein. It may seem surprising at first that the number of modes required for describing the conformational change exceeds the number of modes used in the fitting procedure. There are three reasons for this: (1) The fitting procedure consists of a sequence of small displacements during which the normal modes are recalculated regularly, whereas the cumulative projection is calculated from a single mode set for the total difference vector. (2) The fitting procedure includes a correction of nearest-neighbour distances which in normal mode space corresponds to small-amplitude localized modes. (3) The dynamical domains are only approximately rigid; there are also internal conformational changes inside each domain. It becomes clear from Fig.Â€3 that the 60–70 modes used in the fitting procedure describe most of the rearrangements for domains N and A (80% or more) and an important part for domain P (60% or more), but only about 10% for the transmembrane helices. This is not surprising, considering that the cytoplasmic domains undergo large-amplitude movements, whereas the displacements of the transmembrane helices are smaller and more localized. We can, thus, expect that the rearrangements of the domains are much better represented in our fitted structures than the rearrangements in the helix region. The topic of our second analysis is the amplitudes of the domain displacements between the initial and the final conformation during the fitting process. We wish to calculate for each domain the rigid-body translation (the displacement of the centre of mass), the angle of rotation, and the amount of internal deformation, measured by the RMSD between the two domain conformations. The displacement amplitudes obviously depend on the relative location and orientation of the two protein structures as a whole. Since Ca-ATPase is a membrane protein, it is reasonable to consider the transmembrane part fixed in space. We thus align the two structures by an optimal superposition of the transmembrane helices before calculating the displacement amplitudes. The calculations are performed by the script DomainMotion Amplitudes that comes with DomainFinder. The command line for our example is DomainMotionAmplitudes 1SU4.pdb fit_1SU4_caatpase_ m42.pdb 49-78,89-121,248-274,289-329,740-779,789-808, 831-852,896-914,931-949,965-985 1-40+124-243 360-604 330-359+605-737 The first two arguments are the PDB file names for the initial and final structure. The third argument is the specification of the alignment region, using the same notation as for the script

From Electron Microscopy Maps to Atomic Structures Using Normal Mode-Based Fitting

247

RigidBodyMotionByMode. The remaining three arguments are the specifications of the three domains. The results are shown in TableÂ€2. The low internal deformation (RMSD variation less than 5.3â•›Å) confirms that our domains behave indeed very much like rigid bodies. The amplitudes (translation and rotation) show once again that our 12 structures can be divided into four groups, in agreement with the classification previously observed in TableÂ€1 and corresponding to different reaction intermediates. 3.3. Analysis of the Fitted Structures

Next, we look at the final structures resulting from the 12 different input structures. TableÂ€3 shows the RMS distances between them. Two low-RMSD groups can be distinguished, one resulting from Ca-ATPase structures without calcium and the other from structures with calcium. The remaining structure (1SU4) has calcium bound but its cytoplasmic domains are largely open compared to the other structures. This might be the reason for the higher RMSD values relative to the other calcium containing structures. FigureÂ€ 4a shows the superposition of the best fits obtained from the various initial structures. With the exception of the fit obtained from 1SU4, all structures are in good agreement. FigureÂ€ 4b shows the final structure obtained from 1WPG (the best fit overall) together with spheres placed at the centres of mass of the three cytoplasmic domains (A, N, and P) for all 12 fitted structures. This shows that the placement of the domains depends only very weakly on the input structure; the distances between the spheres for each domain are smaller than 1â•›Å. FigureÂ€4c shows the centers of mass of the transmembrane helices in the same way, illustrating that the structural agreement is much worse in the transmembrane part of the protein. For a quantitative comparison of the structures, we again used the script DomainMotionAmplitudes as explained in the last subheading, but using pairs of fitted structures as inputs. TableÂ€4 shows the RMSD values between all pairs of final structures for the A and N domains. It shows that the internal structure of the N domain is almost identical for all structures (the RMSD varies from 1 to 3â•›Å), whereas the A domain exhibits a higher variation (the RMSD varies from 1 to 6â•›Å). This suggests that there is almost no deformation of the domains during the fitting process. However, analysis of the relative orientation of the domains between the different models revealed important rotations for the A domain (up to 75°). The rotations of the P domain are smaller (up to 20°) and the N domain does not show any significant rotation (at most 9°). Again there are two distinct groups of structures (with and without calcium bound) within which the rotation amplitudes are very similar. FigureÂ€5 shows best fitted structure (obtained from 1WPG) within the EM map. FigureÂ€5a demonstrates that the model fits

5

5

5

5

4

4

10

9

8

10

8

2AGV

2C8L

2C88

2C8K

1XP5

1WPG

1SU4

1VFP

1T5T

1WPE

1T5S

25

20

27

31

34

3

4

12

11

12

12

11

5.3

5.3

5.2

5.2

3.6

1.4

1.3

2.0

1.7

2.0

1.7

1.7

16

14

17

16

11

2

2

17

16

17

17

17

47

45

46

48

55

3

3

18

18

18

18

18

Rotation (°)

3.2

3.6

3.4

3.8

3.4

2.3

2.0

2.8

2.9

2.8

3.0

3.0

RMSD (Å)

2

2

2

2

5

3

3

3

3

3

4

4

Translation (Å)

Domain P

9

12

10

12

25

5

4

11

11

11

11

12

Rotation (°)

4.4

4.2

4.3

4.0

3.6

1.5

1.5

1.8

1.8

1.9

1.9

1.7

RMSD (Å)

Domain definitions are: actuator (A, residues 1–40 and 124–243); nucleotide (N, residues 360–604) and phosphorylation (P, residues 330–359 and 605–737). Translations have been calculated between the centres of mass. Rotations and RMSD have been calculated from an optimal superposition of the domains in the initial and final structures. The structures were aligned by an optimal superposition of the transmembrane helices before calculating the domain movements

5

1IWO

Translation (Å)

RMSD (Å)

Translation (Å)

Rotation (°)

Domain N

Domain A

TableÂ€2 Analysis of the domain movements in the course of the fitting procedure

248 Hinsen et al.

0.00

1IWO

0.00

1.43

5.20

67

2AGV

0.00

1.38

1.74

5.08

66

2C8L

0.00

0.86

1.44

1.59

5.18

64

2C88

0.00

0.76

0.65

1.38

1.67

5.12

66

2C8K

0.00

2.25

2.30

2.44

2.33

2.36

5.04

66

1XP5

0.00

1.70

2.15

2.14

2.21

2.33

2.63

4.85

65

1WPG

0.00

7.29

7.20

7.39

7.42

7.42

7.46

7.43

6.46

67

1SU4

0.00

5.14

7.10

7.11

7.06

7.15

7.10

7.16

7.19

5.79

65

1VFP

0.00

2.03

5.15

7.23

7.25

7.23

7.30

7.27

7.31

7.36

5.99

64

1T5T

0.00

2.15

1.93

5.23

7.47

7.47

7.41

7.48

7.45

7.51

7.51

5.89

63

1WPE

0.00

1.87

1.23

2.17

5.18

7.28

7.27

7.29

7.36

7.34

7.38

7.40

5.96

62

1T5S

“Modes” corresponds to the number of modes used in the fitting procedure. “Delta” is the fit error for the final structure. Two low-RMSD groups are observed and correspond to Ca-free and Ca-bound protein

1T5S

1WPE

1T5T

1VFP

1SU4

1WPG

1XP5

2C8K

2C88

2C8L

2AGV

5.16

Delta (10 )

68

Modes

−5

1IWO

Model

TableÂ€3 RMSD between final fitted structures obtained from different PDB entries (Ca only)

From Electron Microscopy Maps to Atomic Structures Using Normal Mode-Based Fitting 249

250

Hinsen et al.

Fig.Â€4. Visual inspection of the fitted structures. (a) Superposition of the fitted structures obtained from the different initial conformations. (b) Superposition of the centres of mass of the cytoplasmic domains (black spheres) corresponding to the various fitted structures. (c) Superposition of the centres of mass of the transmembrane helices (black spheres) from the various fitted structures. The Ca trace in panels (b) and (c) corresponds to the best overall fit, obtained from the atomic structureÂ€1WPG.

well into the EM map. This is further confirmed looking at crosssections using the PyMol visualization program. A cross-section at the level of D351, the covalently phosphorylated residue (Fig.Â€5b), and another one close to the middle of the transmembrane helices (Fig.Â€5c) indicate the quality of the fit. 3.4. Localization of Intramolecular Decavanadate

A very high density is clearly observed within the EM map. This high density implies the presence of heavy metal which suggests that it corresponds to a decavanadate molecule added during the crystallization process. This is reinforced by the fact that its size corresponds to the known decavanadate structure. Analysis of its location in the atomic structure of the best fitted model shows the proximity of residues from the different structural domains. The identification of the amino acid present in a sphere of radius 1.5Â€nm around the centre of mass of the decavanadate (see Note 3) permits to characterize several positively charged residues (arginines and lysines) as good candidates to interact with the negatively charged decavanadate. Comparison of the atomic structure obtained from the various initial structures reveals that decavanadate is surrounded by a constant group of amino acid from the different domains (Fig.Â€ 5d). Some of these residues have been

0.00

1.08

1.98

1.89

1.95

2.46

2.95

3.60

3.55

3.16

3.48

3.13

1IWO

2AGV

2C8L

2C88

2C8K

1XP5

1WPG

1SU4

1VFP

1T5T

1WPE

1T5S

3.05

3.40

2.98

3.43

3.61

2.76

2.31

1.62

1.72

1.67

0.00

0.94

2AGV

2.96

3.08

2.83

2.92

3.58

2.24

2.43

0.58

0.67

0.00

0.70

0.92

2C8L

3.04

3.19

2.89

3.02

3.79

2.43

2.56

0.74

0.00

0.62

0.75

0.74

2C88

2.85

3.03

2.73

2.87

3.62

2.30

2.33

0.00

0.57

0.49

0.74

0.91

2C8K

2.65

3.31

2.79

3.35

3.45

1.50

0.00

1.98

1.80

2.11

1.90

1.77

1XP5

2.73

3.14

2.70

3.16

3.56

0.00

1.00

2.10

1.90

2.13

1.91

1.87

1WPG

3.49

3.67

3.77

3.84

0.00

4.08

4.20

4.17

4.10

4.10

3.98

4.10

1SU4

1.77

1.64

1.83

0.00

5.72

5.06

5.19

4.88

4.99

4.82

4.88

4.83

1VFP

Upper diagonal, actuator domain (A, residues 1–40 and 124–243) and lower diagonal, nucleotide domain (N, residues 360–604)

1IWO

Model

TableÂ€4 RMSD between domains of best fitted structures starting from different PDB structures

0.00 1.44

1.23

2.16

2.27

5.60

5.34

5.48

4.97

5.06

4.86

4.89

4.96

1WPE

1.79

0.00

1.69

5.89

5.03

5.13

4.89

4.98

4.86

4.89

4.83

1T5T

0.00

2.03

0.99

1.76

6.01

5.07

5.19

4.86

4.95

4.81

4.87

4.84

1T5S

From Electron Microscopy Maps to Atomic Structures Using Normal Mode-Based Fitting 251

252

Hinsen et al.

Fig.Â€5. The EM map of Ca-ATPase with the best fit obtained from the atomic structureÂ€1WPG. (a) Side view with the EM map in grey mesh and the atomic structure in rainbow-coloured ribbon. (b, c) Cross-sections through the cytoplasmic and transmembrane regions, respectively. The positions of the cross-sections are indicated in (a) by the black lines. (d) Section around the highest density, attributed to the decavanadate (dark blue). The closest positively charged residues from the different cytoplasmic domains are indicated.

described as functionally crucial (27, 40, 41) for either nucleotide binding and catalysis or inter-domain interaction occurring during conformational changes. 3.5. Localization of Intermolecular Decavanadate After Tubular Reconstruction

The formation of the tubular crystals of Ca-ATPase has been induced by the addition of decavanadate. The tubes consist of ribbons of Ca-ATPase dimers (32) and decavanadate might participate in the formation of the dimers and/or the ribbons. In order to identify the role that decavanadate plays in crystal formation, we reconstructed an atomic model for the tube. The procedure we used is difficult to generalize, therefore we only summarize it. The starting point was the previously described fitted monomer, from which a dimer was constructed by applying a twofold symmetry rotation. The position and orientation of the symmetry axis were adjusted to produce a good fit of the dimer into the EM map. In the next step, a ribbon was formed from dimers by applying iteratively a transformation consisting of a translation along the tube axis and a rotation around it; the parameters of this transformation, which include the location of the tube axis, were again adjusted. Finally, six ribbons were assembled to make the

From Electron Microscopy Maps to Atomic Structures Using Normal Mode-Based Fitting

253

complete tubular crystal. The transformation mapping a ribbon to its neighbour has no freely adjustable parameters any more, because it is completely defined by the number of ribbons and the number of dimers in one turn. These are integer numbers for which only one value set is compatible with the experimental data. FigureÂ€6a shows the global arrangement of the Ca-ATPase molecules within a tube section. It makes it possible to analyse interprotein contacts within the tubular crystal at different levels within and above the membrane region. On the outside of the tubular crystal, interactions occur between Ca-ATPases forming the dimers within a ribbon and involve the domains A and N (see circles in Fig.Â€ 6a). In the membrane region, interactions occur between Ca-ATPases from different ribbons and involve transmembrane helix M3 and the cytosolic loop linking M7–M8 (Fig.Â€6b). A high electron density between two Ca-ATPases from two different ribbons is observed within the crystal suggesting the presence of heavy metal. This density does not correspond to any part of the fitted protein and can be attributed to the decavanadate molecule (Fig.Â€6c, d). However, this density is much more elongated than a decavanadate molecule, suggesting an uncertainty in its location and/or the presence of multiple decavanadate molecules (42). An uncertainty might be due to the flexibility of the interacting regions of the protein or to conformational variations within the crystal. Four Ca-ATPase monomers surround the decavanadate density (Fig.Â€6c). Considering the inevitable uncertainties in the positioning of the proteins relative to the map, two hypotheses for the interaction between the protein molecules and decavanadate can be suggested: (1) decavanadate interacts with four Ca-ATPases near the residues indicated by the two upper ellipses in Fig.Â€6c and (2) only two Ca-ATPases interact with decavanadate, as indicated by the single drawn-out ellipse at the bottom of Fig.Â€6c.

4. Conclusions The method presented here permits to generate atomic models for low-resolution structures obtained from electron microscopy. It can be applied using the atomic structure of very different conformational states of the same protein. The present chapter gives instructions for performing standard computations and general guidelines for the visual interpretation of the results. The resulting atomic structures make it possible to analyse protein–protein interactions within the crystal. Our study reveals that Ca-ATPase dimers are stabilized by cytoplasmic interactions, whereas neighbouring ribbons are held together by interactions between the transmembrane regions. The Ca-ATPase dimers are

254

Hinsen et al.

Fig.Â€6. The tubular crystal of Ca-ATPase. (a) Top view of the EM map (grey mesh) filled with the best fitted structure. Three ribbons of Ca-ATPase dimers are shown as Ca traces: the top and bottom ribbons in green, the central ribbon in orange with the central dimer of Ca-ATPase highlighted as blue and black monomers. Domains A and N of each monomer within a dimer are indicated by ellipses. Intramolecular decavanadates are coloured in red, whereas intermolecular decavanadates are shown in green. (b) Side view of Ca-ATPases from neighbouring ribbons as indicated by the yellow line in panel (a). (c, d) Top and side views of two Ca-ATPases dimers from a single ribbon as defined by the yellow box in panel (a). The intermolecular decavanadate (green surface) is located at the interface of four monomers (shown in different colours). The residues close to the decavanadate are indicated.

linked along the ribbons by decavanadate, whose location is not very well defined. An analysis of the surrounding residues might explain or suggest which ATPases from the same family can form 2D crystals in the presence of decavanadate. Conversely, the decavanadate molecule located within each Ca-ATPase is well characterized and involves the three cytoplasmic domains A, N, and P.

From Electron Microscopy Maps to Atomic Structures Using Normal Mode-Based Fitting

255

From a functional point of view, decavanadate is an ATP competitor and its location in comparison to the location of ATP analogues in other functional states might reveal a pathway to the ATP binding site and/or the activation site. An analysis of the surrounding amino acids might suggest mutagenesis experiments. The atomic structures obtained from our fitting procedure can also serve as the starting point for molecular dynamics simulations. Such simulations permit explicit modelling of the side chains and more detailed insight into the dynamic processes that enable proteins to function.

5. Notes 1. Ca vs. all-atom models in the fitting procedure: The fitting procedure described in SubheadingÂ€ 3.1 uses an elastic network model for the Ca atoms of the protein. The peptide planes and side chains can be added again to the final fitted structure (see step 5), but they never enter in any comparison between the model and the experimental data. The resulting all-atom structure is thus compatible with the experimental data, but it is not legitimate to attribute the side chain conformations to the EM map. This makes the resulting structure less suitable for certain uses such as the identification of protein–ligand interactions. Since normal mode calculations can be and have been performed for all-atom models, it might seem better to use such a model in the fitting procedure. There are, however, two reasons why this is not a good idea in most cases. First, the requirements in CPU time and working memory make all-atom normal mode calculations prohibitive for proteins above approximately 500 residues. Second, the resolution of the EM data is usually not sufficient to provide information about the placement of side chains, a map resolution of about 4â•›Å being the minimum for the location of side chains. 2. Choice of the number of modes in the fitting procedure: The energy factor parameter in the fitting procedure, which determines the number of normal modes used, is a critical parameter for obtaining good results. Normal modes are a set of coordinates that describe conformations and conformational changes in proteins. Their specificity is that the first (“low-energy”, often called “low-frequency” because of the origins of normal mode analysis in small-molecule chemistry) modes describe collective motions, whereas the following modes describe ever more localized motions. The more modes are used in the fitting procedure, the more flexible the model

256

Hinsen et al.

of the protein becomes, making more detailed rearrangements possible. The natural limit to increasing the number of modes is set by the resolution of the EM maps; when the newly added modes describe conformational changes that are too localized to be visible in the EM data, they are no longer useful in interpreting the results. As the behavior of the fitting error with increasing mode number (Fig.Â€1) shows, adding modes up to about 65 modes leads to a systematic decrease of the fit error. Beyond this limit, the fit error behaves erratically. This is a symptom of overfitting; the newly added modes are not fitted to meaningful data but to experimental noise and numerical inaccuracies. 3. Localization of the intramolecular decavanadate in the EM map and identification of the residues in its vicinity: This calculation is specific to our example and no ready-made program is available to do it. We wrote a short Python script using the Molecular Modelling Toolkit (MMTK) explicitly for our analysis. The decavanadate position was identified by searching for the maximum of the EM map and then the distances of the residues from this point were calculated. This procedure is difficult to generalize to other applications and illustrates the utility of easy-to-learn scripting languages such as Python and corresponding simulation libraries such as MMTK.

Acknowledgment The authors would like to thank Professor D.L. Stokes for generously providing the EM map. This work was supported by CNRS (Centre National de la Recherche Scientifique) and CEA (Commissariat à l’Energie Atomique). References 1. Electron Microscopy Data Bank. http://www. emdatabank.org/ 2. Volkmann N, Hanein D (1999) Quantitative fitting of atomic models into observed densities derived by electron microscopy. J Struct Biol 125:176–184 3. Roseman AM (2000) Docking structures of domains into maps from cryo-electron microscopy using local correlation. Acta Crystallogr D Biol Crystallogr 56:1332–1340 4. Rossmann MG (2000) Fitting atomic models into electron-microscopy maps. Acta Crystallogr D Biol Crystallogr 56: 1341–1349

5. Wriggers W, Chacon P (2001) Modeling tricks and fitting techniques for multiresolution structures. StructureÂ€9:779–788 6. Chacon P, Wriggers W (2002) Multiresolution contour-based fitting of macromolecular structures. J Mol Biol 317:375–384 7. Navaza JJ, Lepault F, Rey A, Alvarez-Rua C, Borge J (2002) On the fitting of model electron densities into EM reconstructions: a reciprocal-space formulation. Acta Crystallogr D Biol Crystallogr 58:1820–1825 8. Wu X, Milne JLS, Borgnia MJ, Rostapshov AV, Subramaniam S, Brooks BR (2002) A core-weighted fitting method for docking

From Electron Microscopy Maps to Atomic Structures Using Normal Mode-Based Fitting

9.

10.

11.

12.

13.

14.

15.

16.

17.

18. 19.

20.

atomic structures into low-resolution maps: application to cryo-electron microscopy. J Struct Biol 141:63–76 Tama F, Miyashita O, Brooks CL (2004) Flexible multi-scale fitting of atomic structures into low-resolution electron density maps with elastic network normal mode analysis. J Mol Biol 337:985–999 Hinsen K, Reuter N, Navaza J, Stokes DL, Lacapère J-J (2005) Normal mode-based fitting of atomic structure into electron density maps: application to sarcoplasmic reticulum Ca-ATPase. Biophys J 88:818–827 Schröder GF, Brunger AT, Levitt M (2007) Combining efficient conformational sampling with a deformable elastic network model facilitates structure refinement at low resolution. StructureÂ€15:1630–1641 Topf M, Lasker K, Webb B, Wolfson H, Chiu W, Sali A (2008) Protein structure fitting and refinement guided by cryo-EM density. StructureÂ€16:295–307 Jolley CC, Wells SA, Fromme P, Thorpe MF (2008) Fitting low-resolution cryo-EM maps of proteins using constrained geometric simulations. Biophys J 94:1613–1621 Møller JV, Jull B, Le Maire M (1996) Structural organization, ion transport, and energy transduction of P-type ATPases. Biochim Biophys Acta 1286:1–51 Forge V, Mintz E, Guillain F (1993) Ca binding to sarcoplasmic reticulum ATPase revisited. I. Mechanism of affinity and cooperativity modulation by H and Mg. J Biol Chem 268:10953–10960 Yu X, Carroll S, Rigaud J, Inesi G (1993) H countertransport and electrogenicity of the sarcoplasmic reticulum Ca pump in reconstituted proteoliposomes. Biophys J 64: 1232–1242 Buoninsegni TF, Bartolommei G, Moncelli MR, Inesi G, Guidelli R (2004) Timeresolved charge translocation by sarcoplasmic reticulum Ca-ATPase measured on a solid supported membrane. Biophys J 86: 3671–3686 Stokes DL, Green NM (2003) Structure and function of the calcium pump. Annu Rev Biophys Biomol Struct 32:445–468 Toyoshima C, Inesi G (2004) Structural basis of ion pumping by Ca2+-ATPase of the sarcoplasmic reticulum. Annu Rev Biochem 73:269–92 Inesi G, Lewis DMH, Prasad A, Toyoshima C (2006) Concerted conformational effects of Ca and ATP are required for activation of sequential reactions in the Ca-ATPase

257

(SERCA) catalytic cycle. Biochemistry 45:13769–13778 21. Toyoshima C, Nakasako M, Nomura H, Ogawa H (2000) Crystal structure of the calcium pump of sarcoplasmic reticulum at 2.6â•›Å resolution. Nature 405:647–655 22. Sorensen T, Moller JV, Nissen P (2004) Phosphoryl transfer and calcium ion occlusion in the calcium pump. Science 304: 1672–1675 23. Toyoshima C, Mitzutani T (2004) Crystal structure of the calcium pump with a bond ATP analogue. Nature 430:529–535 24. Toyoshima C, Nomura H, Tsuda T (2004) Lumenal gating mechanism revealed in calcium pump crystal structures with phosphate analogues. Nature 432:361–368 25. Toyoshima C, Nomura H (2002) Structural changes in the calcium pump accompanying the dissociation of calcium. Nature 418:605–611 26. Obara K, Miyashita N, Xu C, Toyoshima I, Sugita Y, Inesi G, Toyoshima C (2005) Structural role of countertransport revealed in Ca-ATPase pump crystal structure in the absence of Ca. Proc Natl Acad Sci USA 102:14489–14496 27. Jensen A-M, Sorensen T-L, Olesen C, Moller JV, Nissen P (2006) Modulatory and catalytic modes of ATP binding by the calcium pump. EMBO J 25:2305–2314 28. Olesen C, Sorensen T, Nielsen RK, Moller JV, Nissen P (2004) Dephosphorylation of the calcium pump coupled to counterion occlusion. Science 306:2251–2255 29. Toyoshima C, Sasabe H, Stokes DL (1993) Three-dimensional cryo-electron microscopy of the calcium ion pump in the sarcoplasmic reticulum membrane. Nature 362: 467–471 30. Zhang P, Toyoshima C, Yonekura K, Green NM, Stokes DL (1998) Structure of the calcium pump from sarcoplasmic reticulum at 8â•›Å resolution. Nature 392:835–839 31. Xu C, Rice WJ, He W, Stokes DL (2002) A structural model for the catalytic cycle of Ca-ATPase. J Mol Biol 316:201–211 32. Stokes DL, Delavoie F, Rice WJ, Champeil P, McIntosh D, Lacapère J-J (2005) Structural studies of a stabilized phosphoenzyme intermediate of Ca-ATPase. J Biol Chem 280:18063–18072 33. McIntosh D, Montigny C, Champeil P (2008) Unexpected phosphoryl transfer from Asp351 to fluorescein attached to Lys515 in sarcoplasmic reticulum Ca-ATPase. Biochemistry 47:6386–6393

258

Hinsen et al.

34. Hinsen K (2000) The molecular modeling toolkit: a new approach to molecular simulations. J Comp Chem 21:79–85 35. Hinsen K, Petrescu AJ, Dellerue S, BellissentFunel M-C, Kneller GR (2000) Harmonicity in slow protein dynamics. Chem Phys 261:25–37 36. Hinsen K (1997) ScientificPython. http:// dirac.cnrs-orleans.fr/ScientificPython/ 37. Hinsen K, Thomas A, Field MJ (1999) Analysis of domain motions in large proteins. Proteins 34:369–382 3 8. Hinsen K (1998) DomainFinder. http:// dirac.cnrs-orleans.fr/DomainFinder/ 39. Reuter N, Hinsen K, Lacapere J-J (2003) Transconformations of the SERCA1 Ca-ATPase: a normal mode study. Biophys J 85:2186–2197

40. Ma H, Lewis D, Xu C, Inesi G, Toyoshima C (2005) Functional and structural roles of critical amino acids within the “N”, “P”, and “A” domains of the Ca-ATPase (SERCA) headpiece. Biochemistry 44: 8090–8100 41. Clausen JD, McIntosh DB, Anthonisen AN, Woolley DG, Vilsen B, Andersen JP (2007) ATP-binding modes and functionally important interdomain bonds of sarcoplasmic reticulum Ca2+-ATPase revealed by mutation of glycine 438, glutamate 439, and arginine 678. J Biol Chem 282:20686–97 42. Ferreira da Silva JL, Minas da Piedade MF, Duarte MT (2003) Decavanadates: a building block for supramolecular assemblies. Inorganica Chim Acta 356:222–242

as

Part IV Nuclear Magnetic Resonance

Chapter 14 Determination of Membrane Protein Structures Using Solution and Solid-State NMR Pierre Montaville and Nadège Jamin Abstract NMR is an essential tool to characterize the structure, dynamics, and interactions of biomolecules at an atomic level. Its application to membrane protein (MP) structure determination is challenging and currently an active and rapidly developing field. Main difficulties are the low sensitivity of the technique, the size limitation, and the intrinsic motional properties of the system under investigation. Solution and solid-state NMR (ssNMR) have common and own specific requirements. Solution NMR requires a careful choice of the detergent, elaborated stable isotope labelling schemes to overcome signal overlaps and to collect distance restraints. Excessive spectra crowding hampered large MP structure determination by ssNMR, and so far only high resolution structure of small or fragments of MP have been determined. However, ssNMR provides the unique opportunity to obtain atomic level information of MP in phospholipid bilayers such as orientation of the protein in the membrane. Specific and careful sample preparations are required in combination with uniformly and partially labelled protein for ssNMR spectra assignment. Distance restraints measurements benefit from methodologies currently developed for small soluble proteins in micro-crystalline state. Recent advances in the field increased the releasing rate of high resolution MP structures, providing unprecedented structural and dynamics information making NMR a powerful tool for structural and functional membrane protein studies. Key words: Bicelle, Lipid bilayer, Membrane protein, Micelle, Solid-state NMR, Solution NMR, Structure

1. Introduction Both solution and solid-state NMR methods have been used for a long time to gain insights into the structure, dynamics, and interactions of membrane peptides and proteins (MPs) as well as of its main partners within the membrane, i.e., lipids. But, it is only in the past two decades that NMR reveals its Jean-Jacques Lacapère (ed.), Membrane Protein Structure Determination: Methods and Protocols, Methods in Molecular Biology, vol. 654, DOI 10.1007/978-1-60761-762-4_14, © Springer Science+Business Media, LLC 2010

261

262

Montaville and Jamin

abilities to determine MPs structures. This chapter focuses on the methodologies currently used for membrane structure determination as well as on the most recent advances in that field. The main advantage of NMR is the variety of environments that can be used: from native membrane environment to various membrane mimetic environments including planar bilayers, bicelles (1), micelles, and organic solvents, at different salt concentration, temperature, and pH values. NMR studies of MPs in their native membrane is only feasible if large amount of proteins (mg) can be purified which has been found for few MPs like bacteriorhodopsin. Among membrane mimetic environments, planar synthetic bilayer and bicelles more closely mimic the membrane environment. Detergent micelles provide good approximations of the interfacial and hydrophobic regions of membrane but do not account for the bidimensional geometry, local curvature, heterogeneity, and dynamics of biological membranes. Organic solvent mixtures are isotropic media that can mimic the hydrophobic region but cannot account for the lipid–water interface. They have been essentially used for the studies of transmembrane (TM) peptides. The main criterion that distinguishes solution versus solidstate NMR is the motional properties of the protein–lipid sample. As a matter of fact, large proteins or macromolecular assemblies (corresponding to MW greater than ~40Â€kDa) are challenging to study using solution NMR methods due to their slow overall rotational correlation time. Typically, assignment and NMR structure of monomeric membrane protein close to 30Â€ kDa can be achieved. Eighty-five percent signal assignments of the tetrameric KcsA potassium channel in DPC micelles (apparent molecular weight in the range of 130Â€kDa) have been obtained. On contrast, the protein size is in theory not a limiting factor for solid-state NMR (ssNMR) but in practice, spectral resolution impaired studies of large MPs and specific ssNMR strategies including specific stable-isotope labelling of proteins are currently under development focusing on this limiting factor. Both solution and solid-state NMR methods are currently used for the structural studies of small membrane peptides but are still very challenging for MPs. The main factors responsible for these difficulties are (1) sample preparations including large amount of pure protein in an appropriate membrane mimetic model and (2) signal overlaps due to the amino acid sequences of MPs composed generally of high repetitiveness of hydrophobic amino acids and to the presence of one dominant type of secondary structure either b-strands or a-helices. a-Helical MPs which represent the greatest number of MPs, tend to have more crowded and overlapping spectra compared to b-barrel MPs due to the similar chemical shift environment induced by the helical structures, the greater proportion of hydrophobic amino acids, and

Determination of Membrane Protein Structures Using Solution and Solid-State NMR

263

the lower variability of amino acid types found in transmembrane helices compared to b-barrels and probably also by the greater flexibility of helical bundles. Twenty-four unique MP structures including six b-barrel MPs, have been determined so far using NMR to a resolution sufficient to have resulted in a file deposited at the RCSB Protein Data Bank (http://www.drorlist.com/nmr/ MPNMR.html). Most of these structures were determined using solution NMR (14 in micelles and four in isotropic lipid bicelles), and the six structures determined using ssNMR were in planar synthetic bilayers. In this chapter, we will introduce two approaches to determine structures of MPs that are currently used and under development: 1. Solution NMR for small membrane protein in micelles or fast tumbling bicelles. 2. Solid-state NMR applied to MPs in lipid bilayers, large bicelles, or as microcrystals. Solution NMR methods applied to MPs involved essentially the methodology developed for large soluble proteins. They have been successfully applied to MP-containing detergent micelles leading not only to membrane protein structures but also to atomic resolution, dynamics, and interactions with partners (2). Solid-state NMR methods are much more challenging as reflected by the few published data reporting on full signal assignments and few deÂ€ novo structures of small MPs. Nevertheless, these methods are the subject of an active and rapidly developing field (3–6).

2. Samples Whatever the NMR method used, membrane protein sample quality is a key issue of a successful study and is first evaluated using solution NMR: pure, homogeneous, and stable isotope labelled protein should be available in large amount (i.e., mg) and its functionality in the membrane mimetic environment assessed. Moreover, the sample should present a good stability over time (typically weeks) to allow multidimensional spectra recording. Feasibility tests include checks on the structural homogeneity of the sample as it determines line widths and spectral quality (signalto-noise ratio, line widths, and spectral resolution). 2.1. Over-expression

The high amount of protein needed requires high expression level and as a consequence, bacterial expression in E. coli is the method of choice. MPs use the translocation machinery and have to be membrane inserted conversely to soluble proteins. Therefore, in

264

Montaville and Jamin

order to obtain milligrams quantities required for structural studies, high-throughput overproduction studies tend to rationalize MP expression. Recent studies showed that auto induction based protocols work better than IPTG triggered over-expression (7). However, this system requires depletion of glucose in the culture medium for the target protein expression to occur and thus, specific protocols had to be developed for 13C isotope labelling using 13 C glycerol as the main carbon source (8). However, in many cases, membrane protein expression levels in organisms are too low for structural investigations mainly due to the toxic effect of membrane protein insertion into the host membranes. An interesting alternative to the bacterial system is the cell free expression procedure (9). The inÂ€vitro use of cellular extracts containing the transcription and translation machinery prevents the unfavourable impact of protein expression arising from the host organism metabolism. In addition, membrane Â�protein over-produced in cell free system will not be stored in inclusion bodies, which require denaturation purification protocols, especially not convenient for a-helical membrane proteins. Instead, the protein will aggregate allowing subsequent solubilisation by mild detergents. Alternatively, cell free expression Â�system may also be used for directly getting the solubilised membrane protein by adding the suitable detergent to the mixture. Moreover, these efficient cell-free expression systems have low levels of Â�metabolic conversion and therefore are quite useful for selective isotope labelling. 2.2. Tag

Affinity tags are very useful to isolate the over-expressed protein and to perform buffer exchanges. Large tags such as maltose binding protein and glutathione S-transferase often improve the solubilisation of the membrane protein. The fusion protein is purified in its native state and then a specific proteolytic cleavage step is required to obtain the protein. However, due to its simple and straightforward purification protocol, expressing the membrane protein attached to a poly-histidine tag is the most common system used. Moreover, the small size of this tag limits the crowding of the NMR spectra, and its polar character reduces the risk of interference with the hydrophobic core of the protein; consequently, this tag is usually not cleaved off to reduce the number of purification steps and increase the final yield. However, an interaction of the tag with polar loops or water exposed domains cannot be excluded, and great care should be taken to the possible function alteration of the tagged membrane protein. In addition, this small unstructured tag allows denaturation purification protocols by the use of SDS or urea containing buffers. This feature is especially attractive for b-barrel membrane proteins obtained in inclusion bodies. However, it should be kept in mind that the positioning and the length of the histidine tag may strongly

Determination of Membrane Protein Structures Using Solution and Solid-State NMR

265

affect the expression level of membrane proteins independently of the membrane protein topology. 2.3. Labelling

Preparation of uniformly 15N and 13C labelled protein samples is needed both for solution and solid state NMR. A protocol for such labelling and that can be applied to other protein is described and discussed in detail in the case of TSPO (see Chapter 3). Amino acid type selective labelling is often required for improving resonance assignment completion by simplifying NMR spectra. Combinatorial 15N and 13C labelling of the six specific amino acid types (AFGILV) predominantly found in the transmembrane parts of MPs using a cell free based expression system constitute an improvement of the standard selective labelling approach (10). Other strategies for the labelling of MP have been developed associated with the specific NMR techniques used, i.e., solution or solid-state and are presented below. Moreover, novel strategies are currently under development, and the best scheme should be chosen depending on the protein amino acid sequence.

3. Methods 3.1. Solution NMR of Membrane Proteins 3.1.1. Sample Preparation 3.1.1.1. Choice of Detergent/Phopholipids

Solution NMR studies require the solubilisation of the membrane protein in detergent micelles. Several considerations should support the choice of the detergent. Above all, the solubilised membrane protein should be kept in its native functional state. When possible, functional assays should be performed in presence of the chosen detergent. The rotational overall correlation time of the protein-containing micelle should be kept as small as possible to yield well-resolved spectra. The protein should stay stable as long as possible in the presence of micelles, typically a life time in the weeks range is required for a complete set of NMR experiments to be recorded. The nature and the concentration of the detergent have to preserve the highest possible homogeneity of the protein population in the sample to avoid peak broadening or doubling. The system of choice for conducting NMR studies of a given MP will be a compromise of all the previously mentioned points. The samples are evaluated using 15N labelled protein and two-dimensional 15N–1H heteronuclear single-quantum correlation (HSQC). Because no ideal detergent can be, a priori, selected for a particular membrane protein, a HSQC based screening approach comparing the sample spectral quality across a range of detergents at different temperatures and in various buffers should be performed (11–13). Successful detergent used so far for membrane

266

Montaville and Jamin

protein solution NMR studies are dodecylphosphocholine (DPC), dihexanoyl-phosphatidylcholine (DHPC), N-octyl-b-dglucopyranoside (bOG), n-dodecyl-b-d-maltoside (DDM), lauryl dimethylamine oxide (LDAO), lyso-phosphatidylglycerol (LPPG) and sodium dodecyl sulfate (SDS) (14). The use of deuterated detergent is generally not necessary for most experiments. Indeed, the increased relaxation rate of the protein backbone resonances surrounded by detergent protons is not significant due to the highly dynamic character of micelles. Small isotropic bicelles are very attractive for studying membrane protein by solution NMR because they constitute the closest membrane like environment with motional properties still compatible with this technique. Therefore, this system has been suggested as a benchmark for 15N–1H HSQC based detergent screen (15). Due to their larger size compared to micelles, bicelles have been used so far only for small membrane peptides. A recent structure of the transmembrane domain dimer Bnip3 (16) is the largest MP structure in bicelles obtained so far by solution NMR. Different protocols have been proposed to prepare MP-containing bicelles: adding the long-chain phospholipid to a solution of protein solubilised in the short-chain phospholipid micelles, adding the short-chain phospholipid to a solution of protein reconstituted in the long-chain phospholipid vesicles, reconstitution of the protein in the bicelles directly from the lyophilised protein, from the detergent solubilised protein or from the protein solubilised in organic solvent. The use of nonhydrolysable ether-linked lipids to prepare stable bicelles increases their long term stability. 3.1.1.2. Labelling

Apart from fully uniformly (13C, 15N)- and amino acid-specific labelling and for large systems (typically over 30Â€ kDa), high degree of deuteration is required for recording good quality multidimensional spectra. This labelling is achieved by adding 15N and 13C labeled compounds as protein synthesis sources to a D2O based culture medium (see chapter 4). In most cases, fractionally deuterated samples are sufficient for protein backbone assignment. Growing cells in D2O based culture medium leads to about 80% deuteration. Perdeuteration is required for side chain and methyl group assignment, for this purpose, the additional use of the expensive perdeuterated 13C glucose is necessary. Lower than 80% level of deuteration would be required for aliphatic 1H detection; however, at 50% deuteration, fast magnetization relaxation leads to poor quality spectra. Deuteration is the main reason for the lack of side chain 1H assignment for membrane proteins and consequently, the reduced number of 1H–1H distance constraints derived from NOEs used to calculate NMR structures. Due to the high deuteration level required, specific labelling schemes have been developed for side chain assignment of large protein and applied to MPs (2). Addition of 13C, 2H, 1H methyl

Determination of Membrane Protein Structures Using Solution and Solid-State NMR

267

a-ketoisovalerate, and a-ketobutyrate in media suitable for 15N, 13 C, and 2H uniform protein labelling provides samples for which the three magnetically equivalent protons of methyl groups of Ile(d1), Leu, and Val in addition to the backbone 1HN are not deuterated. A specific set of NMR experiments has been developed in combination with this specific labelling scheme (see below). Such a labelling approach is determinant for improving the resolution of b-barrel membrane proteins NMR structures (Fig.Â€1a). For a-helical MPs, the methyl protons chemical shift assignment is not always possible due to the high degeneracy of 13 CO, 13Ca, and 1H13C methyl chemical shifts. However, only few methyl–methyl long range NOEs are sufficient to determine the tertiary fold and the final quality of the structure mainly arises from the accuracy of the secondary structure elements (17) (Fig.Â€1b). This method has been successfully included to determine the tertiary structure of DsbB in DPC micelles (18). Besides stable isotope labelling, other type of labelling may be required to obtain long range constraints such as the sulfhydrospecific coupling of a “spin label” or a paramagnetic metal ion chelating agent to a single cysteine introduced by site directed mutagenesis into the protein sequence (19, 20). 3.1.2. Methodology 3.1.2.1. Transverse Relaxation Optimized Spectroscopy (TROSY)

3.1.2.2. Backbone Chemical Shifts Assignments

The methodology used for MPs essentially comes from major advances achieved in solution NMR spectroscopy studies of large proteins. In particular, TROSY based NMR experiments have been developed to reduce transverse relaxation, which is responsible for signal loss and line broadening, and are thus systematically employed (21). The basic principle is the selection of the narrowest component (i.e., most slowly relaxing) of the multiplet signals of the 15N–1H moiety at high magnetic field. This selection improves the sensitivity and the resolution of “out and back” type experiments starting and ending on 1H amide. The TROSY effect is most efficient for deuterated proteins for which the dipolar 13C–H interaction is greatly reduced, increasing the magnetization lifetime and the magnetization transfert efficiency. Moreover, the TROSY effect is field dependent, for the 1H–15N pair, a magnetic field strength corresponding to a 1H resonance frequency between 900 and 1,100Â€ MHz would be optimum. Therefore, recording NMR spectra at high magnetic field is required (i.e., 1H resonance frequency ³700Â€MHz). Chemical shifts backbone assignment is a prerequisite to any protein NMR study. Because highly deuterated samples are required, 15 N–1H based NMR experiments are performed for assigning the spectra. The TROSY versions of 3D experiments HNCA, HN(CO) CA, HNCACB, CBCA(CO)NH, HN(CA)CO, and HNCO are the basic experiments recorded for assigning backbone chemical shifts. For helical proteins, due to intense 1HNi–1HNi±1 NOE,

268

Montaville and Jamin

Fig.Â€1. Constraints used for the determination of MPs structures using solution NMR. Schematic representation of the influence of backbone 1H15N NOE distance restraints (i), 1H methyl based NOE distance restraints (ii), and PRE based distance restraints (iii) in the process of obtaining a NMR structure of a b-barrel MP (a) or a a-helical MP (b). For illustration

Determination of Membrane Protein Structures Using Solution and Solid-State NMR

269

the combined use of TROSY HNCA and 15N edited NOESY TROSY could be a successful strategy as demonstrated in the case of Mistic (19) and KcsA (22). MPs backbone chemical shifts assignments benefit from recently developed strategies for high molecular weight soluble proteins. A remarkable example is the NMR study of the 82Â€kDa (723 residues) soluble protein malate synthase G (MSG) (23). The combination of a set of 4D TROSY spectra (4D TROSY HNCACO, 4D TROSY HNCOCA, 4D TROSY HNCOi−1CAi, and 4D 15N, 15N-edited NOESY) and the conventional TROSYbased 3D NMR experiments yield an almost complete (>95%) reliable backbone assignment. Moreover, the authors proposed a variant of the TROSY HNCACB experiment for the specific assignment of 15N, 1HN correlations of Ala and Alai+1 residues. As alanine is one of the most abundant residues found in hydrophobic stretches of MPs, the use of such pulse sequences seems very attractive. 3.1.2.3. Long Range Restraints

Calculation of protein global fold by NMR spectroscopy requires NOE-derived distance restraints as well as chemical shift derived dihedral angles restraints. In the case of b-barrel proteins, backbone 1HN–1HN NOEs are sufficient for getting the overall fold. For a-helical proteins, these NOEs are usually short to medium range distance restraints and consequently define only the secondary structure elements (TableÂ€1). The lack of long range distance restraints arise from the use of deuterated samples. Therefore, alternative strategies should be implemented to collect distance restraints in order to increase the resolution of NMR structures of membrane proteins (Fig.Â€1). Long range methyl–methyl and methyl–amide distance constraints are derived from NOEs between methyl/amide protons and methyl protons of leucine, valine, and isoleucine(d1) residues in a perdeuterated environment. Such NOEs derived constraints have been shown to significantly improve the resolution of the NMR structure of OmpX (24). However, for the a-helical membrane protein DAGK, this approach was hampered by the lower dispersion of chemical shifts (1H methyl, 13CO, 13Ca) compared to b-barrel proteins (14). In this case, a deuterated detergent is favoured to avoid signal overlap between the aliphatic resonances of the generally highly concentrated detergent and the aliphatic resonances of the protein.

Fig.Â€1. (continued) purpose, one structure of VDAC (PDB entry: 2k4t) has been used for (a) and one structure of DsbB (PDB entry: 2k73) has been used for (b). Note that for (a, b) (i) and (ii), subsets of experimental distance restraints are shown in black dashed lines, whereas for (a, b) (iii) magenta dashed lines do not account for published experimental data.

270

Montaville and Jamin

TableÂ€1 Number of NOEs distance restraints for a b-barrel protein (VDAC) and a a-helical protein (DsbB) NOE distance restraints

VDAC (278 residues)(2k4t)

Intraresidual

DsbB (176 residues)(2k73)

69

41

199

191

72

216

Long range (|iâ•›−â•›j|â•›>â•›4)

272

39

Total

612

487

Sequential (|iâ•›−â•›j|â•›=â•›1) Medium range (3â•›<â•›|iâ•›−â•›j|â•›<â•›5)

The large amount of long range NOEs observed for b-barrel proteins are sufficient for defining their 3D fold, whereas the large amount of medium range NOEs distance restraints allow a precise structure of the secondary structure elements for a-helical membrane proteins, and only a few long range NOE restraints are required to obtain the 3D fold of the protein

In the case of membrane protein oligomer systems, distinction between intra and inter monomeric NOE correlations is required to define the proper quaternary structure of the system. This distinction is obtained by differential subunit labelling protocols (25). A second set of long range distance restraints may be obtained by the use of the paramagnetic relaxation enhancement (PRE) effect. The covalent linkage of an unpaired electron containing tag to a specific location within the protein result in a dramatic sp increase of the transverse relaxation rate ( R2 ) of the neighbouring nuclear spin according to the inter-spin distance ( R2sp a r6). Comparing peak intensities and linewidths in TROSY spectra of spin labelled and corresponding diamagnetic proteins provide distance restraints up to 30â•›Å. This method has been used to refine the NMR structure of OmpA (20) and to determine the overall fold of the a-helical membrane protein Mistic (19). In order to delineate the protein parts embedded within the micelle and the regions of the MP which are solvent exposed, the combined use of spin labelled detergent or lipophilic paramagnetic probes and soluble paramagnetic compounds has been successfully employed (26, 27). 3.1.2.4. Residual Dipolar Couplings (RDCs)

The orientational constraints deduced from the measurement of RDC are widely used to refine structures of soluble proteins by NMR. In order to measure RDCs, several media can be chosen to induce the weak alignment of the protein sample such as the filamentous phage Pf1, bicelles, and compressed or stretched polyacrylamide gels. For MP-containing micelles characterized by high detergent concentrations, only the latter method is suitable

Determination of Membrane Protein Structures Using Solution and Solid-State NMR

271

and has been applied to measure RDCs for OmpA (20) and phospholamban (28). In order to increase the dataset of RDCs, these gels may be positively or negatively charged to induce distinct alignments of the protein (29). HN, C¢Ca, and NC¢ dipolar couplings may be obtained from IPAP-HSQC and TROSY HNCO experiments. Although not tried yet for membrane proteins, partial protein sample alignment could also be achieved using the covalent attachment of a paramagnetic ion chelating tag. This tagging method is similar to the procedure employed for measuring PREs and could be an alternative to the polyacrylamide gels approach (30). Moreover, different alignments could be achieved varying the lanthanide ion, and therefore, more RDCs could be measured. 3.1.2.5. New Methodological Approaches

C direct detection has been recently proposed as an alternative to the classical more sensitive 1H detection based NMR experiments (31). The slower transverse relaxation rate of 13C, especially upon deuteration labelling, compared to 1H makes the detection of this nucleus very attractive for large biomolecular systems. A whole set of 2D and 3D protonless experiments have been designed to allow backbone and side chain chemical shifts assignments. Identification of 13C resonances within each residue is obtained by 2D/3D CACO, CBCACO, CCCO, and CGCB experiments using various homonuclear decoupling schemes (IPAP, S3E, DIPAP). CON-IPAP, CBCACON-IPAP, and CCCON-IPAP experiments correlate the 13C resonances to the 15 N backbone resonance, and CANCO-IPAP experiments Â�correlate resonances from sequential residues. In the case of Smr solubilised in bicelles containing solution, the combined use of the 2D CACO and TROSY modified HNCA and HNCO Â�experiments yields to more than 80% of COCA correlations assignments (15). One approach to deal with signal overlap is to increase the dimensionality of the NMR spectra without reduction of signal and resolution. Such an approach has been recently proposed and applied to assign the resonances of subunit c of E. coli F1F0 ATP synthase (32). This method, based on the phase sensitive joint sampling of indirect spectral dimensions, allows recording high dimensionality NMR spectra without reduction of resolution compared to their conventional 3D counterparts and within less measurement time. Application to a-helical membrane proteins NMR studies seems particularly suitable considering extensive signal overlap as well as the generally time limited stability of membrane protein NMR samples.

3.1.2.6. Toward MP Dynamics and Mechanisms

One of the main advantages of solution NMR studies of proteins is the ability of probing dynamics for a wide timescale range (typically from pico to milliseconds) at the atomic level. Two recent

13

272

Montaville and Jamin

examples in the literature illustrate how NMR spectroscopy can provide a description of key functional dynamics processes at atomic level. A detailed NMR work performed on the integral membrane enzyme PagP in a detergent compatible with its function, provided a high resolution structure of PagP and established that this protein is in equilibrium between two conformational states (33–35). Chemical shift data indicate that the main structural and dynamics changes between these states involved residues surrounding the catalytic site. It has been proposed that the R state (the more flexible state) may facilitate the ligand entry in the b-barrel core and the more rigid T state may be important for catalysis. Moreover, this enzyme has been further characterized as the relative population of R and T states as well as the exchange rate between R and T and the equilibrium constant were quantified at different temperatures using NMR data. The functional implications of this detailed structural and dynamics description of PagP remain to be investigated. A NMR study of the KcsA potassium channel in DPC micelles at pH 4 (open state) and pH 7 (closed state) provided hints concerning the role of local dynamics in the channel function (22). The NMR signals of residues located in two spatially distinct but functionally important regions of the protein (the selectivity filter and the C-terminal part of TM2) were monitored at different pH values (between 4 and 7). The structural changes observed at different pH values are correlated with the pH dependent open probability profile of the channel measured by functional tests. Occupancy and transition rates between both conformations for residues belonging to the selectivity filter and the C-terminal part of TM2 were measured and appeared to be different at pH 4 and pH 7. Combining these local dynamics data provided a qualitative model of KcsA function in agreement with its simultaneous gating and ion selectivity properties. 3.2. Solid-State NMR of Membrane Proteins

Two different approaches yield high-resolution solid-state NMR spectra. Oriented-sample ssNMR exploits the orientationdependence of the anisotropic interactions. The unique orientation simplifies the spectra and the magnitude and orientation of chemical shift and dipolar coupling tensors provide structural information. Rather than exploiting sample orientation, magic-angle spinning (MAS) NMR averages orientation-dependent interactions through fast sample rotation at the magic angle (54.7°) relative to the external magnetic field while selectively recoupling the interactions needed to obtain structural information during specific periods of the experiment. Structures of small helical MPs have already been obtained using oriented ssNMR (see TableÂ€2). MAS NMR has been applied with success for the determination

18

25

25

36

50

a-Receptor M6

AchR M2 channel

Influenza M2 channel

HIV Vpu channel

fd coat protein

2 (39)

1 (18)

1 (25)

1 (15)

2 (14)

No. of helicesa

POPC/POPG

DOPC/DOPG

DMPC

DMPC

DMPC

Lipid bilayerb

8/80

3.5/75

20/38.5

2/40

4/50

4

5

10

? d

10

400

700

b

a

1MZT

1PJE

1NYJ

1CEK

450 and 500 300 and 400

1PJD

PDB code

700

Number of residues in helices is in parenthesis All samples were aligned on glass plates c Apart from uniformly 15N-labelled sample, selectively 15N-labelled samples were used for both the assignment and structure determination d The number of selectively labelled samples used for the assignment and structure determination is not published

No. of a.a.

Protein

No. of selectively Protein/lipid mg/ 15N-labelled Magnetic field mg samplesc strength (MHz)

TableÂ€2 Small helical MPs which structure have been determined using solid-state NMR

Determination of Membrane Protein Structures Using Solution and Solid-State NMR 273

274

Montaville and Jamin

of 3D structure of soluble protein prepared as microcrystalline sample, and its application to MP is foreseen in a near future. 3.2.1. High Resolution Solid-State of Static Aligned Sample

This method provides very precise orientational constraints for structure determination in membrane bilayers as well as the orientation of the MP in the membrane. All the structures of small membrane peptides/proteins in the PDB determined using ssNMR have been obtained using this method. The largest MP structure is that of fd coat protein, composed of two transmembrane domains. However, its application to larger MP is challenging due to the difficulties to obtain uniform aligned MPs and resonance overlaps (3, 36).

3.2.1.1. Aligned Samples

Sample alignment is a prerequisite of this method and can be very challenging especially for MPs (3, 4). It can be obtained either by mechanical alignment of bilayers on glass plates or by magnetic alignment of bicelles. Mechanical alignment of lipid bilayers.â•… The first step is to prepare proteoliposomes via either organic solvent (for TM peptides) or detergent mediated (for TM peptides and MPs) reconstitution (see Chapter 18 for an example). The choice of organic solvent, detergent, lipids as well as the molar ratio protein to lipid will depend on the peptide/MP and previous synthesis and purification steps. The second step is the preparation of the oriented bilayers on glass plates. Aliquots of a solution of proteoliposomes is spread onto thin glass plates (30–50) and slowly evaporated. The glass plates are stacked, inserted into a sample cell, and are then rehydrated by incubation in a chamber where the humidity is controlled by a saturated salt solution at a temperature above the gel to liquid crystalline phase transition temperature. Oriented bilayers then form after equilibrating the sample in this chamber, and care should be taken to maintain sample hydration during the NMR experiment. Bicelles.â•… Protein-containing bicelles for which the long chain to short-chain phospholipid ratio is between 2.8 and 6.5 will spontaneously align in a magnetic field at temperatures above the gelto-liquid crystalline phase transition temperature of the long-chain phospholipids such that the bilayer normal is perpendicular to the magnetic field. Protocols for preparing stable protein-containing bicelles samples that yield high-resolution ssNMR spectra have recently been published in details (37). The spectral resolution can be improved by using bicelles that align with their bilayer normal parallel to the magnetic field as a result of an increase by a factor two of the spectral range of the chemical shift. This can be obtained with conventional bicelles samples by the addition of lanthanide ions or as demonstrated recently with bicelles containing a lipid with a biphenyl group in one of the acyl chains mixed with DHPC (38).

Determination of Membrane Protein Structures Using Solution and Solid-State NMR

275

The direction and extent of the alignment of the bilayer and the formation of bicelles can be verified with 31P. Protein alignment is checked with a 1D 15N ssNMR spectrum. This spectrum should display significant resolution with identifiable peaks and no evidence of underlying “powder pattern” intensity which would indicate the presence of unoriented material (37). The main advantage of mechanically over magnetically alignment is the choice of lipid composition. Disadvantages of planar bilayers are the difficulties to change easily the environment, to maintain the sample hydration in the presence of high decoupling power, and the necessity to adapt the RF coil to the geometry of the sample. The main disadvantages of bicelles are the limited choice of lipid combinations and the temperature range defined by its lipid composition. 3.2.1.2. Resonance Assignments and Structure Determination

ssNMR on uniaxially oriented samples enables direct correlation between the resonance frequencies of the protein backbone nuclei and the orientation of the peptide plane to which these nuclei belong relative to the bilayer normal. The PISEMA (polarization inversion spin exchange at the magic angle) spectra correlate 15N chemical shift of the backbone amides and 15N–1H dipolar couplings. These spectra exhibit characteristic wheel-like patterns of resonances, PISA wheels (polarity index slant angle), that reflect both the protein structure and orientation in the membrane (3, 4). The analysis of these spectra has been first developed for small helical proteins and was recently applied to a b-barrel MP (39). For uniform helices, the resonances within the PISEMA spectra form PISA wheels with 3.6 resonances per turn and the shape and position of these PISA wheels depend on the helix geometry and its orientation relative to the membrane. PISA wheels can be complemented by dipolar waves, which reflect the periodic wave-like variations of the magnitude of 15N–1H dipolar couplings as a function of residue number. These dipolar waves provide the number of residues in the helix as well as its tilt and rotation angle and are useful to identify helical kinks or curvature. The strategy developed on small MPs relies on the spectra of one uniformly 15N labelled sample and several selectively 15N labelled samples combined with the symmetry properties of the PISA wheel to enable simultaneous resonance assignment and determination of orientational constraints (40). In the PISEMA spectra, resonances from transmembrane helix, from helix parallel to the bilayer plane and from loop and terminal regions without regular secondary structure are in distinct parts of the spectra (Fig.Â€2). An initial model of the MP with the different helices and their orientation relative to the bilayer plane is generated from pattern recognition of the resonances in the different regions of the PISEMA spectrum of a uniformly 15N-labelled sample. This model is then further refined by rotating separately each

276

Montaville and Jamin

Fig.Â€2. Ideal PISA wheels representation of an ideal a-helix tilted at 30° (a) and 90° (b) relative to the direction of the applied magnetic field B0.

helix around its axis until the resonances patterns of the calculated PISA wheel match the resonances of the experimental spectra of selectively labelled samples. This protocol yields the rotation of the helices and the assignments of the resonances. In the next step, the 15N chemical shift and 15N–1H dipolar frequencies are measured in the PISEMA spectra and are used to determine the polar angles that describe the peptide plane orientation in the magnetic field and provide input for structure calculation. For more complex small MPs, each resonance must be assigned before the analysis of the dipolar waves and the structure calculation (41). Other strategies have been developed such as fitting the spectrum to a model structure assuming a constant peptide plane geometry (4, 42). Current developments include the design of tripleresonance experiments for resolution of the overlap of resonances and for measuring additional orientational constraints (43). 3.2.2. High Resolution MAS NMR

Although no MP 3D structure have been determined using this technique, sample preparation and assignments strategies have been published and are briefly summarized below (44, 45).

3.2.2.1. Sample Preparation

MPs can be studied as microcrystal or proteoliposomes. Microcrystal may be obtained from precipitates of solubilized MPs in detergent micelles or bicelles. There is no general procedure for optimal sample preparation for ssNMR and varied precipitation conditions are generally tested to obtain the highest spectral quality (signal-tonoise ratio, resolution) as assessed by 2D 13C–13C homonuclear spectra. Typical narrow line widths of ~0.5–0.7Â€ppm for 13C and ~0.7–5Â€ppm for 15N are expected to give high resolution spectra.

Determination of Membrane Protein Structures Using Solution and Solid-State NMR

277

Detergent-purified membrane proteins can also be reconstituted into lipid bilayers to form proteoliposomes or 2D crystals. The sample is then packed into a rotor (about 40Â€ml standard volume). Due to the low sensitivity of the heteronuclear 3D NMR experiments, large quantity of pure and conformationally homogeneous protein sample are needed. In the case of proteoliposomes, the difficulty of getting a large quantity of proteins into the rotor is linked to the presence of lipids as typical 1:100 molar protein/ lipid ratios are used to avoid protein aggregation. The use of crystalline samples overcomes this concentration problem but at the detriment of the bilayer environment. Finally, the sample should be stable for several days due to the necessity to record several 2D and 3D experiments and compatible with sample heating that can result from strong proton decoupling during acquisition and sample spinning at high speed (~10–20Â€kHz). It has been noticed that 2D crystals seem to yield better resolved lines than proteoliposomes (46). 13 C and 15N signals are detected in ssNMR as 1H signals give large line widths (several kilohertz) at the standard spinning rates used. Therefore, 15N and 13C labelled protein samples should be prepared. Apart from fully uniformly (13C, 15N)-labeled proteins samples, different labelling strategies have been proposed for both signal assignments and structure calculation. In order to reduce spectral crowding, either amino acid-specific and/or reverse labelling can be used. Reverse 13C and 15N-labelling is obtained by adding amino acids of natural 13C and 15N abundance to the expression medium and therefore, the resulting spectra are devoid of 13C and 15N signals from these amino acids. Other specific labelling schemes have been developed in combination with specific NMR techniques. In particular, in order to avoid the strong couplings between directly bonded spins that would interfere with the weak coupling yielding distance information, reduced 13C-block-labelling schemes are used (47). This reduced labelling or spin dilution can be obtained by using [1, 3-13C]-glycerol and [2-13C]-glycerol as the sole nutrient carbon source in bacteria medium and results in an alternating labelling scheme where carbon sites are labelled but not in adjacent positions. In combination with this labelling scheme, long-range distances constraints are obtained using a broad-band recoupling method like the proton-driven spin diffusion (PDSD) mixing scheme (48). 3.2.2.2. Resonance Assignments

The full sequence-specific assignment of membrane protein by ssNMR remains a difficult challenge. It concerns 13C and 15N resonances since the proton resonances are not sufficiently well-resolved. The strategy relies on well dispersed resonances in the 15N dimension and on 13C or 15N heteronuclei correlations, employing NCACX and NCOCX type experiments. NCACX experiments correlate 15 N, CA with 13C of the side chain while NCOCX experiments correlate 15N, CO, and 13C of the backbone/side chains (49).

278

Montaville and Jamin

However, these two experiments are generally not sufficient for uniformly labelled proteins especially for proteins that contain a high number of hydrophobic residues in helical conformation due to the degeneracy of 13C and 15N chemical shifts. For samples containing large quantity of protein, two additional experiments can be recorded, CAN(CO)CX and CON(CA)CX 3D which provide unambiguous interresidue 13C–13C correlations (44). In case of low resolution, dispersion, and sensitivity of NCACX and NCOCX spectra, the assignments would mainly rely on homonuclear carbon correlations and specific amino acid labelling schemes. Similar to solution NMR, the assignment protocol involved first the sequential assignment which is composed of the following steps: spin system identification, linking the spin systems and amino acid type identification. 3.2.2.3. Distance Restraints and Structure

Structure-determination protocols are currently developed for and with small soluble proteins in micro-crystalline state. As a matter of fact, five 3D structures of soluble proteins in microcrystalline state have been determined using MAS (50). Two different strategies have been proposed. The first one is based on the estimation of a large number of 13C–13C distances using protondriven spin-diffusion (PDSD) experiments performed on reduced 13 C-block-labelled samples (51). It necessitates extensive labelling strategies, removing one-bond couplings by partial labelling and has been used for two proteins (spectrin SH3 domain, 62 residues and ubiquitin, 76 residues). The second strategy involves the indirect measurement of proton distances using proton mediated 13C/15N detected experiments performed on a uniformly 13 C/15N labelled sample (13C,13C-encoded 1H,1H mixing, i.e., CHHC and 15C,13C-encoded 1H,1H mixing, i.e., NHHC experiments) (52). This strategy has the advantage that only one sample is required but the experiments are not very sensitive and large amount of protein is necessary. In addition to distance constraints, empirical chemical shifts restraints (torsion angle likelihood obtained from shift and sequence similarity (TALOS)) are combined to improve the structure calculations.

4. Conclusion and Perspectives The success of both solution and ssNMR applied to the determination of MP 3D structure is first highly correlated to the improved protocols for obtaining high amount of homogeneous and functional protein in a membrane mimic environment which yield high spectral quality.

Determination of Membrane Protein Structures Using Solution and Solid-State NMR

279

Membrane protein solution NMR is still a challenging field despite the increased releasing rate of structures published in the last years. The field of large soluble protein structural investigations by NMR mainly contributed to the early successful studies of membrane proteins. The use of specific approaches such as paramagnetic reagents or specific labelling of hydrophobic residues predominantly found in the transmembrane domains provided essential tools for structural determinations of membrane proteins. Recent methodological developments in the field of large protein systems such as 13C direct detection include nowadays MPs as one of their main applications. Solution NMR has already addressed various aspects of membrane protein function such as protein–protein interactions (53), protein–detergent interactions (13, 26), drug screening (54). The ability to study MPs in bicelles (the most native like environment still compatible with solution NMR), will open the application of this technique to more specific aspects such as MP–lipids interactions, MP–MP interactions or membrane protein folding. Recent advances in ssNMR have provided 3D structures of small MPs in phospholipids bilayers. Besides 3D structures, ssNMR already provides essential structural and dynamics data on large MPs (55, 56), as well as more specific information related to MP–protein interactions (57), seven TM helices GPCR topology (58), MP–MP interaction (59, 60), and orientation of MPs in the bilayer (39). Increasing resolution and sensitivity of ssNMR technique is now essential for studying at atomic level more complex functional MPs systems in their native environment. Thanks to recent developments, both solid-state and solution state NMR are essential tools that, when used in combination with other experimental structural techniques such as X-ray crystallography, electron microscopy, AFM, and EPR, provide a powerful set of structural techniques for unprecedented detailed description of membrane proteins systems. References 1. Vold RR, Prosser RS (1996) Magnetically oriented phospholipid bilayered micelles for structural studies of polypeptides. Does the ideal bicelle exist? J Magn Res B 113: 267–271 2. Tugarinov V, Hwang PM, Kay LE (2004) Nuclear magnetic resonance spectroscopy of high-molecular-weight proteins. Annu Rev Biochem 73:107–146 3. Page RC, Li C, Hu J, Gao FP, Cross TA (2007) Lipid bilayers: an essential environment for the understanding of membrane proteins. Magn Reson Chem 45:S2–S11

4. De Angelis AA, Jones DH, Grant CV, Park SH, Mesleh MF, Opella SJ (2005) NMR experiments on aligned samples of membrane proteins. Methods Enzymol 394:350–382 5. Watts A, Straus SK, Grage SL, Kamihira M, Lam YH, Zhao X (2004) Membrane protein structure determination using solid-state NMR Methods. Mol Biol 278:403–473 6. Nielsen N, Malmendal A, Vosegaard T (2004) Techniques and applications of NMR to membrane proteins. Mol Membr Biol 21:129–141 7. Gordon E, Horsefield R, Swarts HG, de Pont JJ, Neutze R, Snijder A (2008) Effective

280

8.

9.

10.

11.

12.

13.

14. 15.

16.

Montaville and Jamin high-throughput overproduction of membrane proteins in Escherichia coli. Protein Expr Purif 62:1–8 Tyler RC, Sreenath HK, Singh S, Aceti DJ, Bingman CA, Markley JL, Fox BG (2005) Auto-induction medium for the production of [U-15N)- and [U-13C, U-15N)-labeled proteins for NMR screening and structure determination. Protein Expr Purif 40:268–278 Koglin A, Klammt C, Trbovic N, Schwarz D, Schneider B, Schafer B, Lohr F, Bernhard F, Dotsch V (2006) Combination of cell-free expression and NMR spectroscopy as a new approach for structural investigation of membrane proteins. Magn Reson Chem 44: S17–S23 Reckel S, Sobhanifar S, Schneider B, Junge F, Schwarz D, Durst F, Lohr F, Guntert P, Bernhard F, Dotsch V (2008) Transmembrane segment enhanced labeling as a tool for the back bone assignment of alpha-helical membrane proteins. Proc Natl Acad Sci U S A 105:8262–8267 Krueger-Koplin RD, Sorgen PL, KruegerKoplin ST, Rivera-Torres IO, Cahill SM, Hicks DB, Grinius L, Krulwich TA, Girvin ME (2004) An evaluation of detergents for NMR structural studies of membrane proteins. J Biomol NMR 28:43–57 Page RC, Moore JD, Nguyen HB, Sharma M, Chase R, Gao FP, Mobley CK, Sanders CR, Ma L, Sonnichsen FD, Lee S, Howell SC, Opella SJ, Cross TA (2006) Comprehensive evaluation of solution nuclear magnetic resonance spectroscopy sample preparation for helical integral membrane proteins. J Struct Funct Genomics 7:51–64 Zhang Q, Horst R, Geralt M, Ma X, Hong WX, Finn MG, Stevens RC, Wuthrich K (2008) Microscale NMR screening of new detergents for membrane protein structural biology. J Am Chem Soc 130:7357–7363 Sanders CR, Sonnichsen F (2006) Solution NMR of membrane proteins: practice and challenges. Magn Reson Chem 44:S24–S40 Poget SF, Girvin ME (2007) Solution NMR of membrane proteins in bilayer mimics: small is beautiful, but sometimes bigger is better. Biochim Biophys Acta 1768:3098–3106 Bocharov EV, Pustovalova YE, Pavlov KV, Volynsky PE, Goncharuk MV, Ermolyuk YS, Karpunin DV, Schulga AA, Kirpichnikov MP, Efremov RG, Maslennikov IV, Arseniev AS (2007) Unique dimeric structure of BNip3 transmembrane domain suggests membrane permeabilization as a cell death trigger. J Biol Chem 282:16256–16266

17. Shih SC, Stoica I, Goto NK (2008) Investigation of the utility of selective methyl protonation for determination of membrane protein structures. J Biomol NMR 42:49–58 18. Zhou Y, Cierpicki T, Flores Jimenez RH, Lukasik SM, Ellena JF, Cafiso DS, Kadokura H, Beckwith J, Bushweller JH (2008) NMR solution structure of the integral membrane enzyme DsbB: functional insights into DsbBcatalyzed disulfide bond formation. Mol Cell 31:896–908 19. Roosild TP, Greenwald J, Vega M, Castronovo S, Riek R, Choe S (2005) NMR structure of Mistic, a membrane-integrating protein for membrane protein expression. Science 307:1317–1321 20. Cierpicki T, Liang B, Tamm LK, Bushweller JH (2006) Increasing the accuracy of solution NMR structures of membrane proteins by application of residual dipolar couplings. High-resolution structure of outer membrane protein. A J Am Chem Soc 128:6947–6951 21. Fernandez C, Wider G (2003) TROSY in NMR studies of the structure and function of large biological macromolecules. Curr Opin Struct Biol 13:570–580 22. Baker KA, Tzitzilonis C, Kwiatkowski W, Choe S, Riek R (2007) Conformational dynamics of the KcsA potassium channel governs gating properties. Nat Struct Mol Biol 14:1089–1095 23. Tugarinov V, Muhandiram R, Ayed A, Kay LE (2002) Four-dimensional NMR spectroscopy of a 723-residue protein: chemical shift assignments and secondary structure of malate synthase g. J Am Chem Soc 124: 10025–10035 24. Fernandez C, Hilty C, Wider G, Guntert P, Wuthrich K (2004) NMR structure of the integral membrane protein OmpX. J Mol Biol 336:1211–1221 25. Oxenoid K, Chou JJ (2005) The structure of phospholamban pentamer reveals a channellike architecture in membranes. Proc Natl Acad Sci U S A 102:10870–10875 26. Hilty C, Wider G, Fernandez C, Wuthrich K (2004) Membrane protein–lipid interactions in mixed micelles studied by NMR spectroscopy with the use of paramagnetic reagents. Chembiochem 5:467–473 27. Beel AJ, Mobley CK, Kim HJ, Tian F, Hadziselimovic A, Jap B, Prestegard JH, Sanders CR (2008) Structural studies of the transmembrane C-terminal domain of the amyloid precursor protein (APP): does APP function as a cholesterol sensor? Biochemistry 47:9428–9446

Determination of Membrane Protein Structures Using Solution and Solid-State NMR 28. Mascioni A, Eggimann BL, Veglia G (2004) Determination of helical membrane protein topology using residual dipolar couplings and exhaustive search algorithm: application to phospholamban. Chem Phys Lipids 132:133–144 29. Cierpicki T, Bushweller JH (2004) Charged gels as orienting media for measurement of residual dipolar couplings in soluble and integral membrane proteins. J Am Chem Soc 126:16259–16266 30. Kamen DE, Cahill SM, Girvin ME (2007) Multiple alignment of membrane proteins for measuring residual dipolar couplings using lanthanide ions bound to a small metal chelator. J Am Chem Soc 129:1846–1847 31. Bermel W, Bertini I, Felli IC, Lee YM, Luchinat C, Pierattelli R (2006) Protonless NMR experiments for sequence-specific assignment of backbone nuclei in unfolded proteins. J Am Chem Soc 128: 3918–3919 32. Zhang Q, Atreya HS, Kamen DE, Girvin ME, Szyperski T (2008) GFT projection NMR based resonance assignment of membrane proteins: application to subunit C of E. coli F(1)F(0) ATP synthase in LPPG micelles. J Biomol NMR 40:157–163 33. Hwang PM, Choy WY, Lo EI, Chen L, Forman-Kay JD, Raetz CR, Prive GG, Bishop RE, Kay LE (2002) Solution structure and dynamics of the outer membrane enzyme PagP by NMR. Proc Natl Acad Sci U S A 99:13560–13565 34. Hwang PM, Bishop RE, Kay LE (2004) The integral membrane enzyme PagP alternates between two dynamically distinct states. Proc Natl Acad Sci U S A 101:9618–9623 35. Hwang PM, Kay LE (2005) Solution structure and dynamics of integral membrane proteins by NMR: a case study involving the enzyme PagP. Methods Enzymol 394: 335–350 36. Li C, Gao P, Qin H, Chase R, Gor’kov PL, Brey WW, Cross TA (2007) Uniformly aligned full-length membrane proteins in liquid crystalline bilayers for structural characterization. J Am Chem Soc 129:5304–5305 37. De Angelis AA, Opella SJ (2007) Bicelle samples for solid-state NMR of membrane proteins. Nat Protoc 2:2332–2338 38. Park SH, Loudet C, Marassi FM, Dufourc EJ, Opella SJ (2008) Solid-state NMR spectroscopy of a membrane protein in biphenyl phospholipid bicelles with the bilayer normal parallel to the magnetic field. J Magn Reson 193:133–138

281

39. Mahalakshmi R, Marassi FM (2008) Orientation of the Escherichia coli outer membrane protein OmpX in phospholipid bilayer membranes determined by solid-state NMR. Biochemistry 47:6531–6538 40. Marassi FM, Opella SJ (2003) Simultaneous assignment and structure determination of a membrane protein from NMR orientational restraints. Protein Sci 12:403–411 41. De Angelis AA, Howell SC, Nevzorov AA, Opella SJ (2006) Structure determination of a membrane protein with two trans-membrane helices in aligned phospholipid bicelles by solid-state NMR spectroscopy. J Am Chem Soc 128:12256–12267 42. Achuthan S, Asbury T, Hu J, Bertram R, Cross TA, Quine JR (2008) Continuity conditions and torsion angles from ssNMR orientational restraints. J Magn Reson 191:24–30 43. Sinha N, Grant CV, Park SH, Brown JM, Opella SJ (2007) Triple resonance experiments for aligned sample solid-state NMR of (13)C and (15)N labeled proteins. J Magn Reson 186:51–64 44. Li Y, Berthold DA, Gennis RB, Rienstra CM (2008) Chemical shift assignment of the transmembrane helices of DsbB, a 20-kDa integral membrane enzyme, by 3D magicangle spinning NMR spectroscopy. Protein Sci 17:199–204 45. Huang L, McDermott AE (2008) Partial sitespecific assignment of a uniformly (13)C, (15) N enriched membrane protein, light-harvesting complex 1 (LH1), by solid state NMR. Biochim Biophys Acta 1777:1098–1108 46. Lorch M, Fahem S, Kaiser C, Weber I, Mason AJ, Bowie JU, Glaubitz C (2005) How to prepare membrane proteins for solid-state NMR: a case study on the alpha-helical integral membrane protein diacylglycerol kinase from E. coli. Chembiochem 6:1693–1700 47. Pauli J, van Rossum B, Forster H, de Groot HJ, Oschkinat H (2000) Sample optimization and identification of signal patterns of amino acid side chains in 2D RFDR spectra of the alpha-spectrin SH3 domain. J Magn Reson 143:411–416 48. Manolikas T, Herrmann T, Meier BH (2008) Protein structure determination from 13C spin-diffusion solid-state NMR spectroscopy. J Am Chem Soc 130:3959–3966 49. Pauli J, Baldus M, van Rossum B, de Groot H, Oschkinat H (2001) Backbone and sidechain 13C and 15N signal assignments of the alpha-spectrin SH3 domain by magic angle spinning solid-state NMR at 17.6 Tesla. Chembiochem 2:272–281

282

Montaville and Jamin

50. http://www.drorlist.com/nmr/SPNMR. html 51. Castellani F, van Rossum B, Diehl A, Schubert M, Rehbein K, Oschkinat H (2002) Structure of a protein determined by solid-state magicangle-spinning NMR spectroscopy. Nature 420:98–102 52. Lange A, Becker S, Seidel K, Giller K, Pongs O, Baldus M (2005) A concept for rapid protein-structure determination by solid-state NMR spectroscopy. Angew Chem Int Ed Engl 44:2089–2092 53. Malia TJ, Wagner G (2007) NMR structural investigation of the mitochondrial outer membrane protein VDAC and its interaction with antiapoptotic Bcl-xL. Biochemistry 46:514–525 54. Assadi-Porter FM, Tonelli M, Maillet E, Hallenga K, Benard O, Max M, Markley JL (2008) Direct NMR detection of the binding of functional ligands to the human sweet receptor, a heterodimeric family 3 GPCR. J Am Chem Soc 130:7212–7213 55. Etzkorn M, Kneuper H, Dunnwald P, Vijayan V, Kramer J, Griesinger C, Becker S, Unden G, Baldus M (2008) Plasticity of the PAS domain and a potential role for signal transduction in the histidine kinase DcuS. Nat Struct Mol Biol 15:1031–1039

56. Ader C, Schneider R, Hornig S, Velisetty P, Wilson EM, Lange A, Giller K, Ohmert I, Martin-Eauclaire MF, Trauner D, Becker S, Pongs O, Baldus M (2008) A structural link between inactivation and block of a K+ channel. Nat Struct Mol Biol 15:605–612 57. Lange A, Giller K, Hornig S, Martin-Eauclaire MF, Pongs O, Becker S, Baldus M (2006) Toxin-induced conformational changes in a potassium channel revealed by solid-state NMR. Nature 440:959–962 58. Etzkorn M, Martell S, Andronesi OC, Seidel K, Engelhard M, Baldus M (2007) Secondary structure, dynamics, and topology of a sevenhelix receptor in native membranes, studied by solid-state NMR spectroscopy. Angew Chem Int Ed Engl 46:459–462 59. Seidel K, Andronesi OC, Krebs J, Griesinger C, Young HS, Becker S, Baldus M (2008) Structural characterization of Ca(2+)ATPase-bound phospholamban in lipid bilayers by solid-state nuclear magnetic resonance (NMR) spectroscopy. Biochemistry 47: 4369–4376 60. Traaseth NJ, Verardi R, Torgersen KD, Karim CB, Thomas DD, Veglia G (2007) Spectroscopic validation of the pentameric structure of phospholamban. Proc Natl Acad Sci U S A 104: 14676–14681

Chapter 15 Membrane Protein Fragments Reveal Both Secondary and Tertiary Structure of Membrane Proteins Philip L. Yeagle and Arlene D. Albert Abstract Structural data on membrane proteins, while crucial to understanding cellular function, are scarce due to difficulties in applying to membrane proteins the common techniques of structural biology. Fragments of membrane proteins have been shown to reflect, in many cases, the secondary structure of the parent protein with fidelity and are more amenable to study. This chapter provides many examples of how the study of membrane protein fragments has provided new insight into the structure of the parent membrane protein. Key words: Membrane protein, Structure, Peptide, Secondary structure, NMR, Loop, Transmembrane helix, Protein fragment, Structural biology

1. Introduction Membrane proteins enable and regulate a wide variety of cell functions such as cell signaling, membrane transport, biosynthesis, and cell morphology. Membrane proteins are major targets for drug development. Three-dimensional structure is as important for membrane proteins as it is for soluble proteins in understanding function, regulation, and external control by pharmaceuticals. A severe deficit of membrane protein structures at the atomic level, however, hampers our understanding of the function of these proteins in cellular membranes. The protein structure databases contain tens of thousands of atomic structures of proteins, but less than 0.5% of the structures are of membrane proteins. One database lists fewer than 200 unique membrane protein structures determined to atomic level resolution (1). This information deficit delineates the frontier of structure biology today: developing new approaches and refining existing Jean-Jacques Lacapère (ed.), Membrane Protein Structure Determination: Methods and Protocols, Methods in Molecular Biology, vol. 654, DOI 10.1007/978-1-60761-762-4_15, © Springer Science+Business Media, LLC 2010

283

284

Yeagle and Albert

techniques to increase the breadth of structural knowledge of membrane proteins. It is only with such structural information that mechanistic understanding of cellular processes can be expressed at the atomic level. And it is such structural information that informs most effectively the development of new interventions to modify function, such as therapies for diseases. The major methodology behind the available structural information for membrane proteins is X-ray crystallography, just as it is for soluble proteins. Many examples will be seen in other chapters in this volume. However, X-ray crystallography has generated far fewer structures of membrane proteins than it has for soluble proteins because membrane proteins, which are not soluble in aqueous media, are not readily amenable to the commonly used technologies for crystal growth required by X-ray crystallography. Therefore, alternative methodologies of structural biology have been explored in the search for new membrane protein structural information. The major alternative methodology for determining highresolution structures of soluble proteins is nuclear magnetic resonance (NMR). Modern NMR approaches have led to an explosion of structural data derived from proteins in solution. This is particularly important because it circumvents the crystallization requirement. A number of investigators have worked to translate this approach from soluble protein structural determination to the determination of the structures of membrane proteins. Unfortunately, the insolubility of membrane proteins in aqueous media has inhibited the application of solution NMR methodology. Isolated membrane proteins reconstituted into a lipid bilayer form a complex too large to experience rotational diffusion rapid enough for solution NMR. To address this problem, membrane proteins have been solubilized in detergent micelles, which are much smaller than reconstituted bilayers and hence experience much more rapid rotational diffusion. By using detergent solubilized proteins, structures for several b-barrel membrane proteins from bacteria (porins) have now been obtained by NMR. Despite the success achieved with b-barrel proteins, only one complete high-resolution structure has been published strictly from NMR data for a membrane protein containing a transmembrane a-helical bundle. Given the importance of this class of transmembrane proteins (likely the majority of transmembrane proteins), it was critical to develop other methodologies to obtain structural information. This imperative drove the development of an alternative approach to obtain structural information for membrane proteins built around an a-helical bundle. This approach that is based on determining the structures of fragments of membrane proteins is the subject of this chapter. This alternative approach was founded on the hypothesis that a-helices in a transmembrane bundle exhibit stable secondary

Membrane Protein Fragments Reveal Both Secondary and Tertiary Structure

285

structure independent of the remainder of the protein (short-range interactions, including hydrogen bonds within the helix, stabilize this element of secondary structure). The hypothesis was extended to the secondary structure of short loops, such as a b-turn connecting two transmembrane helices, which also contain internal hydrogen bonds. Since loops and a-helices are stabilized by shortrange interactions, peptide fragments of the membrane protein corresponding to an individual loop or a-helix may well form the same secondary structure found in the intact protein. This chapter begins with peptide structures that correspond to small domains of membrane proteins. These peptides exhibit well-defined structures and in some cases can be shown to possess biological activity. Biological activity of the fragment lends credence to the deduction that these peptides are forming solution structures similar to what these fragments form in the intact protein. This section illustrates how important information of secondary structure can be obtained even from relatively small fragments of membrane proteins. This chapter continues with recent studies on fragments that contain two transmembrane helices connected by a loop. These studies reveal information about tertiary structure of the proteins, as well as the protein secondary structure because they show the packing of one transmembrane helix against the other. Next to be discussed is a set of membrane proteins for which structures of more than one domain have been determined based on peptide studies. These studies show that the structure of a membrane protein can be extensively defined by determining the secondary structures of the individual helices and loops. Finally, since structures of both helices and loops are independently stabilized by short-range interactions even when separate from the intact protein, one could conceive of an approach leading to the structure of an entire transmembrane helical bundle protein using the structures of these peptide fragments. This chapter ends with a description of the methodology that allows the assembly of the structures of the fragments into a structure of the whole protein. To do so, long distance constraints from measurements on intact proteins are exploited that reflect the threedimensional packing of the elements of secondary structure of the protein. A remarkably useful structure of the whole protein, entirely determined by experimental constraints, can be obtained (2, 3). This methodology was confirmed using bacteriorhodopsin for which several X-ray crystal structures were available for verification of the result. The success of this approach has important implications for membrane protein folding and stability. These concepts form the basis for a rapidly growing series of studies on a wide variety of membrane proteins, many of them G-protein coupled receptors (GPCR), using protein fragments to determine the secondary structure of the intact protein. The earliest

286

Yeagle and Albert

of these studies have been reviewed previously and will only be summarized here (4, 5). In most cases, the structures of these fragments provide the first atomic level insight into the structures of the corresponding membrane proteins.

2. Materials Peptide fragments are designed to encompass a loop, a transmembrane helix, or a helix-turn-helix of the parent protein. Standard structure prediction techniques are used to estimate the portion of the primary sequence of the protein that defines the loop, the transmembrane helix, or the helix-turn-helix. The hydrophobicity plot is the primary tool, with the addition where appropriate of factors like the preferred location of tryptophans at the interface between the hydrophobic and hydrophilic domains of a membrane and the inclusion on the end of fragments representing transmembrane helices of any charged residues in the immediate sequence to assist in solubility. The peptides are synthesized by solid-phase peptide synthesis. If desired, stable isotopes can be included at specific positions. As will be seen in the many examples given below, loop fragments can often be examined in aqueous buffer. However, transmembrane fragments are insoluble in water and must be studied in detergent micelles or in organic solvent. The most generally useful solvent that does not appear to perturb the secondary structure is dimethylsulfoxide. All these structural studies are performed using high resolution, high field NMR, with standard two- and three-dimensional techniques.

3. Methods 3.1. Secondary Structures of Fragments of Membrane Proteins Corresponding to Single Loops or to Single Transmembrane Helices

Structures of over 100 peptide fragments of helical transmembrane proteins have now been determined. In many cases, these peptides exhibit biological activity. It is reasonable to propose that the structure of the peptide in solution that is biologically active closely resembles the structure of the corresponding domain in the intact protein. A case in point is provided by the cytoplasmic loops of bovine rhodopsin (a GPCR) that exhibited both biological activity in solution (6) and structure in solution (7). In some cases, subsequent X-ray crystal structure results demonstrated unequivocally that many peptide fragments of membrane proteins are able to independently assume the structural motifs of helices and turns characteristic of the intact protein. This is

Membrane Protein Fragments Reveal Both Secondary and Tertiary Structure

287

consistent with the stabilization of secondary structures through short-range interactions. These kinds of studies demonstrate that it is feasible to obtain relevant structural information for discrete domains of membrane proteins through experimentally accessible approaches on designed peptides. The following will discuss first transmembrane fragments that form helices, followed by fragments that form loops that in turn connect transmembrane helices. 3.1.1. Transmembrane Helical Fragments

In the following, many of the studies on single transmembrane fragments of membrane proteins are briefly summarized to provide an overview of what has been reported to date. At the end of this section, some general features of these structures are discussed. A number of studies have examined fragments of GPCRs. Peptide fragments were chosen based partly on hydropathy plots and were designed to encompass the entire sequence of the putative transmembrane helix. For example, a peptide fragment corresponding to the seventh transmembrane domain (TM) of the tachykinin NK-1 receptor was synthesized, and an a-helical solution structure was reported (8). Similar studies have been reported on fragments from the human adenosine A2a receptor (9), the a-factor receptor (Ste2p) from Saccharomyces cerevisiae (10), and bovine rhodopsin (11). Studies have reported a-helical structures from fragments encompassing transmembrane segments of transport proteins. Peptides corresponding to the transmembrane domain of the monomer of three forms of the potassium ion channel, Shaker, ROMK1, and minK, were each predominantly helical (12). Two of the putative transmembrane segments of the human erythrocyte anion transporter, band 3, were studied and predominantly a-helical structures were reported (13). A peptide containing the putative transmembrane domain of Isk, a voltage-gated potassium channel (residues 42–68), promoted ion transport and solution NMR studies showed an a-helical structure (14). A peptide corresponding to the putative TM7 of the yeast vacuolar H+ ATPase was synthesized, its solution structure determined by NMR, and a predominantly helical structure was observed (15, 16). A peptide containing the sequence of the seventh transmembrane segment of the Na+/H+ exchanger was synthesized. A helical structure, with a break at the double glycine in the helix, was observed (17). Almost all of the transmembrane segments of lactose permease have also been studied by NMR, and a-helical structures were reported (18). Several other examples of peptide fragments of the transmembrane domain of other transmembrane proteins have been reported to form a-helical structures as well. Peptides corresponding to two putative TM from phosphatidylglycerophosphate synthase (E. coli) formed stable helices in solution and in

288

Yeagle and Albert

SDS micelles (19). A peptide containing the sequence of the first TM domain of nAChR b2 subunit was studied in dodecylphosphocholine micelles, and while predominantly helical, the study revealed nonhelical sections with possible functional significance (20). A peptide corresponding to the transmembrane segment of human erythrocyte glycophorin was synthesized, and a helical structure was reported (21). A peptide containing the sequence of the transmembrane segment of FKBP-like protein twisted dwarf 1 (TWD1) from Arabidopsis thaliana was studied by NMR and found to be largely helical (22). NMR studies show a helical conformation for the transmembrane segment of PMP1, a transmembrane protein with a single transmembrane domain that regulates H+-ATPase (23). The Na+K+-ATPase regulatory protein FXYD1 was studied by NMR in micelles and a predominantly helical, though bent, structure was found (24). Subunit b of the F1F0ATP synthase shows an a-helix representing a TM in the amino terminus of this protein (25). The structures obtained from these fragments of the transmembrane domains of a wide variety of membrane proteins reveal that these fragments form stable a-helical structures. In a few cases, comparison with subsequently reported X-ray crystal structures became possible, and good structural correspondence was observed. One of the general characteristics of these studies is that these a-helical structures in solution showed greatest stability in the center of the fragment. The ends of these peptide structures are commonly frayed (see Fig.Â€ 1). This phenomenon is widely observed for structures of peptides of these sizes and suggests that these peptide fragments must be designed to be longer than the sequence for which a stable structure is desired. 3.1.2. Fragments Corresponding to Loops of Membrane Proteins

Loop regions of membrane proteins correspond to domains exterior to the membrane. Together the loops can form the site for interaction of biological ligands or drug molecules with the receptor. Individually, these loops may exhibit partial activity for binding ligands or other proteins. For example, the third extracellular loop of the cholecystokinin-2 receptor (26–31) binds a ligand for the receptor. Therefore, the structures of membrane proteins loops are of particular interest. Many studies have been reported, and are noted below, followed by some general observations at the end of this section. Peptide fragments from several GPCR’s have been studied. In some of these cases, a peptide fragment was designed to encompass a loop bounded by a short stretch of helix on either end. This fragment design has been called a helix-turn-helix motif based on the structural results (see below). Two fragments of the parathyroid hormone receptor have been studied by solution NMR techniques in detergent micelles. One peptide containing the amino acid sequence of the first

Membrane Protein Fragments Reveal Both Secondary and Tertiary Structure

289

Fig.Â€1. Family of structures derived from solution NMR data for a peptide fragment corresponding to transmembrane helix 7. See ref. 19 for details. The figure demonstrates that the stable portions of these fragments are in central portion of the peptide and the ends of the solution structure are typically frayed.

extracellular loop showed a loop structure. The peptide was sufficiently long to contain a loop bounded by two short helices, forming a helix-turn-helix motif (32). A peptide containing the sequence of the third extracellular loop of the cholecystokinin-2 receptor (residues 352–379) revealed a helix-turn-helix motif (28). A 44-residue peptide of the human cannabinoid 1 receptor containing the amino acid sequence of the third cytoplasmic loop of this GPCR is biologically active and exhibits a helix-turn-helix motif (33) mimicking the turn found in the intact protein. A 34residue peptide fragment corresponding to the second intracellular loop of the bradykinin B2 receptor revealed a helix-turn-helix motif (27, 28). A peptide fragment containing the sequence of the second cytoplasmic loop, iC2, of the V1A vasopressin receptor adopted a helix-turn-helix motif, consistent with a turn in the protein connecting two transmembrane helices (34). The structure was very similar to the structure of the corresponding loop on rhodopsin (35). Other peptide fragments corresponding to loops in GPCRs were not long enough to form the helix-helix-turn motif, but nevertheless formed a stable structure in solution; in several cases,

290

Yeagle and Albert

a turn was observed. The peptide corresponding to the third cytoplasmic loop of the parathyroid hormone receptor formed a loop structure (36, 37). Peptide fragments of the first and second extracellular loops of the thromboxane A2 receptor were synthesized. In each case, loop structures were obtained (38, 39), consistent with the connection of these fragments to two transmembrane helices. A peptide fragment from the cytoplasmic face of the parathyroid hormone/parathyroid hormone related protein receptor was structured in solution (37). Two peptides were synthesized that spanned the third cytoplasmic loop of the rat angiotensin II AT1A receptor, a GPCR (40). Some helix was observed, a portion of which corresponds to the connection to a TM. The fragment peptide corresponding to the first extracellular loop forms a type 2 b turn (41). Peptide fragments of the turkey b-adrenergic receptor corresponding to the third intracellular loop showed helical structure, likely corresponding to the beginning of TM6 (42). The same region of the human b-adrenergic receptor was also found to be helical in detergent micelles but in water the peptide was disordered (43). The second cytoplasmic loop of the a2A adrenergic receptor was studied using a peptide fragment in DPC micelles. This peptide proved to be predominantly helical (44). A peptide fragment of the amino terminus of the neurokinin-1 receptor was synthesized, and the structure was determined. Some helix was observed, likely a portion of TM1. A 27 residue peptide corresponding to the third extracellular loop of this GPCR was made, and its structure showed a largely helical structure, consistent with a helix-turn-helix motif (33). The crystal structure of the GPCR rhodopsin shows a helical segment at the amino terminus that lies parallel to the plane of the bilayer and is referred to as helix 8. A peptide fragment corresponding to the putative helix 8 of the human cannabinoid 1 receptor was found to be predominantly helical (45). The putative helix 8 of the cannabinoid 2 receptor has also been synthesized and the structure determined, and an a-helix was observed in the presence of detergent micelles (46). A predominantly a-helical conformation was observed from a peptide containing the sequence of the corresponding region of the turkey b-adrenergic receptor (47). The collection of work on the helix 8 region of GPCRs suggested that this region is helical in the presence of a membrane mimetic but does not form a stable helix in the absence of the membrane mimetic (48). This could have important consequences in the activation of the receptor. In most cases, peptide fragments corresponding to loops (connecting two transmembrane helices) of membrane proteins formed loop structures in solution, with a helix-turn-helix motif when the peptide fragment was long enough. The short helices at the end of the fragment correspond to the ends of the transmembrane

Membrane Protein Fragments Reveal Both Secondary and Tertiary Structure

291

helices. These studies suggest that loops connecting TM in transmembrane bundles may contribute to the stability of membrane proteins. For the most part, these solution structures are independent of the solution conditions. As in the case of the transmembrane fragments discussed above, the ends of the NMR-derived solution structures of these loops are frayed. 3.1.3. General Features of Membrane Protein Structure Accessible from Studies of Membrane Protein Fragments

The available structural literature on membrane protein fragments: 1. Reports structures that report with fidelity on the secondary structure of the intact protein 2. Shows structures of a transmembrane helix or a loop connecting two transmembrane helices 3. Is dominated by solution NMR studies, often in solvents like DMSO that dissolve hydrophobic peptides without perturbing their secondary structure 4. Hydropathy plots have long been used to predict the helical and loop regions of membrane proteins. While these plots can underestimate the length of the transmembrane helix, they have proved to be an important guide in choosing the peptide sequences to investigate. Another factor that is helpful in determining the appropriate length is the pattern of tryptophans locating in the interfacial region of the membrane, and thus occurring near the end of a transmembrane helix (49) 5. Structures obtained can be independent of solvent; some sequences have been examined in both DMSO and water and found to have the same structure (48, 50) These results suggest that it is feasible to explore the secondary structure of membrane proteins using this fragment approach, discovering, among other features, the stop and start points of helices, break points in helices, and the structures of loops connecting transmembrane helices. Under the most favorable circumstances, studies of the structure of membrane fragments can provide a nearly complete survey of the secondary structure of a membrane protein providing that the protein is built around a transmembrane helical bundle. (In the case of bacterial porins that are formed of b-barrels, studies of membrane fragments would not be expected to produce useful information since b-barrels require long-range interactions for stability.) Importantly, the available literature does not contain an example of a membrane protein fragment where the structure reflected incorrectly the secondary structure of the intact membrane protein. Of course some cases exist in which a fragment is

292

Yeagle and Albert

disordered in solution and thus provides no information on secondary structure. Consequently, studies on the structures of over 100 of these fragments of membrane proteins have not yet misled the investigator about the structure in the protein from which the fragments were derived. 3.2. Two-Helix Fragments

The work on fragments encompassing a single TM of a membrane protein does not provide information on the packing of one helix against another in a transmembrane helical bundle. A step in the direction of obtaining information on helix packing can be taken by studying a two-helix bundle connected by a loop. This helix-loop-helix structure can be expected to express some of the helix–helix packing constraints inherent in the intact transmembrane helical bundle. Several investigators have now explored the structure of such two-helix bundles. The oldest example of a structure of a two-helix bundle is not strictly speaking a member of this family of structures, but is very informative, nonetheless. The protein is human erythrocyte glycophorin and the two-helix bundle is not connected by a peptide loop because glycophorin has a single transmembrane domain and forms a stable dimer in the membrane. In this pioneering NMR structure, both the crossing angle of the two helices (−40°) and the interhelical packing can be determined, as well as the structure of the helix itself. Glycines are found to play an important role in the packing of the helices (21). Several such studies have focused on GPCRs. Extensive studies have been published on the structure of fragments of the single GPCR found in yeast, Ste2p (51–54). Several fragments containing two TM have been studied. Recent studies on the TM1-loop-TM2 fragment and the TM6-loop-TM7 fragment showed strong evidence for a helix-loop-helix structure, though a high resolution structure is yet to be reported (55). An NMR study of a fragment of the human cannabinoid CB2 receptor encompassing the first and second transmembrane domains of this protein reported an extensive assignment of the expressed fragment, but an atomic structure was not determined (56). More detailed structural information was obtained from other membrane proteins. The second and third transmembrane domains of the human glycine receptor were expressed as a TM-turn-TM fragment. The NMR structure of this fragment confirmed the anticipated helix-turn-helix motif and revealed a crossing angle of 44° between the two helices (57). Subunit c of the F1F0ATP synthase forms a helix-turn-helix as a monomer in organic solution. This NMR structure is consistent with known structural information from the intact transmembrane region. In this case, the helix-turn-helix motif represents the entire subunit (58). NMR studies have been reported recently on the transmembrane domain of the ErbB2 receptor that consists of two TM.

Membrane Protein Fragments Reveal Both Secondary and Tertiary Structure

293

A helix-turn-helix motif was found, and the crossing angle of the two helices is −42° with glycines playing a role in helix–helix packing (59). A recent NMR study of the BNip3 transmembrane domain in lipid bicelles showed a helix-turn-helix motif and a helix-crossing angle of −45° (60). The transmembrane domain of the T cell receptor was studied as a zeta–zeta dimer in detergent micelles by NMR. A crossing angle of 23° was observed between the two helices (61). In summary, studies show that TM-loop-TM peptides from the transmembrane domains of membrane proteins constructed of helical bundles provide more information than can be obtained from studies of single TM. These polypeptides form the expected helix-turn-helix motif in micelles or bicelles. Furthermore, information about helix–helix packing that likely reflects the structure of the intact protein, including the helix crossing angles, is revealed. 3.3. Full Scan of Secondary Structure Using Membrane Protein Fragments

The studies described above encouraged a more systematic analysis of an entire membrane protein. The well-studied membrane protein, bacteriorhodopsin, was chosen. An overlapping set of peptide fragments that spanned the entire primary sequence of the protein was synthesized, and each fragment of bacteriorhodopsin was studied individually by solution NMR. The structures of peptides corresponding to transmembrane helical segments were a-helical as protein fragments. Peptides corresponding to loops formed loops as protein fragments. Comparison of these peptide structures to the crystal structure of bacteriorhodopsin revealed that the structures adopted by these fragments of bacteriorhodopsin reported faithfully on the secondary structure of the intact protein (3). These results suggest the hypothesis that scanning the entire primary structure of an integral membrane protein with overlapping fragments and determining the secondary structure of the fragments will provide information on the secondary structure of the whole protein. Several examples exist in which this approach has proven useful. EmrE is a multidrug resistance protein in E. coli. Models for the structure of this protein have included a helical bundle and a b-barrel. Four overlapping peptides were synthesized and their structures determined in SDS micelles, and the authors determined that the helical bundle was the better model for the transmembrane domain of this membrane protein (62). Human adenosine A2a receptor is a GPCR, and thus expected to contain a transmembrane bundle of seven helices. Seven peptides were synthesized, each one designed to encompass one putative transmembrane segment of the protein. Circular dichroism studies revealed that five of these peptides were extensively a-helical, consistent with the hypothesis of the structural model

294

Yeagle and Albert

for this protein. Two of the peptides showed considerably less helical content, suggesting a reduced stability of secondary structure in those transmembrane segments (9). The translocator protein TSPO binds cholesterol and has been suggested to contain five TM. Five peptides encompassing these five TM were synthesized, and their structures were determined by NMR. In each case, the putative TM exhibited largely helical conformation, thus supporting the model of a five TM bundle of helices for the structure of this protein (63). A prior report from the same lab revealed the interesting observation that a fragment containing the C-terminal region of the protein formed a helical structure capable of binding cholesterol (64). The a-factor receptor (Ste2p) from S. cerevisiae has been extensively studied using this fragment approach. Fragments corresponding to each of the transmembrane segments have been synthesized, and structures were determined (10). More recently, a much larger fragment containing the putative third cytoplasmic loop, TM7, and a substantial portion of the C terminus has been studied. Both the helical nature of TM7 was observed as was the putative eighth helix of this GPCR (65). Much has been learned about the structure of this GPCR from these studies, including a complete survey of the secondary structure of this membrane protein using fragments of the receptor (51–55). The data are consistent with a transmembrane helical bundle typical of other GPCRs. A set of 18 peptides spanning the entire sequence of rhodopsin was synthesized and the structure of each of the peptides was individually determined by NMR. Although these peptides were studied before any X-ray crystal structure was available, comparison of these structures to the crystal structures when they became available showed fidelity in reporting of the secondary structure of the protein by the structures of the fragments (11, 49, 50, 66–71). The secondary structure of the 12 TM protein, lactose permease, has been studied using a set of 12 peptides, each encompassing one of the TMs of this protein. As in the case of other membrane proteins described above, most of these 12 fragments of lactose permease showed substantial helical content. Interestingly some did not. Molecular dynamics calculations produced results consistent with the experimental NMR results; that is, some TMs were more stable than others. These data suggested that while the protein contained a transmembrane bundle of 12 helices, some of these helices are more stable than others within the tertiary structure of the protein. When mapped on the recent crystal structure of the protein, a remarkable asymmetry was observed within the protein with respect to stability that may have functional consequences (18, 72).

Membrane Protein Fragments Reveal Both Secondary and Tertiary Structure

295

These studies have demonstrated that considerable information about membrane protein secondary structure (for membrane proteins with an a-helical transmembrane domain) can be obtained from structural analysis of peptide fragments of those membrane proteins. 3.4. ThreeDimensional Structures from Protein Fragments

It is possible to create a set of overlapping peptides that span the sequence of a membrane protein (built around a helical transmembrane domain) in which each peptide encompasses an individual transmembrane segment or a loop connecting two transmembrane segments, as introduced above. If each of the fragments expressed stable secondary structure, the opportunity would present to assemble the three-dimensional structure of the protein from the fragments if information existed to organize the elements of secondary structure in three dimensions. In 2001, bacteriorhodopsin was used to test this idea. An assumption was made that the secondary structure determined from studies of the fragments was the same as in the intact bacteriorhodopsin. This assumption was verified by comparing the structures of the fragments determined in solution by NMR to the available X-ray crystal structures of the intact bacteriorhodopsin. The methodology relied on experimental data from the intact protein that defined the arrangement of the transmembrane helices in three dimensions. These data consisted of point-to-point distances determined experimentally by a variety of techniques. A preliminary construct of the protein was then made by assembling the (overlapping) structures of the fragments into a continuous polypeptide spanning the whole sequence of bacteriorhodopsin. This construct expressed the secondary structure but not the tertiary structure of the protein. All of the available experimental distance constraints were then applied to the construct, including the NMR constraints from the studies of the fragments and the longdistance constraints from other experiments that defined the organization of the protein in three dimensions. This construct was then subjected to simulated annealing to allow an evolution of the structure to achieve a match between the structure and the experimental distance constraints. Finally, the whole structure was minimized with respect to energy under the experimental distance constraints. The result was a structure that agreed remarkably well with X-ray crystal structures of the protein (see Fig.Â€2), and could be characterized as a medium-resolution structure of bacteriorhodopsin. It should be emphasized that this successful result is entirely an experimental structure, driven by individual distance constraints from a variety of experiments (3). A more theoretical foundation to this approach was provided recently in which the process was defined by two steps: (1) “search the conformational space of membrane folds to find those matching a given set of distance constraints and (2) refine the helical bundles

296

Yeagle and Albert

Fig.Â€2. Three-dimensional structure of bacteriorhodopsin determined as described in the text (red) superimposed on one of the crystal structures (yellow) of bacteriorhodopsin (2BRD). (Figure prepared with VMD (79))

from step 1 using a Monte Carlo simulated annealing protocol designed for local minimization of a custom penalty function.” A key finding in this study is that the number of long distance constraints required to adequately organize in three dimensions a transmembrane helical bundle is surprisingly small (73). This result helped to explain the success of the work with bacteriorhodopsin described above, which also utilized a limited number of distance constraints defining the tertiary structure of the transmembrane bundle. This sparse constraints approach was further exploited in assembling a model of the transmembrane domain of lactose permease in which 99 interhelical distance constraints were utilized to define the helix packing of the transmembrane helical bundle of this protein (74). The process outlined above was utilized for the development of an experimentally based structure for the GPCR, rhodopsin. The entire sequence of rhodopsin was scanned with a set of overlapping peptides that encompassed either a putative TM or a loop connecting two TM. The secondary structure of each of these peptides was determined by NMR. (Later, when a crystal structure

Membrane Protein Fragments Reveal Both Secondary and Tertiary Structure

297

became available, the assumption that the secondary structure in the peptides matched the structure in the intact protein was confirmed.) Because the peptides overlapped in sequence, it was possible to build a construct for the entire protein by linking together the fragment structures using the regions of overlap. A substantial number of long-range distance constraints from a variety of independent experiments on the intact protein were then written onto the construct, along with all the short-range constraints from the secondary structure determination in the fragments. This construct was then subjected to simulated annealing and finally to energy minimization. The resulting structure was eventually compared with the crystal structure of the protein and remarkable agreement was noted, best characterized as a medium resolution structure. This structure was determined entirely from experimental distance constraints (75). This approach was extended to the structure of the activated form of rhodopsin, called Meta II. This is the form that binds the G protein to activate the latter and is a transient species. Using a different set of long-range distance constraints obtained from experiments on the trapped activated form of the receptor, a new structure was determined. This structure was determined before any other atomic level structure for the activated form was published and provided a view into the changes that activation causes on the receptor structure (76). These studies reveal that experimentally based membrane protein structures are accessible by an alternative approach. The interactability of many membrane proteins to structure determination is the driving force for the studies described here. While high-resolution structures determined by X-ray crystallography and NMR remain the goal of membrane protein structural biology, alternative approaches have provided considerable information in the absence of X-ray crystal structures and NMR structures of the whole protein. These studies also reveal that the stability of elements of secondary structure are even more important to transmembrane proteins than they are already known to be for the stability of soluble proteins. Folding of newly synthesized transmembrane proteins likely proceeds through the formation of secondary structure, in particular, the transmembrane helical bundle, as a stable core of the protein as soon as the polypeptide chain encounters the hydrophobic interior of the membrane. 3.5. NMR Studies on Intact Membrane Proteins

It is appropriate to close this chapter with a brief consideration of the determination of the complete structure of a membrane protein with a helical transmembrane domain by NMR. For one membrane protein, diacylglycerol kinase, a complete assignment of the NMR data has been achieved (77). For several others, a partial assignment of NMR data from the intact protein has been

298

Yeagle and Albert

reported, including translocator protein (63) and signal peptidase (78). Clearly, determination of intact membrane protein structure for membrane proteins with a transmembrane helical bundle from NMR data alone is yet to come into its own, which emphasizes the importance of the methods reviewed in this chapter.

4. Notes Experimental difficulties arise in using fragments to study membrane protein secondary structure. The greatest difficulty can arise if nonperturbing conditions cannot be found in which the fragment exhibits stable secondary structure. Using strong helixstabilizing solvents such as trifluoroethanol to get around this problem is not very useful because it may be stabilizing a helix where none exists naturally. For example, in the study of the lactose permease by fragments, it was discovered that some fragments had little or no stability as separate fragments and that likely reflected on the internal stability of these transmembrane segments in the native protein (18). The structures themselves are often difficult to obtain because of chemical shift redundancies made worse by the lack of tertiary structure. References 1. http://blanco.biomol.uci.edu/Membrane_ Proteins_xtal.html 2. Katragadda M, Alderfer JL, Yeagle PL (2000) Solution structure of the loops of bacteriorhodopsin closely resemble the crystal structure. Biochim Biophys Acta 1466:1–6 3. Katragadda M, Alderfer JL, Yeagle PL (2001) Assembly of a polytopic membrane protein structure from the solution structures of overlapping peptide fragments of bacteriorhodopsin. Biophys J 81:1029–1036 4. Yeagle PL, Albert AD (2002) Use of nuclear magnetic resonance to study the three-Â� dimensional structure of rhodopsin. Methods Enzymol 343:223–231 5. Yeagle PL, Albert AD (2007) G-protein coupled receptor structure. Biochim Biophys Acta 1768:808–824 6. Konig B, Arendt A, McDowell JH, Kahlert M, Hargrave PA, Hofmann KP (1989) Three cytoplasmic loops of rhodopsin interact with transducin. Proc Natl Acad Sci U S A 86: 6878–6882 7. Yeagle PL, Alderfer JL, Albert AD (1995) Structure of the carboxyl terminal domain

8.

9.

10.

11.

12.

of bovine rhodopsin. Nat Struct Biol 2: 832–834 Berlose J, Convert O, Brunissen A, Chassaing G, Lavielle S (1994) Three dimensional structure of the highly conserved seventh transmembrane domain of G-protein-coupled receptors. FEBS Lett 225:827–843 Lazarova T, Brewin KA, Stoeber K, Robinson CR (2004) Characterization of peptides corresponding to the seven transmembrane domains of human adenosine A2a receptor. Biochemistry 43:12945–12954 Arshava B, Taran I, Xie H, Becker JM, Naider F (2002) High resolution NMR analysis of the seven transmembrane domains of a heptahelical receptor in organic–aqueous medium. Biopolymers 64:161–176 Katragadda M, Chopra A, Bennett M, Alderfer JL, Yeagle PL, Albert AD (2001) Structures of the transmembrane helices of the G-protein coupled receptor, rhodopsin. J Pept Res 58:79–89 Haris PI (1988) Synthetic peptide fragments as probes for structure determination of potassium ion-channel proteins. Biosci Rep 18:299–312

Membrane Protein Fragments Reveal Both Secondary and Tertiary Structure 13. Gargaro AR, Bloomberg GB, Dempsey CE, Murray M, Tanner MJ (1994) The solution structures of the first and second transmembrane-spanning segments of band 3. Eur J Biochem 221:445–454 14. Aggeli A, Bannister ML, Bell M etÂ€al (1998) Conformation and ion-channeling activity of a 27-residue peptide modeled on the singletransmembrane segment of the IsK (minK) protein. Biochemistry 37:8121–8131 15. Duarte AM, de Jong ER, Wechselberger R, van Mierlo CP, Hemminga MA (2007) Segment TM7 from the cytoplasmic hemichannel from VO-H+-V-ATPase includes a flexible region that has a potential role in proton translocation. Biochim Biophys Acta 1768:2263–2270 16. Duarte AM, Wolfs CJ, van Nuland NA etÂ€al (2007) Structure and localization of an essential transmembrane segment of the proton translocation channel of yeast H+-V-ATPase. Biochim Biophys Acta 1768:218–227 17. Ding J, Rainey JK, Xu C, Sykes BD, Fliegel L (2006) Structural and functional characterization of transmembrane segment VII of the Na+/H+ exchanger isoform 1. J Biol Chem 281:29817–29829 18. Bennett M, D’Rozario R, Sansom M, Yeagle PL (2006) Asymmetric stability among the transmembrane helices of lactose permease. Biochemistry 45:8088–8095 19. Morein S, Trouard TP, Hauksson JB, Rilfors L, Arvidson G, Lindblom G (1996) Twodimensional 1H-NMR of transmembrane peptides from Escherichia coli phosphatidylglycerophosphate synthase in micelles. Eur J Biochem 241:489–497 20. Bondarenko V, Xu Y, Tang P (2007) Structure of the first transmembrane domain of the neuronal acetylcholine receptor beta2 subunit. Biophys J 92:1616–1622 21. MacKenzie KR, Prestegard JH, Engelman DM (1997) A transmembrane helix dimer: structure and implications. Science 276:131–133 22. Scheidt HA, Vogel A, Eckhoff A, Koenig BW, Huster D (2007) Solid-state NMR characterization of the putative membrane anchor of TWD1 from Arabidopsis thaliana. Eur Biophys J 36:393–404 23. Mousson F, Beswick V, Coic YM, HuynhDinh T, Sanson A, Neumann JM (2001) Investigating the conformational coupling between the transmembrane and cytoplasmic domains of a single-spanning membrane protein. A 1H-NMR study. FEBS Lett 505: 431–435

299

24. Teriete P, Franzin CM, Choi J, Marassi FM (2007) Structure of the Na, K-ATPase regulatory protein FXYD1 in micelles. Biochemistry 46:6774–6783 25. Dmitriev O, Jones PC, Jiang W, Fillingame RH (1999) Structure of the membrane domain of subunit b of the Escherichia coli F0F1 ATP synthase. J Biol Chem 274: 15598–15604 26. Pellegrini M, Mierke DF (1999) Molecular complex of cholecystokinin-8 and N-terminus of the cholecystokinin A receptor by NMR spectroscopy. Biochemistry 38: 14775–14783 27. Giragossian C, Mierke DF (2003) DetermiÂ� nation of ligand–receptor interactions of cholecystokinin by nuclear magnetic resonance. Life Sci 73:705–713 28. Giragossian C, Schaschke N, Moroder L, Mierke DF (2004) Conformational and molecular modeling studies of beta-cyclodextrin– heptagastrin and the third extracellular loop of the cholecystokinin 2 receptor. Biochemistry 43:2724–2731 29. Giragossian C, Sugg EE, Szewczyk JR, Mierke DF (2003) Intermolecular interactions between peptidic and nonpeptidic agonists and the third extracellular loop of the cholecystokinin 1 receptor. J Med Chem 46: 3476–3482 30. Giragossian C, Mierke DF (2002) IntermolecÂ� ular interactions between cholecystokinin-8 and the third extracellular loop of the cholecystokinin-2 receptor. Biochemistry 41: 4560–4566 31. Giragossian C, Mierke DF (2001) IntermolecÂ� ular interactions between cholecystokinin-8 and the third extracellular loop of the cholecystokinin A receptor. Biochemistry 40: 3804–3809 32. Piserchio A, Bisello A, Rosenblatt M, Chorev M, Mierke DF (2000) Characterization of parathyroid hormone/receptor interactions: structure of the first extracellular loop. Biochemistry 39:8153–8160 33. Ulfers AL, Piserchio A, Mierke DF (2002) Extracellular domains of the neurokinin-1 receptor: structural characterization and interactions with substance P. Biopolymers 66:339–349 34. Demene H, Granier S, Muller D etÂ€al (2003) Active peptidic mimics of the second intracellular loop of the V(1A) vasopressin receptor are structurally related to the second intracellular rhodopsin loop: a combined 1H NMR and biochemical study. Biochemistry 42:8204–8213

300

Yeagle and Albert

35. Yeagle PL, Alderfer JL, Albert AD (1997) The first and second cytoplasmic loops of the G-protein receptor, rhodopsin, independently form b-turns. Biochemistry 36:3864–3869 36. Mierke DF, Royo M, Pelligrini M, Sun H, Chorev M (1996) Third cytoplasmic loop of the PTH/PTHrP receptor. J Am Chem Soc 118:8998–9004 37. Pellegrini M, Royo M, Chorev M, Mierke DF (1996) Conformational characterization of a peptide mimetic of the third cytoplasmic loop of the G-protein coupled parathyroid hormone/parathyroid hormone related protein receptor. Biopolymers 40:653–666 38. Ruan KH, So SP, Wu J, Li D, Huang A, Kung J (2001) Solution structure of the second extracellular loop of human thromboxane A2 receptor. Biochemistry 40:275–280 39. Wu J, So SP, Ruan KH (2003) Solution structure of the third extracellular loop of human thromboxane A2 receptor. Arch Biochem Biophys 414:287–293 40. Franzoni L, Nicastro G, Pertinhez TA etÂ€ al (1999) Structure of two fragments of the third cytoplasmic loop of the rat angiotensin II AT1A receptor. Implications with respect to receptor activation and G-protein selection and coupling. J Biol Chem 274:227–235 41. Nicastro G, Peri F, Franzoni L, de Chiara C, Sartor G, Spisni A (2003) Conformational features of a synthetic model of the first extracellular loop of the angiotensin II AT1A receptor. J Pept Sci 9:229–243 42. Jung H, Windhaber R, Palm D, Schnackerz KD (1995) NMR and circular dichroism studies of synthetic peptides derived from the third intracellular loop of the beta-adrenoceptor. FEBS Lett 358:133–136 43. Katragadda M, Maciejewski MW, Yeagle PL (2004) Structural studies of the putative helix 8 in the human beta(2) adrenergic receptor: an NMR study. Biochim Biophys Acta 1663:74–81 44. Chung DA, Zuiderweg ER, Fowler CB, Soyer OS, Mosberg HI, Neubig RR (2002) NMR structure of the second intracellular loop of the alpha 2A adrenergic receptor: evidence for a novel cytoplasmic helix. Biochemistry 41:3596–3604 45. Choi G, Guo J, Makriyannis A (2005) The conformation of the cytoplasmic helix 8 of the CB1 cannabinoid receptor using NMR and circular dichroism. Biochim Biophys Acta 1668:1–9 46. Choi G, Landin J, Xie XQ (2002) The cytoplasmic helix of cannabinoid receptor CB2, a conformational study by circular dichroism

47.

48.

49. 50.

51.

52.

53.

54.

55.

56.

57.

and (1)H NMR spectroscopy in aqueous and membrane-like environments. J Pept Res 60:169–177 Jung H, Windhaber R, Palm D, Schnackerz KD (1996) Conformation of a beta-adrenoceptor-derived signal transducing peptide as inferred by circular dichroism and 1H NMR spectroscopy. Biochemistry 35:6399–6405 Yeagle PL, Alderfer JL, Albert AD (1996) Structure determination of the fourth cytoplasmic loop and carboxyl terminal domain of bovine rhodopsin. Mol Vis 2. http://www. molvis.org/molvis/v2/p12/ Wallace BA, Janes RW (1999) Tryptophans in membrane proteins. Adv Exp Med Biology 467:789–799 Yeagle PL, Danis C, Choi G, Alderfer JL, Albert AD (2000) Three dimensional structure of the seventh transmembrane helical domain of the G-protein receptor, rhodopsin. Mol Vis. http://www.molvis.org/molvis/v6/ a17/ Arshava B, Liu SF, Jiang H, Breslav M, Becker JM, Naider F (1998) Structure of segments of a G protein-coupled receptor: CD and NMR analysis of the Saccharomyces cerevisiae tridecapeptide pheromone receptor. Biopolymers 46:343–357 Xie HB, Ding FX, Schreiber D etÂ€ al (2000) Synthesis and biophysical analysis of transmembrane domains of a Saccharomyces cerevisiae G protein-coupled receptor. Biochemistry 39:15462–15474 Estephan R, Englander J, Arshava B, Samples KL, Becker JM, Naider F (2005) Biosynthesis and NMR analysis of a 73-residue domain of a Saccharomyces cerevisiae G protein-coupled receptor. Biochemistry 44:11795–11810 Naider F, Khare S, Arshava B, Severino B, Russo J, Becker JM (2005) Synthetic peptides as probes for conformational preferences of domains of membrane receptors. Biopolymers 80:199–213 Cohen LS, Arshava B, Estephan R etÂ€al (2008) Expression and biophysical analysis of two double-transmembrane domain-containing fragments from a yeast G protein-coupled receptor. Biopolymers 90:117–130 Zheng H, Zhao J, Sheng W, Xie XQ (2006) A transmembrane helix-bundle from G-protein coupled receptor CB2: biosynthesis, purification, and NMR characterization. Biopolymers 83:46–61 Ma D, Liu Z, Li L, Tang P, Xu Y (2005) Structure and dynamics of the second and third transmembrane domains of human glycine receptor. Biochemistry 44:8790–8800

Membrane Protein Fragments Reveal Both Secondary and Tertiary Structure 58. Girvin ME, Rastogi VK, Abildgaard F, Markley JL, Fillingame RH (1998) Solution structure of the transmembrane H+transporting subunit c of the F1F0 ATP synthase. Biochemistry 37:8817–8824 59. Bocharov EV, Mineev KS, Volynsky PE etÂ€al (2008) Spatial structure of the dimeric transmembrane domain of the growth factor receptor ErbB2 presumably corresponding to the receptor active state. J Biol Chem 283: 6950–6956 60. Bocharov EV, Pustovalova YE, Pavlov KV etÂ€ al (2007) Unique dimeric structure of BNip3 transmembrane domain suggests membrane permeabilization as a cell death trigger. J Biol Chem 282:16256–16266 61. Call ME, Schnell JR, Xu C, Lutz RA, Chou JJ, Wucherpfennig KW (2006) The structure of the zetazeta transmembrane dimer reveals features essential for its assembly with the T cell receptor. Cell 127:355–368 62. Venkatraman J, Nagana Gowda GA, Balaram P (2002) Structural analysis of synthetic peptide fragments from EmrE, a multidrug resistance protein, in a membrane-mimetic environment. Biochemistry 41:6631–6639 63. Murail S, Robert JC, Coic YM etÂ€ al (2008) Secondary and tertiary structures of the transmembrane domains of the translocator protein TSPO determined by NMR. Stabilization of the TSPO tertiary fold upon ligand binding. Biochim Biophys Acta 1778:1375–1381 64. Jamin N, Neumann J-M, Ostuni MA, Thi Kim Ngoc Vu, Yao ZX, Murail S, Robert J-C, Giatzakis C, Papadopoulos V, Lacapere J-J (2005) Characterization of the cholesterol recognition amino acid consensus sequence of the peripheral-type benzodiazepine receptor. Mol Endocrinol 19(3):588–594 65. Neumoin A, Arshava B, Becker J, Zerbe O, Naider F (2007) NMR studies in dodecylphosphocholine of a fragment containing the seventh transmembrane helix of a G-proteincoupled receptor from Saccharomyces cerevisiae. Biophys J 93:467–482 66. Yeagle PL, Alderfer JL, Albert AD (1995) Structure of the third cytoplasmic loop of bovine rhodopsin. Biochemistry 34: 14621–14625 67. Dorey M, Hargrave PA, McDowell JH etÂ€ al (1999) Effects of phosphorylation on the

68.

69. 70.

71.

72.

73.

74.

75.

76.

77.

78.

79.

301

structure of the G-protein receptor, rhodopsin. Biochim Biophys Acta 1416:217–224 Yeagle PL, Alderfer JL, Albert AD (1997) Three dimensional structure of the cytoplasmic face of the G protein receptor rhodopsin. Biochemistry 36:9649–9654 Albert AD, Yeagle PL (2000) NMR analysis of the three dimensional structure of rhodopsin domains. Methods Enzymol 315:107–115 Chopra A, Yeagle PL, Alderfer JA, Albert A (2000) Solution structure of the sixth transmembrane helix of the G-protein coupled receptor, rhodopsin. Biochim Biophys Acta 1463:1–5 Yeagle PL, Salloum A, Chopra A etÂ€al (2000) Structures of the intradiskal loops and amino terminus of the G-protein receptor, rhodopsin. J Pept Res 55:455–465 Bennett M, Yeagle JA, Maciejewski M, Ocampo J, Yeagle PL (2004) Stability of loops in the structure of lactose permease. Biochemistry 43:12829–12837 Sale K, Faulon JL, Gray GA, Schoeniger JS, Young MM (2004) Optimal bundling of transmembrane helices using sparse distance constraints. Protein Sci 13:2613–2627 Sorgen PL, Hu Y, Guan L, Kaback HR, Girvin ME (2002) An approach to membrane protein structure without crystals. Proc Natl Acad Sci U S A 99:14037–14040 Yeagle PL, Choi G, Albert AD (2001) Studies on the structure of the G-protein coupled receptor rhodopsin including the putative G-protein binding site in unactivated and activated forms. Biochemistry 40:11932–11937 Choi G, Landin J, Galan JF, Birge RR, Albert AD, Yeagle PL (2002) Structural studies of metarhodopsin II, the activated form of the G-protein coupled receptor, rhodopsin. Biochemistry 41:7318–7324 Oxenoid K, Kim HJ, Jacob J, Sonnichsen FD, Sanders CR (2004) NMR assignments for a helical 40Â€kDa membrane protein. J Am Chem Soc 126:5048–5049 Musial-Siwek M, Kendall DA, Yeagle PL (2008) Solution NMR of signal peptidase, a membrane protein. Biochim Biophys Acta 1778:937–944 Humphrey W, Dalke A, Schulten K (1996) VMD – visual molecular dynamics. J Mol GraphÂ€14:33–38

as

Chapter 16 What Can We Learn from a Small Regulatory Membrane Protein? Gianluigi Veglia, Kim N. Ha, Lei Shi, Raffaello Verardi, and Nathaniel J. Traaseth Abstract This chapter reviews the molecular biology, biochemical, and NMR methods that we used to study the structural dynamics, membrane topology, and interaction of phospholamban (PLN), a small regulatory membrane protein involved in the regulation of the sarcoplasmic reticulum Ca-ATPase (SERCA). In particular, we show the progression of our research from the initial hypotheses toward understanding the molecular mechanisms of SERCA’s regulation, including the effects of PLN oligomerization and posttranslational phosphorylation. Finally, we show how the knowledge of the molecular mechanism of the structural dynamics and topology of free and bound proteins can lead to the rational design of PLN analogs for possible use in gene therapy. Key words: Phospholamban, SERCA, Ca-ATPase, Structural dynamics, Topology, Membrane proteins, NMR, PISEMA

1. Introduction Phospholamban (PLN) is a small integral membrane protein (52 residues) localized in the sarcoplasmic reticulum (SR) membrane that regulates the flow of calcium ions into the SR lumen of cardiac muscle (1). Specifically, PLN binds to SR calcium ATPase (SERCA), the enzyme responsible for calcium re-uptake into the SR lumen. In its unphosphorylated form, PLN inhibits SERCA by reducing its affinity for calcium. Upon b-adrenergic stimulation, PLN is phosphorylated by protein kinase A at S16 with concomitant relief of SERCA inhibition and restoration of SERCA activity. These cyclic events account for the relaxation (diastolic phase) of the heart muscle, which if disrupted, may evolve into heart Jean-Jacques Lacapère (ed.), Membrane Protein Structure Determination: Methods and Protocols, Methods in Molecular Biology, vol. 654, DOI 10.1007/978-1-60761-762-4_16, © Springer Science+Business Media, LLC 2010

303

304

Veglia et al.

failure (2). Due to their central role in cardiac muscle contractility, PLN and SERCA have become targets for new, alternative therapies for heart failure (3). What can be learned from the study of the structural and dynamic transitions of PLN free and bound to SERCA?, More importantly, can these studies contribute in the development of new clinical approaches to heart failure? Much progress has been made toward understanding the molecular mechanism of SERCA within the enzymatic cycle (4, 5). Chapters 9 and 15 of this volume give exquisite portraits of the latest findings, outlining the successes of x-ray crystallography in reconstructing the structural transitions of SERCA during enzyme turnover. Undeniably, the first crystal structure of SERCA (E1-Ca2, calcium-bound form) solved by Toyoshima and co-workers paved the way for the atomic level understanding of calcium translocation in the SR (6). To date, there are several crystal structures of SERCA under different experimental conditions that are allowing enzymologists to model the major conformational states of the enzyme during turnover (7). However, very few studies have been dedicated to the determination of the complexes between SERCA and its endogenous inhibitor PLN. This is probably due to the inherent dynamics of the SERCA/PLN complex that might complicate the formation of large, diffracting crystals. This hypothesis is reinforced by several fluorescence and EPR studies published by the Thomas and Squier Laboratories underscoring the dynamic nature of the SERCA/PLN complex (8–14). So far, the labs of Stokes and Young (15, 16) have produced the best images of this complex although the quality of the crystals does not define the interaction surface between the two proteins at the atomic level. NMR studies, on the other hand, have been focused on the structural dynamics of PLN in the unbound form. This is mainly due to the difficulties in producing isotopically labeled SERCA and analyzing its complex with PLN by solution-state NMR (Mwâ•›~â•›116Â€kDa). Our laboratory, in collaboration with the Thomas laboratory, has used a combination of solution and solid-state NMR methods to map the structural dynamics, topology, and interactions between PLN and SERCA. Based on our chemical shift perturbation data from solution NMR, we have constructed an allosteric model for SERCA regulation by PLN (11, 13, 17–19). This regulatory model has been now adopted by several groups in the fields of structural biology and muscle physiology to explain the effects of naturally occurring and engineered mutations in PLN. More importantly, this model has allowed us to rationally manipulate the structural dynamics of PLN and design new mutants that are possible candidates for gene therapy in heart failure treatment. In the following sections, we briefly review the molecular biology, biochemistry, and NMR approaches that we have used to study SERCA’s regulation by looking at PLN free (monomeric

What Can We Learn from a Small Regulatory Membrane Protein?

305

and pentameric) and bound to SERCA. We show the progression of our research from our initial hypotheses on the mechanism of SERCA regulation by PLN to our initial attempts to control PLN structural dynamics to tune SERCA’s function.

2. Materials 2.1. Expression and Purification of Monomeric and Pentameric PLN

1. pMal c2E maltose-binding protein vector (New England Biolabs, Ipswich, MA).

2.1.1. Plasmid Construction and Mutagenesis of AFA–PLN

3. Pfu Turbo DNA polymerase (Stratagene, La Jolla, CA).

2. Oligonucleotide primers (Biomedical Genomics Center, University of MN). 4. dNTP mix (Promega, Madison, WI). 5. T4 DNA ligase (Promega, Madison, WI). 6. 1× Luria–Bertani (LB) media. 7. Qiaprep Spin Miniprep Kit (Qiagen, Valencia, CA).

2.1.2. Expression of Unlabeled and 15N Uniformly Labeled PLN

1. BL21(DE3) strain E. coli cells (Novagen, Gibbstown, NJ). 2. 1× Luria–Bertani broth media. 3. M9 minimal media containing 12.8Â€ g/L Na2HPO4·7H2O, 3Â€ g/L KH2PO4, 0.5Â€ g/L NaCl, and 1Â€ g/L 15N-labeled NH4Cl, 2Â€g/L glucose, and 0.1% w/v ampicillin. 4. 1Â€M IPTG (Gold Biotechnology, St. Louis, MO). 5. Mineral cocktail: 6Â€ mg/L CaCl2, 6Â€ mg/L FeSO4, 1Â€ mg/L MnCl2, 0.8Â€ mg/L CoCl2, 0.7Â€ mg/L ZnSO4, 0.3Â€ mg/L CuCl2, 0.02Â€mg/L H3BO3, 0.25Â€mg/L (NH4)6MO7O24, and 5Â€mg/L EDTA. 6. Vitamin cocktail: 1Â€ mg/L calcium pentothenate, 1Â€ mg/L biotin, 1Â€mg/L folic acid, 1Â€mg/L niacinamide, and 1Â€mg/L pyridoxal phosphate.

2.1.3. Purification of PLN

1. Lysis buffer: 20Â€ mM PBS pH 7.3, 120Â€ mM NaCl, 8Â€ mM EDTA, 0.1Â€mM DTT, 52.6Â€mM glycerol, 0.5Â€mg/mL pepstatin A, 0.5Â€mg/mL leupeptin, 2.5Â€mM lysozyme, and 4.5Â€mM Tween-20. 2. Amylose affinity chromatography resin (New England Biolabs, Ipswich, MA). 3. Wash buffer: 20Â€mM PBS pH 7.3, 120Â€mM NaCl. 4. Maltose elution buffer: 46Â€mM maltose in 20Â€mM PBS pH 7.3, 120Â€mM NaCl, 1Â€mM EDTA, and 3.1Â€mM NaN3. 5. Recombinant tobacco etch virus serine protease (Invitrogen, Carlsbad, CA or expressed and purified in our own lab) (see Note 2).

306

Veglia et al.

6. Dialysis membranes with 1,000-kDa molecular weight cutoff. 7. HPLC purification: H2O with 0.1% trifluoroacetic acid and isopropanol. 2.1.4. Quantification

1. Bio-Rad gradient 10–15% polyacrylamide gel (Bio-Rad, Hercules, CA). 2. Protein standard.

2.1.5. Phosphorylation of PLN at S16 by cAMP-Dependent Protein Kinase A

1. Phosphorylation buffer: 20Â€ mM MOPS (pH 7.3), 1% b-Octylglucoside, 1Â€mM ATP, and 1Â€mM MgCl2.

2.2. SERCA Preparation

1. White muscle from the back and hind legs of rabbit.

2. cAMP-dependent protein kinase A catalytic subunit (Sigma) or recombinant kinase as described in ref. 20.

2. Extraction buffer: 20Â€mM MOPS and 0.1Â€M KCl, pH 7. 3. Sucrose buffer: 20Â€mM MOPS, 0.3Â€M sucrose, 1Â€mM NaN3. 4. Reactive red affinity resin (Sigma). 5. SERCA buffer: 5Â€ mM DPC, 1Â€ mM CaCl2, 1Â€ mM MgCl2, 20Â€mM MOPS (pH 7.0), 20% glycerol, 0.25Â€mM DTT, and 4Â€mM ADP.

2.3. Functional Assays 2.3.1. SERCA PLN Co-reconstitution

1. 1,2-Dioleoyl-sn-glycero-3-phosphocholine (DOPC) and 1,2-dioleoyl-sn-glycero-phosphoethanolamine (DOPE) (Avanti Polar Lipids, Alabaster, AL). 2. Biobeads (Bio-Rad, Hercules, CA). 3. b-Octylglucoside (Sigma).

2.3.2. ATPase Activity Measurements

1. Coupled enzyme assay mix containing 0.5Â€mM phosphoenol pyruvate, 2.5Â€mM ATP, 0.2Â€mM NADH, 2Â€IU of pyruvate kinase, 2Â€ IU of lactate dehydrogenase, and 1–2Â€ mg calcium ionophore (A23187).

2.3.3. 31P NMR Activity Assays

1. Coupled enzyme assay mix containing a final concentration of 10Â€ mM dodecylphosphocholine, 1Â€ mM MgCl2, 80Â€ mM ATP, 20Â€mM MOPS (pH 7.0), and 0.2% C12E8.

2.4. NMR Spectroscopy

1. Dodecylphosphocholine (DPC) (Avertec or Cambridge Isotope Laboratories).

2.4.1. Solution-State Samples for Titration

2. Solution-state NMR sample buffer containing 20Â€ mM Na2PO4, 120Â€mM NaCl, and 0.1% NaN3. 3. Mini-dialysis chambers (Millipore, Billerica, MA). 4. Guanidinium–HCl (Sigma). 5. Dialysis buffer: 20Â€mM PBS and 50Â€mM b-mercaptoethanol.

What Can We Learn from a Small Regulatory Membrane Protein? 2.4.2. Oriented Samples and ssNMR Spectroscopy

307

1. 10% (w/v) sodium dodecyl sulfate (SDS). 2. DOPC and DOPE lipids. 3. Ultra filtration concentrator (Amicon, Millipore Corporation, Bedford, MA). 4. Glass plates, 8â•›×â•›12Â€ mm (Marienfeld GmbH & Co, LaudaKonigshofen, Germany). 5. Trifluoroethanol. 6. Extruder (Northern Lipids Inc. Burnaby, BC, Canada).

3. Methods 3.1. Expression and Purification of Monomeric and Pentameric PLN

For structural studies of membrane proteins by NMR, it is necessary to find a robust expression system that allows for the purification of milligram quantities of pure protein. For this purpose, we optimized a fusion protein construct for both monomeric (cysteine null PLN with the following mutations C36A, C41F, and C46A, which we refer to as AFA–PLN) and pentameric species (21).

3.1.1. Construction of PLN and Mutant PLN Expression Plasmids

To maximize the expression, all of the codons were optimized for E. coli. The nucleic acid sequence was divided into four overlapping primers and three polymerase chain reactions (PCRs) were performed, with incorporation of a TEV protease cleavage site between MBP and PLN, an EcoRI restriction site at the 5¢ end, and a HindIII site at the 3¢ end (a map of the plasmid is available from ref. 21). Each PCR was performed with a total volume of 50Â€mL containing 0.06Â€ mM of each oligonucleotide, 80Â€ mg of parental template, 200Â€mM dNTP mix, and 5.0Â€U Pfu Turbo DNA polymerase. Each PCR was carried out in an Eppendorf Master Cycler system with standard temperature cycling suggested for Pfu Turbo DNA polymerase. AFA–PLN PCR product was amplified and then ligated into EcoRI and HindIII sites on the pMal c2E vector using T4 DNA ligase (Rationale for fusion partner is described in Note 1). The ligated product was used to transform XL1-Blue competent cells with transformants selected by growth on LB/ampicillin plates. For single-site mutations of PLN, the forward and reverse primers were designed with optimized codons for E. coli three residues before and after the target residue to be mutated. PCR was performed with Pfu DNA polymerase using the protocol described above. PCR products were then digested with Dpn-I in order to remove nonmutated methylated dsDNA and then were transformed into XL-Blue competent cells. Constructs were amplified and extracted using the Qiaquick Spin Miniprep kit.

3.1.2. PLN Expression

E. coli BL21(DE3) cells were transformed with MBP–PLN fusion constructs. One hundred milliliters of LB growth media with

308

Veglia et al.

50Â€ mg/mL ampicillin was inoculated with a single colony and grown ~16Â€h at 25°C shaking at 250Â€rpm to reach an OD600 of ~1.0. The growth was then diluted (1:50) into M9 media containing mineral and vitamin cocktails and grown at 37°C to an OD600 of ~1.0. Protein expression was induced with 1Â€mM IPTG for 4.5Â€ h at 37°C. All cells were harvested by centrifugation at 6,370â•›×â•›g for 20Â€min at 4°C. Pellets were stored at −20°C. 3.1.3. PLN Purification

PLN pellets (~10–15Â€g) were resuspended in ~400Â€mL of lysis buffer and blended using a Sorval Omni-Mixer 17105. Cell lysates were then sonicated on ice for 15Â€min using a Branson Sonifier 450 sonicator at an output setting of 4 and a duty cycle of 45%. After sonication, lysates were spun down at 45,700â•›×â•›g for 20Â€ min at 4°C. The supernatant containing MBP/PLN fusion protein was applied to an amylose resin affinity chromatography column at 4°C. The column was washed with 12 column volumes of wash buffer until the OD280 was <0.05. The protein was eluted using a maltose solution, which MBP preferentially binds, releasing the protein from the amylose resin. PLN was cleaved from MBP with TEV protease at 100Â€U/mg fusion protein at 30°C for ~5–8Â€h. At this point, PLN can be further purified in two ways: (1) an S-100 sephacryl gel filtration column using an AKTAprime liquid chromatography system (Amersham Pharmacia Biotech) or (2) using reversed-phased HPLC. While we originally used the gel filtration method, we have abandoned this method in favor of the faster and higher resolving HPLC method as reported by Young and co-workers (22). For the HPLC method, the cleavage reaction was dialyzed against 4Â€L of ddH2O in a 1,000-Da cutoff membrane until PLN precipitated (~12Â€h). The suspension was centrifuged at 6,370â•›×â•›g for 20Â€ min at 4°C. The supernatant containing soluble proteins (MBP, TEV Â�protease, uncleaved fusion protein) was discarded. Pellets were Â�solubilized in 1–2Â€mL of 10% (w/v) SDS and stored at 4°C. This crude Â�solution was purified by reversed phase HPLC using an Agilent 1100 system (Agilent Technologies, Inc. Santa Clara, CA, USA) equipped with a Vydac 208TP C8 monomeric reversed phase column (250â•›×â•›1.0Â€mm i.d. ×10Â€mm particle size, Grace Vydac, Hesperia, CA, USA). The mobile phases used were: A – H2O with 0.1% TFA and B – isopropanol. The column (incubated at 50°C) was equilibrated with 10% solvent B and proteins were eluted at 2Â€mL/min flow rate using the following gradient: 10–30% solvent B in 10Â€min, 30–50% in 20Â€min, 50–70% in 30Â€min, and 70–100% in final 10Â€min. The column was re-equilibrated with 10% B for 10Â€ min after each run. With these conditions, the retention time for PLN was ~59Â€min. PLN fractions were pooled and isopropanol was evaporated under a stream of N2. The resulting suspension was lyophilized and stored at −20°C.

What Can We Learn from a Small Regulatory Membrane Protein?

309

3.1.4. Quantification of Protein

Protein concentration at each purification step was determined using protein absorbance at 280Â€nm, or assayed with densitometry measurements of Coomassie or silver stained gels (Fig.Â€ 1a). Densitometry data was collected on a Bio-Rad Molecular Imager FX using Bio-Rad Quantity One quantitation software. Recombinant protein concentrations were determined by comparison to a standard curve of 1–3Â€ mg synthetic PLN peptides previously quantified by amino acid analysis. PLN was also identified with MALDI-TOF MS and amino acid sequence analysis.

3.1.5. Phosphorylation of PLN at S16 by PKA

HPLC-purified PLN was reconstituted to a concentration of 0.25Â€mg/mL for phosphorylation. One milligram PLN was phosphorylated with 1,000Â€ U of protein kinase A catalytic subunit. Phosphorylation was confirmed using a gel-shift from a Western blot with anti-pS16 PLN antibody 285 and by MALDI-TOF MS.

3.2. SERCA Preparation

SERCA was extracted from rabbit skeletal muscle, purified, and tested for activity as reported previously (15, 17, 18). A solution containing 0.45Â€mM of SERCA, 5Â€mM DPC, 1Â€mM CaCl2, 1Â€mM MgCl2, 20Â€mM MOPS, 20% glycerol, 0.25Â€mM DTT, 0.4Â€mM ADP, pH of 7.0 was used for titration experiments (see below).

3.3. Functional Assays

Lyophilized PLN protein was solubilized in 240Â€mL of chloroform containing 2.4Â€mg of lipids in a 4:1 molar ratio of 1,2-dioleoyl-snglycero-3-phosphocholine (DOPC) and 1,2-dioleoyl-sn-glycero3-phosphoethanolamine (DOPE). Samples were then dried by removing the organic solvent under a stream of N2 gas. The dried film of lipid and PLN was then hydrated in 120Â€ mL of 25Â€ mM imidazole (pH 7.0) by vortexing followed by brief (30–60Â€s) water bath sonication. The lipid/PLN vesicles were then adjusted to contain a final concentration of 20Â€mM imidazole (pH 7.0), 0.1Â€M KCl, 5Â€mM MgCl2, and 10% glycerol. 4.8Â€mg of b-octylglucoside was added, followed by 60Â€mg of purified SERCA with the final volume adjusted to 300Â€mL with buffer. Removal of the detergent was performed by incubation with 120Â€mg of wet Biobeads for 3Â€h at 25°C. The SERCA/PLN lipid vesicles were separated from Biobeads using a gel-loading tip and micropipette and immediately assayed for function.

3.3.1. SERCA/PLN Co-reconstitution

3.3.2. ATPase Activity Measurements

SERCA activity was assayed by an enzyme-linked assay (23) performed in a 96-well plate. Each assay was performed at 37°C in a Thermomax microplate reader (Molecular Devices) in triplicate at different free calcium concentrations in a total volume of 175Â€mL. Between 1 and 3Â€mg of SERCA was added to the samples to start the assay, and the absorbance of NADH was monitored at 340Â€nm to determine the rate of ATP hydrolysis. Data were plotted in Origin 7.5 and fit using the Hill equation (see below) to determine the shift in pKCa (Fig.Â€1b).

310

Veglia et al.

Fig.Â€1. Purification and functional assays of PLN and SERCA. (a) Purification gels showing expression of fusion protein (before and after induction with IPTG), selection on amylose affinity column, and cleavage by TEV protease (left gel). Middle gel shows PLN monomer and pentamer reconstituted in DOPC/DOPE lipid bilayers and DPC detergent micelles. Right gel shows purification of SERCA from crude SR; light SR refers to purified SERCA used for activity assays and NMR experiments. (b) Normalized SERCA activity in the absence (squares) and presence of PLN (circles), and in the presence of phosphorylated PLN at S16 (triangles). This activity assay was performed in lipid bilayers using the coupled enzyme assay in a microplate reader. (c) SERCA activity monitored using 31P NMR spectroscopy in DPC micelles. (d) Example of the 31P NMR activity assay where the inorganic phosphate signal (highlighted in gray) is monitored as a function of time. Middle gel in (a) is reproduced with permission from ref. 37 Copyright 2007 National Academy of Sciences, USA. Panel (b) is reproduced with permission from ref. 19 Copyright 2008 American Chemical Society. Panels (c, d) are reproduced with permission from ref. 18 Copyright 2006 Elsevier.

3.3.3. 31P NMR Activity Assays

To measure the hydrolysis of ATP by SERCA under solution NMR conditions (DPC detergent micelles), assays were performed by monitoring the inorganic phosphate 31P NMR signal. The assay mixture consisted of 6Â€ mM SERCA, 137Â€ mM PLN, assay mix, and concentrations of CaCl2 required to reach free

What Can We Learn from a Small Regulatory Membrane Protein?

311

Ca2+ concentration calculated based on the method of Fabiato and Fabiato (24). Upon each addition of nucleotide, eight free induction decays (FIDs) were collected, each consisting of 16 single pulse transients on 31P. Data were collected on a Varian Inova spectrometer operating at a 1H Larmor frequency of 500Â€MHz (Fig.Â€1). After monitoring the rate, the inorganic phosphate signal was normalized to the known amount of DPC in the sample. This normalized amount was then plotted (V vs. pCa) and fit by the Hill equation (Fig.Â€1c).

V = V max / [1 + 10n (pKCa - pCa ) ]

(1)

where V is the initial ATPase rate and n is the Hill coefficient. The data were normalized to the maximal rate, Vmax, which was obtained from the fit, and then replotted to determine the shift in pKCa (Fig.Â€1b). 3.4. NMR Spectroscopy 3.4.1. Preparation of Solution NMR Samples and Titrations

3.4.2. Kd Measurements by NMR

NMR samples were prepared by dissolving isotopically labeled PLN in NMR sample buffer containing 300Â€mM DPC and 10% D2O to a final concentration of 0.23Â€mM. For the preparation of wt-PLN NMR samples (or other “difficult” samples; see Note 3), a slightly different procedure was used: wt-PLN was solubilized in 300Â€mM DPC, 20Â€ mM PBS (pH 6.0), and 6Â€ M guanidinium–HCl. The sample was dialyzed overnight in a mini-dialysis chamber against 1Â€L dialysis buffer. For the NMR titration, SERCA at a concentration of 0.51Â€mM was added incrementally to the PLN sample as previously reported (17). To correct for this dilution, and potential pH and salt effects, the same titration experiment was performed with SERCA buffer in the absence of enzyme. After each SERCA addition, a 1H–15N heteronuclear single quantum coherence (HSQC) experiment was collected on a Varian Inova spectrometer operating with a 1H Larmor frequency of 600Â€MHz at 37°C, using an inverse detection triple-resonance and triple-axis gradient probe. The HSQC pulse program was equipped with pulsed field gradients for both coherence selection and sensitivity enhancement (25). For measuring the binding constants, we assumed 1:1 molar ratio between SERCA and PLN, and that the intensity reduction (Iretention) of the transmembrane domain residues in Fig.Â€ 2c was directly related to the fraction of PLN bound to SERCA (fb). For a complete derivation of Eq.Â€3 see ref. 18. f b = 1 - I retention

fb =

K d + [SERCA]t + [PLN]t - (K d + [SERCA]t + [PLN]t )2 - 4[SERCA]t [PLN]t 2[PLN]t

(2)

(3)

The dissociation constant (Kd) was calculated using a nonlinear fit to Eq.Â€3.

312

Veglia et al.

Fig.Â€2. Solution NMR results on AFA–PLN/SERCA complex in detergent micelles. (a) The intensity retention plotted for the T state peaks of residues 2, 3, 6, and 8 in p16 PLN (white) and PLN (black) that disappear with increasing additions of SERCA. (b) The percentage of PLN in the R state as a function of the SERCA/PLN molar ratio for p16 PLN (white) and PLN (black) for residues undergoing slow exchange (residues 10–12, 15–17, and 22). A value of 100% indicates that the peak (residue) has completely exchanged to the R state. (c) The intensity retention for residues in the transmembrane domain (28–52) that decrease, but do not abolish in intensity. (d) EPR spectrum of TOAC-labeled A11 PLN in lipid bilayers in the absence of PLN. Since EPR probes a faster timescale, both the T and R states are observed in lipids. (e) Representative 1D and 2D spectra of the residues with abolished intensity (top, corresponds to a), residues undergoing slow exchange (middle corresponds to b), and residues with reduced intensity (bottom, corresponds to c). Coordinates for the PLN/ SERCA molecular model were generously provided by Drs. MacLennan and Toyoshima. Panels (a–c), and (e) were adapted with permission from ref. 18 Copyright 2006 Elsevier. Panel (d) was reproduced with permission from ref. 17 Copyright 2005 National Academy of Sciences, USA.

3.4.3. Preparation of Oriented Samples and Solid-State NMR Spectroscopy

PLN was reconstituted into lipid vesicles using two different preparations: (1) detergent mediated reconstitution and (2) organic solvent mediated reconstitution. Detergent mediated reconstitution:â•‡ 80Â€mg (4/1, w/w) of DOPC/ DOPE were dissolved in chloroform and thoroughly vortexed. Chloroform was evaporated under a stream of N2, after which the lipids were resuspended in 40Â€mL of ddH2O. The lipid suspension was sonicated on ice until small unilamellar vesicles (SUVs) were formed (the suspension became transparent). After sonication, the lipids were centrifuged (6,370â•›×â•›g for 10Â€min at 4°C) to remove larger vesicles and metal particles. PLN (~4Â€mg) was solubilized in 1Â€mL of SDS (10%, w/v), dissolved in the lipid mixture prior, and subjected to one freeze–thaw cycle. Samples were extensively dialyzed against ddH2O to remove the detergent and Â�subsequently

What Can We Learn from a Small Regulatory Membrane Protein?

313

concentrated to 2Â€ mL using an ultrafiltration device (10Â€ kDa Â� cutoff membrane). Approximately 100Â€µL of sample was transferred onto each of 20 glass plates. Samples were slowly dried at 40°C, rehydrated, and finally sealed in a rectangular glass cell. Organic solvent mediated reconstitution: DOPC/DOPE 80Â€ mg (4/1, w/w) were dissolved in chloroform and thoroughly mixed. PLN (~4Â€ mg) was solubilized in 50Â€µL of trifluoroethanol and added to the lipid mixture. Solvents were first evaporated under a stream of N2 and then lyophilized overnight to ensure complete removal of organic solvents. The lipid/protein mixture was resuspended in 40Â€mL of ddH2O and carefully vortexed. Small unilamellar vesicles were prepared by repeatedly extruding the lipid mixture through polycarbonate filters of decreasing pore size (200, 100, 50Â€nm) using a bench-top extruder. This SUV suspension was concentrated to 2Â€mL with samples dried and rehydrated as reported above. The final molar ratio of lipid/protein for all samples was ~200/1. NMR spectroscopy: The 2D polarization inversion spin exchange at the magic angle (PISEMA) was performed (26, 27) with TPPM decoupling during acquisition (28). A phase modulated LeeGoldburg (PMLG) sequence was used in the indirect dimension to decouple 1H–1H interactions and allow for the evolution of 1 H–15N dipolar coupling (29). The initial 90° 1H pulse, crosspolarization, PMLG (1H effective field), and TPPM decoupling during acquisition were applied at ~60Â€ kHz RF field strength. Spectra of uniformly 15N samples were acquired with 1Â€ k scans and 30t1 increments, while selectively labeled samples required 4–12Â€ k scans and ~8–16 increments. All experiments were performed at the National High Magnetic Field Laboratory at a 14.1-T magnetic field strength (1H frequency of 600.1Â€ MHz) equipped with a Bruker DMX spectrometer using a low-E probe built by the RF program (30). 3.4.4. Structure Determination of Monomeric PLN Using a Hybrid Solution and Solid-State NMR Method

Solution and solid-state NMR methods have been used as a complementary approach to the structure determination of small and medium-size membrane proteins (31, 32). For solution NMR, the proteins are reconstituted in detergent micelles with a meticulous choice of experimental conditions, which is based on the NMR spectral features as well as preservation of SERCA function. For solid-state, we reconstitute membrane proteins in synthetic lipid bilayers, which are more amenable to enzyme function. The major advantage of solution NMR is to offer high-resolution spectra (membrane proteins up to 100Â€kDa have now been Â�studied and assigned) and to obtain secondary and tertiary structures through NOE measurements. On the other hand, when membrane proteins are solubilized in detergent micelles, the topological information of the protein within the lipid bilayer is lost.

314

Veglia et al.

This orientation information regarding the molecular topology can be recovered by solid-state NMR measurements using aligned samples. Recent reports verify that several membrane proteins adopt a similar structure in both lipid bilayers and detergent micelles; therefore, we decided to pool the information derived from these two techniques and combine the structural restraints into a unique molecular modeling protocol (33, 34). The total potential energy function (Etotal) is defined as a combination of an empirical energy function (Echem) and two penalty functions that include solution (EsolNMR) and solid-state NMR (EssNMR) restraints (34):

E total = E chem + E solNMR + E ssNMR

(4)

E chem = E bonds + E angles + E torsion + E improper + E vdw

(5)

E solNMR = E NOEs + E CDIH + E HBOND + E RDC + E PRE + E CSP

(6)

E ssNMR = E CSA + E DC

(7)

where

The energy terms of Echem are included into standard force field of XPLOR-NIH (35). EssNMR contains dipolar coupling and anisotropic chemical shift values derived from PISEMA or HETCOR experiments. As a result, the simulated annealing protocol minimizes the energy function for both distance and angular restraints as well as orientational restraints derived from solid-state NMR PISEMA experiments. FigureÂ€ 3d shows the average structures from the lowest energy conformational ensemble for the PLN monomer (33). 3.5. Conclusions 3.5.1. Allosteric Model of SERCA Regulation by PLN

The outcomes of our investigation are summarized within the schematic in Fig.Â€4. In agreement with both inÂ€vitro and inÂ€vivo studies, we found that PLN adopts a pentameric, L-shaped conformation both in micelles and in lipid bilayers. In particular, both solution and solid-state NMR data show that the cytoplasmic domain (which is amphipathic) interacts with the surface of the membrane-mimicking environment (both for micelles and lipid bilayers), while the transmembrane domain forms a tight hydrophobic bundle stapled together by a Â�leucine–isoleucine zipper. PLN depolymerization into active monomers occurs in the presence of SERCA. In the monomeric state (prelude to the interaction with SERCA), PLN is predominantly in a resting T state with a small population in the R state. Using EPR spectroscopy, Thomas and co-workers have estimated that ~16% of PLN exists in the R state in lipid bilayers at 4°C (13). In the presence of SERCA, the equilibrium

What Can We Learn from a Small Regulatory Membrane Protein?

315

Fig.Â€3. Solid-state NMR spectra with overlaid ensemble PLN structures. PISEMA spectra for uniformly 15N-labeled PLN monomer (a) and selectively labeled PLN pentamer ((b) transmembrane domain residues and (c) cytoplasmic domain residues). Note that dipolar coupling values within the spectra are scaled by the theoretical 0.82 factor for the PISEMA experiment (26). Panel (d) shows the overlay of the structural ensemble for monomeric PLN determined using a hybrid of solution and solid-state NMR restraints (PDB 2KB7) (33). Panel (e) shows a cartoon representation of the PLN pentamer indicating the pinwheel assembly (37). Panels (a–c) are reproduced with permission from refs. 33, 37. Copyright 2007 and 2009 National Academy of Science, USA.

is shifted toward the R state that is probably selected to bind the enzyme. While this initial detection of PLN ground (T) and excited (R) states shed light into the dynamic regulatory process of SERCA, more experiments are needed to characterize the excited states both structural and dynamically. Our ongoing investigation is revealing a more complex equilibrium than previously anticipated. Although dynamic in nature, PLN forms a stable inhibitory complex with SERCA (Kd in DPC detergent micelles ~60Â€ mM). This inhibition is reversed by phosphorylation at S16, which changes the conformational dynamics and results in a partial unwinding of the cytoplasmic helix (20). In the presence of SERCA, phosphorylation at S16 PLN causes a slight rearrangement within domain Ib (residues 23–30) and changes the dynamics at the binding interface (13, 18). These structural dynamics affect the conformational transitions of the enzyme, and ultimately calcium translocation. 3.5.2. Rational Design of PLN Mutants as Possible Candidates for Gene Therapy

A fundamental aim of structural biology is to move from understanding structure and dynamics to controlling molecular function. With this in mind, we attempted to manipulate the structural dynamics of PLN and to promote the formation of the R state

316

Veglia et al.

Fig.Â€4. Allosteric model of PLN/SERCA interaction showing depolymerization of pentameric PLN into monomeric units that undergo exchange between an ordered T state and a disordered R state(s) (11, 13, 17–19).

through mutagenesis. Recently, we found that by mimicking the S16 phosphorylated state of PLN (i.e., enhancing the local dynamics in the hinge region), it is possible to tune the extent of SERCA inhibition (36). This work was originally inspired by inÂ€ vivo studies carried out by Chien and co-workers, who demonstrated that a pseudo-phosphorylated variant of PLN can relieve SERCA inhibition and increase heart contractility, ultimately reversing the damages of myocardial infarction. We found that by changing Pro21 to Gly it is possible to generate a loss-of-function mutant with characteristics similar to S16E (pseudo-phosphorylated species tested by Chien), preserving the posttranslational control by b-adrenergic stimuli (36). This new generation of mutants possesses a residual inhibitory power that could be important in adjusting the pathophysiology of diseased hearts leading to reversal or hindrance of cardiac remodeling. We are currently working on different sites to improve the characteristics of loss-of-function mutants.

4. Notes 1. For small membrane peptides, choice of fusion partner and cleavage protease can be crucial for successful expression and purification. Expression of small membrane peptides can

What Can We Learn from a Small Regulatory Membrane Protein?

317

induce the formation of inclusion bodies, or lead to cell death. We attempted cloning PLN with several fusion partners and found MBP to offer the best degree of solubility and expression Â�levels. For choosing a cleavage protease, scanning for secondary cleavage sites is also very important. For PLN, cleaving with thrombin and other proteases led to secondary cleavage. TEV has a longer recognition sequence (seven residues), making it very specific. 2. TEV can be easily expressed in E. coli and purified using a 6× His and Ni-NTA resin chromatography within the laboratory. A construct of TEV that is particularly effective which we use is a S219V mutant with a poly-arginine tail kindly provided to us by the Gorelick Laboratory. 3. The method of unfolding the protein sample with guanidineHCl and refolding via dialysis used to prepare the wt-PLN solution-state NMR samples described in SubheadingÂ€3.4.1 is useful for other membrane protein NMR samples that are prone to aggregation or those which give nonuniform HSQC spectra. We have found that other “difficult” samples (such as pS16 PLN) give better spectra when prepared in this manner.

Acknowledgment This work was supported by grants to GV from the National Institutes of Health (GM64742, HL80081, GM072701) and NJT (AHA 0515491Z). PISEMA spectra were acquired at the NHMFL, Tallahassee, FL (DMR-0084173). NMR instrumentation at the University of Minnesota High Field NMR Center was funded by the National Science Foundation (BIR-961477) and the University of Minnesota Medical School. References 1. Macâ•›Lennan DH, Kranias EG (2003) Phospholamban: a crucial regulator of cardiac contractility. Nat Rev 4:566–577 2. Kranias EG, Bers DM (2007) Calcium and cardiomyopathies. Subcell Biochem 45:523–537 3. Hoshijima M, Knoll R, Pashmforoush M, Chien KR (2006) Reversal of calcium cycling defects in advanced heart failure toward molecular therapy. J Am Coll Cardiol 48:A15–A23 4. Inesi G, Lewis D, Ma H, Prasad A, Toyoshima C (2006) Concerted conformational effects of Ca2+ and ATP are required for activation of sequential reactions in the Ca2+ ATPase (SERCA) catalytic cycle. Biochemistry 45:13769–13778

5. Toyoshima C, Inesi G (2004) Structural basis of ion pumping by Ca2+-ATPase of the sarcoplasmic reticulum. Annu Rev Biochem 73:269–292 6. Toyoshima C, Nakasako M, Nomura H, Ogawa H (2000) Crystal structure of the calcium pump of sarcoplasmic reticulum at 2.6â•›Å resolution. Nature 405:647–655 7. Inesi G, Prasad AM, Pilankatta R (2008) The Ca2+ ATPase of cardiac sarcoplasmic reticulum: physiological role and relevance to diseases. Biochem Biophys Res Commun 369:182–187 8. Negash S, Chen LT, Bigelow DJ, Squier TC (1996) Phosphorylation of phospholamban

318

Veglia et al.

by cAMP-dependent protein kinase enhances interactions between Ca-ATPase polypeptide chains in cardiac sarcoplasmic reticulum membranes. Biochemistry 35:11247–11259 9. Negash S, Sun H, Yao Q, Goh SY, Bigelow DJ, Squier TC (1998) Cytosolic domain of phospholamban remains associated with the Ca-ATPase following phosphorylation by cAMP-dependent protein kinase. Ann N Y Acad Sci 853:288–291 10. Negash S, Yao Q, Sun H, Li J, Bigelow DJ, Squier TC (2000) Phospholamban remains associated with the Ca2+- and Mg2+-dependent ATPase following phosphorylation by cAMPdependent protein kinase. Biochem J 351:195–205 11. Mueller B, Karim CB, Negrashov IV, Kutchai H, Thomas DD (2004) Direct detection of phospholamban and sarcoplasmic reticulum Ca-ATPase interaction in membranes using fluorescence resonance energy transfer. Biochemistry 43:8754–8765 12. Kirby TL, Karim CB, Thomas DD (2004) Electron paramagnetic resonance reveals a large-scale conformational change in the cytoplasmic domain of phospholamban upon binding to the sarcoplasmic reticulum Ca-ATPase. Biochemistry 43:5842–5852 13. Karim CB, Zhang Z, Howard EC, Torgersen KD, Thomas DD (2006) Phosphorylationdependent conformational switch in spinlabeled phospholamban bound to SERCA. J Mol Biol 358:1032–1040 14. Thomas DD, Reddy LG, Karim CB etÂ€ al (1998) Direct spectroscopic detection of molecular dynamics and interactions of the calcium pump and phospholamban. Ann N Y Acad Sci 853:186–194 15. Stokes DL, Green NM (1990) Threedimensional crystals of CaATPase from sarcoplasmic reticulum. Symmetry and molecular packing. Biophys J 57:1–14 16. Stokes DL, Pomfret AJ, Rice WJ, Glaves JP, Young HS (2006) Interactions between Ca2+-ATPase and the pentameric form of phosÂ� pholamban in two-dimensional co-crystals. Biophys J 90:4213–4223 17. Zamoon J, Nitu F, Karim C, Thomas DD, Veglia G (2005) Mapping the interaction surface of a membrane protein: unveiling the conformational switch of phospholamban in calcium pump regulation. Proc Natl Acad Sci U S A 102:4747–4752 18. Traaseth NJ, Thomas DD, Veglia G (2006) Effects of Ser16 phosphorylation on the allosteric transitions of phospholamban/ Ca(2+)-ATPase complex. J Mol Biol 358: 1041–1050

19. Traaseth NJ, Ha KN, Verardi R etÂ€al (2008) Structural and dynamic basis of phospholamban and sarcolipin inhibition of Ca(2+)ATPase. Biochemistry 47:3–13 20. Metcalfe EE, Traaseth NJ, Veglia G (2005) Serine 16 phosphorylation induces an orderto-disorder transition in monomeric phospholamban. Biochemistry 44:4386–4396 21. Buck B, Zamoon J, Kirby TL etÂ€ al (2003) Overexpression, purification, and characterization of recombinant Ca-ATPase regulators for high-resolution solution and solid-state NMR studies. Protein Expr Purif 30:253–261 22. Douglas JL, Trieber CA, Afara M, Young HS (2005) Rapid, high-yield expression and purification of Ca2+-ATPase regulatory proteins for high-resolution structural studies. Protein Expr Purif 40:118–125 23. Madden TD, Chapman D, Quinn PJ (1979) Cholesterol modulates activity of calciumdependent ATPase of the sarcoplasmic reticulum. Nature 279:538–541 24. Fabiato A, Fabiato F (1979) Calculator programs for computing the composition of the solutions containing multiple metals and ligands used for experiments in skinned muscle cells. J Physiol 75:463–505 25. Kay LE, Keifer E, Saarinen T (1992) Pure absorption gradient enhanced heteronuclear single quantum correlation spectroscopy with improved sensitivity. J Am Chem Soc 114:10663–10665 26. Wu CH, Ramamoorthy A, Opella SJ (1994) High-resolution heteronuclear dipolar solidstate NMR spectroscopy. J Magn Reson 109:270–272 27. Ramamoorthy A, Wei Y, Lee D (2004) PISEMA solid-state NMR spectroscopy. Annu Rep NMR Spectrosc 52:1–52 28. Bennett AE, Rienstra CM, Auger M, Lakshmi KV, Griffin RG (1995) Heteronuclear decoupling in rotating solids. J Chem Phys 103:6951–6958 29. Vinogradov E, Madhu PK, Vega S (1999) High-resolution proton solid-state NMR spectroscopy by phase-modulated Lee–Goldburg experiment. Chem Phys Lett 314:443–450 30. Gor’kov PL, Chekmenev EY, Li C etÂ€al (2007) Using low-E resonators to reduce RF heating in biological samples for static solid-state NMR up to 900Â€MHz. J Magn Reson 185:77–93 31. Opella SJ, Marassi FM (2004) Structure determination of membrane proteins by NMR spectroscopy. Chem Rev 104:3587–3606 32. Gao FP, Cross TA (2005) Recent developments in membrane–protein structural genomics. Genome Biol 6:244

What Can We Learn from a Small Regulatory Membrane Protein? 33. Traaseth NJ, Shi L, Verardi R, Mullen DG, Barany G, Veglia G (2009) Structure and topology of monomeric phospholamban in lipid membranes determined by a hybrid solution and solid-state NMR approach. Proc Natl Acad Sci U S A 106: 10165–10170 34. Shi L, Traaseth NJ, Verardi R, Cembran A, Gao J, Veglia G (2009) A refinement protocol to determine structure, topology, and depth of insertion of membrane proteins using hybrid solution and solid-state NMR restraints. J Biomol NMR 44:195–205

319

35. Schwieters CD, Kuszewski JJ, Tjandra N, Clore GM (2003) The Xplor-NIH NMR molecular structure determination package. J Magn Reson 160:65–73 36. Ha KN, Traaseth NJ, Verardi R etÂ€ al (2007) Controlling the inhibition of the sarcoplasmic Ca2+-ATPase by tuning phospholamban structural dynamics. J Biol Chem 282:37205–37214 37. Traaseth NJ, Verardi R, Torgersen KD, Karim CB, Thomas DD, Veglia G (2007) Spectroscopic validation of the pentameric structure of phospholamban. Proc Natl Acad Sci U S A 104:14676–14681

as

Chapter 17 Solution-State NMR Spectroscopy of Membrane Proteins in Detergent Micelles: Structure of the Klebsiella pneumoniae Outer Membrane Protein A, KpOmpA Marie Renault, Olivier Saurel, Pascal Demange, Valérie Reat, and Alain Milon Abstract Structure determination of integral membrane proteins is one of the most important challenges of structural biology. Over the last decade, solution-state NMR spectroscopy has become an increasingly useful approach for 3D structure determination and dynamical analysis of membrane proteins solubilized in detergent micelles. We describe herein an ensemble of methods, including isotopic labelling, in vitro refolding procedure, and state-of-the-art NMR experiments typically applied for the structure determination of high molecular weight molecular complexes. Furthermore, the basic principles of spectrum interpretation and 3D structure calculation are reported. This approach is illustrated by a case study on the transmembrane domain of the outer membrane protein A from Klebsiella pneumoniae (KpOmpA). Key words: Outer membrane protein, Detergent, Perdeuteration, Selective methyl protonation, TROSY, Structure determination, 13C, 15N, 2H stable isotope labelling

1. Introduction Solution-state NMR spectroscopy is becoming an increasingly useful approach for structural and dynamical analysis of large molecular complexes, such as membrane proteins (MPs) solubilized in detergent micelles (1). Such complexes with large apparent molecular weight (i.e. typically above 50Â€kDa) exhibit slow molecular tumbling, resulting in fast transverse relaxation mechanisms, leading to increased linewidth and reduced sensitivity compared to spectra of smaller protein routinely solved by solution-state NMR

Jean-Jacques Lacapère (ed.), Membrane Protein Structure Determination: Methods and Protocols, Methods in Molecular Biology, vol. 654, DOI 10.1007/978-1-60761-762-4_17, © Springer Science+Business Media, LLC 2010

321

322

Renault et al.

spectroscopy. To minimize the deleterious effects of nuclear spin relaxation on solution NMR spectra, the use of the so-called transverse relaxation-optimized spectroscopy (TROSY) (2) at very high field combined with a high level of protein deuteration (>80% 2H) provide large gains in both sensitivity and resolution (3). Using this approach, complete or nearly complete backbone resonance assignments of integral membrane proteins (up to 280 residues) were obtained, leading to the determination of several b-barrel global folds over the last decade (4–8). To circumvent the limited amount of long range 1H–1H NOEs in perdeuterated proteins, the development of efficient tools for obtaining additional conformational restraints such as restraints obtained from residual dipolar couplings (9, 10), paramagnetic relaxation enhancements experiments (11), or selective methyl protonated otherwise perdeuterated samples (12, 13), were reported for the structure refinement of membrane proteins (14–18). These approaches are particularly useful in the case of helical membrane proteins, which exhibit only few long-range NOEs, which are crucial for the global fold determination (15, 19, 20). This chapter highlights the techniques and experimental protocols used for the investigation of the molecular structure of the 216-residue transmembrane domain of the Outer Membrane Protein A from Klebsiella pneumoniae (KpOmpA) in detergent micelles by liquid-state NMR spectroscopy (21). Among the complete set of state-of-the-art NMR methods for backbone assignment of large protein, we performed a 3D 13C-TOCSY-(15N,1H)-TROSY (22) for aliphatic carbon side chain assignment, a 4D 15N,15N separated NOESY (23) for unambiguous assignment of HN–HN NOEs, and selective methyl protonation approaches for the measurement of distance constraints involving methyl groups. Because structure determination of membrane protein placed strong constraints on the expression system efficiency (with the requirement of tens of milligram amount of pure MPs labelled with three low abundant isotopes) and on protein stability (in the presence of highly concentrated detergent molecules), the overexpression of KpOmpA in E. coli inclusion bodies with different 2 H,13C,15N isotope labelling patterns, its inÂ€vitro refolding in nondenaturing detergent micelles, and an efficient and fast detergent screening method for NMR sample preparation are developed.

2. Materials 2.1. Cells Culture and Isotope Labelling

1. Freshly transformed BL21(DE3) cells harboring the plasmid pET21c-KpOmpA. The cDNA for the transmembrane domain of KpOmpA (Met1-Glu207) with a C-terminal

Solution-State NMR Spectroscopy for Investigating 3D Structure of Membrane Proteins

323

hexahistidine tag was subcloned by using standard PCR methods into the NheI/HindIII sites of pET21c expression vector (Novagen, France) from the complete KpOmpA sequence into pVALP40 vector (24). 2. Filter-sterilized solutions of 1Â€M isopropyl b-d-thiogalactopyranoside (IPTG) and of the appropriate antibiotic stored in single use aliquots at −20°C. Working concentration for media are typically 50Â€mM ampicillin and 1Â€mM IPTG. 3. Sterile Luria–Bertani (LB) (Euromedex, France) growth medium. 4. Sterile M9 minimal medium: For 1Â€ L of growth medium, 12.8Â€g of Na2HPO4, 3.0Â€g of KH2PO4, 0.5Â€g of NaCl, 2Â€g of NH4Cl, 2Â€g of D-glucose, 2Â€mM of MgSO4, 0.1Â€mM of CaCl2, vitamins (0.01Â€ mg of cobalamine, 0.1Â€ mg of pyridoxal and folic acid, 0.5Â€mg of riboflavine, 1Â€mg of biotin, pantothenic acid and niacinamide, 25Â€mg of thiamine), traces (50Â€mg of EDTA, 60Â€mg of CaCl2, 60Â€mg of FeSO4, 4.4Â€mg of CuSO4, 12Â€mg of MnCl2, 0.2Â€mg of H3BO3, 7Â€mg of ZnSO4, 2Â€mg of ascorbic acid). Filter-sterilize. 5. Isotope labelled compounds (Cambridge Isotope Laboratory, USA): [15N]–NH4Cl (15N 99%); [13C6]-D-glucose (13C 99%); [13C6, 2H7]-D-glucose (13C 99%, 2H 97–98%); D2O (2H 99.98%); [3,3−2H2, 13C]-a-ketobutyrate (13C 98%, 3–3 2H2 98%); [3−2H, 13C]-a-ketoisovalerate (13C 97–98%, 2 H 98%). 6. All cultures were performed in 2-L flasks containing 500-ml culture medium. Incubation at 37°C at 220Â€rpm. 2.2. In Vitro Refolding and Protein Purification

1. Buffers: (A) 50Â€mM Tris/HCl, 200Â€mM NaCl, 5Â€mM EDTA, 0.05% Tween 20 (w/v) (Anatrace, USA) at pHâ•›=â•›8.5, (B) 50Â€ mM Tris/HCl, 200Â€ mM NaCl, 5Â€ mM EDTA, 0,05% Tween 20 (w/v) (Anatrace, USA), 2Â€M Urea at pHâ•›=â•›8.5, (C), denaturing buffer) 25Â€mM Tris/HCl, 5Â€ mM EDTA, 6Â€M guanidine hydrochloride (Gdn/HCl) at pHâ•›=â•›8.5, (D, refolding buffer) 25Â€ mM Tris/HCl, 150Â€ mM NaCl, 1% Anzergent 3-14 (w/v) (Anatrace, USA) identical to zwitergent 3-14(ZW 3-14) at pHâ•›=â•›8.5, (E, purification buffer) 25Â€mM Tris/HCl, 150Â€mM NaCl, 0.1% Anzergent 3-14(w/ v) (Anatrace, USA) at pHâ•›=â•›8.5. 2. Ni-NDA resin (Amersham, USA), conditioned according to the manufacturer’s procedure. 3. Dialysis membrane (MWCO 12,000–14,000Â€Da, SpectraPor, USA).

2.3. Detergent Screening and NMR Sample Preparation

1. Pure fraction of isotope labelled KpOmpA in 25Â€mM Tris/ HCl, 150Â€ mM NaCl, 0.1% ZW 3-14 (w/v), pHâ•›=â•›8.5, at a protein final concentration of about ~15Â€mg/mL.

324

Renault et al.

2. Solutions for protein precipitation: Cold ethanol (−20°C), solution of (4Â€ M) NaCl sterilized by filtration through 0.2Â€mm filter. 3. NMR buffer: 20Â€mM phosphate, NaCl 100Â€mM at pH 6.5. 4. Detergents: Dihexanoylphosphatidylcholine (DHPC, powder, Avanti Polar Lipids, USA), deuterated dihexanoylphosphatidylcholine (DHPC-d22, powder, CIL, USA), dodecylphosphocholine (DPC, powder, Anatrace, USA), Anzergent 3-14 (ZW 3-14, Anatrace, USA), tetraethylene glycol monooctyl ether (C8E4, Anatrace, USA). 2.4. NMR Spectroscopy, Data Analysis, and Structure Calculation

1. Four radiofrequency channel Bruker spectrometers using 1H (13C, 15N) triple resonance cryoprobe, and equipped with z pulsed field gradient coils. 2. Software: NMRpipe for NMR data processing (http://spin. niddk.nih.gov/NMRPipe); NMRView for spectra analysis, NMR assignment, and NOEs analysis (http://www.onemoonscientific.com/nmrview); CNS for structure calculation (http://cns-online.org/v1.21); PROCHECK for structure validation (http://www.biochem.ucl.ac.uk/~roman/ procheck/procheck.html); PYMOL for structure visualization (http://pymol.org); and other software with similar functions can be used.

3. Methods 3.1. Overexpression and Isotope Labelling Strategies of KpOmpA

The transmembrane domain of KpOmpA was conveniently overexpressed as a recombinant protein of 216 residues (with a C-terminus hexahistidine tag) in inclusion bodies after IPTG induction 1Â€mM at 37°C for 4Â€h in E. coli BL21(DE3) bacteria. Typical yields of KpOmpA are about 220Â€mg/L of rich medium after refolding and purification steps. Incorporation of labelled isotopes into a recombinant protein expressed in E. coli requires minimal growth media containing 15N, 13C, and 2H labeled metabolic sources. Although high deuteration levels are tolerated by the bacteria, a significant reduction of the growth rate and biomass yields is usually observed. Higher protein yields are usually obtained if bacteria are must be progressively adapted to the deuterium content for further improvement of protein production as described below. Protein yields were reported for each isotopic labelling patterns of KpOmpA (see TableÂ€1)

3.1.1. Uniform ( 2H,13C,15N) Labelling

1. Inoculate 3Â€ ml culture of LB-rich medium with a freshly transformed colony of BL21(DE3) cells and incubate the culture in a shaking incubator (220Â€ rpm) at 37°C during 3Â€ h (OD600 is about 2.0)

Solution-State NMR Spectroscopy for Investigating 3D Structure of Membrane Proteins

325

TableÂ€1 Protein yields for different isotope labelling patterns Labelling

Yields (mg/l)a

No

>200

[u−2H,13C,15N]

â•…â•›80

[u−2H,13C,15N/I(d1)LV−13CH3]

â•…â•›40

[u−15N/10%13C]

â•…â•›80

Quantity of KpOmpA per liter of growth medium after refolding and Â� purification steps

a

2. Spin down the cells (1,000â•›×â•›g at room temperature) and resuspend the pellet in 50Â€ml of labeled M9/H2O minimal medium (containing 2Â€ g/L of [15N]-NH4Cl and 2Â€ g/L of [13C6]-D-glucose) to achieve an initial OD600 of 0.08. The culture was grown at 37°C until reaching an OD600 of 0.8. 3. Spin down the cells (1,000â•›×â•›g at room temperature) and resuspend a small amount of cells in 150Â€ml of labeled M9/ D2O minimal medium (containing 99% D2O, 2Â€ g/L of [15N]-NH4Cl and 2Â€g/L of [13C6, 2H7]-D-glucose) to achieve a starting OD600 of 0.08. Incubate the culture at 37°C until reaching an OD600 of 2.0. 4. Dilute the culture to a final volume of 1Â€L with “fresh” 13C, 15 N-M9/D2O medium and incubate at 37°C until the culture reaches an OD600 of 0.5. 5. Induce protein overexpression with IPTG (1Â€mM) for 4 to 30Â€h at 37°C (see Note 1). 6. Collect the cells by centrifugation (5,000â•›×â•›g, at 4°C for 15Â€min) 3.1.2. Selective protonation of Isoleucine, Leucine, and Valine (ILV) Methyl Groups in a Perdeuterated Background

1. Use uniform (2H, 13C, 15N) labelling protocol described above until the step 4. 2. Approximately 1Â€ h before induction (OD600 is about 0.3), add the isotope labeled precursors in the following proportions: 50Â€ mg/L of [3,3−2H2, 13C]-a-ketobutyrate and 100Â€mg/L of [3−2H, 13C]-a-ketoisovalerate as final concentrations. Maintain the culture at 37°C for about 1Â€h until the culture reaches an OD600 of 0.5. 3. Induce protein overexpression with IPTG (1Â€mM) and collect cells after 3 to 4 hours in order to prevent generation of undesired isotopomers.

3.1.3. Fractional 13C (10%) labelling

1. Inocculate cells a 50 ml culture of LB-rich medium with freshly transformed cells and incubate the culture at 37 (220 rpm) incubated in a shaking incubator (220Â€rpm) at 37°C until reaching an OD600 of about 2.0.

326

Renault et al.

2. Spin down the cells (1,000â•›×â•›g at room temperature) and resuspend the appropriate amount of cells in 1Â€L of labeled M9 minimal medium (containing 2Â€ g/L of [15N]-NH4Cl, 0.2Â€ g/L of [13C6]-D-glucose and 1.8Â€ g/L of unlabelled D-glucose as main nitrogen and carbon sources, respectively) to achieve an initial OD600 of 0.08. Incubate the culture at 37°C until reaching an OD600 of 0.8. 3. Induce the protein overexpression with IPTG 1Â€mM as final concentration and continue the growth for 3 to 4Â€ h at 37°C. 3.2. In Vitro Refolding from E. coli Inclusion Bodies

Overexpression in D2O leads to the production of fully deuterated proteins with 2H incorporation also at exchangeable backbone and side chain sites, resulting in severe losses of NMR signal. To ensure a complete 2H/1H exchange for amides in very stable structural motif, deuterated proteins must be unfolded and subsequently refolded in protonated medium. 1. Spin down the cells by centrifugation at 5,000â•›×â•›g for 10Â€min at 4°C, Resuspend the pellet in the buffer A (10Â€mL per g of material), and disrupt the cells by sonication on ice, until obtaining a clear and homogeneous suspension (15Â€ to 30 min, at 50Â€W using 50% duty cycle). 2. Harvest E. coli inclusion bodies by centrifugation at 10,000â•›×â•›g at 4°C for 30Â€ min) and resuspend the pellet in buffer B (10Â€ mL/g of material). To ensure a complete removal of membrane contaminants from the inclusion bodie fraction, repeat this procedure three times and collect cleaned inclusions bodies by centrifugation (10,000â•›×â•›g, at 4°C, for 30Â€min) 3. Solubilize resuspend the pellet the inclusion bodies in a denaturing buffer (buffer C, 4Â€ mL/g of material) and incubate under vigorous stirring at room temperature until obtaining a complete complete solubilization of the protein (12Â€h). 4. Refold the protein by fast dilution into 15 volumes of “Gdn/ HCl-free” refolding buffer (buffer D) and incubate under gentle stirring at ambient temperature for 16Â€ h. Remove insoluble material by centrifugation (10,000â•›×â•›g at 4°C for 30Â€min) and retain the supernatant. 5. Dialyse the supernatant (16Â€ h at 4°C) against purification buffer (buffer E). 6. Load the sample onto a Ni-NDA resin (Amersham) equilibrated in the same buffer E. Incubate under gentle stirring at ambient temperature for 2Â€h. Proteins were eluted by adding stepwise (0Â€mM, 20Â€mM, 400Â€mM, and 1Â€M) imidazole in the purification buffer (buffer E) (6Â€volumes of each eluant buffer at 1.0Â€ml/min).

Solution-State NMR Spectroscopy for Investigating 3D Structure of Membrane Proteins

327

7. Pool the pure protein fractions and dialyse against purification buffer (buffer E) for 16Â€h at 4°C to remove residual imidazole. Protein purity (typicallyâ•›>â•›95%) was checked on an SDS/ PAGE acrylamide gels and Commassie blue staining. Protein concentrations of the purified fraction were determined by OD280 measurement, using a molar extinction coefficient of 49,500/M−1·cm−1. 3.3. Detergent Screening and NMR Sample Preparation

A prerequisite for solution NMR structure determination is to find conditions allowing millimolar concentration solubility, conformational homogeneity, and high stability of the membrane protein (for days to weeks). So far, detergent micelles are the most efficient membrane mimetic for solution NMR structure determination purposes, capable of maintaining a native-fold over extended periods and retaining an overall size of MP/detergent complexes compatible with solution NMR techniques. However, NMR suitability of detergent molecules and membrane protein behaviour in different detergent remains poorly predictable. Screening the nature of detergents for proper and stable fold of the membrane protein is an essential step for structure determination by NMR spectroscopy. We introduced in the KpOmpA refolding protocol a precipitation step in cold ethanol to easily sceen several NMR buffers and detergents. Sample properties in each detergent were judged on the quality of 2D [15N, 1H]-TROSY spectrum of a 15N/2H-labeled protein 1. Precipitate 12Â€mg of isotope labelled protein by adding successively 0.15Â€volume of 4Â€M NaCl and 20Â€volumes of cold ethanol (−20°C) into an aliquot of the purified fraction KpOmpA/ZW 3-14 complexes. This procedure assumes the use of pure fractions of KpOmpA/ZW 3-14 complexes in 25Â€ mM Tris/HCl, 150Â€ mM NaCl at pHâ•›=â•›8.5 (purification buffer), at a final protein concentration of about 15Â€mg/ml. Vortex and leave the mixture at−20°C for 4–16Â€h. 2. Harvest the protein precipitate by centrifugation (15,000â•›×â•›g at 4°C for 10Â€min). Resuspend the pellet in 2Â€ml of deionized water and homogenize the suspension by vigorous stirring at 4°C. Spin down the protein precipitate (15,000â•›×â•›g at 4°C for 10Â€min) and remove the supernatant. To ensure a complete removal of residual salts and detergent molecules, repeat this procedure three times. 3. Resuspend the pellet in the desired NMR buffer to achieve a final protein concentration of about 1Â€mM. Incubate under gentle stirring at room temperature until complete dissolution of protein aggregates (for 4–16Â€h). If any, remove insoluble protein aggregates by centrifugation (10,000â•›×â•›g) and concentrate the soluble fraction using concentrator (Vivaspin, 10Â€kD cutoff) until obtaining a final protein concentration of 1Â€mM.

328

Renault et al.

4. Record 2D [15N, 1H]-TROSY HSQC spectra to assess the quality of each sample, in terms of spectral dispersion, information completeness (i.e. check that the number of observed peaks is close to the expected amide resonances) and stability over several weeks. Four detergents were tested: ZW 3-14, C8E4, DPC, and DHPC. Among them, ZW 3-14, DPC, and DHPC solubilized properly the ethanol precipitate (Fig.Â€1). Detailed analysis of the [15N,1H]-TROSY spectra revealed a better spectral dispersion for the DHPC sample (KpOmpA 1Â€mM, DHPC 300Â€mM, phosphate buffer 20Â€mM, pH 6.5, NaCl 100Â€mM). 3.4. Solution-State NMR Experiments for Resonance Assignments and Structure Determination

The structure elucidation of KpOmpA has involved three distinct 2H, 13C, 15N isotope labelling patterns, nine NMR samples, five NMR spectrometers at different frequencies (500, 600, 700, 800, and 900Â€ MHz) and 11 multidimensional heteronuclear experiments. For the sake of clarity, all the experimental parameters are reported in TableÂ€2 and further commented below (see Notes 2–5). All NMR data were processed with NMRPipe, using linear prediction to improve the resolution in all indirectly detected dimensions, and analysed with NMRView.

Fig.Â€1. Detergent screening for KpOmpA NMR studies. Contour plots of 2D [15N,1H]-TROSY spectra of u-[13C,15N/80%2H]-KpOmpA solubilized in different types of detergent: 300Â€mM ZW 3-14, 300Â€mM DPC, and 300Â€mM DHPC. The protein concentration was ~1Â€mM dissolved in a 20-mM phosphate buffer (pH 6.5) containing 100Â€mM NaCl, 10% D2O. The NMR spectra were recorded at 313Â€K at a 1H Larmor frequency of 700Â€MHz using identical acquisition and process parameters.

Solution-State NMR Spectroscopy for Investigating 3D Structure of Membrane Proteins 3.4.1. Sequential Assignment

329

1. Sequential assignment using typo error Ca and typo error Câ•›b connection pathways. The 3D [15N,1H]-TROSY HNCACB, which detects both intraresidual and sequential correlation peaks with 13Ca,bâ•›typo error and 13Ca,b(i−1) at 15N(i)/1HN(i) position of residue i, was recorded at a proton Larmor frequency of 900Â€MHz to maximize N–H TROSY effect and cross-peaks dispersion. The 3D [15N,1H]-TROSY HN(CO)CACB, which detects only the sequential correlation peaks 13Ca(i−1) and 13Cb(i−1) at 15N(i)/1HN(i) position of residue i, was recorded at lower field (see Note 3). This experiment complements the [15N,1H]-TROSY HNCACB and allows the distinction between 13C(i−1) and 13C(i) peaks in [15N,1H]TROSY HNCACB. The combination of intraresidual and sequential connectivities between (15N, 1HN) and (13Ca, 13 b C ) pairs provides the basis for the main chain assignment procedure (Fig.Â€2a). 2. Sequential assignment using 13C¢ connection pathways. Similarly, the 3D [15N,1H]-TROSY HNCO and 3D [15N,1H]TROSY HN(CA)CO provide the assignment of C’ of both intra and interresidue (13C(i−1)) with (15N(i)/1HN(i)) pair (Fig.Â€2c) and alleviate most of the residual ambiguities of the 13 a 13 b C / C sequential assignment procedure, notably for cluster involving glycine residues with degenerated 13Ca resonances. 3. Sequential assignment using side chain 13Caliph connection pathway. The 3D (HNCA)CC-TOCSY-(CA)-[15N,1H]-TROSY experiment provides 13C aliphatic side assignment of both sequential (13Caliph(i−1)) and intraresidue (13Caliph(i)) with (15N(i)/1HN(i)) pair. The magnetization transfer along the side chain is operated via an out and back FLOPSY-8 spin lock scheme (13C–13C isotropic mixing period of 16.4Â€ms). This experiment provided additional sequential assignment of KpOmpA by direct and unambiguous identification of each residue (Fig.Â€2b), in particular when 13Ca/13Cb were degenerated.

3.4.2. Measurement of Backbone HN–HN NOEs Using a 4D 15N,15N Separated NOESY-TROSY Experiment

For perdeuterated b-barrel membrane proteins, NOEs between amide protons are essential distance restraints for the global fold determination. Perdeuteration of the protein strongly reduces the occurrence of spin diffusion, allowing detection of amide proton NOEs between protons separated by more than 5â•›Å. Instead of the 3D 15N separated NOESY-TROSY experiments, we preferred to perform the 15N,15N-separated HMQC-NOESY-TROSY (mixing time 250Â€ms), which provided again an unambiguous assignment of HN–HN NOEs by the introduction of an additional 15 N dimension without excessive loss of sensitivity as clearly shown in Fig.Â€ 3a, b. This experiment yielded 173 backbone HN–HN NOEs, including 81 cross-stranded long-range NOEs

10.0

–

36.5

6.3

85.5

–

200

184

1,536

–

sw(F3) (ppm)

sw(F4) (ppm)

t1 max (ms)

t2 max (ms)

t3 max (ms)

t4 max (ms)

t1 (data points)

t2 (data points)

t3 (data points)

t4 (data points)

F1 carrier (ppm) 117.5

117.5

–

1,024

104

200

–

85.2

5.4

54.8

–

10.0

64.0

117.5

–

1,024

64

200

–

85.2

17.3

54.8

–

10.0

12.3

117.4

–

1,024

156

74

–

61.1

27.7

16.8

–

12.0

16.0

31.0

117.3

–

1,024

208

168

–

64.0

8.1

34.6

–

10.0

63.7

29.9

117.3

–

1,024

72

160

–

64.0

10.1

33.0

–

10.0

17.8

30.0

800

117.2

–

640

80

128

–

64.0

36.0

42.3

–

10.0

2.2

30.0

500

18.0

–

–

1,024

512

–

–

56.1

58.1

–

–

13.0

25.0

700 (*)

8.8

512

64

32

56

54.2

13.0

9.3

10.5

11.8

29.5

20.0

3.2

800

A4

64.0

30.0

800

D

sw(F2) (ppm)

30.0

700 (*)

B2

30.0

600

B1

sw(F1) (ppm)

600

A3

900

A2

H frequency (MHz)

1

A1

A1

NMR sample

A1

HNCACB HN(CO) HN(CA) HNCO (HNCA)CC(CA) (H)CC(CA) H(CCCO) CT-HSQC 4D 15N,15NCACB CO (*) NH-TOCSY NH-TOCSY NH-TOCSY (*) separated HMQC-NOESYTROSY

Parameter

Experiment

117.5

–

1,536

104

172

–

71.3

14.3

32.0

–

12.0

16.0

30.0

900

B3

3D 15N,13Cseparated HMQC-NOESYTROSY

18.0

–

1,024

104

96

–

60.7

14.4

13.2

–

9.4

16.0

16.0

900

C

3D 13C,13Cseparated HMQCNOESY-HMQC

Table 2 Heteronuclear experiments, acquisition parameters and NMR samples used for chemical shifts assignments and structure calculation of kpOmpA in DHPC micelles

330 Renault et al.

4.59

–

8

F3 carrier (ppm) 4.57

F4 carrier (ppm) –

Scans number

3/21

Total measuring 6/20 time (days/h)

4/21

1.4

16

–

4.59

174.7

2/14

1.5

16

–

4.59

174.5

6/14

1.5

8

–

4.58

42.8

5/13

2.0

16

–

4.58

18.3

5/16

1.2

32

–

4.51

1.0

18Â€h

1.5

80

–

–

4.55

9/16

1.14

4

4.58

116.9

116.1

6/18

1.5

4

–

4.57

18.0

5/15

1.1

32

–

4.57

18.0

NMR samples and corresponding replica are noted as follows: (A1–An) uniformly [2H,13C,15N]-KpOmpA/DHPC; (B1–Bn) [u-2H, 13C, 15N/Leu, Val, Ile(d1)-13CH3]-KpOmpA/ DHPC; (C) [u-2H, 13C, 15N/Leu, Val, Ile(d1)- 13CH3]-KpOmpA/DHPC-d22; (D) u-[15N, 10% 13C]-KpOmpA/DHPC-d22. In each sample, the concentration of the protein was ~1Â€mM, dissolved in 20Â€mM phosphate buffer (pH 6.5) containing 100Â€mM NaCl, 300Â€mM DHPC, 10% D2O. All NMR experiments were carried out at 313Â€K on four radiofrequency channel BRUKER spectrometers equipped with triple resonance cryoprobes (unless indicated *) and z-gradient coils

1.4

Recycle delay (s) 1.4

8

43.5

F2 carrier (ppm) 43.5

Solution-State NMR Spectroscopy for Investigating 3D Structure of Membrane Proteins 331

332

Renault et al.

Fig.Â€2. Sequential assignment of KpOmpA using backbone and aliphatic side-chain carbon connection pathways. Selected 13 C(w1)/1HN(w3) strips of the (a) 3D [15N,1H]-TROSY HNCACB, (b) 3D (HNCA)CC-TOCSY-(CA)-[15N,1H]-TROSY, and (c) 3D 15 1 [ N, H]-TROSY HN(CA)CO showing sequential assignment of Tyr111 to Arg113 protein segment, through the use of backbone and side-chain carbon chemical shifts (highlighted by horizontal and vertical dashed lines). Strips were extracted at the 15N(w2) chemical shifts indicated at the top of the each panel.

(i−jâ•›³â•›5). The identification of HN–HN cross-stranded NOEs for each pair of neighbouring b-strands established clearly the antiparallel eight-stranded b-barrel arrangement (Fig.Â€3c). 3.4.3. Selective Methyl Protonation Approach to Structure Determination

1. The first step consists in the sequence-specific assignment of carbon and proton resonances of isoleucine (d1), leucine (d1,d2), and valine (g1,g2) methyl groups. 13C and 1H chemical shifts were obtained from two separated 3D experiments recorded on a u-[2H,13C,15N]/Ile(d1),Leu,Val-[13CH3]labelled KpOmpA/DHPC sample. The 3D (H)C(CC)TOCSY-(CA)-[15N,1H]-TROSY provides methyl 13C chemical shifts from both intraresidual and sequential methyl carbon to amide correlations (Fig.Â€ 4a). Proton chemical shifts can be independently obtained from sequential methyl proton to backbone amide correlations identified in the 3D H(CCCO) NH-TOCSY spectrum (Fig.Â€ 4b). Combined analysis of the two spectra resulted in a nearly complete sequence-specific assignment of proton and carbon resonances of isoleucine (d1), leucine (d1,d2), and valine (g1,g2) methyl groups of kpOmpA.

Solution-State NMR Spectroscopy for Investigating 3D Structure of Membrane Proteins

333

Fig.Â€ 3. Measurement of unambiguous backbone amide NOEs. (a) 2D 15N(w2)/1HN(w3) plane extracted from the 3D 15N-separated NOESY-TROSY at w1(1H)â•›=â•›9.01Â€ppm corresponding to the proton amide resonance of T159 (coloured in grey). (b) Corresponding 2D 15N(w3)/1HN(w4) plane extracted from the 4D 15N,15N-separated HMQC-NOESY-TROSY spectrum at w1(1H)â•›=â•›9.01Â€ ppm and w2(15N)â•›=â•›114.4Â€ ppm at the proton and nitrogen amide frequencies of T159 (coloured in grey) providing unambiguous assignment of sequential and long-range HN–HN NOE correlations. The most intense correlation that identifies cross-stranded HN–HN NOE (underlined residues) establishes an anti-parallel orientation of sequential b-strands. (c) Topological representation of KpOmpA. Crossstranded HN–HN NOEs are represented with transversal dashed lines. Loops (L) and turns (T) are numbered from N to C-terminus. Residues located in b-strands are shown in squares, those that are partially assigned are coloured in light grey and residues that remain unassigned are coloured in dark grey.

Only methyl resonances of residues whose intraresidual and sequential 1H/15N backbone amide resonances are broadened beyond detection and remain unassigned, due to molecular motions at ms time scale.

334

Renault et al.

Fig.Â€4. Sequence specific and stereospecific assignment of 13C and 1H methyl resonances: the example of Leucine 66. (a) Section of 13C(w1)/1HN(w3) slices from 3D (H)C(CC)-TOCSY-(CA)-[15N,1H]-TROSY spectrum corresponding to intraresidual correlation of L66. (b) Section of 1H(w1)/1HN(w3) slices from 3D H(CCCO)NH-TOCSY spectrum corresponding to sequential correlation of L66. (c) Expanded region of 2D [13C,1H]-CT-HSQC corresponding to proR and proS methyl groups of L66 (positive and negative contour plots are coloured in black and grey, respectively).

2. Stereospecific assignment of 1H and 13C methyl resonances of valine (g1,g2) and leucine (d1,d2) prochiral methyl groups is a prerequisite for straightforward analysis of methyl-derived NOE restraints. They can be obtained using a u-[15N, 10% 13 C]-labelled KpOmpA/DHPCd22 sample (see Note 6). Under these circumstances, the pro-R and pro-S methyl groups became reliably distinguishable on a 2D [13C,1H]- constant time HSQC spectrum by the presence or the absence of 1 JCC coupling, respectively. With the use of a constant time period of 26.6Â€ms (1/Jcc), 13C/1H correlations derived from 13 C methyl directly bound to a 13C partner (pro-R) give opposite phase relative to cross-peaks from isolated 13C spins (pro-S) (Fig.Â€ 4c). Stereospecific assignment of methyl resonances is directly obtained by examination of the resulting spectrum. Only the isopropyl methyl groups showing a complete degeneracy of both carbon and proton frequencies remained indistinguishable. 3. Methyl-to-backbone-amide NOE (Me–HN NOE) and methylto-methyl NOE (Me–Me NOE) measurements are usually achieved using standard 3D 15N-separated and 13C-separated [1H,1H]-NOESY, respectively. However, the strong degeneracy of methyl proton frequencies seriously hampered the

Solution-State NMR Spectroscopy for Investigating 3D Structure of Membrane Proteins

335

assignment of such NOEs (Fig.Â€ 5a). For an unambiguous identification of Me–HN and Me–Me NOE correlations, information about 13C chemical shifts should be preferred since the latter offers higher chemical shift dispersion as compared with proton. The 3D 15N, 13C-separated HSQC-NOESY-TROSY experiment provided a non-ambiguous assignment of each Me–HN NOE correlations (243 NOEs) as shown in Fig.Â€5b. 4. Similarly, the Me–Me NOE measurement was achieved using a 3D 13C, 13C-separated HMQC-NOESY-HMQC instead of standard 13C-separated [1H,1H]-NOESY. In addition to Me–HN NOEs, this experiment provided a large number of NOE contacts between proximal methyl group, and hence abundant distance restraints homogeneously spread over the hydrophobic core of KpOmpA (Fig.Â€6). 3.5. Structure Calculation

Structural calculations for KpOmpA were carried out using HN–HN NOEs, Me–HN NOEs, Me–Me NOEs, dihedral angle restraints, and hydrogen bond distances restraints. 1. Assigned NOEs’ peaks were integrated individually and converted to distances using NOE analysis module of NMRView. For KpOmpA structure calculation, the distances were separated

Fig.Â€5. Profit of the 3D 15N, 13C-separated HMQC-NOESY-TROSY in the assignment of methyl NOEs. Strips extracted form 3D 15N-separated HMQC-NOESY-TROSY (a) and 3D 15 N, 13C-separated HMQC-NOESY-TROSY (b ) at proton and nitrogen amide resonances of Y102.

336

Renault et al.

Fig.Â€6. 3D structure calculation of KpOmpA. (Left ) Cartography of Me–HN and Me-Me NOEs-derived distance restraints involving methyl groups of Ile, Leu, and Val residues (denoted by dashed lines) onto the NMR structure of KpOmpA represented by the superimposition of the 20 lowest-energy CNS conformers in the b-barrel region (Right ). Ribbon representation of the average structure of KpOmpA derived from the 20 lowestenergy CNS conformers with a stick representation of the two aromatic girdles located at membrane interfaces. Structures were generated with PyMOL software.

into three classes, as having strong, medium, or weak NOEs, with upper limits of 3.4â•›Å, 4.7â•›Å, 8.0â•›Å for HN–HN NOEs, and 4.5â•›Å, 7.0â•›Å, 10.0â•›Å for both Me–HN and Me–Me NOEs. 2. Hydrogen bond distance restraints derived from the identification of slowly exchanging amide protons and from characteristics HN–HN NOEs. Experimentally, validated hydrogen bonds were implemented by means of two distance restraints with upper limits of 2.4â•›Å for O–HN and 3.2â•›Å for O–N. 3. Backbone f and j dihedral angle restraints were obtained by using the TALOS program, using chemical shift assignments corrected from shifts due to 2H isotopic and TROSY effect in NMRview. 4. Solution NMR structures of KpOmpA were calculated using CNS (v.1.1) simulated annealing protocols with 2,000 steps for the high temperature annealing stage, 6,000 steps for the first cooling stage using torsion angle dynamics, and 5,000 steps for the second cooling stage using Cartesian dynamics. An ensemble of 200 structures was generated, and the 20 lowest overall violation-energy conformers were selected to represent the 3D structure of KpOmpA in DHPC micelles.

Solution-State NMR Spectroscopy for Investigating 3D Structure of Membrane Proteins

337

5. The structural quality of the ensemble was analysed using PROCHECK, which gave 70.5% of residues in most favoured regions, 26.7% in additional allowed regions, 2.6% in generously allowed regions, and 0.2% in disallowed regions.

4. Notes 1. Since KpOmpA is expressed as intracellular inclusion bodies, the induction time (i.e. 30Â€ h) was optimized in order to enhance the protein expression yields. 2. For standard out-and-back type triple resonance experiments used for backbone and side chain resonance assignments, maximum sensitivity is achieved with 100% deuteration. In these experiments, magnetization starts and finishes on the amide proton, which relaxes more slowly and the attenuation due to 13Ca transverse relaxation is more favourable when the protein is perdeuterated. 3. Because of large chemical shift anisotropy of carbonyl carbon and resulting fast transverse relaxation of 13C¢ nuclei at high fields, the 3D [15N,1H]-TROSY HN(CO)CACB, HNCO, and HN(CA)CO experiments should be performed at a 1H frequency equal or less than 700Â€ MHz. All other NMR experiments should be preformed at the highest magnetic field available in order to expect the better sensitivity and resolution. 4. Standard 3D multidimensional NMR experiments, especially those in TROSY detection scheme, are usually written with a 15 N evolution in constant time mode in order to minimize the relaxation contribution. The evolution in a constant time manner limits the number of time domain data points in the 15 N dimension and hence the final spectral resolution (i.e. FID resolution about 30Â€Hz/point), which is a crucial parameter for the structure determination of macromolecule with high molecular weight. Therefore, each 3D multidimensional experiment described above was performed with a 15N evolution in semi-constant time manner in order to record 200 time domain data points in 15N dimension (FID resolution about 14Â€Hz/point). 5. Since the sample contains 300Â€ mM of DHPC, echo/antiecho gradient field pulses for the 15N coherence selection should be adjusted carefully to remove efficiently the 1H signals of DHPC, particularly the intense signal arising from the choline methyl groups. 6. Samples for the 3D 13C,13C-separated HMQC-NOESY-HMQC and 2D [13C,1H]- constant time HSQC were prepared with

338

Renault et al.

deuterated DHPC (i.e. DHPC-d22) to prevent undesirable 1H signal from the methyl groups of the detergent acyl chains.

Acknowledgements These studies and the IPBS NMR equipment were financed by the French research ministry, CNRS, Université Paul Sabatier, the Région Midi-Pyrénées and European structural funds. Extended access to the EU-NMR facility in Frankfurt 6th Framework Program of the EC (contract number RII3-026145) has a crucial role in the development of this project and is duly acknowledged. We thank, in particular, Dr C. Richter, and Professor H. Schwalbe. The CIPF (P. Fabre) is acknowledged for providing the pVALP40 plasmid of KpOmpA. References 1. Tamm LK, Liang BY (2006) NMR of membrane proteins in solution. Prog Nucl Magn Reson Spectrosc 48:201–210 2. Pervushin K, Riek R, Wider G, Wuthrich K (1997) Attenuated T2 relaxation by mutual cancellation of dipole-dipole coupling and chemical shift anisotropy indicates an avenue to NMR structures of very large biological macromolecules in solution. Proc Natl Acad Sci U S A 94:12366–12371 3. Pervushin K (2000) Impact of transverse relaxation optimized soectroscopy (TROCSY) on NMR as a technique in structural biology. Q Rev Biophys 33:161–197 4. Arora A, Abildgaard F, Bushweller JH, Tamm LK (2001) Structure of outer membrane protein A transmembrane domain by NMR spectroscopy. Nat Struct Biol 8:334–338 5. Fernandez C, Adeishvili K, Wuthrich K (2001) Transverse relaxation-optimized NMR spectroscopy with the outer membrane protein OmpX in dihexanoyl phosphatidylcholine micelles. Proc Natl Acad Sci U S A 98: 2358–2363 6. Hwang PM, Choy WY, Lo EI, Chen L, Forman-Kay JD, Raetz CRH, Prive GG, Bishop RE, Kay LE (2002) Solution structure of the outer membrane enzyme PagP by NMR. Proc Natl Acad Sci U S A 99: 13560–13565 7. Johansson MU, Alioth S, Hu KF, Walser R, Koebnik R, Pervushin K (2007) A minimal

8.

9.

10. 11.

12.

13.

14.

transmembrane beta-barrel platform protein studied by nuclear magnetic resonance. Biochemistry 46:1128–1140 Liang BY, Tamm LK (2007) Structure of outer membrane protein G by solution NMR spectroscopy. Proc Natl Acad Sci U S A 104:16140–16145 Tjandra N, Bax A (1997) Direct measurement of distances and angles in biomolecules by NMR in a dilute liquid crystalline medium. Science 278:1111–1114 Prestegard JH, Bougault CM, Kishore AI (2004) Chem Rev 104:3519–3540 Battiste JL, Wagner G (2000) Utilization of siet-directed spin labelling and high resolution heteronuclear NMR for global fold determination of large proteins with limited nuclear overhauser effect data. Biochemistry 39:5355–5365 Rosen MK, Gardner KH, Willis RC, Parris WE, Pawson T, Kay LE (1996) Selective methyl group protonation of perdeuterated proteins. J Mol Biol 263:627–636 Goto NK, Gardner KH, Mueller GA, Willis RC, Kay LE (1999) A robust and cost-effective method for the production of Val, Leu, Ile (delta 1) methyl-protonated 15N-, 13C-, 2H-labeled proteins. J Biomol NMR 13: 369–374 Fernandez C, Hilty C, Wider G, Guntert P, Wuthrich K (2004) NMR structure of the integral membrane protein OmpX. J Mol Biol 336:1211–1221

Solution-State NMR Spectroscopy for Investigating 3D Structure of Membrane Proteins 15. Roosild TP, Greenwald J, Vega M, Castronovo S, Riek R, Choe S (2005) NMR structure of mistic, a membrane-integrating protein for membrane protein expression. Science 307: 1317–1321 16. Liang BY, Bushweller JH, Tamm LK (2006) Site-directed parallel spin-labelling and paramagnetic relaxation enhancement in structure determination of membrane proteins by solution NMR spectroscopy. J Am Chem Soc 128:4389–4397 17. Cierpicki T, Liang BY, Tamm LK, Bushweller JH (2006) Increasing the accuracy of solution NMR structures of membrane proteins by application of residual dipolar couplings. High-resolution structure of outer membrane protein A. J Am Chem Soc 128:6947–6951 18. Hiller S, Garces RG, Malia TJ, Orekhov VY, Colombini M, Wagner G (2008) Solution structure of the integral human membrane VDAC-1 in detergent micelles. Science 321:1206–1210 19. Teriete P, Franzin CM, Choi J, Marassi FM (2007) Structure of the Na, K-ATPase regulatory protein FXYD1 in micelles. Biochemistry 46:6774–6783

339

20. Shih SC, Stoica I, Goto NK (2008) Investigation of the utility of selective methyl protonation for determination of membrane protein structures. J Biomol NMR 42:49–58 21. Renault M, Saurel O, Czaplicki J, Demange P, Gervais V, Löhr F, Réat V, Piotto M, Milon A (2009) Solution state NMR structure and dynamics of KpOmpA, a 210 residue transmembrane domain possessing a high potential for immunology applications. J Mol Biol 385:117–130 22. Lohr F, Ruterjans H (2002) Correlation of backbone amide and side-chain (13)C resonances in perdeuterated proteins. J Magn Reson 156:10–18 23. Xia YL, Sze KH, Zhu G (2000) Transverse relaxation optimized 3D and 4D 15N/15N separated NOESY experiments of 15N labeled proteins. J Biomol NMR 18:261–268 24. Haeuw JF, Rauly I, Zanna L, Libon C, Andreoni C, Nguyen TN, Baussant T, Bonnefoy JY, Beck A (1998) The recombinant Klebsiella pneumoniae outer membrane protein OmpA has carrier properties for conjugated antigenic peptides. Eur J Biochem 255:446–454

as

Chapter 18 NMR Spectroscopy of Lipid Bilayers Axelle Grélard, Cécile Loudet, Anna Diller, and Erick J. Dufourc Abstract Knowledge of lipid structure and dynamics in a membranous environment is of first importance for deciphering cellular function. Sterols and sphingolipids are key molecules in maintaining membrane integrity and are the building blocks of membrane domains, such as “rafts”. Phosphatidyl inositols are crucial in signalling pathways as they are recognition sites at the membrane surface. Other lipids such as Phosphatidylethanolamines, Cardiolipins, or diacylglycerols are essential in fusion processes. It is fundamental to have techniques that can resolve the structure and dynamics of various classes of lipids in a membrane environment. Solid state NMR with its high resolution and wide line facets is a very powerful tool for such determinations. Here it is shown that multinuclear solid state NMR provides information on the nature of the membrane phase (bicelle, lamellar, hexagonal, micelle, cubic, etc.), its dynamics (fluid or gel, or liquid-ordered with cholesterol), and the molecular structure of embedded lipids when using the magic angle sample spinning (MAS) apparatus. Typical examples of relatively simple experiments are shown both with high resolution MAS and wide line NMR of lipids. Relaxation time measurements are also described to measure lipid motional processes from the picosecond to the second timescale. Key words: Solid state 1H–2H–14N–31P-NMR, Bicelles, Liposomes, Micelles, Gel and fluid phases, Hexagonal phases, Liquid-ordered state, Cholesterol, Sphingolipid, Diacylphosphatidylcholine, Diacylphophatidylethanolamine, Deuterium-labelled lipids, Magic angle sample spinning, Wide line spectra, Relaxation times, Order parameters

1. Introduction Lipids are one of the building blocks in cell biology and are present in many cellular locations. They may be found as small aggregates allowing their transport in water media or may be embedded in cellular membranes where their dynamics are much more reduced. In this case, they constitute the cement of the bilayer membrane and also play many signalling roles. For instance, cholesterol and sphingolipids are key molecules in maintaining membrane integrity and are the building blocks of membrane liquid-ordered phases, Jean-Jacques Lacapère (ed.), Membrane Protein Structure Determination: Methods and Protocols, Methods in Molecular Biology, vol. 654, DOI 10.1007/978-1-60761-762-4_18, © Springer Science+Business Media, LLC 2010

341

342

Grélard et al.

also named “rafts” (1, 2). Lipids such as phosphatidylinositol (PtdIns) are essential in signalling pathways as they are recognition sites at the membrane surface (3). In order to decipher their structure in their native membrane environment, Nuclear Magnetic Resonance is one of the most powerful techniques because it can act in media of reduced dynamics such as membranes or aggregates. NMR relies on the presence of active nuclei in atoms that constitute the lipid molecules. A lipid naturally contains protons (1H), carbon (only the 13C is active, but in low natural abundance), phosphorus (31P), oxygen (only the 17O is magnetically active but has a very low natural abundance) and nitrogen (14N). Some lipids may be chemically labelled with the hydrogen isotope, deuterium (2H), with fluorine (19F), carbon-13 (13C), or with other nuclei of interest for structural biology. Lipids are not water soluble; as a consequence, their structure and dynamics must be evaluated in a hydrated model membrane state (bicelles, liposomes, bilayers vesicles of various sizes, etc.) (4) or in real membranes (envelopes of living cells) (5).

2. Materials 2.1. Samples, Preparation Tools and Measuring Cells

1. Ninety percent ultra pure water (miliQ-system)â•›+â•›10% deuterated water (2H2O), v:v (Eurisotop, Saint-Aubin, France). 2. Deuterium-depleted water, 1H2O (Isotec-Sigma-Aldrich, France). 3. Lipids: 1,2-Dipalmitoyl-2H62-sn-glycero-3-phosphocholine (16:0/16:0 DPPtdCho-2H62), 1,2-dimyristoyl-sn-glycero-3phosphocholine (14:0/14:0 DMPtdCho), 1,2-dicaproyl-snglycero-3-phosphocholine (6:0/6:0 DCPtdCho), Liver Phosphatidylethanolamine (PtdEth), Liver Phosphatidylcholine (PtdCho), brain Sphingomyelin (SM) and cholesterol are from Avanti Polar Lipids (USA). 1-Tetradecanoyl-2-(4-(4biphenyl)butanoyl)-sn-glycero-3-phosphocholine (14:0/BB TBBPtdCho) and 1,2-dimyristoyl-sn-glycero-3-phosphocholine 2H72 (14:0/14:0 DMPtdCho-2H72) were synthesized in the laboratory (6, 7). 4. Table top centrifuge Eppendorf (VWR, France). 5. Two millilitre micro centrifuge tubes (VWR, France). 6. MS2 Minishaker (VWR, France). 7. Water bath, thermostated at 50°C. 8. Liquid nitrogen, 50Â€ml in a 500Â€ml Dewar. 9. Zirconia (ZrO2) 4Â€mm diameter magic angle spinning (MAS) rotor of HR-MAS type (50Â€ml) and of CP-MAS type (100Â€ml) (Cortec, France).

NMR Spectroscopy of Lipid Bilayers 2.2. NMR Instrumentation for 31P-NMR

343

1. NMR Spectrometer equipped for solid state: Bruker Avance 300 operating at 121.5Â€MHz (Bruker, Wissembourg, France). 2. NMR Probe: 1H/X, 4Â€mm CP-MAS probe tuned at 121.5Â€MHz (31P channel) and 300.13Â€MHz (1H channel); equipped with temperature regulation (thermocouple and heater). 3. PC station with TopSpin software v2.0 (Bruker, Wissembourg, France). 4. Bruker BCU (air cooling) and VT (variable temperature) units to control sample temperature using dry airflow.

2.3. NMR Instrumentation for 2H-NMR

1. NMR Spectrometer equipped for solid state: Bruker Avance 500 operating at 76.8Â€MHz (Bruker, Wissembourg, France). 2. NMR Probe: Low Gamma 1H/X, 4Â€ mm CP-MAS probe tuned at 76.8Â€ MHz; equipped with temperature regulation (thermocouple and heater). 3. PC station with TopSpin software v2.1 (Bruker, Wissembourg, France). 4. Bruker BCU (air cooling) and VT (variable temperature) units to control sample temperature using dry airflow.

2.4. NMR Instrumentation for 14N-NMR

1. NMR Spectrometer equipped for solid state: Bruker Avance 500 operating at 36.1Â€MHz (Bruker, Wissembourg, France). 2. NMR Probe: PE triple 1H–X–Y CP probe tuned at 36.1Â€MHz on X channel; 50Â€W on the Y and 1H channels; equipped with temperature regulation (thermocouple and heater). Homemade 4Â€mm horizontal coil. 3. PC station with TopSpin software v2.1 (Bruker, Wissembourg, France). 4. Bruker BCU (air cooling) and VT (variable temperature) units to control sample temperature using dry airflow.

2.5. NMR Instrumentation for HR-MAS 1H-NMR

1. NMR Spectrometer equipped for solid state: Bruker Avance 500 operating at 500.16Â€ MHz (Bruker, Wissembourg, France). 2. NMR Probe: 1H/13C, 2H-lock, z-gradient, 4Â€mm HR-MAS probe tuned at 500.16Â€ MHz, equipped with temperature regulation (thermocouple and heater). 3. PC station with TopSpin software v2.1 (Bruker, Wissembourg, France). 4. Bruker BCU (air cooling) and VT (variable temperature) units to control temperature on sample using dry airflow. 5. Bruker pneumatic MAS unit to control sample rotation at the magic angle.

344

Grélard et al.

2.6. Data Analysis

1. PC computer loaded with Topspin 2.1 software (Bruker, Wissembourg, France) and Origin software V7.5 (OriginLab Corporation, Massachusetts, USA). 2. NMR Friend V1.2 plug-in software to import and treat NMR spectra (see Note 1) on Origin software developed by Sébastien Buchoux, UMR5248 CNRS-University Bordeaux 1, France (see Note 2).

3. Methods The membrane that defines the cell entity or the cell organelles is by its very nature a medium that is half way between a liquid and a solid. This state called soft matter is by definition a liquid crystalline medium whose anisotropic properties are essential for membrane function and cells. Molecules there embedded, such as lipids, proteins, drugs may undergo many dynamic processes such as lateral diffusion in the bilayer plane, rotational diffusion around the bilayer normal or transverse diffusion from one membrane leaflet to the other. They may also group as ordered patches in the membrane plane that are named “rafts” to picture the rigidity of these membrane domains that swim in a “sea” of more fluid lipids and proteins. Understanding structure and dynamics of membrane components will provide insight towards deciphering complex biological processes such as cell fusion, trafficking, apoptosis, energetic, signal transduction, etc. Solid state NMR is the only non-destructive, non-invasive and quantitative spectroscopy that can measure membrane structure and dynamics. There are two data categories that can be obtained using NMR, spectra and relaxation times. Spectra are in turn of two types “wide line” and “high resolution”. Wide line NMR (spectra may span several hundredths of kilohertz) can be sensitive to membrane macroscopic orientation and symmetry (Fig.Â€ 1) and molecular motions (Fig.Â€ 2). It is a useful tool to determine the nature of membrane phases (bicelles, lamellar, hexagonal, isotropic (micelles or cubic), etc.) (8, 9) and track membrane dynamics (membrane fluidity, fusion, gel (solidordered), liquid-ordered, and fluid (liquid-disordered) states). The well-known regulating effect of cholesterol on membrane phases (increasing fluidity of solid-ordered phases and decreasing that of liquid-disordered phases) can easily be monitored by using phospholipids or sterols that are deuterated (2, 10–12). Analysis of spectral moments or quadrupolar splittings allows describing membrane dynamics either globally (bilayer/molecule fluctuations) or locally by measuring the reorientation of individual chain segments in the bilayer core (Fig.Â€ 2). By taking advantage of

NMR Spectroscopy of Lipid Bilayers

345

Fig.Â€1. Nitrogen-14 and Phosphorus-31 solid state NMR spectra of lipids embedded in different membrane phases. Top row: spectra obtained for a mixture of large micrometric liposomes and nanometric micelles made of phosphatidylcholine lipids. Middle-top row: spectra of DMPtdCho/DCPtdCho bicelles in 80% D2O oriented such that the normal to the disc (400Â€nm) plane is perpendicular to the magnetic field direction, B0. Middle-bottom row: spectra of TBBPtdCho/DCPtdCho bicelles (biphenyl bicelles) in 80% D2O oriented such that the normal to the disc (800Â€nm) plane is parallel to the magnetic field direction, B0. Bottom row: spectra of phopholipid micelles (nm size). On the right hand side are represented cartoons to picture the nature of the lipid systems. L, I, B and B//, B0, and n stand respectively for lamellar, isotropic, bicelle normal perpendicular and parallel, magnetic field, and bilayer normal.

Â� magnetic field orientational dependence, wide line NMR spectra can also probe average orientations of molecules embedded in membranes (membrane topology) (10, 11, 13) when used in conjunction with X-rays or neutrons structural information. The 3D structure of molecules in membranes is also obtained by making use of MAS, a technique by which the sample is rapidly spun at an angle of 54.7° with respect to the magnetic field (Fig.Â€4d) leading to pseudo “high-resolution” spectra (sharp lines of less than 1Â€Hz width), as in the liquid state, the sample being still in the membranous “liquid-crystalline” state (Fig.Â€4). All multidimensional NMR techniques used in solution NMR can hence be applied bringing assignment of resonances to molecular structure (TableÂ€1, Fig.Â€4). Relaxation times (T1 and T2) are obtained by making use of specific pulse sequences (see Note 10) and bring

346

Grélard et al.

a

c +30% Chol (lo)

Gel (so)

b

d

+30% Chol (lo)

Fluid(ld)

60

40

20 2H

0

−20 −40

−60 −80 80

40

20 2H

Frequency (kHz)

e First spectral moment (kHz)

60

0

−20 −40

−60 −80

Frequency (kHz)

f

62.5

so

140

lo

120

lo

100

50.0

+30% Chol +30% Chol

80

37.5 25.0

ld

60

12.5

ld 40

Quadrupolar splitting (kHz)

80

0.0 0

10

20

30

40

Temperature (°C)

50

60

2

4

6

8

10

12

14

16

Labelled carbon position on lipid acyl chain

Fig.Â€2. Solid-state wide line 2H-NMR spectra of DPPtdCho-2H62 with and without cholesterol dispersed in water (liposomes). (a) DPPtdCho, Tâ•›=â•›10°C, gel (so) phase; (b) DPPtdCho, Tâ•›=â•›40°C, fluid (ld) phase; (c) DPPtdCho/cholesterol (2/1 molar), Tâ•›=â•›10°C, lo phase; (d) DPPtdCho/cholesterol (2/1 molar), Tâ•›=â•›40°C, lo phase. (e) First spectral moment as a function of temperature for DPPtdCho (filled symbol) and DPPtdCho/cholesterol (empty symbol) spectra. (f) Quadrupolar splittings for DPPtdCho acyl chains at 40°C in the presence (empty symbols) and absence (filled symbols) of cholesterol.

values ranging from microseconds to minutes (Fig.Â€5). Analysis of relaxation times in the frame of proper motional modes leads to the calculation of speed of molecular motion and activation energies. This allows describing membrane dynamics from the atomic level where intra-molecular motions dominate (nano-to-Â� picoseconds timescale), to the cell level where membrane hydrodynamic modes of motion play an important role (seconds time scale) (14–16).

NMR Spectroscopy of Lipid Bilayers

3.1. Determination of Lipid Phase Nature (Lamellar (Liposomes), Oriented-Bicelle, Isotropic) Using Wide Line 31P-NMR

347

1. Appropriate amounts of phospholipids are weighed (see Note 11), and a suitable volume of water is added to obtain a lipid hydration (water mass/total mass) 80–95% (w/w). Lipid concentration is of the order 10–200Â€mM. For bicelle samples, 14:0/BB TBBPtdCho and 14:0/14:0 DMPtdCho represent about 80Â€mol% of the total lipid, 6:0/6:0 DCPtdCho the remaining 20%. 2. The hydrated sample is then vigorously shaken in a vortex mixer, frozen in liquid nitrogen, and heated to 50°C for 10Â€min in a water bath. This cycle of liposome (multilamellar vesicles of micrometre size) or bicelle (bilayer discs of 400–800Â€nm diameters) formation is repeated three to five times until a homogeneous preparation is obtained at room temperature. 3. The resulting sample (may appear translucent when in bicelle phase) is transferred (see Note 12) into a 4-mm NMR ZrO2 rotor (100Â€ml) that is placed in the magnetic field. The measuring probe is properly tuned to 121.5Â€MHz (31P channel) and 300.13Â€MHz (1H channel) and the space homogeneity (“shimming”) is adjusted at best. 4. Using the variable temperature set up (VT and BCU units), samples are allowed to equilibrate 15–30Â€min at a given temperature before the time dependent NMR signal (FID) is acquired; the temperature is regulated to ±1°C (see Note 9). 5. 31P-NMR spectroscopy is performed at 121.5Â€MHz. FID are acquired using a phase-cycled Hahn-echo pulse sequence with gated broad-band proton decoupling (17). Typical acquisition parameters are as follows: p/2 pulse width (P1) of 14.5Â€µs (see Notes 3 and 10) spectral window (SW) of 32Â€kHz, interpulse delay (D6) of 50Â€µs and a repetition delay (D1) of 5Â€s (see Note 4). Typically, 512 scans are recorded in 42Â€ min (see Note 5). 6. Processing parameters: noise filtering of the FID with an exponential window (see Note 7), characterized by the Lorentzian line-broadening factor (LBâ•›=â•›10–50Â€Hz, see Note 6), is used. Fourier Transformation is applied to get the spectrum, and its phase is adjusted to correct for unavoidable time delays in acquisition. Reference is made to 85% H3PO4 (0Â€ ppm). The width of the spectrum is measured and used to calculate the chemical shift anisotropy (CSA). 7. The spectrum (Fig.Â€1, 31P column, top) is characteristic of a non-oriented spectrum (broad and major axially symmetric pattern of 45Â€ppm CSA) superimposed on a small isotropic line (0Â€ppm). This is typical of a mixture of micrometre liposomes and nanometre micelles (sketches on the right).

348

Grélard et al.

TableÂ€1 1 H chemical shifts of liposomal lipids in H2O/D2O (90/10). Ambient temperature, reference TSP (0Â€ppm). The sample is spun at 7Â€kHz at the magic angle

Head group

Backbone

Fatty acyl chains

N

Sphingomyelin

Liver PtdEth

Liver PtdCho

a

nd

3.97

3.99

b

nd

3.22

3.66

g

3.66

7.87

3.21

g1

nd

nd

nd

g2

5.31

5.29

5.31

g3

4.00

4.03

4.27

1

5.71

1′ 2

5.71

2.40

2.28

2′

2.22/2.31

nd

2.28

3

2.02

1.57

1.57

3′

1.60

nd

1.57

4

1.29

1.24

1.25

4′

1.29

2.00

1.25

5

1.29

1.24

1.25

5′

1.29

nd

1.25

6

1.29

1.24

1.25

6′

1.29

nd

1.25

7

1.29

1.24

1.25

7′

1.29

2.75

1.25

8

1.29

1.24

1.25

8′

1.29

nd

2.02

9

1.29

1.24

1.25

9′

1.29

nd

nd

10

1.29

1.24

1.25

10′

1.29

2.75

nd

11

1.29

1.24

1.25

11′

1.29

nd

2.77

12

1.29

1.24

1.25

(continued)

NMR Spectroscopy of Lipid Bilayers

349

Table 1 (continued) N

Sphingomyelin

Liver PtdEth

Liver PtdCho

12′

1.29

nd

nd

13

1.29

1.24

1.25

13′

1.29

2.75

nd

14

1.29

1.24

1.25

14′

1.29

nd

2.02

15

0.88

1.24

1.25

15′

1.29

nd

1.25

1.24

1.25

2.00

1.25

1.24

1.25

1.24

1.25

0.85

0.86

1.24

0.86

16 16′

1.29

17 17′

1.29

18 18′

0.88

19 19′

1.24

20′

0.85

nd not determined. Numbers separated by “/” (e.g. 2.22/2.31) indicate that there are two lines of indicated chemical shifts

8. The spectrum (Fig.Â€1, 31P column, middle-top) is much narrower and made of two sharp lines representing 14:0/14:0 DMPtdCho (major, −12Â€ ppm) and 6:0/6:0 DCPtdCho (minor, −4Â€ppm) in a bicelle (disc) structure with the bilayer normal oriented at 90° with respect to the magnetic field direction (sketches on the right). 9. The spectrum (Fig.Â€1, 31P column, middle-bottom) is also made of two sharp lines representing 14:0/BB TBBPtdCho (major, 20Â€ppm) and 6:0/6:0 DCPtdCho (minor, 8Â€ppm) in a bicelle (disc) structure with the bilayer normal oriented at 0° with respect to the magnetic field direction (sketch on the right). 10. The spectrum (Fig.Â€1, 31P column, bottom) is characteristic of a phase with isotropic symmetry (cubic, micelles, etc.): a single, very intense sharp line appears and is centred at ca. 0Â€ppm.

350

Grélard et al.

3.2. Determination of Lipid Phase Nature (Lamellar (Liposomes), OrientedBicelle, Isotropic) Using Wide Line 14 N-NMR

1. Step 1 as in SubheadingÂ€3.1. 2. Step 2 as in SubheadingÂ€3.1. 3. The resulting sample (may appear translucent when in bicelle phase) is transferred into a 4-mm NMR ZrO2 rotor (100Â€ml) that is placed in the magnetic field. The measuring probe is properly tuned to 36.1Â€MHz (X-channel). 4. Step 4 as in SubheadingÂ€3.1. 5. 14N-NMR spectroscopy is performed at 36.1Â€ MHz. NMR spectra are acquired using a quadrupolar echo pulse sequence (18). Typical acquisition parameters are as follows: spectral window of 100Â€kHz; p/2 pulse width 10Â€µs (see Note 13), and interpulse delay of 200Â€µs. A recycle delay of 0.2Â€s is used. Typically, 40,000 scans are accumulated in 2.2Â€h. 6. Processing parameters: noise filtering of the FID with an exponential window, characterized by the Lorentzian line-broadening factor (LBâ•›=â•›100Â€Hz), is used. Fourier Transformation is applied to get the spectrum, and its phase is adjusted to correct for unavoidable time delays in acquisition. The spectrum centre is set to 0Â€ppm. 7. The spectrum (Fig.Â€1, 14N column, top) is characteristic of a non-oriented spectrum (broad axially symmetric powder pattern of ca. 9Â€kHz) superimposed on a isotropic line (0Â€Hz). This is typical of a mixture of micrometre liposomes and nanometre micelles (sketches on the right). 8. The spectrum (Fig.Â€ 1, 14N column, middle-top) is made of two sets of sharp doublets representing 14:0/14:0 DMPtdCho (major, ca. 7Â€kHz splitting) and 6:0/6:0 DCPtdCho (minor, ca. 1.5Â€ kHz splitting) in a bicelle (disc) structure with the bilayer normal oriented at 90° with respect to the magnetic field direction (sketches on the right). 9. The spectrum (Fig.Â€ 1, 14N column, middle-bottom) is also made of two sets of sharp doublets representing 14:0/BB TBBPtdCho (major, ca. 13Â€ kHz Splitting) and 6:0/6:0 DCPtdCho (minor, ca. 3Â€ kHz splitting) in a bicelle (disc) structure with the bilayer normal oriented at 0° with respect to the magnetic field direction (sketch on the right). 10. The spectrum (Fig.Â€1, 14N column, bottom) is characteristic of a phase with isotropic symmetry (cubic, micelles, etc.): a single, very intense sharp line appears and is centred at ca. 0Â€Hz.

3.3. Determination of Lipid Dynamics in Non-oriented Multilamellar Bilayers (Gel, Fluid, LiquidOrdered) Using Wide Line 2H-NMR

1. Appropriate amounts of phospholipids are weighed to reach a final concentration of ca. 10–200Â€mM (see Note 11). 2. A suitable volume of deuterium-depleted water, 1H2O is added to obtain a lipid hydration of 80–90% (v/w). Hydration is defined as the mass of water over the total mass of the system (phospholipids and water).

NMR Spectroscopy of Lipid Bilayers

351

3. The hydrated sample is then vigorously shaken in a vortex mixer, frozen in liquid nitrogen, and heated to 50°C for 10Â€min in a water bath. This cycle of liposome (multilamellar vesicles) formation is repeated three to five times until a milky dispersion is obtained at room temperature. 4. The resulting dispersion is transferred into a 4-mm NMR ZrO2 rotor (100Â€ml) that is placed in the magnetic field (see Note 12). The measuring probe is properly tuned to 76.77Â€MHz (2H Channel), and the space homogeneity is adjusted at best. 5. As step 4 in SubheadingÂ€3.1. 6. 2H NMR spectroscopy is performed at 76.77Â€MHz. NMR spectra are acquired using a solid quadrupolar echo pulse sequence (18). Typical acquisition parameters are as follows: spectral window of 500Â€kHz, p/2 pulse width 3Â€µs (see Note 13), and interpulse delay of 30Â€µs. A recycle delay of 1.5Â€s is used. Typically, 512–1,024 scans are accumulated in 10–20Â€min. 7. Processing parameters: noise filtering of the FID with an exponential window, characterized by the Lorentzian linebroadening factor (LBâ•›=â•›50–300Â€ Hz), is used. Fourier Transformation is applied to get the spectrum, and its phase is adjusted to correct for unavoidable time delays in acquisition. The peak-to-peak separation, called the quadrupolar splitting, that can be in principle obtained for non-equivalent C–2H bonds, can be measured on some spectra. Spectral moments can also be used to estimate width and shape changes in spectra (14, 19). 8. The spectrum at 10°C for DPPtdCho-2H62 liposomes (Fig.Â€2a) is characteristic of a gel phase where the quadrupolar interaction occurring for each C–2H bond (there are 62 deuterium labelled positions on the acyl chains of DPPtdCho2 H62) reveals non-axial symmetry due to the very restricted chain dynamics (solid-ordered phase, so) and to the symmetry of the phase. No peak-to-peak measurement can be measured on such a spectrum, except for the central broad doublet assigned to the chain-end deuterated methyl groups. The first moment is calculated using a routine built in the Origin software. 9. The spectrum obtained at 40°C for DPPtdCho-2H62 liposomes (Fig.Â€ 2b) is much narrower along the frequency axis and shows a shape characteristic of a lamellar fluid phase with axial symmetry (liquid-disordered phase, ld). The overall spectrum is nonetheless complex because it represents the superposition of all spectra coming from all C–2H bonds. Several peak-to-peak separations, quadrupolar splittings, can be measured and plotted as a function of the labelled carbon position along the acyl chain (Fig.Â€2f). Spectral Â�deconvolution

352

Grélard et al.

to get oriented-like spectra (20, 21) can be applied here for more resolution (see Note 2). 10. The spectrum at 10°C for DPPtdCho-2H62/cholesterol (2/1) liposomes (Fig.Â€2c) is characteristic of a liquid-ordered phase (lo) where the quadrupolar interaction occurring for each C–2H bond reveals axial symmetry due to the axial diffusion of lipids in the membrane. 11. The spectrum at 40°C for DPPtdCho-2H62/cholesterol (2/1) liposomes (Fig.Â€2d) is characteristic of a liquid-ordered phase with greater dynamics than at 10°C (spectrum of smaller width). Several peak-to-peak separations, quadrupolar splittings, can be measured and plotted as a function of the labelled carbon position along the acyl chain (Fig.Â€2f). 12. Spectra were recorded as a function of temperature, the first moment calculated and plotted against temperature (Fig.Â€2e). Elevated values are found for gel phase temperatures and depict an ordered rigid state (so) whereas smaller values are found above the gel-to-fluid phase transition temperature (ca. 37°C) for pure DPPtdCho-2H62 liposomes and depict a disordered lamellar fluid phase (ld). In case of added cholesterol, the transition is nearly smoothed out demonstrating the regulating effect of cholesterol: it increases the fluidity of ordered phases and decreases that of fluid phases: this is the liquid-ordered state (lo). 13. Peak-to-peak separation, so-called quadrupolar splittings, measured for individual C–2H bonds report on local space and time angular fluctuations (order parameters, see Note 14). When plotted against labelled acyl chain position, they report on local dynamics (order/disorder) (Fig.Â€ 2f ). The greatest the quadrupolar splitting, the highest the order (less bond fluctuations). C–2H bonds close to the glycerol backbone (positions 2–8) are much more ordered than those near the bilayer centre (positions 12–16). 3.4. Determination of Lipid Molecular Structure in a Membrane Environment Using 1 H HR-MAS (Fig.Â€3)

1. Appropriate amounts of phospholipids are weighed to reach a final concentration of ca. 1–10Â€mM (see Note 11). 2. A suitable volume of ultra pure water mixed with deuterated water (90:10, v:v) containing traces of TSP (trimethylsilylpropionic acid) as a reference (0Â€ppm) is added to obtain a lipid hydration of 80–95% (v/w). Hydration is defined as the mass of water over the total mass of the system (phospholipids and water). 3. The hydrated sample is then vigorously shaken in a vortex mixer, frozen in liquid nitrogen, and heated to 50°C for 10Â€min in a water bath. This cycle of liposome (multilamellar vesicles) formation is repeated three to five times until a milky dispersion is obtained at room temperature.

NMR Spectroscopy of Lipid Bilayers

353

O

Liver PC

γ

γ

+

N

β α

γ

O

P

3'

1'

O

2'

H O O

O-

G1

G2

O

G3

5'

4'

2 1

7'

6'

4

9'

8'

6

3

10'

8

5

11'

13'

12'

10

7

12

9

15'

14'

18

11

17'

16'

18'

16

13

15

18 17

O O

Liver PE +

H3N

β

α

1'

O

H O

O P O

G2

O-

G1

2'

3'

O

2 1

G3

4'

5'

6'

4 3

7'

8'

9'

6 5

10'

8

11'

12'

10

7

14'

15'

13'

9

12 11

17'

19'

16' 18

13

18'

16 15

20'

18 17

O

O 3'

1'

Sphingomyelin

γ

γ N+ γ

O β

α

H

O P O O-

G1

2'

HN G2 G3

5' 4'

1

3 2

7' 6' 5

4

9'

7 6

11' 10'

8'

9 8

13' 12' 11

10

15' 14' 13

12

17' 16'

18'

15 14

H OH

Fig.Â€3. Molecular structures of selected lipids with numbering for TableÂ€1.

4. The resulting dispersion is transferred into a 4-mm NMR ZrO2 rotor for “High-Resolution MAS NMR” (50Â€ml) that is placed in the magnetic field. 5. Speed for MAS is set to 7–10Â€kHz using the pneumatic unit device. Using the variable temperature set up (VT and BCU units), samples are allowed to equilibrate 15–30Â€min at 30°C before the time dependent NMR signal (FID) is acquired; the temperature is regulated to ±1°C. The measuring probe is properly tuned to 500.16Â€MHz and the space and time homogeneity is adjusted by a deuterium “lock” system using the deuterated solvent as a reference. 6. 1H-NMR spectroscopy is performed at 500.16Â€MHz. NMR spectra are acquired using a single pulse sequence with water suppression (22). Typical acquisition parameters are a pulse duration (P1) of 5.5Â€ms, a time domain (TD) set to 32Â€K points, a spectral width (SW) of 20Â€ppm and a relaxation delay (D1) of 3Â€ s. The number of scans is set to 16. These conditions determine a total acquisition time of 90Â€s for the time dependent signal (FID). 7. Processing parameters: noise filtering with an exponential window of the FID, characterized by the Lorentzian linebroadening factor (LBâ•›=â•›0.3Â€Hz), is used. This noise filtering generally improves the signal-to-noise ratio, but at the cost of resolution. Fourier Transformation is applied to get the spectrum, and its phase is adjusted to correct for unavoidable time delays in acquisition (Fig.Â€ 4). Referencing (TSP rightmost resonance is set to 0Â€ppm), peak picking (measure of chemical

354

Grélard et al.

shifts in parts per million), and peak integration (measure line areas) are part of the relevant processing parameters. 8. The spectrum obtained in the absence of rotation (Fig.Â€4c) is characteristic of an unresolved proton spectrum of a membrane phase. Very little peak assignment can be made on such a spectrum. 9. The spectrum obtained for a MAS speed of 10Â€kHz (Fig.Â€4b) shows a resolution that is similar to that obtained in liquids (organic solution, Fig.Â€4a), the sample being still under the form of liposomes. Referencing, peak and peak integration can now be applied. 10. The strategy for structure elucidation is the same as for lipids in organic solvent solution; 2D-NMR experiments (COSY, a Chloroform solution

10

5

0

5

0

b Liposomes, Water, MAS B0 54.7 °

Rotor axis

d c 10 Liposomes, Water, Static

10

5

0

ppm (1H)

Fig.Â€4. 1H-NMR spectra of Liver PtdCho. (a) Lipid dissolved in CDCl3 solution. (b) Solid state “high-resolution” magic angle sample spinning (sketch in inset d) of lipid dispersed in water containing 10% D2O, MAS speedâ•›=â•›10Â€ kHz. Water suppression was used during signal acquisition. (c) Spectrum without spinning. Note the change in x-scale from (b) to (c).

NMR Spectroscopy of Lipid Bilayers

355

TOCSY) are often needed for complete assignment. TableÂ€1 reports the structural assignment of 1H chemical shifts for selected lipids (Fig.Â€ 3) under liposomal form, i.e., as water dispersions (see Note 8). 3.5. Determination of Lipid Internal Dynamics in Oriented Bilayers Using 2H-NMR Spin-Lattice Relaxation Time Measurements

1. Appropriate amounts of phospholipids are weighed to obtain a bicelle composition (ca. 80Â€mol% of 14:0/14:0 DMPtdCho2 H72 and 20Â€mol% of 6:0/6:0 DCPtdCho (23)) and dissolved in organic solvent (CHCl3/MeOH), the solvent is evaporated, water is added, the sample shaken in a vortex mixer and lyophilized to remove solvent traces. 2. A suitable volume of deuterium-depleted water is added to obtain a lipid hydration (water mass/total mass) 80–95% (w/w). Total lipid concentration is of the order of 200–50Â€mM. 3. The hydrated sample is then vigorously shaken in a vortex mixer, frozen in liquid nitrogen, and heated to 50°C for 10Â€min in a water bath. This cycle of bicelle (bilayer discs of 400Â€nm diameter) formation is repeated three to five times until a homogeneous translucent preparation is obtained at room temperature. 4. The resulting dispersion is transferred into a 4-mm NMR ZrO2 rotor (100Â€ml) that is placed in the magnetic field (see Note 12). The measuring probe is properly tuned to 76.77Â€ MHz (2H channel) and the space homogeneity is adjusted at best. 5. As step 4 in SubheadingÂ€3.1. 6. 2H NMR spectroscopy is performed at 76.77Â€MHz. A series of ca. 30 NMR signals are acquired using an inversion recovery quadrupolar echo pulse sequence. Typical acquisition parameters are as follows: spectral window of 250Â€kHz, p/2 and p pulse widths of 3 and 6Â€ µs (see Note 13), interpulse delay in the echo sequence of 30Â€µs, variable delays, t1, between the inverting p pulse and the first p/2 pulse of the echo sequence ranging between 1 and 1,500Â€ms. A recycle delay of 1.5Â€s is used and typically, 256 scans are accumulated per spectrum in 6Â€min. The whole experiment lasts for 4.5Â€h. 7. Processing parameters: noise filtering of all FIDs with the same exponential window, characterized by a Lorentzian linebroadening factor (LB) of 50Â€Hz. Fourier Transformation is applied to get the spectrum series. The zero order phase correction is adjusted on the last spectrum of the series and applied to all spectra (Fig.Â€5c). 8. Quadrupolar splitting can easily be measured on oriented Â�spectra and attributed to labelled carbon positions in the moleÂ� cular structure (Fig.Â€ 5b) using previous assignments (6).

356

Grélard et al.

Fig.Â€5. 2H-NMR relaxation of DMPtdCho-2H72 in DMPtdCho/DCPtdCho bicelles hydrated at 80% with 1H2O. (a) Rate of spin-lattice relaxation (1/T1z) as a function of labelled carbon position. (b) DMPtdCho-2H72 molecule showing carbon numbering. (c) Stack of Fourier transformed NMR signals after an inversion recovery quadrupolar echo sequence. X-axis is the classical frequency axis (kHz) and the t1 axis is that of delays between the inverting pulse and the first pulse of the echo sequence. Analysis of such data yields (a).

The area of each individual splitting is determined using a Bruker routine, and the area variation as function of t1 is fitted against the proper equation for inversion-recovery sequence in the Bruker package to obtain the individual spin lattice relaxation time T1 (see Note 15).

NMR Spectroscopy of Lipid Bilayers

357

9. Spin-lattice relaxation times measured for individual C–2H bonds report on molecular motions occurring at the nanosecond timescale. When plotted against labelled carbon position, they report on local dynamics (correlation times for segment or molecular motions) (Fig.Â€5a). The relaxation rate (inverse of T1) is the largest for C–2H bonds of the glycerol backbone (positions g1, g2, g3) and is much smaller for head group (a, b, g) or chain (2–14) positions. This indicates that the glycerol backbone possesses a very restricted motional freedom and can be considered as a semi-rigid “kneecap” separating two very fluid zones, the phosphocholine head group facing the water medium and the acyl chains constituting the oily bilayer interior.

4. Notes 1. Data may also be treated on the computer that drives the spectrometer. Most spectrometer-linked computers are now PCs. 2. Spectral moment calculations or spectral de-Pake-ing may not be available on computers driving spectrometers. They can nonetheless be implemented in PC by obtaining the source code from the authors (14, 21). 3. P1 is determined with a routine provided on all spectrometers (POPT on Bruker software). It must be determined especially for samples with different solvents (because of the different dielectric constants that may modify the P1 value). Note that elevated salt concentration may alter the P1 duration. 4. Relaxation delays are also determined using a routine provided on spectrometers, which allows calculation of spin lattice relaxation time, T1. D1 is usually set to 5â•›×â•›T1 for optimum relaxation delay. 5. The number of scans is chosen such that the signal-to-noise ratio is about or greater than 50. 6. Line broadening is set in order to filter spectral noise without marked modification of peak line width. 7. More sophisticated window filtering can be used as they are available on most spectrometer software. 8. Many groups can be identified by their 1H chemical shifts using chemical shift tables (24) and coupling constants (multiplet separation). Two-Dimensional NMR experiments (COSY, TOCSY) are sometimes needed for complete assignment. 9. Experiments may sometimes be run overnight implementing a temperature variation from low to high temperatures and back to low to detect possible sample hysteresis.

358

Grélard et al.

10. A pulse sequence is a series of radiofrequency impulsions separated by specific delays, which allow selecting one or several magnetic or electric molecular interactions to be observed. The time dependent signals (FID) that are detected following a pulse sequence must be Fourier Transformed to obtain spectral data. 11. In case of lipid mixtures (DPPC/Cholesterol), lipids are dissolved in organic solvent (CHCl3/MeOH), the solvent is evaporated, water is added, the sample shaken in a vortex mixer and lyophilized to remove solvent traces All these experiments can be run with DPPtdCho-2H31 (less expensive) instead of DPPtdCho-2H62, the numbers of scans must then be increased by a factor 2. 12. In order to facilitate the transfer into the rotor, it may be wise to vary (increase or decrease) the temperature to be in a fluid state. The sample may then be easily poured into the rotor. 13. The pulse must be very short to acquire very wide spectra (hundredths of kilohertz). Using longer pulse widths will lead to intensity loss in the outer verges of the spectrum and hence errors in spectral moment calculation. 14. Order parameters, S, represent a time and space average of molecules or of molecular bonds with respect to the bilayer normal, z, in an axially symmetric situation, as in lipid bilayers. The SZ order parameter may vary from 1 (M1 of ca. 150Â€ kHz or C–D2 quadrupolar splitting of 62.5Â€ kHz) for fully rigid solid state systems to 0 for liquid systems. 15. Mz(t1)â•›=â•›Mz(t1â•›= ∞)â•›×â•›(1â•›−â•›2â•›×â•›exp(–t1/T1z)) where Mz(t1) is the z-magnetization at time t1 obtained by straight integration of individual lines; Mz(t1â•›= ∞) is the Boltzmann equilibrium magnetization; t1 is the delay between the inverting p pulse and the first p/2 pulse of the echo sequence and T1z is the spinlattice (longitudinal) NMR relaxation time. References 1. Simons K, Ikonen E (2000) How cells handle cholesterol. Science 290:1721–1726 2. Beck JG, Mathieu D, Loudet C, Buchoux S, Dufourc EJ (2007) Plant sterols in “rafts”: a better way to regulate membrane thermal shocks. FASEB J 21:1714–1723 3. Byrne RD, Barona TM, Garnier M, Koster G, Katan M, Poccia DL, Larijani B (2004) Nuclear envelope assembly is promoted by phosphoinositide-specific PLC with selective recruitment of phosphatidylinositol enriched membranes. Biochem J 387:393–400 4. Dufourc EJ (2009) NMR for lipids and biomembranes. In: Wiley encyclopedia of

chemical Biology. Wiley, Chichester, England DOI: 10.1002/9780470048672. wecb 9780470048389 5. Gamier-Lhomme M, Grelard A, Byrne RD, Loudet C, Dufourc EJ, Larijani B (2007) Probing the dynamics of intact cells and nuclear envelope precursor membrane vesicles by deuterium solid state NMR spectroscopy. Biochim Biophys Acta Biomembr 1768:2516–2527 6. Aussenac F, Laguerre M, Schmitter JM, Dufourc EJ (2003) Detailed structure and dynamics of bicelle phospholipids using selectively and perdeuterated labels. A 2H-NMR and molecular mechanics study. Langmuir 19:10468–10479

NMR Spectroscopy of Lipid Bilayers 7. Loudet C, Manet S, Gineste S, Oda R, Achard MF, Dufourc EJ (2007) Biphenyl bicelle disks align perpendicular to magnetic fields on large temperature scales: a study combining synthesis, solid-state NMR, TEM, and SAXS. Biophys J 92:3949–3959 8. Marinov R, Dufourc EJ (1995) Cholesterol stabilizes the hexagonal type II phase of 1-palmitoyl-2-oleoyl sn glycero-3-phosphoethanolamine. A solid state 2H and 31P NMR study. J Chim Phys 92:1727–1731 9. Marinov R, Dufourc EJ (1996) Thermotropism and hydration properties of POPE and POPEcholesterol systems as revealed by solid state 2 H and 31P-NMR. Eur Biophys J 24:423–431 10. Aussenac F, Tavares M, Dufourc EJ (2003) Cholesterol dynamics in membranes of raft composition: a molecular point of view from 2 H and 31P solid state NMR. Biochemistry 42:1383–1390 11. Dufourc EJ, Parish EJ, Chitrakorn S, Smith ICP (1984) Structural and dynamical details of cholesterol-lipid interaction as revealed by deuterium NMR. Biochemistry 23: 6063–6071 12. Pott T, Maillet JC, Dufourc EJ (1995) Effects of pH and cholesterol on DMPA membranes: a solid state 2H- and 31P-NMR study. Biophys J 69:1897–1908 13. Marsan MP, Muller I, Ramos C, Rodriguez F, Dufourc EJ, Czaplicki J, Milon A (1999) Cholesterol orientation and dynamics in dimyristoylphosphatidylcholine bilayers: a solid state deuterium NMR analysis. Biophys J 76:351–359 14. Dufourc EJ (2006) Solid state NMR in biomembranes. In: Larijani B, Woscholski R, Rosser CA (eds) Chemical biology. Wiley, London, pp 113–131 15. Dufourc EJ, Mayer C, Stohrer J, Althoff G, Kothe G (1992) Dynamics of phosphate head

359

groups in biomembranes. Comprehensive analysis using phosphorus-31 nuclear magnetic resonance lineshape and relaxation time measurements. Biophys J 61:42–47 16. Dufourc EJ, Smith ICP (1986) A detailed analysis of the motions of cholesterol in biological membranes by 2H-NMR relaxation. Chem Phys Lipids 41:123–135 17. Rance M, Byrd RA (1983) Obtaining highfidelity spin-1/2 powder spectra in anisotropic media: phase-cycled Hahn echo spectroscopy. J Magn Reson 52:221–240 18. Davis JH (1979) Deuterium magnetic resonance study of the gel and liquid crystalline phases of dipalmitoylphosphatidylcholine. Biophys J 27:339–358 19. Davis JH, Jeffrey KR, Bloom M, Valic MI, Higgs TP (1976) Quadrupolar echo deuteron magnetic resonance spectroscopy in ordered hydrocarbon chains. Chem Phys Lett 42:390–394 20. Bloom M, Davis JH, Mackay AL (1981) Direct determination of the oriented sample NMR spectrum for systems with local axial symmetry. Chem Phys Lett 80:198–201 21. Sternin E, Bloom M, MacKay AL (1983) De-Pake-ing of NMR Spectra. J Magn Reson 55:274–282 22. Piotto MVS, Sklenar V (1992) Gradienttailored excitation for single-quantum NMR spectroscopy of aqueous solutions. J Biomol NMR 2:661–666 23. Raffard G, Steinbruckner S, Arnold A, Davis JH, Dufourc EJ (2000) Temperaturecomposition diagram of dimyristoyl-dicaproyl phosphatidylcholine “bicelles” self-orienting in the magnetic field. A solid state 2H and 31PNMR study. Langmuir 16:7655–7662 24. Pretsch E, Bühlmann P, Affolter C (2000) Structure determination of organic compounds. Springer, Berlin

as

Part V Molecular Modelling

as

Chapter 19 Critical Review of General Guidelines for Membrane Proteins Model Building and Analysis Catherine Etchebest and Gaelle Debret Abstract Membrane proteins play major roles in many biological processes such as signalling, transport, etc. They have been shown to be involved in the development of many diseases and have become important drug targets per se. The understanding of their functional properties may be facilitated if a 3D structure is available. However, in the case of membrane proteins, only a few 3D structures have been solved to date. Bioinformatics and molecular modelling approaches are thus powerful alternatives to fill the gap between the sequence and the structure. Here, a review of the most recent approaches is proposed together with guidelines on how to use them. In addition, insofar as important biological processes require conformational changes, we discuss some interesting methods aimed at exploring the dynamic behaviour of proteins in their membrane environment. The paper ends with a brief description of useful approaches for determining oligomerisation or ligand binding sites. Key words: Membrane proteins, Structure prediction, Bioinformatics, Molecular dynamics, Normal mode analysis, Docking, Conformational changes

1. Introduction For biologists working with membrane proteins, molecular modelling techniques have become a true companion. They are valuable tools that complement biophysical approaches by yielding, at an atomic level, structural and dynamic details for a given system. Their use extends from 3D structure prediction to fine structural and energetic details of ions transport as well as the study of conformational changes of proteins in the lipid environment. Even though the concepts are quite identical to those for soluble proteins, when dealing with membrane proteins, important differences exist in their practical usage. Indeed, the membrane environment, which represents an anisotropic medium compared to water, requires specific Jean-Jacques Lacapère (ed.), Membrane Protein Structure Determination: Methods and Protocols, Methods in Molecular Biology, vol. 654, DOI 10.1007/978-1-60761-762-4_19, © Springer Science+Business Media, LLC 2010

363

364

Etchebest and Debret

methods, protocols, and careful controls of different parameters for the characterisation of biologically pertinent features. This chapter describes the most used (and useful, it is hoped,) techniques for simulating membrane protein systems. We restrict our topic to the a-helical class of membrane proteins. The structural modelling of b-barrels proteins requires specific approaches that will not be detailed here. Indeed, b-strands are generally more difficult to predict because they are stabilised by long-range interactions. The first section is dedicated to tools aimed at constructing membrane protein structures from sequence information that could be guided by experimental data. We describe different strategies that can yield complete 3D-models when sequence relationship exists with known structures (homology modelling) or alternatively, we provide partial structural information such as the location of the transmembrane segments and/or the orientation of the N and C termini with respect to the intra/extra cellular medium. Some examples that illustrate the similarities and differences between the different methods are given. The second section focuses on structural analysis using sophisticated molecular simulation techniques, i.e., molecular dynamics (MD), normal modes analysis (NMA), steered molecular dynamics (SMD), driven molecular dynamics for deducing, and/or understanding functional characteristics of membrane proteins. These approaches enable, for instance, the exploration of the mechanical features of the protein, among which are domain flexibility, pathway of ions along channels with the corresponding free energies, etc. More importantly, molecular dynamics simulations are widely used for solving structures with nuclear magnetic resonance data (see chapters in Part IV of this volume). Steered molecular dynamics simulations also have become a true partner to Atomic Force Microscopic (AFM) experiments, providing further insight into the impact of the forces on the structure at an atomistic level. Even though most methods are practised by experts in the field, strong efforts are realised to make these tools as convivial as possible. For instance, the first difficulty lies in the elaboration of the system to be simulated, i.e., a protein embedded in a membrane in the presence of a solvent. Useful graphical interfaces are now available that allow the construction of a complete system. Lastly, we also introduce briefly the problem of constructing multimeric assemblies and ligand–protein complexes via docking methods. Generally, a supervised procedure is necessary because the membrane, which restrains the accessible regions for ligands, is frequently not considered. This issue is addressed in the light of a well-documented example. A rapid methodology survey is given for the different techniques, with a discussion of their capabilities and their limits. A few examples illustrate the topic with advices and recommendations.

Critical Review of General Guidelines for Membrane Proteins Model Building and Analysis

365

2. Materials A PC-computer with an internet connection is required. Most tools described below are available on web-servers and/or may be downloaded (freely). When dealing with molecular dynamics simulations, the calculations are CPU-time consuming. Therefore, a powerful computer is required in this case. Two main molecular dynamics softwares are freely available, namely, Gromacs and NAMD, and can be easily installed locally (see Note 1). They are particularly efficient on parallel architecture as developed on PC-clusters for instance. Charmm and Amber are also very popular programs, but they require an end-user licence and fee. Lastly, a graphical software is a valuable tool for exploring and analysing the 3D structure. Many programs exist, but the most popular ones in the community are undoubtedly PyMol and Vmd. Unfortunately, latest versions of Pymol are no longer available for free, and the free older versions (pre 1.0) are not maintained. Besides the program, numerous extensions developed by the community can be used via the Tk console for VMD or added through the plugin facilities for Pymol (see Note 2).

3. Methods 3.1. Tools for Bridging Modelling Structure from Sequence 3.1.1. With Available Known Homologues

Homology modelling is the most successful method for obtaining a reliable structure if a homolog exists in the Protein Data Bank. However, as a matter of fact, the number of membrane protein 3D structures solved at an atomistic level is considerably smaller compared to soluble proteins. Consequently, this approach is considerably limited to the few membrane protein families (see http:// blanco.biomol.uci.edi) available in the dataset. As an example, until recently, the G protein coupled receptor family, one of the largest membrane protein families with crucial functional properties, was represented by only one structure, the bovin rhodopsin (code PDB: 1F88 and 1HZX) (1, 2) with a rather weak resolution (2.8â•›Ǻ). Since, the protein was solved in different species (3, 4) and in different stages, i.e., free from ligand (5) or complexed with ligand. These last 2 years have seen the release of structures for two different GPCR members, the turkey b1-adrenergic GPCR (6) and the humanb2-adrenergic GPCR (7), and their engineered form (8, 9). Comparison of the structures of the rhodopsin and b1 adrenergic receptor clearly shows a strong similarity in the transmembrane (TM) domain (Fig.Â€1). Accordingly, a structural model of rhodopsin based on the structure of b1 adrenergic receptor (or reversely) might be reliable. However, this model strongly depends on the accuracy of the

366

Etchebest and Debret

Fig.Â€ 1. Superimposition of the structures of the bovine rhodopsin (pdb:1U19) in red and the b2 adrenergic receptor (pdb:2R1H). Left and middle: the aligned transmembrane core domains along two different views. Right: a view including the nonaligned regions in yellow. The T4-lysozyme region is omitted for clarity.

alignment. As an illustration, in Fig.Â€ 2, are given the sequence alignments between the two proteins, one using a powerful tool, Muscle (10, 11) for multiple sequence alignment (MSA) (see Note 3) and the other one deduced from Rapido (12), a new elegant software for structure superimposition (see Note 4). The alignments are far from being identical, but it may be noticed that in the transmembrane core, the alignments are very similar. In the present case, both alignments are difficult to obtain because the b2-adrenergic receptor is fused with T4-lysozyme. The use of a large set of homologous sequences permits generally to trap the main conserved (or similar) regions and yields an improvement of the alignment. As an advice, it is recommended to perform multiple sequences alignment for the target and for the template sequences separately and to correct the pairwise alignment between the template and the target accordingly. Information based on secondary structures and/or TM regions predictions are also valuable for improving the detection of similar regions. Numerous tools exist for the prediction of transmembrane domains mainly for a-helical bundles. Most of them rely on learning and data mining methods relying on databases dedicated to TM proteins structures (13–18). The most popular methods are based on sophisticated learning procedures such as Hidden Markov Models, Neural Networks, or Support Vectors Machine (HMMTop (19), TMHMM (20), PolyPHOBIUS (21), TMPpro (22), MEMSAT3 (23), SVMTop (24), k4HTM (25), and MemBrain (26)). Significant advances have recently been achieved, thanks to the incorporation of evolutionary information. One of the most interesting approaches based on physical principles demonstrates a success comparable with that of the best methods based on machine-learning techniques (27). This method relies on experimental data that describe the energetic of insertion into the endoplasmic reticulum membrane of a single TM segment embedded within a larger protein.

Critical Review of General Guidelines for Membrane Proteins Model Building and Analysis

367

Fig.Â€ 2. Sequence alignments between b2-adrenergic receptor (2RH1, first line) and rhodopsin (1U19, second line). (a) Alignment from MUSCLE and (b) alignment deduced from structural superimposition with RAPIDO (see SubheadingÂ€4).

Most approaches predict the number of TM segments, their limits, as well as the topology, i.e., the orientation of the protein with respect to the intra- or extra-cellular side. The prediction takes profit from a bias in the residues distribution in the loops located in the intracellular side. Indeed, it has been observed that these loops

368

Etchebest and Debret

are enriched in positive residues while the loops located on the extracellular side are depleted. This observation is known as the so-called “positive inside rule” (28). All these methods nevertheless suffer from the sparseness of the available structural data. Their performances tend to decrease when they are re-evaluated on new and larger datasets. Hence, a regular checking of the literature is recommended before selecting a method that claims to be the most accurate one. In addition, it has to be kept in mind that even with an accurate alignment, a structural homologous model requires further investigations and refinement to take into account possible distortions and/or slight displacements of the helices compared to the template structure. These differences in the transmembrane core domain, as well as the external regions, frequently nonconserved, correspond to the true signature of the sequence. They may be of considerable importance when dealing with fine recognition process and ligands specificity investigation (29). 3.1.2. Without Known Homologues

When no template structure is available or easy to detect, alternative approaches are required. These methods may be divided into two categories: threading (or fold recognition) for which a structural database is needed, and ab initio/deÂ€novo methods. In the last case, the structural modelling may be achieved by combining fragments for instance or by applying physical principles. In many cases, numerous models are constructed and sorted according to an appropriate scoring function. These approaches were initially developed and extensively tested on soluble proteins. They are currently being adapted for the prediction of transmembrane proteins. Threading requires a set of 3D structures and a scoring function that estimates the compatibility between the sequence and a 3D structure. This scoring function is frequently on the basis of statistical potentials learnt from a large databank of 3D structures. In case of membrane proteins, amino acids distribution in the transmembrane region is somewhat different and thus, statistical potentials from soluble proteins might not be appropriate. The TASSER approach, which combines threading and fragments combination, developed successfully for soluble proteins, was recently extended with specific adaptation for modelling the structure of all identified G Protein-coupled receptors in the human genome (30). The procedure was thus able to trap the seven-helices bundle topology but cannot be readily extended to other folds of membrane proteins. Similarly, a threading approach using the 3D structure of lactose permease from E. coli (LacY) was applied and permitted to detect proteins similar to LacY in prokaryotes and eukaryotes (31). Finally, when no putative template structure (evolutionary related or not) may be detected, ab initio methods are definitely required. Ab initio (or deÂ€ novo) methods are supposed to be much easier to develop for transmembrane (TM)

Critical Review of General Guidelines for Membrane Proteins Model Building and Analysis

369

domains of these proteins. Indeed, it has been argued that the constraint of the membrane strongly limits the diversity of possible folds compared to soluble proteins. However, such methods have been essentially applied for the prediction of proteins mainly composed of a-helices. The protocol is generally hierarchical and consists (1) in predicting the TM helical segments, (2) in determining the relative positions of each helical axis, (3) in establishing the orientation of the helical faces (lipid vs. protein interior), (4) in positioning the side chains and the connecting loops conformations, and (5) in refining all the structure (32). Steps 1 and 3–5, even though difficult, are tractable. Step 2 is clearly the most difficult step because one needs to know which helix is close to which other one. For this purpose, experimental data, for instance, are required that can come from low-resolution electron microscopy maps (Chapter 13 this book: by Hinsen etÂ€ al., “from electron microscopy maps to atomic structures using normal mode-based fitting”), and/or from any indirect biochemical or biophysical experiments, that could inform about the proximity between the helical segments. Since the pioneering work performed by Herzyk and Hubbard (33) and Baldwin (34, 35), many new protocols have been developed. However, they mainly focus on GPCR families (36–39). In these methods, the main prevailing hypotheses are roughly, the role of hydrophobicity, the role of the hydrophobic moment, and a specific “lipidic” moment that would point towards the lipid bilayer. The conservation patterns also help to orient the faces of the helices, the most conserved residues being supposed to point towards the interior of the protein (40). For instance, the LIPS server (41) (see Note 5) allows the prediction of the faces turned towards the lipid environment with a good efficiency. Finally, special care is taken for considering the specific packing forces that drive to the full folded structure. Preferred contacts in membrane protein structures were thus analysed and used to deduce statistical potentials. With their help and various rules, constraints are applied that guide the a-helix assemblies (42). Their success is still limited but quite promising because it is the only issue when no structural template exists. Most approaches consider helices as rigid bodies. Noteworthy, some methods begin to deal with local deformations of helices, which are frequently encountered in membrane proteins. Nowadays, such local deformations can be easily studied with synthetic peptides coupled to NMR studies (see above). Molecular mechanics computations can also be performed for studying the conformation of helices and per se, their putative deformations in an environment representative of the membrane properties (see next section). Lastly, interesting and promising success was obtained with Robetta method developed by Baker’s group for soluble proteins, and adapted for membrane proteins, with no need of any experimental data (43).

370

Etchebest and Debret

In summary, the recent literature shows some significant progress, but it is generally limited to few membrane protein families, mainly G coupled protein receptor for instance. Their success is high as the procedure takes advantage of experimental data to drive the conformational space exploration and to guide the finding of the 3D structure (44). An example of such an approach is described in detail in Chapter 20 in a concrete case with some advices. A systematic check of the literature is required before going further into the exploration of the modelled structure. 3.2. Tools for Exploring Function from Structure

When a 3D-structure is available, the next important step consists in exploring its functional properties. Different levels of exploration may be distinguished. The simplest one consists in analysing the structure itself and then its relation with the environment. Finally, the furthest step deals with the dynamical behaviour of the structure.

3.2.1. Structural Analysis

As in the case of soluble proteins, surface properties are useful information to understand functional properties. For instance, for receptor proteins, pockets have to be identified to propose putative binding sites. Different tools are available for determining pockets in proteins. CastP (45) is a method that provides us with different information such as the size of the pockets, the atoms that delineate the pockets and cavities, as well as the mouths of these pockets (available online, see Note 6). Naccess software developed by Hubbard and Thornton (46) is another useful tool that computes the accessibility of the residues. The accessible surface area S is generally measured by considering a probe with a 1.4â•›Ǻ radius i.e., a water molecule (but the value can be modified, see Note 7). The accessibility of a residue X is then defined as the ratio between the surface S and the accessible surface computed for the same residue in a tripeptide GXG, in an extended conformation. A residue is considered as buried when this ratio is smaller than 10% (25% is also an acceptable threshold for defining the burying limit). The state of the residue in the model structure may be compared with data deduced from experiments such as antigen–antibody recognition for instance. Indeed, if no important conformational rearrangement occurs during the recognition process, the residues have to be accessible before making the interaction. Accordingly, some models may be discarded or privileged. Finally, a convenient way to analyse a channel pore is to use the “Hole” approach developed by Smart etÂ€al. (47). The dimensions of the pore are obtained using a Monte Carlo simulated annealing procedure that finds the optimal pathway for a sphere with variable radius to squeeze through the channel (see Note 8).

3.2.1.1. Surface and Pockets

3.2.1.2. Helices Conformation and Relative Positioning

A first geometrical analysis is generally dedicated to the determination of the limits of the secondary structures. This is performed using for instance, DSSP (48) or Stride (49). Once the extents of

Critical Review of General Guidelines for Membrane Proteins Model Building and Analysis

371

the helices are obtained, their relative positioning may be defined with different geometrical parameters: distances between centres of mass, for instance, tilt angles or W angles that necessitate the helical axis computation (see Fig.Â€3). If the helix is not deformed, the calculation of the axis is obtained by diagonalisation of the inertia matrix of the structured segment defined in the previous step. The axis may be computed by considering the polypeptide backbone atoms or alternatively, the sole Ca atoms (see Note 9). Then it is possible to measure the tilt, i.e., the angle between the helical axis and the normal to the membrane plane. The main problem consists in determining this normal because the location of the membrane is only hypothesised. In many cases (but not in all cases), the principal inertia axis of the whole TM domain is aligned with this normal. Yet, it is sometimes necessary to apply transformations on the coordinates (e.g., rotations) to make the appropriate alignment. If the membrane protein is an oligomer, the membrane normal is generally considered as the rotation axis for superimposing one monomer on another one. In the absence of symmetry, the problem is much more complicated. A visual inspection and a human supervision may be sometimes necessary to retrieve the plane. The location of aromatic residues, which generally delimit the interfacial region and lie roughly in the same plane, can be a useful guide to define the membrane plane. Finally, Lomize etÂ€al. (13, 50) proposed a method that predicts, from the structure, the orientation of the protein in the membrane. It is determined by minimising the transfer energy, DGtransfer, from water to membrane interior, taking account different variables, and in a coordinate system whose axis Z coincides with the bilayer normal. The results for all membrane proteins are available online in a database (see Note 10). The calculation of the crossing W angle (also called packing angle) may be obtained using the approach defined initially by

Fig.Â€3. Definition of the tilt (t), azimuthal rotation (r), and crossing (W) angles.

372

Etchebest and Debret

Chothia etÂ€al. (51) or with more recent tools such as HA2 (see Note 11, (52)). The main steps consist in computing the mass centres of each helix in contact with another one, defining the vector that joins these two points, and then computing the dihedral angle defined by the two helical axes with respect to this vector. On the basis of the angle distribution in membrane proteins, significant differences were observed with packing angles in soluble proteins. However, when looking at structural homologues in soluble protein, similarities seem to exist that can be useful for deducing sequence–structure relationship (53). Besides, a recent study shows some differences between packing angles in channels and membrane bundles (also called membrane coils in the publication) (54). The determination of these parameters requires that the helices are mainly straight. Yet, in many cases, the helices are slightly deformed, curved, or even kinked. These deformations induce modifications in the local helical parameters that may deviate significantly from ideal a-helix characteristics. These parameters can be computed, for instance, with HELANAL (55, 56) that provides us with a detailed description of the helix properties: bending angles, average helical radius, twist angles, etc. (see Note 12). The Qhelix computational tool (57) complements this analysis by measuring the inter-helical angles (see Note 13). In our recent study (58), we also examined kink angles in TM helices and provided a method that permits to locate automatically the residue(s) responsible for the deformation as well as the amplitudes and the direction of the kinks (see Note 14). 3.2.2. Dynamics and Conformational Changes

The different approaches described above allow construction and analysis of model structure. They provide first elements to understand function and to delineate functional sites. Yet, in many cases, protein function is accomplished through conformational changes that range from a few angströms (e.g., side chain rotation) to tens of angströms displacements (domain motions) and extend from few picoseconds to seconds. Experimental approaches yield in most cases indirect information on the structures and the dynamics of the protein. In some cases, they can take profit from simulation and modelling techniques to interpret at an atomic scale, the data they observe (see others chapters in Parts III and IV in this volume).

3.2.2.1. Molecular Dynamics Simulations and Related Approaches

There are nowadays the most popular in silico approaches for exploring the time behaviour of a protein and its putative conformational changes with atomic details. However, the time scale that can be simulated is far from being comparable with the experiment time scale (see Chapter 21 for the concepts and a detailed discussion). For membrane proteins, the main problem relates to the lipid environment which increases considerably the

Critical Review of General Guidelines for Membrane Proteins Model Building and Analysis

373

number of atoms included in the system, and per se limits the time scale that could be simulated. A second problem lies in the representation of the environment. If experiments are conducted in a cellular medium, the environment is so complex that it can’t be truly represented in the simulations. In this way, direct comparison is quite limited and even prohibited. For inÂ€vitro experiments, the medium is simpler but not necessarily reproducible directly in the simulations. Indeed, most simulation methods require the computation of energies between chemical groups encountered in proteins, lipids, and any component participating in the system (e.g., prosthetic groups, ions, quinones, organic molecules, detergents, etc). These interaction energies are on the basis of empirical force fields with parameters issued from a fine calibration aiming at reproducing experimental data and/or sophisticated quantum mechanics computations. Nowadays, two main families of force fields are currently available. The most standard ones are founded on a full (or almost full) atomic description while the most recent ones reduce the number of particles by using a coarse-grained description (see Chapter 22). The last ones are frequently calibrated on the first ones. Most chemical groups, required for the simulation of proteins in a pure lipid bilayer, are present in the most exercised force fields (see ref. 59 for a review). For instance, the Charmm force field (revisited recently for some lipids (60–63)) contains, besides classical amino acid residues, parameters for the head groups of the major lipids. Similar lipid molecules can be approached with AMBER force field and can also be treated in Gromacs package through the parameterisation developed by Berger (see Chapter 21 for advices). However, the large diversity of lipids, or detergents used in experiments, in particular, the nature of the polar headgroup, necessitates the fine calibration of chemical groups, which are “unusual” compared to those encountered in proteins or DNA molecules. Consequently, before starting any simulation, it is necessary to verify that all the chemical groups represented in the system are available in the force field. A careful reading of the outputs of the construction of the system (topology, force field implementation) is thus required even if most packages alert the user of the lack of parameters (see Note 15). The second step consists in constructing the membrane and placing the protein in an appropriate position. Different pre-equilibrated (pure) solvated membranes (“lipid box”) are currently available and easily usable (see Note 16). It is however, generally necessary to enlarge the lipid box to accommodate a large protein such as the MscL (Mechanosensitive Channel of Large conductance) for instance. This is generally performed by duplicating the box and translating the generated copy along appropriate axes. This enlarged lipid box is afterwards equilibrated for a short period of time (few nanoseconds). Then, the protein is

374

Etchebest and Debret

placed in the centre of the resulting lipid box. The position of the protein along the normal axis of the membrane normal is the most difficult to determine. Besides experimental data, aromatic residues are a good help for determining the appropriate location along this axis. Indeed, they frequently anchor the protein at the membrane interface. In their absence, the most hydrophobic region of the membrane can be located and aligned with the region of the lipid acyl chains. Finally, the system lipids–protein is solvated. This can be done using a box of solvent (mainly water) pre-equilibrated (see Note 17). Once the system is defined, the simulation can be started, choosing appropriate conditions (for practical recommendations, see Chapter 21). Finally, analysis of different structural as well as energetic properties can be achieved along the time course of the simulations (see SubheadingÂ€3.2.1). The amplitude of motions strongly depends on the simulation time and consequently, large conformational changes are rarely observed. Essential dynamics is a convenient way to highlight relevant collective motions (64). The method is on the basis of the principal component analysis of the covariance matrix (see Note 18). Once the directions of the movement are obtained, trajectories can be generated along these privileged directions. It is therefore possible to construct new structures along a putative conformational pathway and to cross barriers using slightly larger amplitudes compared to the MD average fluctuations. Deformations along principal directions are closer to the “natural” deformation tendencies of the protein in comparison to classical steered molecular dynamics that generally require assumptions on the points to apply the forces and the directions of these forces. The main limitation in case of essential dynamics is of course to have beforehand classical MD simulations for obtaining the principal directions. In contrast, steered molecular dynamics are extremely useful for obtaining results comparable to AFM experiment results (65). Brownian dynamics (BD), in which a stochastic force replaces the explicit solvent, is a very interesting alternative for studying large systems and simulating long time processes. Few studies on membrane systems have been reported until now, but their applications are quite important (66). Besides BD simulations, other approximations may be introduced and consist mainly in simplifying the lipid environment representation. The simplifications range from coarse-grained models for lipids (see Chapter 22) to implicit membrane models. In this last case, the lipid bilayer is represented by slabs with different thickness and physical properties (67, 68). 3.2.2.2. Normal Mode Analysis

This approach, introduced in details in Chapter 13, enables to trap the main tendencies of a given structure to accomplish conformational changes. The procedure relies on the harmonic approximation of the energy surface. For a few years, crude energy models have been introduced. They do not necessitate large

Critical Review of General Guidelines for Membrane Proteins Model Building and Analysis

375

computer memory (as required with classical all-atoms force field) or significant computer time. We applied such an approach for studying the mechanical properties of MscL and obtained quite interesting results that pointed out putative regions involved in the opening process as well as the relevant directions of motions to reach an open state (see Note 19). 3.2.3. Tools for Predicting Multimeric Assemblies, Membrane Protein Complexes

In many cases, the tools for predicting multimers or more generally complexes involving protein–protein/protein–ligands interactions are on the basis of classical docking approaches similar to those used for soluble proteins (69). Docking procedures may be decomposed into different steps: the first stage of the process consists in exploring different positions of one protein (“the ligand”) with respect to the surface of the target (“the receptor”). Each position is scored with the aid of a simple energy function. At this stage, the partners are generally considered as rigid. In the second stage, a first set of optimal configurations (generally more than a thousand) is selected and refined with a more sophisticated scoring function. The flexibility of the ligand may be introduced as well as some local conformational changes in the receptor. The solutions are then ranked, and few complexes are finally considered. The docking procedures are all the more efficient that some experimental data are used to guide the exploration. In the absence of such data, predictions based on sequences features (mainly based on multiple sequence alignment) can be used. Illustrative of the application of such approaches is the recent oligomeric structure prediction of G protein coupled receptor (70). In this case, the interaction site is located inside the membrane domain while in the case of ligands, the interaction frequently occurs with regions close to or in contact with the solvent (see Note 20 and Chapter 23). As an example, we recently examined the binding sites of the DARC protein (Duffy antigen/ receptor for chemokine) supposed to interact with loops and/or termini regions (see Note 21 and Chapter 20). Once the ligand binding site is located, dedicated tools for drug design may be applied, namely screening libraries for detecting new leads. Finally, coupled with methods described above, alternative conformations may be explored to identify new putative binding sites. In this context, an interesting work, very recently published, explores direct and also indirect interactions on recognition controlled from distant sites in proteins, e.g., by changes in protein conformation and mobility. The approach, called proteochemometrics, seems to successfully predict the location of indirect effects on ligand recognition arising from distant sites in the receptors (71).

3.3. C onclusion

In summary, molecular modelling can be used to successfully construct models of membrane proteins either by complete or

376

Etchebest and Debret

partial homology or by ab initio strategies and thus, be relevant for a broad range of applications. It also explores conformational changes of these proteins either by placing atomic structure in lipid environment or by making a dynamic bridge between static view of membrane proteins gained from X-ray, EM, or NMR data. Its main limits reside in time scale reached (still far from classical enzymatic reaction (milliseconds)) and in the accuracy of the physical atomic description used (force field). However, immense progress has been made these last few years on both programs and equipments that should permit to overcome these difficulties. Finally, it is the method of choice for drug design as long as atomic structures are reliable.

4. Notes 1. NAMD may be downloaded from the URL: http://www. ks.uiuc.edu/Development/Download/download. cgi?PackageName=NAMD. A large number of pre-compiled binaries are available, for different architectures, including Windows and MacOS. In Windows environment, NAMD must be run from the MS-DOS command line. Cygwin package, a Unix-like environment for Windows, (http://www.cygwin. com/) may be installed to facilitate the usage of the program. NAMD can also be compiled from source files. Force field data, i.e., topology files and parameters, are also required. Various versions of the Charmm force field are available at http:// mackerell.umar yland.edu/CHARMM_ff_params.html. NAMD is also compatible with Amber (http://ambermd.org/ dbase.html) and X-PLOR force fields. Preparation of some input files are facilitated by the use of VMD that is, besides its graphical capacities, an excellent partner for NAMD. â•… Gromacs is nowadays widely used in the membrane protein modellers’ community (see Chapters 21 and 22). Source codes are freely available at http://www.gromacs.org/ and pre-compiled binaries also exist for different architectures. Gromacs has become very popular because it is extremely fast and, importantly, is provided with numerous useful tools dedicated to the analysis of proteins structures, membrane characteristics, as well as many other physical properties such as diffusion coefficients, order parameter, radial distribution function, etc. Parallel computations may be performed using MPI library. The LAM (Local Area Multicomputer) package is generally used on local workstation networks.

Critical Review of General Guidelines for Membrane Proteins Model Building and Analysis

377

â•… For some computations, the FFTW (Fast Fourier Transform) library is required. It may be downloaded at http://www. fftw.org/. It has to be installed before configuring Gromacs. Both NAMD and Gromacs necessitate a C compiler and eventually an f77 compiler. 2. For VMD, the Tk console allows extending the possibilities of computations without modifying the program itself. A library of scripts (Python or Tcl) is available at http:// www.ks.uiuc.edu/Research/vmd/script_library/. The users’ community also provides us with many useful additional tools, the main difficulty being to locate them. â•… For Pymol, Tcl/Python scripts are made available under the Plugin menu once the files are copied in the PyMol modules/pmg_tk/startup directory. Some useful scripts can be found in the PyMol wiki (http://pymolwiki.org). 3. Muscle is available at http://www.ebi.ac.uk/Tools/muscle/ index.html. It is included in Seaview (http://pbil.univ-lyon1. fr/software/seaview.html), a graphical multiple sequence alignment editor, that also proposes ClustalW, the most popular program for multiple sequence alignment. Seaview binaries may be downloaded for different computer architectures (MacOS, Windows and Linux). The source code is freely available. 4. Rapido is a recent approach (http://webapps.embl-hamburg. de/rapido/) that permits the superimposition of 3D structures of proteins. It is able to take into account large structural changes such as hinge motions between domains. It is a very convenient way to locate rigid bodies vs. flexible regions. Besides its own Java applet (Jmol), it provides rasmol and Pymol scripts for easily visualising the results with the corresponding softwares. 5. The LIPS method is on the basis of a canonical model of the heptad repeat originally developed for coiled coils. The LIPS server (http://gila.bioengr.uic.edu/lab/larisa) requires a multiple sequence alignment for the TM domain studied. The helical domain is divided into seven surfaces. For each surface, are given a list of residues with different scores and three average values: the lipophilicity surface score, the entropy (a measure related to the conservation of a position in the multiple sequence alignment), and the LIPS score, the product of the two previous values. For the TM2-MscL, the LIPS score didn’t correctly assign the lipid-oriented face but the lipophilicity surface score was quite relevant. The entropy values strongly depend on the quality of the alignment and the diversity of the sequences used. A careful analysis of the MSA is thus recommended.

378

Etchebest and Debret

6. CastP server (http://sts-fw.bioengr.uic.edu/castp/calculation.php) uses a PDB file as input. The results are displayed through a quite convenient interface. It is possible to select a given pocket, to obtain different information such as its delimiting residues or its area and volume. If the user email is given (highly suggested), a set of files is sent, which contains all the useful results. In particular, the parameter Nmouth is indicated in the .pocInfo file. If the value is 0, it means that the pocket is a cavity fully buried in the protein. Distinction between pockets and cavities is important. In the first case, it could be a ligand binding site while in the second case (cavity), the void space could facilitate internal motions. A Pymol plugin may be downloaded (http://sts-fw.bioengr.uic.edu/ castp/pymol.php) and installed as described in Note 2. The plugin itself does not make CastP computations but enables to retrieve the results using the job identification number of a previous CastP request. 7. Naccess is available at http://www.bioinf.manchester.ac.uk/ naccess/. It can be easily installed on many platforms because the source files are given. The default probe radius is 1.4â•›Å (a water molecule radius). In some publications on membrane proteins, the probe radius is increased to 1.9â•›Å to better account for the accessible surface of atoms to CH2 groups of lipids. We developed a program in the context of Gromacs, to easily compute Naccess accessibilities along the time course of a Gromacs MD simulation. Some data for lipids have been included in the radii file. The program is available on request. We also considered a 3â•›Å probe radius when we tested the accessibility values for the Ca’s in a bundle. This raw probing enables to better compare the lipid accessible faces predicted from the sequences to those computed from the 3D structure. 8. The Hole program may be downloaded at (http://hole.biop. ox.ac.uk/hole). The main inputs are an initial position and a direction corresponding roughly to the channel axis (typically 0â•›0â•›1 if the channel is oriented along the Z axis). Apart from numerical results, the program can also generate some files for visualising the results with VMD. 9. The inertia axis for a given helical segment may be computed with a Python script that takes profit from the Python numerical libraries, i.e., Numpy or Scipy, that include linear algebra calculations. A part of a simple script developed by P. Poulain in our group is given below. The full script, available on request, also includes some tricks to visualise the axis with Pymol.

Critical Review of General Guidelines for Membrane Proteins Model Building and Analysis

379

10. The OPM database can be found at http://opm.phar.umich. edu/. It is not possible to make its own computation, but it is possible to send a request by mail to the authors. 11. The program HA2 can be downloaded at http://bioinfo.tau. ac.il/~sarel/HelAna.html and easily installed on a PC-Linux with Perl language programming available. However, it requires MatLab package that is not a freeware, even for academics. Similar routines could exist in the R package (freeware) and can be appropriately called in the main Perl routines. 12. Helanal (http://nucleix.mbu.iisc.ernet.in/helanal/helanal. shtml) only needs a pdb file and by default, it uses the information of the field “HELIX” in the pdb file (location and limits of the helical secondary structure), but it is possible to explicitly introduce these limits. Helanal yields much useful information about the helical geometrical parameters (twist, rise, pitch, etc.). It distinguishes linear, curved, and kinked helices. 13. QHelix (http://compbio.sookmyung.ac.kr/~qhelix/) yields interhelical axis values. The inputs are a pdb file, the position of the first residue in each helical segment, and the length of the helix. It is not possible to consider multichains complex unless some modifications are made in the pdb file, i.e., a renumbering of the residues. In contrast to Helanal, it does not consider the field HELIX as a default. A Python script may easily extract this information from the pdb file and writes in a correct way, the input for Qhelix (available on request). The angles are calculated for all pairs of helices. 14. The program, written in C Language, includes a loop running on the number of residues in the helix. First, a residue is selected along the helix and for each helical sub-domain this residue delimits, an inertia axis is computed. Then, the angle

380

Etchebest and Debret

between these two axes is calculated. The kink angle is defined as the angle with the largest value, when all the residues along the helix have been tested. The position of the kink corresponds to the residue where the angle is maximal. 15. In case of absence of parameters, it is possible to develop new parameters by transferring values from close fragments and adjusting them to reproduce experimental observables. A more rigorous approach consists in dividing the molecule of interest into small representative fragments, in performing sophisticated ab initio quantum chemistry computations, and in deducing empirical parameters, such as charges, coefficients for Lennard–Jones potentials, and torsion barriers, etc, which should reproduce the quantum computations (see ChapterÂ€24). Then the parameters for fragments have to be combined, generally slightly modified when they are transferred to the whole molecule. This new set of parameters is then assessed on experimental measurements. Yet, it is important to keep in mind that the parameters for lipids generally are calibrated to reproduce experimental data measured on pure membranes that include one type of lipids at a time. The use of data measured on biological membranes i.e., a mixture of lipids with different polar head groups and different alkyl chains could thus be inappropriate. In the same spirit, experimental information on membranes (even pure) that accommodate a protein, are very difficult to obtain. 16. Different coordinates and topologies files for membranes preequilibrated with Gromacs software may be downloaded at Peter Tieleman’s web site (http://moose.bio.ucalgary.ca/ index.php?page=Structures_and_Topologies). Valuable information is also available from Dr. M. Karttunen’s web site (http://www.apmaths.uwo.ca/~mkarttu/downloads. shtml). In particular, a mixed membrane of POPG/POPE, as well as the parameters for a SDS micelle, is provided. The crude placement of the protein in the membrane leads to numerous overlaps with lipid atoms. In most cases, the lipids that overlap with any atoms of the proteins are removed. Roux’s group (http://thallium.bsd.uchicago.edu/RouxLab/ method.html) developed a slightly different strategy for embedding a protein in a membrane. The lipids from a preequilibrated and pre-hydrated set are randomly placed around the protein. The number of core–core overlaps between heavy atoms is reduced through convenient rotations and translations of the lipids and protein. To provide the initial positions in the membrane plane for each lipid, a simplified representation of the full lipid molecule is used. Then the packing of the particles with the protein is obtained with a MD simulation using appropriate restraints.

Critical Review of General Guidelines for Membrane Proteins Model Building and Analysis

381

17. With the genbox tool from Gromacs, water molecules are placed in any free volumes, whose size is sufficient to accommodate a water molecule. Consequently, it is sometimes necessary to remove (by hand) some water molecules located in the hydrophobic membrane domain. In case of MscL, we chose firstly to fully solvate the channel. We performed a MD simulation where the periplasmic loops and the side chains bordering the channel pore were free to move, while restraining all the TM domains of the protein. Afterwards, we placed the protein plus the equilibrated water molecules contained in the channel vestibule in a solvated membrane. 18. Principal component analysis can be performed with Gromacs tool. It is important to check that the cosine content of eigenvectors is different from one to differentiate true motions from random diffusion. The relevance of the analysis can be checked by comparing the directions computed for different parts of the simulation. If the directions are conserved whatever be the set of time frames chosen, the corresponding directions can be considered as significant. Finally, to focus on significant directions, it is also convenient to compute the contribution of each direction to the total fluctuations. In most cases, the very first directions contain the largest part of the motions. 19. A main question in NMA relates to the choice of modes for exploring a conformational change. The modes with the lowest frequencies are obviously the most appropriate ones because they are associated with the largest amplitudes of motions. Yet, in our study, we also found that the collectivity measure is a powerful index for selecting interesting and significant modes. This index evaluates the collective protein motions within a given mode or eigenvector (i.e., the number of atoms significantly affected). It is rarely used but we suggest considering it systematically. We chose the definition proposed by Bruschweiler (72).

k =

 N  1 exp  − ∑ aAi2 log aAi2   i =1  N

where Ai is the amplitude of displacement and a is a normalisation factor such as ∑ aA 2 = 1. The conformational change i is maximal for a value of 1 and minimal for a value of 1/N . 20. In the case of ligands, the flexibility of the receptor may play a major role and has to be taken into account. However, most docking approaches only include flexibility at a final refinement stage. Hence, if the interaction requires complex conformational changes, the prediction could be completely misleading. New approaches aim at better introducing the

382

Etchebest and Debret

flexibility (73) but actually they do not take into account the conformational changes all along the docking process. 21. We have elaborated an original protocol for studying the DARC protein (see Chapter 20). The DARC protein is an erythrocyte seven TM-receptor that binds chemokines and also malaria parasites (Plasmodium vivax and Plasmodium knowlesi). Biological data show that the four extracellular domains (ECDs) of DARC are essential for its interaction with chemokines (CXC-L8 for instance), while the first (ECD1) is sufficient for the interaction with malaria erythrocyte-binding protein (duffy binding protein DBP). Importantly, ECD1 domain is a highly flexible 50 residues peptide with two sulfated tyrosines. When starting from a 3D structural model of DARC (44), including ECD1 in a free state structure, and using a rigid-body docking approach, we failed to find a solution compatible with all the experimental data for the binding sites of P. vivax DBP and CXC-L8. We finally combined semi-flexible and semi-rigid dockings taking into account ECD1 flexibility: (1) the flexible ECD1 domain is independently docked on a rigid ligand using the flexible docking tools of ICM (74), (2) in parallel, the rigid ligand is docked on the rigid TM domain, and (3) solutions from both steps are combined and only those compatible with bond constraints between ECD1 and TM1 are selected. This approach yields a 3D-model for DARC–CXC-L8 and DARC– DBP complexes, both consistent with available experimental data. Such models permitted us to identify and predict potential residue targets for mutagenesis, and to define the area of interaction for receptor and both ligands. References 1. Teller DC, Okada T, Behnke CA, Palczewski K, Stenkamp RE (2001) Advances in determination of a high-resolution three-dimensional structure of rhodopsin, a model of G-proteincoupled receptors (GPCRs). Biochemistry 40(26):7761–7772 2. Palczewski K, Kumasaka T, Hori T, Behnke CA, Motoshima H, Fox BA etÂ€al (2000) Crystal structure of rhodopsin: a G protein-coupled receptor. Science 289(5480):739–745 3. Murakami M, Kouyama T (2008) Crystal structure of squid rhodopsin. Nature 453(7193):363–367 4. Shimamura T, Hiraki K, Takahashi N, Hori T, Ago H, Masuda K etÂ€al (2008) Crystal structure of squid rhodopsin with intracellularly extended cytoplasmic region. J Biol Chem 283(26):17753–17756

5. Park JH, Scheerer P, Hofmann KP, Choe H, Ernst OP (2008) Crystal structure of the ligand-free G-protein-coupled receptor opsin. Nature 454(7201):183–187 6. Warne T, Serrano-Vega MJ, Baker JG, Moukhametzianov R, Edwards PC, Henderson R etÂ€ al (2008) Structure of a beta1-adrenergic G-protein-coupled receptor. Nature 454(7203):486–491 7. Rasmussen SGF, Choi H, Rosenbaum DM, Kobilka TS, Thian FS, Edwards PC etÂ€ al (2007) Crystal structure of the human beta2 adrenergic G-protein-coupled receptor. Nature 450(7168):383–387 8. Cherezov V, Rosenbaum DM, Hanson MA, Rasmussen SGF, Thian FS, Kobilka TS etÂ€ al (2007) High-resolution crystal structure of an engineered human beta2-adrenergic G

Critical Review of General Guidelines for Membrane Proteins Model Building and Analysis

9.

10.

11.

12.

13.

14.

15.

16.

17.

18. 19. 20.

21.

protein-coupled receptor. Science 318(5854): 1258–1265 Hanson MA, Cherezov V, Griffith MT, Roth CB, Jaakola V, Chien EYT etÂ€ al (2008) A Â�specific cholesterol binding site is established by the 2.8Â€ A structure of the human beta2adrenergic receptor. StructureÂ€ 16(6): 897–905 Edgar R (2004) MUSCLE: a multiple sequence alignment method with reduced time and space complexity. BMC Bioinformatics 5(1):113 Edgar RC (2004) MUSCLE: multiple sequence alignment with high accuracy and high throughput. Nucleic Acids Res 32(5): 1792–1797 Mosca R, Schneider TR (2008) RAPIDO: a web server for the alignment of protein structures in the presence of conformational change. Nucleic Acids Res 36(Suppl 2):W42–W46 Lomize MA, Lomize AL, Pogozheva ID, Mosberg HI (2006) OPM: orientations of proteins in membranes database. Bioinformatics 22(5):623–625 Tusnády GE, Dosztányi Z, Simon I (2005) PDB_TM: selection and membrane localization of transmembrane proteins in the protein data bank. Nucleic Acids Res 33(Database issue):D275–D278 Tusnády GE, Kalmár L, Simon I (2008) TOPDB: topology data bank of transmembrane proteins. Nucleic Acids Res 36(Database issue):D234–D239 Tusnády GE, Kalmár L, Hegyi H, Tompa P, Simon I (2008) TOPDOM: database of domains and motifs with conservative location in transmembrane proteins. Bioinformatics 24(12):1469–1470 Tusnády GE, Dosztányi Z, Simon I (2004) Transmembrane proteins in the Protein Data Bank: identification and classification. Bioinformatics 20(17):2964–2972 Jayasinghe S, Hristova K, White SH (2001) MPtopo: a database of membrane protein topology. Protein Sci 10(2):455–458 Tusnády GE, Simon I (2001) The HMMTOP transmembrane topology prediction server. Bioinformatics 17(9):849–850 Krogh A, Larsson B, von Heijne G, Sonnhammer EL (2001) Predicting transmembrane protein topology with a hidden Markov model: application to complete genomes. J Mol Biol 305(3):567–580 Käll L, Krogh A, Sonnhammer ELL (2005) An HMM posterior decoder for sequence feature prediction that includes homology information. Bioinformatics 21(Suppl 1):i251–i257

383

22. Ganapathiraju M, Balakrishnan N, Reddy R, Klein-Seetharaman J (2008) Transmembrane helix prediction using amino acid property features and latent semantic analysis. BMC Bioinformatics 9(Suppl 1):S4 23. Jones DT (2007) Improving the accuracy of transmembrane protein topology prediction using evolutionary information. Bioinformatics 23(5):538–544 24. Lo A, Chiu H, Sung T, Lyu P, Hsu W (2008) Enhanced membrane protein topology prediction using a hierarchical classification method and a new scoring function. J Proteome Res 7(2):487–496 25. Kitsas IK, Hadjileontiadis LJ, Panas SM (2008) Transmembrane helix prediction in proteins using hydrophobicity properties and higher-order statistics. Comput Biol Med 38(8):867–880 26. Shen H, Chou JJ (2008) MemBrain: improving the accuracy of predicting transmembrane helices. PLoS One 3(6):e2399 27. Bernsel A, Viklund H, Falk J, Lindahl E, von Heijne G, Elofsson A (2008) Prediction of membrane-protein topology from first principles. Proc Natl Acad Sci U S A 105(20): 7177–7181 28. von Heijne G, Gavel Y (1988) Topogenic signals in integral membrane proteins. Eur J Biochem 174(4):671–678 29. Kobilka B, Schertler GFX (2008) New G-protein-coupled receptor crystal structures: insights and limitations. Trends Pharmacol Sci 29(2):79–83 30. Zhang Y, Devries ME, Skolnick J (2006) Structure modeling of all identified G proteincoupled receptors in the human genome. PLoS Comput Biol 2(2):e13 31. Kasho VN, Smirnova IN, Kaback HR (2006) Sequence alignment and homology threading reveals prokaryotic and eukaryotic proteins similar to lactose permease. J Mol Biol 358(4):1060–1070 32. Etchebest C, Popot J (1997) Packing transmembrane a-helices into bundles: computational vs experimental approaches. In: von Heijne G (ed) Membrane protein assembly. Chapman & Hall, New York, pp 221–250 33. Herzyk P, Hubbard RE (1998) Combined biophysical and biochemical information confirms arrangement of transmembrane helices visible from the three-dimensional map of frog rhodopsin. J Mol Biol 281(4): 741–754 34. Baldwin JM (1993) The probable arrangement of the helices in G protein-coupled receptors. EMBO J 12(4):1693–1703

384

Etchebest and Debret

35. Baldwin JM, Schertler GF, Unger VM (1997) An alpha-carbon template for the transmembrane helices in the rhodopsin family of G-protein-coupled receptors. J Mol Biol 272(1):144–164 36. Elofsson A, von Heijne G (2007) Membrane protein structure: prediction versus reality. Annu Rev Biochem 76:125–140 37. Dastmalchi S, Church WB, Morris MB (2008) Modelling the structures of G protein-coupled receptors aided by three-dimensional validation. BMC Bioinformatics 9(Suppl 1):S14 38. Trabanino RJ, Hall SE, Vaidehi N, Floriano WB, Kam VWT, Goddard WA (2004) First principles predictions of the structure and function of G-protein-coupled receptors: validation for bovine rhodopsin. Biophys J 86(4):1904–1921 39. Vaidehi N, Floriano WB, Trabanino R, Hall SE, Freddolino P, Choi EJ etÂ€ al (2002) Prediction of structure and function of G protein-coupled receptors. Proc Natl Acad Sci U S A 99(20):12622–12627 40. Park Y, Helms V (2006) Assembly of transmembrane helices of simple polytopic membrane proteins from sequence conservation patterns. Proteins 64(4):895–905 41. Adamian L, Liang J (2006) Prediction of transmembrane helix orientation in polytopic membrane proteins. BMC Struct Biol 6:13 42. McAllister SR, Floudas CA (2008) Alphahelical topology prediction and generation of distance restraints in membrane proteins. Biophys J 95(11):5281–5295 43. Yarov-Yarovoy V, Schonbrun J, Baker D (2006) Multipass membrane protein structure prediction using Rosetta. Proteins Struct Funct Bioinf 62(4):1010–1025 44. de Brevern AG, Wong H, Tournamille C, Colin Y, Le Van Kim C, Etchebest C (2005) A structural model of a seven-transmembrane helix receptor: the Duffy antigen/receptor for chemokine (DARC). Biochim Biophys Acta 1724(3):288–306 45. Binkowski TA, Naghibzadeh S, Liang J (2003) CASTp: computed atlas of surface topography of proteins. Nucleic Acids Res 31(13):3352–3355 46. Hubbard S, Thornton J (1996) Naccess V2.1.1: atomic solvent accessible area calculations. http://www.bioinf.manchester.ac.uk/ naccess/ 47. Smart OS, Neduvelil JG, Wang X, Wallace BA, Sansom MS (1996) HOLE: a program for the analysis of the pore dimensions of ion channel structural models. J Mol GraphÂ€14(6):354–360, 376

48. Kabsch W, Sander C (1983) Dictionary of protein secondary structure: pattern recognition of hydrogen-bonded and geometrical features. Biopolymers 22(12):2577–2637 49. Heinig M, Frishman D (2004) STRIDE: a web server for secondary structure assignment from known atomic coordinates of proteins. Nucleic Acids Res 32(Web server issue): W500–W502 50. Lomize AL, Pogozheva ID, Lomize MA, Mosberg HI (2006) Positioning of proteins in membranes: a computational approach. Protein Sci 15(6):1318–1333 51. Chothia C, Levitt M, Richardson D (1981) Helix to helix packing in proteins. J Mol Biol 145(1):215–250 52. Fleishman SJ, Ben-Tal N (2002) A novel scoring function for predicting the conformations of tightly packed pairs of transmembrane (alpha)-helices. J Mol Biol 321(2):363–378 53. Gimpelev M, Forrest LR, Murray D, Honig B (2004) Helical packing patterns in membrane and soluble proteins. Biophys J 87(6): 4075–4086 54. Hildebrand PW, Gunther S, Goede A, Forrest L, Frommel C, Preissner R (2008) Hydrogenbonding and packing features of membrane proteins: functional implications. Biophys J 94(6):1945–1953 55. Bansal M, Kumar S, Velavan R (2000) HELANAL: a program to characterize helix geometry in proteins. J Biomol Struct Dyn 17(5):811–819 56. Kumar S, Bansal M (1996) Structural and sequence characteristics of long alpha helices in globular proteins. Biophys J 71(3):1574–1586 57. Lee HS, Choi J, Yoon S (2007) QHELIX: a computational tool for the improved measurement of inter-helical angles in proteins. Protein J 26(8):556–561 58. Debret G, Valadié H, Stadler AM, Etchebest C (2007) New insights of membrane environment effects on MscL channel mechanics from theoretical approaches. Proteins 71(3):1183–1196 59. Guvench O, MacKerell AD (2008) Comparison of protein force fields for molecular dynamics simulations. Methods Mol Biol 443:63–88 60. Sonne J, Jensen MO, Hansen FY, Hemmingsen L, Peters GH (2007) Reparameterization of all-atom dipalmitoylphosphatidylcholine lipid parameters enables simulation of fluid bilayers at zero tension. Biophys J 92(12):4157–4167 61. Davis JE, Warren GL, Patel S (2008) Revised charge equilibration potential for liquid alkanes. J Phys Chem B 112(28):8298–8310

Critical Review of General Guidelines for Membrane Proteins Model Building and Analysis 62. Rosso L, Gould IR (2008) Structure and dynamics of phospholipid bilayers using recently developed general all-atom force fields. J Comput Chem 29(1):24–37 63. Högberg C, Nikitin AM, Lyubartsev AP (2008) Modification of the CHARMM force field for DMPC lipid bilayer. J Comput Chem 29(14):2359–2369 64. Kitao A, Go N (1999) Investigating protein dynamics in collective coordinate space. Curr Opin Struct Biol 9(2):164–169 65. Isralewitz B, Gao M, Schulten K (2001) Steered molecular dynamics and mechanical functions of proteins. Curr Opin Struct Biol 11(2):224–230 66. Haddadian EJ, Cheng MH, Coalson RD, Xu Y, Tang P (2008) In silico models for the human alpha4beta2 nicotinic acetylcholine receptor. J Phys Chem B 112(44): 13981–13990 67. Sperotto MM, May S, Baumgaertner A (2006) Modelling of proteins in membranes. Chem Phys Lipids 141(1–2):2–29

385

68. Feig M (2008) Implicit membrane models for membrane protein simulation. Methods Mol Biol 443:181–196 69. Ritchie DW (2008) Recent progress and future directions in protein–protein docking. Curr Protein Pept Sci 9(1):1–15 70. Reggio PH (2006) Computational methods in drug design: modeling G protein-coupled receptor monomers, dimers, and oligomers. AAPS J 8(2):E322–E336 71. Prusis P, Uhlén S, Petrovska R, Lapinsh M, Wikberg JES (2006) Prediction of indirect interactions in proteins. BMC Bioinformatics 7:167 72. Bruschweiler R (1995) Collective protein dynamics and nuclear spin relaxation. J Chem Phys 102(8):3396–3403 73. Andrusier N, Mashiach E, Nussinov R, Wolfson HJ (2008) Principles of flexible protein– protein docking. Proteins 73(2):271–289 74. Fernández-Recio J, Totrov M, Abagyan R (2003) ICM–DISCO docking by global energy optimization with fully flexible sidechains. Proteins 52(1):113–117

as

Chapter 20 3D Structural Models of Transmembrane Proteins Alexandre G. de Brevern Abstract Transmembrane proteins are macromolecules implicated in major biological processes and diseases. Because of their specific neighborhood, few transmembrane protein structures are currently available. The building of structural models of transmembrane proteins is a major research area. Because of the lack of available 3D structures, automatic homology modeling is not an efficient way of proposing pertinent structural models. Hence, most of the structural models of transmembrane proteins are developed through a more complex protocol that comprises the use of secondary structure prediction to complete the comparative modeling process. Then, refinement and assessment steps are performed go often to a novel comparative modeling process. Nowadays, it is also possible to take attention to the helix–helix and helix–lipid interactions, and even build quaternary structures. In all cases, the most important factor when proceeding to correct structural models is taking the experimental data into account. Key words: Transmembrane protein structures, Multiple sequence alignment, Structural models, Comparative modeling, Homology modeling, Secondary structure prediction, Helix–helix interaction, Helix–lipid interaction, Bilayer membrane, Protein docking

1. Introduction Transmembrane proteins represent ~25% of proteins coded by genomes. They are composed of two major classes: all-a, e.g., rhodopsin and all-b, e.g., outer membrane proteins. They support essential biological functions as receptors, transporters, or channels. They are embedded in lipid membrane that constitutes a very specific neighborhood. As a result of this particularity, obtaining experimental 3D transmembrane structures is difficult. The total number of transmembrane proteins in the Protein DataBank (1) is limited, comprising ~1% of available structures (2). The design of structural models becomes an important axis of research. Indeed, more than two-thirds of the marketed drugs target a transmembrane protein and 50%, specifically a GPCR (3). Jean-Jacques Lacapère (ed.), Membrane Protein Structure Determination: Methods and Protocols, Methods in Molecular Biology, vol. 654, DOI 10.1007/978-1-60761-762-4_20, © Springer Science+Business Media, LLC 2010

387

388

de Brevern

Thus, most of the time it is not possible to work with an experimental structure, and so, the 3D structural model is one of the important research fields for understanding biological mechanisms and interactions (4). We present in this chapter the classical pipeline to build 3D structural models. The most common way to propose 3D structural models is on the basis of a comparative modeling process coupled with transmembrane segment predictions. Nonetheless, the principle goes far beyond the classical homology modeling as often the target structure is not directly related to the query sequence, i.e., it is not possible to simply align the sequence of the protein queries and targets. Thus, it is an iterative process mainly based – when possible – on multiple sequence alignments, bibliographic and web researches, molecular refinement, and helix–helix and helix–lipid interaction prediction tools. Two papers can be read to have pertinent examples of such an approach (5, 6). The chosen structural models must encompass most of the biochemical features and reflect the known experimental data. They may be used to analyze functional interaction properties.

2. Materials A recent computer with an internet connection is sufficient. Numerous software programs are available on-line, but some must be locally installed. Most of these software programs can be used with the Windows Operating System (OS) while a few work only with Linux OS. Tables 1–3 summarize some of the available tools that can be used, e.g., the secondary structure and interaction prediction methods. Updated and additional links can be found at http://www. dsimb.inserm.fr/~debrevern/TM/index.html.

3. Methods The main principle of the approach is described in Fig. 1. The first step consists in the true knowledge of the protein of interest (POI). Three complementary approaches are important: multiple sequence alignment, bibliographic and on-line research, and secondary structure prediction. Then, the alignment of the POI (query) sequence with a (target) sequence of a protein with an available (template) 3D structure must be done. This information is sufficient to build protein structures; thanks to the comparative modeling approach, different structural models can be obtained. They must be refined and evaluated. Appropriate corrections must be done to the original

3D Structural Models of Transmembrane Proteins

389

Table 1 Web services and softwares Name

Purpose

Soft

Url

Protein DataBank

Database of available protein structures

On-line

http://www.rcsb.org/pdb/ home/home.do

Stephen White Laboratory

Transmembrane protein structures

On-line

http://blanco.biomol.uci.edu/ Membrane_Proteins_xtal.html

UniProt KB/ Swiss-Prot

Protein sequences

On-line

http://www.expasy.ch/sprot/

PDBTM

Information on transmembrane proteins

On-line

http://pdbtm.enzim.hu/

OMP

Orientations of proteins in membranes (OPM) database

On-line

http://opm.phar.umich.edu/

GPCRdb

Information on GPCR

On-line

http://www.gpcr.org/7tm/

PSI-BLAST

Search protein database

Linux/on-line

http://blast.ncbi.nlm.nih.gov/ Blast.cgi

Modeller

Homology modeling

Windows/Linux http://salilab.org/modeller/ modeller.html

Swiss model

Homology modeling (and other)

Windows/Linux http://swissmodel.expasy.org/ SWISS-MODEL.html

Homer

Homology modeling

On-line

http://protein.cribi.unipd.it/ Homer/

Consensus

Homology modeling

On-line

http://structure.bu.edu/ cgi-bin/consensus/consensus. cgi

Proteus 2

Homology modeling

On-line

http://wishart.biology.ualberta. ca/proteus2

HOMA

Homology modeling (and other)

On-line

http://www.nmr.cabm.rutgers. edu/HOMA/

Clustalw

Sequence alignment

All

http://www.ebi.ac.uk/Tools/ clustalw2/index.html

T-Coffee

Sequence alignment

Linux/on-line

http://www.ebi.ac.uk/Tools/tcoffee/

Muscle

Sequence alignment

Linux/on-line

http://www.drive5.com/ muscle/

rasmol

Protein visualization

Windows/Linux http://www.rasmol.org/

VMD

Protein visualization

Windows/Linux http://www.ks.uiuc.edu/ Research/vmd/

PyMol

Protein visualization

Windows/Linux http://pymol.sourceforge.net/ (continued)

390

de Brevern

Table 1 (continued) Name

Purpose

Soft

Url

WhatCheck

Structural model assessment

Linux

http://swift.cmbi.ru.nl/gv/ whatcheck/

ProCheck

Structural model assessment

Linux

http://www.biochem.ucl.ac. uk/~roman/procheck/ procheck.html

SCWRL

Side-chain replacement

Linux/on-line

http://dunbrack.fccc.edu/ SCWRL3.php http://www1.jcsg.org/prod/ scripts/scwrl/serve.cgi

Gromacs

Molecular modeling

Linux

http://www.gromacs.org/

Table 2 Transmembrane repetitive structure prediction web services Web services

Year

Url

TMbase

1993

http://www.ch.embnet.org/software/TMPRED_form. html

TopPred II

1994

http://mobyle.pasteur.fr/cgi-bin/MobylePortal/portal. py?form=toppred

TMAP

1994

http://bioinfo4.limbo.ifm.liu.se/tmap/index.html

MEMSAT

1994

http://saier-144-37.ucsd.edu/memsat.html

PredictProtein (include PhDTm)

1995

http://www.predictprotein.org/

TSEG

1998

http://www.genome.ad.jp/SIT/tsegdir/

TM-Finder

1999

http://www.ccb.sickkids.ca/tools/tmfinder/cgi-bin/ TMFinderForm.cgi

PRED-TMR 2

1999

http://athina.biol.uoa.gr/PRED-TMR2/

HMMTOP 2.0

2001

http://www.enzim.hu/hmmtop/

OrienTM

2001

http://o2.biol.uoa.gr/orienTM/

DAS-TMfilter (DAS 2.0)

2002

http://mendel.imp.ac.at/sat/DAS/DAS.html

SOSUI

2002

http://bp.nuap.nagoya-u.ac.jp/sosui/sosui_submit.html

SPLIT 4.0

2002

http://split.pmfst.hr/split/4/

THUMBUP-UMDHMM TMHP

2003

http://sparks.informatics.iupui.edu/Softwares-Services_ files/umdhmm.htm

All-a

(continued)

3D Structural Models of Transmembrane Proteins

391

Table 2 (continued) Web services

Year

Url

Phobius (new TMHMM)

2004

http://phobius.sbc.su.se/

ConPred II

2004

http://bioinfo.si.hirosaki-u.ac.jp/~ConPred2/

PRODIV-TMHMM

2004

http://www.pdc.kth.se/~hakanv/prodiv-tmhmm/

SVMtm

2004

http://ccb.imb.uq.edu.au/svmtm/svmtm_predictor. shtml

TUPS

2005

http://sparks.informatics.iupui.edu/Softwares-Services_ files/tups.htm

Localizome

2006

http://localodom.kobic.re.kr/LocaloDom/index.htm

MINNOU

2006

http://minnou.cchmc.org/

MEMSAT3

2007

http://bioinf.cs.ucl.ac.uk/memsat/

MemBrain

2008

http://chou.med.harvard.edu/bioinf/MemBrain/

BDM

2002

http://gpcr.biocomp.unibo.it/biodec/

TBBpred

2004

http://www.imtech.res.in/raghava/tbbpred/

PROFtmb

2004

http://www.rostlab.org/services/ProfTMB/index.html

BOMP

2004

http://www.bioinfo.no/tools/bomp

PRED-TMBB

2004

http://bioinformatics2.biol.uoa.gr/PRED-TMBB/ index.jsp

TMBETA-DISC

2005

http://psfs.cbrc.jp/tmbetadisc/

TMBETA-SVM

2005

http://tmbeta-svm.cbrc.jp/

TMBETA-NET

2005

http://psfs.cbrc.jp/tmbeta-net/

TMB-HMM

2006

http://bmbpcu36.leeds.ac.uk/~andy/betaBarrel/ TMB_Hunt_2/TMB_HMM.cgi

transFold

2006

http://bioinformatics.bc.edu/clotelab/transFold/

All-b

alignment until a structural model is found interesting. Then, many approaches can be performed to go deeper into the protein structural models (see the other chapters). 3.1. Sequence Analyses: Multiple Sequence Alignment and Bibliographic Researches

1. The correct sequence of the (query) POI must be found. When alternative splicing or mutants are common, careful attention must be paid to the selection of the right sequence. The expasy server (Uniprot) is a valuable web server as it is highly manually curated. 2. Related protein sequences can be found – thanks to PSIBLAST. As a first step, this search is performed using only the

392

de Brevern

Table 3 Helix–helix, helix–lipid and miscellaneous web services Web services

Year

Url

Helix analysis computational tool

2002

http://bioinfo.tau.ac.il/~sarel/HelAna.html (Not web service)

kprot

2004

Not reachable

TMLIP-H

2005

Not simply usable

ProperTM

2004

http://icb.med.cornell.edu/services/propertm/start

TMLIP-C

2005

Not simply usable

LIPS

2006

http://gila.bioengr.uic.edu/lab/larisa/lips.html

TMX

2007

http://service.bioinformatik.uni-saarland.de/tmx/ about.htm

TMRPres2D

2004

http://biophysics.biol.uoa.gr/TMRPres2D/

Archdb

2004

http://sbi.imim.es/cgi-bin/archdb/loops.pl

Triton

2008

http://ncbr.chemi.muni.cz/triton/

Helix–helix

Helix–lipid

Misc.

Protein DataBank. If one hit appears with a correct sequence identity, it must be used as the structural template. However, this is not often the case. 3. As the second step, related sequences must be found to analyze evolutionary information. The search is so done on the Uniprot and on the non-redundant databank of NCBI. As PSI-BLAST can take very redundant sequences and also select short sequences, the researcher must carefully select interesting proteins. The NCBI web service of PSI-BLAST is well conceived for such pruning schema (see Note 1). 4. PSI-BLAST has not been created to optimally align sequences, but to search for sequences related to one query sequence. The selected related proteins must then be aligned into a multiple alignment with dedicated software like Clustalw, T-Coffee, or Muscle. It allows locating conserved and non conserved regions of the protein sequences, i.e., more selective pressure can be expected to conserved regions. They are important for the protein fold/function. In contrast, nonconserved regions can be potentially more flexible. A phylogenetic approach – with a dendogram – is an efficient way to analyze the raw data (see Note 2).

3D Structural Models of Transmembrane Proteins

393

Fig. 1. Pipeline to build a structural model of a transmembrane protein structures. The protein sequence of interest is analyzed, thanks to multiple sequence analysis, bibliographic and on-line researches, and secondary structure prediction methods. Then a comparative modeling with constraints is performed. The structural models are refined and analyzed. It is an iterative process; the local modification of alignment gives new structural models which must be analyzed.

5. A bibliographic research on the POI is obvious, but it is also important to search information on related protein sequences. In the same way, a good knowledge of related protein folds is essential. For this purpose, dedicated transmembrane protein websites, e.g., Stephen White, PDBTM, and OMP, have a lot of pertinent information, both for the specialists and non- specialists. If the protein is a GPCR, GPCRdb is also an interesting tool to obtain information of this kind of protein. 6. Biological and experimental data are essential to build pertinent structural models. Summarize the biological data in a table with the amino acid position, its kind, and the corresponding biological property(ies) (e.g., see Table 1 of (5)) (see Note 3). 3.2. Of Helices and Transmembrane Segments

1. Table 2 summarizes different and recognized transmembrane secondary prediction methods. These web services are on the basis of different approaches (hydrophobicity, sequence alignment, artificial neural networks…) and of different protein structure databanks (ranging from the oldest one to the most recent ones). The principle is to put the sequence in a form and wait for the results.

394

de Brevern

2. Results can be given in different ways ranging from the simple delimitation of the transmembrane repetitive secondary structure to the localization of extra- and intra-cellular loops and even a confidence index to the quality of the predictions (see Note 4). 3. As the number of available transmembrane protein structures is limited, the prediction rate is to be taken carefully when no homologous sequences is associated to a known structure. Summarize the results in a table (e.g., see Table 2 of (5)). The high number of different approaches allows finding average helix number and positions; they will serve to start the comparative modeling. Bibliographic data could also eliminate a method which seems inadequate for this protein, e.g., a method which misses half of the helices. In the same way, the confidence index can explain quantitatively why some regions are predicted very differently by the prediction methods. Indeed, some regions are hard-to-predict and require a special care (see Note 5). 4. An interesting point is to predict the secondary structure of the target structure, i.e., the structure of the sequence that will be aligned with the query sequence. It permits to look at the behaviors of each secondary structure prediction method. 3.3. Comparative Modeling and Other Approaches

1. Some web services able to propose structural models are available, e.g., Homer, Consensus, Proteus 2, or HOMA. As in the case of the other web services, the sequence must be put in a form. Numerous methods are not dedicated to transmembrane proteins and cannot find correct transmembrane models. Nonetheless, they must be tried as they can give some hints about the compatibility folds. 2. Comparative modeling can be done with Modeller software, which is the most known and most used homology modeling software. It uses an alignment between one query and one target sequence; the target sequence must have a known available (template) structure. The crucial part is the quality of the alignment. Concerning our purpose, it is also the converging of repetitive secondary structures. Figure 2 summarizes the principle: (a) The POI (query sequence) must be cut in transmembrane domain, i.e., the region with the transmembrane repetitive secondary structure (TRSS) and connecting loops, and the N and C termini regions. In Subheading 3.2, initial delineation of repetitive secondary structure has been performed. (b) The TRSSs of the target structure must be precisely located. (c) These TRSSs must be aligned with the corresponding TRSSs of the target structure. Same thing must be done for the loops, resulting in a global alignment of

3D Structural Models of Transmembrane Proteins

395

Fig. 2. Construction of the alignment of the transmembrane protein. The protein sequence must be cut into three parts: transmembrane domain and the N and C termini regions. (a) The predicted repetitive structures are placed on the sequence. (b) The repetitive structures are assigned on the template structure. (c) The correspondence between predicted and assigned regions must be made out. Each local alignment is optimized with dedicated software such as Clustalw. (d) The N and C termini regions can be treated as globular proteins, and specific predictions can so be made (or not). (e) A complete alignment is done corresponding to the entire protein. >P1;1F88 structureX:1F88: 1 :A : 348 : :RHODOPSIN : BOS TAURUS: ----MNGTE-GP--------NFYVPFSNKTGVVRSPFEAPQYYL---A-EP WQ----FS--MLAAYMFLLIMLGFPINFLTLY-VTVQHKKLR-TP-LNYIL LNLAV-ADLF-MVFGGFTTTLYTSLHGYFVFGPTGC-NLE--GFFATLGGE IALWSLVVLAIERYVVVCKPMSNFRFGENH---AIMGVAF-TWVMALACAA PPLVGWSRYIPEGMQCSCGIDYYTPHEETNNE---SFVIYMFVVHFIIPLI VIFFCYGQLVFTVKEAAA----SATTQKAEKEVTRMVIIMVIAFLICWLPY AGVAFYI--FTHQ-----GSDFGP-----IFMTIPAFFAKTSAVYNPVIYI -MMNKQFRNCMVTTLCCGKNP--------------------* >P1;DARC_H1 sequence:DARC_H1: 1 : : 336 : : DARC HUMAN :: : MGNCLHRAELSPSTENSSQLDFEDVWNSSYG-VNDSFPDGDYDANLEAAAP CHSCNLLDDSALP-FFILTSVLGILASSTVLFMLFRPLFRWQLCPGWP-VL AQLAVGSALFSIVVPVLAPG----------LGSTRSSALCSLGYCVWYGSA FA--QALLLG-------CHASLGHRLGAGQVPGLTLGLTVGIWGVAALLTL P-VT-LASGASGGL---CTLIYSTELKA----LQATHTVACLAIFVLLPLG --LFG-AK---GLKKALGMGPGPW-------------MNILWAWFIFWWPH -GVVLGLDFLVRSKLLLLSTCLAQQALDL-LLNLAEALAILHCVATPLLLA LFCHQATR-TLLPSL-----PLPEG-WSSHLDTLGSKS---*

2.80 :-1.0

Fig. 3. Example of protein sequence alignment usable by Modeller. The alignment was done between the sequence query DARC and the sequence of the target (structural template), the rhodopsin. The alignment file is named darc.ali.

transmembrane domain. (d) The N and C termini regions can be considered as globular. So, classical homology modeling can be done with software of Table 2. Depending on the size of and compatibility with related structures, it can be less or more complicated. (e) An entire alignment can also be done. Figure 3 shows an example of an alignment usable by Modeller

396

de Brevern

(PIR format). This example is the alignment of (target) rhodopsin with (query) DARC. 3. Modeller needs not only this alignment, but also the corresponding template structure and a simple script which summarizes all this information. The template structure needs to be strictly equivalent to the sequence of the structure, i.e., the sequence present in the PDB file. Indeed, it is classical to have not a complete equivalence of protein sequence and the resolved protein. For instance, the PDB structure of the rhodopsin (PDB code 1F88) has missing atoms within a helix. The correspondence between sequence and structure of rhodopsin is not complete. Figure 4 gives an example of a Modeller script. 4. A series of models (at least 100) must be generated. It corresponds to the a.ending in the Modeller script. The recent Modeller software use Python language (see Note 6). 5. It is often interesting to explicitly add constraints on TRSSs, i.e., to add in the script the position of TRSSs to force the conservation of repetitive structures in the generated models. In the same way, constraints on known disulfide bridges or distances between residues lead to enhanced structural models. In Fig. 4, the disulfide bond constraint has been added (class MyModel) for disulfide bonds 51–276 and 129–195 (see Note 7). # Homology modeling by the automodel class from modeller import * # Load standard Modeller classes from modeller.automodel import * # Load the automodel class # Redefine the special_patches routine to include the additional disulfides # (this routine is empty by default): class MyModel(automodel): def special_patches(self, aln): # A disulfide between residues 51 and 276: self.patch(residue_type='DISU', residues=(self.residues['51'], self.residues['276'])) # A disulfide between residues 129 and 195: self.patch(residue_type='DISU', residues=(self.residues['129'], self.residues['195'])) log.verbose() env = environ()

# request verbose output # create a new MODELLER environment to build this model in

# directories for input atom files env.io.atom_files_directory = ['.', '../atom_files'] a = automodel(env, alnfile = 'darc.ali', knowns = '1F88', sequence = 'DARC_H1') a.starting_model= 1 # a.ending_model = 100 # # # a.make() #

Fig. 4. Example of Modeller script.

# alignment filename # codes of the templates # code of the target index of the first model index of the last model (determines how many models to calculate) do the actual homology modeling

3D Structural Models of Transmembrane Proteins

3.4. Structural Models: Assessment and Refinement

397

1. Modeller or equivalent approaches are quite good to define a correct topology. However, the structural models must be refined to obtain a better geometry. Thus, a first step often consists in selecting few models from the generated ones. It is possible to only look at the objective and DOPE functions given by Modeller for each of the generated models. Nonetheless, these functions can be used carefully as they were not fully tested with transmembrane proteins. Moreover, these functions can be highly sensitive to flexible regions. For instance, the DARC protein has a long flexible region (5), namely ECD1; these functions were only underlining the different conformations of this loop (only 20% of the protein). They were not discriminating (see Note 8). 2. Side-chains must be replaced, thanks to a performing method, e.g., SCWRL. 3. A minimization of the generated structural models is also a good requirement. The most powerful software today is Gromacs (see Chapter 21 for details). 4. Then, a precise analysis of the models must be performed. Visualization of the protein structures with dedicated software, e.g., rasmol, PyMol, or VMD, is a first obligated step. 5. Recognition of different types of errors in 3D models can be done by different software, e.g., verify 3D or Prosa. They are on the basis of sequence–structure relationship statistics deduced from non-redundant databank. Moreover, no equivalent method is available for transmembrane proteins. Thus, only geometry of the structural models can be checked, e.g., atom distances, angle values… The most-used dedicated software are ProCheck and WhatCheck. They give numerous values with summaries. They can highlight part of the structural models with wrong geometry or particular residues with a strange conformation. An important point is also to look at the target structure. Indeed, the template structure can already have some local un-canonical conformation. 6. The chosen structural models must encompass most of the biochemical features and reflect the known experimental data. If experimental data have shown that some residues are accessible, they must be found accessible in the selected models. Otherwise, the alignment must be corrected and the process of comparative modeling and analyzes done again. For instance, the building of DARC (5) has needed ten consecutive rounds with manual corrections of the alignment (see Note 9). 7. Other information can be used (see the sections below).

3.5. Helix–Helix Interactions

1. Interaction between helices are well-described for globular proteins while for transmembrane some interesting works have been published. To have an idea of the interaction zones

398

de Brevern

between helices is a major advantage in proposing pertinent structural models. Table 3 shows the available methods of helix–helix and helix–lipid interactions. 2. A very elegant research (7) has shown the possibility to compute the compatibility between two helices and deduced rules. Specific researches have quantified these compatibility statistics. In this limited field, kPROT is the most renowned approach. Nonetheless, it is not sometimes reachable. Other approaches are not fully automated and so researchers need a strong background in bioinformatics to deal with this kind of approach. 3.6. Helix–Lipid Interactions

1. These interactions are also essential for the protein folding and biological functions. It was less analyzed than helix–helix interactions, but nowadays, the number of available methods is higher than for helix–helix interactions. Often, e.g., LIPS, they also provide hints to regions of interaction with other helices. 2. For LIPS, a multiple sequence alignment must be given, but the alignment concerns only one helix and must not contain a gap. So, important further work is required to prepare the data. 3. As a result, the web server does not give a single result, but the quantification of each possible face (7). Thus, a careful research of correspondence between these predictions and the selected structural models can be performed to improve the quality of the models.

3.7. Of Loops and Mutants

1. Specific research can be done on extra- and intra-cellular loops. It is especially interesting in the case of GPCRs. Indeed, it is a common hypothesis that the seven transmembrane helices have a strong conservation among their entire folds. This hypothesis is strongly supported by biochemical, biophysical, and biological experiments. Hence, a major divergence between the GPCRs is the conformation of loops. A database like ArchDB can be used to select (manually) some alternative potential conformations. Then, building of an alternative conformation can be generated, thanks to comparative modeling. The simplest approach is to use the selected models as the query and the template. In the target sequence, place gaps at the position of the loops. In the example of Fig. 3, it could correspond to putting two times the DARC_H1 sequence, one as the query sequence, one as the target sequence, and putting the obtained structural models as the template. In the target sequence, put gaps at the position of the desired loop. Add a new sequence of the length of query sequence, but only with gaps. Add also the name of this sequence in the Modeller script (line sequence).

3D Structural Models of Transmembrane Proteins

399

At the position, change the gaps by the sequence of the loop. In the same way, add its PDB file into the script (to the line knows). It is possible to add constraints to all the rest of the protein. Molecular dynamics (or a simulated annealing) can help to analyze the flexibility of the loops (see Chapter 21). Recently, a novel class (MyLoop) in Modeller has been proposed to refine the loops; it is on the basis of the optimization of Modeller’s DOPE function. 2. Another potentiality of in silico building of transmembrane protein structural models is the proposition of “supervised” mutants. As the final selected models reflect the biological data, an analysis of the surface underlines the important positions. It is possible to mutate in silico the residues (a specific class exist in Modeller) and analyze a possible consequence of this mutation. A simple idea is to observe the electrostatic properties of the protein. Electrostatic potential mapped on the molecular surface can be coarsely done with Swiss-PdbViewer (aka DeepView) from SwissModel. It allows ranking of the mutation to be proposed. 3. Quaternary structures can be computed in silico. Two approaches exist: (a) from a known available complex or (b) by docking. For the first possibility, it consists – roughly – in the comparative/homology modeling of the different partners. The hypothesis is that the interaction regions are re-conserved and so, must also be in the structural model complex. The second case is most common. If the protein complex is an homomer with a known symmetry, it is a constraint that can be directly used during the comparative modeling approach. Otherwise, each partner must be built and docking approach must be used. As for the building of one protein structural model, the building of a complete quaternary structure is greatly enhanced if biological data are included.

4. Notes 1. For the multiple sequence alignment, it is important to control the redundancy of the data as the length of the sequences. (a) A cluster of highly redundant sequences is not informative and bias the alignment analysis. (b) PSI-BLAST often keep small sequence fragments; they are not pertinent in the case of a protein multiple sequence alignment. 2. With a pertinent multiple sequence alignment, it is also possible to locate important residues in related sequences. These positions can be used as biological constraints, the most important ones.

400

de Brevern

3. Summarize the data in tables, for the prediction and for the important residues. Note that in a mutation abolishing a binding, an interaction is not always present directly at the binding site, but it can also be a crucial residue implicated in the maintaining of the fold (important is not always accessible). 4. The prediction index is a great tool to analyze the difficulty of prediction. Some regions are clearly repetitive structures embedded in the bilayer or loop swimming in water while others are complicated. These latter must be carefully checked at each step of the process. 5. As the number of available protein structure is quite low, it is very important to analyze properly the known folds and the rules (8) that govern this kind of protein folds. However, do not blindly follow the common features of transmembrane proteins. For instance, Tryptophan is known to be preferred at the lipid− water interface, but often it is not the case. 6. Multiple structural models must be developed to find at least a pertinent one. In the same way, the testing of alternative alignment is an essential task. Another interesting approach is to predict also structural models of related proteins. In the same way, if one structural model of a related protein is available, it can be used as a structural template. 7. Proline and kinks in repetitive structures are always a problem. For instance, rhodopsin has three kinked helices and the K+ channel one as essential features. 8. N and C termini regions of the transmembrane protein can be long fragments. In this case, it is possible to use classical methods available on the web, e.g., Proteus 2, and use the results as novel template only for one part of the template (see Subheading 3.7). In the same way, the use of multiple templates is an excellent tool to sample the potential conformations of the structural models. 9. As presented in this chapter, it must be noticed that the building of a pertinent transmembrane protein structural model is mainly a manual approach and that the most important constraint is the biological (experimental) one.

Acknowledgments This work was supported by French National Institute for Blood Transfusion (INTS) and French Institute for Health and Medical Care (INSERM) and University Paris 7 – Denis Diderot.

3D Structural Models of Transmembrane Proteins

401

References 1. Berman HM, Westbrook J, Feng Z, Gilliland G, Bhat TN, Weissig H, Shindyalov IN, Bourne PE (2000) The Protein Data Bank. Nucleic Acids Res 28:235–242 2. Fleishman SJ, Unger VM, Ben-Tal N (2006) Transmembrane protein structures without X-rays. Trends Biochem Sci 31: 106–113 3. Klabunde T, Hessler G (2002) Drug design strategies for targeting G-protein-coupled receptors. Chembiochem 3:928–944 4. Radestock S, Weil T, Renner S (2008) Homology model-based virtual screening for GPCR ligands using docking and targetbiased scoring. J Chem Inf Model 48: 1104–1117

5. de Brevern AG, Wong H, Tournamille C, Colin Y, Le Van Kim C, Etchebest C (2005) A structural model of a seven-transmembrane helix receptor: the Duffy antigen/receptor for chemokine (DARC). Biochim Biophys Acta 1724:288–306 6. Hazai E, Bikadi Z (2008) Homology modeling of breast cancer resistance protein (ABCG2). J Struct Biol 162:63–74 7. Fleishman SJ, Ben-Tal N (2002) A novel scoring function for predicting the conformations of tightly packed pairs of transmembrane alpha-helices. J Mol Biol 321:363–378 8. von Heijne G (2007) The membrane protein universe: what’s out there and why bother? J Intern Med 261:543–557

as

Chapter 21 Molecular Dynamics of Membrane Peptides and Proteins: Principles and Comparison to Experimental Data Patrick F.J. Fuchs Abstract Molecular dynamics (MD) simulation is a standard tool used to assess the motion of biomolecules at atomic resolution. It requires a so-called “force field” that allows the evaluation of an empirical energy from the 3D coordinates of the atoms in the system. In this chapter, the application of MD simulations to membrane proteins and peptides is described with a particular emphasis on the comparison of MD results to experimental data. Such a comparison can be used either for (1) validating the results of a simulation, (2) interpreting an experiment at the atomic level, or (3) calibrating the force field. This last step is particularly important for the use of MD as a predictive tool. As an illustration, a comparison of 2H NMR experiments to MD simulations of a transmembrane peptide is presented and discussed. Key words: Molecular dynamics, Membrane protein, Membrane peptide, Comparison to Â�experimental data, WALP23, 2H NMR

1. Introduction Molecular dynamics (MD) simulations have become a powerful tool for assessing the motion of biomolecules at atomic or molecular resolution (1–3). Using an empirical energy function, the basic principle of MD is to integrate Newton’s equations, giving thus a collection of coordinates of all the particles of the system (e.g., a membrane protein within a bilayer of lipids surrounded by water molecules) as a function of time (i.e., a trajectory). With the explosion of computer capabilities as well as improvements of computing methodologies and algorithms, it is now possible to simulate systems of hundreds of thousands of atoms over tens (sometimes hundreds) of nanoseconds (this aspect is further developed in the next chapter). MD applied to membrane objects (peptides and proteins) has started in the 1990s (4) and is now Jean-Jacques Lacapère (ed.), Membrane Protein Structure Determination: Methods and Protocols, Methods in Molecular Biology, vol. 654, DOI 10.1007/978-1-60761-762-4_21, © Springer Science+Business Media, LLC 2010

403

404

Fuchs

routinely used with an increasing number of studies (reviewed, for example, in refs. 5, 6 for membrane proteins). Most of the time, MD simulations complement one (or many) experimental work(s) (7, 8), allowing an atomic/molecular description of the process under study, such as a specific function of a membrane protein (e.g., channeling of ions across the membrane, signal transduction, etc.), the binding to the membrane of a peptide (e.g., binding of an antimicrobial peptide). This comparison is not always easy given the variety of information experiments can bring, such as functional data (e.g., effect of mutations on the protein activity) or biophysical data (e.g., orientations of chemical bonds derived from nuclear magnetic resonance). The approach also works the other way round, that is, MD can suggest new experiments by bringing a molecular description of the problem. In this chapter, I briefly describe the tools used for performing MD simulations specifically on membrane proteins/peptides and emphasize the strategies, allowing a sound comparison of simulations to experimental data. As a recent issue of Methods in Molecular Biology (Vol.â•›443) was entirely devoted to computer simulations, I just recall to the reader the basics of MD and then concentrate on the comparison between experimental data and simulations.

2. Theory Even if we manipulate “virtual” objects in the field of computer simulations (usually referred to as in silico), I briefly overview here what is needed to perform an MD simulation in terms of hardware, software, and other tools/algorithms/concepts. 2.1. Computers

The basic need when one wants to perform an MD simulation is a suitable computer or access to a computing facility. This can be (1) a single workstation, (2) a cluster of computers (many computers working together), or (3) a bunch of Central Processing Units (CPUs) on a super computer (see Note 1). Point (1) is the easiest way of performing an MD; it can even be done on a home computer (under the appropriate operating system, usually Unix or Unix-like). Current 2Â€GHz (gigahertz) computers with 64 bits multicore CPUs (of course, the higher the number of CPUs, the faster the simulation, thus the longer the trajectory one can simulate), such as Intel or AMD, work very well. Solution (2) is on the one hand more expensive and requires more complicated hardware and software administration than (1), but on the other hand, it gives really much more computing power. Solution (3) needs to write a (small) scientific project to be granted some hours of computation on a super computer. A list of the top 500 world

Molecular Dynamics of Membrane Peptides and Proteins

405

super computers is available at the following URL: http://www. top500.org/. Last but not least, MD simulations can generate many and very large files (to store the trajectories) and thus require a large amount of storage (at least a few hundreds of GigaBytes, even more depending on how big the system is, and how long the trajectory one wants to simulate). 2.2. Force Field

MD simulations require a so-called “force field” in order to be performed. A force field is a set of parameters to which a functional form of the potential energy is associated (9–12). Basically, it allows the calculation of the (empirical) energy of a molecular system from its 3D coordinates, as well as the forces acting on each atom using the first derivative of the energy with respect to the coordinates. The most commonly used force fields for simulations of peptides and proteins are AMBER (13), CHARMM (14), GROMOS (15), and OPLS (16) see e.g., refs. 3, 9–13 for more details). Simulations of membrane peptides/proteins within their native environment also require the use of a force field for phospholipids. Currently, two main force fields are available, the set of Berger (17) and that of CHARMM (18) recently reparameterized (19) (see Note 2). Berger lipids work with an old version of GROMOS (20) for the protein and have been modified in order to work with OPLS (21). A discussion on how to combine those two force fields has been recently published (22). Obviously, CHARMM lipids (18, 19) work with the CHARMM force field for proteins (14). Recently, another alternative became available with the development of phospholipid parameters compatible with the AMBER force field (23). Last, we need a model to simulate water. The two standard models are SPC (24) (working with GROMOS and Berger lipids) and TIP3 (25) (working with CHARMM, AMBER, and OPLS). Note that the use of the lipid force fields described above implies that we simulate explicitly their presence, that is, we construct the coordinates of each one of them and calculate explicitly the microscopic interactions between them and the protein/peptide; the same applies for water. The other alternative is to use an implicit representation of the membrane (26) where this latter is treated as a continuum (as well as for water). This has the advantage to speed up the calculation time, but the main drawbacks are the overestimated kinetics and the loss of microscopic details of the interactions between the protein/peptide with lipids and water. In the following, I focus only on explicit simulations, which are nowadays the standard approach for membrane systems.

2.3. Software

The most widely used software for MD of membrane peptides/ proteins are GROMACS (27) and NAMD (28). GROMACS natively implements the GROMOS force field and the Berger lipids have been imported there (29), whereas NAMD natively

406

Fuchs

implements the CHARMM force field. The advantage of these two programs is the high speed of the code and the efficiency of the parallelization (the efficiency to run the program on several CPUs) (see Note 3). There are some other alternatives such as CHARMM or AMBER, which are programs that implement their own force field, but they are generally slower than GROMACS or NAMD. 2.4. Protein/Peptide 3D Structure

MD simulations of a membrane protein or peptide of course require the 3D structure of the object one wants to embed in the bilayer. For membrane proteins, the first source is the protein data bank (30); Stephen White’s lab also maintains an up to date list of membrane proteins of known 3D structure (31). When no 3D structure of the protein of interest exists, it is possible to build a (homology) model based on a homolog which shares sufficient sequence identity (this topic is largely developed in the previous chapter). For peptides, it is generally easier as they are most of the time alpha-helical. One has just to take care of the partitioning of the peptide within the bilayer, that is, whether it is a transmembrane (that spans across the bilayer) or interfacial (that lies at the interface between the phospholipids and the aqueous phase) peptide. The initial structure can be easily constructed with the appropriate f/y angles of an a-helix using structure generators such as Pymol (32, 33), Swiss-Pdb viewer (34, 35), or some web services like Basic Builder hosted on the RPBS platform (36, 37).

2.5. Preparation of the System

Various strategies have been developed to embed the protein/ peptide within the bilayer (38–42). In the 1990s, the first membrane protein (or peptide) simulations used tricky strategies to shorten as much as possible the equilibration phase (4). Nowadays, with the speed of current computers and available MD codes, the fastest strategy just requires a pre-equilibrated bilayer of phospholipids, in which one can make a hole (slightly larger than the protein) (38) and/or adapt the lipids position around the protein (42). Then, an appropriate equilibration of a few nanoseconds (usually maintaining the protein fixed by position restraints) with pressure coupling will shrink the box and pack the lipids against the protein. Most of the time, this step still requires a manual input even if efforts are underway to help build protein/peptide/ membrane systems (40, 41).

2.6. Simulation Parameters

Once the object is embedded in the bilayer, it is time to run the MD using the appropriate algorithms and parameters. Two issues are of critical importance when dealing with membrane simulations, the way of controlling the pressure of the system and computing electrostatics (43). Nowadays, membrane simulations

Molecular Dynamics of Membrane Peptides and Proteins

407

are most of the time run under the so-called “NPT” ensemble, that is, the number of molecules (N) is fixed, and we couple the system to a thermostat and barostat to get the temperature (T) and pressure (P) constant, respectively; the other alternative is the use of the NPgT ensemble, where g stands for constant surface tension (see Note 2). Coupling the system to a pressure bath basically needs to scale the dimensions of the box (x,y,z), which can be done using the same scaling factor for all directions (isotropic coupling), the same scaling factor for x and y directions but a different one for the z direction (semi-isotropic coupling), or independently in all directions (anisotropic coupling). In all these possibilities, it is recommended that semi-isotropic coupling (see Note 4) be used when simulating lamellar systems. One popular algorithm for applying pressure coupling is the one of Berendsen (44) (see Note 5). The other important issue in membrane protein simulation is the way of computing electrostatics when evaluating the energy of the system. Nonbonded interactions (including van der Waals and electrostatics) represent the most expensive computational burden (see Note 3). To alleviate the computing effort, various strategies were created in the early time of MD in the 1970s and improved so far. They generally rely on the so-called cutoff. This latter stands for an atom–atom distance beyond which the interaction is considered equal to 0; since both electrostatics and van der Waals are inverse power functions (of the distance between atoms), they quickly tend to 0 at long distances. Although this approximation can be acceptable for simple Â�systems, it has been demonstrated to cause numerous artifacts on highly charged systems such as ionic solutions. It leads to a wrong radial distribution of the ions, which tend to be separated by the cutoff distance while they should not (45). To circumvent this problem, two main techniques/algorithms are used in biomolecular simulations, the particle-mesh-Ewald (PME) (46, 47) and the reaction field (48). On pure phospholipids, the use of cutoff gives the wrong area per lipid; they indeed tend to be too packed, thus the membrane thickness is overestimated (49, 50). For simulations of membrane proteins (or peptides) within a bilayer of phospholipids, the use of cutoff has been shown to not affect directly the protein nor the water, but only the phospholipids (51). Nonetheless, it is still highly recommended not to use cutoff schemes if one wants to avoid artifacts due to a wrong thickness/area per lipid, which may cause conformational consequences due to different matching/mismatching conditions of the environment around the protein (or peptide). Currently, PME is the most correct and most used technique for computing electrostatics on membrane systems (49, 50) (see Note 6).

408

Fuchs

3. Comparison of MD Simulations to Experimental Data

3.1. General Considerations on the Comparison of MD to Experiments

In this section, I first highlight some issues to consider when one wants to compare an MD trajectory with experiments, notably sampling. I then briefly describe and discuss the practical aspects of comparing a simulation to both biochemical and biophysical data. I then give a complete example of such a comparison in a study of the orientation of a helical peptide within a phospholipid bilayer. Once the production run has been done, it is time to analyze all the relevant data one can extract from the trajectory(ies). Sooner or later, the results will have to be compared to experimental data, either to validate the simulation or to interpret the data at atomic/ molecular level (7, 8, 43, 52). This is not an easy task. The main reason of this comes from a difference of scale, both in time and space, of many orders of magnitude. The first issue is the quantity of molecules. During an experiment in the real world, one deals with a number of molecules in the sample that is on the order of the Avogadro number (~6.1023 molecules), whereas the simulation box represents a microscopic view of the system (103–105 molecules), with a single molecule of solute if we consider for example a membrane protein within a bilayer (surrounded by water). One of the drawbacks is that we neglect the interactions between the multiple copies of the object (dimerization, aggregation…). The second issue is the time scale. A typical experiment generally ranges from microseconds to seconds depending on the technique; we thus make the approximation that the quantity of interest averages in the same way during the full time of the experiment than in the simulated time. Consequently, we have to ensure that the simulation time is greater (at least of one order of magnitude) than the correlation time of the measured quantity, in order for this quantity to relax (see below). Some of the experimental results are even harder to predict since most of the biologically relevant motions take place on longer time scales (2, 53). However, there is an overlap that begins to arise between experiments and simulations (around the microsecond) (2), notably using coarse-grained models (54) (see next chapter for more considerations on this topic). What links both experiments (macroscopic) and simulations (microscopic) is statistical mechanics (see ref. 55 for a brief introduction, and all references therein), that is, in the limit of infinite sampling of simulations, macroscopic (bulk) properties can be explained by the individual interactions between the atoms of the system (thus on a microscopic description of the system). Of course, there is a first inherent limitation to this, since we sample a system in a simulation over a finite time (a few tens to

Molecular Dynamics of Membrane Peptides and Proteins

409

hundreds of nanoseconds with current computers). The ergodic hypothesis states that averaging a property over an infinite time is equivalent to averaging over an appropriate statistical ensemble of configurations of the system. Such an ensemble is most of the time impossible to obtain on complex biological macromolecules using standard MD simulations. At least, one has to make sure that the property of interest is at equilibrium by plotting its autocorrelation function (43); usually, an exponential decay is observed, which allows the extraction of the relaxation time of the observable. One has then to verify that the simulated time is (at least) greater than one order of magnitude of the relaxation time. It allows one to get good statistics to calculate a reliable average of the observable. If a given event is observed once in a simulation, it is highly recommended to run many clones of the same simulation (at least three using different initial conditions). It enhances the sampling and ensures that this event of interest does not depend on the initial conditions and thus occurs by chance. If the event is very rare, it might never occur in a classical MD simulation because the free energy barrier associated to it is too high. In such a case, free energy techniques are adapted to force the system to go from one state to another or along a given reaction coordinate (e.g., the position of a small solute along the main axis of a channel) even if a significant free energy barrier is present (55). Last, one limitation to bear in mind before starting any simulation is that it cannot reproduce the experimental complexity of the biological system under study (medium, molecules around, etc.). One spectacular example is the constitution of biological membranes, which contain several different types of lipid (phospholipids, sphingolipids, cholesterol, etc.) and are sometimes asymmetric such as in red blood cell membranes (56); simulations can hardly model this complexity, and we usually simplify the model by taking into account only the most represented lipid. 3.2. Comparison of MD to Biochemical Data

There is a great variety of information that can be extracted from biochemical experiments, which can give useful insights such as (structural) stability, ability to bind some ligands or to accomplish a given function (e.g., catalysis), etc. Usually, the experimentalist will observe the consequence of a given change in the “normal” system on a given observable. The purpose of this chapter is not to enumerate exhaustively all the possible experiments and consequences, but we can cite a few of them, e.g., the mutation of one (or a few) residue(s) according to the wild type protein, addition of a product/cofactor/inhibitor/drug, modification of a physical condition (temperature, pressure, ionic force), etc. For instance, if some parameters are available, it is possible to make a simulation of a protein with another molecule added to the system that has been shown experimentally to perturb the protein. As a response

410

Fuchs

to the change of condition, the experimentalist will often observe a qualitative or quantitative variation of the activity of the protein such as ligand binding, conductance of ions, ability to transmit the signal across the membrane etc. This can also be at higher level, such as the cellular level. Of course MD simulations cannot reproduce the complexity of the experiment but might be able to give some clues on the early molecular event(s) generated by the change of condition. For example, in the case of mutagenesis data, it is trivial to mutate in silico a residue and assess the effect of the mutation on the structure by running simulations of the wild type and of the mutant. One useful modeling technique for the study of mutation is the evaluation in silico of the free energy of a (bio)chemical process, for example, the variation of the free energy of binding of a ligand to a receptor between the wild type protein and one of its mutants; Chipot gives an example of such a calculation applied to the binding of a ligand to a GPCR (55). I now give two examples taken from recent literature that illustrate the comparison of experiments to MD simulations. The first deals with a very recent work on the SecYE translocon (57), which is involved in the translocation of proteins across the endoplasmic reticulum membrane. The authors could solve the crystal structure of an open state of SecYE by cocrystallizing the protein with an antibody. Additionally, they performed disulfide mapping experiments, which suggested that the closed form is more stable when nothing is bound to the protein, whereas the open form occurs when a ligand is bound. To assess this hypothesis, they ran an MD simulation of 100Â€ns starting from the open state (with nothing bound to it) embedded in the lipid bilayer. The MD confirmed the hypothesis since the protein reached the closed state over the 100Â€ns. This interesting example shows the synergy between MD and a typical biochemical experiment. A second example deals with the permeation of aquaporins and aquaglyceroporins (58). These latter allow an efficient and selective permeation of small solutes across the membrane. In this work, the authors studied their selectivity by computer simulations, using umbrella sampling (59). This technique permits the evaluation of the potential of mean force (PMF) along a chosen reaction coordinate, giving an estimation of the free energy barrier. In this case, the PMF was evaluated along the z position of the solute within the pore and gave the free energy barrier encountered by the solute when passing through the porin. Interestingly, the authors could evaluate that ammonia crosses a barrier of 12.5Â€kJ/mol, which is in line with permeation experiments. Additionally, this work allowed the authors to identify which region in the pore acts as a selectivity filter. In this example, the experiment validated the simulation, and this latter gave meaningful atomic/molecular details. More generally, PMF is

Molecular Dynamics of Membrane Peptides and Proteins

411

now a standard computational tool to examine permeation through protein channels and is often compared to binding and conductance experiments (60). For more examples on the comparison of experiments with MD simulations, one can read the following two references. Corry and Martinac showed how the computational studies complemented some experimental techniques (EPR, FRET spectroscopy, Patch Clamp) on the comprehension of bacterial mechanosensitive channels (MscL and MscS) (61). Dodson etÂ€al. give some more examples of successful comparisons between MD simulations and experiments that help understand the underlying biological mechanism under study (1). 3.3. Comparison of MD to Biophysical Data

I consider the case of biophysical data separated from biochemical data because of the nature of information that can be injected into or compared to the simulations. Usually, biophysical data can lead to fairly precise structural observable(s) and is thus very useful for e.g., lipid force field calibration (e.g., area per lipid from SAXS or NMR), or can very precisely be compared to simulation results. Moreover, it is often possible (and desirable) to back-calculate directly the experimental observable and compare it directly with the simulations results (e.g., a quadrupolar splitting evaluated from the orientation of a bond vector, see next section for an example). Nonetheless, comparing MD simulations to biophysical data is very challenging because of the difference of time scale. Usually, experiments based on NMR or X-ray and neutron scattering (SAXS, WAXS, SANS…) give access to useful observables. The main difficulty is that we generally have no idea of the shape of the distribution of these observables (8) as the experimental measurement gives a unique value (or set of values), which represents an average over the time scale of the experiment. Different distributions may be interpreted in different ways and lead to confusions. In the next section, I present an example that illustrates this problem; I notably emphasize the need to take into account the dynamics of the observable during the time course of the experiment, in order to interpret the experimental data at the structural point of view. Last, it is important to critically assess the source of experimental data, especially the limits of the technique of measurement and the conditions in which it has been measured. For instance, Nagle described in 2000 the variety of published values of the area per lipid of DPPC in a fluid phase (62), which was greater than 15â•›Å2. Knowing that lipid force fields are sensitive to differences of 1 to a few Å2, one has to do a critical choice of the source of experimental data before establishing any comparison. I now give a few examples of biophysical techniques that give useful information on pure lipid systems, which can directly be compared to simulations. From NMR 13C and 2H spin-lattice

412

Fuchs

relaxation experiments, it is possible to derive the orientation and dynamics of the C–H bond along the aliphatic chains (63, 64). From the simulation, it just requires the computation of the correlation function describing the orientation of a chosen C–H bond vector along the aliphatic chain of the lipid. This is then possible to evaluate the spectral density function from the Fourier transform of the correlation function and compare it directly to experiment (43). Lateral self-diffusion of lipids can be probed by fluorescence recovery after photobleaching (FRAP). From simulations, diffusion is usually evaluated using the mean square Â�displacement (MSD) and the Einstein relation. Before evaluating the MSD, Wohlert and Edholm recommend that the motion of each leaflet relative to the other (as well as the motion of each leaflet relative to the solvent), which are artifacts due to the finite size of a simulation, be subtracted (64). They also propose a fitting procedure to extract two diffusion coefficients from the simulation: one for the fast diffusion of lipids at short time (<5Â€ns) and another one at longer time scale that can be compared to FRAP data (64). In general, it is important to have fairly long simulations (of hundred(s) of ns) when one wants to evaluate the diffusion. X-ray and neutron scattering techniques can bring useful information on the structure of lipid bilayers. The comparison of scattering data with MD necessitates the calculation of the electron density profile (along the axis parallel to the normal which is usually z) as a time average over all conformations of the simulation. From this, it is possible to evaluate the structure factor (65) or the form factor (66) and compare it directly to the experimental ones. Last but not least, a very important data used to validate lipid force fields is the evaluation of the order parameter of lipid acyl chains (67). Basically, it gives some information on the orientation and motion of the aliphatic chains. Experimentally, order parameters are determined by 2H NMR using phospholipids labeled by deuterons on the acyl chains. From the simulations, it is calculated from the orientation of a specific vector with respect to a reference axis (see ref. 67 for more details). In general, the agreement between experimental and simulated order parameters is rather good in the middle of the bilayer but less satisfactory in the interfacial region (67). If we now focus on membrane proteins (or peptides), the experimental techniques of choice for studying their structure are those described in the other sections of this book, that is, X-ray crystallography, electron microscopy, and NMR. I thus refer the reader to these sections for more details on these topics. Overall, we see that many biophysical techniques give insights on pure (hydrated) bilayers (see above) or on membrane proteins (or peptides). For example, membrane proteins are crystallized within detergents and not in their native environment (i.e., phospholipid bilayer).

Molecular Dynamics of Membrane Peptides and Proteins

413

Recently, Sapay and Tieleman pointed out the lack of experimental data, describing the interactions of proteins (and peptides) with lipids (22) that would allow a sound validation of membrane protein simulations. This aspect is mandatory for the use of computational techniques as a predictive tool. However, improvement of computer hardware as well as recent progresses in computing methodologies, notably the simplification of biomolecule representations using coarse-grained models (see next chapter), open the road for a fruitful future for simulations next to experiments (see next chapter). 3.4. One Guided Example: Comparison of MD Simulations to 2H NMR Data on a Transmembrane Peptide

As an illustration, I present an example dealing with the interpretation of solid state NMR data using molecular dynamics (68). The question was to assess the orientation of a synthetic peptide WALP23 (a poly-alanine peptide flanked by two pairs of tryptophans) in dimyristoylphosphatidylcholine (DMPC or di-C14:0-PC), which gives a slight positive mismatch (i.e., the hydrophobic length of the peptide is slightly longer than that of the bilayer). The peptide embedded in the bilayer surrounded by water is presented in Fig.Â€1. Experimentally, this system was studied by 2H solid state NMR experiments where each alanine of the sequence is selectively labeled with deuterons (71, 72). In this technique, a socalled quadrupolar splitting (QS) is measured; it is related to the orientation of the C–D bonds of the labeled residue with respect to the magnetic field. As the three alanine hydrogens are replaced by deuterons, one cannot distinguish between them because of the fast methyl rotation. Thus, the QS is sensitive to the orientation of the Ca–Cb bond, which makes an angle qi with the magnetic field: ∆Vqi =

3 · K (3cos2 q i − 1) 4

(1)

where Dniq is the quadrupolar splitting on alanine i, and K is a constant with a frequency dimension, the angular brackets mean a time and ensemble average. In the 2H NMR case, it means that this average is a result over the time scale of the experiment (~10−5â•›s) as well as over all molecules labeled with 2H in the sample; in the simulation, as we have only one peptide, this average is evaluated over all the conformations of the trajectory(ies). To extract the orientation of the peptide from the QS, Killian and collaborators developed the Geometric Analysis of Labeled Alanine (GALA) (71, 72). The principle is to calculate the tilt (angle between the helix axis and the normal to the bilayer plane) and the azimuthal rotation (angle describing the rotation about the helix axis, between a vector orthogonal to the helix axis that crosses the Ca of Gly1 and the direction of the tilt, which is a

414

Fuchs

Fig.Â€1. Snapshot of the WALP23 peptide (spheres) embedded in a DMPC bilayer (sticks) surrounded by water (wireframes) taken from the semiPME trajectory at 100Â€ns (61). This image was rendered using tachyon within vmd (62).

vector orthogonal to the helix axis in the tilt plane; this latter is defined by the normal to the membrane and the helix axis) using a fitting procedure from the QS, assuming a static rigid a-helix (see ref. 68 for details). Using this method, the authors found a tilt of 5.2° for WALP23 in DMPC (71). In contrast, MD simulations gave an average tilt angle of 33.5° on hundreds of nanoseconds trajectories (68); this greater extent of tilt found by simulations was in accordance with other computational works (73). To understand this discrepancy, the idea was to back-calculate directly the QS from the trajectory using Eq. (1) and to compare the values to those determined experimentally in oriented samples (NMR experiments described above make use of macroscopically oriented samples with their normal oriented parallel to the magnetic field). Any single trajectory gave a very big difference with experimental values (26Â€kHz on average), while if the six trajectories were concatenated into a single one (and the QS evaluated on this concatenated trajectory), the accordance was by far better (6Â€kHz). It shows that it is important to consider several trajectories, thus to have as much sampling as possible, in order to better fit the experimental QS. Finally, to fully explain the discrepancy on the extent of tilt, this latter was evaluated from the QS back-calculated from the MD trajectories and injected in the GALA method as if they were measured experimentally. Surprisingly, a small tilt of 6.5° was found though the real tilt in the trajectories was equal on average to 33.5°. The reason of this comes from averaging effects

Molecular Dynamics of Membrane Peptides and Proteins

415

as illustrated for one alanine (Ala9) in Fig.Â€2 (similar effects are observed for the other alanines of WALP23). Each trajectory gives on average an important discrepancy (21Â€kHz) compared to the experimental value, whereas the concatenated trajectory gives a QS very close to the experimental one. The explanation is rather straightforward. If one considers a single MD simulation, the peptide samples only a limited range of azimuthal rotation, which depends primarily on the initial conditions (68); in such a case, the discrepancy is important. In contrast, the peptide visits almost all the range of possible azimuthal rotation in the concatenated trajectory, which has the effect to broaden the distribution of QS (see panel CONC in Fig.Â€2). As QS can be either positive or negative, they tend to be averaged to a value that is closer to 0Â€kHz (than in any single trajectory). It thus shows that highly tilted peptides can fit the set of QS determined from 2H NMR. It also demonstrates that it is mandatory to consider the dynamics of the peptide over the time scale

Fig.Â€2. Distribution of the QS of alanine 9 back-calculated from the MD simulations. Solid lines represent the experimental value and dashed lines the mean of the distribution. In the last panel, CONC means concatenated trajectory, whereas the other panels refer to the name of each trajectory in TableÂ€1 of refs. 68. In each plot, the y axis represents the normalized frequency of each class of the histogram.

416

Fuchs

of the experiment, which typically is of a few tens of microseconds for 2H NMR (in fact the inverse of the QS). Interestingly, the same phenomenon has been observed for other artificial transmembrane peptides WLP23 and KLP23 (73). Since highly tilted peptides (68) as well as moderately tilted ones (71) can both fit the experimental values, the question remains, what is the real extent of tilt? In fact, deriving the orientation from QS is an underdetermined problem. We need other data to answer this question. Özdirekcan etÂ€al. proposed to use the PISEMA (Polarization Inversion with Spin Exchange at Magic Angle, which is a 15N solid state NMR technique) approach complementarily to 2H NMR (68). In a PISEMA experiment, the chemical shift of the 15N (of each labeled residue) as well as the 15 N–1H dipolar coupling are reported on a 2D spectrum, the signal is globally sensitive to the angle between the 15N–1H bond vector with respect to the magnetic field. Very recently, the groups of Koeppe and Opella tested the GWALP23 (same as WALP23 but the outer tryptophans on both sides are replaced by glycines) peptide in dilaureylphosphatidylcholine (DLPC, or di-C12:0-PC) using both techniques (74). They found a tilt of 12.6° using the GALA approach from 2H NMR and 10.8° from PISEMA. These values are on the same order than for WALP23, KALP23, WLP23, and KLP23 (KALP23 and KLP23 are similar to WALP23 and WLP23 except the tryptophans that are replaced by lysines) determined from 2H NMR using GALA (71). Clearly, we expect higher values from simulations since it has been reported angles around 30° for a peptide of the same hydrophobic length (WALP23 or WLP23), but in longer lipids (DMPC) (68, 73). Among the possible reasons of this discrepancy, Vostrikov etÂ€al. proposed several explanations (74): (1) more experiments on GWALP23 are needed to validate the above-mentioned results, (2) refinements of the computational methods are needed, (3) different level of hydration may play a role, and (4) averaging effects could in principle occur both in 2H NMR as well as in PISEMA. This last point has been tested by Straus etÂ€al. who simulated a variety of PISEMA spectra using uniform and Gaussian distributions to account for tilt and rotation motions (they also evaluated the influence of librational motions of the peptide plane) (75). They found that fluctuations of both tilt and rotation do not affect very much the spectra. Nevertheless, they only tested distribution widths of 10° and 20°, whereas larger ones were observed during MDs (68, 73), especially for azimuthal rotation. One interesting way of checking this would be to back-calculate 15N chemical shifts and 15 N–1H dipolar couplings from simulations to reconstruct a PISEMA spectrum, which would be a way of taking reasonable fluctuations of motion into account. One other aspect that might be important deals with the fact that PISEMA experiments are usually recorded at high P/L ratio (e.g., 1/20 for GWALP23),

Molecular Dynamics of Membrane Peptides and Proteins

417

thus aggregation is likely. Moreover, hydrophobic mismatch, which is rather important for GWALP23 in DLPC, has been shown to promote aggregation (76). Thus, there is probably an effect of peptide self-association at such ratios and in such mismatching conditions, which would probably tend to lower the extent of tilt. We are still far from fully understanding the orientation and dynamics of single helix transmembrane peptides. Nevertheless, we see in this example that both experimental and computational communities are bridging the gap by working together and suggesting new experiments or simulations to each other. The next challenging goal will probably be the study of these aspects in larger objects such as membrane proteins.

4. Notes 1. Important for the speed of the calculation are (a) the power of each CPU (i.e., measured as a frequency), (b) the total number of CPUs, and (c) the communication between each CPU. It is indeed possible with the actual available MD codes (i.e., softwares) to run simulations in parallel (each CPU is running a part of the calculation, speeding up the whole process) (see Note 3). For point (a), current cheap CPUs available such as AMD or Intel perform very well. For point (b), it of course depends on the amount of money that is available when one buys the computer(s). Last, to improve point (c), the use of shared memory (like on super computers or on current multicore CPU computers) is preferable. All these computers need to be run under an appropriate operating system (such as Unix/Linux), allowing the compilation of the chosen MD code. Distributed computing (i.e., the use of home computers of volunteers) is another alternative for reaching very long simulation times, such as the folding@home project (77). However, this is not possible to use it routinely, unless one collaborates with the investigator of the project. 2. It is to be noted that CHARMM lipids were initially parameterized under the NPgT ensemble, where g stands for surface tension (18). It means that the system has to be stretched on the lateral directions (x and y the plane of the bilayer) in order to obtain a fluid phase. Recently, the charges of CHARMM lipids have been reparameterized which allow the simulation under the NPT ensemble, thus at zero surface tension (19) (see Note 4). 3. GROMACS (27) is probably the fastest code on cheap CPUs (Intel/AMD), notably due to the writing of inner loops (during

418

Fuchs

the evaluation of nonbonded interactions, which is the most time consuming part of the calculation) in assembler code. For example, GROMACS 3.3.1 can run 3–4Â€ ns a day on a system of 70,000 atoms (using a computer with two 64 bits Intel quad-core CPUs of 2.66Â€GHz, with the so-called PME algorithm for computing electrostatics (46, 47)). Recent improvements in the 4.0 version of GROMACS speeds it up even more (78). On the other hand, NAMD (28) has proven to work very well on super computers. 4. Usually, the bilayer stands in the box along the xy plane (with the normal along z). Isotropic pressure coupling should be avoided because it creates an artificial surface tension due to the same scaling in the three dimensions. Anisotropic or semiisotropic coupling are more correct in this respect since they allow an independent scaling of the z dimension. Last, anisotropic coupling can lead to a deformation of the box on long simulations, with the bilayer forming a rectangle instead of square. This can cause artifacts especially if there are not enough lipids next to the protein in the direction that was too much reduced. Thus, it is recommended that semi-isotropic coupling be used to handle pressure coupling on a membrane system (49) 5. The most widely used algorithm for pressure coupling is that of Berendsen (44). Although it does not generate the strict NPT ensemble (the same holds for the thermostat), it is very efficient for equilibrating the density of a system in case of “vacuum defects.” The algorithm shrinks the box until the correct density is reached; the ability to reproduce the correct density then depends primarily on the force field. This is of critical importance since reaching the correct density should give the correct area per lipid. 6. PME (39, 40) is a smart and fast way of evaluating the so-called Ewald summation, which is basically a method for calculating electrostatics in a crystal. When we use PME in biomolecular simulations, we thus consider the system as an infinite crystal, that is, we replicate the simulation box infinitely in all directions. Although this is the most correct way to handle electrostatics, it has been shown to induce artificial periodicities on the system (79). The other alternative is called reaction field (41) and consists in considering a homogenous medium of fixed permittivity beyond the cutoff. It is well suited for simulations of a unique molecule in a liquid solvent, such as a globular protein in water. Although the approximation of a homogenous medium is no longer valid for a bilayer surrounded by water, it has been shown to work fairly well on pure membrane simulations, notably for avoiding the artifacts generated from the use of cutoff (42). Anyhow, PME is the most used algorithm for computing electrostatics in phospholipids systems.

Molecular Dynamics of Membrane Peptides and Proteins

419

References 1. Dodson GG, Lane DP, Verma CS (2008) Molecular simulations of protein dynamics: new windows on mechanisms in biology. EMBO Rep 9(2):144–150 2. Lindahl ER (2008) Molecular dynamics simulations. Methods Mol Biol 443:3–23 3. van Gunsteren WF, Bakowies D, Baron R etÂ€al (2006) Biomolecular modeling: goals, problems, perspectives. Angew Chem Int Ed Engl 45(25):4064–4092 4. Woolf TB, Roux B (1994) Molecular dynamics simulation of the gramicidin channel in a phospholipid bilayer. Proc Natl Acad Sci U S A 91(24):11631–11635 5. Ash WL, Zlomislic MR, Oloo EO, Tieleman DP (2004) Computer simulations of membrane proteins. Biochim Biophys acta 1666(1–2):158–189 6. Gumbart J, Wang Y, Aksimentiev A, Tajkhorshid E, Schulten K (2005) Molecular dynamics simulations of proteins in lipid bilayers. Curr Opin Struct Biol 15(4):423–431 7. van Gunsteren WF, Dolenc J (2008) Biomolecular simulation: historical picture and future perspectives. Biochem Soc Trans 36(Pt 1):11–15 8. van Gunsteren WF, Dolenc J, Mark AE (2008) Molecular simulation as an aid to experimentalists. Curr Opin Struct Biol 18(2):49–53 9. Guvench O, MacKerell AD Jr (2008) Comparison of protein force fields for molecular dynamics simulations. Methods Mol Biol 443:63–88 10. Jorgensen WL, Tirado-Rives J (2005) Potential energy functions for atomic-level simulations of water and organic and biomolecular systems. Proc Natl Acad Sci U S A 102(19):6665–6670 11. Mackerell AD Jr (2004) Empirical force fields for biological macromolecules: overview and issues. J Comput Chem 25(13):1584–1604 12. Ponder JW, Case DA (2003) Force fields for protein simulations. Adv Protein Chem 66:27–85 13. Case DA, Cheatham TE 3rd, Darden T etÂ€al (2005) The Amber biomolecular simulation programs. J Comput Chem 26(16): 1668–1688 14. MacKerrel AD Jr, Bashford D, Bellott M etÂ€al (1998) All-atom empirical potential for molecular modeling and dynamics studies of proteins. J Phys Chem B 102:3586–3616 15. Oostenbrink C, Villa A, Mark AE, van Gunsteren WF (2004) A biomolecular force field based on the free enthalpy of hydration

16.

17.

18.

19.

20. 21.

22. 23.

24.

25.

26. 27.

and solvation: the GROMOS force-field parameter sets 53A5 and 53A6. J Comput Chem 25(13):1656–1676 Jorgensen WL, Maxwell DS, Tirado-Rives J (1996) Development and testing of the OPLS all-atom force field on conformational energetics and properties of organic liquids. J Am Chem Soc 118:11225–11236 Berger O, Edholm O, Jahnig F (1997) Molecular dynamics simulations of a fluid bilayer of dipalmitoylphosphatidylcholine at full hydration, constant pressure, and constant temperature. Biophys J 72(5):2002–2013 Feller S, MacKerell AD Jr (2000) An improved empirical potential energy function for molecular simulations of phospholipids. J Phys Chem B 104(31):7510–7515 Sonne J, Jensen MO, Hansen FY, Hemmingsen L, Peters GH (2007) Reparameterization of all-atom dipalmitoylphosphatidylcholine lipid parameters enables simulation of fluid bilayers at zero tension. Biophys J 92(12):4157–4167 van Gunsteren WF, Berendsen HJC (1987) Groningen Molecular Simulation (GROMOS) library manual. Biomos, Groningen Tieleman DP, MacCallum JL, Ash WL, Kandt C, Xu Z, Monticelli L (2006) Membrane protein simulations with a united-atom lipid and all-atom protein model: lipid–protein interactions, side chain transfer free energies and model proteins. J Phys Condens Matter 18(28):S1221–S1234 Sapay N, Tieleman DP (2008) Molecular dynamics simulation of lipid–protein interactions. Curr Top Memb 60:111–130 Siu SW, Vacha R, Jungwirth P, Bockmann RA (2008) Biomolecular simulations of membranes: physical properties from different force fields. J Chem Phys 128(12):125103 Berendsen HJC, Postma JPM, van Gunsteren WF, Hermans J (1981) Interaction models for water in relation to protein hydration. In: Pullman B (ed) Intermolecular forces. Reidel, Dordrecht, pp 331–342 Jorgensen WL, Chandrasekha J, Madura JD, Impey RW, Klein ML (1983) Comparison of simple potential functions for simulating Â�liquid water. J Chem Phys 79(2):926–935 Feig M (2008) Implicit membrane models for membrane protein simulation. Methods Mol Biol 443:181–196 Van Der Spoel D, Lindahl E, Hess B, Groenhof G, Mark AE, Berendsen HJ (2005) GROMACS: fast, flexible, and free. J Comput Chem 26(16):1701–1718

420

Fuchs

28. Phillips JC, Braun R, Wang W etÂ€ al (2005) Scalable molecular dynamics with NAMD. J Comput Chem 26(16):1781–1802 29. h t t p : / / m o o s e . b i o . u c a l g a r y. c a / i n d e x . php?page=Structures_and_Topologies 30. http://www.rcsb.org/ 31. http://blanco.biomol.uci.edu/Membrane_ Proteins_xtal.html 32. DeLano WL (2002) The PyMOL molecular graphics system. DeLano Scientific, Palo Alto, CA 33. http://www.pymol.org 34. Guex N, Peitsch MC (1997) SWISS-MODEL and the Swiss-PdbViewer: an environment for comparative protein modeling. Electrophoresis 18(15):2714–2723 35. http://www.expasy.org/spdbv/ 36. Alland C, Moreews F, Boens D etÂ€al (2005) RPBS: a web resource for structural bioinformatics. Nucleic Acids Res 33(Web Server issue):W44–W49 37. http://bioserv.rpbs.jussieu.fr/RPBS/cgibinRessource.cgi?chzn_lg=an& chzn_rsrc=BasicBuilder 38. Biggin PC, Bond PJ (2008) Molecular dynamics simulations of membrane proteins. Methods Mol Biol 443:147–160 39. Faraldo-Gomez JD, Smith GR, Sansom MS (2002) Setting up and optimization of membrane protein simulations. Eur Biophys J 31(3):217–227 40. Jo S, Kim T, Im W (2007) Automated builder and database of protein/membrane complexes for molecular dynamics simulations. PLoS ONE 2(9):e880 41. Jo S, Kim T, Iyer VG, Im W (2008) CHARMM-GUI: a web-based graphical user interface for CHARMM. J Comput Chem 29(11):1859–1865 42. Kandt C, Ash WL, Tieleman DP (2007) Setting up and running molecular dynamics simulations of membrane proteins. Methods 41(4):475–488 43. Feller SE (2007) Molecular dynamics simulations as a complement to nuclear magnetic resonance and X-ray diffraction measurements. Methods Mol Biol 400:89–102 44. Berendsen HJC, Postma JPM, DiNola A, Haak JR (1984) Molecular dynamics with coupling to an external bath. J Chem Phys 81:3684–3690 45. Auffinger P, Beveridge DL (1995) A simple test for evaluating the truncation effects in simulations of systems involving charged groups. Chem Phys Lett 234(4–6): 413–415

46. Darden T, York D, Pedersen L (1993) Particle mesh Ewald: an N [center-dot] log(N) method for Ewald sums in large systems. J Chem Phys 12:10089–10092 47. Essmann U, Perera L, Berkowitz ML, Darden T, Lee H, Pedersen LG (1995) A smooth particle mesh Ewald method. J Chem Phys 103(19):8577–8593 48. Tironi IG, Sperb R, Smith PE, van Gunsteren WF (1995) A generalized reaction field method for molecular dynamics simulations. J Chem Phys 102:5451–5459 49. Anézo C, de Vries AH, Höltje H-D, Tieleman DP, Marrink SJ (2003) Methodological issues in lipid bilayer simulations. J Phys Chem B 107(35):9424–9433 50. Patra M, Karttunen M, Hyvonen MT, Falck E, Lindqvist P, Vattulainen I (2003) Molecular dynamics simulations of lipid bilayers: major artifacts due to truncating electrostatic interactions. Biophys J 84(6):3636–3645 51. Cordomi A, Edholm O, Perez JJ (2007) Effect of different treatments of long-range interactions and sampling conditions in molecular dynamic simulations of rhodopsin embedded in a dipalmitoyl phosphatidylcholine bilayer. J Comput Chem 28(6):1017–1030 52. van Gunsteren WF, Mark AE (1992) On the interpretation of biochemical data by molecular dynamics computer simulation. Eur J Biochem 204(3):947–961 53. Lacapere JJ, Pebay-Peyroula E, Neumann JM, Etchebest C (2007) Determining membrane protein structures: still a challenge! Trends Biochem Sci 32(6):259–270 54. Monticelli L, Kandasamy S, Periole X, Larson R, Tieleman DP, Marrink SJ (2008) The MARTINI coarse grained force field: extension to proteins. J Chem Theory Comput 4:819–834 55. Chipot C (2008) Free energy calculations applied to membrane proteins. Methods Mol Biol 443:121–144 56. Yawata Y (ed) (2004) Composition of normal red cell membranes. In: Cell membrane. Wiley, Weinheim, pp 27–46 57. Tsukazaki T, Mori H, Fukai S etÂ€ al (2008) Conformational transition of Sec machinery inferred from bacterial SecYE structures. Nature 455(7215):988–991 58. Hub JS, de Groot BL (2008) Mechanism of selectivity in aquaporins and aquaglyceroporins. Proc Natl Acad Sci U S A 105(4):1198–1203 59. Torrie GM, Valleau JP (1977) Nonphysical sampling distributions in Monte Carlo

Molecular Dynamics of Membrane Peptides and Proteins Â� free-energy estimation: umbrella sampling. J Comp Phys 23(2):187–199 60. Allen TW, Andersen OS, Roux B (2006) Molecular dynamics – potential of mean force calculations as a tool for understanding ion permeation and selectivity in narrow channels. Biophys Chem 124(3):251–267 61. Corry B, Martinac B (2008) Bacterial mechanosensitive channels: experiment and theory. Biochim Biophys Acta 1778(9):1859–1870 62. Nagle JF, Tristram-Nagle S (2000) Structure of lipid bilayers. Biochim Biophys Acta 1469(3):159–195 63. Pastor RW, Venable RM, Feller SE (2002) Lipid bilayers, NMR relaxation, and computer simulations. Acc Chem Res 35(6):438–446 64. Wohlert J, Edholm O (2006) Dynamics in atomistic simulations of phospholipid membranes: nuclear magnetic resonance relaxation rates and lateral diffusion. J Chem Phys 125(20):204703 65. Benz RW, Castro-Roman F, Tobias DJ, White SH (2005) Experimental validation of molecular dynamics simulations of lipid bilayers: a new approach. Biophys J 88(2): 805–817 66. Klauda JB, Kucerka N, Brooks BR, Pastor RW, Nagle JF (2006) Simulation-based methods for interpreting x-ray data from lipid bilayers. Biophys J 90(8):2796–2807 67. Vermeer LS, de Groot BL, Reat V, Milon A, Czaplicki J (2007) Acyl chain order parameter profiles in phospholipid bilayers: computation from molecular dynamics simulations and comparison with 2H NMR experiments. Eur Biophys J 36(8):919–931 68. Ozdirekcan S, Etchebest C, Killian JA, Fuchs PF (2007) On the orientation of a designed transmembrane peptide: toward the right tilt angle? J Am Chem Soc 129(49):15174–15181 69. Humphrey W, Dalke A, Schulten K (1996) VMD: visual molecular dynamics. J Mol GraphÂ€14(1):33–38, 27–28 70. http://www.ks.uiuc.edu/Research/vmd/

421

71. Ozdirekcan S, Rijkers DT, Liskamp RM, Killian JA (2005) Influence of flanking residues on tilt and rotation angles of transmembrane peptides in lipid bilayers. A solid-state 2H NMR study. Biochemistry 44(3):1004–1012 72. Strandberg E, Ozdirekcan S, Rijkers DT etÂ€al (2004) Tilt angles of transmembrane model peptides in oriented and non-oriented lipid bilayers as determined by 2H solid-state NMR. Biophys J 86(6):3709–3721 73. Esteban-Martin S, Salgado J (2007) The dynamic orientation of membrane-bound peptides: bridging simulations and experiments. Biophys J 93(12):4278–4288 74. Vostrikov VV, Grant CV, Daily AE, Opella SJ, Koeppe RE 2nd (2008) Comparison of “Polarization inversion with spin exchange at magic angle” and “geometric analysis of labeled alanines” methods for transmembrane helix alignment. J Am Chem Soc 130(38):12584–12585 75. Straus SK, Scott WRP, Watts A (2003) Assessing the effects of time and spatial averaging in 15N chemical shift/15N-1H dipolar correlation solid state NMR experiments. J Biomol NMR 26(4):283–295 76. Sparr E, Ash WL, Nazarov PV etÂ€ al (2005) Self-association of transmembrane alphahelices in model membranes: importance of helix orientation and role of hydrophobic mismatch. J Biol Chem 280(47): 39324–39331 77. Shirts M, Pande VS (2000) Computing: screen savers of the world unite! Science 290(5498):1903–1904 78. Hess B, Kutzner C, Van Der Spoel D, Lindahl E (2008) GROMACS 4: algorithms for highly efficient, load-balanced, and scalable molecular simulation. J Chem Theory Comput 4(3):435–447 79. Hunenberger PH, McCammon JA (1999) Effect of artificial periodicity in simulations of biomolecules under Ewald boundary conditions: a continuum electrostatics study. Biophys Chem 78(1–2):69–88

as

Chapter 22 Membrane Protein Dynamics from Femtoseconds to Seconds Christian Kandt and Luca Monticelli Abstract Membrane proteins play a key role in energy conversion, transport, signal recognition, transduction, and other fundamental biological processes. Despite considerable progress in experimental techniques, the determination of structure and dynamics of membrane proteins still represents a great challenge. Computer simulation methods are becoming an increasingly important tool not only in the interpretation of experiments but also in the prediction of membrane protein dynamics. In the present review, we give a brief introduction to molecular modeling techniques currently used to explore protein dynamics on time scales ranging from femtoseconds to microseconds. We then describe a few recent example applications of these techniques to membrane proteins. In conclusion, we also discuss some of the newest developments in simulation methodology that have the potential to further extend the time scale accessible to explore (membrane) protein dynamics. Key words: Membrane protein, Protein dynamics, Computer simulation, Molecular mechanics, Quantum mechanics, Molecular dynamics, Force field, Coarse-grain

1. Introduction 1.1. Membranes and Membrane Proteins

Biological membranes are thin films consisting primarily of lipid and protein molecules held together mainly by non-covalent interactions. Essentially acting like two-dimensional fluids, they generate confined and self-contained reaction spaces that are a fundamental precondition of life. The most common structural organization of biomembranes is that of a continuous double layer of lipids. Lipid monolayers do occur as well, for instance in lung surfactant where they are vital for the breathing process (1). Beyond their basic role as a barrier, biomembranes facilitate a vast palette of other functions mainly determined by the type

Jean-Jacques Lacapère (ed.), Membrane Protein Structure Determination: Methods and Protocols, Methods in Molecular Biology, vol. 654, DOI 10.1007/978-1-60761-762-4_22, © Springer Science+Business Media, LLC 2010

423

424

Kandt and Monticelli

of proteins associated with or embedded in the membrane. Located at the interface of cell or organelle inside and outside, membrane proteins are key players in fundamental processes such as energy conversion, transport, signal recognition, and transduction. Though the last years have seen substantial progress in the determination of protein high-resolution structures by means of X-ray crystallography (see sectionÂ€2 of this book), electron microscopy (see sectionÂ€3) and NMR spectroscopy (see sectionÂ€4), the wealth of known structures of membrane proteins in the Brookhaven Protein Data Bank (PDB) (2–4) is still far behind the amount of available structures of water-soluble proteins. At the time of writing, the PDB holds a total of 51,155 structures, 158 of which are of membrane proteins (5). Gaining a deeper understanding of the function and architecture of membrane proteins is clearly a major challenge in modern structural biology. Next to the on-going improvement of experimental techniques, computer simulations have become powerful tools contributing to an increase in our understanding of (membrane) protein structure and dynamics.

2. Methods 2.1. Molecular Modeling

Proteins are flexible linear macromolecules, capable of a wide range of conformational changes. In the majority of cases, this dynamics of motion is the foundation enabling proteins to carry out their physiological functions. The successful determination of a protein’s high-resolution structure is a landmark on the way to understanding its function. However, another key element is still missing and that is the dynamic of motion. Characteristic time scales for protein motions range from femtoseconds to hours (see TableÂ€1). Information on protein dynamics can be obtained using a number of experimental techniques (e.g., NMR, EPR, IR, or fluorescence spectroscopy), but this information is often indirect and can be difficult to interpret. Moreover, the formation of unstable species like transitions states and reaction intermediates is difficult or impossible to observe experimentally. Computer simulations can provide detailed information on protein dynamics on a broad range of time scales, from femtoseconds to milliseconds. Different methodological approaches are employed to model the dynamics of biological macromolecules depending on the length and time scale of the process of interest. These approaches can be classified in two groups: molecular mechanical (MM) and quantum mechanical (QM) methods. While molecular mechanical simulations do not consider the

Membrane Protein Dynamics from Femtoseconds to Seconds

425

TableÂ€1 Characteristic length and time scales for protein dynamics Dynamic event

Amplitude (nm)

Typical time scale (s)

Bond vibration

0.001–0.01

10−14–10−13

Side chain rotation (solvent-exposed)

0.5–1.0

10−11–10−10

Relative motions of globular domains

0.1–0.5

10−10–10−7

Rigid domain motions

1.0–5.0

10−9–10−6

Helix-coil transition

>3.0

10−7–10−5

Beta-sheet formation

>2.0

10−6–10−3

Local denaturation

0.5–1.0

10−5–101

Protein aggregation

>3.0

10−5–104

electrons in a molecule directly and focus on the dynamics of entire atoms, QM simulations concentrate on the dynamics of electrons and chemical bonds. This includes electronic distributions, excitations events, transition states, or the formation and breaking of bonds in chemical reactions. 2.2. Quantum Mechanics

Both MM and QM simulations on a molecular scale describe atoms as point masses. However, while MM simulations resort to an empirically derived energy function to model the electronic contributions, QM simulations consider some or all electrons in the system explicitly, using terms of spatial probability distributions to describe each electron by its wave function. This is done by solving or rather approximating the Schrödinger equation, for which an exact solution is only possible for the hydrogen atom. More complex systems are described through a linear combination of many hydrogen orbitals (6, 7). The approximation quality employed to solve the Schrödinger equation determines the range of system size and simulation time accessible by QM techniques. If a low level of approximation quality is used, one time-step can be performed to describe the electronic structure of a small protein, like crambin or a hemoglobin monomer. On a high level of approximation quality, characteristics such as bond lengths, vibrations, and absorption spectra can be computed from a triatomic system with an accuracy exceeding that of any direct experimental measurements today (8). While QM can also be used to derive the forces acting on each atom and thus predicting atomic motions – as in Born–Oppenheimer MD or Car–Parrinello MD (CPMD) (9) – the method’s main focus is on the high-detail description of the nature of chemical

426

Kandt and Monticelli

bonds, which does not necessarily require a time scale aspect. QM methods are also a crucial means in parameterizing new models for MM simulations. Solving the Schrödinger equation can be done in a timedependent or time-independent way. Both ways can be applied to compute a system’s dynamics over time as the terms of time dependency refer only to the way the wave function is used: in the time-dependent variant, the wave function is built only once at the beginning of the simulation and is later on propagated through time, whereas in the time-independent approach, the wave function is built several times throughout a simulation in a predefined interval of computation steps (10). Both approaches can be carried out using different sets of simulation techniques. Ab initio QM solves the Schrödinger equation without simplifying any of the mathematical operations necessary to compute the wave function, whereas semi-empirical QM employs mathematical simplifications and uses predetermined parameters in the calculation of the wave function interactions. In terms of complexity, computational cost and accuracy density functional theory (DFT) can be considered intermediate between ab intio and semi empirical QM. It also follows a different strategy to describe electronic structure by using terms of electron densities instead of electronic wave functions (11). Of these techniques, ab initio allows for the greatest accuracy but on the other hand is limited to the smallest system sizes of tens of atoms. DFT methods are currently among the most popular QM methods. Computationally less expensive than ab initio, they still retain a good accuracy on a number of systems of biological interest, like metalloenzymes (12), that is, systems up to hundreds of atoms. Very recently, it has become feasible to calculate whole proteins by means of DFT. More approximate semi-empirical methods can be used to model even larger systems, but can be rather inaccurate in more difficult cases such as the calculation of reaction energies. Ab initio, DFT, and semi-empirical techniques can be combined with molecular mechanical simulations in so-called hybrid methods, where part of the system of interest is treated using a QM method and the rest is treated using MM (typically with a molecular dynamics approach). Review articles on QM/ MM methods can be found in (12, 13). The QM region would normally include only the active site of the enzyme and the substrate, while the MM region would consist of the rest of the protein and the environment around it (membrane and solvent). QM/MM methods were introduced by Warshel and Levitt in 1976 (14) and, since then, have become increasingly popular in theoretical studies of enzymatic catalysis. The major advantage of QM/MM methods compared to plain QM methods lies in the description of the chemical environment around the site where the reaction takes place.

Membrane Protein Dynamics from Femtoseconds to Seconds

427

2.3. Molecular Mechanics

Molecular mechanics (MM) refers to a large number of simulation methods in which molecular systems are modeled using simple, empirical potential energy functions. In molecular mechanics, atoms are considered as point masses moving in an effective energy field, commonly referred to as force field. Electrons are not considered explicitly, instead their effect is accounted for indirectly. Force fields usually describe molecules in terms of elastic springs, bond lengths, bond angles, torsion angles, and point charges (although simpler forms are also used in some cases). MM includes a very diverse range of simulation techniques. Some of them – such as energy minimization, normal mode analysis, or Monte Carlo simulations – do not rely on equations of motion and therefore do not provide direct information on the dynamics of the system. In other methods – for example, Langevin dynamics (LD), Brownian dynamics (BD), dissipative particle dynamics (DPD), and molecular dynamics (MD) – the positions and momenta of all the atoms (or particles) are calculated as a function of time by integrating an equation of motion in discrete time steps. The result is referred to as “trajectory” and, for biomolecular systems, it typically consists of millions of steps (15) (i.e., nanoseconds of simulation time). The technique that is most widely used for the description of membrane protein dynamics is MD (15, 16). It is still rare that the entire time scale of the process of interest can be covered by MD simulations. However, the dynamics of motion, even on a relatively short time scale, can reveal trends of conformational or energetic changes that provide insight into protein function. Next to classical atomistic force fields, where each atom is treated explicitly, simplified representations can also be used, in which groups of atoms are represented by single interaction sites (beads). Such coarse-grained (CG) approaches are often used in combination with elastic network models, DPD and MD. Coarsegrained models were initially developed in the 1970s (17) and then applied mainly to lipids, surfactants, polymers, and proteins. Numerous CG representations have been developed for membrane proteins, reviewed in (18–21). Dynamics in CG simulations is often faster compared to the atomistic case, and has to be interpreted with care (22). Coarse-grained MD simulations provide a means toward mesoscale simulations, where system sizes on a magnitude of 106 particles and simulation times of microseconds are feasible. As the representation of the system’s electronic configuration does not change in the course of a molecular mechanics simulation, chemical reactions cannot be simulated directly (see Fig.Â€1).

2.4. Limitations of Computer Simulations

In principle, all properties of a system that depend on positions and velocities of the atoms can be calculated from a trajectory, using the tools of statistical mechanics. In practice, limitations are present in any simulation, related mainly to two aspects: (1)

428

Kandt and Monticelli

Fig.Â€1. While QM simulations concentrate on electron distributions and the dynamics of chemical bonds (a), molecular dynamics simulations do not consider the electrons directly and focus on the dynamics of entire atoms and molecules (b, c). Atomistic MD simulations consider all atoms in the system whereas in coarse-grained MD simulations groups of atoms are condensed into effective interaction sites (c).

the time and length scale of the simulation and (2) the accuracy in the calculation of the interactions. The limited size and time scale of the simulations is often referred to as the “sampling problem”: in order to sample correctly motions that take place on a certain time scale, the simulations are required to be at least as long as the time scale of the phenomena, and typically much longer. Moreover, the size of the modeled systems should be larger than the length scale of the motion one wants to characterize. As a consequence, it is still difficult to use computer simulations to study transformations and motions that take place on time scales beyond the microsecond and lengths beyond tens of nanometers. Unfortunately, this involves many biologically interesting phenomena. Techniques are available to overcome some of these problems by accelerating molecular motions and transformations. The development of these techniques is a very active area in the field of molecular simulations. In addition to the sampling problem, computer simulations are also limited by the accuracy of the calculations of the interactions between atoms. This problem is particularly important in MM simulations, which are based on force fields. Force fields are simplified descriptions of the interactions between atoms and contain a number of approximations. Comparison with experiments is normally used both in the development and in the validation of the force field. Once a certain force field has been shown to reproduce realistically some quantities experimentally measured, we can expect it to perform reasonably well in predicting properties that have not been measured, as long as the simulation conditions are similar. 2.5. Applications

In this section, we will discuss recent examples of QM and MM computer simulations of membrane proteins investigating biological phenomena occurring on a time scale of femtoseconds

Membrane Protein Dynamics from Femtoseconds to Seconds

429

to milliseconds. As the field of molecular modeling of membrane proteins is growing rapidly, giving a complete overview is not possible in the framework of this chapter. We will therefore limit our presentation to a few selected examples published in the last few years. Further literature reviewing computer simulations of membrane proteins can be found in references (15, 23–26). 2.5.1. Femtosecond to Picosecond Dynamics: Quantum Mechanics Simulations

Combined QM/MM methods have been widely used to address questions on the mechanism of specific enzymatic reactions and on the general principles of enzymatic catalysis, much debated in recent years. Most studies published so far using QM/MM techniques deal with the dynamics of water-soluble enzymes, while fewer studies have been published on membrane proteins. One example from the recent literature is about the KcsA potassium channel. Potassium channels are membrane proteins responsible for the transmission of electrical pulses in the nervous system (for a more detailed discussion of ion channels, see the next section). They conduct K+ ions at near diffusion limit (108 ions s−1 for each channel) while not allowing the conduction of other monovalent ions. Although the crystal structure of a few K+ channels has been solved (27–31), questions still remain on the origin of the high selectivity, as well as about the mechanism of opening and closing of the pore. The region responsible for the high selectivity of K+ channels is know as the “selectivity filter” of the protein. In KcsA, two ionizable residues are found close to the selectivity filter. In order to explore the molecular determinants of ion channel selectivity using computer simulations, it is necessary to model accurately the selectivity filter and the adjacent region, including the protonation state of all residues. The group of Rothlisberger used CPMD to investigate the KcsA potassium channel, focusing on the electronic structure of the selectivity filter, including polarization effects, charge transfer (32), and the protonation state of specific residues (33). It was found that two acidic residues in the proximity of the selectivity filter of KcsA, namely Glu71 and Asp80, share one proton. The proton exchange occurs on the picosecond time scale, a phenomenon that cannot be observed experimentally nor through classical MD simulations. In combination with experiments, the theoretical investigation suggested that the occupancy and structure of the ion selectivity filter, as well as ion translocation, depend on the protonation state of these residues.

2.5.2. Picosecond to Microsecond Dynamics: Molecular Mechanics Simulations

G protein-coupled receptors (GPCRs) form the largest known protein superfamily. Its members are of critical importance in a wide range of eukaryotic signaling processes (34). Sensing molecules or other stimuli outside the cell, GPCRs activate signal transduction cascades based on the second messenger mechanism. To date, the structures of two GPCRs are known at atomic resolution: the beta2 adrenergic receptor (35, 36) and rhodopsin (37). All the recent simulation studies have focused on rhodopsin, which

2.5.2.1. G Protein-Coupled Receptors

430

Kandt and Monticelli

is the primary light receptor in animal visual systems. Like in the photosynthetic bacteriorhodopsin (38, 39), retinal acts as a molecular light sensor, undergoing a cis to trans transition in its polyene chain within 200Â€ fs upon light absorption (40, 41), which ultimately induces the closure or cGMP-gated ion channels within micro to milliseconds after the light stimulus (42, 43). Simulation studies of rhodopsin have focused on how lightinduced changes in the retinal chromophor propagate through the protein (44–48). They also investigated protein–lipid interactions (49, 50) and rhodopsin oligomerization behavior (51, 52). To determine which residues couple the changes in retinal configuration to the changes in the cytoplasmic transducin binding site, Kong and Karplus computed the correlation of residue–residue and residue–retinal interaction energies in multi-copy MD simulations of rhodopsin with and without simulated chromophor isomerization (48). Although the simulated time of 2–3Â€ ns per run is too short compared to the microsecond scale of the actual structural changes, the authors did observe changes in residue interaction energy preceding the actual conformational transitions, based on which potential mutagenesis candidates were identified and cross-linking experiments proposed. MartinezMayorga etÂ€ al. focused on the retinal micro-environment and how different protonation states affect the chromophor dynamics after light isomerization (46). Two opposing functional models have been proposed regarding the protonation state of two glutamates in the retinal binding pocket. According to the first model, Glu-181 would be protonated prior to isomerization but deprotonated after the isomerization (proton transfer to Glu-113) (53); the second model proposes that both glutamates remain deprotonated (54). To understand which functional model is more likely, rhodopsin was simulated in a bilayer environment for 1,500 and 2,000Â€ns after retinal isomerization at two different protonation states. From the trajectories, solid state ²H NMR spectra of the retinal methyl groups were computed and compared to experimental data. The simulated spectra were found in remarkable agreement with the second model. 25.2.2. Transport Across the Membrane: Transporters and Channels

Biological membranes are largely impermeable to ions and polar molecules. To achieve controlled permeability for such compounds, living cells employ two types of membrane proteins: channels and transporters. While channels facilitate substrate diffusion down its electrochemical gradient (passive transport), transporters draw from a source of energy – such as light, ATP hydrolysis, or concentration gradient – to drive thermodynamically uphill transport of the substrate against its electrochemical gradient (active transport). Transporters can be classified as either uniporters – transporting one substrate in one direction, symporters – two substrates in the same direction or antiporters – two

Membrane Protein Dynamics from Femtoseconds to Seconds

431

substrates in opposite directions. Numerous simulation studies of transporters were performed in the last few years on uniporters (55, 56), antiporters (57–59), and symporters (60). 2.5.2.3. Transporters

ABC transporters are uniporters found in all forms of life. Powered by ATP hydrolysis, they transport a broad range of substrates across the membrane, ranging from small molecules such as ions, sugars, and amino acids to larger compounds such as pharmaceutics, lipids, and oligopeptides (61, 62). In order to understand the mechanism of functioning of ABC transporters, Sonne etÂ€al. combined elastic network normal mode analysis (63) with a series of steered MD simulations on the vitamin B12 importer BtuCD in a POPE bilayer (56). Using the motor domains of another transporter trapped in the presence and absence of nucleotide as a template, the BtuCD motor domains were pushed together and pulled apart while monitoring the conformational response in the trans-membrane domains. Both simulation techniques found the same trends in conformational response, clearly supporting one functional model while contradicting the other. The favored model has later been confirmed by further ABC transporter crystal structures (62). A critical component for Escherichia coli to maintain cellular salt and pH homeostasis is the Na+/H+ antiporter NhaA (64). Though the crystal structure of the protein was determined in an inactive and thus sodium-free state, two aspartates located halfway through the protein had been proposed as Na+ binding sites based on earlier mutagenesis experiments (65). As the protonation state of these two aspartates was unknown, Arkin etÂ€al. simulated all four possible protonation scenarios, with NahA embedded in a POPE bilayer for 12–100Â€ ns each (57). Monitoring water accessibility from each side of the membrane and sodium diffusion behavior after initial placement next to one of the aspartates, the authors were able to deduct a complete Na+/H+ antiporting cycle for NhaA.

2.5.2.4. Channels

Self-assembling in the host cell membrane into a heptameric pore, alpha hemolysin is a bacterial toxin that causes cell death by introducing unregulated membrane channels in the target membrane (66). Patch clamp experiments have shown that various solutes (with sizes up to 1,300 nucleotides single DNA or RNA) can translocate through the alpha hemolysin channel by applying a transmembrane electric potential (67). Wells and coworkers chose this channel to introduce a new method of steered MD that makes it possible to simulate in tens of nanoseconds transport events that naturally occur on a millisecond time scale (68). Applying a transmembrane electric potential acting on the solute only, the authors were able to simulate DNA and peptide translocation through alpha hemolysin. The simulations reproduced a number

432

Kandt and Monticelli

of experimental results, including ratios of DNA translocation velocities and amplitude of the applied potential. OpcA is an outer membrane channel in Neisseria meningitides – a major pathogen for inflammatory diseases such as meningtitis and septicaemia – and a key player in facilitating cell adhesion and the internalization of the bacterium within the host cells. Two crystal structures of the protein have been determined, revealing substantially different conformations in two of the extracellular loop regions in a region of the protein implicated in proteoglycan binding (69, 70). Luan etÂ€al. performed 20Â€ns MD simulation of each crystal structure, embedded in a POPC bilayer and in the unit cell of their respective crystal lattice (71). They could show that the different loop conformations arose from the different crystal packings, and that within the membrane environment the X-ray structures showed large conformational changes. These findings were combined into a new model, based on parts of the X-ray structures that were least affected by crystal contacts. The new model, subjected to another 20Â€ns simulation, proved to be more stable than the original crystal structures. 2.5.2.5. Ion Channels

Ion channels act as extremely selective gateways, allowing specific ions to pass in and out of the cell in response to various signals. They are fundamental for physiological processes such as the formation and transduction of nerve impulses, muscle contraction, and osmoregulation. Recent computational studies included computer simulations of potassium channels, mechanosensitive, and ligand-gated ion channels. The first X-ray crystal structure of an ion channel was determined for the potassium channel KcsA in 1998 (27). While KcsA has a high selectivity for potassium, the structurally similar NaK channel is able to conduct both K+ and Na+ ions (72). Noskov and Roux investigated the physical foundation for ion selectivity in these channels using a combination of simple dynamic models of a cation surrounded by different ligand components – such as carbonyl functions or water – as well as free energy calculations based on atomistic MD simulations of NaK embedded in a DPPC bilayer (73). They found the selectivity of the central ion binding sites in KcsA and NaK to be largely controlled by the hydration of the cation inside the channel. Oxygen atoms lining the ion pore were found to play an important role in this, having a different effect on ion selectivity, depending on whether an oxygen is donated by a carbonyl group or a water molecule. A central ion binding site dominated by carbonyl functions – as in KcsA – is found K+-selective, whereas the binding site in NaK is not, as there the cation is liganded predominantly by water. Being slightly wider in this region, the ion pore in NaK does not remove the ion’s hydration shell as efficiently as KcsA does.

Membrane Protein Dynamics from Femtoseconds to Seconds

433

2.5.3. Microsecond Dynamics: Coarse-Grained Simulations

A wide range of biologically interesting phenomena occurs on time scales that are currently out of reach for atomistic models. Vesicle fusion, self-assembly of large protein complexes, and signal transduction take place on microsecond or longer time scales. Simplification of the model is required to simulate these motions. In recent years, the use of coarse-grained models has become increasingly popular, and a large number of membrane protein systems has been investigated. Many different approaches have been developed to study lipid–protein interactions, reviewed in (19–21). It is impossible to give a broad overview that will do justice to any of these simulations. Some CG simulation studies of membrane peptides and proteins are selected here for a brief discussion.

2.5.3.1. Hydrophobic Mismatch

Membrane proteins normally present large hydrophobic surfaces in contact with lipid membranes. To minimize exposure of hydrophobic residues to the water environment, it is necessary that the hydrophobic thickness of the lipids surrounding the protein matches the hydrophobic thickness of the protein. It has been proposed that hydrophobic matching has an important role in several fundamental processes in cell membranes, e.g., in the secretory pathway in the Golgi, in lipid sorting around membrane proteins and in sequestering proteins with long transmembrane regions into lipid rafts (sphingolipid and cholesterol rich membrane patches). Hydrophobic mismatch has been proposed to result in protein-induced bilayer deformations, lipid-induced protein tilting, and aggregation. For many years, coarse-grained simulations have been used to explore the consequences of hydrophobic mismatch in proteincontaining membranes. Using Monte Carlo CG simulations, Sperotto and Mouritsen found that the perturbation of the bilayer decays exponentially with the distance from the protein (74). More detailed CG models were used by Smit and coworkers in DPD simulations to describe the effects of hydrophobic mismatch as a function of the size of the protein (75, 76). Small transmembrane peptides were predicted to display much larger tilt angles compared to larger proteins. These results were recently confirmed by experimental investigations by Marsh and coworkers (77).

2.5.3.2. Protein Aggregation and ProteinMembrane Self-Assembly

Among the numerous CG models recently published, the one by Marrink and coworkers (coined the MARTINI force field (22, 78, 79)) has been used to study protein aggregation and lipid– protein interactions. Based on the current state of knowledge, most membrane proteins appear to be oligomeric (80). Self-assembly of membrane proteins probably plays a role in sorting of membrane components. Rhodopsin, the light receptor involved in vision, also seems to form dimers in native membranes, according to recent experimental evidence (81). Periole used the MARTINI model to

434

Kandt and Monticelli

study the aggregation of rhodopsin in model membranes (51). The simulated systems contained up to 16 copies of the GPCR and 1,600 lipids, in addition to water, and were simulated for up to 8Â€ ms. Spontaneous protein aggregation was observed to be dependent on the thickness of the lipid membrane, with hydrophobic mismatch promoting aggregation, in agreement with theoretical models. The bilayer was found to adapt to the local hydrophobic thickness of the protein, and the persistence length of the thickness alterations was 1–2Â€nm. Protein aggregation was found to proceed via a multi-stage mechanism, with the formation of an encounter complex followed by a rearrangement leading to the fully bound state. Shape complementarity was suggested to play an important role in the second stage of the formation of the complex. Mechanosensitive channels are membrane proteins that open in response to tension in cell membranes. They are involved in various physiological processes, like touch, hearing, and osmoregulation. When the tension in the membrane reaches a threshold value (gating tension), the channel opens a large pore with a diameter up to 4Â€nm for Eco-MscL (82). The threshold value can change in proteins from different organisms, and the time scale for the gating process is normally between milliseconds and seconds. The crystal structure of the mechanosensitive channel Tb-MscL, solved a few years ago (83), has been used by several groups as a starting point to simulate the gating mechanism and to predict the channel structure in the open state. Since the time scales of the biological process are not within reach of molecular simulation techniques, different methods have been used to speed up the conformational change. Marrink and coworkers used a CG force field to simulate the gating process of Tb-MscL and a mutant (84). In their simulations, high negative pressures were applied in the membrane plane in order to promote the channel opening. Both proteins were found to open in response to increasing membrane tension with an iris-like mechanism. The protein conformational changes were observed on the microsecond time scale, and the Tb-MscL mutant showed a more pronounced expansion of the pore, in good agreement with experimental measures. Folding and self-assembly of membrane proteins are fundamental problems in modern biology. For simple helical peptides, folding can take place on time scales as short as 10–100Â€ns and has been simulated in atomistic detail (85–87). For proteins including multiple helical regions or beta-sheets and combinations of different secondary structure elements, the folding process requires time scales of microseconds and beyond (88). The mechanism of insertion of proteins into membranes is thought to be different for helical and beta-sheet membrane proteins: for the helical ones, transmembrane helices can insert independently into

Membrane Protein Dynamics from Femtoseconds to Seconds

435

the membrane and then self-assemble (89), while for beta-barrels folding and insertion would occur at the same time (90). In order to explore this issue, Bond and Sansom simulated the insertion of glycophorin (alpha helical) and OmpA (beta-barrel) in detergent micelles and in lipid bilayers (91). Multiple CG simulations were carried out starting from random distributions of lipids and proteins. Lipids were observed to self-assemble spontaneously into bilayers both in the presence and in the absence of proteins (91). In the case of glycophorin, interfacial partitioning of the peptide was followed by spontaneous insertion into the bilayer and dimerization. This mechanism is consistent with the much debated two-state model of membrane protein folding. In the case of OmpA, the protein was quickly surrounded by lipids during the formation of the bilayer so that once the bilayer was formed the protein was already inserted in it. Comparison with atomistic simulations (92) showed good agreement in terms of lipid–protein interactions and tilt of the beta barrel relative to the bilayer normal. These results show that CG simulations can be used to insert a folded membrane protein into a lipid bilayer, and can therefore, provide starting structures for atomistic simulations to explore finer details of protein dynamics. 2.5.4. Beyond the Microsecond

Two factors contribute to the continuous increase in the time scale accessible to molecular simulations: the progress in computer technology and the development of new simulation techniques. In the following paragraph, we will briefly discuss some of the methodological advances that might allow us in the near future to explore protein dynamics on time scales well beyond the microsecond. The success of CG simulations paves the way to new developments in computer simulations of membrane proteins, with multiscale approaches among the most promising. In multiscale approaches, different resolutions can be applied in sequence or at the same time. The first case is more simple, since there is no interaction between descriptions at different levels of resolution. The second is more difficult to implement but also potentially more powerful, in that it allows to explore protein dynamics with the accuracy of all-atoms representations on length and time scales typical of coarse-grained representation. Different methodologies have recently been developed to simulate biological macromolecules simultaneously at different length and time scales. In the first approach, part of the system is treated with an atomistic representation and part with a less detailed model. This includes methodologies in which an atomistic protein is embedded in a continuum membrane (93, 94) and methods in which some molecules are described with atomistic detail and others at the coarse-grained level (95–97). In principle, different resolutions could also be used on different

436

Kandt and Monticelli

parts of the same molecule, like in the QM–MM approach. A large gain in the simulation speed is achieved through the reduction of degrees of freedom in the “less interesting” parts of the system, while higher accuracy is maintained in the “more interesting” parts. Another approach, developed by Lyman and Zuckerman, is based on the resolution exchange method: simulations are performed at the same time at different resolutions, from all-atoms to increasingly coarse-grained, and an exchange is attempted at discrete time intervals (98, 99). Energy barriers for large conformational changes are crossed easily in simulations running at lower resolution, while canonical sampling is achieved at the atomistic level. In all these approaches, the main difficulty lies in modeling correctly the interaction between components with different resolution. While it is difficult to quantify the sampling efficiency and the speed-up in the protein dynamics, multi-scale techniques allow us to explore conformational transformations that would take extremely long times, presently out of reach for traditional molecular dynamics methods. 2.6. Concluding Remarks

Computer simulations have become increasingly important tools in structural biology and particularly in the field of membrane protein studies. They are used both as an aid in the interpretation of experimental data and in the prediction of structure and dynamics. Depending on the details of the problem of interest, and particularly on the length scale and the time scale of the phenomena, a vast range of techniques can be employed, ranging from quantum mechanics to mesoscale approaches. In the present work, we have reviewed the basic methodological aspects of some of the most common simulation techniques, and described a few examples of their application to real-life problems. While current techniques allow us to explore membrane protein dynamics on time scales up to tens of microseconds, progress in computer technology and the ongoing development of multi-scale approaches bear the possibility to extend simulations beyond this limit in the near future.

Acknowledgments We thank Frank Wennmohs and Emppu Salonen for their fruitful discussions. This work is supported by the Academy of Finland and its Center of Excellence, and by the Ministerium für Innovation, Wissenschaft, Forschung und Technologie des Landes Nordrhein-Westfalen. CK is a junior research group leader funded by the NRW Rückkehrerprogramm.

Membrane Protein Dynamics from Femtoseconds to Seconds

437

References 1. Veldhuizen EJ, Haagsman HP (2000) Role of pulmonary surfactant components in surface film formation and dynamics. Biochim Biophys Acta 1467:255–270 2. Bernstein FC, Koetzle TF, Williams GJ etÂ€al (1977) The Protein Data Bank: a computerbased archival file for macromolecular structures. J Mol Biol 112:535–542 3. Berman HM, Westbrook J, Feng Z etÂ€ al (2000) The Protein Data Bank. Nucleic Acids Res 28:235–242 4. Berman H, Henrick K, Nakamura H (2003) Announcing the worldwide Protein Data Bank. Nat Struct Biol 10:980 5. White SH (2004) The progress of membrane protein structure determination. Protein Sci 13:1948–1949 6. Leach AR (2001) Molecular modelling: principles and applications. Pearson Education, Harlow 7. Pauling L, Wilson EBJ (1985) Introduction to quantum mechanics with applications to chemistry. Courier Dover, New York 8. Neese F (2003) An improvement of the resolution of the identity approximation for the formation of the Coulomb matrix. J Comput Chem 24:1740–1747 9. Car R, Parrinello M (1985) Unified approach for molecular dynamics and density-functional theory. Phys Rev Lett 55:2471–2474 10. Messiah A (1963) Quantum mechanics. Wiley, New York 11. Jensen F (2006) Introduction to computational chemistry. Wiley, New York 12. Mulholland AJ (2007) Chemical accuracy in QM/MM calculations on enzyme-catalysed reactions. Chem Cent J 1:19 13. Mulholland AJ (2005) Modelling enzyme reaction mechanisms, specificity and catalysis. Drug Discov Today 10:1393–1402 14. Warshel A, Levitt M (1976) Theoretical studies of enzymic reactions: dielectric, electrostatic and steric stabilization of the carbonium ion in the reaction of lysozyme. J Mol Biol 103:227–249 15. Kandt C, Ash WL, Tieleman DP (2007) Setting up and running molecular dynamics simulations of membrane proteins. Methods 41:475–488 16. Tieleman DP, Marrink SJ, Berendsen HJC (1997) A computer perspective of membranes: molecular dynamics studies of lipid bilayer systems. Biochim Biophys Acta 1331:235–270

17. Levitt M, Warshel A (1975) Computer simulation of protein folding. Nature 253: 694–698 18. Shillcock JC, Lipowsky R (2006) The computational route from bilayer membranes to vesicle fusion. J Phys Condens Matter 18:S1191–S1219 19. Sperotto MM, May S, Baumgaertner A (2006) Modelling of proteins in membranes. Chem Phys Lipids 141:2–29 20. Venturoli M, Sperotto MM, Kranenburg M, Smit B (2006) Mesoscopic models of biological membranes. Phys Rep 437:1–54 21. Müller M, Katsov K, Schick M (2006) Biological and synthetic membranes: what can be learned from a coarse-grained description? Phys Rep 434:113–176 22. Monticelli L, Kandasamy SK, Periole X, Larson RG, Tieleman DP, Marrink SJ (2008) The MARTINI coarse-grained force field: Extension to proteins. J Chem Theory Comput 4:819–834 23. Grossfield A, Feller SE, Pitman MC (2007) Convergence of molecular dynamics simulations of membrane proteins. Proteins 67:31–40 24. Gumbart J, Wang Y, Aksimentiev A, Tajkhorshid E, Schulten K (2005) Molecular dynamics simulations of proteins in lipid bilayers. Curr Opin Struct Biol 15:423–431 25. Kandt C, Mátyus E, Tieleman DP (2008) Protein lipid interactions from a molecular dynamics simulation point of view. In: Nag K (ed) Structure & dynamics of membranous interfaces. Wiley-Interscience, Hoboken, NJ, pp 267–282 26. Lindahl E, Sansom MS (2008) Membrane proteins: molecular dynamics simulations. Curr Opin Struct Biol 18:425–431 27. Doyle DA, Morais Cabral J, Pfuetzner RA etÂ€ al (1998) The structure of the potassium channel: molecular basis of K+ conduction and selectivity. Science 280:69–77 28. Jiang YX, Lee A, Chen JY, Cadene M, Chait BT, MacKinnon R (2002) Crystal structure and mechanism of a calcium-gated potassium channel. Nature 417:515–522 29. Jiang YX, Lee A, Chen JY etÂ€al (2003) X-ray structure of a voltage-dependent K+ channel. Nature 423:33–41 30. Kuo AL, Gulbis JM, Antcliff JF etÂ€al (2003) Crystal structure of the potassium channel KirBac1.1 in the closed state. Science 300:1922–1926

438

Kandt and Monticelli

31. Long SB, Campbell EB, MacKinnon R (2005) Crystal structure of a mammalian voltagedependent Shaker family K+ channel. Science 309:897–903 32. Bucher D, Raugei S, Guidoni L etÂ€al (2006) Polarization effects and charge transfer in the KcsA potassium channel. Biophys Chem 124:292–301 33. Bucher D, Guidoni L, Rothlisberger U (2007) The protonation state of the Glu-71/Asp-80 residues in the KcsA potassium channel: A first-principles QM/MM molecular dynamics study. Biophys J 93:2315–2324 34. Gether U (2000) Uncovering molecular mechanisms involved in activation of G protein-coupled receptors. Endocr Rev 21:90–113 35. Cherezov V, Rosenbaum DM, Hanson MA etÂ€al (2007) High-resolution crystal structure of an engineered human beta2-adrenergic G protein-coupled receptor. Science 318:1258–1265 36. Rasmussen SGF, Choi H-J, Rosenbaum DM etÂ€ al (2007) Crystal structure of the human beta2 adrenergic G-protein-coupled receptor. Nature 450:383–387 37. Luecke H, Schobert B, Lanyi JK, Spudich EN, Spudich JL (2001) Crystal structure of sensory rhodopsin II at 2.4 angstroms: insights into color tuning and transducer interaction. Science 293:1499–1503 38. Kandt C, Gerwert K, Schlitter J (2005) Water dynamics simulation as a tool for probing proton transfer pathways in a heptahelical membrane protein. Proteins 58:528–537 39. Kandt C, Schlitter J, Gerwert K (2004) Dynamics of water molecules in the bacteriorhodopsin trimer in explicit lipid/water environment. Biophys J 86:705–717 40. Zgrabli G, Voãtchovsky K, Kindermann M, Haacke S, Chergui M (2005) Ultrafast excited state dynamics of the protonated Schiff base of all-trans retinal in solvents. Biophys J 88:2779–2788 41. Schenkl S, van Mourik F, van der Zwan G, Haacke S, Chergui M (2005) Probing the ultrafast charge translocation of photoexcited retinal in bacteriorhodopsin. Science 309:917–920 42. Okada T, Ernst OP, Palczewski K, Hofmann KP (2001) Activation of rhodopsin: new insights from structural and biochemical studies. Trends Biochem Sci 26:318–324 43. Meng EC, Bourne HR (2001) Receptor activation: what does the rhodopsin structure tell us? Trends Pharmacol Sci 22:587–593

44. Lemaitre V, Yeagle P, Watts A (2005) Molecular dynamics simulations of retinal in rhodopsin: from the dark-adapted state towards lumirhodopsin. Biochemistry 44:12667–12680 45. Isin B, Rader AJ, Dhiman HK, KleinSeetharaman J, Bahar I (2006) Predisposition of the dark state of rhodopsin to functional changes in structure. Proteins 65:970–983 46. Martinez-Mayorga K, Pitman MC, Grossfield A, Feller SE, Brown MF (2006) Retinal counterion switch mechanism in vision evaluated by molecular simulations. J Am Chem Soc 128:16502–16503 47. Crozier PS, Stevens MJ, Woolf TB (2007) How a small change in retinal leads to G-protein activation: initial events suggested by molecular dynamics calculations. Proteins 66:559–574 48. Kong Y, Karplus M (2007) The signaling pathway of rhodopsin. StructureÂ€15:611–623 49. Grossfield A, Feller SE, Pitman MC (2006) A role for direct interactions in the modulation of rhodopsin by omega-3 polyunsaturated lipids. Proc Natl Acad Sci U S A 103:4888–4893 50. Cordomi A, Perez JJ (2007) Molecular dynamics simulations of rhodopsin in different one-component lipid bilayers. J Phys Chem B 111:7052–7063 51. Periole X, Huber T, Marrink SJ, Sakmar TP (2007) G protein-coupled receptors selfassemble in dynamics simulations of model bilayers. J Am Chem Soc 129:10126–10132 52. Niv MY, Filizola M (2008) Influence of oligomerization on the dynamics of G-protein coupled receptors as assessed by normal mode analysis. Proteins 71:575–586 53. Yan ECY, Kazmi MA, Ganim Z etÂ€ al (2003) Retinal counterion switch in the photoactivation of the G protein-coupled receptor rhodopsin. Proc Natl Acad Sci U S A 100:9262–9267 54. Ludeke S, Beck R, Yan ECY, Sakmar TP, Siebert F, Vogel R (2005) The role of Glu181 in the photoactivation of rhodopsin. J Mol Biol 353:345–356 55. Ivetac A, Campbell JD, Sansom MSP (2007) Dynamics and function in a bacterial ABC transporter: simulation studies of the BtuCDF system and its components. Biochemistry 46:2767–2778 56. Sonne J, Kandt C, Peters GH, Hansen FY, Jensen M, Tieleman DP (2007) Simulation of the coupling between nucleotide binding and transmembrane domains in the ATP binding cassette transporter BtuCD. Biophys J 92:2727–2734

Membrane Protein Dynamics from Femtoseconds to Seconds 57. Arkin IT, Xu H, Mù J etÂ€al (2007) Mechanism of Na+/H+ antiporting. Science 317:799–803 58. Law CJ, Almqvist J, Bernstein A etÂ€al (2008) Salt-bridge dynamics control substrate-induced conformational change in the membrane transporter GlpT. J Mol Biol 378:826–837 59. Lu WC, Wang CZ, Yu EW, Ho KM (2006) Dynamics of the trimeric AcrB transporter protein inferred from a B-factor analysis of the crystal structure. Proteins 62:152–158 60. Holyoake J, Sansom MSP (2007) Conformational change in an MFS protein: MD simulations of LacY. StructureÂ€15:873–884 61. Davidson AL, Dassa E, Orelle C, Chen J (2008) Structure, function, and evolution of bacterial ATP-binding cassette systems. Microbiol Mol Biol Rev 72:317–364 62. Moussatova A, Kandt C, O’Mara ML, Tieleman DP (2008) ATP-binding cassette transporters in Escherichia coli. Biochim Biophys Acta 1778:1757–1771 63. Tirion MM (1996) Large amplitude elastic motions in proteins from a single-parameter, atomic analysis. Phys Rev Lett 77:1905–1908 64. Hunte C, Screpanti E, Venturi M, Rimon A, Padan E, Michel H (2005) Structure of a Na+/H+ antiporter and insights into mechanism of action and regulation by pH. Nature 435:1197–1202 65. Inoue H, Noumi T, Tsuchiya T, Kanazawa H (1995) Essential aspartic acid residues, Asp-133, Asp-163 and Asp-164, in the transmembrane helices of a Na+/H+ antiporter (NhaA) from Escherichia coli. FEBS Lett 363:264–268 66. Song L, Hobaugh MR, Shustak C, Cheley S, Bayley H, Gouaux JE (1996) Structure of staphylococcal alpha-hemolysin, a heptameric transmembrane pore. Science 274:1859–1866 67. Akeson M, Branton D, Kasianowicz JJ, Brandin E, Deamer DW (1999) Microsecond time-scale discrimination among polycytidylic acid, polyadenylic acid, and polyuridylic acid as homopolymers or as segments within single RNA molecules. Biophys J 77:3227–3233 68. Wells DB, Abramkina V, Aksimentiev A (2007) Exploring transmembrane transport through alpha-hemolysin with grid-steered molecular dynamics. J Chem Phys 127:125101–125110 69. Prince SM, Achtman M, Derrick JP (2002) Crystal structure of the OpcA integral membrane adhesin from Neisseria meningitidis. Proc Natl Acad Sci U S A 99:3417–3421 70. Cherezov V, Liu W, Derrick JP etÂ€al (2008) In meso crystal structure and docking simulations suggest an alternative proteoglycan

71.

72. 73.

74.

75. 76.

77.

78. 79.

80. 81.

82.

83.

439

binding site in the OpcA outer membrane adhesin. Proteins 71:24–34 Luan B, Caffrey M, Aksimentiev A (2007) Structure refinement of the OpcA adhesin using molecular dynamics. Biophys J 93:3058–3069 Shi N, Ye S, Alam A, Chen L, Jiang Y (2006) Atomic structure of a Na+- and K+-conducting channel. Nature 440:570–574 Noskov SY, Roux B (2007) Importance of hydration and dynamics on the selectivity of the KcsA and NaK channels. J Gen Physiol 129:135–143 Sperotto MM, Mouritsen OG (1991) MonteCarlo simulation studies of lipid order parameter profiles near integral membrane-proteins. Biophys J 59:261–270 Kranenburg M, Venturoli M, Smit B (2003) Molecular simulations of mesoscopic bilayer phases. Phys Rev E 67:060901 Venturoli M, Smit B, Sperotto MM (2005) Simulation studies of protein-induced bilayer deformations, and lipid-induced protein tilting, on a mesoscopic model for lipid bilayers with embedded proteins. Biophys J 88:1778–1798 Ramakrishnan M, Qu J, Pocanschi CL, Kleinschmidt JH, Marsh D (2005) Orientation of beta-barrel proteins OmpA and FhuA in lipid membranes. chain length dependence from infrared dichroism. Biochemistry 44:3515–3523 Marrink SJ, de Vries AH, Mark AE (2004) Coarse grained model for semiquantitative lipid simulations. J Phys Chem B 108:750–760 Marrink SJ, Risselada HJ, Yefimov S, Tieleman DP, de Vries AH (2007) The MARTINI forcefield: coarse grained model for biomolecular simulations. J Phys Chem B 111:7812–7824 Engelman DM (2005) Membranes are more mosaic than fluid. Nature 438:578–580 Fotiadis D, Liang Y, Filipek S, Saperstein DA, Engel A, Palczewski K (2003) Atomic-force microscopy: rhodopsin dimers in native disc membranes. Nature 421:127–128 Cruickshank CC, Minchin RF, Le Dain AC, Martinac B (1997) Estimation of the pore size of the large-conductance mechanosensitive ion channel of Escherichia coli. Biophys J 73:1925–1931 Chang G, Spencer RH, Lee AT, Barclay MT, Rees DC (1998) Structure of the MscL homolog from Mycobacterium tuberculosis: a gated mechanosensitive ion channel. Science 282:2220–2226

440

Kandt and Monticelli

84. Yefimov S, van der Giessen E, Onck PR, Marrink SJ (2008) Mechanosensitive membrane channels in action. Biophys J 94:2994–3002 85. Duan Y, Pa K (1998) Pathways to a protein folding intermediate observed in a 1-microsecond simulation in aqueous solution. Science 282:740–744 86. Zagrovic B, Snow CD, Shirts MR, Pande VS (2002) Simulation of folding of a small alphahelical protein in atomistic detail using worldwide-distributed computing. J Mol Biol 324:1051 87. Monticelli L, Tieleman DP, Colombo G (2005) Mechanism of helix nucleation and propagation: microscopic view from microsecond time scale MD simulations. J Phys Chem B 109:20064–20067 88. Munoz V, Thompson PA, Hofrichter J, Eaton WA (1997) Folding dynamics and mechanism of beta-hairpin formation. Nature 390:196–199 89. Popot JL, Engelman DM (1990) Membrane protein folding and oligomerization – the 2-stage model. Biochemistry 29:4031–4037 90. Tamm LK, Hong H, Liang BY (2004) Folding and assembly of beta-barrel membrane proteins. Biochim Biophys Acta 1666:250–263 91. Bond PJ, Sansom MSP (2006) Insertion and assembly of membrane proteins via simulation. J Am Chem Soc 128:2697–2704

92. Deol SS, Domene C, Bond PJ, Sansom MSP (2006) Anionic phospholipid interactions with the potassium channel KcsA: simulation studies. Biophys J 90:822–830 93. Im W, Feig M, Brooks CL III (2003) An implicit membrane generalized born theory for the study of structure, stability, and interactions of membrane proteins. Biophys J 85:2900–2918 94. Ayton G, Bardenhagen SG, McMurtry P, Sulsky D, Voth GA (2001) Interfacing continuum and molecular dynamics: an application to lipid bilayers. J Chem Phys 114:6913–6924 95. Villa E, Balaeff A, Mahadevan L, Schulten K (2004) Multiscale method for simulating protein–DNA complexes. Multiscale Model Simul 2:527–553 96. Villa E, Balaeff A, Schulten K (2005) Structural dynamics of the lac repressor–DNA complex revealed by a multiscale simulation. Proc Natl Acad Sci U S A 102:6783–6788 97. Shi Q, Izvekov S, Voth GA (2006) Mixed atomistic and coarse-grained molecular dynamics: simulation of a membrane-bound ion channel. J Phys Chem B 110:15045–15048 98. Lyman E, Ytreberg FM, Zuckerman DM (2006) Resolution exchange simulation. Phys Rev Lett 96:028105 99. Lyman E, Zuckerman DM (2006) Resolution exchange simulation with incremental coarsening. J Chem Theory Comput 2:656–666

Chapter 23 The Family of G Protein-Coupled Receptors: An Example of Membrane Proteins Irina G. Tikhonova and Daniel Fourmy Abstract The G protein coupled receptors belong to the largest group of membrane proteins that regulates many essential physiological properties and represents an important class of drug targets. In this chapter, we show how the synergy between a laboratory experiment and computational modeling leads to structural delineation of the ligand binding pocket and how the knowledge of ligand–protein interactions is used for rational local and global drug design in which the structural knowledge of a particular receptor and its ligands is used for drug design of this particular GPCR and others. Key words: G protein coupled receptor, Cholecystokinin, Molecular modeling, Binding site, Site-directed mutagenesis, Drug design

1. Introduction The G protein-coupled receptors (GPCRs) are the largest group of cell surface membrane proteins, which form a transmembrane bundle composed of seven membrane-spanning alpha helices connected by loop regions. Binding with endogenous mediators causes conformational changes in GPCRs (1, 2) that lead to high affinity interaction of GPCRs with the cognate G protein and in turn initiate numerous downstream signaling pathways in cells (3, 4). GPCRs are signal transmitters for hormones, neuromediators, cytokines, lipids, peptides, small molecules, and various sensory exogenous stimuli, such as light, odors, and taste. Consequently, GPCRs are regulators of many life important cell processes and universal drug targets against various diseases. Because of structure-based drug design is a rational way to design novel small molecule ligands and to improve binding and Jean-Jacques Lacapère (ed.), Membrane Protein Structure Determination: Methods and Protocols, Methods in Molecular Biology, vol. 654, DOI 10.1007/978-1-60761-762-4_23, © Springer Science+Business Media, LLC 2010

441

442

Tikhonova and Fourmy

selectivity of old drugs, many laboratories have been working on structural delineation of the ligand binding site of GPCRs. As membrane proteins, GPCRs have difficulties to be expressed, purified and crystallized in a large scale. Therefore, there is a limited number of high-resolution GPCR structures today: bovine and squid rhodopsin (5–9) opsin (10), b2 and b1 adrenergic (11–14), and A2A adenosine (15) receptors. For decades, the structural insight of GPCRs ligand binding site has being mainly gained using indirect methods such as receptor mutagenesis, ligand structure– activity relationships (SAR), and receptor modeling. Thus, till 2007, only the crystal structure of bovine rhodopsin was available and used to construct the low-resolution homology-based models of GPCRs. Although, the sequence homology between light-activated rhodopsin and ligand-activated GPCRs is low (about 11–17%) and the second extracellular loop (EL2) of rhodopsin structure buries deeply into the helical bundle and closes a putative ligand binding cavity, several modeling strategies have been developed to delineate the ligand binding site of ligand-activated GPCRs based on the rhodopsin structure. In one approach, the EL2 was removed before the docking of known ligands and added back when the ligand interactions with transmembrane helices were defined by molecular docking and receptor mutagenesis (16, 17). In another approach, forced molecular dynamics simulations were applied straight away to drive the ligand–protein interactions in the initial homology model according to experimental data (18–20). In addition, molecular simulations of the rhodopsin-based homology model in water– lipid environment were used for the formation of a binding cavity by initially added spheres with a flexible Van der Waals radius, and the formed cavity then were employed for docking of ligands (21). In all these approaches, the iterative application of computational and experimental studies to establish ligand–protein interactions was important, in which molecular modeling results were used to generate hypothesis that were then validated by receptor mutagenesis, and the mutagenesis results were used, in turn, to refine computational models. Moreover, the docking of different available ligands into the binding cavity and its experimental validation allowed better to shape and optimize the binding site. The recently published first crystal structures of beta 1,2-adrenergic receptors (11–14) have proved the earlier anticipated larger volume of the binding cavity and the solvent exposed conformation of the EL2 in the ligand-activated GPCRs. The homology modeling based on these structures will simplify the construction of the ligand binding sites of GPCRs; hence, the ligand–protein contacts are still required to be proved by mutagenesis data, especially when the sequence homology of residues in the binding cavity between a template structure and a modeled GPCR is very low.

The Family of G Protein-Coupled Receptors: An Example of Membrane Proteins

443

In this review, we show the case study of the ligand binding site delineation for cholecystokinin receptor named as CCK1 by the International Union of Pharmacology that has been done in our laboratory for the past ten years and others, and then we provide the rational strategies for local and global GPCRs drug discovery in which the information of the ligand binding site of a particular GPCR is used for drug design of the particular GPCR and others.

2. Materials Cholecystokinin (CCK) is a regulatory peptide having high affinity for CCK receptors. Cholecystokinin shares its carboxylterminal pentapeptide sequence with gastrin, another regulatory peptide. Cholecystokinin and gastrin differ in their selectivity for the two CCK receptor subtypes, the CCK1 (CCK1R) and the CCK2 (CCK2R) receptors, on the basis of tyrosine sulfation at the seventh position (CCK) or at the sixth position (gastrin) from the carboxyl-terminus (22). CCK1R are mainly found in the periphery where they regulate pancreatic secretion, gallbladder, and gastrointestinal motility, but are also found in some areas of central nervous system where they regulate satiety and analgesia. At present, a large set of converging data related to binding sites of CCK1R is currently available, giving a good picture of the binding mode of natural and synthetic ligands to this receptor. The data were provided using essentially four complementary approaches, site-directed mutagenesis, photoaffinity labeling, NMR-NOE transfer, and three-dimensional modeling. In the laboratory, we have combined the use of site-directed mutagenesis and molecular modeling to delineate the binding sites of the CCK1R. A model of the CCK1R was constructed by homology modeling and refined on the basis of site-directed mutagenesis data as the homology model of the CCK1R constructed on the basis of rhodospin template could not accommodate the peptidic ligand CCK into its binding site (23).

3. Methods 3.1. Mapping of CCK1R Binding Site: Synergy Between Laboratory and In Silico Experiments

The first contact point between the CCK1R and CCK was defined on the basis of photoaffinity labeling results (24). Indeed, in the course of biochemical studies of the pancreatic CCK1R, we identified a CCK1R truncated from its terminal moiety, using a peptidic antagonist (JMV 179) as a photoaffinity label.

444

Tikhonova and Fourmy

As this truncated CCK1R was not labeled by a CCK-derived photoaffinity label, we hypothetized that the lacking region in the truncated CCK1R contained the attachment site of the CCKderived photoaffinity label. Moreover, in a competition assay, the labeled truncated CCK1R seemed capable of interacting with CCK, but with a lower affinity than did the intact CCK1R. These biochemical results were a first indication that residue(s) located between the N-terminus of the CCK1R and top of TM I were probably in interaction with the N-terminal moiety of CCK through which the photoaffinity group was attached (24). The next step was the construction and analysis of the N-terminally truncated CCK1R, which led to the identification of a region at the junction between the transmembrane helix 1 and the N-terminal moiety, between residues 38–42, which was involved in the binding of CCK (25). Residues Trp39 and Gln40 of the receptor were then shown to be important for recognition of the C-terminal nonapeptide of CCK as Trp39Phe and Gln40Asn mutants demonstrated parallel decreases in both binding affinity, and potency to induce accumulation of inositol phosphates (26). To determine which region of CCK interacts with Trp39 and Gln40 at the binding site, we compared binding affinities of Trp39Phe and Gln40Asn mutants for CCK analogs modified at their C- and N-terminal ends. We postulated that only peptides containing residues capable of interacting with aminoacids Trp39 and Gln40 of the receptor would bind to the mutated receptors Trp39Phe and Gln40Asn with decreased affinities relative to the wild-type receptor. Such experiments identified residues at the N-terminal of the nonapeptide of CCK as likely in interaction with Trp39 and Gln40 (26). At this stage of the work, a three dimensional model of the CCK1R was constructed using transmembrane helical arrangement found in the bacteriorhodopsin crystal structure as starting point since at this time, the high resolution crystal of rhodopsin was not available yet (27). The model was then modified according to the Baldwin model for G-protein coupled receptors and to the mutant data-base “input/output” information scheme defined in the Viseur program (28). Extracellular and intracellular loops connecting the helices were then added to the preliminary 7-helix bundle, and the structural model was optimized by the use of simulated annealing procedures. The entire system was finally relaxed and submitted to one nanosecond molecular dynamics with possible translational and rotational movements of individual TM helices taken into account. The positioning of the CCK peptide in the modeled CCK1R was achieved using a docking model in which the C-terminal moiety of the CCK-related ligands was placed in the middle of the receptor transmembrane region, while the N-terminal part was positioned in its extracellular region near the entrance of the putative binding pocket.

The Family of G Protein-Coupled Receptors: An Example of Membrane Proteins

445

Fig.Â€1. Ligands of the CCK1 receptor.

By doing so, it appeared that the N-terminal residues of CCK nonapeptide (Fig.Â€ 1) could be easily connected to the receptor through a strong hydrogen bond and salt bridge network involving residues Trp39 and Gln40 identified experimentally in our site-directed mutagenesis-based study (26). Based on this first set of data, subsequent studies were devoted to identification of determinants of the receptor in interaction with other key residues of CCK. We constrained the N-terminus of CCK in interaction with Trp39 and Gln40, while the rest of CCK was positioned inside the receptor grove, allowing inspection of the molecular electrostatic potentials. Then, the docking was improved by simulated annealing calculations. The resulting structure obtained for the ligand/receptor complex was further refined using molecular dynamics and energy minimization. In a second step, experiments were performed by mutating candidate residue(s) of the receptor binding site, which had been depicted in the 3D model, and analyzing extensively effects of mutation(s). By doing so, several critical contacts between the CCK1R binding site and CCK were successively discovered: Met195 and Arg197, located in the second extracellular loop, were shown to interact with the sulfated tyrosine (29, 30); Arg336 and Asn333 at the top of helix VI were demonstrated to pair with the Asp carboxylate and the C-terminal amide of CCK, respectively (31). Subsequently, a network of hydrophobic residues from helices III, V, VI, and VII all forming a binding pocket for the CCK C-terminal region was identified. This binding mode of the C-terminus of CCK into CCK1R thus obtained appeared in agreement with an NMR study of the interactions between CCK and a fragment of CCK1R comprising the top portion of helix VI and the third extracellular loop as well as a fragment including amino acids at the top of transmembrane segment I (32, 33). Using photoaffinity labeling, two

446

Tikhonova and Fourmy

hits in the CCK1R were identified. The first was a Trp at the top of TM I using a photoprobe with the reactive moiety within the C-terminal Phe of CCK, and the second was a His within the third extracellular loop using a probe with a benzophenone in the place of the Gly of CCK (Fig.Â€2) (34, 35). Accordingly, a model of binding of CCK to the CCK1R was proposed in which the C-terminus of CCK and the tyrosine sulfate were in interaction with Trp39 and Arg197, respectively, and the N-terminal moiety was in contact with the third extracellular loop of the receptor (36). This

Fig.Â€2. Serpentine representation of the CCK1R depicting residues involved in the binding site of the full agonist CCK (a). Schematic representation of the CCK1R binding site with docked CCK (b).

The Family of G Protein-Coupled Receptors: An Example of Membrane Proteins

447

second model for CCK binding mode into the CCK1R binding site is somewhat divergent from that obtained on the basis of sitedirected mutagenesis results (19). Further key support for the model placing the C-terminal end of CCK in an hydrophobic cavity formed by helices III, V, VI, and VII came from a study dedicated to the understanding of the molecular basis for partial agonist activity of JMV 180 (Fig.Â€1), a CCK analog having the C-terminal amidated phenylalanine substituted by a phenylethyl ester (37). In this study combining site-directed mutagenesis experiments, the use of CCK-related peptides modified at their C-terminus and molecular modeling, we demonstrated that partial agonism of JMV180 was due to flexibility of the phenylethyl ester moiety of the ligand which could not allow optimal interaction with key aromatic residue for receptor activation in helices V and VI (38). The key role in the process of receptor activation of such aromatic residue (especially Trp6.48) is a general feature of the group I of GPCRs, as it was explored for instance in rhodopsin and more recently in the CCK2R (39, 40). The structure of CCK1R.CCK complex obtained on the basis of site-directed mutagenesis data was used to study the binding site of several synthetic nonpeptide agonists and antagonists. According to docking and experimental data, nonpeptide agonists and antagonists most likely occupy a region in CCK receptors that interacts with the amidated tetrapeptide of the C-terminal part of CCK, i.e., Trp-Met-Asp-Phe-NH2, and moreover, there is an overlap between agonist and antagonist binding sites. For example, we showed that structurally related compounds, the agonist SR-146131 and the antagonist SR-27,897 of CCK1R (Fig.Â€1), interact with Arg336(6.58) and Asn333(6.55) and locate in two hydrophobic sub-pockets composed of amino acids of helices III/VI and I/III/VII, respectively (19, 41). Interestingly, in the case of agonist SR-146,131, there is a hydrophobic interaction between cyclohexane moiety of the ligand and Leu356(7.39) of the CCK1R, which is lacking in the binding mode of the antagonist SR-27897 (19, 41). We further combined site-directed mutagenesis studies, SAR, and dynamic-based docking in order to identify the binding site of pyridopyrimidine-derived antagonists of the CCK1R and understand their selectivity towards this receptor type versus the CCK2R (42). This study, again, provided unambiguous evidence that the binding site of these antagonists is overlapping that of the C-terminal tetrapeptide of CCK. However, in the course of this study, several orientations for the ligand within the binding site were found, with an identical probability of occurrence. Only combination of experimental and SAR data enabled us to propose a very likely binding mode (42). Collectively, data provided by these studies validated the binding site of the CCK1R.

448

Tikhonova and Fourmy

A similar strategy as that used for the CCK1R was applied to the CCK2R leading to a complete delineation of its binding site (43–46). Moreover, on this latter CCK receptor subtype, we worked at identification of networks involved in receptor switching from the inactive state to the active state (40). The computational approach consisted in modeling the active conformation from the inactive-rhodopsin-based model, using sets of experimental restraints, which characterize the active conformation in family A, and CCK interactions, identified by biophysical and site directed mutagenesis techniques, respectively (40). The exercise of comparison of both conformations in the CCK2R led to the identification of different network rearrangements, which have been experimentally validated to be involved in the control of the equilibrium between the two conformations and therefore in the mechanism of activation. In this study, conversion of the CCK2R from the inactive to active conformation was studied by targeted molecular dynamics, a method based on MD previously applied to other protein families (47). This method was then applied to complexes formed between the CCK2R and two structurally very closely related nonpeptide ligand having either partial agonist activity or inverse agonist activity. Targeted molecular dynamics was able to discriminate the two compounds; however, correct docking of the compounds into the CCK2R binding pocket required experimental investigations based on site-directed mutagenesis (48). 3.2. Local GPCR Drug Design

The general flowchart of structure-based GPCR drug discovery is presented in Fig.Â€3. In the previous paragraph, we show on the example of CCK1R how the structural knowledge of the ligand binding site of a GPCR was provided using a synergy of modeling and experimental data. Once the ligand binding site is delineated, a structure-based lead search campaign can be carried out using

Fig.Â€3. The G-protein coupled receptors structure-base local drug design flowchart.

The Family of G Protein-Coupled Receptors: An Example of Membrane Proteins

449

high-throughput docking or/and receptor-based pharmacophore search. In the high-throughput docking, each small molecule of chemical libraries is placed inside of the binding cavity based on preferable electrostatic and Van der Waals interactions with a receptor, and the selection of a focused library for the experimental test is performed using a chosen threshold value of scoring functions, which is established with docking results of known ligands. Since scoring functions are simplified and rapid estimation of a binding energy and homology-based 3D structures of GPCRs ligand binding sites are low-resolution models, it has been shown that rescoring of docking results by different force field-based, knowledge-based and, empirical scoring functions and/or their consensus scoring improve the discrimination of known compounds from the random decoy in the retrospective virtual screenings for dopamine D3, muscarinic M1, vasopressin V1a (18), cannabinoid 2 receptor (49), 5-hydroxytryptamine receptors 5-HT2c (50), and chemokine CCR1 (51). Moreover, recently applied for ranking of the metabotropic glutamate receptor (mGluR) subtype 5 and b2 adrenergic receptor ligands, ligand-receptor interaction fingerprint-based similarity has shown higher enrichment rates relative to the scoring functions (52, 53). When a GPCR has already plenty of diverse ligands, for example, biogenic GPCRs, the application of docking-based virtual screening with ligand-based methods, such as ligand-based pharmacophores and QSAR (quantitative structure activity relationships) models, increases significantly the enrichment rates in the retrospective virtual screenings (16). The docking-based virtual screening for alpha 1 adrenergic receptor (54), melanin-concentrating hormone receptor (55), serotonin 5HT1 and 5HT4, tachykinin NK1 receptor, dopamine D2 receptor, chemokine CCKR3 (56) and CCKR5 (57) receptor, free fatty acid receptor 1 FFAR1 (58), and thyrotropin-releasing hormone receptor TRH1 (59) lead to the identification of novel ligands with micro/nano molar activity. In receptor-based pharmacophore search, the ligand binding site is represented as a 3D-pharmacophore, in which important for ligand binding amino acid residues are coded as acceptor, donor, hydrophobic, or other features, and the shape of binding cavity is described by a molecular surface and formed a spatial constraint on the possible ligand atom location. The screening library of small molecules is flexibly aligned to the pharmacophore, and the focused library is selected using the defined threshold value of root-mean square distance (RMSD) between the query features and matching ligand atoms, and of Van der Waals (VdW) radii of selected atoms for spatial constraints. The threshold value of RMSD and VdW radii has to preliminary set using benchmark calculation of a set of known ligands in a random decoy.

450

Tikhonova and Fourmy

The application of this search was shown in retrospective screening of the metabotropic glutamate receptors (60). Since this method is too coarse-grained, it provides more rough and large libraries, which can be minimized by a subsequent docking procedure. Indeed, a receptor-based pharmacophore filter was used to create a focused library for TRH1 receptor in the multistep virtual screening leading to identification of diverse high potent antagonists (59). The structure-based optimization campaign is devoted to improve the binding properties of pharmacologically confirmed lead compounds or already known ligands by creating structural analogs using the information about the ligand binding site. The accurate docking of lead compounds provides visual information about what chemical groups can be incorporated to improve the affinity. 3D-quantitative structure–activity relationship models (QSAR) can be built using the set of known analogs with experimentally measured binding affinities and knowledge of ligand– protein interactions to predict in silico affinities of novel analogs. As learnt from previous paragraphs, the GPCR ligand binding site can be used for the discovery of novel ligands and improvement of the binding affinity. Can we use the ligand binding site to predict the efficacy of the ligands? Intuitively, to predict the agonistic and antagonistic properties, the knowledge of different GPCR conformations is required, especially the conformations of active and inactive states. It has been shown that the ligand binding site optimized based on the full potent agonist allows to better discriminate the agonists in retrospective screening than the ligand binding site optimized based on antagonists (18, 53). Indeed, the virtual screening of the FFA1 agonist binding site led to the discovery of full and partial agonists (58). 3.3. Global GPCRs Drug Design

For decades, many laboratories have been working to integrate chemistry, biology, and medicine of GPCRs with the aim of rational discovery of novel potent drugs. As a result of this work, huge amount of data related to structure-function relationships of receptors and SAR of the small molecules have been produced for many GPCRs. As gaining from the previous sections, structural information of the ligand binding site of GPCRs has been learnt indirectly through synergetic application of site-directed mutagenesis and molecular modeling. The information about GPCR mutagenesis data, sequences, and modeling is collected in several web pages (Fig.Â€4). A large number of small molecules were designed using medicinal chemistry, high-throughput and virtual screening approaches and tested for GPCR activity. The information about known ligands and their biological activity was used in turn to predict novel molecules using receptor and ligand based

The Family of G Protein-Coupled Receptors: An Example of Membrane Proteins

451

Fig.Â€4. G protein-coupled receptors internet resources.

Fig.Â€5. The G-protein coupled receptors structure-base global drug design flowchart.

molecular modeling approaches. Moreover, small molecule knowledge allowed to determine substructures or so-called privileges structures that are active in a particular receptor subtype (20). A database of GPCR ligands with the vast biological data is publicly available (Fig.Â€4). All these accumulated chemical and biological knowledge promote novel global GPCR drug design strategies (61) in which the knowledge gained from well-characterized receptors with plenty known ligands can be applied for the search of novel ligands of similar GPCRs but poor-characterized structurally and with a few known ligands (62) (Fig.Â€5). Thus, the sequence analysis of residues in the ligand binding sites of GPCRs can be done to find the closest homologous receptors to the receptor of interest, then known ligands for this receptor can be used as lead structures to generate high-potent ligands for the GPCR of interest. This approach was used to discover the first nonpeptide

452

Tikhonova and Fourmy

ligands for the somastotin receptor subtype 5 SST5 (63). A virtual screening methodology using the information about protein sequences and known nonnative ligands was developed to predict ligands for orphan GPCR receptors (64). References 1. Tikhonova IG, Best RB, Engel S, Gershengorn MC, Hummer G, Costanzi S (2008) Atomistic insights into rhodopsin activation from a dynamic model. J Am Chem Soc 130(31):10141–10149 2. Kobilka BK, Deupi X (2007) Conformational complexity of G-protein-coupled receptors. Trends Pharmacol Sci 28(8):397–406 3. Johnston CA, Siderovski DP (2007) Receptormediated activation of heterotrimeric G-proteins: current structural insights. Mol Pharmacol 72(2):219–230 4. Thompson MD, Cole DE, Jose PA (2008) Pharmacogenomics of G protein-coupled receptor signaling: insights from health and disease. Methods Mol Biol 448:77–107 5. Palczewski K, Kumasaka T, Hori T etÂ€ al (2000) Crystal structure of rhodopsin: A G protein-coupled receptor. Science 289(5480):739–745 6. Teller DC, Okada T, Behnke CA, Palczewski K, Stenkamp RE (2001) Advances in determination of a high-resolution three-dimensional structure of rhodopsin, a model of G-proteincoupled receptors (GPCRs). Biochemistry 40(26):7761–7772 7. Li J, Edwards PC, Burghammer M, Villa C, Schertler GF (2004) Structure of bovine rhodopsin in a trigonal crystal form. J Mol Biol 343(5):1409–1438 8. Okada T, Sugihara M, Bondar AN, Elstner M, Entel P, Buss V (2004) The retinal conformation and its environment in rhodopsin in light of a new 2.2Â€ A crystal structure. J Mol Biol 342(2):571–583 9. Murakami M, Kouyama T (2008) Crystal structure of squid rhodopsin. Nature 453(7193):363–367 10. Park JH, Scheerer P, Hofmann KP, Choe HW, Ernst OP (2008) Crystal structure of the ligand-free G-protein-coupled receptor opsin. Nature 454(7201):183–187 11. Warne T, Serrano-Vega MJ, Baker JG etÂ€al (2008) Structure of a beta1-adrenergic G-proteincoupled receptor. Nature 454(7203):486–491 12. Rosenbaum DM, Cherezov V, Hanson MA etÂ€ al (2007) GPCR engineering yields highresolution structural insights into beta2-adrenergic receptor function. Science 318(5854): 1266–1273

13. Cherezov V, Rosenbaum DM, Hanson MA etÂ€al (2007) High-resolution crystal structure of an engineered human beta2-adrenergic G protein-coupled receptor. Science 318(5854):1258–1265 14. Rasmussen SG, Choi HJ, Rosenbaum DM etÂ€ al (2007) Crystal structure of the human beta2 adrenergic G-protein-coupled receptor. Nature 450(7168):383–387 15. Jaakola VP, Griffith MT, Hanson MA etÂ€ al (2008) The 2.6Â€Angstrom crystal structure of a human A2A adenosine receptor bound to an antagonist. Science 322(5905):1211–1217 16. Evers A, Hessler G, Matter H, Klabunde T (2005) Virtual screening of biogenic aminebinding G-protein coupled receptors: comparative evaluation of protein- and ligand-based virtual screening protocols. J Med Chem 48(17):5448–5465 17. Tikhonova IG, Sum CS, Neumann S etÂ€ al (2007) Bidirectional, iterative approach to the structural delineation of the functional “chemoprint” in GPR40 for agonist recognition. J Med Chem 50(13):2981–2989 18. Bissantz C, Bernard P, Hibert M, Rognan D (2003) Protein-based virtual screening of chemical databases. II. Are homology models of G-Protein coupled receptors suitable targets? Proteins 50(1):5–25 19. Archer-Lahlou E, Tikhonova I, Escrieut C etÂ€al (2005) Modeled structure of a G-proteincoupled receptor: the cholecystokinin-1 receptor. J Med Chem 48(1):180–191 20. Klabunde T, Hessler G (2002) Drug design strategies for targeting G-protein-coupled receptors. Chembiochem 3(10):928–944 21. Krystek SR Jr, Kimura SR, Tebben AJ (2006) Modeling and active site refinement for G protein-coupled receptors: application to the beta-2 adrenergic receptor. J Comput Aided Mol Des 20(7–8):463–470 22. Dufresne M, Seva C, Fourmy D (2006) Cholecystokinin and gastrin receptors. Physiol Rev 86(3):805–847 23. Archer E, Maigret B, Escrieut C, Pradayrol L, Fourmy D (2003) Rhodopsin crystal: new template yielding realistic models of G-proteincoupled receptors? Trends Pharmacol Sci 24(1):36–40

The Family of G Protein-Coupled Receptors: An Example of Membrane Proteins 24. Poirot SS, Escrieut C, Dufresne M etÂ€al (1994) Photoaffinity labeling of rat pancreatic cholecystokinin type A receptor antagonist binding sites demonstrates the presence of a truncated cholecystokinin type A receptor. Mol Pharmacol 45(4):599–607 25. Kennedy K, Escrieut C, Dufresne M, Clerc P, Vaysse N, Fourmy D (1995) Identification of a region of the N-terminal of the human CCKA receptor essential for the high affinity interaction with agonist CCK. Biochem Biophys Res Commun 213(3):845–852 26. Kennedy K, Gigoux V, Escrieut C etÂ€al (1997) Identification of two amino acids of the human cholecystokinin-A receptor that interact with the N-terminal moiety of cholecystokinin. J Biol Chem 272(5):2920–2926 27. Henderson R, Baldwin JM, Ceska TA, Zemlin F, Beckmann E, Downing KH (1990) Model for the structure of bacteriorhodopsin based on high-resolution electron cryo-microscopy. J Mol Biol 213(4):899–929 28. Campagne F, Jestin R, Reversat JL, Bernassau JM, Maigret B (1999) Visualisation and integration of G protein-coupled receptor related information help the modelling: description and applications of the Viseur program. J Comput Aided Mol Des 13(6):625–643 29. Gigoux V, Maigret B, Escrieut C etÂ€al (1999) Arginine 197 of the cholecystokinin-A receptor binding site interacts with the sulfate of the peptide agonist cholecystokinin. Protein Sci 8(11):2347–2354 30. Gigoux V, Escrieut C, Silvente-Poirot S etÂ€al (1998) Met-195 of the cholecystokinin-A receptor interacts with the sulfated tyrosine of cholecystokinin and is crucial for receptor transition to high affinity state. J Biol Chem 273(23):14380–14386 31. Gigoux V, Escrieut C, Fehrentz JA etÂ€ al (1999) Arginine 336 and asparagine 333 of the human cholecystokinin-A receptor binding site interact with the penultimate aspartic acid and the C-terminal amide of cholecystokinin. J Biol Chem 274(29):20457–20464 32. Giragossian C, Mierke DF (2001) Intermolecular interactions between cholecystokinin-8 and the third extracellular loop of the cholecystokinin A receptor. Biochemistry 40(13):3804–3809 33. Giragossian C, Sugg EE, Szewczyk JR, Mierke DF (2003) Intermolecular interactions between peptidic and nonpeptidic agonists and the third extracellular loop of the cholecystokinin 1 receptor. J Med Chem 46(16):3476–3482 34. Ji Z, Hadac EM, Henne RM, Patel SA, Lybrand TP, Miller LJ (1997) Direct identification of a distinct site of interaction between

35.

36.

37.

38.

39.

40.

41.

42.

43.

44.

453

the carboxyl-terminal residue of cholecystokinin and the type A cholecystokinin receptor using photoaffinity labeling. J Biol Chem 272(39): 24393–24401 Hadac EM, Pinon DI, Ji Z etÂ€al (1998) Direct identification of a second distinct site of contact between cholecystokinin and its receptor. J Biol Chem 273(21):12988–12993 Ding XQ, Pinon DI, Furse KE, Lybrand TP, Miller LJ (2002) Refinement of the conformation of a critical region of charge–charge interaction between cholecystokinin and its receptor. Mol Pharmacol 61(5):1041–1052 Galas MC, Lignon MF, Rodriguez M etÂ€ al (1988) Structure–activity relationship studies on cholecystokinin: analogues with partial agonist activity. Am J Physiol 254(2 Pt 1): G176–G182 Archer-Lahlou E, Escrieut C, Clerc P etÂ€ al (2005) Molecular mechanism underlying partial and full agonism mediated by the human cholecystokinin-1 receptor. J Biol Chem 280(11):10664–10674 Crocker E, Eilers M, Ahuja S etÂ€ al (2006) Location of Trp265 in metarhodopsin II: implications for the activation mechanism of the visual receptor rhodopsin. J Mol Biol 357(1):163–172 Marco E, Foucaud M, Langer I, Escrieut C, Tikhonova IG, Fourmy D (2007) Mechanism of activation of a G protein-coupled receptor, the human cholecystokinin-2 receptor. J Biol Chem 282(39):28779–28790 Escrieut C, Gigoux V, Archer E etÂ€al (2002) The biologically crucial C terminus of cholecystokinin and the non-peptide agonist SR-146, 131 share a common binding site in the human CCK1 receptor. Evidence for a crucial role of Met-121 in the activation process. J Biol Chem 277(9):7546–7555 Martin-Martinez M, Marty A, Jourdan M etÂ€al (2005) Combination of molecular modeling, site-directed mutagenesis, and SAR studies to delineate the binding site of pyridopyrimidine antagonists on the human CCK1 receptor. J Med Chem 48(15):4842–4850 Silvente-Poirot S, Escrieut C, Gales C etÂ€ al (1999) Evidence for a direct interaction between the penultimate aspartic acid of cholecystokinin and histidine 207, located in the second extracellular loop of the cholecystokinin B receptor. J Biol Chem 274(33): 23191–23197 Gales C, Poirot M, Taillefer J etÂ€ al (2003) Identification of tyrosine 189 and asparagine 358 of the cholecystokinin 2 receptor in direct interaction with the crucial C-terminal amide of cholecystokinin by molecular modeling,

454

45.

46.

47.

48.

49.

50.

51.

52.

53.

54.

Tikhonova and Fourmy site-directed mutagenesis, and structure/affinity studies. Mol Pharmacol 63(5):973–982 Langer I, Tikhonova IG, Travers MA etÂ€ al (2005) Evidence that interspecies polymorphism in the human and rat cholecystokinin receptor-2 affects structure of the binding site for the endogenous agonist cholecystokinin. J Biol Chem 280(23):22198–22204 Foucaud M, Tikhonova IG, Langer I etÂ€ al (2006) Partial agonism, neutral antagonism, and inverse agonism at the human wild-type and constitutively active cholecystokinin-2 receptors. Mol Pharmacol 69(3):680–690 Rodriguez-Barrios F, Balzarini J, Gago F (2005) The molecular basis of resilience to the effect of the Lys103Asn mutation in nonnucleoside HIV-1 reverse transcriptase inhibitors studied by targeted molecular dynamics simulations. J Am Chem Soc 127(20): 7570–7578 Foucaud M, Marco E, Escrieut C, Low C, Kalindjian B, Fourmy D (2008) Linking nonpeptide ligand binding mode to activity at the human cholecystokinin-2 receptor. J Biol Chem 283(51):35860–35868 Chen JZ, Wang J, Xie XQ (2007) GPCR structure-based virtual screening approach for CB2 antagonist search. J Chem Inf Model 47(4):1626–1637 Bissantz C, Schalon C, Guba W, Stahl M (2005) Focused library design in GPCR projects on the example of 5-HT(2c) agonists: comparison of structure-based virtual screening with ligand-based search methods. Proteins 61(4):938–952 Vaidehi N, Schlyer S, Trabanino RJ etÂ€ al (2006) Predictions of CCR1 chemokine receptor structure and BX 471 antagonist binding followed by experimental validation. J Biol Chem 281(37):27613–27620 Radestock S, Weil T, Renner S (2008) Homology model-based virtual screening for GPCR ligands using docking and target-biased scoring. J Chem Inf Model 48(5):1104–1117 de Graaf C, Rognan D (2008) Selective structure-based virtual screening for full and partial agonists of the beta2 adrenergic receptor. J Med Chem 51(16):4978–4985 Evers A, Klabunde T (2005) Structure-based drug discovery using GPCR homology modeling:

55.

56.

57.

58.

59.

60.

61. 62. 63.

64.

successful virtual screening for antagonists of the alpha1A adrenergic receptor. J Med Chem 48(4):1088–1097 Cavasotto CN, Orry AJ, Murgolo NJ etÂ€ al (2008) Discovery of novel chemotypes to a G-protein-coupled receptor through ligandsteered homology modeling and structurebased virtual screening. J Med Chem 51(3):581–588 Becker OM, Marantz Y, Shacham S etÂ€ al (2004) G protein-coupled receptors: in silico drug discovery in 3D. Proc Natl Acad Sci U S A 101(31):11304–11309 Kellenberger E, Springael JY, Parmentier M, Hachet-Haas M, Galzi JL, Rognan D (2007) Identification of nonpeptide CCR5 receptor agonists by structure-based virtual screening. J Med Chem 50(6):1294–1303 Tikhonova IG, Sum CS, Neumann S etÂ€ al (2008) Discovery of novel agonists and antagonists of the free fatty acid receptor 1 (FFAR1) using virtual screening. J Med Chem 51(3):625–633 Engel S, Skoumbourdis AP, Childress J etÂ€al (2008) A virtual screen for diverse ligands: discovery of selective G protein-coupled receptor antagonists. J Am Chem Soc 130(15):5115–5123 Kratochwil NA, Malherbe P, Lindemann L etÂ€ al (2005) An automated system for the analysis of G protein-coupled receptor transmembrane binding pockets: alignment, receptor-based pharmacophores, and their application. J Chem Inf Model 45(5): 1324–1336 Rognan D (2007) Chemogenomic approaches to rational drug design. Br J Pharmacol 152(1):38–52 Klabunde T (2007) Chemogenomic approaches to drug discovery: similar receptors bind similar ligands. Br J Pharmacol 152(1):5–7 Martin RE, Green LG, Guba W, Kratochwil N, Christ A (2007) Discovery of the first nonpeptidic, small-molecule, highly selective somatostatin receptor subtype 5 antagonists: a chemogenomics approach. J Med Chem 50(25):6291–6294 Bock JR, Gough DA (2005) Virtual screen for ligands of orphan G protein-coupled receptors. J Chem Inf Model 45(5):1402–1414

Index

A

C

Absorption................................................. 5–10, 16, 32, 38, 66, 81, 93, 150, 425, 430 ADP/ATP transporter (AAC, ANT, ANCP) expression................................................................... 26 inhibitors bonkrelic acid (BA)........................26, 107, 112, 115 carboxyatractyloside (CATR)........................ 20, 21, 24–27, 107, 109, 110, 112, 115 purification..........................................20, 107, 113, 115 Alpha helices break points.............................................................. 291 helix–helix packing........................................... 292, 293 helix-turn-helix.......................... 286, 288–290, 292, 293 interactions with lipid............................................... 279 structure............................................................ 284–285 TM-turn-TM........................................................... 292 Amphiphile...................................................................... 80 Aquaporin (AQP)................... 171, 172, 174, 181–183, 410 ATP binding cassette transporter (ABCG2) expression.......................................................53, 55, 58, 59, 62, 65, 66, 68, 70, 71 inhibitors Hoechst.............................................33, 342, 59–61 vanadate.....................................................64, 67, 72 purification........................ 48–49, 62–64, 66–69, 72, 73

Calcium ATPase (Ca-ATPase) inhibitor, Thapsigargin..............................124, 131, 238 Carbon films...............................................6, 181, 189–190, 194–197, 203, 204 Cells epithelial........................................................... 172, 179 HT-29............................................................. 222–224 Sf9............................................................ 47, 48, 55–58 Channel...................................................... 84, 90, 98, 131, 171, 172, 183, 262, 272, 273, 287, 324, 343, 347, 350, 351, 355, 364, 370, 372, 373, 378, 381, 387, 400, 409, 411, 429–432, 434 Chimera software................................................... 223, 231 Circular dichroism....................................... 4, 51, 52, 64, 93 ClustalW software...................................377, 389, 392, 395 Column affinity (Ni-NTA)....................................................... 30 gel filtration........................................................ 63, 308 hydroxyapatite............................................................ 25 reverse phase............................................................. 308 size exclusion............................................................ 161 Cryoelectronmicroscopy (cryo-EM)....................... 187, 196 Crystallization................................................16, 20, 23, 24, 79–100, 107, 108, 110, 111, 115, 123, 124, 131, 144, 146–150, 153–157, 161–165, 187–204, 250, 284 Crystallography................................................79–100, 106, 110, 122, 161, 183, 187–204, 207–209, 217, 238, 239, 284, 297, 304, 412, 424

B Bacteria expression.........................6, 9, 29–31, 33, 263, 322, 337 inclusion bodies...................................... 6, 9, 31, 35–38, 322, 326–327, 337 Bacteriorhodopsin (BR).................................84–85, 87–90, 98–100, 187, 285, 293, 295, 296, 430, 444 Baculovirus/insect cell systems............................. 48, 67–71 Beta barrels.............................................210, 216, 262–264, 267–270, 272, 275, 284, 291, 293, 322, 329, 332, 336, 364, 435 Beta sheet..........................................................66, 425, 434 Bio Beads................................................ 5, 9–13, 16, 23–25, 107, 189, 192–193, 202–203, 211, 306, 309 Bioinformatics................................................................ 398

D Data collection (data set collection)........................100, 144, 146, 150, 159–160, 166, 218 Detergents cholamidopropyl dimethylammonio hydroxy propanesulfonate (CHAPSO).................. 89, 90 cholamidopropyl dimethylammonio propanesulfonate (CHAPS)....................................39, 67, 69, 203 dihexanoylphosphatidylcholine (DHPC).................. 32, 39, 89, 90, 266, 328, 338 dodecylmaltopyranoside (DDM)..................39, 62, 266

Jean-Jacques Lacapère (ed.), Membrane Protein Structure Determination: Methods and Protocols, Methods in Molecular Biology, vol. 654, DOI 10.1007/978-1-60761-762-4, © Springer Science+Business Media, LLC 2010

455

Membrane Protein Structure Determination 456â•› Index

â•›â•›

Detergents (Continuedâ•›) dodecylphosphocholine (DPC)........................5–10, 12, 13, 39, 41, 42, 310, 315, 328 fos-choline 16 (FC-16)............................................... 67 lauroylsarcosine (sarcosyl)..................................... 31, 39 laurylamido dimethylpropylaminoxide (LAPAO).................................................. 20, 25 lauryldimethylamine oxide (LDAO).................................................. 39, 266 lyso-phosphatidylglycerol (LPPG)......................................................... 266 micelle.......................................................72, 81, 90, 97, 262, 263, 265, 276, 284, 286, 288, 290, 293, 310, 312–315, 321, 322, 327, 435 octaethylene glycol monododecyl ether (C12E8).................................................. 39, 202 octylglucopyranoside (OG)................................ 85, 100 perfluoro-octanoic acid (PFO)............................. 67, 69 sodium dodecyl sulfate (SDS)............................... 5–11, 13, 16, 31, 35, 39, 41, 69 tetraethylene glycol monooctyl ether (C8E4)......................................... 324, 328 triton X-100 (TX-100)............................................... 39 zwittergent 3–14 (ZW 3–14)........................... 324, 328 Dialysis buttons.............................................................. 193, 194 membranes................................ 189, 193, 194, 306, 323 Diffraction electron............................................................... 92, 108 neutron....................................................................... 97 X-ray.........................................................................91 Docking..........................................................134, 135, 239, 364, 375, 381, 382, 399, 442, 445, 447–450 DomainFinder software..................................240, 244–246 Dynamics.................................................. 93, 105, 112, 113, 115, 255, 261–263, 271–272, 279, 294, 304, 305, 315, 316, 336, 341, 342, 344, 346, 350–352, 355–357, 364, 365, 372, 374, 399, 403–418, 423–436, 442, 444, 445, 448

E Electron microscopy cryo...................................................175, 177, 181–182, 191, 196, 207, 208, 237 crystallography................... 183, 187–204, 207–209, 217 density maps..................................................... 239–241 fitting procedure............................................... 240–244 freeze fracture............................ 173, 177, 178, 181, 184 grids...........................................189–191, 195–197, 224 missing wedge.......................................................... 218 negative staining................................178–179, 190, 208 section............................................................... 177–180 tomography....................................................... 221–234 Electrostatic interactions.................................109, 127, 128

Escherichia coli (E. Coli)....................................... 6, 9, 29, 30, 33–35, 39, 42, 48, 49, 53, 71, 91, 189, 191, 210, 263, 271, 287, 293, 305, 307, 317, 322, 324, 326, 368, 431 Expression..............................................4, 5, 29–31, 33–35, 39, 43, 48, 53–55, 58, 59, 65, 68, 70, 124, 190, 210, 263–265, 277, 305, 307–308, 310, 316, 322, 323, 337

F Fluorescence (fluorescence recovery after photobleaching, FRAP)........................... 51, 60, 66, 146, 150–153, 156, 162, 164, 412

G G coupled protein receptor (GPCR)......................... 39, 89, 99, 141–166, 279, 285–294, 296, 365, 369, 387, 389, 393, 398, 410, 429, 434, 441–443, 447–452

H His-tag proteins................................................72, 161, 163 Hydrogen bonds...............109, 129, 134, 285, 335, 336, 445 Hydrophobic clusters.............................................. 134, 136

I ImageJ software................190, 201, 223, 227–229, 231, 234 Image processing.....................................182–183, 223, 226 Immobilized metal ion affinity chromatography (IMAC).............................................. 26, 35–40 IMOD software..............................................213, 215, 218 Inflammation.......................................................... 222, 432 Ion transport, ion channel.............................................. 287 Isopropyl thio galactopyranoside (IPTG)................. 33–35, 53–55, 264, 305, 308, 310, 323–326

K Klebsiella pneumoniae (KpOmpA)........................... 321–338

L Labelling carbon (12C and 13C).......................................34, 43, 44, 264–267, 271, 276–279, 322–326, 328, 329, 332, 334, 335, 411 hydrogen (1H and 2H)..................................... 266–269, 271, 277, 278, 311, 313, 322, 324, 326–330, 334–338, 342, 343, 347–354 nitrogen (14N and 15N)....................................34, 43, 44, 265–269, 271, 273, 275–278, 305, 311, 313, 315, 322–335, 337, 342, 343, 350, 416 perdeuteration............................................266, 329–332 phosphorus (31P).......................................275, 306, 310, 311, 342, 343, 345, 347, 349 selective methyl protonation..................................... 322

seleno methionine....................................................... 91 stable isotope labelling.............................................. 267 Lactose permease.............................................107, 294, 368 Lipid analysis.............................. 149, 188–189, 191–192, 207 azolectine.............................................................. 67, 72 bicelles..................................................89–90, 143, 262, 263, 266, 274–275, 279, 293, 344, 345, 347–350, 355, 356 bilayer.................................................... 6, 11, 80, 84, 90, 146, 148, 171, 187, 188, 191, 209, 210, 212, 215–217, 274, 277, 284, 310, 312–314, 341–359, 369, 373, 374, 410, 412, 435 cholesterol.................................................147, 165, 342, 344, 346, 352, 358, 409, 433 cubic phase (LCP)..........................................81, 83–85, 87, 88, 100, 107, 143–157, 159, 161–163, 165 detergent extraction..................................... 4, 5, 107, 209, 215 removal...................................................5, 9–13, 16, 188, 189, 191, 193–194, 209–211, 309 gel and fluid phase....................... 85, 351, 352, 411, 417 hexagonal phase.........................................156, 157, 162 lamellar phase.............................................98, 100, 156, 162, 163, 347, 350, 352 liposome.........................................................5, 11, 107, 210, 342, 347–350, 352, 354 liquid-ordered phase......................................... 341, 352 membrane.................................................11, 48, 69, 71, 81, 113, 207–209, 212, 387, 433, 434 monoolein (MO)............................................81, 84–86, 98, 100, 144, 147, 151, 163, 164, 189 oriented............................. 193, 347, 350–352, 355, 377 reconstitution......................................72, 144, 146–148, 153, 188, 191, 192, 202, 203, 207, 209, 210, 266, 274, 306, 309, 313 sphingolipids.............................................341, 409, 433 sponge phase..................................................86, 87, 165 vesicle multilamellar........................ 5, 67, 72, 347, 351, 352 unilamellar.................................................. 312, 313 Liquid chromatography fast protein liquid chromatography (FPLC).................................... 30, 31, 36, 37, 44 high performance liquid chromatography (HPLC)................... 44, 124, 144, 306, 308, 309 Loop.......................................................108, 113, 133–135, 143, 158–160, 163, 183, 253, 264, 275, 285–296, 333, 367–369, 375, 379, 394, 397–400, 417, 432, 441, 442, 444, 446

M Mass spectroscopy (MALDI-TOF)..................4, 5, 30, 309

Membrane Protein Structure Determination 457 Index â•›â•›â•›â•› Medium for bacterial culture Luria Bertani (LB)...................................30, 31, 33–35, 43, 49, 53, 54, 324, 325 minimal (M9)........................ 34, 43, 236, 305, 323, 325 Membrane..............................................3–17, 19–27, 29–44, 47–73, 79–100, 105–111, 113, 115, 120–122, 124–126, 128–136, 141–166, 171–185, 187–204, 207–218, 221–234, 238, 244–247, 250, 252, 253, 261–279, 283–298, 303–317, 321–338, 341, 342, 344–346, 352–355, 363–382, 387–390, 393–395, 397–400, 403–418, 423–436, 441–452 Merging data.................................................................. 160 Metal fluoride................................................................. 123 Mitochondria inner membrane...................................19, 105, 106, 233 intermembrane space (IMS)......................107–111, 113 outer membrane........................................................ 106 proteins (mitochondrial carrier family (MCF)).............................................4, 5, 19–27, 30, 106–108, 110, 111, 113, 115, 222 ultra-structure............................................221, 224–225 Modeller software................................................... 394, 396 Modelling ab initio.............................................368, 376, 380, 426 comparative...............................................388, 393–399 homology.................................. 114, 115, 143, 364, 365, 376, 388, 389, 394, 395, 399, 406, 442, 443, 449 Molecular dynamics............................................... 112–113, 115, 255, 294, 364, 365, 372–374, 399, 403–418, 426–428, 436, 442, 444, 445, 448 Molecular modelling...............................240, 256, 363, 375 Molecular refinement..................................................... 388 Multi drug efflux pump (MexA, MexB, OprM)................................. 208 Mutagenesis (mutants).........................................21, 22, 26, 49, 122, 132, 143, 255, 267, 305, 316, 382, 410, 430, 431, 442, 445, 447, 448, 450

N Neutron diffraction................................................................... 97 diffusion.........................................................88, 90, 412 scattering..........................88, 90, 99, 100, 121, 411, 412 Nuclear magnetic resonance (NMR) chemical shift...........................................262, 267–269, 271, 272, 274, 275, 304, 314, 330, 336, 337, 347, 355, 357, 416 correlation spectroscopy (COSY )..................... 354, 357 distance restrains................................268–270, 278, 336 heteronuclear multiple-quantum coherence (HMQC).. 329, 333, 335, 337 heteronuclear single-quantum coherence (HSQC )..............................265, 266, 271, 311, 317, 328, 330, 334, 335, 337

Membrane Protein Structure Determination 458â•› Index

â•›â•›

Nuclear magnetic resonance (NMR) (Continuedâ•›) high resolution magic angle spinning (HR-MAS )...................................342, 343, 352 magic angle...............................................272, 275, 313, 342, 343, 348, 354, 416 nuclear overhauser enhancement (NOE)................... 267–270, 313, 333–335, 443 nuclear overhauser enhancement spectroscopy (NOESY )..................... 269, 322, 329–335, 337 oriented sample......................... 272, 275, 312–313, 414 paramagnetic relaxation enhancement (PRE)............................................268, 270, 322 phase modulated Lee–Goldburg scheme (PMLG )...................................................... 313 polarised inversion spin exchange at the magic angle (PISEMA)..................... 275, 276, 313–315, 416 polarity index slant angle (PISA)..................... 275, 276 quadrupolar splitting........................................344, 346, 351, 352, 355, 358, 411, 413 relaxation time...................................344–346, 355–358 residual dipolar coupling (RDC)...............270–271, 322 softwares NMRpipe................................................... 324, 328 NMRview....................................324, 328, 335, 336 TALOS...................................................... 278, 336 solid state (ssNMR)............ 89, 261–279, 304, 312–315, 343–346, 354, 358, 413, 416, 430 solution...............................................43, 261–279, 284, 287–289, 291–293, 295, 304, 306, 310–315, 317, 321–338, 345, 354 total correlation spectroscopy (TOCSY )................. 322, 329, 330, 332, 334, 355, 357 transverse relaxation-optimized spectroscopy (TROSY ) HNCA(CO)................................267, 269, 271, 330 HNCACB...................................267, 269, 329, 332 HN(CO)CACB......................................... 329, 337 two pulse phase modulated sequence (TPPM)........................................................ 313 wide line spectra............................................... 344–352 Normal modes................................................113, 237–256, 364, 369, 374–375, 427, 431

P Peptide backbone....................................................266, 275, 371 plane.................................................242, 255, 275, 276, 290, 413, 414, 416–418, 434 side chain...........................................242, 255, 266, 369 Phospholamban (PLN)...................................271, 303–317 Phospholipids dimyristoil phosphatidyl choline (DMPC).......................... 5, 11, 89, 90, 273, 416

dimyristoil phosphatidyl ethanolamine (DMPE)..................................................... 5, 11 dioleyl phosphatidyl (choline, ethanolamine) (DOPC, DOPE)..................................124, 209, 211, 273, 306, 307, 309, 310, 312, 313 egg yolk phosphatidic acid........................................ 189 egg yolk phosphatidyl choline................................... 189 egg yolk phosphatidyl ethanolamine......................... 189 1-palmitoyl-2-oleoylphosphatidyl choline (POPC)................................................ 273, 432 1-palmitoyl-2-oleoylphosphatidyl glycerol (POPG)................................................ 273, 380 phosphatidylinositol................................................. 342 tetradecanoyl phosphocholine.................................. 342 Plasmids.........................................................21–22, 30, 33, 48–50, 54, 71, 305, 307, 322 PROCHECK software.......................................... 324, 397 Protein analysis (absorption spectra)........................7, 9, 32, 425 characterisation analytical centrifugation.................................. 82, 93 light scattering.....................................11, 12, 82, 93 SDS-gels............................................................... 81 concentration determination..................................... 5–9 concentrators..................................................50, 72, 82, 149, 161, 307, 327 data bank............................. 79, 122, 263, 365, 406, 424 docking.....................................................364, 375, 381, 382, 399, 442, 444, 447–450 dynamics...................................................51, 63, 64, 93, 113, 115, 122, 135, 142, 161, 233, 240, 244, 255, 261–263, 266, 271–272, 279, 294, 304, 305, 315, 336, 344, 363–365, 372–374, 376, 399, 403–418, 423–436, 442, 444, 445, 447, 448 fragments..................................................21, 22, 31, 83, 98, 144, 283–298, 368, 380, 399, 400, 445 over expression....................................................4, 6, 29, 48, 53–59, 70, 71, 173, 263–264, 322, 324–326 polymers................................................42, 43, 221, 427 recombinant..................................................5, 6, 19–27, 29, 30, 33–35, 43, 306, 309, 324 refolding.............................. 39, 317, 322–324, 326, 327 sequence alignment...........................................366, 367, 375, 377, 388, 391–393, 395, 398, 399 Psi-blast software.............................................389, 392, 399 P-type ATPases...............................................120, 123, 238 Purification............................................... 4–7, 9, 11, 16, 20, 22–25, 29–32, 35–42, 44, 48, 50, 62–64, 72, 73, 80, 81, 83, 92, 107, 108, 113, 120, 124, 161, 176, 210, 221, 264, 274, 305–310, 316, 323–327 PyMol software......................................................240, 250, 324, 336, 365, 377, 378, 389, 397, 406

R Receptors.................................................... 5, 30, 39, 89, 99, 141–144, 147, 156, 163, 273, 285, 287–290, 292–294, 297, 365–368, 370, 375, 381, 382, 387, 410, 429–430, 433, 441–451 Resin embedded material............................................... 233 Rhodopsin.................................98, 142, 144, 286, 287, 289, 290, 294, 296, 297, 365–367, 387, 395, 396, 400, 429, 430, 433, 434, 442, 444, 447, 448

S Saccharomyces cerevisiae.....................................................287 Sarcoplasmic reticulum...........................................120, 121, 123, 124, 130, 239, 240, 303, 304, 310, 447 Screening....................................................26, 93, 143–146, 148–153, 156, 157, 161, 162, 165, 190, 198–202, 204, 265, 279, 322–324, 327–328, 375, 449, 450, 452 Secondary structure..............................................64–66, 72, 93, 262, 267, 269, 275, 285–298, 366, 370, 379, 388, 393, 394, 434 Signal transduction.........................................142, 143, 161, 344, 404, 429, 433 Single particle analysis............................................ 207–218 Solubilization.................................................25, 31, 35–36, 41, 44, 50, 62–64, 67–69, 72, 81, 82, 92, 106, 124, 192, 202, 210, 326 Spider software............................................................... 182 Sterols............................................................................. 344 Structure evaluation......................................................... 410, 412 function relationship........................................... 26, 450 prediction............................................ 65, 286, 328, 363, 366, 368, 369, 375, 388, 390, 393–395, 400, 436 protein data bank................. 79, 122, 263, 365, 406, 424

Membrane Protein Structure Determination 459 Index â•›â•›â•›â•› Transmembrane beta barrel................................................................. 435 domain................................................ 48, 126, 216, 266, 274, 279, 280, 287, 288, 292, 293, 295–297, 311, 312, 314, 315, 322, 324, 365, 366, 368–369, 394, 395 fragment...................................................109, 113, 244, 253, 275, 285–287, 289–294, 298, 388, 406, 417, 444, 445 helix........................................... 284–292, 295, 298, 445 TSPO expression................................................30, 33–35, 265 ligands benzodiazepine................................................. 5, 30 PK 11195.................................................... 6, 13–16 purification.........................5, 7, 9, 16, 29–32, 35–42, 44 Two-dimensional crystals sheet (mono, multilayered)............................... 111, 423 tubular....................................................................... 188

V Vapour diffusion hanging................................................................. 83, 87 sitting drops.......................................................... 83, 87 Vectors pET15.................................................................. 30, 33 pET21b...........................................................49, 53, 54 pET21c............................................................. 322, 323 pMalc2E........................................................... 305, 307 Vibrational dynamics...................................................... 122 Voltage dependant anionic channel (VDAC)............................ 26, 90, 106, 269, 270

W Water channel................................................................ 172

T

X

Tertiary structure............................. 107, 267, 283–298, 313 Three-dimensional crystals............................................... 83 Three-dimensional reconstruction.................................. 233 Tomograms algorithm.......................................................... 218, 230 backprojection...................................209, 218, 229–230 weighted back projection...................209, 218, 228, 230 TomoJ software...............................................223, 226–231 Topology.........................................................128, 265, 279, 304, 314, 345, 367, 368, 373, 376, 397 Transfection bacteria....................................................................... 33 cells..................................................................33, 56, 57

Xmipp software...................................................... 214, 216 X-ray crystallography..............................................80, 83, 106, 110, 122, 183, 188, 201, 202, 238, 279, 284, 297, 304, 412, 424 diffraction......................................................91, 98, 122 scattering................................. 86, 88, 90, 100, 411, 412

Y Yeast expression......................................................26, 48, 124 strain.................................................... 21, 22, 26, 48, 49