Protein Modules and Protein-Protein Interactions, Volume 61 (Advances in Protein Chemistry)

ADVANCES IN PROTEIN CHEMISTRY Volume 61 Protein Modules and Protein±Protein Interaction This Page Intentionally Left ...

Author: Joel Janin

54 downloads 1268 Views 7MB 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

ADVANCES IN PROTEIN CHEMISTRY Volume 61 Protein Modules and Protein±Protein Interaction

This Page Intentionally Left Blank

ADVANCES IN PROTEIN CHEMISTRY EDITED BY FREDERIC M. RICHARDS

DAVID S. EISENBERG

Department of Molecular Biophysics and Biochemistry Yale University New Haven, Connecticut

Department of Chemistry and Biochemistry University of California, Los Angeles Los Angeles, California

JOHN KURIYAN Department of Molecular and Cellular Biology University of California, Berkeley Berkeley, California

VOLUME 61

Protein Modules and Protein±Protein Interaction EDITED BY JOEÈL JANIN Laboratoire d'Enzymologie et Biochimie Structurales C.N.R.S., Gif-sur-Yvette, France

SHOSHANA J. WODAK Unite de Conformation de MacromoleÂcules Biologique UniversiteÂ Libre de Bruxelles, Brussels, Belgium

Amsterdam Boston London New York Oxford Paris San Diego San Francisco Singapore Sydney Tokyo

This book is printed on acid-free paper.

⬁

Copyright 2003, Elsevier Science (USA). All rights reserved. No part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording, or any information storage and retrieval system, without permission in writing from the Publisher. The appearance of the code at the bottom of the ®rst page of a chapter indicates the Publisher's consent that copies of the chapter may be made for personal or internal use of speci®c clients. This consent is given on the condition, however, that the copier pay the stated per copy fee through the Copyright Clearance Center, Inc. (www.copyright.com) for copying beyond that permitted by Sections 107 or 108 of the U.S. Copyright Law. This consent does not extend to other kinds of copying, such as copying for general distribution, for advertising or promotional purposes, for creating new collective works, or for resale. Copy fees for pre-2003 chapters are as shown on the title pages. If no fee code appears on the title page, the copy fee is the same as for current chapters. 0065-3233/03 $35.00 Explicit permission from Academic Press is not required to reproduce a maximum of two ®gures or tables from an Academic Press chapter in another scienti®c or research publication provided that the material has not been credited to another source and that full credit to the Academic Press chapter is given. Academic Press An imprint of Elsevier Science 525 B Street, Suite 1900, San Diego, California 92101-4495, USA http://www.academicpress.com Academic Press 84 Theobald's Road, London WC1X 8RR, UK http://www.academicpress.com International Standard Serial Number: 0065-3233 International Standard Book Number: 0-12-034261-8 PRINTED IN THE UNITED STATES OF AMERICA 02 03 04 05 06 07 9 8 7 6 5 4 3 2 1

CONTENTS

Introduction JOEÈL JANIN AND SHOSHANA J. WODAK I. Protein Modules and Protein±Protein Interaction: Toward a Global View . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Web Sites . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1 5 6

Structural Basis of Macromolecular Recognition SHOSHANA J. WODAK AND JOEÈL JANIN I. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . II. Geometric and Chemical Features of Macromolecular Recognition. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . III. Classi®cation of Protein±Protein and Protein±DNA Complexes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . IV. Energetics of Macromolecular Recognition. . . . . . . . . . . . . . . . . . . . . . . V. Computational Approaches for Predicting and Simulating Protein±Protein Interaction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . VI. Concluding Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Web Sites . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

9 11 32 40 50 64 65 66

Sequence Analysis of Multidomain Proteins: Past Perspectives and Future Directions RICHARD R. COPLEY, CHRIS P. PONTING, JOÈRG SCHULTZ, AND PEER BORK I. II. III. IV.

Identi®cation of Novel Protein Domain Families . . . . . . . . . . . . . . . . . Methods for Classifying Protein Domain Families . . . . . . . . . . . . . . . . From Domain Classi®cation to Domain Context . . . . . . . . . . . . . . . . . . Genome-Wide Analysis: New Quality in Domain Research . . . . . . . . v

75 81 85 89

vi

CONTENTS

V. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

96 96

Identi®cation of Transiently Interacting Proteins and of Stable Protein Complexes BERTRAND SEÂRAPHIN I. II. III. IV. V. VI.

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Genetic Approaches . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Biological Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Biochemical Methods. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Validation of Interactions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Characterization of Interacting Subunits and Dissection of Interaction Domains . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

99 100 103 105 113 114 115

Molecular Recognition in Antibody±Antigen Complexes ERIC J. SUNDBERG AND ROY A. MARIUZZA I. II. III. IV. V.

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Structure of Antibody±Antigen Interfaces. . . . . . . . . . . . . . . . . . . . . . Antibody Cross-Reactivity and Molecular Mimicry . . . . . . . . . . . . . Thermodynamic Mapping of Antigen±Antibody Interfaces . . . . . Dissection of Binding Energetics in Antigen±Antibody Interfaces Using Double-Mutant Cycles . . . . . . . . . . . . . . . . . . . . . . . VI. Accommodation of Mutations in Antigen±Antibody Interfaces . . VII. Functional Roles for Protein Plasticity in Antigen Recognition . . VIII. Conclusions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

119 121 125 132 136 144 148 156 157

Molecular Recognition by SH2 Domains J. MICHAEL BRADSHAW AND GABRIEL WAKSMAN I. II. III. IV.

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Single SH2 Domain Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Functional Analysis of Single SH2 Domain Binding . . . . . . . . . . . . Structure and Function of SH2 Domains in the Context of Other Protein Modules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

161 164 172 184

CONTENTS

V. Unusual SH2 Domain Interactions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . VI. SH2 Domains as Drug Targets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . VII. Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

vii 197 202 204 205

How SH3 Domains Recognize Proline ANDREA MUSACCHIO I. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . II. Proline and Polyproline Type II Helices . . . . . . . . . . . . . . . . . . . . . . . . III. The SH3 Domain: A Model System to Understand Interactions Mediated by Proline-Rich PPH Helices . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

211 212 215 255

Structural Biology of eIF4F: mRNA Recognition and Preparation in Eukaryotic Translation Initiation JOSEPH MARCOTRIGIANO AND STEPHEN K. BURLEY I. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . II. Structural Features of Eukaryotic mRNAs. . . . . . . . . . . . . . . . . . . . . . . III. General Mechanisms of Cellular, Cap-Dependent Translation Initiation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . IV. HEAT Repeats within eIF4G Direct Assembly of Translation Machinery . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . V. Preparation of the mRNA 5'-UTR for Small Ribosomal Subunit Binding. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . VI. Conclusion and Perspectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

AUTHOR INDEX . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . SUBJECT INDEX . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

269 271 274 283 288 293 294

299 327

This Page Intentionally Left Blank

INTRODUCTION È L JANIN* AND SHOSHANA J. WODAK² BY JOE * Laboratoire d'Enzymologie et Biochimie Structurales, CNRS UPR9063, 91198 Gif-sur-Yvette, France, and ² Unite de Conformation de MacromoleÂcules Biologique, UniversiteÂ Libre de Bruxelles CP 160/16, 1050 Brussels, Belgium

I. Protein Modules and Protein±Protein Interaction: Toward a Global View . . . . . . Web Sites. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1 5 6

I. PROTEIN MODULES AND PROTEIN±PROTEIN INTERACTION: TOWARD A GLOBAL VIEW As this volume was being assembled, the completion of the DNA sequence of the human genome was reported by the International Human Genome Sequencing Consortium (2001) and by Venter et al. (2001). This is a major hallmark of biology, which brings to the forefront the vast new potential and momentous challenges that the knowledge of whole genome sequences is offering. A hundred or so complete genomes, of species ranging from bacteria to human, have now been sequenced, and many more are in the pipeline (NCBI Genomic Biology Web site). This ¯ow of information is changing the way in which research in all ®elds of biology is performed. Until recently most biochemists and molecular biologists focused on the properties of single genes and proteins involved in individual biological processes. Now it becomes possible to study how individual genes and gene products cooperate to build up complex cellular structures and to perform all the elaborate processes that enable cells and organisms to live and reproduce themselves. Before this vast new potential can be exploited, however, the genome sequence information must be decoded in terms of biological function. This endeavor, termed functional genomics, is the new challenge that immediately follows that of sequencing the human genome. It deals with gene products rather than genes, and because nearly all the gene products are proteins, the center of interest is shifting from DNA to RNA to proteins. After a major effort in DNA sequencing, we are hence witnessing a major effort in proteomics, the analysis of the proteome, de®ned as the complement of proteins in a cell. Proteomics encompasses many aspects, ranging from the identi®cation and quanti®cation of proteins at the cellular level to the analysis of their functional and physical interactions, as well as their molecular structure and function (Fields, 2001). 1 ADVANCES IN PROTEIN CHEMISTRY, Vol. 61

Copyright 2003, Elsevier Science (USA). All rights reserved. 0065-3233/03 $35.00

2

È L JANIN AND SHOSHANA J. WODAK JOE

One key result of the genome sequencing work is the relatively small number of genes needed to build a living organism. An organism can live a parasitic life with only a few hundred genes (Mycoplasma). It can perform elaborate chemistry and physiology with a few thousands (Escherichia coli Bacillus subtilis, Saccharomyces cerevisiae, and many other unicellular species) and is able to develop into a complete plant or animal with a few tens of thousands (Arabidopsis thaliana, Caenorhabditis elegans, or Drosophila melanogaster). In the human genome, no more than 30,000± 40,000 protein-coding genes can be identi®ed at present, a mere onethird more than the number of genes in the worm C. elegans (C. elegans Sequencing Consortium, 1998). This clearly suggests that the complexity of living organisms is not simply correlated to the number of protein-coding genes. We know of mechanisms that increase the number of chemically distinct proteins beyond the number of genes. Alternative splicing of messenger RNA is one; posttranslational modi®cations of the polypeptide chains may be another. But the major source of complexity in an organism must be combinatorial, rather than proportional to the complexity of its genome. Combinatorial complexity is an expected consequence of the implication of many gene products in each physiological process, with elaborate regulations at both the gene and protein levels, which remain largely unexplored. At the molecular level, complexity in the living cell derives ®rst from the many different interactions that proteins can undergo. The immense variety of biological functions performed by proteins essentially depends on the molecular interactions that they make and on the cellular context in which they ®nd themselves. Each protein is designed to speci®cally bind one or several small molecules, nucleic acids, and other proteins. Thus, the analysis of protein interactions, a major theme of this volume, is becoming a main focus of attention in the postgenomic era. Efforts in this area include the systematic identi®cation of all the constituents of large protein complexes (Neubauer et al., 1998; Peltier et al., 2000; Rout et al., 2000; Gaven et al., 2002; Ho et al., 2002) and attempts to map by genetic assays the complete repertoire of protein±protein interactions that occur in a cell. The already classical yeast two-hybrid method (Fields and Song, 1989) is being complemented by others, the split ubiquitin system (Stagljar et al., 1998), for instance. The biochemical approach combines established experimental techniques, such as chromatography and gel electrophoresis for preparing complexes, with ultrasensitive methods of mass spectrometry to identify their components (Link et al., 1999; Washburn et al., 2001). These techniques, discussed in the chapter by Seraphin in this volume, are being combined in new ways, scaled up, and overhauled to face the challenge of identifying components of large assemblies available in trace amounts.

INTRODUCTION

3

Large sets of data are being generated by these genetic and biochemical methods and specialized databases are being developed to manage them. These include the Database of Interacting Proteins developed by Eisenberg and colleagues in the United States (Xenarios et al., 2000) and the Biomolecular Interaction Network Database developed by Hogue and collaborators in Canada (see Bader et al., 2001). For instance, systematic screens of interacting proteins in the yeast S. cerevisiae performed by different authors have produced networks containing thousands of interacting pairs (Schwikowski et al., 2000; Uetz et al., 2000; Ito et al., 2001). The same approach is being pursued in other organisms (Walhout et al., 2000; Rain et al., 2001). In yeast, there is very little overlap between the networks of interactions derived by different authors, and one likely reason is that the number of interactions that actually exist in the cell is so large that each experiment detects only a small fraction of them (Hazbun and Fields, 2001). The challenge is thus once again in the interpretation of the data. Are all the detected interactions biologically and functionally meaningful? And, if not, how can we ef®ciently single out the meaningful ones from the others? Are we missing key interactions that go undetected by these approaches? Answering these questions requires some understanding of the basic physical principles that govern protein±protein recognition and macromolecular recognition in general at the molecular and atomic level. Many systems have already been studied in great detail. Of particular interest is the recognition of a protein antigen by an antibody. This binary interaction is analyzed in the chapter by Sundberg and Mariuzza in this volume. The antibody moiety of antigen±antibody recognition always involves the same protein component made of a pair of variable immunoglobulin domains. Signal transduction in eukaryotic cells also relies on binary protein±protein recognition, but it involves a cascade of recognition steps and many different types of proteins: receptors, enzymes, adaptors, transcription factors, etc. The chapter on Sarc-homology 2 (SH2) domains by Bradshaw and Waksman and the chapter on polyproline recognition by Musacchio describe some of the steps in signal transduction and the proteins that perform these steps. In antigen±antibody recognition and in signal transduction, a large body of data has been derived from the structural, biochemical, and theoretical analyses of protein±protein complexes. In other processes, the information may be equally rich, but far less complete. Gene transcription in eukaryotic cells is an example. It involves several levels of regulation and is performed by large macromolecular assemblies, which we are just barely starting to understand. This is a very active ®eld of study, illustrated in this volume in the chapter by Marcotrigiano and Burley on initiation factor eIF4F. In all these systems, the knowledge of the three-dimensional structure

4

È L JANIN AND SHOSHANA J. WODAK JOE

derived from X-ray crystallography is the starting point for biochemical and physical chemical studies, and, thanks to these studies, recognition can be said to be understood at the atomic level. There are many other cellular processes for which this information is not yet available. Nevertheless, biologists at large should be aware of this work and use it to correctly extrapolate from the in vitro observations to the in vivo properties and to design better experiments for systematically analyzing interactions in a cell. Protein modules, another focus of this volume, are involved in all of the above-mentioned interactions and processes. The word ``module'' is a general designation for recurrent protein fragments that have a distinct structure, function, and/or evolutionary history. Examples are the immunoglobulin domains or the SH2 domains. The chapter by Copley and collaborators in this volume reviews the fruitful interplay of studies in protein sequence, three-dimensional structure, and function, which leads to the description of proteins as modular entities. Structural domains were initially de®ned as segments of the polypeptide chain that fold into globular units, but they may also carry out specialized molecular functions (Wetlaufer, 1973; Rose, 1979; Janin and Wodak, 1983). When the same or a closely similar domain is found in many different proteins, it becomes a module. Whereas the three-dimensional structure is conserved in these different contexts, the aminoacid sequence, and occasionally the function, may differ substantially. Recent analyses of known genomes con®rm that organisms as diverse as bacteria and humans share many proteins and protein domains and give strong support to the view the that total number of gene/protein modules is small. Before any genome sequence was completed, Chothia estimated the total number of different folds that protein domains can adopt to be no more than 1000 (Chothia, 1992). More recent estimates are in the range 1000±6000 (Orengo et al., 1994; Brenner et al., 1997). Many of these domains are highly recurrent and may therefore be called modules. Modules can exist as such, forming a small single-domain protein. Most of the time, however, they are part of larger polypeptide chains, assembled by successive events of gene fusion. Combining a speci®c set of modules within a single polypeptide chain ensures that they are expressed together and localized in the same cells or cellular compartments (Tsoka and Ouzounis, 2000). Alternatively, the modules may participate in the same cellular process and sometimes interact physically, forming speci®c protein±protein complexes, without being linked covalently. Both the fused and the separated arrangements exist in nature. While the former is observed in some organisms, the latter is likely to be used in others. This is the basis of a method for detecting protein±protein inter-

INTRODUCTION

5

actions proposed by Marcotte and collaborators at UCLA (Marcotte et al., 1999a, b). Ouzounis and collaborators at EBI-EMBL (Hinxton, UK) have performed a similar analysis (Enright et al., 1999). Although it is at present not clear what fraction of the interactions detected by these approaches represent actual physical interactions between the protein modules, there is mounting evidence that these methods detect their involvement in common functional processes. Whether the genes are fused or separate, there is a very close interplay between protein modules and their interactions at the functional or physical level. Our understanding of evolution at the molecular and cellular levels and our ability to understand, modify, and one day simulate cellular function will crucially depend on our knowledge of the rules of this game. This is prompting attempts to capture some of the global patterns of the protein±protein interaction networks determined from the twohybrid screens (Schwikowski et al., 2000; Jeong et al., 2001) to map data from these experiments onto the network of cellular processes or to link information on the function and evolution of individual protein modules with that on the physical interactions between these modules (Park et al., 2001). Figure 1 illustrates graphically one such attempt. Once the networks are established, atomic structures and thermodynamic parameters for the interacting protein pairs will still be required to validate data produced by genome±scale experiments and to give a molecular basis to the networks. Protein±protein interaction is becoming a major theme of the postgenome era, and its study constitutes an essential component of functional genomics. We hope that this volume will help bring biologists and physical chemists together. Their respective approaches are complementary and should thus never be more productive than when they are combined. WEB SITES Biomolecular Interaction Network Database (Toronto, Ontario, Canada), http:// www.bind.ca. Database of Interacting Proteins (UCLA, CA), http://www.doe-mbi.ucla.edu. Macromolecular Structure Database (EBI-EMBL, Hinxton, UK), http://pdb-browsers.ebi. ac.uk. NCBI Genomic Biology (National Center for Biotechnology Information, Bethesda, MD), http://www.ncbi.nlm.nih.gov/Genomes. Protein Data Bank (Research Collaboratory for Structural Bioinformatics, Rutgers University, New Brunswick, NJ), http:// www.rcsb.org/pdb. Structural Classi®cation of Proteins (MRC-LMB, Cambridge, UK), http://scop.mrc-lmb. cam.ac.uk/scop/.

6

È L JANIN AND SHOSHANA J. WODAK JOE

FIG. 1. Global network of protein±protein interaction in yeast built from data on two-hybrid screens [reproduced with permission from Jeong et al. (2001) ]. Nodes are colored according to the phenotypic effect of removing the corresponding protein from the organism (red, lethal; green, nonlethal; orange, slow growth; yellow, unknown).

REFERENCES Bader, G. D., Donaldson, I., Wolting, C., Ouellette, B. F., Pawson, T., and Hogue, C. W. (2001). BINDÐThe Biomolecular Interaction Network Database. Nucleic Acids Res. 29(1), 242±245. Brenner, S. E., Chothia, C., and Hubbard, T. J. (1997). Population statistics of protein structures: Lessons from structural classi®cations. Curr. Opin. Struct. Biol. 7(3), 369±376. C. elegans Sequencing Consortium (1988). Genome sequence of the nematode C. elegans: A platform for investigating biology. Science 282(5396), 2012±2018. Chothia, C. (1992). Proteins: One thousand families for the molecular biologist. Nature 357(6379), 543±544. Enright, A. J., Iliopoulos, I., Kyrpides, N. C., and Ouzounis, C. A. (1999). Protein interaction maps for complete genomes based on gene fusion events. Nature 402(6757), 86±90. Fields, S. (2001). Proteomics. Proteomics in genomeland. Science 291(5507), 1221±1224.

INTRODUCTION

7

Fields, S., and Song, O. (1989). A novel genetic system to detect protein±protein interactions. Nature 340(6230), 245±624b. Gavin, A. C., Bosche, M., Krause, R., Grandi, P., Marzioch, M., Bauer, A., Schultz, J., Rick, J. M., Michon, A. M., Cruciat, C. M., Remor, M., Hofert, C., Schelder, M., Brajenovic, M., Ruffner, H., Merino, A., Klein, K., Hudak, M., Dickson, D., Rudi, T., Gnau, V., Bauch, A., Bastuck, S., Huhse, B., Leutwein, C., Heurtier, M. A., Copley, R. R., Edelmann, A., Querfurth, E., Rybin, V., Drewes, G., Raida, M., Bouwmeester, T., Bork, P., Seraphin, B., Kuster, B., Neubauer, G., Superti-Furga, and G. (2002). Functional organization of the yeast proteome by systematic analysis of protein complexes. Nature 415, 141±147. Hazbun, T. R., and Fields, S. (2001). Networking proteins in yeast. Proc. Natl. Acad. Sci. USA 98(8), 4277±4278. Ho, Y., Gruhler, A., Heilbut, A., Bader, G. D., Moore, L., Adams, S. L., Millar, A., Taylor, P., Bennett, K., Boutilier, K., Yang, L., Wolting, C., Donaldson, I., Schandorff, S., Shewnarane, J., Vo, M., Taggart, J., Goudreault, M., Muskat, B., Alfarano, C., Dewar, D., Lin, Z., Michalickova, K., Willems, A. R., Sassi, H., Nielsen, P. A., Rasmussen, K. J., Andersen, J. R., Johansen, L. E., Hansen, L. H., Jespersen, H., Podtelejnikov, A., Nielsen, E., Crawford, J., Poulsen, V., Sorensen, B. D., Matthiesen, J., Hendrickson, R. C., Gleeson, F., Pawson, T., Moran, M. F., Durocher, D., Mann, M., Hogue, C. W., Figeys, D., and Tyers, M. (2002). Systematic identi®cation of protein complexes in Saccharomyces cerevisiae by mass spectrometry. Nature 415, 180±183. International Human Genome Sequencing Consortium (2001). Initial sequencing and analysis of the human genome. Nature 409(6822), 860±921. Ito, T., Chiba, T., Ozawa, R., Yoshida, M., Hattori, M., and Sakaki, Y. (2001). A comprehensive two-hybrid analysis to explore the yeast protein interactome. Proc. Natl. Acad. Sci. USA 98(8), 4569±4574. Janin, J., and Wodak, S. J. (1983). Structural domains in proteins and their role in the dynamics of protein function. Prog. Biophys. Mol. Biol. 42(1), 21±78. Jeong, H., Mason, S. P., Barabasi, A. L., and Oltvai, Z. N. (2001). Lethality and centrality in protein networks. Nature 411(6833), 41±42. Link, A. J., Eng, J., Schieltz, D. M., Carmack, E., Mize, G. J., Morris, D. R., Garvik, B. M., and Yates, J. R. (1999). Direct analysis of protein complexes using mass spectrometry. Nat. Biotechnol. 17(7), 676±682. Marcotte, E. M., Pellegrini, M., Ng, H. L., Rice, D. W., Yeates, T. O., and Eisenberg, D. (1999a). Detecting protein function and protein±protein interactions from genome sequences. Science 285(5428), 751±753. Marcotte, E. M., Pellegrini, M., Thompson, M. J., Yeates, T. O., and Eisenberg, D. (1999b). A combined algorithm for genome-wide prediction of protein function. Nature 402(6757), 83±86. Neubauer, G., King, A., Rappsilber, J., Calvio, C., Watson, M., Ajuh, P., Sleeman, J., Lamond, A., and Mann, M. (1998). Mass spectrometry and EST-database searching allows characterization of the multi-protein spliceosome complex. Nat. Genet. 20(1), 46±50. Orengo, C. A., Jones, D. T., and Thornton, J. M. (1994). Protein superfamilies and domain superfolds. Nature 372(6507), 631±634. Park, J., Lappe, M., and Teichmann, S. A. (2001). Mapping protein family Interactions: intramolecular and intermolecular protein family interaction repertoires in the PDB and yeast. J. Mol. Biol. 307(3), 929±938. Peltier, J. B., Friso, G., Kalume, D. E., Roepstorff, P., Nilsson, F., Adamska, I., and van Wijk, K. J. (2000). Proteomics of the chloroplast: Systematic identi®cation and targeting analysis of lumenal and peripheral thylakoid proteins. Plant Cell 12(3), 319±341.

8

È L JANIN AND SHOSHANA J. WODAK JOE

Rain, J. C., Selig, L., De Reuse, H., Battaglia, V., Reverdy, C., Simon, S., Lenzen, G., Petel, F., Wojcik, J., and Schachter, V. (2001). The protein±protein interaction map of Helicobacter pylori. Nature 409(6817), 211±215. Rose, G. D. (1979). Hierarchic organization of domains in globular proteins. J. Mol. Biol. 134(3), 447±470. Rout, M. P., Aitchison, J. D., Suprapto, A., Hjertaas, K., Zhao, Y., and Chait, B. T. (2000). The yeast nuclear pore complex: Composition, architecture, and transport mechanism. J. Cell Biol. 148(4), 635±651. Schwikowski, B., Uetz, P., and Fields, S. (2000). A network of protein±protein interactions in yeast. Nat. Biotechnol. 18(12), 1257±1261. Stagljar, I., Korostensky, C., Johnsson, N., and te Heesen, S. (1998). A genetic system based on split-ubiquitin for the analysis of interactions between membrane proteins in vivo. Proc. Natl. Acad. Sci. USA 95(9), 5187±5192. Tsoka, S., and Ouzounis, C. A. (2000). Prediction of protein interactions: Metabolic enzymes are frequently involved in gene fusion. Nat. Genet. 26(2), 141±142. Uetz, P., Giot, L., Cagney, G., Mans®eld, T. A., Judson, R. S., Knight, J. R., Lockshon, D., Narayan, V., Srinivasan, M., and Pochart, P. (2000). A comprehensive analysis of protein± protein interactions in Saccharomyces cerevisiae. Nature 403(6770), 623±627. Venter, J. C., Adams, M. D., Myers, E. W., Li, P. W., Mural, R. J., Sutton, G. G., Smith, H. O., Yandell, M., Evans, C. A., and Holt, R. A. (2001). The sequence of the human genome. Science 291(5507), 1304±1351. Walhout, A. J., Sordella, R., Lu, X., Hartley, J. L., Temple, G. F., Brasch, M. A., ThierryMieg, N., and Vidal, M. (2000). Protein interaction mapping in C. elegans using proteins involved in vulval development. Science 287(5450), 116±122. Washburn, M. P., Wolters, D., and Yates, J. R. (2001). Large-scale analysis of the yeast proteome by multidimensional protein identi®cation technology. Nat. Biotechnol. 19(3), 242±247. Wetlaufer, D. B. (1973). ``Nucleation, rapid folding, and globular intrachain regions in proteins.'' Proc. Natl. Acad. Sci. USA 70(3), 697±701. Xenarios, I., Rice, D. W., Salwinski, L., Baron, M. K., Marcotte, E. M., and Eisenberg, D. (2000). DIP: The database of interacting proteins. Nucleic Acids Res. 28(1), 289±591.

STRUCTURAL BASIS OF MACROMOLECULAR RECOGNITION È L JANIN² BY SHOSHANA J. WODAK* AND JOE *Unite de Conformation de MacromoleÂcules Biologique, UniversiteÂ Libre de Bruxelles CP 160/16, 1050 Brussels, Belgium, and ² Laboratoire d'Enzymologie et Biochimie Structurales, CNRS UPR9063, 91198 Gif-sur-Yvette, France

I. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . II. Geometric and Chemical Features of Macromolecular Recognition. . . . . . . . . . . A. Interface Area . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B. Geometric Complementarity and Atomic Packing . . . . . . . . . . . . . . . . . . . . . . C. Chemical Composition and Polar Interactions . . . . . . . . . . . . . . . . . . . . . . . . . D. Wet and Dry Interfaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . E. Conformational Changes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . III. Classi®cation of Protein±Protein and Protein±DNA Complexes . . . . . . . . . . . . . . A. Standard Size Protein±Protein Interfaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B. Small Interfaces and Low-Af®nity Complexes . . . . . . . . . . . . . . . . . . . . . . . . . C. Large Interfaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . D. Classi®cation of Protein±DNA Complexes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . IV. Energetics of Macromolecular Recognition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A. Protein±Protein Recognition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B. Protein±DNA Recognition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . C. Statistical Mechanics of Speci®city . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . V. Computational Approaches for Predicting and Simulating Protein±Protein Interaction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A. Docking Procedures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B. Simulations of Protein±Protein Association . . . . . . . . . . . . . . . . . . . . . . . . . . . . VI. Concluding Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Web Sites. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

9 11 11 16 21 25 29 32 32 35 36 38 40 40 46 47 50 50 59 64 65 66

I. INTRODUCTION The function of nearly all proteins is mediated by their interaction with partners that are small molecules or, very often, other biological macromolecules: proteins, DNA, and RNA. Protein±protein and protein± nucleic acid interactions are ubiquitous and essential to all known cellular and physiological processes. At a time when the decoding of whole genomes provides the molecular biologist and the biochemist with easy access to the complete set of the individual proteins that can be present in a given organism, the challenge now lies in understanding molecular assemblies and integrated systems. The analysis of pairwise interactions is a ®rst step in that direction. Several recent reviews and books (Phizicky and Fields, 1995; Schwehm and Stites, 1998; Srere, 1999) are devoted to protein±protein recognition. They cover techniques, from computer 9 ADVANCES IN PROTEIN CHEMISTRY, Vol. 61

Copyright 2003, Elsevier Science (USA). All rights reserved. 0065-3233/03 $35.00

10

È L JANIN SHOSHANA J. WODAK AND JOE

science to genetics to physical chemistry, that have been developed to detect and characterize pairwise interactions both in vivo and in vitro. A number of specialized databases, for instance, the Database of Interacting Proteins (see Web Sites) (Xenarios et al., 2000), make the results of these experiments accessible on the Web. Here we consider the problem from a structural point of view. X-Ray crystallography has determined the atomic structure of many protein± protein complexes and protein±DNA complexes. This has helped illustrate the role of macromolecular recognition in processes such as the regulation of enzyme activity, gene expression, signal transduction, and the immune response. It has also provided a wealth of chemical and geometrical data made easily accessible through the Protein Data Bank and Web servers such as the Protein±Protein Interaction Server and the Protein±Nucleic Acid Interaction Server of the Thornton laboratory at University College, London, United Kingdom, or the Surface Properties of Interfaces Database of the Honig laboratory at Columbia University, New York. Chapters of this book deal with individual systems. They provide many examples of biochemical experimentsÐand biological conclusionsÐ based on the accurate description by X-ray crystallography of the molecular surfaces which mediate recognition. In contrast, this chapter focuses on surveys of protein±protein recognition ( Jones and Thornton, 1996; Lo Conte et al., 1999) and protein± DNA recognition ( Jones et al., 1999; Nadassy et al., 1999) in many systems. The aim of these surveys is to identify features which are common to most, if not all, of the analyzed examples. Some of these features are even common to both recognition processes, which suggests that they characterize the stability of macromolecular association in general. Therefore, the chemical and geometrical properties of protein± protein and protein±DNA interfaces may be expected to correlate with the thermodynamic and kinetic parameters, which characterize the same systems in solution. Extensive sets of experimental data are available for several protein±protein complexes, both for wild-type proteins and for designed mutants [(Bogan and Thorn, 1998; Thorn and Bogan, 2001) and their ASEdb database of alanine-scanning mutants], offering the opportunity for evaluating such correlation systematically. In systems where the correlation holds, physical chemists can analyze the energetics of association and perform computer simulations of the recognition process based on this analysis. While still far from satisfactory, structure-based theoretical approaches to macromolecular recognitionÐ another subject of our chapterÐrepresent an active ®eld of research with an important future in the postgenomic era. In a matter of years, largescale efforts in protein structure determination are expected to provide

STRUCTURAL BASIS OF MACROMOLECULAR RECOGNITION

11

structural information on most of the protein folds (Vitkup et al., 2001). But information on all the possible modes of speci®c protein±protein or protein±DNA interactions, whose number is orders of magnitude larger than the number of folds, is being generated much more slowly. We hence believe that docking algorithms and computer simulations based on these approaches will be valuable predictive tools despite the approximations they make. They therefore deserve consideration alongside the more established experimental techniques with which their results may be compared. In addition to being a challenging test of our understanding of the physical chemistry of biological interaction, reliable methods for predicting protein±protein interaction may therefore be of great consequences to biology as a whole and could prove to be even more useful than methods for predicting protein folds from their amino acid sequence.

II. GEOMETRIC AND CHEMICAL FEATURES OF MACROMOLECULAR RECOGNITION A. Interface Area The interface area is a geometric quantity related to the solvent accessible surface area (Lee and Richards, 1971; Chothia and Janin, 1975). It can be easily derived from atomic coordinates of a complex between two macromolecules and provides a natural way to estimate the extent of their contact. It requires evaluating ®rst the solvent accessible surface area A12 of the complex, then A1 and A2 , the surface areas of dissociated components: B A1 A2

A12 :

(1)

The rolling sphere algorithm of Lee and Richards (1971) and a number of related algorithms yield the solvent accessible surface areas Ê for water. The (Fig.1). The radius of the solvent probe is near 1.5 A interface area B is the area of the protein surface that becomes buried at the interface when the two molecules associate, but, as the calculation uses only coordinates of the complex, it ignores conformation changes which may affect the accessible surface area of the components. The contribution of each molecule to B can be evaluated separately and is approximately equal to B/2. Thus, other authors may prefer to quote values of B/2 ( Jones and Thornton, 1995, 1996). However, when the surfaces in contact have a strong curvature, the convex side tends to contribute more interface area than the concave side because accessible surface areas are measured one probe radius away from

12

È L JANIN SHOSHANA J. WODAK AND JOE

FIG. 1. Accessible surface and interface in macromolecular complexes. Positions of the center of the solvent probe W de®ne the solvent accessible surface (shaded) of molecules 1 and 2. When they form a complex, W is expelled from the interface and some of the accessible surface is lost. The area of the buried surface is the sum of the accessible surface areas of the two molecules less that of the complex.

the van der Waals surface of the molecules. Protein±protein interfaces are rather ¯at (Argos, 1988; Goodsell and Olson, 1993; Jones and Thornton, 1996) and curvature effects are minimal, except in some protease±inhibitor complexes where the enzyme active site is concave and the inhibitor surface is convex. 1. Crystal Contacts The Protein Data Bank contains many examples of intermolecular contacts for which interface areas can be measured. The most abundant type of contact is due to crystal packing (Janin and Rodier, 1995; Carugo and Argos, 1997; Dasgupta et al., 1997). It forms a category of low-af®nity, nonspeci®c interaction, which may serve as a background for the study of biologically relevant interactions. Figure 2 shows the distribution of the interface areas of the 1260 pairwise crystal packing contacts observed in a sample of 152 crystal forms of monomeric proteins ( Janin and Rodier, Ê 2, 1995). Most of the pairwise interfaces are small and cover about 600 A 2 Ê per molecule, only a few percent of its accessible surface that is, 300 A

13

STRUCTURAL BASIS OF MACROMOLECULAR RECOGNITION

20 Complexes

>3000

2900

2500

2100

1700

1300

900

0

500

10

100

Interfaces

Crystals

2

Interface area (A )

FIG. 2. Histogram of the interface area in protein crystals and protein complexes. The sample includes 75 interfaces in protein±protein complexes (Lo Conte et al., 1999) and 1260 pairwise interfaces observed in 152 crystal forms of monomeric proteins (Janin and Rodier, 1995); for the second type of interface, the vertical scale should be multiplied by 20.

area. However, each protein molecule forms 6 to 12 such contacts and the total buried surface area of any protein molecule can be large, sometimes larger than half of the whole protein surface area, even though pairwise contacts are of limited size. Ê 2 are nevertheless A few pairwise interfaces in the range 1300±1900 A observed in crystals. Almost all result from the presence of twofold and other point-group symmetry elements, which are relatively uncommon in crystals of monomeric proteins. Their occurrence suggests that the formation of dimers or other small oligomers in solution precedes crystallization under the conditions where these particular crystals are obtained at protein concentrations typically in the range 10 5 10 3 M. The large majority of the crystal contacts are associated with lattice translations and screw rotations not found in oligomeric proteins. Their size distribution resembles that of the transient interfaces created by the random collision of two small proteins simulated in the computer

14

È L JANIN SHOSHANA J. WODAK AND JOE

discussed in Section IV,C. Thus, this type of crystal contact can also be viewed as resulting from random protein association ( Janin, 1997). 2. Protein±Protein and Protein±DNA Complexes Together with the histogram of crystal contact interface areas, Fig. 2 shows the distribution of the interface areas which are observed in a sample of 75 protein±protein complexes (Lo Conte et al., 1999). It is Ê 2 , has a well-de®ned maximum around very different; it starts at 1100 A 2 Ê Ê 2 . The lack of B 1500 A , and contains some large values above 3000 A 2 Ê interfaces below 1100 A indicates that the formation of a stable and speci®c complex between two proteins requires making a suf®cient number of contacts and removing water from part of the protein surface. This requirement sets a lower limit to the interface area. The presence of a peak in the histogram suggests that there is a preferred or ``standard size'' for protein±protein interfaces. In the sample of Lo Conte et al. Ê 2, (1999), 70% of the complexes had interfaces in the range 1600400 A which we take to de®ne this standard size, and 27% had interfaces larger Ê 2. than 2000 A Remarkably, the rules governing the minimum and preferred sizes of protein±protein interfaces also apply to protein recognition of doublestranded DNA, and they probably have the same physical origin. No protein±DNA interface in a sample of 65 complexes analyzed by Nadassy Ê 2 . Most are much larger and bury up to et al. (1999) buries less than 1100 A Ê 2 , forming a broad distribution with a peak around B 3000 A Ê2 6000 A Ê 2 (Fig. 3, top). Protein±DNA interfaces are larger on rather than 1500 A average than protein±protein interfaces, but many DNA-binding proteins are dimers or oligomers and form two or more interfaces with DNA. Others, for instance, the zinc ®nger-containing transcription factors, are tandem-repeat proteins. The complexes with DNA generally contain several binding units which are related by a twofold symmetry or by a screw symmetry in the case of tandem repeats. This was taken into account by Nadassy et al. (1999), who calculated the histogram of the buried surface area per binding unit (B/unit) shown in Fig. 3 (bottom). Ê 2 in this histogram. There are some interfaces with B/unit less than 1100 A Ê2 All but two are with the zinc ®ngers, which bury approximately 900 A per unit. Zinc ®ngers are very small protein motifs, presumably too small to make a larger interface with DNA. They always occur as repeats and there is no evidence that an isolated zinc ®nger can form a stable complex with DNA. Of the remainder, half of the complexes have B/unit in the Ê 2 , as for a standard size protein±protein intersame range, 1600400 A face. This range includes almost all the binding units in transcription factors other than zinc ®ngers, but few of the interfaces in enzymes acting on DNA. Enzymes such as DNA polymerase or recombinases are

STRUCTURAL BASIS OF MACROMOLECULAR RECOGNITION

Miscellaneous Zn fingers TBP Eukaryotic TF Prokaryotic TF Enzymes

8

Protein-DNA interfaces

15

6

4

2

Interface area B (A )

Miscellaneous TBP Zn fingers Eukaryotic TF Prokaryotic TF Enzymes

12

Protein-DNA interfaces

>5000

4500

4100

3700

3300

2900

2500

2100

1700

1300

900

500

0

100

2

8

4

2900

>3300

2

2500

2100

1700

1300

900

500

100

0

B per unit (A )

FIG. 3. Protein±DNA interface areas. The sample of protein±double-stranded DNA complexes analyzed by Nadassy et al. (1999) comprises 65 complexes, 16 with enzymes, 9 with prokaryotic transcription factors (TF), 37 with eukaryotic transcription factors, and 3 with miscellaneous proteins. (Top) Histogram of the interface areas. (Bottom) When the proteins are homodimers or direct repeats (of zinc ®ngers, for instance), which is the case for most proteins in this sample, the complexes contain two or more binding units corresponding to a monomer. The histogram is drawn for these bindings units. The ®ve DNA complexes with zinc ®ngers and four with TBP are set apart from other eukaryotic transcription factors.

16

È L JANIN SHOSHANA J. WODAK AND JOE

multidomain units, with several domains interacting with DNA in addition to their active site, and the interface tends to be larger than in transcription factors (see Section III,D). B. Geometric Complementarity and Atomic Packing 1. Evaluating the Geometric Complementarity of Two Macromolecular Surfaces The geometric complementarity of the two surfaces in contact, which optimizes van der Waals contacts, is a major element of the recognition process between two molecules. For proteins, geometric complementarity has been estimated in a number of different ways. Lawrence and Colman (1993) describe the two molecular surfaces as a set of closely spaced points x and evaluate the function: S(x) cos u exp (

wd2 ):

(2)

Here, d is the distance from grid point x of one surface to the nearest grid point on the other surface, and u is the angle of the normal vectors to the surfaces at these two points. The grid points are de®ned with the program MS of Connolly (1983) and the weight factor w is adjusted Ê 2 . A protein±protein interface contains many thousands of to 0.5 A grid points and the values of S(x) spread between 1 and 1. Most negative and low positive values originate from grid points at the edge Ê are of the contact region; those in a peripheral band of width 1.5 A removed from the distribution. The remainder has a peak in the range 0.6±1.0, which indicates that, when the molecular surfaces are in contact (small d), their normal vectors are in general approximately parallel (small u), yielding a natural way to de®ne complementarity. However, the observed distribution of S(x) in actual protein±protein interfaces is wide and skewed and is thus dif®cult to describe by a single number. To compare the interfaces in different systems, Lawrence and Colman (1993) select the median value of the distribution as a shape correlation index Sc . They ®nd Sc to be in the range 0.71±0.76 in a sample of four protease±inhibitor complexes where the complementarity of the interacting surfaces is visually excellent. In ®ve antigen±antibody inhibitor complexes, Sc is signi®cantly lower: 0.65±0.67. Thus, the shape correlation index suggests that the geometric match of an antigen combining site with the cognate epitope surface is less perfect than between an enzyme active site and an inhibitor surface which has undergone a much longer selection process than in the immune system. Jones and Thornton (1995, 1996) use a gap index to characterize complementarity on the basis of the compactness of the interface. The

STRUCTURAL BASIS OF MACROMOLECULAR RECOGNITION

17

gap index is the distance de®ned by the ratio of the gap volume Vgap to the interface area B: igap Vgap =B:

(3)

The gap volume between the two protein surfaces is estimated with the program SURFNET (Laskowski, 1995). The program places a set of spheres in contact with atoms on the solvent accessible surfaces of both components of the complex, removes the spheres with diameters less Ê or larger than 10 A Ê , and sums the volumes of the remainthan 1 A ing spheres. The spheres in contact with atoms at the edge of the interface tend to be larger than those in the middle and, therefore, to contribute more to Vgap . Thus, the gap index is more representative of the packing at the periphery of the interface than in the core region. This may be the reason for its wide variability, much greater than for Sc and other complementarity indices. In a sample of 27 protein±protein Ê ( Jones and complexes, the value of igap was in the range 0.35±5.2 A Ê Thornton, 1996) and the overall mean was 2.5 A. A somewhat lower Ê ) was observed in enzyme±inhibitor complexes, and a mean value (2.2 A Ê ) was observed in six antigen±antibody complexes. higher value (3.0 A The calculation was also performed on a sample of 26 protein±DNA complexes ( Jones et al., 1999), which also covered a wide range of igap Ê ). values (0.8±4.3 A 2. Voronoi Volumes and the Atomic Packing at Protein±Protein Interfaces An alternative approach to geometric complementarity relies on an analysis of the atomic packing. Complementary surfaces must form compact interfaces with close-packed atoms and few cavities. The packing density may be estimated by measuring the volumes of the Voronoi polyhedron drawn around each protein atom and comparing them to a set of reference volumes (Fig. 4). The use of Voronoi volumes to analyze the packing of protein atoms was proposed by Richards (1974) and by Finney (1975). Its application to protein±protein interfaces was ®rst performed by Janin and Chothia (1976). Updated sets of atomic volumes in proteins have recently been published by Tsai et al. (1999). They are averages of the Voronoi volumes of atoms buried inside globular proteins, which are smaller by 0±10% than in crystals of small organic molecules (Harpaz et al., 1994). They must represent close-packed atoms despite the occasional occurrence of internal cavities. The compactness of a set of atoms relative to this reference can be estimated from the ratio of the sum V of the Voronoi volumes to the sum V0 of the reference volumes. Any set of atoms with V=V0 1 should be closepacked, whereas a larger V=V0 ratio indicates a looser packing.

18

È L JANIN SHOSHANA J. WODAK AND JOE

A W

FIG. 4. Voronoi polyhedra and packing volumes. The polygon drawn around atom A is the equivalent in two dimensions of the three-dimensional Voronoi polyhedon. Atom A is accessible to the solvent and the right-most edge of the polygon is de®ned by the presence of water molecule W. The position of W must be known in order to draw the polygon.

The Voronoi calculation can be performed on protein atoms buried at interfaces as well as inside proteins. However, the procedure has a serious limitation: a Voronoi polyhedron can be drawn around an atom only if it is completely surrounded by other atoms. At interfaces, only about onethird of the atoms that contribute to the interface area B have zero accessible surface area. These atoms are located mostly at the center of the interface, which biases the V=V0 ratio in an opposite way to the gap index, which is biased toward the periphery. However, highresolution X-ray structures usually report positions for immobilized water molecules, which are abundant at interfaces (see Section II,D). These molecules may also be used to close the polyhedra, making the evaluation of Voronoi volumes possible for atoms which are surrounded by both protein atoms and immobilized water molecules (Fig. 4). On average, there are as many such interface atoms as there are completely buried atoms. Thus, a Voronoi calculation taking into account the crystallographic water molecules applies to two-thirds of the interface atoms on average instead of only one-third and up to 90% in speci®c cases (Lo Conte et al., 1999).

STRUCTURAL BASIS OF MACROMOLECULAR RECOGNITION

19

In the 75 protein±protein complexes of Lo Conte et al. (1999), 96% of the interfaces have V=V0 in the range 0.97±1.06. Thus, the packing of atoms buried at protein±protein interfaces is very similar to that of the protein interior. In 36 complexes with X-ray structures at a resolution of Ê and better, the V=V0 ratios calculated in the presence of water 2.5 A molecules were distributed over a narrow range of 0.97±1.03 (Fig. 5, top). Therefore, their interfaces are packed like the protein core, except that water, which is almost entirely excluded from the protein core, makes an important contribution to the packing at protein±protein interfaces. There is one exception to this rule in the sample analyzed by Lo Conte et al. (1999): the complex between cytochrome peroxidase and cytochrome c [PDB code, 1ccp (Pelletier and Kraut, 1992) ]. Its interface is small and has only a few buried atoms and a large volume ratio (1.07). In contrast, the 19 protease±inhibitor and the 19 antigen±antibody complexes of this sample have mean V=V0 ratios of 1.00 and 1.01, respectively. Thus, unlike Sc and the gap index, the volume ratio indicates that these two types of interfaces are close-packed and shows no difference in their packing density, at least for their buried atoms. 3. Packing at Protein±DNA Interfaces The Voronoi volume calculation was also carried out for protein atoms at protein±DNA interfaces in 28 high-resolution structures of protein± DNA complexes by Nadassy et al. (1999). The coordinate ®les also contain water positions, and volume calculations were performed in the presence of these water molecules. Fifty-seven percent of the interface protein atoms could be included and the V=V0 ratios were found to spread over a narrow range of 0.97±1.04, with a mean of 1.01 (Fig. 5, bottom). Thus, the atomic packing is very similar to that found for protein±protein interfaces. Omitting water molecules from the calculations yielded a broader range (0.94±1.1) and many fewer buried interface atoms. Thus, water molecules once more play a key role in fostering ef®cient packing at protein±nucleic acid interfaces. Volume calculations can also be performed for DNA atoms, but no reference volumes have been reported for these atoms and their packing at the interface could not be evaluated. Recently, Nadassy et al. (2001) measured the atomic volumes in crystals of B-form DNA and recomputed the volume ratios at protein±DNA interfaces taking B-DNA volumes as a reference. In a sample of 25 protein±DNA complexes, the ratios were also found to be close to unity (1.01 0.03) for the DNA atoms. Thus, they pack at protein±DNA interfaces as they do in crystals of B-DNA. In the same study, the number and volume of cavities located at the protein±DNA were also surveyed. It was found that cavities are frequent at protein±DNA interfaces. Most contain water molecules and

20

È L JANIN SHOSHANA J. WODAK AND JOE

Protein-protein

< 60% > 60%

Interfaces

15

10

5

1.07

1.06

1.05

1.04

1.03

1.02

1.01

1.00

0.99

0.98

0.97

0.96

0.95

0.94

0 Packing ratio (with water) 12

Protein-DNA

10

Interfaces

8

6

4

2

1.10

1.08

1.06

1.04

1.02

1.00

0.98

0.96

0.94

0 Volume ratio (with water)

FIG. 5. Atomic packing at protein±protein and protein±DNA interfaces. Histograms of the V/V0 packing ratio, where V is the sum of the Voronoi volume of all interface atoms in a complex and V0 is the sum of reference volumes for the same atoms. The reference is a set of atoms buried inside globular proteins. Crystallographic water molecules are taken into account when evaluating Voronoi polyhedra. (Top) HistoÊ gram of the packing ratio in 36 protein±protein complexes determined at 2.4-A resolution or better (Lo Conte et al., 1999). Black columns are the 17 complexes where more than 60% of the interface has zero accessibility in the presence of crystallographic water; empty columns have less than 60% of the interface having zero accessibility. The interface with V/V0 1:07 is in cytochrome c±cytochrome oxidase. (Bottom) Histogram of the packing ratio for interface protein atoms in 24 protein± Ê resolution or better (Nadassy et al., 1999). DNA complexes determined at 2.4-A

STRUCTURAL BASIS OF MACROMOLECULAR RECOGNITION

21

the fraction of water-containing cavities is markedly higher at protein± DNA interfaces than inside globular proteins. This provides further evidence of the important role played by solvent molecules in these interfaces. 4. Shape Complementarity vs Atomic Packing Different shape complementarity indices yield different views of the interfaces between macromolecules. We attribute the difference between the volume ratio and the gap index calculations to the fact that the latter overemphasizes the contribution from peripheral regions of the interfaces. In antigen±antibody or protein±DNA complexes, the periphery may be poorly packed relative to the central region. Also, water molecules play a more important role at these interfaces than in most protease± inhibitor interfaces. Crystallographic water was not considered in the gap index calculation, but its incorporation signi®cantly improves the shape correlation index. A comprehensive analysis of the relationship between the gap volume index of Laskowski (1995), the shape correlation index of Lawrence and Colman (1993), and the atomic volume ratios in protein±protein complexes remains to be performed. The three parameters were nevertheless estimated by Nadassy et al. (2001) for the same sample of 25 protein± DNA complexes. Their values are very poorly correlated: the linear correlation coef®cients between the volume ratio on one hand and the igap and Sc values, on the other hand, were 0.4 and 0.5, respectively; the correlation coef®cient between igap and Sc was even lower, 0.2. Thus, the information given by each of the three parameters must be incomplete and biased to some extent by the way it is derived from the atomic coordinates. C. Chemical Composition and Polar Interactions 1. Composition of Protein±Protein Interfaces The average chemical composition of the regions of the protein surface that forms the interface of protein±protein complexes is listed in Table I. On average, interfaces comprise 56% nonpolar carbon-containing groups, 29% neutral polar groups (all groups with N and O atoms not carrying a full electric charge), and 15% charged groups. These contributions are evaluated as percentages of the total interface area. This result indicates that protein±protein interfaces are similar to the accessible surface of proteins. Other types of buried surfaces, such as the interface between subunits in oligomeric proteins, tend to be more hydrophobic. The solvent accessible surface of the average small globular protein is 57% nonpolar, 24% neutral polar, and 19% charged (Miller et al., 1987);

22

È L JANIN SHOSHANA J. WODAK AND JOE

TABLE I Polar/Nonpolar Character of Protein±Protein and Protein±DNA Interfaces Surface/interface Protein surfaceb Protein±protein interfacesc

Number

Nonpolara

Polara

Chargeda

37 75

57 (4) 56 (6)

24 (5) 29 (6)

19 (5) 15 (6)

Protease±inhibitor

19

61 (4)

29 (3)

9 (4)

Antigen±antibody

19

51 (4)

33 (3)

15 (4)

Others

37

56 (4)

26 (4)

17 (4)

Protein-to-DNA interfacesd

65

52 (8)

24 (7)

25 (9)

DNA-to-protein interfacesd

65

41 (7)

16 (4)

43 (8)

DNA surfacee

65

47 (3)

19 (3)

34 (4)

a Percentage of the area of the solvent accessible surface or the interface contributed by nonpolar (carbon containing) and neutral or charged polar (nitrogen/oxygen containing) groups with the standard deviation in parentheses. bAverage area composition of the solvent accessible surface of 37 small globular proteins analyzed by Miller et al. (1987). c Average area composition of protein±protein interfaces in Lo Conte et al. (1999). dAverage area compositions of the protein side and the DNA side of protein±DNA interfaces of Nadassy et al. (1999). The 25% charged surface includes 23% positive and 2% negative charges. e Average area composition of the solvent accessible surface of DNA of Nadassy et al. (1999). The charged surface is all negatively charged.

the average subunit interface of an oligomeric protein is 65% nonpolar, 22% neutral polar, and 13% charged ( Janin et al., 1988). Jones andThornton (1996) reached the same conclusion based on a hydrophobicity scale where the average protein interior is near 0.25 and the average protein surface near 0.25. On that scale, the average subunit interface in a homodimer is at 0.12 and the average interface in protein±protein complexes is at 0.14. Thus, a subunit interface in an oligomeric protein has a hydrophobic character that protein±protein interfaces, in general, do not have. This difference in composition is clearly related to the different status of the two types of interfaces: an oligomeric protein is a permanent assembly, whereas the two components of a complex exist as separate entities that may not carry large hydrophobic patches on their surface. There are also signi®cant differences from this point of view between different complexes or categories of complexes (Table I). With 61% nonpolar area and a hydrophobicity 0 in the scale of Jones and Thornton (1996), protease±inhibitor interfaces are more hydrophobic on average than other types of protein±protein interfaces, and some are as hydrophobic as those in homodimers. In addition, they have very little charged area. In contrast, the average antigen±antibody interface has a

23

STRUCTURAL BASIS OF MACROMOLECULAR RECOGNITION

hydrophobicity of 0.22 and 51% nonpolar area. Extreme values of the nonpolar contribution to all protein±protein interfaces are 42 and 71%. The area contribution of the charged surface, 15% on average, varies even more widely from one interface to another, from near zero in some protease±inhibitor complexes to 25±27% in several antigen±antibody and other types of complexes. 2. Polar Interactions at Protein±Protein Interfaces Polar groups at interfaces form hydrogen bonds either between themselves (direct H-bonds) or with water molecules. On average, there are about 10 direct interface H-bonds per complex. However, the range is very wide, from only 1 to as many as 34 H-bonds, in the sample analyzed by Lo Conte et al. (1999). The estimated number of H-bonds is sensitive to both the geometric parameters of the bonds and the accuracy of the structure. In addition, a detailed analysis of their geometry (Xu et al., 1997) suggests that on average H-bonds at protein±protein interfaces are less optimal and possibly weaker than those formed intramolecularly. Table II lists values for a subset of high-resolution X-ray structures where H-bond attribution should be reliable. There is a correlation between the number of H-bonds with the interface area B, but it is mediocre. In high-resolution structures, there is on average one HÊ 2 of B, with a correlation coef®cient of 0.84, and one bond per 170 A Ê 2 of polar area, with a slightly better correlation coef®cient. The per 72 A former number is equivalent to that quoted in Table 2 of Jones and Ê 2 of B/2. In comparison, Thornton (1996): 1.13 H-bonds per 100 A Ê 2 of B/2. A comparison of dimeric proteins have 0.7 H-bonds per 100 A TABLE II Interface H-Bonds and Water Molecules Type of complex

Protein±protein

Protein±DNA

Number of complexesa

22

H-bondsb

11 (4)

24 (8)

Water moleculesc

22 (11)

28 (11)

Ê 2) Interface area B (A 2 Ê B per H-bond (A ) Ê 2) B per water (A

1820 (640)

8

2990 (680)

170

135

115

115

a Ê Protein±protein and protein±DNA complexes with X-ray structures at 2-A resolution or better listed in Lo Conte et al. (1999) and Nadassy et al. (1999). bAverage number of protein±protein and protein±DNA H-bonds determined by the program HBPLUS (McDonald and Thornton, 1994) with the default geometric parameters. Standard deviations are in parentheses. c Average number of crystallographic water positions at a distance less than Ê from atoms of both components. Standard deviations are in parentheses. 3.5 A

24

È L JANIN SHOSHANA J. WODAK AND JOE

the two values indicates that interface H-bonds are 60% more abundant per unit area in complexes than in dimers, in line with the lesser hydrophobic character of the interfaces. 3. Composition of Protein±DNA Interfaces Somewhat surprisingly at ®rst glance, the chemical composition of interfaces contributed by the protein component in a sample of 65 protein complexes with double-stranded DNA (Table I) is not so different from that of a protein±protein interface and therefore from the solvent accessible surface of small globular proteins. The protein surface in contact with DNA is slightly less nonpolar, 52% instead of 57%, than the surface in contact with water. It has the same neutral polar component (24%) and a larger charged component, 25% instead of 19%. There is, however, a clear difference in the sign of the charged component. On the average protein surface, positive and negative charges approximately balance each other, whereas interfaces with DNA are highly enriched in positive charges from lysine and arginine side chains and are almost entirely devoid of negative charges from carboxylates. The abundance of positive charges on the protein side of the interfaces counterbalances the dominance of the negatively charged phosphate group on the DNA side, which carries no positively charged group at all. The negatively charged component represents one-third of the accessible surface of double-stranded DNA and an even larger fraction (43%) of the surface in contact with the proteins (Table I). Whereas nucleic acid bases contribute equally (about 27%) to both the interfaces and the DNA accessible surface, the sugar moiety contributes less to interfaces than to the surface. The sugar surface of DNA has no free hydroxyl and is comparatively nonpolar. The nonpolar groups from the sugars and bases are partly excluded from contact with the protein, and the DNA side of the interfaces, which is 41% nonpolar, is even less hydrophobic than the 47% nonpolar accessible surface of DNA. In addition, the DNA side of the interfaces is less hydrophobic than the protein surface which is in contact with it and much more charged: 43% instead of 25%. The actual excess in electric charge is real, but not as important as the fractional interface areas suggest. The average protein±DNA interface contains 15 negative charges from phosphates and 12 positive charges from lysines or arginines. Presumably, metal cations ensure electrostatic neutrality, but very few are located in electron density maps and reported in the deposited structures. 4. Polar Interactions in Protein±DNA Complexes The many polar groups buried at protein±DNA interfaces form polar interactions bridging the protein and the nucleic acid moieties, either

STRUCTURAL BASIS OF MACROMOLECULAR RECOGNITION

25

directly or via water molecules. The number of hydrogen bonds per complex, evaluated by Nadassy et al. (1999), with the default parameters of HBPLUS (McDonald and Thornton, 1994) varies widely, from 2 between thrombin and the single-stranded DNA aptamer (1hut) to 47 in the complex with the RAP1 telomere-binding protein (1ign). On average, there are 22 H-bonds in complexes with double-stranded DNA. A cursory analysis, limited to the survey of the donor±acceptor distances, suggests that their geometry is in general comparable to that found inside proteins. The average number of H-bonds listed in Table II for a high-resolution subset of X-ray structures is slightly larger. In general, the number of protein±DNA H-bonds increases with the size of the interface, but, as in protein±protein interfaces, the correlation is mediocre. In X-ray structures Ê or better, there is on average one hydrogen bond per at a resolution of 2 A 2 Ê 135 A of interface area, 25% more than in protein±protein interfaces. Ninety percent of these H-bonds have the donor group on the protein and the acceptor group on the DNA or RNA, a direct consequence of the chemical nature of the interacting macromolecules. In double-stranded DNA, the phosphate group contributes 60% of the bonds, and the bases contribute 34%. On the protein side, the major H-bond donors are main chain NH groups and the charged side chains of lysine and arginine. These three groups account for 60% of the H-bonds to DNA and 65% of those to the phosphates. A variety of polar amino acid side chains provide the remainder. Bonds to the phosphate group are not sequence speci®c, and the speci®city of protein±nucleic acid interactions is believed to result largely from direct and water-mediated base recognition by protein groups. In double-stranded DNA such recognition can occur in the major or the minor groove. Direct base recognition also takes place in single-stranded DNA or RNA, but statistics are insuf®cient to derive general rules. Polar interactions in the major groove, which involve groups at positions 6 and 7 of purines or at position 4 of pyrimidines, represent 80% of the hydrogen bonds to the bases and an average of 6 bonds per complex. Minor groove interactions with positions 2 and 3 of purines and position 2 of pyrimidines are four times less abundant (Nadassy et al., 1999). D. Wet and Dry Interfaces As Table II indicates, water is present in abundance at protein±protein and protein±DNA interfaces. To analyze interactions with water, the data set must be restricted to high-resolution X-ray structures because water positions are often not reported in medium-resolution coordinate sets. The number of interactions with water is probably still underestimated at high resolution, and the large standard deviation is partly artifactual, for

26

È L JANIN SHOSHANA J. WODAK AND JOE

there is still no established practice among crystallographers for describing the solvent structure. In Table II, there is an average of one water Ê 2 of interface area, in both protein±protein and molecule per 115 A protein±DNA complexes. Although the correlation between the number and the area is even weaker than for H-bonds, the ratio suggests that about 1 in 12 of the solvent molecules which hydrate the protein or DNA surface remains at the interface when the complex forms. As nearly all reported interface waters are involved in bridging hydrogen bonds, the data establish that protein±protein and protein±DNA interfaces contain at least as many water-mediated interactions as direct hydrogen bonds or salt bridges. Water is therefore a major player in the polar interactions that stabilize the complexes. Figure 6 illustrates how water molecules are distributed at several protein±protein interfaces. At the chymotrypsin±ovomucoid interface [PDB code, 1cho (Fujinaga et al., 1987) ], a typical protease±inhibitor interface, water molecules line the edge of the interface and form a ring around a dry central patch. In the FvD1.3±FvE5.2 antigen±antibody complex [PDB code, 1dvf (Braden et al., 1996) ], the interface appears wet throughout. The amount of buried surface and the number of water molecules are approximately the same in the two interfaces, and the different distribution is a consequence of their chemical composition. The D1.3±E5.2 antigen±antibody interface is only 42% nonpolar and much less hydrophobic than the 64% nonpolar chymotrypsin±ovomucoid interface. Homodimer interfaces, which are generally at least as hydrophobic as chymotrypsin±ovomucoid, also tend to form dry patches surrounded by rings of water molecules [see for instance Larsen et al. (1998), and images on the Visual Survey of Homodimeric Proteins Web site of The Scripps Research Institute (see Web Sites) ]. Whether dry or wet, all these interfaces are close-packed, and therefore water molecules must completely ®ll a set of cavities at the D1.3±E5.2 antigen±antibody interface. Although antigen±antibody interfaces are generally less hydrophobic and protease±inhibitor interfaces are more hydrophobic than the average protein surface, dry antigen±antibody interfaces also exist, and there are examples of wet interfaces in enzyme± inhibitor complexes, the barnase±barstar complex described in Section III being an example. Moreover, they are standard size interfaces made of a single patch on the component surfaces. In contrast, the larger Ga±Gbg interface in transducin (Lambright et al., 1996) and the b-lactamase interface with the BLIP inhibitor (Strynadka et al., 1996) form two patches, which are essentially dry and lined with water molecules. Protein±DNA interfaces, which are more polar due to the phosphate groups of DNA and the abundance of positively charged protein groups, are generally ``wet'' like the D1.3±E5.2 antigen±antibody interface. This is

STRUCTURAL BASIS OF MACROMOLECULAR RECOGNITION

27

FIG. 6. Water at protein±protein interfaces. Four protein±protein complexes are shown with one component in front of the other and the protein backbones drawn as a tube. The gray surface belongs to the back protein and is in contact with the front protein. Red spheres represent interface water molecules. The chymotrypsin±ovomucoid (1cho) complex has a ``dry'' protease±inhibitor interface. The interface between the Fv fragments of antibodies D1.3 and E5.2 is a ``wet'' antibody±antigen complex (1dvf). These two interfaces are standard size, whereas the b-lactamase±BLIP inhibitor interface and the Ga±Gbg interface of transducin (1got) are large interfaces. Figure taken from Lo Conte et al. (1999) and drawn with GRASP (see Web Sites).

illustrated in Fig. 7 by the l phage repressor [1lmb (Beamer and Pabo, 1992) ] and the papilloma E2 protein [2bop (Hegde et al., 1992) ]. In these complexes, the DNA double helix is bent, but it retains the standard B conformation and the contacts are mostly in the major groove. Nevertheless, there are some dry protein±DNA interfaces. For instance, water is excluded from the central surface patch of the

28

È L JANIN SHOSHANA J. WODAK AND JOE

FIG. 7. Water at protein±DNA interfaces. Four protein±DNA complexes are shown with the DNA placed in front of the molecular surface of the protein, colored according to the electrostatic potential (red, negative; blue, positive). The complexes are with the lambda repressor dimer (1lmb); three zinc ®ngers from the Zif 268 transcription factor (1aay); the human TATA box-binding protein (1cdw); and the dimeric E2 domain of papilloma virus (2bop). Red spheres represent interface water molecules. Figure taken from Nadassy et al. (1999) and drawn with GRASP (see web sites).

TATA box-binding protein [TBP, PDB code, 1cdw (Nikolov et al., 1996) ], where mostly nonpolar protein side chains contact the bases and sugars in the minor groove of DNA. A large distortion of the double helix pushes

STRUCTURAL BASIS OF MACROMOLECULAR RECOGNITION

29

the phosphate backbone to the edge of the interface. In general, water molecules tend to follow the phosphate backbone of DNA and hydrogen bond to phosphate oxygens, but water-mediated hydrogen bonds to the bases are also very frequent. The average protein±DNA interface contains about 13 phosphate±water±protein bonds and 6 base±water±protein bonds (Nadassy et al., 1999), and these numbers are likely underestimates.

E. Conformational Changes 1. Protein±Protein Association An essential feature of protein±protein association is the way it affects the conformation of the protein components. When X-ray structures of the components are known independently of that of the complex, the extent and nature of the conformational changes can be assessed. For moderate changes, a convenient measure of the extent of the changes is the root-mean-square distance (rmsd) of equivalent main chain atoms after least-squares superposition of the bound and free components. Ê. Many protein±protein complexes display a rmsd in the range 0.5±1 A Ê At the interface, the main chain typically moves by 1±2 A and a few surface side chains reorient. In solution, proteins probably undergo ¯uctuations of the same type and amplitude. Local movements of up to Ê and side chain rotations are also seen at crystal contacts when 1 A different crystal structures of the same protein are compared. Thus, the formation of these complexes has the same sorts of (minor) effects on conformation as packing forces in crystals, and it can be described to a good approximation as a rigid-body association mechanism. Many enzyme±inhibitor complexes assemble as rigid bodies in this way. An example is the trypsin±pancreatic trypsin inhibitor (PTI) system, where X-ray structures of the complex and of its components display a rmsd of Ê (Huber et al., 1974; Janin and Chothia, 1990). A majority of less than 1 A the protein antigens also undergo main chain movements no larger than Ê as they bind antibodies. On the antibody side of antigen±antibody 1±2 A complexes, the CDR3 loop of the heavy chain (H3 loop) can undergo Ê , whereas the rest of large changes, displacing some residues by up to 10 A the structure is maintained (Davies and Cohen, 1996). Ê and There are nevertheless complexes where the rmsd is well over 2 A large movements take place. The conformation change modi®es the shape and chemical character of regions of the protein surfaces, which are not preformed and complementary before they start interacting. The conformational change may be localized at the interface or affect the whole protein. Loops of polypeptide chain can move like the H3 loop

30

È L JANIN SHOSHANA J. WODAK AND JOE

of antibodies; an a-helix may unwind or change its orientation. A frequent feature is the movement of whole domains. Also, a protein may change its quaternary structure on association with another protein. An example is the human growth hormone receptor, which dimerizes on binding the hormone (de Vos et al., 1992). Of particular signi®cance are disorder-to-order transitions related to association: part or all of the polypeptide chain in one component is unfolded in the free form and becomes structured in the complex. We took the trypsin±PTI complex as an example of rigid-body association. The closely related trypsinogen±PTI complex is an example of a disorderto-order transition. Its structure is essentially identical to that of trypsin± PTI, yet, in free trypsinogen, loops amounting to about one-quarter of the polypeptide chain are fully disordered. In the presence of PTI, they TABLE III Conformation Changes in Protein±Protein Complexes with a Large Interfacea Complex Large protease complexes 1bth Thrombin E192Q±PTI 4htc Thrombin±hirudin

Ê 2) B (A

Type of change

2380

Large loop movements in thrombin

3350

C-terminal tail of hirudin becomes ordered

Other enzyme±inhibitor complexes b-Lactamase±BLIP

2560

Saddle-shaped BLIP inhibitor bends

1dfj RNase A±RNase inhibitor

2600

Horseshoe-shaped inhibitor opens up

1dhk a-Amylase±bean inhibitor

3080

Large loop movements in a-amylase

G proteins, signal transduction 1tx4 Rho±Rho GAP

2280

Loops becomes ordered in Rho and in GAP

1efu EFtu±EFts, Escherichia coli 1aip EFtu±EFts, Thermus thermophilus

3660 2940

Domains and loops move in EFtu In EFtu, domains move, an a-helix and loops shift; in EFts, N-terminal residues become ordered

1 got Transducin Gta Gtbg

2500

In Ga , a a-helix rotates, loops move; N- and C-terminal residues become ordered or refold; loops move in Gbg

2trc Gtbg -phosducin

4660

1®n CDK2±cyclin A

3400

Loops move in Gbg ; likely domain movements in phosducin, an extended molecule that wraps around Gbg In the kinase, domains move, an a-helix rotates, and a loop moves

2btf Actin±pro®lin 1hwg HGH receptor±HGH

2090 4200

Domains move in actin Receptor dimerizes; large helix movements in HGH

a Ê 2 for which the structure of at least one of the two componComplexes with B > 2000 A ents is known independently of that of the complex. Adapted from Table 2 of Lo Conte et al. (1999).

STRUCTURAL BASIS OF MACROMOLECULAR RECOGNITION

31

become ordered and take the same conformation as in the mature enzyme (Marquart et al., 1983). Others examples are cited in Section III,C, and Table III lists complexes taken from Lo Conte et al. (1999) showing large conformational changes. It suggests that essentially all Ê 2 for which the free components structure complexes with B > 2000 A is known undergo large changes or disorder-to-order transitions on association. 2. Protein±DNA Association Conformational changes play an even more important role in protein± DNA recognition. As in protein±protein interactions, extensive changes accompany the formation of large interfaces: disordered segments of the polypeptide chain become ordered, and whole subunits and domains move and loops rearrange on the protein surface (Nadassy et al., 1999). In protein±DNA interactions, the abundance of disorder-to-order transitions is probably underestimated, because many proteins that undergo such transitions do not yield useful crystals or NMR spectra in the free state. In addition to conformational changes, the interaction with DNA or RNA tends to reduce the mobility of the protein atoms. Evidence for this was obtained by computing the average crystallographic temperature factors of the interface atoms in the complex and the free protein, and comparing the difference to that observed for the core atoms (Nadassy et al., 1999). Large, differential reductions in mobility often also accompany large conformational changes. Changes that affect the DNA component also correlate to some extent with the size of the interface (Jones et al., 1999; Nadassy et al., 1999). Half of the protein±DNA complexes analyzed by Nadassy et al. (1999) undergo deformations of limited amplitude and of various types. The most frequent are bending or unwinding of the double helix and the widening of the major groove, where most of the interactions with the protein take place. Another quarter displays speci®c large local distortions, such as kinks, which sometimes severely disrupt the path of the helix (Jones et al., 1999; Nadassy et al., 1999). In most of these cases the distortions play a major functional role (Werner and Burley, 1997). It was also observed that DNA bending on complex formation can take either of two forms (Jones et al., 1999). In one form, the DNA bends toward the major groove, resulting in the compression of this groove and the widening of the opposite minor groove. In the other form, bending occurs toward the minor groove, resulting in the compression of this groove and the widening of the opposite major groove. This latter form is observed mainly in the so-called ``double-headed'' complexes (see below), where two protein subunits simultaneously bind the DNA.

32

È L JANIN SHOSHANA J. WODAK AND JOE

III. CLASSIFICATION OF PROTEIN±PROTEIN AND PROTEIN±DNA COMPLEXES A. Standard Size Protein±Protein Interfaces The histogram of the interface areas observed in a sample of 75 protein±protein complexes analyzed by Lo Conte et al. (1999) is shown in Fig. 8. In this sample, the majority of the complexes (70%) have interfaces in Ê 2 . This range includes 19 of 24 protease±inhibitor the range 1200±2000 A complexes, 18 of 19 antigen±antibody complexes, and 15 of 32 other types of protein±protein complexes in the sample analyzed by Lo Conte et al. (1999), who de®ne a standard size interface as having B 1600400 Ê 2 . General characteristics of such interfaces are listed in Table IV; two A typical examples are shown in Fig. 6: the chymotrypsin±ovomucoid inhibitor and the FvD1.3±Fv E5.2 interfaces. A standard size interface

Others Antibody-Antigen Protease-Inhibitor

16

Interfaces

12

8

4

>4500

4100

3700

3300

2900

2500

2100

1700

1300

900

500

100

0

Interface area (A2) FIG. 8. Histogram of interface areas in protein±protein complexes. The sample of 75 interfaces in protein±protein complexes (Lo Conte et al., 1999) includes 24 protease±inhibitor complexes, 19 antigen±antibody complexes, and 32 other types of Ê 2 , which de®nes complexes. The horizontal arrow marks the range B 1600 400 A standard size interfaces. This range includes 19 of the protease±inhibitor complexes, all but one of the 19 antigen±antibody complexes, and 15 of the others. Three Ê 2 . Twenty complexes have large interfaces with small interfaces have B 1150 A Ê 2: areas in the range 2000±5000 A

STRUCTURAL BASIS OF MACROMOLECULAR RECOGNITION

33

typically contains 9 H-bonds. The range in high-resolution structures is 6 to 13 H-bonds, and there is no signi®cant difference between the protease±inhibitor and the antigen±antibody interfaces from this point of view, even though the ®rst tends to be less polar than the second. Each Ê 2 and contributes about 90 side of these interfaces buries 800200 A interface atoms belonging to some 24 amino acid residues. An interface Ê 2 of surface area, but only about 30 such atom loses on average 9.2 A atoms become fully buried, plus an additional 30 when crystallographic water positions are taken into account. A standard size interface suf®ces to achieve very high stability: the trypsin±PTI complex has a dissociation constant (Kd ) below 10 13 M (Vincent and Lazdunski, 1972). It also suf®ces to achieve the high speci®city required by the immune system for antigen±antibody recognition. The range of interface areas is especially narrow in antigen±antibody Ê 2 , between Fab DeÂsireÂcomplexes. The largest interface area is 2340 A 1 and a TCR Fv domain [PDB code, 1kb5 (Housset et al., 1997) ]. The TABLE IV Characteristics of Standard Size Interfaces Characteristics Examples

Protein±protein

Protein±DNA

Protease±inhibitor, antigen±antibody

Recognition module

70%

50%

Fraction of samplea Ê2) Interface area (A

1600 400

1600 400

Interface residues (per component)b

24

24 amino acids 12 nucleotides

Interface atoms (per component)c Buried Buried with water

30 30

Accessible

27

Total

87

Polar interactions H-bonds Water mediated a

94 15

13 15

The sample of Lo Conte et al. (1999) comprises 75 protein±protein complexes; that of Nadassay et al. (1999) comprises 65 complexes with double-stranded DNA. bThe average number of amino acids or nucleotides that lose accessible surface in the complex is given per component molecule. A protein±protein interface contains twice that number. c The average number of atoms that lose accessible surface in the complex is given per component molecule. A protein±protein interface contains twice that number. These interface atoms are buried if they have zero accessible surface area in the complex (but nonzero in the isolated components), buried with water if they have zero accessible surface area in the complex, taking crystallographic water molecules into account, and accessible otherwise.

34

È L JANIN SHOSHANA J. WODAK AND JOE

Ê 2 , between lysozyme and the Fv fragment of antibody smallest is 1260 A D11±15 [PDB code, 1jhl (Chitarra et al., 1993) ]. The interface between Fab 27 and the core domain of HIV gp120 in complex with CD4 (1gc1) is Ê 2 [recalculated with the same parambelow the range, having B 1100 A eters as in Lo Conte et al. (1999); the original paper by Kwong et al. (1998) cites an even lower value]. However, the core domain lacks several surface loops that should contribute to the epitope recognized by Fab 27 and increase the interface area. Thus, the complete interface is likely to be standard size. Outside the protease±inhibitor and antigen±antibody categories, standard size interfaces are found in several types of enzyme±inhibitor complexes. An excellent example is the barnase±barstar complex. Barnase is a ribonuclease excreted by Bacillus amyloliquefaciens, which also produces barstar to block its activity in the cytoplasm, where it would be lethal to the bacterium (Hartley, 1993). The two proteins are globular and small (10 kDa), and their structure and their mode of recognition are known from high-resolution X-ray studies. Neither component undergoes signi®cant conformational changes upon association, and the inhibition is entirely attributable to the occlusion by barstar of the binding site for the RNA substrate and especially its interaction with the catalytic His-102 side chain (Mauguen et al., 1982; Guillet et al., 1993; Buckle et al., 1994). The thermodynamics and kinetics of the association, extensively studied by Fersht and collaborators, indicate that the complex is extremely stable, with Kd below 1 picomolar. Moreover, association is very fast at moderate ionic strength, due to a large electrostatic enhancement of the rate [Schreiber and Fersht 1993, 1995, 1996); see Section IV,A). Ê 2 equally distributed The barnase±barstar interface covers 1570 A between the two protein components. It is close-packed as shown by the Voronoi volume ratio V=V0 1:00. It contains 13 direct H-bonds or ion Ê X-ray structure, 32 water molecules bridge the two pairs and, in the 2.0-A molecules [PDB code, 1brs (Buckle et al., 1994) ]. These water molecules spread through the entire interface and play a major part in the packing. Nineteen residues of barnase and 18 of barstar contribute to the interface. The largest contributors in terms of buried area are Arg-59 and His102 of barnase and Tyr-29, Asp-35, and Asp-39 of barstar. Together, these 5 residues account for 36% of the interface area. Site-directed mutagenesis indicates that they all make large contributions to the enthalpy and free enthalpy of association. The substitution by alanine (alanine scanning) of each one of them reduces DGd by DDGd 3 to 8 kcal mol 1 and DHd by 5 to 14 kcal mol 1 (Frisch et al., 1997). Only one other residue, Arg-87 of barnase, is of comparable importance in the alanine-scanning study. This residue is completely buried in the complex, but it is already almost buried in free barnase, and its contribution to B is small.

STRUCTURAL BASIS OF MACROMOLECULAR RECOGNITION

35

In addition to these 6 residues which constitute ``hot spots'' for the interaction, the interface contains seven sites where DDGd is less than 3 kcal mol 1 . Alanine substitutions of most of the 25 other residues that contribute to the interface have no signi®cant effect on af®nity. Thus, the ``functional epitope'' de®ned by site-directed mutagenesis is much less extensive than the physical interface. This remark is valid in many other systems, but it should be recalled that only interactions involving side chains beyond the Cb (and in residues other than Gly, Ala, and Pro) can be tested by mutation. The numerous van der Waals contacts and H-bonds involving main chain groups escape analysis entirely. B. Small Interfaces and Low-Af®nity Complexes Three interfaces in the sample of 75 complexes of Lo Conte et al. Ê 2 , just below the range 1200±2000 A Ê 2 de®ning (1999) have B 1150 A the standard size. The distinction may be spurious, due to experimental error or to missing contacts. In two of the crystal structures, one of the components is not the complete protein, and, therefore, the interface may be incomplete as in the HIV gp120±CD4±Fab 27 crystal. These are the complex of a domain of Che A with Che Y [PDB code, 1a0o (Welch et al., 1998) ], implicated in the two-component system of bacterial signal transduction, and the complex of a domain of the HIV capsid protein with cyclophilin [PDB code, 1ak4 (Gamble et al., 1996) ]. The third complex with a small interface is between cytochrome c peroxidase and cytochrome c [PDB code, 2pcc (Pelletier and Kraut, 1992) ]. It is a fully functional enzyme±substrate complex, yet it has peculiar features. Its small interface contains fewer than 10 buried atoms and only one direct H-bond, but it also has 15 water molecules. The atomic packing is poor as shown by the exceptionally high V=Vo ratio of 1.07 mentioned above (this ratio is calculated in the presence of interface water). Moreover, two slightly different orientations of cytochrome c relative to the enzyme are observed in two high-resolution structures of the complex (Pelletier and Kraut, 1992). Thus, there is some ¯exibility at the interface in line with functional studies which suggest that electron transfer between the hemes of the cytochrome and of the peroxidase does not require an exact positioning of the two groups. This complex differs in structural and functional ways from another enzyme±substrate complex, trypsin±PTI, in which the geometry of binding at the protease active site is precisely determined. Once PTI binds trypsin, no turnover can take place due to the very tight binding and extremely slow (months) release of the inhibitor (Vincent and Lazdunski, 1972). To be functional, the cytochrome c peroxidase complex must be much less stable and release the product in a fraction of a second.

36

È L JANIN SHOSHANA J. WODAK AND JOE

Electron transfer, which has much less demanding geometric requirements than most reactions catalyzed by enzymes, may be compatible with a loose mode of association. Other biological recognition processes also tolerate a ¯exible geometry and require dissociation to be fast. In the immune system, cell adhesion between T lymphocytes and antigenpresenting cells is a process where ``imperfect interfaces'' (Ysern et al., 1998) play a central role. The X-ray structure of a complex between CD2 and CD58, two proteins that mediate this process, also has a small interÊ 2 ) in line with a Kd above micromolar values (Wang et al., face (B 1200 A 1999). There is a growing scienti®c interest in low-af®nity protein± protein recognition not only in cell adhesion, but also in signal transduction and a variety of regulatory processes. Thus, we expect to see more examples in coming years of protein±protein complexes with interfaces Ê 2 , poor atomic packing, few direct H-bonds, between 1100 and 1200 A and no signi®cant conformational change in the component proteins. Alternatively, a complex may have a standard size interface and a low af®nity if it displays conformational changes. An example is the trypsinogen±PTI complex, which has the same size interface as trypsin±PTI, but a Kd six orders of magnitude larger (Marquart et al., 1983), which represents the cost of the disorder-to-order transition induced in trypsinogen. C. Large Interfaces One-quarter of the interfaces in the sample analyzed by Lo Conte et al. Ê 2 . This includes several enzyme± (1999) have areas larger than 2000 A inhibitor complexes and most of the complexes involving guanine nucleotide-binding proteins (G proteins). Examples of the ®rst category are the complex between b-lactamase and the BLIP inhibitor (Strynadka et al., 1996) illustrated in Fig. 6 and the two complexes between the blood protease thrombin and the inhibitors ornithodorin and hirudin [PDB code, 1toc (van de Locht et al., 1996); PDB code, 4htc (Rydel et al., 1991) ] illustrated in Fig. 9. Thrombin is homologous to trypsin and ornithodorin is essentially a duplicated pancreatic trypsin inhibitor. While one of the PTI-like domains of ornithodorin blocks the active site of the protease as in trypsin±PTI, the other domain occupies a secondary or ``exo'' site on the enzyme, doubling the size of the interface. Hirudin from the medicinal leech is unrelated to PTI and ornithodorin. It has an ordered domain that binds at the active site and a disordered C-terminal tail that binds at the exo site (Rydel et al., 1991). Thus, both inhibitors make a large interface with thrombin by binding at two distinct sites. An example of the second category of large interfaces is found in the trimeric G-protein transducin discussed below. Despite their name, trimeric G proteins are made of two components, the Ga component, which

STRUCTURAL BASIS OF MACROMOLECULAR RECOGNITION

2PTC Trypsin-PTI

1TOC Thrombin-ornithodorin

4HTC Thrombin-hirudin

1430 A2

3500 A2

3400 A2

37

FIG. 9. Large interfaces in serine protease±protease inhibitor complexes. Trypsin and thrombin are evolutionarily related serine proteases of approximately the same size. The 58-residue PTI binds at the active site and makes with trypsin a standard size interface. Ornithodorin and hirudin are thrombin inhibitors that make a large interface with the enzyme. Ornithodorin is a duplicate of PTI; hirudin has an extended Cterminal tail that is disordered in solution.

is a GTPase, and the regulatory Gbg component. They associate or dissociate depending on the presence of other partners. Transducin is an example of a G protein, where the interface between Ga and Gbg buries Ê 2 . Figure 7 shows that it is made of two patches: a central patch 2500 A about the same size as the chymotrypsin±ovomucoid interface and a Ê 2 and involves the lateral patch which buries an additional 1000 A N-terminal helix on the Ga side. The interface between thrombin and ornithodorin and the Ga Gag interface in transducin are comparable in size to subunit interfaces in oligomeric proteins which are permanent assemblies (Jones and Thornton, 1995, 1996; Ford et al., 1998). The average homodimer in the sample Ê 2 , and the average analyzed by Jones and Thornton (1996) buries 3370 A 2 Ê antigen±antibody complex buries 1550 A (note that these authors quote interface areas per subunit, which is B/2 in our notation). The assembly of oligomeric proteins is usually accompanied by major conformational changes and/or disorder-to-order transitions. These events also accompany the formation of protein±protein complexes with large interfaces.

38

È L JANIN SHOSHANA J. WODAK AND JOE

Thus, hirudin, as shown in Fig. 9, undergoes a disorder-to-order transiÊ 2 interface with thrombin, as NMR studies of tion when forming a 3300-A the free protein indicate that the C-terminal tail is disordered in solution. The structure of free ornithodorin is not known, but the two PTI-like domains are unlikely to adopt the same relative position as in Fig. 9. Conformational changes induced by association are of major signi®cance in G proteins which are implicated in cellular responses to signals originating at the cell membrane. In the retina rod cell, association± dissociation of transducin Ga with Gbg carries the signal from the photoreceptor rhodopsin in the membrane to the phosphodiesterase that hydrolyzes the chemical mediator cyclic GMP. Two regions of Ga are involved in the two surface patches that contact Gbg . The N-terminal ahelix, which contributes the lateral patch, is disordered in free Ga and it undergoes a disorder-to-order transition on association. The central patch involves a pair of loops located near the GTPase active site; the loops are denoted Switch I and Switch II because they change their conformation as GTP is hydrolyzed and GDP released (Noel et al., 1993). On Gbg binding, the loops undergo a different and even larger conformational change (Lambright et al., 1994, 1996) which completely remodels the protein surface. Less dramatic changes take place on the Gbg side of the interface, yet a comparison with the free structure shows Ê (Sondek et al., 1996). some loop movements of up to 6 A Like the trypsinogen±PTI complex cited above, the Ga Gbg complex has a relatively low af®nity in solution despite its large interface (it may, however, be more stable when the components are attached to the membrane). These two examples, and many others, show that there is no simple relationship between af®nity and the size of the interface, largely for the reason that large conformational changes accompany the formation of large interfaces. D. Classi®cation of Protein±DNA Complexes The picture concerning the different types of protein±DNA complexes is much less clear. One recent attempt at classifying these complexes describes a total of three classes on the basis of the pattern of DNA backbone and base contacts ( Jones et al., 1999), as illustrated in Fig. 10. These classes were denoted ``single-headed,'' ``double-headed,'' and ``enveloping.'' The single-headed are de®ned as those with a single cluster of residues contacting both the DNA backbone and bases. The doubleheaded feature two distinct clusters of residues contacting the DNA backbone and a number of unclustered base-contacting residues, whereas the enveloping ones are a distinct class in which a large protein cleft forms extensive contacts with the DNA backbone. It was

STRUCTURAL BASIS OF MACROMOLECULAR RECOGNITION

39

FIG. 10. Protein and DNA footprints (Jones et al., 1999). The protein footprint differentiates between the residues contacting the sugar±phosphate backbone and those contacting the bases. Protein residues that make no contact with the DNA are colored blue. Those connecting the sugar±phosphate backbone are colored red, and those making base contacts are colored yellow. (a) Proteins with a single binding head; (b) proteins with a double binding head; and (c) proteins with an envelope mode of binding.

40

È L JANIN SHOSHANA J. WODAK AND JOE

noted that the enveloping mode is predominantly exhibited by DNAbinding enzymes, whereas transcription factors bind in a single- or double-headed fashion. Nadassy et al. (1999) prefer to describe the same complexes in terms of the interaction modules mentioned above. These modules can occur in isolation or more often as pairs, forming, respectively, the single- and double-headed complexes of Jones et al. (1999). Thus DNA complexes with transcription factors can all be described in terms of the binding modules. This includes those of prokaryotic origin, containing the classical helix-turn-helix motif, the homeodomains, various types of leucine zippers, hormone receptor-type zinc modules, NF-kB, and more. The presence of binding modules can also be recognized in complexes with the TATA box-binding protein (TBP). This single-chain transcription factor protein, classi®ed by Jones et al. (1999) as belonging to the enveloping category, has in fact an internal duplication not seen in the sequence (Nikolov and Burley, 1994). Each half is the size of a standard interaction module, but there is no evidence that a half-TBP can be isolated and that it would bind DNA. The much larger interfaces belonging to the enveloping category and formed in complexes with enzymes could also be described in terms of interaction modules. But those would be larger and bury more surface area in the complex than the transcription factor modules (Nadassy et al., 1999). For example, DNA polymerases are multidomain proteins shaped like a hand. At least three domains, called ``®ngers,'' ``palm,'' and ``thumb,'' contact DNA, with the palm domain carrying the polymerase active site (Ollis et al., 1985a,b). As in TBP, each domain could be considered as part of a different interaction module, although, here too, it is unlikely that this module is stable and functional on its own. Attempts to use other features to classify the interfaces, such as the types of structural motifs of the protein (helix, beta, helix-turn-helix) or of the DNA (minor or major groove) involved in binding, have not yielded clear patterns thus far. More data on the structures of protein±DNA complexes are clearly needed in order to provide a better basis for performing classi®cations of the different types of protein±DNA complexes.

IV. ENERGETICS OF MACROMOLECULAR RECOGNITION A. Protein±Protein Recognition 1. Thermodynamic Parameters Characterizing the forces that stabilize protein±protein interactions and understanding how these forces operate to yield complexes with

STRUCTURAL BASIS OF MACROMOLECULAR RECOGNITION

41

widely different stability ranges whose formation can be regulated by different exogenous factors are of considerable biological importance. Over the past 10 years, signi®cant progress has been achieved in this regard. A series of careful experimental studies, some of which are reported in other chapters of this book, has produced values for the thermodynamic parameters for complex formation. The most important parameters are DGd , the change in the standard state free enthalpy (or Gibbs energy) of the system on dissociation, its enthalpy component DHd , and its entropy component DSd : DGd DHd TDSd :

(4)

The value of DGd is commonly derived from that of the equilibrium constant Kd by applying the classical relation: DGd

RT ln

Kd : c0

(5)

c0 is the concentration de®ning the standard state; by convention, c0 1 mol L 1 . If Kd has been measured as a function of temperature, DHd and DSd can be derived from its temperature dependence by applying the Vant'Hoff law. In recent years, isothermal titration microcalorimetry (ITC) has also been used for that purpose. As reviewed by Fisher and Singh (1995) and Ladbury and Chowdhry (1996), microcalorimetry gives direct access to DHd and, also, if the titration is performed at several temperatures, to the heat capacity change DCd at constant pressure. The ®rst quantity is proportional to the heat evolved on mixing known amounts of the two components of a complex, and the latter is equal to DCd

d(DHd ) d(DSd ) T : dT dT

(6)

Theoretical analysis aims to rationalize the value of these thermodynamic parameters in terms of the different physical forces involved, at least qualitatively (Gilson et al., 1997). On the other hand, site-directed mutagenesis and related techniques provide useful estimates of the effect of perturbations, typically by following the removal of individual chemical groups at the interface. Free enthalpy changes can easily be estimated by comparing the dissociation constant of the wild-type complex Kd with that of the variant Kd (which is usually larger than Kd ): DDGd 1

RT ln Kd =Kd :

(7)

An accuracy of 0.2 kcal mol can easily be achieved for DDGd, whereas DDHd values derived from ITC measurements on the wild-type and mutant complexes have a larger error.

42

È L JANIN SHOSHANA J. WODAK AND JOE

2. Hydrophobic vs Polar Interactions As for protein folding, the physicochemical basis for the stability of protein complexes results from a delicate balance between several contributions. The formation of a complex involves the dehydration of the protein groups at the interface. Dehydrating aliphatic and other nonpolar groups has a thermodynamically favorable entropy and free enthalpy, at least in the temperature range 5±1008C (Makhatadze and Privalov, 1996). Nonpolar dehydration, generally ascribed to the hydrophobic effect, is therefore a factor stabilizing the complex. Dehydration is accompanied by the formation of van der Waals and electrostatic interactions between protein groups at the interface. This by itself should be stabilizing, but the same groups also make nonbonded interactions with the solvent in the free proteins. The net contribution of the bonds is therefore a balance between the protein±solvent and protein±protein interactions. In general, van der Waals interactions are believed to be neutral or slightly stabilizing, because protein atoms pack more densely at interfaces than water molecules do in liquid water. In contrast, electrostatic interactions are often destabilizing due to the high cost of dehydrating charged groups. These concepts are illustrated in Fig. 11, which shows the energetics of a protein antigen, hen lysozyme, binding to the Fab fragment of the HyHel5 monoclonal antibody and of an inhibitor, barstar, binding to an enzyme, barnase. The two complexes have interfaces of a similar (standard) size, and their formation involves small conformational changes. Both interfaces are about 50% nonpolar and contain 11±13 hydrogen bonds, but barnase±barstar has more charged groups making these bonds. This may be one reason that the ®rst complex is more stable, its Kd being 10 13 M and that of HyHel5±lysozyme 10 10 M. Both systems have been studied by isothermal titration calorimetry. For barnase±barstar at 258C, the enthalpy and free enthalpy changes are almost equal (Frisch et al., 1997) and the entropy change DSd is 0 (in the 1 M standard state, which has no practical relevance). For HyHel5±lysozyme, DHd is signi®cantly larger than DGd (Hibbits et al., 1994) and the standard state entropy change is unfavorable. Therefore, both complexes are energy-driven near 258C and more so for HyHel5±lysozyme, with the reservation that the large heat capacity change DCd (near 0.35 kcal mol 1 K 1 in both cases) makes DHd and DSd strongly temperature dependent. These remarks apply to many other protein±protein complexes. The DH and DG values of Fig. 11 may be compared with these data. They were derived from a relatively unsophisticated attempt to reproduce in the computer the energetics of the HyHel5±lysozyme and barnase±barstar systems, based on the atomic coordinates of the two complexes

STRUCTURAL BASIS OF MACROMOLECULAR RECOGNITION

43

200

kcal / mol

100

− T∆S gas polar elec vdw non-polar

0

−100

∆H

∆G

−200 HyHEL5lysozyme

barnasebarstar

HyHEL5lysozyme

barnasebarstar

FIG. 11. Energetics of an antigen±antibody complex and an enzyme±inhibitor complex. The complexes are Fab HyHEL5±lysozyme [2h¯ (Sheriff et al., 1987) ] and barnase±barstar [1bgs (Guillet et al., 1993) ]. Their dissociation enthalpy DH and free enthalpy DG are calculated as the sum of ®ve terms: the energy of van der Waals (vdw ) and electrostatic (elec ) interactions between the two molecules; the hydration enthalpy or free enthalpy of nonpolar and of polar groups at the interface; and a gas-phase entropy change ( TDS gas). The vdw and elec terms are obtained by energy minimization in gas phase with the dielectric constant adjusted to e 3. The hydration terms are proportional to the contribution of nonpolar and of polar groups to the interface area. The gas-phase entropy change includes contributions of the external and internal degrees of freedom. HyEL5±lysozyme Exp

Calc

DHd (kcal mol )

22.6

19

DGd (kcal mol 1 ) DCp (kcal mol 1 K 1 )

14.5 0.34

8

1

Barnase±barstar Exp

Calc

19

82

19 0.38

45

Calculated values are taken from Janin (1995); experimental data at 258C are from Hibbits et al. (1994), and Frisch et al. (1997).

(Janin,1995). The procedure involves two steps. The ®rst step assembles the complexes in the gas phase; in the second step, the system is transferred into aqueous solution and hydrated. In the gas phase, van der Waals and electrostatic interactions are formed. Their energy, evaluated

44

È L JANIN SHOSHANA J. WODAK AND JOE

by molecular mechanics, is stabilizing and large. There is also a gas-phase entropy change, which favors dissociation (Finkelstein and Janin, 1989). It includes contributions of the external degrees of freedom (molecular translation/rotation), of amino acid side chains immobilized in the complex, and of internal vibrations. For the hydration step, enthalpy and free enthalpy changes are obtained by a method pioneered by Privalov and collaborators (see Privalov and Gill, 1988; Makhatadze and Privalov, 1996). It uses empirical hydration coef®cients derived from calorimetric studies of model small molecules, which are applied to protein groups in proportion of their contribution to the interface area. For HyHel5±lysozyme, the calculated enthalpy and free enthalpy changes are in good agreement with the experimental values at 258C. Eelec , the energy of electrostatic interactions in the gas phase, is largeÐ 66 kcal mol 1 , taking the dielectric constant to be e 3 inside the protein. Yet, Eelec is less than the cost of dehydrating polar groupsÐ91 kcal mol 1 , estimated with the parameters of Oobatake and Ooi (1993). On the other hand, van der Waals interactions have a net favorable enthalpy, and dehydrating nonpolar groups, while energetically unfavorable, has a favorable free enthalpy. Overall, the hydrophobic effect and van der Waals interactions are stabilizing, whereas electrostatic interactions and the loss of degrees of freedom on association are destabilizing. With barnase±barstar, the value of Eelec calculated in the gas phase exceeds the enthalpy of polar group dehydration, and the calculation predicts electrostatic interactions to be stabilizing, while other terms are about the same as for HyHel5±lysozyme. However, the calculated DHd and DGd very much exceed the experimental values. Eelec is highly dependent on the assumed value of the dielectric constant, and it has a very large error bar. More accurate estimations use continuum electrostatics models (Honig and Nicholls, 1995) to evaluate both coulombic and dehydration energies in an aqueous environment. These methods, which we discuss below, should be used to decide whether the balance of electrostatics actually favors association in this system. 3. Electrostatics of Protein±Protein Recognition Charged groups are less frequent on average at protein±protein interfaces than elsewhere on the protein surface, but they are much more abundant than inside proteins. Thus, ionic interactions and the dehydration of charged groups may sometimes be ignored in folding studies of small proteins at least. They certainly cannot be ignored at interfaces. The electrostatic aspects of protein±protein interaction have been the subject of many theoretical studies in recent years (Froloff et al., 1997; Reddy et al., 1998; Hendsch and Tidor, 1999; Schapira et al., 1999; and for a more recent review, see Sheinerman et al., 2000). Most suggest that, due to the cost of

STRUCTURAL BASIS OF MACROMOLECULAR RECOGNITION

45

dehydration, ion pairs and H-bonds do not actually provide stability to the complex. However, because any mispairing of electric charges or of donor± acceptor H-bonding couples adds to that cost, the net effect of having many polar groups at protein±protein interfaces is to select one mode of association among many others with comparable characteristics. It is generally believed that polar interactions play a major role in determining the speci®city of the recognition, and therefore, speci®city is achieved in part at the expense of stability. This ambiguous role of polar interactions is best illustrated by mutagenesis studies. A point substitution which deletes a charged group involved in an ion pair or a H-bond in an interface often appears as a hot spot (Clackson and Wells, 1995). The mutation may destabilize the complex far beyond the net contribution of the ion pair or H-bond to DGd , because removing a polar group on one side of an interface leaves another charged or polar group buried and unpaired on the opposite side. The net contribution can be recovered by also removing the partner group in a second mutation and performing a ``double mutant cycle'' (Carter et al., 1984; Horovitz, 1987). Other chapters in this book illustrate this procedure. Although electrostatics are destabilizing overall, there are local examples of polar interactions that continuum electrostatics predicts to favor complex formation. For example, a theoretical analysis of the dimeric GCN4 leucine zipper (Hendsch and Tidor, 1999) locates ion pairs that contribute favorably to dimer stability. Moreover, electrostatic interactions within a subunit may be enhanced in the dimer, due to shielding from the solvent. A survey of protein±protein complexes (Xu et al., 1997) reports a positive correlation between the number of ion pairs at an interface and the binding af®nity. It also suggests that the stabilizing effect involves networks of electrostatic interactions, as individual charged groups often form multiple interactions across the interface. Networks of this type have been proposed to contribute to the enhanced thermal stability of hyperthermophilic proteins (Xiao and Honig, 1999). 4. Electrostatics and Association Kinetics Electrostatic interactions have a much longer range than other interactions and exert attractive (or repulsive) forces over distances comparable to the size of protein molecules. The range is shorter at high ionic strengths, but, at medium and low ionic strength, electrostatic interactions can modulate the rate of collision between two (macro)molecules bearing net electric charges or dipoles (Berg and von Hippel, 1985; Sheinerman et al., 2000). The coulombic attraction between charges of opposite sign makes collisions more frequent, and the repulsion between charges of the same sign makes collisions less frequent. Charge±dipole interactions exert a torque and its steering effect may favor or disfavor productive

46

È L JANIN SHOSHANA J. WODAK AND JOE

collisions. In either case, the rate of complex formation, de®ned by the bimolecular rate constant of association ka , is affected. It becomes ionic strength dependent (Janin, 1997; Camacho et al., 1999; Gabdoulline and Wade, 1999) and sensitive to point mutations that change the net charge. In this case, the mutation site can be anywhere on the protein surface, whereas most mutations outside the interface have a negligible effect and those at the interface affect only the rate of dissociation. Barnase±barstar is a case of electrostatically enhanced protein±protein association, being much faster and more sensitive to ionic strength than, say, trypsin±PTI or antibody±lysozyme association (Schreiber and Fersht, 1996). The last two proceed with ka in the range 105 106 M 1 s 1 . This is the expected range if we assume that the two protein components diffuse and collide freely in solution, but form a complex only if they happen to be in the right orientation when they collide (Northrup and Erickson, 1992; Janin, 1997). Electrostatic steering is unimportant in antibody± lysozyme association, and it possibly makes PTI±trypsin association slower, as the two proteins carry a net positive charge. In contrast, barnase has a positive net change and barstar a negative net charge. Their rate of association ka is in the same range as that for trypsin±PTI at very high ionic strength; it is much faster, on the order of 1010 M 1 s 1 , at medium and low ionic strength. Moreover, ka increases when barnase is made more positive or barstar more negative by point mutations (Schreiber and Fersht, 1996). Brownian dynamics simulations (Gabdoulline and Wade, 1997) or the application of a modi®ed Debye±Hu È ckel model of coulombic interactions (Selzer et al., 2000) reproduces these variations quantitatively (as seen below). In the barnase±barstar system, there is a selective pressure on the kinetics of association, the ribonuclease activity being lethal if expressed in the cell. The genes form an operon and, when both proteins are produced together in the bacterium, barnase must be either excreted or immediately inhibited by barstar. Unlike the selection for tight binding, which operates almost exclusively on residues at the interface, the selection for fast binding acts on all charged residues of the two proteins. B. Protein±DNA Recognition The energetics of protein±DNA recognition and protein±nucleic acid recognition in general are less well understood than those of protein± protein complexes. The association process in these systems generally involves large conformational changes and/or order±disorder transitions that affect both the protein and nucleic acid moieties. It furthermore often involves protein±protein association in addition to protein±DNA interactions. A number of theoretical studies have been devoted to the analysis

STRUCTURAL BASIS OF MACROMOLECULAR RECOGNITION

47

of the electrostatic contributions to the binding free energy in DNA±transcription factor complexes (Misra et al., 1994a,b; 1998; Fogolari et al., 1997), which are among the best studied systems of protein±DNA complexes. The protein±DNA complex observed crystallographically is represented in full atomic detail, whereas the solvent is a continuum with a single dielectric constant which depends on ionic strength. The dif®cult issues involved in dealing with protein folding, protein±protein interactions, and conformational changes that usually occur on protein±DNA complex formation are rarely addressed, for obvious reasons. The prevailing picture provided by these studies [see for example, Misra et al. (1998) ] is that the net electrostatic contribution to protein± DNA association in these systems opposes binding at physiological ionic strength due to the very high cost of burying charged groups, mostly the phosphates of DNA, and other polar groups on the protein. This cost exceeds the favorable contribution of the positively charged protein groups neutralizing the charges on DNA and canceling the long-range coulombic forces arising within the highly charged nucleic acid. Water molecules and cations, bound speci®cally or not at the protein±DNA interface, and the networks of solvent-mediated protein±nucleic acid contacts that they form certainly also contribute [see, for example, Ladbury et al. (1994) and Shakked et al. 1994) ], but these contributions are even more dif®cult to evaluate. This notwithstanding, the general conclusion reached from these studies is that the main driving force for protein±DNA association is most likely nonpolar in nature and similar in kind to that which stabilizes protein±protein interactions. Finally, fast association plays a key role in protein±DNA recognition, and electrostatically enhanced association is a more general phenomenon than for protein±protein recognition. Nevertheless, a detailed analysis of the lac repressor±lac operator system [see von Hippel and Berg (1989), for a review] indicates that attractive electrostatic forces do not fully account for the very high observed value of ka , on the order of 5 1010 M 1 s 1 . Moreover, they do not explain the dependence of the rate of association on the size of the DNA molecule when the lac operator is inserted in fragments of various lengths. Additional effects, such as onedimensional diffusion along DNA and an increased capture radius due to the partial unfolding of the protein prior to binding, have therefore been proposed (Shoemaker et al., 2000). C. Statistical Mechanics of Speci®city Speci®city is much more dif®cult to characterize than af®nity, which the value of the equilibrium constant Kd or the standard state free enthalpy DGd suf®ces to quantify. To test speci®c binding by a protein, a protease

48

È L JANIN SHOSHANA J. WODAK AND JOE

or an antibody, for instance, one often compares its af®nity for two closely related ligands, such as the wild-type and a point mutant of the inhibitor or the protein antigen, by measuring changes in Kd or differences in DGd (DDGd ). This type of study, illustrated in other chapters of this book, is undoubtedly of great value as it details the role of individual side chains in discriminating between ligands. However, as an approach to speci®city in general, it poorly describes situations that prevail in vivo. In the immune system, the cognate epitope competes for binding to an antibody, not against a slightly modi®ed copy of itself, but against the extremely heterogeneous mixture of molecular surfaces that are present in the serum. These non-epitope surfaces may all have a very weak af®nity for the antibody, but they occur in very large number and their transitory interaction with the antibody contributes to the background that hinders the ef®cient recognition of the speci®c epitope. These situations can be analyzed with the methods of statistical physics by enumerating all possible modes of association and listing their energies. The simplest model, with only two states separated by an energy gap D, represents the discrimination between two ligands competing for the same site, and the energy gap is equivalent to DDGd in mutant studies. A more realistic situation has many states with different energies. In the Random Energy Model, originally developed to describe complex physical systems such as spin glasses, the states form a quasi-continuous energy spectrum and their statistical distribution is described by the spectral density function m(E) and the related entropy function: S(E) R ln m(E):

(8)

The Random Energy Model has been applied to protein folding by Bryngelson and Wolynes (1987) and was later extended to protein± ligand recognition (Janin, 1996). In folding studies, any conformation of the polypeptide chain is a state and E its conformational energy. In protein±ligand recognition, any mode of binding is a state. At thermodynamic equilibrium, the modes are distributed following Boltzmann's law. Nonnative modes may be complexes with nonspeci®c ligands or complexes where the ligand binds outside the speci®c site or in an incorrect orientation. Figure 12A describes the distribution of nonnative complexes generated in a computer simulation where hen lysozyme is docked in an arbitrary orientation onto the combining site of Fab HELHy5. We attribute to each docking solution an energy E linearly related to its interface area B. All possible orientations are tested in order to derive the spectral density function m(E) and the corresponding entropy. Most arti®cial complexes generated by docking have the combining site of the antibody contacting regions of the lysozyme surface other than the speci®c epitope. Thus, they represent the sort of weak

STRUCTURAL BASIS OF MACROMOLECULAR RECOGNITION

B

2HFL Lysozyme-antibody complex

106

0

0

S = R ln m(E)

Tc

Native

Number of complexes

A

49

Ts

0

500 1000 1500 2 Interface area B (A )

0

∆ Energy of interaction E

100

FIG. 12. Energy spectrum of a protein±protein complex. The association of hen lysozyme with Fab HyHEL5 is simulated by docking in the computer the antigen against the combining site of the antibody. The docking is performed 8:8 106 times to sample all degrees of translational/rotational freedom of the molecule. The native mode of association is obtained by docking in the orientation found in the X-ray structure of the complex (Sheriff et al., 1987; Cher®l et al., 1991). (A) Histogram of the interface area B achieved on docking. The interface area is the solvent accessible surface area lost by both partners on association. (B) The entropy±energy curve (heavy line) derived from the histogram in (A) by assuming the nonbonded energy of interaction E to be a linear function of B. The native complex was taken to have zero energy by convention, and all other docked complexes are nonnative and their energy E is above the energy gap D. The maximum value of E is D0 taken as that for which B 0. The number of docked complexes with energy in the range (E, E dE) is m(E) dE. For clarity, the energy gap D has been set to a much larger relative value than the docking simulation can actually achieve. The two tangent lines (dashed lines) de®ne transition temperatures Ts and Tc . Figure reproduced with permission from Janin (1996).

interactions the antibody can make as it diffuses and collides with the many different proteins present in serum under physiological conditions. With few exceptions, these complexes have a small interface area B and, therefore, a high energy compared to the native. Nevertheless, the population of nonnative modes of binding starts increasing to a large degree as soon as the temperature reaches a certain Tc value, the spin glass transition temperature of the system. Speci®city is lost completely when they become more abundant than the native state. This takes place above the transition temperature, Ts , given by the slope of the tangent from the origin to the S(E) curve (Fig. 12B). More generally, the ratio r of the number of nonnative to native complexes at equilibrium is equal to the partition function: r

Z1 D

m(E) exp

E dE RT

Z1 D

exp

E

TS dE: RT

(9)

50

È L JANIN SHOSHANA J. WODAK AND JOE

r is a measure of speci®city in a given environment. Its value depends on the shape of the energy spectrum rather than just the gap D as it would in a twostate model. Indeed, the nonnative modes that compete with the native mode are not those of low energy, which are few, but those of low free energy E-TS. The composition of the solution, in other words, the concentration of all chemical species in it, affects the energy spectrum. We may, for instance, consider the ratio r as the noise-to-signal ratio in an immunodetection experiment. We then expect r to increase when the solution becomes more heterogeneous, even if the total concentration of contaminating species is the same, just because the m(E) distribution becomes wider. This analysis can easily be extended to other recognition processes, for instance, to DNA±DNA recognition in DNA chips, where false-positives limit the sensitivity of detection of cognate sequences. V. COMPUTATIONAL APPROACHES FOR PREDICTING AND SIMULATING PROTEIN±PROTEIN INTERACTION Computational approaches for predicting how a ligand binds to a receptor protein are often referred to as the docking problem by analogy to the folding problem, which is the prediction of three-dimensional structures from amino acid sequences. The problem of docking a protein, rather than a small ligand, onto another protein was ®rst considered by Wodak and Janin (1978). It has attracted renewed interest in recent years and is becoming a focus of attention in the postgenomics era. In this review, we discuss protein±protein docking methods and algorithms, which aim to ®nd the stablest mode of association. Then, we brie¯y summarize methods aimed at simulating the process of protein±protein association. A. Docking Procedures Docking procedures start from the atomic coordinates of two proteins, generate putative complexes, and give them a score. The docking problem is most commonly formulated as a search for complementary modes of association between two preformed molecules (Wodak and Janin, 1978; Kuntz et al., 1982). With small molecules, internal degrees of freedom can be taken into account (at considerable expense in terms of computation), but, to date, all protein±protein docking procedures assume rigid-body association. However, the two molecules must be ``soft,'' in order to allow for the small conformational changes and side chain movements that take place in all complexes (Jiang and Kim, 1991). An essential test of a docking procedure is its capacity to handle ``unbound'' molecules, that is, atomic coordinates issued from the crystal structure of the component

STRUCTURAL BASIS OF MACROMOLECULAR RECOGNITION

51

proteins, as well as those of ``bound'' molecules, taken from the crystal structure of the complex itself. The latter are biased toward the native solution, and algorithms that retrieve the native complex only from the bound components obviously have no predictive value. Within the rigid-body approximation, the problem has only six degrees of freedom. Nevertheless, it is of considerable complexity. AsÊ radius on the surface of suming that each protein is a sphere with a 15-A Ê grid, a systematic which atomic scale surface features are drawn on a 1-A 9 search requires probing approximately 10 distinct association modes (Connolly, 1986). This may be feasible with present-day computers for a single pair of proteins, but not on the genomic scale. Surface side chains that rearrange on complex formation introduce additional degrees of freedom. These can be handled by molecular mechanics force ®elds based on detailed atomic models, but the cost of the computation is high, and most docking algorithms resort to some form of data reduction or simplifying representations. 1. Docking Simpli®ed Protein Models The earliest protein docking algorithm (Wodak and Janin, 1978; Janin and Wodak, 1985) mapped the rigid-body search space into ®ve rotational degrees of freedom plus a translation designed to bring the two molecules in direct contact once their orientation has been ®xed (Fig. 13A). All possible orientation modes are tested and given a score based on the surface area buried between the two molecules in contact. As the area calculation is expensive, Wodak and Janin (1980) developed an analytical approximation based on the simpli®ed protein model of Levitt (1976). In this model, each amino acid residue is replaced by a sphere of appropriate radius and a soft repulsive residue-pair potential allows a limited interpenetration of the spheres. The simpli®ed model decreases the number of interaction centers by a factor of about 7, and it makes the energy landscape smoother, enabling soft docking. The analytical approximation authorizes the computation of the derivatives of the interface area relative to the rigid-body degrees of freedom for mathematical optimization. The small number of interaction centers and the ef®cient calculation of a surface area-based score render the procedure relatively fast. Moreover, it performs similarly on the unbound and bound structures of the components, due to the simpli®ed model removing atomic details such as side chain conformations. The algorithm was initially applied to docking PTI on the active side of trypsin, yielding a native-like mode of association among the top 12 solutions (Wodak and Janin, 1978). In ®ve of six antibody±lysozyme and protease±inhibitor complexes, nativelike complexes were retrieved among the 10 highest scoring solutions by Monte Carlo simulated

52

È L JANIN SHOSHANA J. WODAK AND JOE

annealing (Fig. 13B). As the native complex could not be singled out from false-positives using the simpli®ed model, full-atom models of the highranking complexes were generated and subjected to a classical molecular mechanics re®nement procedure. This improved their geometry by removing clashes and fostering the formation of H-bonds, but the re®ned energy was not useful as a criterion for discriminating the native mode of association from the false-positives (Cher®ls et al., 1991). On the other hand, the method provided an ef®cient way to probe the interaction of the two ab dimers of hemoglobin and draw a reaction path between the two allosteric R and T forms of the tetramer (Janin and Wodak, 1985). This application of docking has not been explored using other algorithms. 2. Shape Complementarity The Wodak±Janin procedure entirely ignores the chemical nature of the protein surfaces in contact and searches only for large surface patches with complementary shapes. Many other docking procedures are also based on shape complementarity. The two molecular surfaces can be described by sets of critical points de®ned as ``knobs and holes'' (Lee and Rose, 1985; Connolly, 1986; Zachmann et al., 1992). Trial docking solutions are derived from a combinatorial search for groups of matching critical points. The surface complementarity approach was pioneered by Connolly (1985, 1986), who identi®ed knobs and holes by triangulating the surface of the two component molecules and searched for quartets of complementary knob/hole pairs. Each quartet ®xes a relative orientation of the molecules and yields a candidate solution after steric overlaps are searched for. When the solutions were ranked according to their interface area, the algorithm successfully predicted the native mode of association of the a and b subunits of hemoglobin, but not that of trypsin±PTI. Wang (1991) used a modi®cation of this method in which each knob/hole pair yields a starting orientation, which is locally optimized. The scoring function, evaluated on a grid as in other procedures described below, measures the overlap between a double layer of surface points sampled on one molecule and the interior of the other molecule. Wang's algorithm was successful on trypsin±PTI as well as on the ab hemoglobin dimer. The program DOCK of Kuntz and colleagues (see Shoichet et al., 1992) de®nes shape complementarity in a very different way. The program was developed for docking small ligands into a known binding site on a receptor protein. Candidate solutions are generated by matching a reduced representation of the ligand to a ``negative image'' of the receptor-binding site. Both representations are generated by constructing overlapping spheres that are in contact with Connolly's molecular dot surfaces. On the protein, the spheres are placed on the exterior of the surface; on the ligand, they are placed on the interior. Docking solutions are generated

STRUCTURAL BASIS OF MACROMOLECULAR RECOGNITION

A

Y2

X2

O2

53

Y1

Z2 θ2 ϕ2

ρ

χ θ1

ϕ1

X1

O1 Z1

ξ

B A C

A B

C

Docking solutions

6000

4000

2000

0

0 −4 −8 −12 −17 −21 −25 −29 −33 −37 −42 −46 −50 −54

FIG. 13. Rigid docking procedure with simpli®ed protein model of Wodak and Janin (1978). (A) The polar coordinates used to describe the relative orientations of the two docking molecules. (B) Sketch of the docking procedure: Molecule 2, drawn as a single sphere, moves along the line (dashed line) between the two molecules; as it approaches molecule 1 from the right, it ®rst touches atom B, but must move further to the left to reach its correct docking position corresponding to translation j. (C) Histogram of the Fab HyHEL5±lysozyme docking solutions computed by Cher®ls et al. (1991) as the function of the scoring function EB. A total of 30,784 orientations of lysozyme were generated, and at each of these orientations lysozyme was docked into the combining site of Fab HyHEL5. The arrow points to the complex reconstituted in the native orientation.

54

È L JANIN SHOSHANA J. WODAK AND JOE

by matching sets (cliques) of spheres belonging to the ligand and the protein, with four or more matching pairs de®ning a rigid-body transformation. Version DOCK2 of the algorithm was adapted to docking protein inhibitors onto an enzyme or receptor recognition site (Shoichet and Kuntz, 1991; Shoichet et al., 1992). It groups the spheres into subclusters and bins pairwise distances in order to avoid the combinatorial explosion in matching groups of spheres. The scoring function favors close contacts and penalizes steric overlaps. It is evaluated on a grid constructed around the recognition site, with each receptor grid point being assigned a value according to its distance from ligand atoms. DOCK2 was applied to trypsin±PTI and three other enzyme±inhibitor complexes. A signi®cant proportion of the highest scoring solutions were Ê rms of the crystal structure, and the results were similar with within 2.5 A the unbound and bound components. A number of attempts were then made to re®ne the solutions and devise better criteria for identifying the correct solutions among the many false-positives. The continuum electrostatic calculations implemented in DELPHI (Gilson and Honig, 1988) gave the best result. Energy minimization with AMBER (Weiner et al., Ê 1984) reliably identi®ed low-energy orientations, but only to within 5 A rms from the crystallographic structures (Shoichet and Kuntz, 1991). 3. Geometric Hashing Computer vision algorithms based on ``geometric hashing'' extend the methods based on knobs and holes. Fischer et al. (1993) developed a docking method that uses an algorithm adapted from an object recognition technique in computer vision (Bachar et al., 1993) (Fig. 14a). As in the sphere-matching procedure described above, the relative orientation and translation of the interacting molecules are obtained by matching cliques of surface features. To that end, the molecular surfaces are processed in order to identify triplets of critical points (Fig. 14b), which are indexed by the distance between each pair in the triplet and stored in hash tables. In a ®rst version of the program, the critical points were constructed with the sphere generation module of DOCK (Norel et al., 1994). In its most recent form, the program uses surface ``caps and pits'' (Lin et al., 1994), calculated by a modi®cation of Connolly's algorithm, and docking is performed by matching cliques of ligand caps to receptor pits. A complementarity score, calculated for each orientation, counts overlaps Ê grid on the ligand surface, with receptor grid cells, of dots placed on a 1-A in a manner very similar to that used by Wang (1991). In this case, however, the receptor grid cells are assigned a type (surface, intermediate, interior) and the score is weighted according to the grid cell type (Fischer et al., 1995). The method proved to be very ef®cient starting

STRUCTURAL BASIS OF MACROMOLECULAR RECOGNITION

Caps and pits

55

Belts

FIG. 14. Molecular surface representation using critical points (reproduced with permission of the authors) (Lin et al., 1994). (b) Critical points of the molecular surface of the heme, connected in a triangle mesh: White, caps; blue, pits; red belts.

56

È L JANIN SHOSHANA J. WODAK AND JOE

Ê ) were from bound molecules. Nativelike solutions with a low rms (1.5 A retrieved for each of the 19 complexes tested. In 7 cases, the correct Ê ) had the highest score, and in all but one, it solution (to within 1.5 A ranked within the top 150 solutions. However, the results were signi®cantly poorer when the calculation was applied to the unbound species. The correct solution ranked 1497 with HyHEL5±lysozyme and 4217 with the trypsin±PTI complex, indicating that the method is highly sensitive to small changes in surface features. 4. Docking by FFT Among the methods for de®ning surface complementarity, that of Jiang and Kim (1991) is one of the simplest. A cubic grid is drawn, and each grid point xi is given a weight wA , which is negative and large if the point is inside protein A, equal to zero if it is outside, and equal to 1 if it is near its surface. The same scheme is used for protein B, and the score attributed to a candidate solution is then the sum over all grid points of the products of the two weights. The product is large and positive (unfavorable) where the two molecular volumes overlap and negative and favorable at grid points which belong to the surface of one molecule and the volume of the other. Katchalski-Katzir et al. (1992) noted that when molecule B is translated by u relative to A, this score can be written as a convolution product: P S(u) i wA (xi ) wB (xi u): (10)

Therefore, it can be very ef®ciently calculated by the Fast Fourier Transform (FFT) algorithm, provided u is sampled on the same grid as x (Press et al., 1992). Ef®cient bit handling algorithms are an alternative to the FFT (Palma et al., 2000). The grid must be rede®ned and the translation search repeated for each new orientation to perform sixdimensional docking searches. The FFT approach can accommodate more elaborate score functions than that just described. The weights wA and wB may contain information on surface properties such as hydrophobicity (Vakser and A¯alo, 1994) and on electrostatic and van der Waals interactions (Harrison et al., 1994; Blom and Sygusch, 1997). The weights may be complex numbers, with a real part coding for geometric properties and an imaginary part coding for chemical properties associated with the grid points (Chen and Weng, 2002). In the FTDOCK program developed by Gabb et al. (1997), diagrammed in Fig. 15, a simple electrostatic term, added to the correlation function representing geometric matching, was found to improve the rank of correct solutions when docking 10 enzyme±inhibitor or antigen±antibody complexes. Another advantage of the FFT method is that one can easily adjust the resolution at which the features of the

STRUCTURAL BASIS OF MACROMOLECULAR RECOGNITION

57

FIG. 15. Flow diagram of the Fourier correlation docking procedure of Gabb et al. (1997). (Figure reproduced with permission of the authors.)

58

È L JANIN SHOSHANA J. WODAK AND JOE

interacting molecules are considered by limiting the number of Fourier terms in the summation. Low-resolution searches are very fast, and they may be suf®cient for detecting shape complementary or locating a ligand near a receptor-binding site (Vakser, 1995, 1996). The low resolution blurs atomic details and the small conformational changes which are expected to occur when docking unbound molecules (Gabb et al., 1997). On the other hand, high-resolution searches are required for accurately de®ning the molecular positions and orientations. They must Ê in translation, less than 108 in rotation) and be done on a ®ne grid (1 A represent heavy calculations, possibly several days long when docking large molecules (Gabb et al., 1997). Various strategies have been proposed to make the calculation faster. Meyer et al. (1996) restrict the translation search to a subset of the orientations that have the correct geometry to form two or more interfacial H-bonds, in a manner analogous to the matching pairs of critical points in the procedure of Fischer et al. (1995). This search is then repeated on a ®ner grid for those orientations that score within the top 75% in the ®rst pass. When 45 protein complexes of known structure Ê rms of were tested in this way, the highest ranking solution was within 3 A Ê native structure in each case and often within 1 A rms. This is an impressive achievement, but all calculations were done on bound molecules and it is not known how the method currently performs on unbound species. A second variant of the FFT method is the fast but less accurate algorithm of Ackermann et al. (1998). The molecular surfaces are divided into a small number (5 to 10) of convex or concave elements and pairs of complementary elements are matched, yielding a small set of starting orientations for the FFT search. When applied to 51 protein complexes of Ê known structure, the procedure produced at least one solution within 5 A rms of the native in 30 cases. Better accuracy could be obtained by performing a more exhaustive rotational search, but at a computational cost nearly two orders of magnitude higher. A third and very promising variant was recently proposed by Ritchie and Kemp (2000). It uses spherical polar Fourier expansion coef®cients to accelerate the search for low-energy solutions. In this approach, interaction energies are estimated from an excluded volume model derived from the notion of ``overlapping surface skins,'' augmented by a rigorous but soft model of electrostatic complementarity. Although there is no analogue to the FFT in a spherical polar representation, this approach has many advantages. Namely, a complete search over all six rigid-body degrees of freedom can be performed by rotating and translating only the initial expansion coef®cients. Infeasible orientations may be eliminated rapidly using only low-resolution terms, and the search can be restricted to surface regions near known binding sites when such know-

STRUCTURAL BASIS OF MACROMOLECULAR RECOGNITION

59

ledge is available. Execution times on single processors range from a few minutes to about 2 h for a global search (108 trial orientations). The results obtained with this method on a set of enzyme±inhibitor and antigen±antibody complexes were particularly encouraging. The native solution was frequently identi®ed when redocking bound species, and when starting from the unbound species, a native-like solution ranked within the top 20 in 11 of the 18 complexes tested. 5. Assessing Docking Procedure These recent successes, particularly those of Meyer et al. (1996), Ritchie and Kemp (2000), Palma et al. (2000), Gardiner et al. (2001), and Chen and Weng (2002) (see also the review by Halperin et al., 2002), demonstrate that it is possible to reassemble known protein complexes from their component structures. Although all these methods use the rigidbody approximation and coarse force-®elds for scoring candidate solutions, they require considerable computational resources and it is still not clear whether they are appropriate for predictive docking, which is after all the more biologically relevant problem. Attempts have been made to critically assess docking procedures by comparing their performance in blind trials in which the X-ray structure of the components is known and that of the complex is made available only at the time of evaluation (Strynadka et al., 1996; Dixon, 1997). The ®rst of these docking challenges aimed at deriving the complex of blactamase with the BLIP inhibitor. Six groups using different algorithms were able to propose nativelike solutions. The most accurate was the FFT correlation method of Kachalski-Katzir et al. (1992), which yielded a Ê rms from the crystal structure (Strynadka et al., model at only 1 A 1996). In the second challenge the task was to predict the complex between the trimeric ¯u hemagglutinin and a monoclonal antibody. This proved to be much more dif®cult: the best result, again from the Ê rms from the crystal structure, and the FFT approach, was at 9.5 A epitope was not correctly predicted (Vakser, 1997). The failure to predict the hemagglutinin±antibody complex is due in part to the large size of the hemagglutinin molecule, whereas the success in the case of the blactamase±BLIP derives from the presence of a large interface, which is more easily recognized than a standard size interface by algorithms searching for shape complementarity, and from the nature of the conformational change in BLIP: the saddle-shaped molecule bends by a few degrees on binding b-lactamase, which improves a preexisting complementarity. More blind trials are required to fully assess existing procedures. The CAPRI (Critical Assessment of PRedicted Interactions) experiment has been designed to organized such trials on a worldwide basis (see Web sites).

60

È L JANIN SHOSHANA J. WODAK AND JOE

B. Simulations of Protein±Protein Association The various procedures discussed above address the thermodynamic aspects of protein±protein interactions but say nothing about the dynamical aspects, such as the association kinetics and the mechanism. To gain insights into these latter aspects from computational methods, protein± protein association must actually be simulated at the molecular level. The most successful approach to such simulations has been Brownian dynamics (BD) (for recent reviews, see Gabdoulline and Wade, 1998, and Elcock et al., 2001). Brownian dynamics (BD) models diffusional systems in which the particles undergo Brownian motion. In such systems the particles, whose mass and size are larger than those of the solvent molecules, are subjected to stochastic collisions and to the viscous drag exerted by these molecules. This leads to the apparently random motion of the particles, which is diffusion, ®rst recorded in the 19th century by Brown. The technique of BD models these dynamic effects in an implicit fashion by combining the use of a diffusion coef®cient, suitably scaled for the molecules of interest, which accounts for solvent viscosity, with a random force component that models the effects of collisions with solvent molecules. The dynamics of the diffusional motion are described by the Langevin equation [for review see Madura et al. (1997) ], which can be solved in a number of ways. The solution used for protein±protein association is that of Ermak and McCammon (1978), R(t Dt) R(t)

DFDt S, kT

(11)

where R is the particle position, F is the systematic force acting on the particle due to the interactions with other particles, S is a random displacement vector, whose mean is zero, and Dt is the time step. The D/kT factor, which multiplies F, models the damping effect of solvent friction, with k being the Bolzmann constant and T the temperature. The algorithm provides a simple recipe for continually updating the positions of the diffusing particle so as to simulate its dynamical behavior. It is signi®cantly faster than classical molecular dynamics (MD) simulations. Due in part to the implicit representation of solvent molecules and to the use of much simpler force ®elds, much larger times steps (Dt) can be used than in classical MD simulations. These time steps are typically 1 ps (1 ps 10 12 s), representing a 1000-fold increase over the 1-fs (1fs 10 15 s) time step used in classical MD. The Ermak±McCammon algorithm, or variants thereof, has been extensively used for simulations of colloidal solutions using highly simpli®ed models for interparticle interactions [see, for example, Ravichandran

STRUCTURAL BASIS OF MACROMOLECULAR RECOGNITION

61

and Talbot (2000) ], and for simulations of enzyme±substrate and protein±protein interactions [see Madura et al. (1995) and Wade (1996) ], where both highly simpli®ed models and detailed atomic models have been employed. In simulations of protein±protein interactions, where the proteins are treated as rigid bodies, only two solute particles are considered. Translational motion is then simulated for one of the proteins relative to the position of the other protein, which is kept ®xed (Fig. 16). The displacement of the moving protein is then given by Eq. (11), with D replaced by the relative translation diffusional constant. Usually the effects of hydrodynamic interactions are not considered but can be treated by using tensors for the diffusion coef®cient instead of scalars. When detailed atomic models are used, intermolecular forces are computed as the sum of electrostatic and exclusion forces, while shortrange attractive forces such as hydrogen bonds and van der Waals interactions are not modeled, because they are less important for the diffusional process. In representing electrostatic interactions in BD simulations, models with varying levels of detail have been used. More

FIG. 16. Schematic illustration of the use of Brownian dynamics simulations for the calculation of binding rate constants. (Figure reproduced with permission from Elcock et al., 2001.)

62

È L JANIN SHOSHANA J. WODAK AND JOE

recently, it has become possible to assign partial atomic charges to all atoms in the proteins and to model solvation and ionic strength effects by solving the Poisson±Bolzmann equation numerically. This is, however, too computationally demanding to be applied at every time step of the BD simulations. Instead, the electrostatic potential is computed on a grid only for the ®xed protein, and the moving protein is represented as a collection of either test charges or effective charges (Gabdoulline and Wade, 1996). The intermolecular electrostatic forces are then computed by placing the array of charges of the moving protein on the potential grid of the ®xed protein. Computing rate constants from BD simulations (Madura et al., 1994; Wade, 1996) requires performing a series of BD simulations. In each simulation the moving protein is placed at a given distance b from the ®xed protein but in a randomly chosen direction in space and in a randomly chosen orientation relative to it (Fig. 16). The BD simulation is then run until the two proteins either associate or reach a much further degree of separation (q), at which point they are assumed to have escaped from each other. De®ning when association occurs is an important and as yet not completely resolved issue (Gabdoulline and Wade, 1999), and it has a large impact on the computed rates. In the context of BD, association means the formation of the encounter complex. The latter is viewed as the end-point of the diffusional association phase, from which the proteins can then rearrange into a tightly bound complex. Such tight binding cannot occur in the BD simulations because of the crude model used for representing the interactions. The best performance seems to be obtained when the criterion for association is based on the formation of contacts between certain pairs of atoms, known to be close to each other in the ®nal complex. This amounts to assuming that once the atoms achieve these contacts, subsequent association is ensured. By performing a large number of simulations, usually in the thousands, statistically robust estimates of the probability b that association will occur can be estimated. This probability can then be combined with the analytical expression of Smoluchowski (1917) for the steady state rate constant (kb ) of two proteins achieving the initial separation b, to give the overall rate constant for the association of the two proteins kass : kass kb b:

(12)

The ®rst applications of detailed atomic models in BD simulations of protein±protein association were used to simulate electron transfer rates in cytochrome systems (Northrup et al., 1988). More recently other systems studied comprise the association of the enzyme barnase with its protein inhibitor, barstar (Schreiber and Fersht, 1996; Gabdoulline and Wade, 1997) and that of the enzyme acetylcholinesterase with the snake

STRUCTURAL BASIS OF MACROMOLECULAR RECOGNITION

63

FIG. 17. Schematic illustration of the association of fasciculin (left) with acetylcholinesterase. (Figure reproduced with permission Elcock et al., 2001).

venom toxin fasciculin [Fig. 17, and references Radic et al. (1997) and Elcock et al. (1999) ]. Experiments have shown that in these systems association is a diffusion-limited process, which is dramatically accelerated by electrostatic interactions (see also Section IV,A,4). BD simulations in which the systematic forces acting on the proteins were assumed to be entirely electrostatic and described using the effective charge model were shown to successfully reproduce these results. In the barnase±barstar system, the simulations reproduced the ionic strength dependence of the association rate constants as well as their variation as the result of single and double mutations in the associating proteins. In the case of the acetylcholinesterase± fasciculin association, the simulations showed that accelerated association between the two proteins could occur even when electrostatic interactions do not strongly stabilize the bound complex. In other words, the kinetics can be accelerated even when binding thermodynamics is not signi®cantly improved. More recently, the same methodology was used to gain insight into why actin polymerizes faster at the socalled ``barbed'' end than at the ``pointed'' end of the ®laments (Sept et al., 1999).

64

È L JANIN SHOSHANA J. WODAK AND JOE

Interesting simulation work was also performed on the characterization of the factors responsible for fast protein±protein association in systems where electrostatic interactions are weak (Camacho et al., 2000). These simulations employed force ®elds which combine a surface areadependent desolvation term derived by Zhang et al. (1997) with a very simple approximation to the electrostatic interactions. The particularly noteworthy result was that the desolvation contribution appears to substantially increase the diffusion-limited association rate. This increase is obtained even though the forces averaged over the majority of the interaction geometries are repulsive, a fact which actually prevents nonspeci®c association. The simulations also identify the existence of weakly speci®c pathways leading to a few or possibly one low free-energy association geometry, representing the encounter complex. It is thus encouraging that BD is capable of dealing with a wide range of protein±protein interaction systems. So far, however, it has not been applied to predict the mode of association between two proteins, and this may be a useful direction in which this method could be further developed.

VI. CONCLUDING REMARKS The aim of this chapter was to present an overview of what is known today about the structural and energetic features that characterize protein±protein and protein±DNA associations. Without doubt our current understanding of the principles that govern the speci®city and af®nity of macromolecular recognition is still limited, and the performance of available structure-based theoretical approaches for predicting and simulating macromolecular recognition is far from satisfactory. Nevertheless, what we know and are able to do today should be most helpful in guiding the many efforts that are being launched world-wide for chartering the networks of protein interactions in living cells and for obtaining genome-scale information on the detailed atomic structures of macromolecular complexes. Already, docking procedures ®nd useful applications in building atomic models for large supramolecular complexes, starting from lowresolution electron density maps of the complex obtained from electron microscopy and high-resolution atomic coordinates of the individual components (Wriggers et al., 1999). Indicative of things to come, docking procedures are also being used, albeit in a very exploratory manner, to investigate the association between proteins whose structures have not been determined experimentally, but derived by homology modeling methods. Because modeled structures tend to be rather inaccurate, these exploratory calculations currently aim to roughly localize the surface regions of the protein, which are likely to interact with a given

STRUCTURAL BASIS OF MACROMOLECULAR RECOGNITION

65

partner. In the future, however, they might be applied to actually identify the residues involved in the association and at some point also to predict the 3D structure of a protein±protein complex. For this to be achieved in a reasonably reliable fashion, however, several hurdles need to be overcome. Docking procedures must be capable of modeling backbone and side chain conformational adjustments within practical time frames. With the many degrees of freedom of the polypeptide, this is a considerable challenge, but ways of meeting it are already being outlined. Considering that protein±protein association stabilizes conformational states already visited in the isolated molecular components [for a discussion, see Camacho et al. (1999) ], essential modes of dynamical backbone ¯uctuations in the isolated proteins (Amadei et al., 1993) can be computed and then used to model backbone adjustments during docking (Smith and Sternberg, 2002). A further hurdle that must be overcome is the incorporation of a better representation of the solvent contributions, including those of speci®c interactions with water molecules and counterions. As shown in this chapter, these latter types of interactions play a crucial role in speci®c protein±DNA recognition, and as the repertoire of known 3D structures of protein complexes grows, we may well ®nd them commonly mediating protein±protein association as well. Finally, another very useful approach for probing our understanding of recognition processes is the analysis and prediction of the effects of mutation on the af®nity and speci®city of known modes of protein interactions. This approach, made possible through the now classical protein-engineering techniques, has already provided valuable insights into many systems, as mentioned earlier in Section IV,A,3 and in other chapters of this book. More recently, novel techniques of so-called directed evolution, involving the synthesis of large combinatorial DNA libraries and powerful in vitro selection procedures (Hilvert, 2000), have signi®cantly expanded the scope of these studies. In the future, one would expect further progress by combining these approaches with computational procedures for selecting amino acid sequences likely to stabilize a given fold (Dahiyat and Mayo, 1997; Wernisch et al., 2000) or a speci®c interaction. All this holds greater promise than ever of improving our understanding of how evolution manages to ``design'' protein surface groups so as to minimize nonspeci®c interactions and aggregation, while favoring biologically relevant recognition. WEB SITES Protein±Protein Interaction Server (BSM, University College, London), http://www.biochem. ucl.ac.uk/bsm/PP/server.

66

È L JANIN SHOSHANA J. WODAK AND JOE

Protein±Nucleic Acid Interaction Server (BSM, University College, London), http:// www.biochem.ucl.ac.uk/bsm/DNA/server. SPIN: Surface Properties of Interfaces Database (Columbia University, New York), http:// trantor.bioc.columbia.edu/cgi-bin/SPIN/. Visual Survey of Homodimeric Proteins (Scripps Research Institute, La Jolla, CA), http:// www.scripps.edu/pub/goodsell/interface. CAPRI: Critical Assessment of PRedicted Interactions, http://capri.ebi.ac.uk. GRASP (Columbia University, New York), http://trantor.bioc.columbia.edu/. DIP: Database of Interacting Proteins (Xenarios et al., 2000), http://www.doe-mbi.ucla.edu. ASEdb: Database of Hotspots in Proteins (Thorn and Bogan, 2001), http://www.asedb.org. Web 1, http://www.biochem.ucl.ac.uk/bsm/PP/server/. Web 2, http://trantor.bioc.columbia.edu/GRASS/surfserv_enter.cgi. Web 3, http://honiglab.cpmc.columbia.edu/.

REFERENCES Ackermann, F., Herrmann, G., Posch, S., and Sagerer, G. (1998). Estimation and ®ltering of potential protein±protein docking positions. Bioinformatics 14(2), 196±205. Amadei, A., Linssen, A. B., and Berendsen, H. J. (1993). Essential dynamics of proteins. Proteins 17(4), 412±425. Argos, P. (1988). An investigation of protein subunit and domain interfaces. Protein Eng. 2(2), 101±113. Bachar, O., Fischer, D., Nussinov, R., and Wolfson, H. (1993). A computer vision based technique for 3-D sequence-independent structural comparison of proteins. Protein Eng. 6(3), 279±288. Ê crystal structure of the lambda Beamer, L. J., and Pabo, C. O. (1992). Re®ned 1.8 A repressor±operator complex. J. Mol. Biol. 227(1), 177±196. Berg, O. G., and von Hippel, P. H. (1985). Diffusion-controlled macromolecular interactions. Annu. Rev. Biophys. Biophys. Chem. 14, 131±160. Blom, N. S., and Sygusch, J. (1997). High resolution fast quantitative docking using Fourier domain correlation techniques. Proteins 27(4), 493±506. Bogan, A. A., and Thorn, K. S. (1998). Anatomy of hot spots in protein interfaces. J. Mol. Biol. 280, 1±9. Braden, B. C., Fields, B. A., Ysern, X., Dall'Acqua, W., Goldbaum, F. A., Poljak, R. J., and Mariuzza, R. A. (1996). Crystal structure of an Fv±Fv idiotope±anti-idiotope complex at Ê resolution. J. Mol. Biol. 264, 137±151. 1.9 A Bryngelson, J. D., and Wolynes, P. G. (1987). Spin glasses and the statistical mechanics of protein folding. Proc. Natl. Acad. Sci. USA 84(21), 7524±7528. Buckle, A. M., Schreiber, G., and Fersht, A. R. (1994). Protein±protein recognition: Crystal Ê resolution. Biochemistry 33(30), structural analysis of a barnase±barstar complex at 2.0-A 8878±8889. Camacho, C. J., Kimura, S. R., DeLisi, C., and Vajda, S. (2000). Kinetics of desolvationmediated protein±protein binding. Biophys. J. 78(3), 1094±1105. Camacho, C. J., Weng, Z., Vajda, S., and DeLisi, C. (1999). Free energy landscapes of encounter complexes in protein±protein association. Biophys. J. 76(3), 1166±1178. Carter, P. J., Winter, G., Wilkinson, A. J., and Fersht, A. R. (1984). The use of double mutants to detect structural changes in the active site of the tyrosyl-tRNA synthetase (Bacillus stearothermophilus). Cell 38(3), 835±840. Carugo, O., and Argos, P. (1997). Protein±protein crystal-packing contacts. Protein Sci. 6(10), 2261±2263.

STRUCTURAL BASIS OF MACROMOLECULAR RECOGNITION

67

Chen, R., and Weng, Z. (2002). Docking unbound proteins using shape complementarity, desolvation, and electrostatics. Proteins 47, 281±294. Cher®ls, J., Duquerroy, S., and Janin, J. (1991). Protein±protein recognition analyzed by docking simulation. Proteins 11(4), 271±280. Chitarra, V., Alzari, P. M., Bentley, G. A., Bhat, T. N., Eisele, J. L., Houdusse, A., Lescar, J., Souchon, H., and Poljak, R. J. (1993). Three-dimensional structure of a heteroclitic antigen±antibody cross-reaction complex. Proc. Natl. Acad. Sci. USA 90(16), 7711± 7715. Chothia, C., and Janin, J. (1975). Principles of protein±protein recognition. Nature 256(5520), 705±708. Clackson, T., and Wells, J. A. (1995). A hot spot of binding energy in a hormone±receptor interface. Science 267(5196), 383±386. Connolly, M. L. (1983). Analytical molecular surface calculation. J. Appl. Crystallogr. 16, 548±558. Connolly, M. L. (1985). Molecular surface triangulation. J. Appl. Crystallogr. 18, 499±505. Connolly, M. L. (1986). Shape complementarity at the hemoglobin alpha 1 beta 1 subunit interface. Biopolymers 25(7), 1229±1247. Dahiyat, B. I., and Mayo, S. L. (1997). De novo protein design: Fully automated sequence selection. Science 278(5335), 82±87. Dasgupta, S., Iyer, G. H., Bryant, S. H., Lawrence, C. E., and Bell, J. A. (1997). Extent and nature of contacts between protein molecules in crystal lattices and between subunits of protein oligomers. Proteins 28(4), 494±514. Davies, D. R., and Cohen, G. H. (1996). Interactions of protein antigens with antibodies. Proc. Natl. Acad. Sci. USA 93(1), 7±12. de Vos, A. M., Ultsch, M., and Kossiakoff, A. A. (1992). Human growth hormone and extracellular domain of its receptor: Crystal structure of the complex. Science 255(5042), 306±312. Dixon, J. S. (1997). Evaluation of the CASP2 docking section. Proteins Suppl. 1), 198±204. Elcock, A. H., Gabdoulline, R. R., Wade, R. C., and McCammon, J. A. (1999). Computer simulation of protein±protein association kinetics: Acetylcholinesterase-fasciculin. J. Mol. Biol. 291, 149±162. Elcock, A. H., Sept, D., and McCammon, J. A. (2001). Computer simulation of protein± protein interactions. J. Phys. Chem. B. 105(8), 1504±1518. Ermak, D. L., and McCammon, J. A. (1978). Brownian dynamics with hydrodynamic interactions. J. Chem. Phys. 69, 1352±1360. Finkelstein, A. V., and Janin, J. (1989). The price of lost freedom: Entropy of bimolecular complex formation. Protein Eng. 3(1), 1±3. Finney, J. L. (1975). Volume occupation, environment and accessibility in proteins: The problem of the protein surface. J. Mol. Biol. 96, 721±732. Fischer, D., Lin, S. L., Wolfson, H. L., and Nussinov, R. (1995a). A geometry-based suite of molecular docking processes. J. Mol. Biol. 248, 459±477. Fischer, D., Norel, R., Wolfson, H., and Nussinov, R. (1993). Surface motifs by a computer vision technique: Searches, detection, and implications for protein±ligand recognition. Proteins 16(3), 278±292. Fisher, H. F., and Singh, N. (1995). Calorimetric methods for interpreting protein±ligand interactions. Methods Enzymol. 259, 194±221. Fogolari, F., Elcock, A. H., Esposito, G., Viglino, P., Briggs, J. M., and McCammon, J. A. (1997). Electrostatic effects in homeodomain±DNA interactions. J. Mol. Biol. 267, 368±381. Ford, C. E., Skiba, N. P., Bae, H., Daaka, Y., Reuveny, E., Shekter, L. R., Rosal, R., Weng, G., Yang, C. S., and Iyengar, R. (1998). Molecular basis for interactions of G protein betagamma subunits with effectors. Science 280(5367), 1271±1274.

68

È L JANIN SHOSHANA J. WODAK AND JOE

Frisch, C., Schreiber, G., Johnson, C. M., and Fersht, A. R. (1997). Thermodynamics of the interaction of barnase and barstar: Changes in free energy versus changes in enthalpy on mutation. J. Mol. Biol. 267, 696±706. Froloff, N., Windemuth, A., and Honig, B. (1997). On the calculation of binding free energies using continuum methods: Application to MHC class I protein±peptide interactions. Protein Sci. 6(6), 1293±1301. Fujinaga, M., Sielecki, A. R., Read, R. J., Ardelt, W., Laskowski, M., Jr., and James, M. N. (1987). Crystal and molecular structures of the complex of alpha-chymotrypsin with Ê resolution. J. Mol. Biol. 195, its inhibitor turkey ovomucoid third domain at 1.8 A 397±418. Gabb, H. A., Jackson, R. M., and Sternberg, M. J. (1997). Modelling protein docking using shape complementarity, electrostatics and biochemical information. J. Mol. Biol. 272, 106±120. Gabdoulline, R. R., and Wade, R. C. Effective charges for macromolecules in solvent. (1996). J. Phys. Chem. 100, 3868. Gabdoulline, R. R., and Wade, R. C. (1997). Simulation of the diffusional association of barnase and barstar. Biophys. J. 72(5), 1917±1929. Gabdoulline, R. R., and Wade, R. C. (1998). Brownian dynamics simulation of protein± protein diffusional encounter. Methods 14, 329±341. Gabdoulline, R. R., and Wade, R. C. (1999). On the protein±protein diffusional encounter complex. J. Mol. Recogn. 12(4), 226±234. Gamble, T. R., Vajdos, F. F., Yoo, S., Worthylake, D. K., Houseweart, M., Sundquist, W. I., and Hill, C. P. (1996). Crystal structure of human cyclophilin A bound to the aminoterminal domain of HIV-1 capsid. Cell 87(7), 1285±1294. Gardiner, E. J., Willett, P., and Artymiuk, P. J. (2001). Protein docking using a genetic algorithm. Proteins 44, 44±56. Gilson, M. K., and Honig, B. (1988). Calculation of the total electrostatic energy of a macromolecular system: Solvation energies, binding energies, and conformational analysis. Proteins 4(1), 7±18. Gilson, M. K., Given, J. A., Bush, B. L., and McCammon, J. A. (1997). The statisticalthermodynamic basis for computation of binding af®nities: A critical review. Biophys J. 72, 1047±1069. Goodsell, D. S., and Olson, A. J. (1993). Soluble proteins: Size, shape and function. Trends Biochem. Sci. 18(3), 65±68. Guillet, V., Lapthorn, A., and Mauguen, Y. (1993). Three-dimensional structure of a Ê resolution. FEBS Lett. 330(2), 137±140. barnase±30 GMP complex at 2.2 A Halperin, I., Ma, B., Wolfson, H., and Nissinov, R. (2002). Principles of docking: An overview of search algorithms and a guide to scoring functions. Proteins 47, 409±443. Harpaz, Y., Gerstein, M., and Chothia, C. (1994). Volume changes on protein folding. Structure 2(7), 641±649. Harrison, R. W., Kourinov, I. V., and Andrews, L. C. (1994). The Fourier±Green's function and the rapid evaluation of molecular potentials. Protein Eng. 7(3), 359±369. Hartley, R. W. (1993). Directed mutagenesis and barnase±barstar recognition. Biochemistry 32(23), 5978±5984. Hegde, R. S., Grossman, S. R., Laimins, L. A., and Sigler, P. B. (1992). Crystal structure at Ê of the bovine papillomavirus-1 E2 DNA-binding domain bound to its DNA target. 1.7 A Nature 359(6395), 505±512. Hendsch, Z. S., and Tidor, B. (1999). Electrostatic interactions in the GCN4 leucine zipper: Substantial contributions arise from intramolecular interactions enhanced on binding. Protein Sci. 8(7), 1381±1392.

STRUCTURAL BASIS OF MACROMOLECULAR RECOGNITION

69

Hibbits, K. A., Gill, D. S., and Willson, R. C. (1994). Isothermal titration calorimetric study of the association of hen egg lysozyme and the anti-lysozyme antibody HyHEL-5. Biochemistry 33(12), 3584±3590. Hilvert, D. (2000). Genetic selection as a tool in mechanistic enzymology and protein design. Ernst Schering Res. Found. Workshop(32), 253±268. Honig, B., and Nicholls, A. (1995). Classical electrostatics in biology and chemistry. Science 268(5214), 1144±1149. Horovitz, A. (1987). Non-additivity in protein±protein interactions. J. Mol. Biol. 196, 733±735. Housset, D., Mazza, G., Gregoire, C., Piras, C., Malissen, B., and Fontecilla-Camps, J. C. (1997). The three-dimensional structure of a T-cell antigen receptor Valpha V beta heterodimer reveals a novel arrangement of the V beta domain. EMBO J. 16(14), 4205±4216. Huber, R., Kukla, D., Bode, W., Schwager, P., Bartels, K., Deisenhofer, J., and Steigemann, W. (1974). Structure of the complex formed by bovine trypsin and bovine pancreatic trypsin Ê resolution. J. Mol. Biol. 89, 73±101. inhibitor. II. Crystallographic re®nement at 1.9 A Janin, J. (1995). Elusive af®nities. Proteins 21(1), 30±39. Janin, J. (1996). Quantifying biological speci®city: The statistical mechanics of molecular recognition. Proteins 25(4), 438±445. Janin, J. (1997). Elusive af®nities. Proteins 21, 30±39. Janin, J., and Chothia, C. (1976). Stability and speci®city of protein±protein interactions: The case of the trypsin±trypsin inhibitor complexes. J. Mol. Biol. 100, 197±211. Janin, J., and Chothia, C. (1990). The structure of protein±protein recognition sites. J. Biol. Chem. 265(27), 16027±16030. Janin, J., Miller, S., and Chothia, C. (1988). Surface, subunit interfaces and interior of oligomeric proteins. J. Mol. Biol. 204, 155±164. Janin, J., and Rodier, F. (1995). Protein±protein interaction at crystal contacts. Proteins 23(4), 580±587. Janin, J., and Wodak, S. J. (1985). Reaction pathway for the quaternary structure change in hemoglobin. Biopolymers 24(3), 509±526. Jiang, F., and Kim, S. H. (1991). ``Soft docking'': Matching of molecular surface cubes. J. Mol. Biol. 219, 79±102. Jones, S., and Thornton, J. M. (1995). Protein±protein interactions: A review of protein dimer structures. Prog. Biophys. Mol. Biol. 63(1), 31±65. Jones, S., and Thornton, J. M. (1996). Principles of protein±protein interactions. Proc. Natl. Acad. Sci. USA 93(1), 13±20. Jones, S., van Heyningen, P., Berman, H. M., and Thornton, J. M. (1999). Protein±DNA interactions: A structural analysis. J. Mol. Biol. 287, 877±896. Katchalski-Katzir, E., Shariv, I., Eisenstein, M., Friesem, A. A., A¯alo, C., and Vakser, I. A. (1992). Molecular surface recognition: Determination of geometric ®t between proteins and their ligands by correlation techniques. Proc. Natl. Acad. Sci. USA 89(6), 2195±2199. Kuntz, I. D., Blaney, J. M., Oatley, S. J., Langridge, R., and Ferrin, T. E. (1982). A geometric approach to macromolecule±ligand interactions. J. Mol. Biol. 161, 269±288. Kwong, P. D., Wyatt, R., Robinson, J., Sweet, R. W., Sodroski, J., and Hendrickson, W. A. (1998). Structure of an HIV gp120 envelope glycoprotein in complex with the CD4 receptor and a neutralizing human antibody. Nature 393(6686), 648±659. Ladbury, J. E., and Chowdhry, B. Z. (1996). Sensing the heat: The application of isothermal titration calorimetry to thermodynamic studies of biomolecular interactions. Chem. Biol. 3(10), 791±801. Ladbury, J. E., Wright, J. G., Sturtevant, J. M., and Sigler, P. B. (1994). A thermodynamic study of the trp repressor±operator interaction. J. Mol. Biol. 238, 669±681.

70

È L JANIN SHOSHANA J. WODAK AND JOE

Lambright, D. G., Noel, J. P., Hamm, H. E., and Sigler, P. B. (1994). Structural determinants for activation of the alpha-subunit of a heterotrimeric G protein. Nature 369(6482), 621±628. Lambright, D. G., Sondek, J., Bohm, A., Skiba, N. P., Hamm, H. E., and Sigler, P. B. (1996). Ê crystal structure of a heterotrimeric G protein. Nature 379(6563), 311±319. The 2.0 A Larsen, T. A., Olson, A. J., and Goodsell, D. S. (1998). Morphology of protein±protein interfaces. Structure 6(4), 421±427. Laskowski, R. A. (1995). SURFNET: A program for visualizing molecular surfaces, cavities, and intermolecular interactions. J. Mol. Graph. 13(5), 323±330. Lawrence, M. C., and Colman, P. M. (1993). Shape complementarity at protein/protein interfaces. J. Mol. Biol. 234, 946±950. Lee, B., and Richards, F. M. (1971). The interpretation of protein structures: Estimation of static accessibility. J. Mol. Biol. 55, 379±400. Lee, R. H., and Rose, G. D. (1985). Molecular recognition. I. Automatic identi®cation of topographic surface features. Biopolymers 24(8), 1613±1627. Levitt, M. (1976). A simpli®ed representation of protein conformations for rapid simulation of protein folding. J. Mol. Biol. 104, 59±107. Lin, S. L., Nussinov, R., Fischer, D., and Wolfson, H. J. (1994). Molecular surface representations by sparse critical points. Proteins 18(1), 94±101. Lo Conte, L. L., Chothia, C., and Janin, J. (1999). The atomic structure of protein±protein recognition sites. J. Mol. Biol. 285, 2177±2198. Madura, J. D., Briggs, J. M., Wade, R. C., Davis, M. E., Luty, B. A., Ilin, A., Antosiewicz, J., Gilson, M. K., Bagheri, B., and Scott, L. R. (1995). Electrostatics and diffusion of molecules in solution: simulations with the University of Houston Brownian dynamics program. Comp. Phys. Commun. 91, 57±95. Madura, J. D., Briggs, J. M., Wade, R. C., and Gabdoulline, R. R. (1998). Brownian dynamics. In ``Encyclopedia of Computational Chemistry'' (P. V. R. Schleyer, N. L. Allinger, T. Clark, J. Gasteiger, P. A. Kollman, H. F. Schreiner, and H. F. Schaefer, eds.), vol. 1, pp. 141±154. John Wiley and Sons, UK. Madura, J. D., Davis, M. E., Gilson, M. K., Wade, R. C., Luty, B. A., and McCammon, J. A. (1994). Biological applications of electrostatic calculations and Brownian dynamics simulations. Comp. Chem. Rev. 5, 229±267. Makhatadze, G. I., and Privalov, P. L. (1996). On the entropy of protein folding. Protein Sci. 5(3), 507±510. Marquart, M., Walter, J., Deisenhofer, J., Bode, W., and Huber, R. (1983). The geometry of the reactive site and of the peptide groups in trypsin, trypsinogen and its complexes with inhibitors. Acta Crystallogr. Sect. B. 39, 480±490. Mauguen, Y., Hartley, R. W., Dodson, E. J., Dodson, G. G., Bricogne, G., Chothia, C., and Jack, A. (1982). Molecular structure of a new family of ribonucleases. Nature 297(5862), 162±164. McDonald, I. K., and Thornton, J. M. (1994). Satisfying hydrogen bonding potential in proteins. J. Mol. Biol. 238, 777±793. Meyer, M., Wilson, P., and Schomburg, D. (1996). Hydrogen bonding and molecular surface shape complementarity as a basis for protein docking. J. Mol. Biol. 264, 199±210. Miller, S., Lesk, A. M., Janin, J., and Chothia, C. (1987). The accessible surface area and stability of oligomeric proteins. Nature 328(6133), 834±836. Misra, V. K., Hecht, J. L., Sharp, K. A., Friedman, R. A., and Honig, B. (1994a). Salt effects on protein±DNA interactions. The lambda cI repressor and EcoRI endonuclease. J. Mol. Biol. 238, 264±280. Misra, V. K., Sharp, K. A., Friedman, R. A., and Honig, B. (1994b). Salt effects on ligand± DNA binding. Minor groove binding antibiotics. J. Mol. Biol. 238, 245±263. Misra, V. K., Hecht, J. L., Yang, A. S., and Honig, B. (1998). Electrostatic contributions to the binding free energy of the lambdacI repressor to DNA. Biophys. J. 75(5), 2262±2273.

STRUCTURAL BASIS OF MACROMOLECULAR RECOGNITION

71

Nadassy, K., Tomas-Oliveira, I., Alberts, I., Janin, J., and Wodak, S. J. (2001). Standard atomic volumes in double-stranded DNA and packing in protein±DNA interfaces. Nucleic Acids Res. 29(16), 3362±3376. Nadassy, K., Wodak, S. J., and Janin, J. (1999). Structural features of protein±nucleic acid recognition sites. Biochemistry 38(7), 1999±2017. Ê resolution re®ned structure of a TATA boxNikolov, D. B., and Burley, S. K. (1994). 2.1 A binding protein (TBP). Nat. Struct. Biol. 1(9), 621±637. Nikolov, D. B., Chen, H., Halay, E. D., Hoffman, A., Roeder, R. G., and Burley, S. K. (1996). Crystal structure of a human TATA box-binding protein/TATA element complex. Proc. Natl. Acad. Sci. USA 93(10), 4862±4867. Ê crystal structure of transducinNoel, J. P., Hamm, H. E., and Sigler, P. B. (1993). The 2.2 A alpha complexed with GTP gamma S. Nature 366(6456), 654±663. Norel, R., Fischer, D., Wolfson, H. J., and Nussinov, R. (1994). Molecular surface recognition by a computer vision-based technique. Protein Eng. 7(1), 39±46. Northrup, S. H., and Erickson, H. P. (1992). Kinetics of protein±protein association explained by Brownian dynamics computer simulation. Proc. Natl. Acad. Sci. USA 89(8), 3338±3342. Northrup, S. H., Boles, J. O., and Reynolds, J. C. (1988). Brownian dynamics of cytochrome c and cytochrome c peroxidase association. Science 241, 372±384. Ogata, K., and Wodak, S. J. Conserved water molecules in MHC Class-I molecules and their putative structural and functional Roles. Submitted for publication. Ollis, D. L., Brick, P., Hamlin, R., Xuong, N. G., and Steitz, T. A. (1985a). Structure of large fragment of Escherichia coli DNA polymerase I complexed with dTMP. Nature 313(6005), 762±766. Ollis, D. L., Kline, C., and Steitz, T. A. (1985b). Domain of E. coli DNA polymerase I showing sequence homology to T7 DNA polymerase. Nature 313(6005), 818±819. Oobatake, M., and Ooi, T. (1993). Hydration and heat stability effects on protein unfolding. Prog. Biophys. Mol. Biol. 59(3), 237±284. Palma, P. N., Krippahl, L., Wampler, J. E., and Moura, J. J. (2000). BiGGER: A new (soft) docking algorithm for predicting protein interactions. Proteins 39, 44±56. Pelletier, H., and Kraut, J. (1992). Crystal structure of a complex between electron transfer partners, cytochrome c peroxidase and cytochrome c. Science 258(5089), 1748±1755. Phizicky, E. M., and Fields, S. (1995). Protein±protein interactions: Methods for detection and analysis. Microbiol. Rev. 59(1), 94±123. Press, W. H., Teukolsky, S. A., Vetterling, W. T., and Flannery, B. P. (1992). ``Numerical Recipes in C,'' 2nd ed. Cambridge Univ. Press, Cambridge. Privalov, P. L., and Gill, S. J. (1988). Stability of protein structure and hydrophobic interaction. Adv. Protein Chem. 39, 191±234. Radic, Z., Kirchhoff, P. D., Quinn, D. M., McCammon, J. A., and Taylor, P. (1997). Electrostatic in¯uence on the kinetics of ligand binding to acetylcholinesterase. Distinctions between active center ligands and fasciculin. J. Biol. Chem. 272(37), 23265±23277. Ravichandran, S., and Talbot, J. (2000). Mobility of adsorbed proteins: A Brownian dynamics study. Biophys. J. 78(1), 110±120. Reddy, V. S., Giesing, H. A., Morton, R. T., Kumar, A., Post, C. B., Brooks, C. L., III, and Johnson, J. E. (1998). Energetics of quasiequivalence: Computational analysis of protein± protein interactions in icosahedral viruses. Biophys. J. 74(1), 546±558. Richards, F. M. (1974). The interpretation of protein structures: Total volume, group volume distributions and packing density. J. Mol. Biol. 82, 1±14. Ritchie, D. W., and Kemp, G. J. (2000). Protein docking using spherical polar Fourier correlations. Proteins 39(2), 178±194. Rydel, T. J., Tulinsky, A., Bode, W., and Huber, R. (1991). Re®ned structure of the hirudin± thrombin complex. J. Mol. Biol. 221, 583±601.

72

È L JANIN SHOSHANA J. WODAK AND JOE

Schapira, M., Totrov, M., and Abagyan, R. (1999). Prediction of the binding energy for small molecules, peptides and proteins. J. Mol. Recogn. 12(3), 177±190. Schreiber, G., and Fersht, A. R. (1993). The refolding of cis- and trans-peptidylprolyl isomers of barstar. Biochemistry 32(41), 11195±11203. Schreiber, G., and Fersht, A. R. (1995). Energetics of protein±protein interactions: Analysis of the barnase±barstar interface by single mutations and double mutant cycles. J. Mol. Biol. 248, 478±486. Schreiber, G., and Fersht, A. R. (1996). Rapid, electrostatically assisted association of proteins. Nat. Struct. Biol. 3(5), 427±431. Schwehm, J. M., and Stites, W. E. (1998). Application of automated methods for determination of protein conformational stability. Methods Enzymol. 295, 150±170. Selzer, T., Albeck, S., and Schreiber, G. (2000). Rational design of faster associating and tighter binding protein complexes. Nat. Struct. Biol. 7(7), 537±541. Sept, D., Elcock, A. H., and McCammon, J. A. (1999). Computer simulations of actin polymerization can explain the barbed-pointed end asymmetry. J. Mol. Biol. 294, 1181±1189. Shakked, Z., Guzikevich-Guerstein, G., Frolow, F., Rabinovich, D., Joachimiak, A., and Sigler, P. B. (1994). Determinants of repressor/operator recognition from the structure of the trp operator binding site. Nature 368(6470), 469±473. Sheinerman, F. B., Norel, R., and Honig, B. (2000). Electrostatic aspects of protein±protein interactions. Curr. Opin. Struct. Biol. 10(2), 153±159. Sheriff, S., Silverton, E. W., Padlan, E. A., Cohen, G. H., Smith-Gill, S. J., Finzel, B. C., and Davies, D. R. (1987). Three-dimensional structure of an antibody±antigen complex. Proc. Natl. Acad. Sci. USA 84(22), 8075±8079. Shoemaker, B. A., Portman, J. J., and Wolynes, P. G. (2000). Speeding molecular recognition by using the folding funnel: The ¯y-casting mechanism. Proc. Natl. Acad. Sci. USA 97(16), 8868±8873. Shoichet, B. K., Bodian, D. L., and Kuntz, I. D. (1992). Molecular docking using shape descriptors. J. Comput. Chem. 13(3), 380±397. Shoichet, B. K., and Kuntz, I. D. (1991). Protein docking and complementarity. J. Mol. Biol. 221, 327±346. Smith, G. R., and Sternberg, M. J. (2002). Prediction of protein±protein interactions by docking methods. Curr. Opin. Struct. Biol. 12, 28±35. Smoluchowski, M. V. (1917). Versuch einer mathematischen Theorie der KoagulationsKinetik KolloiderLoeschungen. J. Phys. Chem. 92, 129±168. Sondek, J., Bohm, A., Lambright, D. G., Hamm, H. E., and Sigler, P. B. (1996). Crystal Ê resolution. Nature 379(6563), structure of a G-protein beta gamma dimer at 2.1 A 369±374. Srere, P. A. (1999). Protein interactions. Methods 19(2), 193. Strynadka, N. C., Eisenstein, M., Katchalski-Katzir, E., Shoichet, B. K., Kuntz, I. D., Abagyan, R., Totrov, M., Janin, J., Cher®ls, J., and Zimmerman, F. (1996). Molecular docking programs successfully predict the binding of a beta-lactamase inhibitory protein to TEM1 beta-lactamase. Nat. Struct. Biol. 3(3), 233±239. Thorn, K. S., and Bogan, A. A. (2001). ASEdb: A database of alanine mutations and their effects on the free energy of binding in protein interactions. Bioinformatics 17(3), 284±285. Tovchigrechko, A., Wells, C. A., and Vakser, I. (2001). Docking of protein models. In press. Tsai, J., Taylor, R., Chothia, C., and Gerstein, M. (1999). The packing density in proteins: Standard radii and volumes. J. Mol. Biol. 290, 253±266. Vakser, I. A. (1995). Protein docking for low-resolution structures. Protein Eng. 8(4), 371±377. Vakser, I. A. (1996). Long-distance potentials: An approach to the multiple-minima problem in ligand±receptor interaction. Protein Eng. 9(1), 37±41.

STRUCTURAL BASIS OF MACROMOLECULAR RECOGNITION

73

Vakser, I. A. (1997). Evaluation of GRAMM low-resolution docking methodology on the hemagglutinin±antibody complex. Proteins Suppl. 1), 226±230. Vakser, I. A., and A¯alo, C. (1994). Hydrophobic docking: A proposed enhancement to molecular recognition techniques. Proteins 20(4), 320±329. van de Locht, A., Stubbs, M. T., Bode, W., Friedrich, T., Bollschweiler, C., Hoffken, W., and Huber, R. (1996). The ornithodorin±thrombin crystal structure, a key to the TAP enigma? EMBO J. 15(22), 6011±6017. Vincent, J. P., and Lazdunski, M. (1972). Trypsin±pancreatic trypsin inhibitor association. Dynamics of the interaction and role of disul®de bridges. Biochemistry 11(16), 2967±2977. Vitkup, D., Melamud, E., Moult, J., and Sander, C. (2001). Completeness in structural genomics. Nat. Struct. Biol. 8(6), 559±566. von Hippel, P. H., and Berg, O. G. (1989). Facilitated target location in biological systems. J. Biol. Chem. 264(2), 675±678. Wade, R. C. (1996). Brownian dynamics simulations of enzyme±substrate encounter. Biochem. Soc. Trans. 24(1), 254±259. Wang, H. (1991). Grid-search molecular accessible surface algorithm for solving the protein docking problem. J. Comp. Chem. 12, 746±750. Wang, J. H., Smolyar, A., Tan, K., Liu, J. H., Kim, M., Sun, Z. Y., Wagner, G., and Reinherz, E. L. (1999). Structure of a heterophilic adhesion complex between the human CD2 and CD58 (LFA-3) counterreceptors. Cell 97(6), 791±803. Weiner, S. J., Kollman, P. A., Case, D. A., Singh, U. C., Ghio, C., Alagona, G., Profeta Jr., S., and Weiner, P. (1984). A new force ®eld for molecular mechanical simulation of nucleic acids and proteins. J. Am. Chem. Soc. 106, 765±784. Welch, M., Chinardet, N., Mourey, L., Birck, C., and Samama, J. P. (1998). Structure of the CheY-binding domain of histidine kinase CheA in complex with CheY. Nat. Struct. Biol. 5(1), 25±29. Werner, M. H., and Burley, S. K. (1997). Architectural transcription factors: Proteins that remodel DNA. Cell 88(6), 733±736. Wernisch, L., Hery, S., and Wodak, S. J. (2000). Automatic protein design with all atom force-®elds by exact and heuristic optimization. J. Mol. Biol. 301, 713±736. Wodak, S. J., and Janin, J. (1978). Computer analysis of protein±protein interaction. J. Mol. Biol. 124, 323±342. Wodak, S. J., and Janin, J. (1980). Analytical approximation to the accessible surface area of proteins. Proc. Natl. Acad. Sci USA. 77(4), 1736±1740. Wriggers, W., Milligan, R. A., and McCammon, J. A. (1999). Situs: A package for docking crystal structures into low-resolution maps from electron microscopy. J. Struct. Biol. 125, 185±195. Xenarios, I., Rice, D. W., Salwinski, L., Baron, M. K., Marcotte, E. M., and Eisenberg, D. (2000). DIP: The database of interacting proteins. Nucleic Acids Res. 28(1), 289±291. Xiao, L., and Honig, B. (1999). Electrostatic contributions to the stability of hyperthermophilic proteins. J. Mol. Biol. 289, 1435±1444. Xu, D., Tsai, C. J., and Nussinov, R. (1997). Hydrogen bonds and salt bridges across protein±protein interfaces. Protein Eng. 10(9), 999±1012. Ysern, X., Li, H., and Mariuzza, R. A. (1998). Imperfect interfaces. Nat. Struct. Biol. 5(6), 412±414. Zachmann, C. D., Heiden, W., Schlenkrich, M., and Brickmann, J. (1992). Topological analysis of complex molecular surfaces. J. Comp. Chem. 13(1), 76±84. Zhang, C., Vasmatzis, G., Cornette, J. L., and DeLisi, C. (1997). Determination of atomic desolvation energies from the structures of crystallized proteins. J. Mol. Biol. 267, 707±726.

This Page Intentionally Left Blank

SEQUENCE ANALYSIS OF MULTIDOMAIN PROTEINS: PAST PERSPECTIVES AND FUTURE DIRECTIONS È RG SCHULTZ,*,³ ,1 BY RICHARD R. COPLEY,* CHRIS P. PONTING,² JO AND PEER BORK*,³ * EMBL, 69012 Heidelberg, Germany, ² MRC Functional Genetics Unit, Department of Human Anatomy and Genetics, University of Oxford, Oxford OX1 3QX UK, United Kingdom, and ³ Max-DelbruÈck-Center for Molecular Medicine, Berlin-Buch, Germany

I. Identi®cation of Novel Protein Domain Families . . . . . . . . . . . . . . . . . . . . . . . . . . A. History of Domain Discovery . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B. Domain Discovery Today. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . II. Methods for Classifying Protein Domain Families . . . . . . . . . . . . . . . . . . . . . . . . . III. From Domain Classi®cation to Domain Context . . . . . . . . . . . . . . . . . . . . . . . . . . A. Zooming In: Residue Context and Functional Subtyping . . . . . . . . . . . . . . . . B. Zooming Out: Domain Context within Proteins. . . . . . . . . . . . . . . . . . . . . . . . IV. Genome-Wide Analysis: New Quality in Domain Research. . . . . . . . . . . . . . . . . . A. Domains as a Tool to Aid Gene Prediction . . . . . . . . . . . . . . . . . . . . . . . . . . . . B. Orthology and Paralogy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . C. Comparative Analysis and Evolution of Function . . . . . . . . . . . . . . . . . . . . . . . V. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

75 77 79 81 85 86 86 89 89 91 93 96 96

I. IDENTIFICATION OF NOVEL PROTEIN DOMAIN FAMILIES The reductionist concept of domains in proteins has important roles to play in structural biology, genetics, evolution, and biochemistry. Unfortunately, however, this concept often is used differently in each of these subdisciplines. In structural biology, protein domains are usually de®ned as continuous polypeptide chains that are folded into spatially distinct structural units (e.g., Janin and Chothia, 1985). By contrast, a domain is often de®ned in the biochemical and genetic literature as the minimal fragment of a gene that is still able to perform a certain function, such as that identi®ed in deletion experiments. In sequence analysis, domains are usually de®ned as such only when they are contiguous in sequence and when they are found in different multidomain contexts, for example, when they occur with different ¯anking domains. The situation in which domain homologs are present in proteins with different domain compositions and arrangements is assumed to have arisen through intragenomic duplication and recombination events. Such genetically mobile domains are also sometimes called modules. This term was originally introduced into the protein world in the context of immunoglobulin domains, but was later used to describe packed 1

Present address: CellZome, Meyerhofstrasse 1, 69117 Heidelberg, Germany. 75

ADVANCES IN PROTEIN CHEMISTRY, Vol. 61

Copyright 2003, Elsevier Science (USA). All rights reserved. 0065-3233/03 $35.00

76

RICHARD R. COPLEY ET AL.

supersecondary structure elements within the context of exon-shuf¯ing theories of gene evolution (Go, 1983). Later still, however, the term was used to describe mobile domains in extracellular multidomain proteins (Patthy et al., 1984). Use of this term has increased rapidly since the beginning of the 1990s (e.g., Baron et al., 1991; Bork, 1992). Modules are viewed as evolutionarily independent entities that may be found in single copies. As such they differ from repeats, sequence units that are structurally and functionally interdependent and require multiple copies to form a stable structure which, in turn, may be considered as a domain. A third commonly used term is ``motif.'' This may encompass only a portion of a domain, such as active or binding site residues, or may exist outside of domains in sequence regions that are structured only when bound to a substrate. For further de®nitions of terms, see reviews such as Bork and Koonin (1996). Analysis of completed genomes shows that eukaryotes encode larger numbers of longer proteins (Fig. 1) As structural domains typically have an average size of approximately 100±150 residues, this leads to the conclusion that a greater proportion of proteins found in eukaryotes are multidomain in character when compared to prokaryotes. As each domain is likely to contribute differently to the functional attributes of the protein as a whole, it follows that identi®cation of domains, repeats, and motifs is often an essential part of understanding how higher levels of protein function emerge during the processes of evolution.

3000 2500 E. coli

Count

2000

S. cerevisiae

1500

D. melanogaster 1000

C. elegans

500

>1000

950

850

750

650

550

450

350

250

150

50

0

Length

FIG. 1. Histogram of the lengths of predicted proteins encoded in ths completely sequenced genomes of Esherichia coli, Saccharomyces cerevisiae, Caenorhabditis elegans, and Drosophila melanogaster.

SEQUENCE ANALYSIS OF MULTIDOMAIN PROTEINS

77

Functional attributes of a protein, such as active or binding sites, can be assigned to domains, repeats, and motifs. Other structures, however, also contribute to a protein's functions. These are sequences that target proteins to, for example, membranes (transmembrane helices) or organelles (such as signal peptides or mitochondrial import sequences). Coiled coils can contribute to function by mediating protein±protein interactions or spatially separating functionally distinct protein regions. By understanding the functions of the individual regions of a sequence, such as domains, motifs, coils, transmembrane helices, signal peptides, and other targeting sequences, we aim to move toward an understanding of the protein as a whole. Here we review different strategies of domain identi®cation at the sequence level from a historical perspective, point to some future directions of domain research, and describe domain discovery in the context of genome analysis. To support our points, we provide illustrative examples of domains that are mostly represented in SMART (Simple Modular Architecture Research Tool; Schultz et al., 1998, 2000). Where a SMART domain name is mentioned in the text, we represent it in boldface. A. History of Domain Discovery Following determination of the ®rst crystal structures of proteins, it became obvious that large proteins, such as dehydrogenases, are frequently composed of different structural units or domains (Adams et al., 1970). Although often containing different enzymatic domains, NAD(P)dependent dehydrogenases were found to possess a common dinucleotide-binding domain (Rossmann et al. 1974). From these and similar studies it soon became apparent that pairs of domains do not necessarily always co-occur in proteins. With greater numbers of sequences and structures becoming known in the 1970s, investigators became increasingly reliant on sequence to predict structure. At this time it became a matter of faith that pairs of sequences with considerable similarity possess highly similar structures. Sequence comparisons thus acquired greater signi®cance, and algorithms were devised for alignment (Needleman and Wunsch, 1970; Smith and Waterman, 1981). This gave rise to the thorny question of whether observed sequence similarities imply an evolutionary relatedness or else a chance event (see the review by Altschul and Gish, 1996). This question prompted still ongoing advances in database searching algorithms (Karlin and Altschul, 1990; Altschul et al., 1997; Pearson, 1998; Mott, 2000). These algorithms, in particular the BLAST suite (Altschul et al., 1997), provide reliable and robust alignment score statistics that have held the key to the detection of remote homologies.

78

RICHARD R. COPLEY ET AL.

With the observation of domains in distinct molecular contexts came proposals for genetic mechanisms for their spread. Gilbert (1978, 1985) proposed that individual exons code for domains and that genes represent collections of exons brought together by recombination within intron sequences. For the speci®c case of extracellular proteins, this proposal gained support by constraints on the intron/exon boundaries (Patthy, 1987). There is, however, little evidence that exons code for domains in intracellular proteins (Bork, 1996). The emphasis on exon shuf¯ing in the late 1970s and 1980s was due, in part, to the considerable numbers of extracellular domain families that were ®rst identi®ed in these years (Doolittle, 1985; Patthy, 1985). Extracellular domain families have continued to be discovered, although at a greatly reduced rate in more recent years [for example, the PAN domain (Tordai et al., 1999) ]. To date, there are more than 150 distinct modules, occurring in extracellular portions of proteins, that have been catalogued (see, e.g., the SMART resource; Schultz et al., 2000), and proposals for their nomenclature have been made (Bork and Bairoch, 1995). By contrast, the majority of intracellular domains were only identi®ed in the 1990s (Bork et al., 1997) (Fig. 2a). These domains are mostly involved in signaling, transport, and nuclear processes, but they can also ¯ank the catalytic domains of metabolic enzymes. Identi®cation of intracellular signaling domain families, which are frequently considerably diverged in sequence, was greatly facilitated by improvements in database search algorithms such as BLAST. When one considers, however, the number of proteins currently known to contain these domains (Fig. 2b), it appears that most of the truly widespread cytoplasmic signaling domain families have already been discovered. This is not meant to imply that the era of important domain discoveries is over, only that it may be more pro®table in the future to consider the domains of cellular processes other than cytoplasmic signaling. The above discussion makes a simplistic distinction between extracellular and intracellular domains. Although many domain families such as kringle (KR), epidermal growth factorlike (EGF), and ®bronectin type I and II (FN1, FN2) domains appear to occur only in extracellular proteins, for other domains this is not always the case. Indeed, ®bronectin type III (FN3), von Willebrand factor A (VWA), and immunoglobulin domains (IG), often described as ``extracellular domains,'' are known in intracellular proteins. A more extreme example is that of PDZ domains, most frequently seen in intracellular proteins, but which have also been identi®ed in nuclear and in extracellular proteins (SIP-1 and interleukin-16, respectively).

79

SEQUENCE ANALYSIS OF MULTIDOMAIN PROTEINS

a 80 70

2000

60

1998

1999 1997

50 1996

40

1995

30

1994

20

1992

10

1990

0

1988 1972

1993 1991 1989 1986

b 6000 WD40 5000

Numbers

4000 ANK 3000 2000 1000

SH 2/3

RING

PDZ

PH

Fbox 0 86 89 72 88

90

91 92

93

94

95 Year

96

97

98

99

00

FIG. 2. (a) Proportions of currently known signaling domains discovered in a particular year. (b) Chronology of discovery of cytoplasmic signaling domains. Each year is colored distinctly. The heights of the bars represent the number of proteins that the domain is found in. Particularly common domains are labeled.

B. Domain Discovery Today The rapid expansion of sequence and structure databases is proving to be a great boon to domain discovery and our understanding of domain propagation. Knowledge of complete genome sequences from several representatives of all three forms of cellular life (archaea, bacteria, and eukarya) has focused attention on the presence of prokaryotic homologs

80

RICHARD R. COPLEY ET AL.

of domains previously thought to be speci®c to eukaryotes. From current phylogenetic distributions, it has been possible to infer the presence of the majority of enzymatic domains in the last common ancestor (``cenancestor'') of archaea, bacteria, and eukarya (Ponting et al., 1999). Equally, attempts to identify chordate-speci®c enzymes have produced only a single example thus far (Lander et al., 2001). In contrast, the majority of signaling domains in eukaryotes appear to have no homologs in the prokaryotes (Ponting et al., 1999). Such conclusions may require revision as protein 3D structures are solved, and it becomes possible to identify more distant homologs. Another theme that has recently become apparent is the relatively frequent horizontal transfer of genes encoding signaling domains from eukaryotes to bacteria (but not to archaea) (Ponting et al., 1999). One pressing issue that has arisen from these studies and remains to be resolved is whether such genes have acquired functions in the new bacterial contexts that are distinct from their eukaryotic counterparts. The sequence and structure data explosion is prompting a greater realization that domain families, once thought to be evolutionarily distinct, might merely be separate branches of a greater superfamily tree. Examples of this abound. Extracellular families of tumor necrosis factor (TNF) and complement 1q (C1Q) are now thought to be homologous (Shapiro and Scherer, 1998), as are the intracellular families of WASp-homology 1 (WH1) and Ran-binding domains (RanBD) (Callebaut et al., 1998). Even functionally distinct molecules, such as the extracellular cytokines, interleukins-1, and ®broblast growth factors and the intracellular actin-binding proteins, hisactophilin and fascin, have been shown to be distant homologs (Ponting and Russell, 2000). The ``merging'' of sequence families into superfamilies is also being increasingly seen among the repeats. For example, HAT, protein farnesyl transferase A, and SEL-1 repeats are all now recognized as divergent subfamilies of tetratricopeptide repeats (TPR) (Andrade et al., 2000; Ponting, 2000). Assignment of distant homology is now being greatly assisted by the large numbers of three-dimensional structures of modules being determined, as proteins can share similar structures even though sequence similarity is undetectable with current methods. Several NMR and crystallography groups are specializing in solving structures of modules that are classi®ed in domain databases (e.g., Tsujishita and Hurley, 2000; see below). This has led to the current situation where 65% of the domain families in the SMART database (Schultz et al., 2000) have at least one family member whose structure has been determined and deposited in the PDB.

SEQUENCE ANALYSIS OF MULTIDOMAIN PROTEINS

81

II. METHODS FOR CLASSIFYING PROTEIN DOMAIN FAMILIES The process of collating homologous members of a domain family can be achieved using either automatic clustering ``top-down'' methods or else a ``bottom-up'' method of combining automatic searches with expert assessment of results. The automatic clustering methods build multiple alignments on the basis of ``all-against-all'' sequence comparisons of a given database (or genome). This approach is much faster and more systematic than bottom-up semiautomatic methods but has signi®cant limitations in sensitivity and selectivity since, although many approaches have been tried (Heger and Holm, 2000), they often provide inaccurate domain boundaries and alignments. For a description of a frequently used top-down approach, the reader is referred to a recent report of improvements to the ProDom database (Corpet et al., 2000). Bottom-up semiautomatic approaches curate individual families by establishing signi®cant similarities among a set of sequence regions. These regions are used to build multiple sequence alignments with due attention paid to domain boundaries and other aspects of alignment accuracy, such as gap placement and consistency with predicted secondary structure. Such high-quality sequence alignments are valuable in themselves, as with suitable algorithmic processing they can be used to sensitively search sequence databases and identify further more distantly related homologs. Domain families are then annotated with respect to known structure and function. These approaches can improve on automatic methods if problems associated with sequence errors arising from, for example, misassembly of genes from genomic sequences or sequence fragments are resolved by expert hand-curation. On the other hand, the top-down methods have the advantage of clustering a far greater proportion of sequence databases than semiautomatic methods. The two approaches need not be mutually exclusive. By using a top-down approach on sequences with domains de®ned by a bottom-up approach, it is possible to rapidly screen for novel protein domains (Doerks et al., 2002). In databases built using bottom-up approaches, any computational representation believed to be common to all members of a particular domain family can be used. This representation, in conjunction with appropriate searching software, should optimally be able to distinguish all true family members from the background noise of unrelated proteins stored in sequence databases. This is a challenging problem, tackled with varying degrees of sophistication by different approaches. At the most basic level, the representation can consist of a simple pattern of amino acids common to a particular domain. Such an approach is found in

82

RICHARD R. COPLEY ET AL.

many of the PROSITE database motifs (Hofmann et al., 1999). At the other extreme of sophistication are hidden Markov models (HMMs) and generalized pro®les. The latter are position-speci®c scoring tables calculated from multiple sequence alignments (Gribskov et al., 1987) and are formally equivalent to certain types of HMMs (Bucher et al., 1996). HMM pro®les have a strong theoretical basis for their treatment of gaps (insertion/deletion positions) and for providing robust estimators of the biological signi®cance of sequence similarities. Consequently, these are the representations of choice for the SMART (http:// smart.embl-heidelberg.de/), PFAM (http://www.sanger.ac.uk/Pfam/), and Prosite pro®le (http://www.isrec.isb-sib.ch/software/PFSCAN_form.html) databases. SMART and PFAM also provide additional predictions on their servers that are unrelated to domains. These resources use coiled-coil, signal peptide, transmembrane helix, and compositional bias prediction algorithms to provide added value to their service. Other such ``meta-sites'' will undoubtedly ¯ourish in the coming years. Already, the InterPro project (http://ebi.ac.uk/interpro) has integrated different domain databases and their respective search methods into a uni®ed tool for proteome annotation. Currently, the underlying databases represented are PFAM, PRINTS, PROSITE, ProDom, and SMART, although other databases are likely to be integrated at a later date. The derived InterPro database has already been used to annotate the genome sequences of Drosophila melanogaster (Adams et al., 2000) and Homo sapiens (Lander et al., 2001). A major advantage of InterPro is that it has allowed the pooling of labor-intensive annotation efforts. Thus, literature and biochemical data on a particular domain family can be easily shared irrespective of the computational techniques used for their identi®cation. Another recently provided meta-site is that of CDD (Conserved Domain Database and search service) provided by the National Center for Biotechnology Information (http://web.ncbi.nlm.nih.gov/Structure/ cdd/cdd.shtml). CDD uses SMART- and PFAM-derived multiple alignments, as well as additional alignments provided locally, to generate position-speci®c score matrices. These matrices may be compared against single sequences using a derivative of the popular BLAST suite of programs. The strengths of this approach are threefold: (i) the search algorithm is familiar to users of BLAST; (ii) results are comparable with known tertiary structures using Cn3D; and (iii) results partially complement those produced by PFAM and SMART since a different search algorithm is employed. On the other hand, the HMM-based methodology of PFAM and SMART signi®cantly outperforms that of BLAST for short repeats or domains, and the CDD alignment search set is a subset of the union of the PFAM and SMART set. Consequently, it is recommended

SEQUENCE ANALYSIS OF MULTIDOMAIN PROTEINS

83

that users of these resources search each of them to obtain maximum bene®t. Finally, it is important to emphasize that none of these resources will result in the detection of all homologs and none is guaranteed to discriminate against all nonhomologs. This is because there is always a small but ®nite chance that a nonhomologous sequence will be detected with a relatively high alignment score or, conversely, that a homologous sequence will be detected with a low score. A healthy degree of skepticism, therefore, is appropriate in interpreting the predictions of these resources. Domains are frequently categorized in terms of their structure and evolution, but rarely by function. The structural classi®cation resources such as SCOP (Lo Conte et al., 2000) and CATH (Pearl et al., 2000) are invaluable in deciding which domains possess the same fold and hence, implicitly, which are likely to be homologous. Similarly, homology classi®cation resources, such as SMART, PFAM, and ProDom, predict evolutionary relationships explicitly and structural similarities implicitly (Fig. 3). Although merging domain families into superfamilies is vital to an understanding of evolution, it is important to realize that it says nothing about function explicitly. Thus the broad evolutionary approach of looking for distant homologs must be complemented by attempts to make more precise functional predictions, if such resources are to confer the greatest bene®t to biology. Of course, for close evolutionary relationships, it is reasonable to assume some kind of functional relationship (Wilson et al., 2000). For a family of functionally diverse homologs, predicting function requires the use of more than one domain representation, such as a multiple alignment. Instead, one or both of two stratagems may be employed. First, a multiple alignment might be constructed that represents all branches of the superfamily tree. Partitioning into functional groups might be achieved by considering the conservation, or otherwise, of patterns of important amino acids that act as functional determinants (see below). Second, several multiple alignments that may each represent a major branch of the superfamily tree might be constructed. Function prediction relies on a homolog being more similar to sequence members of one alignment than it is to others from other alignments. These are complementary stratagems and one is not necessarily preferable to another. On the one hand, the pattern-based approach is useful only if the functional determinants for families have been experimentally derived. On the other hand, the multiple alignments approach is useful only if it is assumed that the most sequence-similar proteins possess the most similar functions. In this regard, it is emphasized that a difference of only a single residue, for example, that in an active site, between two

84

RICHARD R. COPLEY ET AL.

a Function driven sub-classification Protein kinase STYKc

Function

Tyr specific

Ser / Thr specific S TKc

TytKc

b Structure driven merging

DED

DEATH

CARD

Structure

DEATH domain superfamily

FIG. 3. (a) Divergent evolution can give rise to scenarious in which homologous domains can have different functions. If different functions are known initially, pro®les or HMMs can be created that re¯ect the speci®c subfunctions of family members. SMARTcontains both Ser/Thr- and Tyr-speci®c kinase HMMs. If a sequence is a strong

SEQUENCE ANALYSIS OF MULTIDOMAIN PROTEINS

85

sequences might dramatically alter function, even if the rest of the proteins' sequences are absolutely conserved. A striking example of this is the serine proteases of the venom of Crotalinae snakes (Deshimaru et al., 1996). These possess enzyme speci®cities that are akin to a broad range of mammalian serine proteases. Their sequences, however, are all similar to only a small subset of these serine proteases. It is likely that considerable selective pressures on these snakes have driven an accelerated evolution of these proteases' sequences. Neither is the converse case necessarily straightforward. If two proteins share similar functions and are known to be homologous, it does not necessarily mean that they have recently diverged. For instance, the Zn2 -peptidase superfamily shows independent evolution of N-deacylation and N-desuccinylation within separate lineages, pointing to functional convergence of homologous proteins (Makarova and Grishin, 1999). Automated prediction resources, such as SMART, currently rely on multiple alignments for predicting functions of superfamily members. This is most apparent for the superfamilies of Ras-like small GTPases and protein kinases. In the future, as annotation resources mature, it is likely that combinations of cross-linked multiple alignments and patterns will be used to partition superfamilies into functionally distinct sets. This will be applicable for almost all domains, the exceptions being domain homologs that are circularly permuted or that are inserted (Russell and Ponting, 1998) and others that contain small regions of signi®cant sequence similarity embedded in different nonhomologous contexts (Lupas et al., 2001).

III. FROM DOMAIN CLASSIFICATION TO DOMAIN CONTEXT In the absence of experimental information, useful clues to predict the functions of the constituent domains of a protein can be acquired from several different sources. First, however, it is instructive to distinguish between functions acting at different linear scales. Predicting domain function requires consideration of a domain's multiple alignment at the residue scale. Predicting protein function, on the other hand, may involve a synthesis of its domains' functions. Conversely, some insight into the functions of a domain family may be gleaned from consideration of coenough match to either of these, it receives a speci®c prediction; otherwise, it is classi®ed into the more generic STYKc class, corresponding to protein kinases with speci®city unassigned. (b) An alternative scenario. For the DED, DEATH, and CARD domains, homology is not apparent from sequence alone. Only with the determination of threedimensional structure is it possible to assign them as members of the same superfamily.

86

RICHARD R. COPLEY ET AL.

occurring domains within the context of the known function of a protein. Each of these directions is worth pursuing and undoubtedly will bring a greater functional insight into current domain collections in the near future. A. Zooming In: Residue Context and Functional Subtyping Since the 1980s it has been standard practice to assemble multiple alignments of superfamilies from as many family members as possible. This increases sequence variability and enables the identi®cation of conserved functional residues. From the 1990s, several small-scale attempts have been made to identify residues characteristic of families that all share a subset of the functions of the superfamily. It has been, however, only from the mid-1990s that automatic and large-scale methods have been developed for functional subtyping (e.g., Casari et al., 1995; Lichtarge et al., 1996; Sjolander, 1998), and further re®nements of such approaches continue to be published (Hannenhalli and Russell, 2000). Such methods typically attempt to correlate speci®c amino acids in a sequence alignment with particular groups of sequences (for instance, with the sequence groupings found in a phylogenetic tree). In the absence of structural information, these methods work best for enzyme families with conserved active sites and a number of known experimental constraints on different substrates. By contrast, most mobile regulatory domain families possess subtle binding determinants that are unable to be discriminated using sequence information alone. Integration of structure with sequence information, however, has led to functional subtyping for a few regulatory domain families, such as RA (Kalhammer et al., 1997), WW (Espanel and Sudol, 1999), src homology 2 (SH2) (Kimber et al., 2000), and pleckstrin homology (PH) (Isakoff et al., 1998) domain families. Three-dimensional structures, if available, provide an even greater potential for identifying functionally related subfamilies from among superfamily homologs. Structures allow predictions using features that cannot be quanti®ed from a multiple alignment, such as electrostatic potentials. This approach was used, for example, to identify a subfamily of classical PH domains unlikely to share the lipid-binding function ascribed to some PH domains (Blomberg and Nilges, 1997). B. Zooming Out: Domain Context within Proteins Valuable functional information may be extracted from the composition and order of domains in proteins. This arises from a basic premise that domains involved in similar cellular functions are more likely to be

SEQUENCE ANALYSIS OF MULTIDOMAIN PROTEINS

87

found together in a multidomain protein than are functionally dissimilar domains. An example of this is the detection of PH domains at the C-termini of 82% of guanine nucleotide exchange factors (GEFs) acting on Rho-type small GTPases. It is predicted that all such RhoGEFs possess C-terminal PH domains, but that for the 18% minority, they are not detectable computationally, as they lack statistically signi®cant sequence similarity. As suggested from the crystal structure of the RhoGEF/PH domain pair (Soisson et al., 1998), the functions of these two domains are likely to be highly cooperative. At the other extreme, some domains are found to be anti-correlated. This demonstrates that some domain types occur only in proteins found in speci®c cellular compartments and supports labels such as ``secreted'' or ``nuclear'' to be assigned to domain families. For example, the secreted domain KR is never found in tandem with a ``cytoplasmic'' PH domain. More surprisingly, some domains found in proteins targeted to the same cellular compartment can be anti-correlated. The most striking example of this relates to SH2 and PDZ (PSD-95, Dlg, ZO-1/2) domains that never co-occur in sequenced proteins. This is despite a relatively high co-occurrence of both domain types with SH3 domains: 24 and 9% of SH3 domain-containing proteins contain SH2 domains or PDZ domains, respectively. The reason for this striking negative correlation remains unknown, particularly as 16 examples are known of PDZ domains cooccurring with PTB (phosphotyrosine-binding) domains that, in some cases, have been shown to possess a function similar to that of SH2 domains. Other negative correlations are simpler to interpret. Two functionally antagonistic enzymes, namely, protein kinases and protein phosphatases, have, to date, not been found in the same protein. Similarly, WW and SH3 domains that both bind the similar polyproline-containing substrates are never found together. This last ®nding, however, is curious since 216 proteins that contain either two or more WW, or two or more SH3, domains are known. Finally, it appears that proteins with domains that bind phosphoserine or phosphothreonine (FHA, forkhead-associated domains) never contain domains that bind phosphotyrosine (SH2, PTB, and PTBI domains). This indicates that cytoplasmic signaling via phosphoserine or phosphothreonine occurs via pathways distinct from those signaling via phosphotyrosine. In 1999, the co-occurrence of domains was used to hint at the cellular functions of proteins (Marcotte et al., 1999a,b; Enright et al., 1999); an example of this is given in Fig. 4. These represented some of the ®rst instances where function was inferred without explicit use of homology. Such approaches result in considerable error rates due to protein modules that can be fused with many other domains. Consequently,

88

RICHARD R. COPLEY ET AL.

Yeast Der3p

R I N G

Human AMFR

R I N G

Human Tollip

C2

C U E

C. elegans Tollip (F25H2.1)

C2

C U FB E OX

C. elegans (F22G12.3)

FB OX

C. elegans (C33F10.8)

FB OX

C U E

Yeast Cuelp

C U E

L I P

L I P L I P

PTP

FIG. 4. The domain organizations of some CUE and LIP domain-containing proteins. Yeast Der3p/Hrd1p and Cue1p are proteins of the endoplasmic reticulum degradation pathway. As human autocrine motility factor receptor (AMFR) contains the same domain organization of a conceptual Der3p/Hrd1p and Cue1p fusion, it is proposed that Der3p/Hrd1p and Cue1p interact physically (Ponting, 2000). The C. elegans sequence most similar to human Tollip contains a C-terminal extension containing an F±box domain and an incomplete LIP domain. Over 190 LIP domains occur in at least 172 C. elegans hypothetical proteins, but have not been observed in other species' sequences; the functions of this domain remain unknown. LIP domains frequently co-occur with F-box domains and in one case (C33F10.8) a protein tyrosine phosphatase-like (PTP) domain.

these ``promiscuous'' domains were discarded in the prediction procedure (Marcotte et al., 1999a). Even when such fusions occur, there may be little that can be usefully said about function. Before the current round of systematic analyses Pekarsky et al. (1998) recognized the signi®cance of Caenorhabditis elegans and Drosophila fusion proteins (since termed ``Rosetta'' sequences) of Fhit, a human tumor suppressor gene of unknown function, and Nit, a member of a protein family homologous to nitrilases. As little was known about this latter family, little light was shed on the function of Fhit. The recently determined crystal structure of the NitFhit fusion protein has not made the situation much clearer (Pace et al., 2000).

SEQUENCE ANALYSIS OF MULTIDOMAIN PROTEINS

89

It remains unclear whether such approaches are truly general, in particular for proteins such as receptors that span different cellular compartments. For example, some receptor tyrosine kinases contain a kringle domain in their extracellular regions. Would such protocols predict common functions for intracellular tyrosine kinases and extracellular kringle-containing proteins, such as those of the blood coagulation pathway? Nevertheless, it is apparent that considerable functional constraints exist for domains to co-occur and that domain combinations are often very limited.

IV. GENOME-WIDE ANALYSIS: NEW QUALITY IN DOMAIN RESEARCH The availability of completely sequenced genomes offers the possibility of studying gene, protein, and domain evolution at the organismal level. Domain-based analyses confer several bene®ts to these studies. Not only can annotations be improved in various ways, but also the evolution of the many multidomain protein families can be traced. The latter requires a careful distinction between homology detection and orthology identi®cation. Orthology identi®cation is crucial for many approaches in comparative genome analysis and should be carried out at both the protein level and the domain level. As the postgenome era of cellular organisms is only a few years old, comparative analysis of entire genomes is in its early stages: domain analysis in this context has only recently been applied in a very rudimentary form. We expect this to change in the near future; the topics below represent only a few examples to indicate the different levels that can be exploited. A. Domains as a Tool to Aid Gene Prediction Domain analysis is particularly important for gene identi®cation and annotation, as well as for the detection of misassembled genes. For example, a C. elegans gene (C35B8.2), predicted by the Genome Sequencing Project to encode the orthologue of human Vav, contains a different domain architecture from Vav (Fig. 5a). Further analysis of the genome sequence reveals that this gene has been misassembled and that a corrected sequence does indeed contain all the domains seen for its human orthologue. Intriguingly, the Drosophila orthologue (CG7893) of this gene appears to be missing the more N-terminal SH3 domain with apparently no room in the genomic sequence to accommodate it (Dekel et al., 2000). Similar phenomena may not be associated with errors,

90

RICHARD R. COPLEY ET AL.

a Worm C35B8.2

CH

Fly

CH

Worm

CH

Human

CH

PH

RboGEF

PH

RboGEF

RboGEF

CL

PH

SH3

SH2

C

PH

RboGEF

SH2

CL

CL

SH3

SH3

SH2

SH2

SH3

SH3

b Toll receptor, Drosophila L L L L L R R R R R

L L L L R R R R

L R

L L L L R R R R

L L L L L R R R R R

LRR CT

LRR NT

L R

L R

LRR CT

LRR CT

TR

Best BLAST hit, C. elegans L R

L R

L R

L L L R R R

C07F11 Genomic DNA

W05D2

TR

FIG. 5. (a) Vav like proteins in C. elegans, D. melanogaster (SP-TREMBL Q9NHV9), and humans (P15498). The predicted C. elegans protein C35B8.2 lacks a C-terminal SH3 domain that is apparent in the genomic sequence. The Drosophila protein lacks the ®rst SH3 domain, and the absence of genomic DNA of suf®cient length suggests that this is not a failure to predict the domain. The ®rst SH3 domain in C. elegans is very divergent and not predicted by SMART. (b) The Drosophila Toll receptor (P08953) was used to identify the best match in the predicted proteins of C. elegans (wormpep id C07F11.1). This predicted protein shows the extracellular leucine-rich repeats (LRR) and associated domains, but does not contain a TIR domain. The protein is found at the end of a genomic DNA clone (C07F11). Analysis of the following clone (W05D2) using genewise revealed the presence of a previously unpredicted TIR domain.

SEQUENCE ANALYSIS OF MULTIDOMAIN PROTEINS

91

but with the complexities of alternatively spliced gene products (Black, 2000). Understanding the domain structure of proteins and an accurate gene prediction process go hand in hand. Gene prediction is greatly assisted by using homology information (i.e., sequence similarity), but in the case of multidomain proteins, similarity may be dif®cult to detect or may not extend over the full length of the protein. The TIR domain is found in a large family of Toll-like receptors in Drosophila and in humans (Fig. 5b). These proteins are believed to be involved in the innate immune response (Aderem and Ulevitch, 2000). The predicted C. elegans protein set (wormpep, http://www.sanger.ac.uk/Projects/C_elegans/wormpep/) at the date of this writing contained only a single example of a TIR domain in a molecular context different from that of the Toll-like receptors (Aravind et al., 1999). However, by identifying the C. elegans peptide with the best match to the extracellular portion of a Drosophila toll receptor and using sensitive software to identify matches between genomic DNA and protein HMMs (genewise, http://www.sanger.ac.uk/Software/Wise2), we were able to con®rm the presence of a previously unpredicted TIR domain (Fig. 5b). The details are important. The value of model organisms lies in their increased simplicity: in this case, only a single Tolllike receptor is present in C. elegans compared to multiple copies in Drosophila and humans. Thus, the phenotype (if any) of knocking out the C. elegans gene is unlikely to be obscured by functional redundancy of close copies. To understand the evolution of function and of multidomain proteins, artifacts arising due to sequence or annotation errors must be distinguished from the genuine products of domain shuf¯ing or acquisition. For example, the C. elegans sequence most similar to that of human Tollip is F25H2.1 (Fig. 4), yet the worm sequence contains an extra domain at its C-terminal end. It remains a matter of conjecture whether it represents an aberrant gene fusion of two distinct genes or whether the worm sequence is genuine. If genuine, then several questions arise. Are these human and C. elegans genes functionally equivalent? Can the function of one be inferred from the other? And, can one decide on a de®nition of whether such genes are ``equivalent'' or not? B. Orthology and Paralogy Equivalent genes in different species have been named orthologues (Fitch, 1970). The value of discriminating between genes that arose from speciation events (orthologues) and genes that arose from intragenomic

92

RICHARD R. COPLEY ET AL.

duplication events (paralogues) was not fully recognized until the mid1990s when the ®rst microbial genomes were sequenced and the extent of intragenomic duplication became apparent. For function prediction, the distinction is crucial: different members of multigene families may possess distinct functions. Indeed, for paralogues to persist in a genome, it is likely that there must be some distinction in function, with subtle changes in either expression or speci®city (Lynch and Conery, 2000, and references therein). Consequently, only orthologues should be used as excellent predictors for the transfer of functional information. In practice, eukaryotic paralogues are often useful in accurately predicting function. This is supported by many experimental observations of eukaryotic paralogues possessing overlapping functions (for example, Manley and Capecchi, 1997; Teglund et al., 1998; Tinsley et al., 1998). Although the concept of orthologues is clear, there are no foolproof methods for their identi®cation. One working de®nition of the orthologues, OA and OB , of two genomes A and B, are the genes for which OA is the most similar to OB in a search of B, and OB is the most similar to OA in a search of A. However, in many instances this de®nition is too simplistic. This is because there are often not one-to-one, but many-to-many relationships between related genes in different organisms arising from distinct gene duplication events in separate evolutionary lineages. Another complication arises from the fusion, ®ssion, deletion, or shuf¯ing of domains. We suggest that the concept of orthology be applicable to both domains and proteins in a similar manner to the current application of homologous domains and proteins. Thus the case might arise whereby orthologous domains in different species co-occur in different nonorthologous multidomain contexts. In this scenario, it is expected that the proteins differ in function despite their constituent orthologous domains possessing equivalent functions. Issues relating to orthology and paralogy are crucial for genome comparison and annotation. This is particularly true for the genomes of multicellular organisms since, with their larger genomes, they encode many duplicated genes, a high proportion of which encode multidomain proteins. However, due to the methodological problems described above, domain analysis in such situations is currently restricted to homology identi®cation and domain counting approaches (The C. elegans Sequencing Consortium, 1998; Rubin et al., 2000; Lander et al., 2001; Ponting et al., 1999). These have proved to be important in highlighting the frequency of lineage-speci®c domain expansions (Fig. 4, and below). However, it is clear that advances in orthology/paralogy identi®cation are required if more complex orthologous relationships are to be recognized. The use of domain analysis can help resolve more complex scenarios as outlined below.

SEQUENCE ANALYSIS OF MULTIDOMAIN PROTEINS

93

C. Comparative Analysis and Evolution of Function As complete genomes become available and orthologues are assigned, it becomes possible to trace the evolution of the domain structure of proteins. Processes of domain loss or domain gain must be reconciled with what is known of species phylogeny. Figure 6a illustrates a potential case of domain gain. The UPF0034 domain from Pfam (Bateman et al., 2000) appears fused with a double stranded RNA-binding motif (DSRM) in certain metazoan (human, Drosophila, and C. elegans) proteins. However, the apparent yeast orthologue of these proteins contains no DSRM domain. This case again illustrates the potential values and dif®culties of using domain fusion to predict function. The UPF0034 domain is poorly characterized, but is probably a phosphate-binding (b=a)8 barrel (Copley and Bork, 2000). The fusion with a DSRM domain suggests that its substrate may be RNA, providing a testable hypothesis where none was present before. Intriguingly, the yeast protein possesses a weak ability to suppress a defect in faulty mitochondrial tRNA Asp processing (Rinaldi et al., 1997). Within the eukaryotic crown group, Viridiplantae (i.e., plants) are believed to be less closely related to the metazoans than fungi (e.g., Baldauf et al., 2000). Thus if a domain combination is shared between plants and metazoa, but not with fungi, the most parsimonious explanation is that it has been altered in the lineage leading to fungi. Figure 6b shows such a situation, with the potential loss of SANT domains from an ancestral gene encoding both DnaJ and SANT domains. In cases where the organismal phylogeny is unknown, it is possible that shared derived domain combinations can be used as synapomorphies (i.e., if it is assumed that the evolution of that particular domain structure is monophyletic) to resolve the branching order of the species. It has been argued that Drosophila and C. elegans can be united in a clade of molting animals, known as ecdysozoa (Aguinaldo et al., 1997). However, some evidence from the domain structure of equivalent proteins found in humans Drosophila, and C. elegans is in con¯ict with this hypothesis (Lander et al., 2001). In the coming years, with the arrival of more complete genome sequences from crown-group eukaryotes, such questions will receive close attention. Evolution reworks the domain structure of proteins, modifying them by the addition and deletion of domains. A parallel evolutionary theme highlighted by comparative analysis is the expansion or contraction of particular domain families within phylogenetic lineages. Analysis of the complete genomes of C. elegans and D. melanogaster reveals striking examples of expansion of different domain families, particularly in the complement of extracellular proteins. As has previously been noted

94

RICHARD R. COPLEY ET AL.

Plants

UPF0034

Fungi

UPF0034

Metazoa

UPF0034

DS RM

(a)

S T

S T

S T

S T

(b)

DnaJ

UPF0034

UPF0034

S T DS RM

SANT

DSRM

FIG. 6. (a) Fusion of double-stranded RNA-binding Motif (DSRM) with UPF0034, an uncharacterized domain from PFAM (Accession No. PF01207). The sequences shown are predicted to be orthologous. Sequence accession numbers used: SP±TREMBL: C. elegans, Q9XWJ9; D. melanogaster, Q9VY45; S. cerevisiae, P53720; H. sapiens, Q9NX74. (b) Potential loss of SANT domains. Sequences: S. cerevisiae, P32527; H. sapiens, 060415; D. melanogaster, Q9VP77; C. elegans, Q94216; Arabidopsis thaliana, Q9SS16, Q9LHS5. The orthologous grouping of the common domains (i.e., UPF0034 or DnaJ) was, in both cases, supported by bootstrapping. However, the branching within the groups is less well supported. Accordingly, the tree represents the standard species phylogeny. Two additional explanations could also account for such apparent rearrangements of domain architecture: (1) invoking an ancestor with both domain structures, followed by multiple gene losses, and (2) aberrant protein prediction from DNA sequences.

(Rubin et al., 2000), the classical trypsin-like serine protease family (Tryp_SPc) is greatly expanded in Drosophila compared to C. elegans. This expanded family includes easter, snake, and gastrulation defective from the Toll pathway, crucial to dorsoventral patterning in Drosophila. Despite the importance of the pathway in Drosophila development, neither these serine proteases nor other components of the pathway have counterparts in C. elegans. Rapid expansion of catalytic sequence families such as this may be partially explained by the evolution and duplication of protein cascades, where similar proteins act on one another to amplify an initial signal (Caffrey et al., 1999). In addition to expansion, these domains provide an example of another general theme in domain evolution: an arrangement of domains unique to a particular phylogenetic lineage. Although they were initially reported to be single-domain proteins (Rubin et al., 2000), close inspection reveals that many Tryp_SPc

95

SEQUENCE ANALYSIS OF MULTIDOMAIN PROTEINS

domain-containing proteins (including snake and easter) have a conserved motif at the N-terminus (CLIP) that appears to be unique to the arthropods (Smith and De Lotto, 1992). A similar motif is also found in four copies in the Drosophila masquerade protein. Interestingly, here the serine protease domain is lacking essential catalytic residues and so is unlikely to function as an enzyme (Murugasu-Oei et al., 1995). In contrast to this relatively simple domain structure, mammalian serine proteases exhibit a greater architectural complexity (despite a lower number of proteases being encoded in the genome). This may be accounted for by far greater demands for regulational complexity related to their roles in, for instance, the regulation of digestion and blood clotting cascades (Fig. 7). Interesting as this may be, and shedding light on some of the most fundamental aspects of molecular evolution, these analyses are, hopefully, only a beginning. The ultimate goal must be to integrate these studies with function, phenotypes, and population dynamics if we are to truly succeed in understanding the molecular nature of biological change.

Human

Drosophila Tryp_SPc

T.

L L AA

FRI

RP T

RP T

RP T

SR

RP T

L A

L A

SEA

Tryp_SPc

SR

L A

SR

CLECT

PAN AP

CLECT

FN2 FRI

F2

KR

KR

LDLa

L A

MAM

MAM

CCP

CUB

BTB BTB

EGF

EG

GLA

GA

ChtBD2

Cht BD2

EGF CA

GPS

GPS

PAN_AP

PAN AP

HormR

Horm R

SEA

SEA

EGF_CA

FL

SR

Tryp_SPc

Tryp_SPc

KR

KR

KR

KR

Tryp_SPc

Tryp_SPc

KR

L L L L A A A A

L L L FRI A A A

SR

Tryp_SPc

Tryp_SPc

YWA

Tryp_SPc

CCP

FN1

EG R EG

Tryp_SPc

Tryp_SPc

BTB

RP T

L A

CUB

AP

EGE BOE GA CA itc

CUB

CLIP

AP

Tryp_SPc

GPS

BTB BTB

AP

AP

Tryp_SPc

CCP CCP CCP

Horm R

C. elegans

APPLE

MAM

F2

FRI

CLECT

CUB

Tryp_SPc

CUB

Cht Cht BD2 BD2

AP

FRI

SR Tryp_SPc TSP1 VWA

SR

Tryp_SPc

TL

VWA

FIG. 7. Trypsin-like serine proteases in C. elegans, D. melanogaster, and humans. The ®gure shows illustrative examples of the distinct domain co-occurrences with the serine protease-like domains in each species. The key shows the SMART (http:// smart.embl-heidelberg.de/) names of the co-occurring domains.

96

RICHARD R. COPLEY ET AL.

V. CONCLUSION Domain identi®cation and analysis are essential for understanding protein structure and evolution and provide the starting point from which comparative analysis of genomes can proceed. Although most of the widespread domain families have likely already been identi®ed, we are far from having a full catalogue of all domains. Such a complete catalogue would include representatives of all sequence families, within the unifying conceptual framework of a structural (and evolutionary) hierarchy, through which the emergence of function could be traced with increasing levels of precision. By applying what is already known, we can begin a systematic classi®cation and quanti®cation of evolutionary events. Linking such computational analyses to the coming waves of data from functional genomics projects, where information is generated for thousands of proteins at a time (e.g., Go Ènzcy et al., 2000; Fraser et al., 2000), will stand us in good stead as we attempt to move ever closer to an understanding of our nature. REFERENCES Adams, M. J., et al., (2000). Science 287, 2185±2195. Aderem, A., and Ulevitch, R. J. (2000). Nature 406, 782±787. Aguinaldo, A. M., Turbeville, J. M., Linford, L. S., Rivera, M. C., Garey, J. R., Raff, R. A., and Lake, J. A. (1997). Nature 387, 489±493. Altschul, S. F., and Gish, W. (1996). Methods Enzymol. 266, 460±480. Altschul, S. F., Madden, T. L., Schaffer, A. A., Zhang, J., Zhang, Z., Miller, W., and Lipman, D. J. (1997). Nucleic Acids Res. 25, 3389±3402. Andrade, M. A., Ponting, C. P., Gibson, T. J., and Bork, P. (2000). J. Mol. Biol. 298, 521±537. Aravind, L., Dixit, V. M., and Koonin, E. V. (1999). Trends Biochem. Sci. 24, 47±53. Baldauf, S. L., Roger, A. J., Wenk-Siefert, I., and Doolittle, W. F. (2000). Science 290, 972±977. Baron, M., Norman, D. G., and Campbell, I. D. (1991). Trends Biochem. Sci. 16, 13±17. Bateman, A., Birney, E., Durbin, R., Eddy, S. R., Howe, K. L., and Sonnhammer, E. L. (2000). Nucleic Acids Res. 28, 263±266. Black, D. L. (2000). Cell 103, 367±370. Blomberg, N., and Nilges, M. (1997). Fold. Des. 2, 343±355. Bork, P. (1992). Curr. Opin. Struct. Biol. 2, 413±421. Bork, P. (1996). Matrix Biol. 15, 311±312. Bork, P., and Bairoch, A. (1995). Trends Biochem. Sci. 20 (Poster Supplement C02). Bork, P., and Koonin, E. V. (1996). Curr. Opin. Struct. Biol. 6, 366±376. Bork, P., Schultz, J., and Ponting, C. P. (1997). Trends Biochem. Sci. 22, 296±298. Bucher, P., Karplus, K., Moeri, N., and Hofmann, K. (1996). Comput. Chem. 20, 3±23. Caffrey, D. R., O'Neill, L. A., and Shields, D. C. (1999). J. Mol. Evol. 49, 567±582. Callebaut, I., Cossart, P., and Dehoux, P. (1998). FEBS Lett. 441, 181±185. Casari, G., Sander, C., and Valencia, A. (1995). Nat. Struct. Biol. 2, 171±178. C. elegans Sequencing Consortium (1998). Science 282, 2012±2018.

SEQUENCE ANALYSIS OF MULTIDOMAIN PROTEINS

97

Copley, R. R., and Bork, P. (2000). J. Mol. Biol. 303, 627±641. Corpet, F., Servant, F., Gouzy, J., and Kahn, D. (2000). Nucleic Acids Res. 28, 267±269. Dekel, I., Russek, N., Jones, T., Mortin, M. A., and Katzav, S. (2000). FEBS Lett. 472, 99±104. Deshimaru, M., Ogawa, T., Nakashima, K., Nobuhisa, I., Chijiwa, T., Shimohigashi, Y., Fukumaki, Y., Niwa, M., Yamashina, I., Hattori, S., and Ohno, M. (1996). FEBS Lett. 397, 83±88. Doerks, T., Copley, R., Schultz, J., Ponting, C. P., and Bork, P. (2002). Genome Res. 12, 47±56. Doolittle, R. F. (1985). Trends Biochem. Sci. 10, 233±237. Enright, A. J., Iliopoulous, I., Kyripides, N. C., and Ouzounis, C. A. (1999). Nature 402, 86±90. Espanel, X., and Sudol, M. (1999). J. Biol. Chem. 274, 17284±17289. Fitch, W. M. (1970). Syst. Zool. 19, 99±113. Fraser, A. G., Kamath, R. S., Zipperlen, P., Martinez-Campos, M., Sohrmann, M., and Ahringer, J. (2000). Nature 408, 325±330. Gilbert, W. (1978). Nature 271, 501. Gilbert, W. (1985). Science 228, 823±824. Go, M. (1983). Proc. Natl. Acad. Sci. USA 80, 1964±1968. Go Ènczy, P., Echeverri, C., Oegema, K., Coulson, A., Jones, S. J. M., Copley, R. R., Duperon, J., Oegema, J., Brehm, M., Cassin, E., Hannak, E., Kirkham, M., Pichler, S., Flohrs, K., Goesson, A., Leidel, S., Alleaume, A-M., Martin, C., Ozlu, N., Bork, P., and Hyman, A. A. (2000). Nature 408, 331±336. Gouzy, J., Corpet, F., and Kahn, D. (1999). Comput. Chem. 23, 333±340. Gribskov, M., McLachlan, A. D., and Eisenberg, D. (1987). Proc. Natl. Acad. Sci. USA 84, 4355±4358. Hannenhalli, S. S., and Russell, R. B. (2000). J. Mol. Biol. 303, 61±76. Heger, A., and Holm, L. (2000). Prog. Biophys. Mol. Biol. 73, 321±337. Hofmann, K., Bucher, P., Falquet, L., and Bairoch, A. (1999). Nucleic Acids Res. 27, 215±219. Isakoff, S. J., Cardozo, T., Andreev, J., Li, Z., Ferguson, K. M., Abagyan, R., Lemmon, M. A., Aronheim, A., and Skolnik, E. Y. (1998). EMBO J. 17, 5374±5387. Janin, J., and Chothia, C. (1985). Methods Enzymol. 115, 420±430. Kalhammer, G., Bahler, M., Schmitz, F., Jockel, J., and Block, C. (1997). FEBS Lett. 414, 599±602. Karlin, S., and Altschul, S. F. (1990). Proc. Natl. Acad. Sci. USA 87, 2264±2268. Kimber, M. S., Nachman, J., Cunningham, A. M., Gish, G. D., Pawson, T., and Pai, E. F. (2000). Mol. Cell 5, 1043±1049. Lander, E. S., et al. (2001). Nature 409, 860±921. Lichtarge, O., Bourne, H. R., and Cohen, F. E. (1996). J. Mol. Biol. 257, 342±358. Lo Conte, L., Ailey, B., Hubbard, T. J., Brenner, S. E., Murzin, A. G., and Chothia, C. (2000). Nucleic Acids Res. 28, 257±259. Lupas, A. N., Ponting, C. P., and Russell, R. B. (2001), J. Struct. Biol. 134, 191±203. Lynch, M., and Conery, J. S. (2000). Science 290, 1151±1155. Makarova, K. S., and Grishin, N. V. (1999). J. Mol. Biol. 292, 11±17. Manley, N. R., and Capecchi, M. R. (1997). Dev. Biol. 192, 274±288. Marcotte, E. M., Pellegrini, M., Ng, H. L., Rice, D. W., Yeates, T. O., and Eisenberg, D. (1999a). Science 285, 751±753. Marcotte, E. M., Pellegrini, M., Thompson, M. J., Yeates, T. O., and Eisenberg, D. (1999b). Nature 402, 83±86. Mott, R. (2000). J. Mol. Biol. 300, 649±659. Murugasu-Oei, B., Rodrigues, V., Yang, X., and Chia, W. (1995). Genes Dev. 9, 139±154. Needleman, S. B., and Wunsch, C. D. (1970). J. Mol. Biol. 48, 443±453. Pace, H. C., Hodawadekar, S. C., Dragenscu, A., Huang, J., Bieganowski, P., Pekarsky, Y., Croce, C. M., and Brenner, C. (2000). Curr. Biol. 10, 907±917.

98

RICHARD R. COPLEY ET AL.

Patthy, L. (1985). Cell 41, 657±663. Patthy, L. (1987). FEBS Lett. 214, 1±7. Patthy, L., Trexler, M., Vali, Z., Banyai, L., and Varadi, A. (1984). FEBS Lett. 171, 131±136. Pearl, F. M., Lee, D., Bray, J. E., Silltoe, I., Todd, A. E., Harrison, A. P., Thornton, J. M., and Orengo, C. A. (2000). Nucleic Acids Res. 28, 277±282. Pearson, W. R. (1998). J. Mol. Biol. 276, 71±84. Pekarsky, Y., Campiglio, M., Siparashvili, Z., Druck, T., Sedkov, Y., Tillib, S., Draganescu, A., Wermuth, P., Rothman, J. H., Huebner, K., Buchberg, A. M., Mazo, A., Brenner, C., and Croce, C. M. (1998). Proc. Natl. Acad. Sci. USA 95, 8744±8749. Ponting, C. P. (2000). Biochem. J. 351, 527±535. Ponting, C. P., Aravind, L., Schultz, J., Bork, P., and Koonin, E. V. (1999). J. Mol. Biol. 289, 729±745. Ponting, C. P., and Russell, R. B. (2000). J. Mol. Biol. 302, 1041±1047. Rinaldi, T., Lande, R., Bolotin-Fukuhara, M., and Frontali, L. (1997). Curr. Genet. 31, 494±496. Rossmann, M. G., Moras, D., and Olsen, K. W. (1974). Nature 250, 194±199. Rubin, G. M., et al. (2000). Science 287, 2204±2215. Russell, R. B., and Ponting, C. P. (1998). Curr. Opin. Struct. Biol. 8, 364±371. Schultz, J., Milpetz, F., Bork, P., and Ponting, C. P. (1998). Proc. Natl. Acad. Sci. USA 95, 5857±5864. Schultz, J., Copley, R. R., Doerks, T., Ponting, C. P., and Bork, P. (2000). Nucleic Acids Res. 28, 231±234. Shapiro, L., and Scherer, P. E. (1998). Curr. Biol. 8, 335±338. Sjolander, K. (1998). I.S.M.B. 6, 165±174. Smith, C. L., and De Lotto, R. (1992). Protein Sci. 1, 1225±1226. Smith, T. F., and Waterman, M. S. (1981). J. Mol. Biol. 147, 195±197. Soisson, S. M., Nimnual, A. S., Uy, M., Bar-Sagi, D., and Kuriyan, J. (1998). Cell 95, 259±268. Teglund, S., McKay, C., Schuetz, E., van Deursen, J. M., Stravopodis, D., Wang, D., Brown, M., Bodner, S., Grosveld, G., and Ihle, J. N. (1998). Cell 93, 841±850. Tinsley, J., Deconinck, N., Fisher, R., Khan, D., Phelps, S., Gillis, J. M., and Davies, K. (1998). Nat. Med. 4, 1441±1444. Tordai, H., Banyai, L., and Patthy, L. (1999). FEBS Lett. 461, 63±67. Tsujishita, Y., and Hurley, J. H. (2000). Nat. Struct. Biol. 7, 408±414. Wilson, C. A., Kreychman, J., and Gerstein, M. (2000). J. Mol. Biol. 297, 233±249.

IDENTIFICATION OF TRANSIENTLY INTERACTING PROTEINS AND OF STABLE PROTEIN COMPLEXES Â RAPHIN BY BERTRAND SE Centre de GeÂneÂtique MoleÂculaire, 91198 Gif sur Yvette, France

I. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . II. Genetic Approaches . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A. Suppressor Analyses. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B. Synthetic Phenotypes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . C. Dosage Effects. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . III. Biological Methods. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A. Two-Hybrid Analyses and Related Strategies . . . . . . . . . . . . . . . . . . . . . . . . . B. Phage Display . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . IV. Biochemical Methods. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A. Biochemical Puri®cation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B. Co-immunoprecipitation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . C. Af®nity Puri®cation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . D. Native Gel Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . E. Cross-Linking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . F. Af®nity Selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . G. Far-Western . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . H. Protein Arrays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . V. Validation of Interactions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . VI. Characterization of Interacting Subunits and Dissection of Interaction Domains . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

99 100 100 102 102 103 103 105 105 105 107 108 109 109 110 112 112 113 114 115

I. INTRODUCTION Proteins were originally puri®ed and identi®ed by following their enzymatic properties. This operational mode of protein identi®cation led to the characterization of numerous enzymes. However, it also became clear that protein functions were not limited to enzymatic activities: abundant proteins were also assigned structural roles in cells and tissues. Overall, these various studies revealed that proteins are endowed with the capacity to interact speci®cally with high af®nity with a multitude of other compounds of varied chemical structure, including other proteins. Consistently, protein±protein interactions are involved in many, if not all, cellular functions. The propensity of proteins to interact speci®cally with other molecules is particularly adapted for the control of physiological parameters. Protein±protein interactions are indeed essential for controlling many processes including cell division, gene expression, and in the case of hormones and receptors, the coordination of activities of distant tissues. Interestingly, recent comparison of the genomic 99 ADVANCES IN PROTEIN CHEMISTRY, Vol. 61

Copyright 2003, Elsevier Science (USA). All rights reserved. 0065-3233/03 $35.00

100

Â RAPHIN BERTRAND SE

sequences revealed that increased biological complexity is associated with only a modest increase in gene number (Rubin, 2001). Complexity is therefore likely to result from the combinatorial use of protein interaction. Indeed, sequence analyses reveal a signi®cant overrepresentation of domains involved in protein interaction (protein kinases, cell adhesion modules, etc.) in more complex organisms (Chervitz et al., 1998). Deciphering the role of proteins identi®ed in genome sequencing projects will therefore require methods to identify interacting partners in transient or stable protein complexes. Furthermore, understanding protein interactions and their physiological roles may open the way for the design of drugs targeting interaction domains of interest. Given the widespread interest in this area, the goal of this chapter is to summarize the advantages and limitations of some of the methods currently available to identify protein interactions and protein complexes.

II. GENETIC APPROACHES While indirect, genetic approaches have several advantages in identifying interacting partners. First, genetic screens reveal only biologically relevant protein interactions. In addition, no preconceived knowledge about the interacting partner is required. Furthermore, it is important to note that this strategy is independent of the abundance of the interacting partners and that it may reveal transient interactions. Genetic approaches also suffer from several disadvantages, however. The main restriction comes from the requirement for a genetically tractable organism. Detecting protein interaction using this approach is therefore limited to some model systems. Nevertheless, given the conservation of the cellular machinery and the availability of genomic sequences, it is often possible to quickly transpose data obtained in a model system to an organism of interest. Another limitation of genetic screens is that effects may be indirect and, therefore, considerable effort may be required to demonstrate that a genetic interaction results from a direct protein interaction. In addition, because the read-out of this assay is a phenotype, genetic screens will require a starting mutant with appropriate properties. Various strategies are available to genetically identify interacting partners. These are detailed below. A. Suppressor Analyses Suppressor analysis is often easy to perform and has therefore been used extensively. The basic idea is that the starting mutant contains a protein carrying an alteration perturbing its association with a natural partner. By

PROTEIN INTERACTION

101

selecting a suppressor in the target protein, interaction between the two proteins will be restored together with the biological function (Fig. 1A). If appropriate selective forces can be applied to the model organism to recover (pseudo-) wild-type revertants, isolation of spontaneous or induced suppressors is often straightforward. Genetic analyses will reveal whether suppression is intragenic (i.e., possibly resulting from interaction between various regions of the target protein) or due to an extragenic mutation. While such extragenic suppressor mutations may affect a direct interaction partner, one should remember that suppression might also result from the activation of an alternative pathway, from the overexpression of the mutant proteins, or from (partial) correction of the original mutation by the translational machinery. Careful genetic and/or biochemical analyses of the suppressor are therefore often required to demonstrate that it directly interacts with the starting protein. It is also noteworthy that, given the complexity of protein interactions, only a

FIG. 1. Principle of the various genetic methods used to identify interacting partners: (A) suppressor screens, (B) aynthetic mutants, and (C) dosage effects.

102

Â RAPHIN BERTRAND SE

limited number of original mutations can be quantitatively corrected by a simple modi®cation(s) of a single interacting partner. While sometimes successful (e.g., Chapon and Legrain, 1992; Legrain et al., 1993), one must keep in mind that suppressor analysis may require a signi®cant investment for a limited output. B. Synthetic Phenotypes Screening for synthetic phenotypes is often more cumbersome than screening for suppressors. Indeed, one is looking for mutants enhancing the phenotype of the original mutation, an effect that can often not be selected for. Work-intensive observation of the mutant can therefore seldom be avoided. However, this is often counterbalanced by the higher selectivity of synthetic mutants. The logic behind synthetic phenotypes is that, under adequate conditions, the original mutation becomes ratelimiting in its biological pathway. Therefore, any mutation reducing even further the ¯ow rate through that pathway will produce a synthetic phenotype (Fig. 1B). In contrast, mutations in unrelated pathways will not act synergistically with the original mutant. Obviously, only a fraction of the mutants displaying a synthetic phenotype conform to this model. Again, careful genetic and/or biochemical analyses of synthetic mutants are required to demonstrate a direct interaction with the original protein. However, it is noteworthy that analysis of synthetic or enhancer mutants has proved very powerful in analyses such as that of the yeast nuclear pore (e.g., Wimmer et al., 1992) or of complex processes in Drosophila (e.g., Greaves et al., 1999). C. Dosage Effects Classical genetic screens can introduce only a limited number of changes in the target organism. To palliate this limitation, screens using arti®cial constructs altering protein expression levels have been developed. A simple example corresponds to multicopy suppression in yeast. As for classical suppressor screens, a simple genetic selection procedure is often available. However, in this case a library of plasmids carrying various DNA fragments covering the genome is introduced in the mutant strain. These plasmids are present at high copy numbers, increasing the expression of the encoded genes. If the original mutation reduces interaction between two partners, one expects that overexpression of one partner will, by a mass-action law, restore interaction at a level compatible with activity (Fig. 1C). Various strategies can be used to affect protein expression levels including the use of multicopy plasmids, the use of heterologous regulated promoters, and, in diploid organisms, the

PROTEIN INTERACTION

103

use of deletions. Obviously, dosage effects can be used to screen for either suppressor or synthetic phenotypes. Because it is technically simple, rapid, and often more direct than classical suppressor screens, multicopy suppression screening is often a method of choice in yeast (e.g., Liang et al., 1995). However, as for any genetic strategy, direct interaction between genes should be con®rmed using another method.

III. BIOLOGICAL METHODS A. Two-Hybrid Analyses and Related Strategies Among the methods making use of in vivo expression of proteins to identify protein interaction in a cellular context, two-hybrid screens are very popular. A major advantage of this strategy is that it is essentially equivalent to a genetic screen allowing for positive selection of interacting partners. The original two-hybrid system was developed in yeast (Fields and Song, 1989). The basic principle behind this strategy was derived from the molecular dissection of transcription activators. These experiments had revealed that several of these factors are modular proteins with independent DNA-binding and transcriptional activation domains, even though both domains are required to produce a biological response. Because no precise architectural relationship is required between the two domains, the physical link maintaining them associated to activate transcription (i.e., the polypeptide chain) can be replaced by a noncovalent interaction mediated by interacting proteins. In the two-hybrid assay, a speci®c DNA-binding domain is fused to a protein of interest to create a bait (Fig. 2). In addition, (potential) coding sequences are fused to a

FIG. 2. Principle of the standard two-hybrid assay. A bait protein is fused to a DNAbinding domain (DBD). A library of fusion between cellular proteins and a transcriptional activation domain (ACT) is introduced together with the bait protein in a yeast strain containing appropriate reporter genes harboring a cognate binding site for the DBD. The prey±ACT fusion interacting with the bait protein will activate Pol IImediated transcription, resulting in the production of the reporter transcript that can be selected for.

104

Â RAPHIN BERTRAND SE

transcription activation domain to create a library. The library is introduced in yeast strains expressing the bait protein. By using a reporter gene(s) under the control of a promoter-containing binding site(s) for the bait protein, one can select for expression of proteins carrying a transcriptional activation domain that associate with the bait. Activating plasmids are recovered, and the sequence of the polypeptide fused to the transcriptional activation domain is deduced from the corresponding DNA sequence. This method is extremely powerful because it allows for the detection of interaction between proteins without ever looking at the proteins themselves, all the work being done at the DNA level. The method is therefore relatively independent of the natural cellular concentration of the interacting partners. It is also noteworthy that partial proteins may be expressed (as a bait or in the library), allowing the identi®cation of interaction domains in modular proteins. As for any strategy, there are, however, limitations to the two-hybrid assay. First, even though the assay is performed in vivo, one should keep in mind that the yeast nucleus might not represent a natural environment for many heterologous proteins. Some proteins may lack a posttranscriptional modi®cation(s), others may interfere with transcription activation, while some may simply prevent entry of the fusion protein in this cellular compartment (e.g., membrane proteins). Depending on the source of the bait and prey proteins, they may also have to compete for interaction with endogenous yeast factors (in particular when using proteins from yeast or related species). Alternatively, some binary interactions may escape detection because of the absence of subunits stabilizing the interaction in the natural host (particularly for heterologous proteins). Finally, (partial) proteins may give a positive signal in the two-hybrid assay even though they are not expressed simultaneously in the natural host or ever located in the same subcellular compartment, producing thereby a background of erroneous interaction(s). Given the potential pitfall of the method, it appears therefore essential to use a second (direct) strategy to validate two-hybrid results. Nevertheless, because many of the molecular biology techniques used for two-hybrid assays can be automated, this strategy was adapted for large-scale screens at the genomic level. In this way, protein interaction maps for yeast (Uetz et al., 2000; Ito et al., 2001) and Helicobacter pylori (Rain et al., 2001) have recently been produced. Intriguingly, only limited overlaps between two yeast protein interaction maps have been detected (Ito et al., 2001), suggesting that results obtained in these screens are incomplete and/or contain some erroneous interactions. Several variants of the two-hybrid strategy have been developed. Some result solely from technical differences (e.g., use of different reporters, DNA-binding domains, fusion of the bait to the activation domain rather than to the DNA-binding domain), while others allow, for example, for

PROTEIN INTERACTION

105

counterscreening (Vidal and Legrain, 1999) or the use of PolIII rather than PolII as the target for activation (Marsolier and Sentenac, 1999). Finally, systems for detection have been developed that do not rely on transcriptional activation as a read-out but rather on protein degradation (the split ubiquitin strategy; Johnsson and Varshavsky, 1994) or activation of Ras at the yeast cellular membrane (Broder et al., 1998). Two-hybrid systems have also been adapted for other host cells including human cells (Maroun and Aronheim, 1999) and bacteria (Dove and Hochschild, 1998; Dove et al., 1997). These have the advantage of allowing protein production in a natural environment with appropriate posttranslational modi®cations (e.g., human proteins in human cells) and/or the use of larger libraries and a faster growing host (bacterial system). B. Phage Display Phage display shares with two-hybrid analysis the fact that, in a library, proteins are physically associated with genomes encoding them, allowing for genetic selection (Parmley and Smith, 1988). In this system, however, the fusion protein is expressed as a fusion with a subunit of a viral capsid. Unlike two-hybrid screens, interaction with the bait protein is selected in vitro rather than in vivo. The bait protein should therefore be available in suf®cient quantities for in vitro selection. Phages with af®nity for the bait protein can be easily recovered and reampli®ed, allowing for rapid successive cycles of selection. The identity of selected interacting partners is revealed by sequencing of the appropriate region of the phage genomes. Because selection is performed in vitro rather than in vivo, conditions can be precisely controlled and a counterselection step(s) can easily be implemented. A major limitation of this strategy is that phages are not always suitable for the expression of large proteins. Therefore, if phage display selections have been extensively used with a large library of random peptides, they are less commonly used than two-hybrid analysis for full-length proteins (see Wojnar et al., 2001, however).

IV. BIOCHEMICAL METHODS A. Biochemical Puri®cation Standard biochemical puri®cation can be used to identify subunits of a complex or protein associated with a protein of interest. In this case an appropriate biochemical or biological assay is used to follow an activity or the presence of the protein of interest (e.g., enzymatic assay, in vivo assays, Western blot). Several reviews have summarized the various procedures

106

Â RAPHIN BERTRAND SE

that can be used to purify the protein of interest (e.g., Deutscher, 1990). Optimal conditions need to be determined in an empirical manner (Fig. 3). This makes this strategy cumbersome and time-consuming, particularly if the assays are not straightforward, if the source of the starting material is limiting, and/or if the activity is labile. Furthermore, interacting subunits not required for the assay may be lost during the puri®cation. Polypeptides present in the puri®ed fraction(s) are usually identi®ed by mass spectrometry (Wilm, 2000) or Edman sequencing after gel electrophoresis. Alternatively, liquid chromatography may be used to fractionate peptides originating from a protein mixture before analysis by

FIG. 3. Puri®cation of complexes assembled in vivo. Depending on the material available, classical biochemical puri®cation (left), co-immunoprecipitation (center), or af®nity puri®cation (right) may be applied. The advantages and inconveniences of the various strategies are described in the text. The material puri®ed using these various strategies is characterized in the same way.

PROTEIN INTERACTION

107

mass spectrometry (Link et al., 1999). Finally, if some guess can be made about the interacting partners, those may be con®rmed by Western blotting (Fig. 3). B. Co-immunoprecipitation If antibodies with high af®nity and strong speci®city for the protein of interest are available, co-immunoprecipitation may be a method of choice to identify interacting partners (Harlow and Lane, 1988). In some ways, co-immunoprecipitation is identical to biochemical puri®cation except that a highly speci®c reagent with strong af®nity for the target protein is used for the puri®cation, reducing thereby the number of puri®cation steps and the need to develop a speci®c strategy (Fig. 3). This procedure requires, however, some prior knowledge of the protein of interest. During co-immunoprecipitation, extracts containing the target protein are incubated with antibodies and the bound fraction is recovered. Elution should be performed under suitable conditions or antibodies should be covalently coupled to the support to avoid the presence of large amounts of contaminating antibodies in the bound fraction. Factors present in this fraction are then analyzed and identi®ed as described above (Section IV,A; Fig. 3). Alternatively, Western blotting may be used if one can make some guess about the nature of the partners and if antibodies are available. The success of this strategy depends essentially on the quantity of the target protein and the quality of the antibodies. If it is often impossible to control the former parameter (except by choice of the best starting biological material); the use of af®nity-puri®ed or monoclonal antibodies may help with the latter parameter. However, one should always remember that high-quality antibodies are of little help if the major epitopes of the target protein are masked in the complex. Af®nity-puri®ed or monoclonal antibodies may also help to reduce nonspeci®c background. Alternative strategies to remove background include preclearing the lysate using unrelated or nonimmune antibodies before the speci®c antibodies to remove background, changing binding or washing buffer to implement more stringent conditions, removing cellular aggregates prior to the experiment, or selecting a different solid support that may be less prone to nonspeci®c interaction. However, it is noteworthy that nonspeci®c interactions may occur in the extract as factors from different subcellular compartments may associate even though they do not naturally interact in the cell. The use of cell fractions rather than whole-cell extracts can be used to reduce this source of background. Overall, co-immunoprecipitation is the method of choice to identify interacting proteins and proteins of interest if a suf®cient amount of (homogenous) starting material and high-quality antibodies are available.

108

Â RAPHIN BERTRAND SE

C. Af®nity Puri®cation Biochemical puri®cation and co-immunoprecipitation result in the puri®cation of endogenous proteins through different means. They require the development of speci®c procedures allowing the selective recovery of this protein or of speci®c reagents directed against the target factor (Fig. 3). Because each protein or protein complex is different, the development of speci®c tools can be time-consuming and may not be successful in every case. An alternative is therefore to add to the target protein some feature(s) allowing its puri®cation and/or precipitation under standard prede®ned conditions. This is usually done by fusing, at the DNA level, the sequence coding for a (short) peptide, a tag, to the coding sequence of the protein of interest (Fig. 3). This requires, obviously, knowledge of this protein. The resulting recombinant DNA is then introduced back into cells that express the fusion of the protein of interest and the tag. Speci®c reagents and procedures can then be used to recover the tagged protein and associated partners. The major advantage of the addition of a tag to the target protein for its puri®cation is that the reagents and procedure used are highly speci®c, rely on high-af®nity interaction, and have been extensively tested (Fig. 3). Indeed, for many tags, speci®c binding and elution procedures have been developed. However, the addition of a tag to the protein of interest may also have some drawbacks. First, the fusion protein may not be (fully) functional or may not interact with its natural partner. The tag itself may associate with endogenous cellular proteins, thereby generating some background. A second problem results from the necessity to reintroduce the construct encoding the tagged protein in the target cell and organism and to express stably or transiently the protein in the appropriate environment. While this is easy with some organisms, it may turn out to be tricky or impossible with others. It is probably preferable, whenever possible, to express the target protein at its natural level to avoid generating undesired phenotypes and to prevent its association with unnatural partners. Indeed, overexpression of a protein will, in most cases, not result in the coordinated overexpression of associated factors. Excess protein is therefore free to interact with other cellular factors such as chaperones or proteases that may obscure a less abundant natural partner(s) once the puri®ed fraction is fractionated on a gel (e.g., Swaf®eld et al., 1995). It is also noteworthy that avoiding overexpression also helps keep the protein in a soluble form. Despite these potential problems, af®nity puri®cation is a very powerful method. Over the years, several tags have been built and used to recover proteins overexpressed in Escherichia coli or other organisms. Several of these tags have been tested for the puri®cation of protein complexes in various host cells or organisms. Recently, combinations of tags allowing

PROTEIN INTERACTION

109

ef®cient puri®cation of proteins of low abundance under gentle conditions have been developed. Among them, the TAP tag, consisting of two IgG-binding modules of protein A from Staphylococcus aureus, a TEV protease cleavage site and a calmodulin-binding peptide, have proved extremely useful in yeast and mammalian cells (Rigaut et al., 1999; Westermarck et al., 2002). Furthermore, combinations of these tags allow the selective retrieval of subcomplexes or complexes sharing some common subunit(s) (Bouveret et al., 2000; Puig et al., 2001). Together with improved strategies for introduction of DNA in recipient cells or organisms and the development of mass spectrometry (Wilm, 2000), it appears therefore that af®nity puri®cation will provide an essential proteomic tool to decipher intricate protein interaction networks occurring in cells. D. Native Gel Analysis A rarely used method to characterize protein complexes is native gel electrophoresis (Schagger et al., 1994). This strategy can be used to fractionate complex protein mixtures in protein assemblies of different sizes. By performing a denaturing electrophoresis step in a second dimension, protein subunits present in a complex may be resolved. A major limitation of this technique for protein identi®cation is that it can be applied only if the protein of interest represents a signi®cant proportion of the sample loaded on the native gel. It is therefore useful only for abundant cellular proteins or for complexes that can easily be selectively enriched (e.g., through cellular fractionation). It is worth mentioning, however, that this strategy is well suited to the analysis of membraneassociated complexes. Indeed, it was extensively used to analyze complexes of the respiratory chain from mitochondrial membranes (Schagger et al., 1994). E. Cross-Linking Because protein interactions may be highly transient, cross-linking reagents able to create a covalent bond between neighboring proteins are used to link interacting subunits. A wide variety of chemical crosslinkers have been developed (Mattson et al., 1993). Those differ by the type of reactions mediating the attachment of the reagent to proteins, the mode of chemical activation (e.g., photoactivation), their speci®city toward proteins or other compounds, their solubility (e.g., in membranes), the possibility of spliting the cross-linker to recover the two interacting subunits, etc. Cross-linking reagents can be used with cells, with puri®ed organelles, or in vitro. In the last two cases, cross-linking reagents may be

110

Â RAPHIN BERTRAND SE

added chemically or during in vitro translation speci®cally to the protein of interest, limiting the generation of background aggregates relative to the cross-link of interest. In most cases, products of a cross-linking reaction need to be puri®ed using another method (e.g., immunoprecipitation, af®nity puri®cation, or biochemical fractionation) before they can be identi®ed unless the starting material is already highly puri®ed. Polypeptides present in the cross-linked species may be identi®ed by direct analysis (Edman sequencing, mass spectrometry) or by Western blotting (Sollner et al., 1992). In some cases, the exact location of the cross-link can be determined. The major problems associated with the use of crosslinking methods are the low yield and, depending on the strategy, the possible generation of a high level of background cross-linked species. Obviously, success will be affected by the quantity of the starting protein as well as by the speci®city and yield of the reaction. Unfortunately, the last two parameters vary in an unpredictable manner. Using partially puri®ed (and therefore less complex) biologically active systems also helps if the level of nonspeci®c aggregates is not too high. It is noteworthy that crosslinking has proved particularly useful for trapping highly transient interactions such as those occurring during the transport of proteins between various cellular compartments (Sollner et al., 1992). F. Af®nity Selection An alternative strategy to identify factors interacting with a protein of interest is to use af®nity selection. This requires reconstituting the interaction in vitro rather than relying on subunits preassembled in vivo (Fig. 4). Suf®cient quantities of the protein of interest are required (e.g., as a recombinant product overexpressed in bacteria). This protein is attached to a solid matrix, e.g., by covalent coupling. Extracts carrying putative interaction partners are incubated with this af®nity medium (Fig. 4). Unbound factors are removed by extensive washing before the release of factors selectively associated with the protein of interest by harsh treatment. Proteins present in the puri®ed fraction are analyzed and identi®ed as described for standard biochemical puri®cation (see Section IV,A, above). The high concentration of the protein of interest in this strategy is used to drive the association of interacting factors. However, because an interacting factor(s) may be present in low concentration, it may be masked in the selected fraction by abundant proteins displaying low af®nity for the protein of interest. Preclearing of the extract using selection on an unrelated protein column may help reduce the background. If available, however, it is therefore advisable to use a closely related speci®city control. This is possible, for example, if mutants of the protein of interest unable to interact with the putative partner have been

PROTEIN INTERACTION

111

FIG. 4. Principle of the af®nity selection strategy.

obtained using a functional assay. Comparing the pattern of proteins associated with the protein of interest and with the mutant form may reveal speci®c interaction partners even if a strong background of a nonspeci®cally associated factor is present. Similarly, if protein association depends on the presence of a speci®c ligand, comparing proteins associated in the presence or in the absence of the ligand provides a strict speci®city control (e.g., Kobayashi et al., 1998). Af®nity selection is particularly easy to perform if a large quantity of interacting partners is present. However, one should remember that interacting partners may already be modi®ed and/or stably bound to the protein of interest or other factor present in the extract and, therefore, unable to interact with the protein attached to the af®nity matrix (Fig. 4). This method may

112

Â RAPHIN BERTRAND SE

therefore be more appropriate for identifying factor(s) interacting transiently with the protein of interest. G. Far-Western An alternative method for identifying interacting partners by reconstituting the interaction in vitro is far-Western analysis. In contrast to af®nity selection, this does not require a large quantity of the protein of interest. The cell extract or fraction containing the putative partners is fractionated by gel electrophoresis. Proteins are then transferred onto a membrane and (partially) renatured. The membrane is subsequently incubated with the protein of interest under conditions favoring the reconstitution of interactions. The positions at which speci®c complexes are formed are then revealed. This is most often accomplished by using a (radioactively) labeled protein of interest (e.g., Coutavas et al., 1993). This method will not reveal directly the identity of the interacting partners, unless one can guess from their sizes and previous biological knowledge and/or one is analyzing a highly puri®ed fraction. In other situations, farWestern analysis provides an assay only to identify and follow interactions, and the biochemical fraction of the starting extracts or fraction is required for identi®cation of the interacting partner. A possibility for identifying the interacting subunit is to screen an expression library using the protein of interest as a probe. In this case, a (cDNA) library from the cells or tissues expressing the interactant (as revealed by far-Western) is prepared, usually in a phage vector. Plaques (or colonies) are prepared and transferred to a solid support. This membrane is then probed with the (radioactively) labeled protein of interest. Plaques or colonies expressing the interacting protein(s) are then puri®ed. The nature of the interacting protein(s) is then determined by sequencing the corresponding clones. Obviously, this strategy is not suitable for the analysis of interactions requiring more than two subunits. Furthermore, depending on the method used to prepare the proteins, molecules serving as the probe and/or present in the library may lack essential modi®cations. This strategy may, nevertheless, identify transiently interacting partners and has proved useful on many occasions (e.g., Coutavas et al., 1993). H. Protein Arrays An emerging strategy for identifying protein interaction is the use of protein arrays. The principle relies on the attachment of a wide variety of proteins (e.g., all proteins of a given organism or organelle) to a solid support. Small quantities of each protein are spotted at prede®ned positions on a regular array format. The technology being currently de-

PROTEIN INTERACTION

113

veloped, the mode of production of these proteins, the spotting conditions, the nature of the solid support, and all other parameters are being investigated and optimized. Once the protein array is built, it can be used for testing several functions of proteins including activity and/or interaction with ligands including other proteins. Preliminary descriptions of systems allowing the identi®cation of protein±protein interactions have been reported in the literature (e.g., MacBeath and Schreiber, 2000). Methodologically, a protein array is very similar to a far-Western analysis as, in both cases, proteins are immobilized on a solid support before being assayed. An important difference is that in the protein array, the nature of the protein present at a given location is most often known from its coordinates. Also, proteins present at each position in an array may be in a native folded state and/or multimeric (depending on the mode of production). This should in principle facilitate the detection of interacting partners. Because the protein array technology is relatively new, its potential for identifying interacting partners will certainly improve dramatically in the near future.

V. VALIDATION OF INTERACTIONS Because none of the methods used to identify protein interaction is foolproof, one should seriously consider using a second method to validate whether an interaction detected in an experiment really occurs in vivo. This is especially important if the interaction has been detected only indirectly (e.g., through genetic means) or in an arti®cial system (e.g., in the two-hybrid assay). However, even if two polypeptides cofractionate during biochemical puri®cation, one should ascertain that they interact in vivo. One should also remember that the puri®ed material may be heterogenous and that a given polypeptide may be present in several complexes harboring speci®c subunits. In principle, any strategy used to identify protein interaction (see Sections II±IV above) can be used to validate the existence of an interaction obtained with another strategy. However, in some circumstances, additional methods can also be used (see below). The choice of the most appropriate method will depend primarily on the tools and material available. However, one should also take into account other parameters including the strength of the experiment (e.g., biological evidence for interaction in vivo versus in vivo reconstitution of the protein assembly), evidence for a direct interaction (use of puri®ed systems), and the possibility of deriving additional information (e.g., de®nition of interacting domains).

114

Â RAPHIN BERTRAND SE

Co-immunoprecipitation is the method of choice to validate a protein interaction identi®ed with another strategy because it is biologically relevant, easy to implement, and sensitive if high-quality antibodies are available. Furthermore, one may obtain rough quantitative parameters related to the interaction (fraction of protein interacting under given conditions and/or subunit stoichiometry). If antibodies are not available, tagged protein may be used, especially if some constructs have been prepared for af®nity puri®cation. In both cases, cross-linking may be used to stabilize the interaction. In principle, protein puri®cation provides an alternative strategy to validate the biological relevance of a protein interaction identi®ed by other means. However, it is seldom used given its complexity compared to other strategies (e.g., co-immunoprecipitation). Furthermore, transiently interacting partners may well be lost during the procedure. Other methods can be used to ascertain the existence of the putative interaction. While they are not biologically relevant (e.g., far-Western, af®nity selection, or two-hybrid assay) they may provide evidence for a direct interaction (e.g., far-Western, af®nity selection) and/or provide an easy way to map interacting domains (e.g., two-hybrid assay). Alternative strategies that may help to validate the biological role of a putative interaction include biocomputing analysis of the protein sequence that may reveal the presence of domains of types previously known to interact (Bateman and Birney, 2000; Ponting et al., 2000). While this may appear trivial, the importance of an in-depth analysis of the protein sequence should not be underestimated. In vivo analysis of the two interacting partners by microscopy can also be used to support or validate an interaction. Obviously, if two polypeptides interact in vivo, one expects that a fraction of the two partners will be located in the same subcellular compartment(s). However, through ¯uorescence resonance energy transfer one can also obtain evidence supporting a direct, biologically relevant interaction in vivo even if the system is highly dynamic (e.g., Damelin and Silver, 2000).

VI. CHARACTERIZATION OF INTERACTING SUBUNITS AND DISSECTION OF INTERACTION DOMAINS Identi®cation of a complex involving several polypeptides raises several questions: What are the subunits in direct contact? What is the stoichiometry of the interaction? What is the nature of the fragment of these proteins mediating the contacts? Answering these questions may provide some tools for understanding the biological function of the identi®ed interactions, e.g., through the construction of dominant-negative mutants. Again, the various

PROTEIN INTERACTION

115

strategies enumerated above may be used for such analyses. However, novel techniques are being developed to facilitate these studies. These techniques include mass spectrometry, which has recently been used to characterize the mass (and therefore stoichiometry) and subunit association in large complexes (Rostom et al., 2000; Rostom and Robinson, 1999). Identifying interacting domains may be a rewarding task. Indeed, while interaction may require the complete polypeptide chain to build the correct fold, it is not rare that independent protein modules contribute to the various activities of a protein, especially in eukaryotes. If the original interaction can be detected using the two-hybrid assay, this system is well suited to this purpose. Biocomputing analysis is also often very informative (Bateman and Birney, 2000). However, detailed understanding will ultimately be obtained using biophysical methods (e.g., using plasmon surface resonance) and structural determination by Xray crystallography and/or NMR.

ACKNOWLEDGMENTS ÂdeÂric Gabriel for critically reading the I thank Fabienne Mauxion, Erwin van Dijk, and Fre Âdicale and an manuscript. This work was supported by the Fondation pour la Recherche Me ATIPE from the CNRS.

REFERENCES Bateman, A., and Birney, E. (2000). Searching databases to ®nd protein domain organization. Adv. Protein Chem. 54, 137±157. Âraphin, B. (2000). A Sm-like Bouveret, E., Rigaut, G., Shevchenko, A., Wilm, M., and Se protein complex that participates in mRNA degradation. EMBO J. 19, 1661±1671. Broder, Y. C., Katz, S., and Aronheim, A. (1998). The ras recruitment system, a novel approach to the study of protein±protein interactions. Curr. Biol. 8, 1121±1124. Chapon, C., and Legrain, P. (1992). A novel gene, spp91±1, suppresses the splicing defect and the pre-mRNA nuclear export in the prp9±1 mutant. EMBO J. 11, 3279±3288. Chervitz, S. A., Aravind, L., Sherlock, G., Ball, C. A., Koonin, E. V., Dwight, S. S., Harris, M. A., Dolinski, K., Mohr, S., Smith, T., Weng, S., Cherry, J. M., and Botstein, D. (1998). Comparison of the complete protein sets of worm and yeast: Orthology and divergence. Science 282, 2022±2028. Coutavas, E., Ren, M., Oppenheim, J. D., D'Eustachio, P., and Rush, M. G. (1993). Characterization of proteins that interact with the cell-cycle regulatory protein Ran/TC4. Nature 366, 585±587. Damelin, M., and Silver, P. A. (2000). Mapping interactions between nuclear transport factors in living cells reveals pathways through the nuclear pore complex. Mol. Cell 5, 133±140. Deutscher, M. (1990). ``Guide to Protein Puri®cation.'' Academic Press, San Diego, CA.

116

Â RAPHIN BERTRAND SE

Dove, S. L., and Hochschild, A. (1998). Conversion of the omega subunit of Escherichia coli RNA polymerase into a transcriptional activator or an activation target. Genes. Dev. 12, 745±754. Dove, S. L., Joung, J. K., and Hochschild, A. (1997). Activation of prokaryotic transcription through arbitrary protein±protein contacts. Nature 386, 627±630. Fields, S., and Song, O. (1989). A novel genetic system to detect protein±protein interactions. Nature 340, 245±246. Greaves, S., Sanson, B., White, P., and Vincent, J. P. (1999). A screen for identifying genes interacting with armadillo, the Drosophila homolog of beta-catenin. Genetics 153, 1753±1766. Harlow, E., and Lane, D. (1988). ``Antibodies: A Laboratory Manual.'' Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY. Ito, T., Chiba, T., Ozawa, R., Yoshida, M., Hattori, M., and Sakaki, Y. (2001). A comprehensive two-hybrid analysis to explore the yeast protein interactome. Proc. Natl. Acad. Sci. USA 98, 4569±4574. Johnsson, N., and Varshavsky, A. (1994). Split ubiquitin as a sensor of protein interactions in vivo. Proc. Natl. Acad. Sci. USA 91, 10340±10344. Kobayashi, K., Kuroda, S., Fukata, M., Nakamura, T., Nagase, T., Nomura, N., Matsuura, Y., Yoshida-Kubomura, N., Iwamatsu, A., and Kaibuchi, K. (1998). p140Sra-1 (speci®cally Rac1-associated protein) is a novel speci®c target for Rac1 small GTPase. J. Biol. Chem. 273, 291±295. Legrain, P., Chapon, C., and Galisson, F. (1993). Interactions between PRP9 and SPP91 splicing factors identify a protein complex required in prespliceosome assembly. Genes Dev. 7, 1390±1399. Liang, C., Weinreich, M., and Stillman, B. (1995). ORC and Cdc6p interact and determine the frequency of initiation of DNA replication in the genome. Cell 81, 667±676. Link, A. J., Eng, J., Schieltz, D. M., Carmack, E., Mize, G. J., Morris, D. R., Garvik, B. M., and Yates, J. R. (1999). Direct analysis of protein complexes using mass spectrometry. Nat. Biotechnol. 17, 676±682. MacBeath, G., and Schreiber, S. L. (2000). Printing proteins as microarrays for highthroughput function determination. Science 289, 1760±1763. Maroun, M., and Aronheim, A. (1999). A novel in vivo assay for the analysis of protein± protein interaction. Nucleic Acids Res. 27, e4. Marsolier, M. C., and Sentenac, A. (1999). RNA polymerase III-based two-hybrid system. Methods Enzymol. 303, 411±422. Mattson, G., Conklin, E., Desai, S., Nielander, G., Savage, M. D., and Morgensen, S. (1993). A practical approach to crosslinking. Mol. Biol. Rep. 17, 167±183. Parmley, S. F., and Smith, G. P. (1988). Antibody-selectable ®lamentous fd phage vectors: Af®nity puri®cation of target genes. Gene 73, 305±318. Ponting, C. P., Schultz, J., Copley, R. R., Andrade, M. A., and Bork, P. (2000). Evolution of domain families. Adv. Protein Chem. 54, 185±244. Puig, O., Caspary, F., Rigaut, G., Rutz, B., Bouveret, E., Bragado-Nilsson, E., Wilm, M., and Âraphin, B. (2001). The Tandem Af®nity Puri®cation (TAP) method: A general procedSe ure of protein complex puri®cation. Methods 24, 218±229.. Rain, J. C., Selig, L., De Reuse, H., Battaglia, V., Reverdy, C., Simon, S., Lenzen, G., Petel, F., Wojcik, J., Schachter, V., Chemama, Y., Labigne, A., and Legrain, P. (2001). The protein± protein interaction map of Helicobacter pylori. Nature 409, 211±215. Âraphin, B. (1999). A Rigaut, G., Shevchenko, A., Rutz, B., Wilm, M., Mann, M., and Se generic protein puri®cation method for protein complex characterization and proteome exploration. Nat. Biotechnol. 17, 1030±1032.

PROTEIN INTERACTION

117

Rostom, A. A., Fucini, P., Benjamin, D. R., Juenemann, R., Nierhaus, K. H., Hartl, F. U., Dobson, C. M., and Robinson, C. V. (2000). Detection and selective dissociation of intact ribosomes in a mass spectrometer. Proc. Natl. Acad. Sci. USA 97, 5185±5190. Rostom, A. A., and Robinson, C. V. (1999). Disassembly of intact multiprotein complexes in the gas phase. Curr. Opin. Struct. Biol. 9, 135±141. Rubin, G. M. (2001). The draft sequences. Comparing species. Nature 409, 820±821. Schagger, H., Cramer, W. A., and von Jagow, G. (1994). Analysis of molecular masses and oligomeric states of protein complexes by blue native electrophoresis and isolation of membrane protein complexes by two-dimensional native electrophoresis. Anal. Biochem. 217, 220±230. Sollner, T., Rassow, J., Wiedmann, M., Schlossmann, J., Keil, P., Neupert, W., and Pfanner, N. (1992). Mapping of the protein import machinery in the mitochondrial outer membrane by crosslinking of translocation intermediates. Nature 355, 84±87. Swaf®eld, J. C., Melcher, K., and Johnston, S. A. (1995). A highly conserved ATPase protein as a mediator between acidic activation domains and the TATA-binding protein. Nature 374, 88±91. Uetz, P., Giot, L., Cagney, G., Mans®eld, T. A., Judson, R. S., Knight, J. R., Lockshon, D., Narayan, V., Srinivasan, M., Pochart, P., Qureshi-Emili, A., Li, Y., Godwin, B., Conover, D., Kalb¯eisch, T., Vijayadamodar, G., Yang, M., Johnston, M., Fields, S., and Rothberg, J. M. (2000). A comprehensive analysis of protein±protein interactions in Saccharomyces cerevisiae. Nature 403, 623±627. Vidal, M., and Legrain, P. (1999). Yeast forward and reverse `n'-hybrid systems. Nucleic Acids Res. 27, 919±929. Westermarck, J., Weiss, C., Saffrich, R., Kast, J., Musti, A. M. Wessely, M., Ansorge, W., Âraphin, B., Wilm, M., Valdez, B. C., and Bohmann, D. (2002). The DEXD/H-box RNA Se helicase RHII/Gu is a co-factor for c-jun-activated transcription. EMBO J. 21, 451±460. Wilm, M. (2000). Mass spectrometric analysis of proteins. Adv. Protein Chem. 54, 1±30. Wimmer, C., Doye, V., Grandi, P., Nehrbass, U., and Hurt, E. C. (1992). A new subclass of nucleoporins that functionally interact with nuclear pore protein NSP1. EMBO J. 11, 5051±5061. Wojnar, P., Lechner, M., Merschak, P., and Redl, B. (2001). Molecular cloning of a novel Lipocalin-1 interacting human cell membrane receptor (LIMR) using phage-display. J. Biol. Chem. 3, e3.

This Page Intentionally Left Blank

MOLECULAR RECOGNITION IN ANTIBODY±ANTIGEN COMPLEXES BY ERIC J. SUNDBERG AND ROY A. MARIUZZA Center for Advanced Research in Biotechnology, University of Maryland Biotechnology Institute, Rockville, Maryland 20850

I. II. III. IV. V.

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Structure of Antibody±Antigen Interfaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Antibody Cross-Reactivity and Molecular Mimicry . . . . . . . . . . . . . . . . . . . . . . Thermodynamic Mapping of Antigen±Antibody Interfaces. . . . . . . . . . . . . . . Dissection of Binding Energetics in Antigen±Antibody Interfaces Using Double-Mutant Cycles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A. The FvD1.3±FvE5.2 Complex . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B. The FvD1.3±HEL Complex. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . VI. Accommodation of Mutations in Antigen±Antibody Interfaces . . . . . . . . . . . . VII. Functional Roles for Protein Plasticity in Antigen Recognition . . . . . . . . . . . . A. Af®nity Maturation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B. Induced Fit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . C. Beyond the Af®nity Ceiling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . VIII. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

119 121 125 132 136 136 141 144 148 148 151 155 156 157

With the numerous detailed molecular descriptions of antibody±antigen interfaces, the structural study of these molecular interactions has evolved from an attempt to understand immunological function to their use as model systems for protein±protein interactions. In this chapter, we describe the structural aspects common to antibody±antigen interfaces and discuss the roles they may play in antibody cross-reactivity and molecular mimicry. More detailed analysis of these interfaces has required the marriage of structural studies with extensive mutagenesis and thermodynamic analysis efforts. Here, we discuss the thermodynamic mapping of interfaces for two model antibody±antigen complexes, including the identi®cation of thermodynamic hot spots in binding and the various mechanisms used to accommodate interface mutations. We also discuss the functional roles for protein plasticity in antigen recognition, including the entropic control of antibody af®nity maturation and the use of induced ®t mechanisms of different types and to varying degrees by mature antibodies in binding their speci®c antigens. I. INTRODUCTION The ability of proteins to form speci®c, stable complexes with other proteins is fundamental to most cellular processes, including signal 119 ADVANCES IN PROTEIN CHEMISTRY, Vol. 61

Copyright 2003, Elsevier Science (USA). All rights reserved. 0065-3233/03 $35.00

120

ERIC J. SUNDBERG AND ROY A. MARIUZZA

transduction, vesicle transport, and cytoskeletal remodeling. Indeed, detailed knowledge of the structural basis of molecular recognition is an essential element in understanding protein function, since virtually all proteins act by combining with other molecules, be they small substrates, nucleic acids, or other proteins. Antibodies may be regarded as products of a protein engineering system developed by nature for the generation of a virtually unlimited repertoire of complementary molecular surfaces and, as such, constitute an excellent model for elucidating the principles governing macromolecular recognition. Antibody molecules are homodimers composed of two identical polypeptide chains of approximately 450 amino acids (the heavy or H chains) covalently linked through disul®de bridges to two identical polypeptide chains of about 250 residues (the light or L chains). Based on amino acid sequence comparisons, the H and L chains may be divided into N-terminal variable (V) and C-terminal constant (C) portions. Each H chain contains four domains (VH , CH 1, CH 2, CH 3) of two anti-parallel b-sheets, while each L chain consists of two such domains (VL , CL ). These b-sheet domains are structurally very similar and hence have been termed the ``immunoglobulin fold'' (Amzel and Poljak, 1979). Each of the VH and VL domains contains three segments, or loops, that connect the b-strands and are highly variable in length and sequence among different antibodies. These so-called complementarity determining regions (CDRs) lie in close spatial proximity on the surface of the V domains and determine the conformation of the combining site. In this way, the CDRs confer speci®c binding activity to the antibody molecule. The central paradigm of antigen±antibody recognition is that the three-dimensional structure formed by the six CDRs recognizes and binds a complementary surface (epitope) on the antigen. X-ray crystallographic studies of antigen±antibody complexes involving protein antigens have provided much valuable information on the molecular architecture of protein±protein interfaces, including the identity of contacting residues, the amount of buried surface area, the number and type of hydrogen bonds, and the magnitude of conformational changes associated with complex formation. However, the basic principles governing antigen±antibody and protein±protein interactions have remained elusive (Chothia and Janin, 1975; Jones and Thornton, 1996; Bogan and Thorn, 1998; Janin, 1999; Lo Conte et al., 1999; Sundberg and Mariuzza, 2000), with important fundamental problems relating to the recognition process still to be solved. What are the relative contributions of hydrophobicity, surface complementarity, and hydrogen bonding to the energetics and mechanism of binding? To what extent do the strengths of individual bonding interactions depend on their local environment and overall location in the interface? What is the role of solvent in complex

MOLECULAR RECOGNITION IN ANTIBODY±ANTIGEN COMPLEXES

121

stabilization? Is water simply displaced from the contacting surfaces or can it also serve to optimize contacts between antigen and antibody? What is the role of conformational ¯exibility, or structural plasticity, in antigen± antibody recognition? Is productive binding mediated by a distinct subset of combining site residues or are complex cooperative interactions involving both contacting and noncontacting residues responsible for the observed af®nities? What determines whether an interface residues is a so-called ``hot spot'' for ligand binding (i.e., a residue that contributes a disproportionately large fraction of the binding free energy)? How are potentially disruptive amino acid changes in the interface (for example, those that create ``holes'') accommodated? What is the structural basis of af®nity maturation, whereby somatic mutations in antibody genes that confer increased af®nity for antigen are selected? Finally, is it possible to predict ab initio the effects of a given amino acid substitution on antibody af®nity and speci®city? In addition to their importance in understanding the physical basis of immunological recognition, these issues are of general relevance to protein±protein association processes. In this chapter, we describe recent attempts to progress from purely anatomical descriptions of antigen±antibody interfaces obtained from X-ray crystallography to a detailed understanding of how structural features contribute to the af®nity and speci®city of the binding reactions.

II. STRUCTURE OF ANTIBODY±ANTIGEN INTERFACES In attempts to determine some basic rules for antigen recognition, a number of studies have sought to elucidate patterns in the amino acid composition of the antibody combining site (Kabat et al., 1977; Padlan, 1990; Janin and Chothia, 1990; Lea and Stuart, 1995; Davies and Cohen, 1996). These studies have shown that antibody±antigen interfaces are signi®cantly richer in aromatic residues, particularly tyrosine and tryptophan, than the average protein surface (Lo Conte et al., 1998). They are depleted in the charged residues aspartate, glutamate, and lysine, but are enriched in arginine. The relative abundance of arginine and aromatic residues in the antibody paratope should not be completely unexpected, however, as an analysis of alanine mutations on binding energies (Bogan and Thorn, 1998) has shown that thermodynamic hot spots in binding are weighted toward these amino acid types, perhaps because they are capable of making multiple favorable interactions. For example, tyrosine offers a large hydrophobic surface, aromatic p-interactions, and the hydrogen bonding potential of its hydroxy group, while arginine has three hydrophobic methylene carbon atoms and a guanidinium moiety that can contribute hydrogen bonds and a salt bridge.

122

ERIC J. SUNDBERG AND ROY A. MARIUZZA

The three-dimensional structures of approximately 30 complexes between antibodies and various protein antigens have been determined in recent years, including hen egg white lysozyme (HEL) (Amit et al., 1986; Sheriff et al., 1987; Padlan et al., 1989; Chitarra et al., 1993; Braden et al., 1994; Kondo et al., 1999; Li et al., 2000), in¯uenza virus neuraminidase (Colman et al., 1987; Tulip et al., 1992; Malby et al., 1994), horse cytochrome c (Mylvaganam et al., 1998), human tissue factor (Huang et al., 1998), Escherichia coli histidine phosphocarrier protein (Prasad et al., 1993), staphylococcal nuclease (Bossart-Whitaker et al., 1995), in¯uenza virus hemagglutinin (Bizebard et al., 1995), human vascular endothelial growth factor (Muller et al., 1998), T cell receptors (Housset et al., 1997; Wang et al., 1998), and antibodies bearing idiotypic determinants (Bentley et al., 1990; Ban et al., 1994; Evans et al., 1994; Fields et al., 1995), plus an ever-expanding database of anti-hapten, -peptide, -nucleic acid, and -carbohydrate antibody±antigen complexes. These studies have permitted a detailed description of antigenic determinants and of antibody combining sites. In addition, they have provided important information on the general characteristics of protein±protein interfaces (Jones and Thornton, 1996; Bogan and Thorn, 1998; Lo Conte et al., 1999). Although CDR loops are hypervariable and confer binding speci®city to the antibody, it is not necessary that all six CDR loops interact with a given antigen. In fact, in a small sampling of antibody±antigen complexes (Wilson and Stan®eld, 1993) only 3 of 13 complexes utilize all of the CDR loops for antigen recognition, although a minimum of four hypervariable loops are used (Chitarra et al., 1993); VL CDR3 and VH CDR3 are implicated in all 13 of these complexes. However, it has recently become evident that even fewer CDR loops are required for molecular recognition of antigens. Camelids can produce antibodies having no light chains (Hamers-Casterman et al., 1993) and can bind protein antigens with nanomolar af®nities using as few as two CDR loops (Decanniere et al., 1999). In a larger sampling of antibody±antigen complexes (Wilson and Stan®eld, 1994), not only did CDR loop usage vary between speci®c associations, but framework regions were commonly invoked in antigen recognition to varying degrees and could constitute up to 15% of the buried surface area of the complex. The VH CDRs, and VH CDR3 in particular, generally make more extensive contacts than VL CDRs, and the geometrical center of the interface tends to lie near VH CDR3. The second CDR of the light chain, VL CDR2, rarely contributes to antigen binding, while VL CDR3 and VH CDR3 tend to dominate the interactions, in terms of buried surface area, although VH CDR3 dominance is more transient due to its highly variable length distribution. There exists a strong correlation between residues that do not form contacts with antigen and those residues that are important in de®ning the canonical

MOLECULAR RECOGNITION IN ANTIBODY±ANTIGEN COMPLEXES

123

backbone structures of the CDR loops (Chothia et al., 1989). These residues tend to pack internally and are therefore less exposed on the antibody combining site surface. Antibody±antigen complexes exhibit a high degree of both shape and chemical complementarity at their interacting surfaces. Hydrophobic patches on the surface of the antigen pack against hydrophobic patches on the surface of the antibody combining site, atoms of polar character interact with atoms of opposite charge across the interface, and proton donors and acceptors form hydrogen bonds. Protruding side chains of one surface ®t into depressions of the other, and numerous van der Waals interactions are interspersed with 5±15 hydrogen bonds and an occasional salt bridge, as in other protein±protein interfaces (Lo Conte et al., 1999). The combined solvent-accessible surfaces buried in anti-protein Ê 2, antibody±antigen complexes range from approximately 1400 to 2300 A with roughly equal contributions from antigen and antibody, while smaller antigens, such as haptens and peptides, generally bury less overall surface area when bound to antibody. To quantitate shape complementarity in protein±protein interfaces, Lawrence and Colman (1993) have de®ned a shape correlation statistic (Sc ) that measures the degree of geometric match between two juxtaposed surfaces. Interfaces with perfect ®ts have an Sc value of 1, whereas interfaces that are topologically uncorrelated have Sc values near zero. The application of this algorithm to oligomeric proteins or protease± protease inhibitor complexes gives Sc values ranging from 0.70 to 0.76. For antigen±antibody interfaces, however, Sc values of 0.64 to 0.68 are obtained, indicating poorer shape correlation, albeit a slightly better topological correlation than for other classes of nonobligatory heterocomplexes ( Jones and Thornton, 1996). These differences in shape complementarity very likely re¯ect the particular biological context in which each type of interface is selected. Thus, protease±protease inhibitor and oligomeric interfaces have coevolved to optimize the ®t (and presumably the af®nity) between the interacting components, whereas antibodies must bind antigens not previously encountered during the evolutionary history of the immune system. Indeed, in a process termed ``af®nity maturation,'' somatic hypermutation of antibody genes is believed to increase af®nity by improving complementarity between antigen and antibody. The surface topography of the antigen-contacting surface, as well as other general structural features, of antibodies can vary signi®cantly according to antigen size (MacCallum et al., 1996). While the percentage of the antigen surface buried in the interface with antibody is always high and their surfaces are complementary, the antibody contact surface becomes more concave as the antigen becomes smaller. Thus, although the

124

ERIC J. SUNDBERG AND ROY A. MARIUZZA

combining sites of antibodies that recognize large protein antigens are generally planar, and are often more planar than a number of other types of protein±protein interfaces (Jones and Thornton, 1996), antibodies that recognize medium-sized antigens, such as peptides, DNA, and carbohydrates, often have a grooved antigen-contacting surface, while even smaller antigens (haptens) are recognized by antibodies with distinct cavities (Webster et al., 1994). A common feature of anti-peptide antibody±antigen interactions is a b-turn motif of the peptide buried deeply in the combining site. A number of examples exist for this general recognition scheme, including type I b-turns (Rini et al., 1992), type II b-turns (Stan®eld et al., 1990), and multiple tight turns (Garcia et al., 1992a). Over the entirety of the antibody combining site the distinctions in planarity become less clear. This is also the case for unbound antibodies, which alludes to the degree of conformational change that is induced increasingly in the antibody as antigen size is reduced (see below). The amount of surface area on the antibody molecule buried by antigen decreases with antigen size, as less of the antibody surface is utilized to envelop the smaller antigens. Large antigens often contact antibody residues at the edge of the combining site and interact with the more apical portions of the CDR loops, while the interactions of smaller antigens are more restricted to the central portion of the antibody combining site (MacCallum et al., 1996). Structures of antibody±antigen complexes illustrate the importance of bound water molecules in mediating antigen±antibody interactions. Indeed, with only one exception (Muller et al., 1998), water molecules have been localized in the interfaces of each of the antigen±antibody complexes whose crystal structures have been determined at suf®ciently Ê ), which allows the identi®cation of ordered high resolution (< 2:5 A waters with a reasonable degree of accuracy (Bhat et al., 1994; Fields et al., 1995; Mylavaganam et al., 1998; Kondo et al., 1999; Li et al., 2000). It therefore appears that water molecules are required to correct imperfections in antigen±antibody interfaces by improving the ®t between the proteins and by neutralizing unpaired hydrogen-bonding groups. Bound waters, acting as molecular adaptors, may compensate for the lack of evolutionary optimization of antigen±antibody interfaces, compared to other protein±protein interfaces in which the interacting surfaces may have coevolved to maximize complementarity (e.g., oligomeric proteins). In detailed structural analyses of the FvD1.3±HEL complex (Bhat et al., 1994; Braden et al., 1995, 1998), it has been shown that many water molecules from the free antibody and antigen structures are positionally conserved in the complex, that there is a recruitment of water molecules from the bulk solvent to the complex interface as demonstrated by a net gain of water molecules in the complex structures relative

MOLECULAR RECOGNITION IN ANTIBODY±ANTIGEN COMPLEXES

125

to the individual component structures, and that water molecules in the interface, regardless of their positional origin, increase the shape and chemical complementarity of the interacting surfaces. With such an apparently important role for water molecules in the molecular interface, it comes as no surprise that the D1.3±HEL association is an enthalpically driven process with some compensating negative entropy component as determined by isothermal titration calorimetry (Bhat et al., 1994). In fact, there is a strong correlation between decreases in water activity and association constants as determined through binding analyses performed in the presence of cosolutes with polarities lower than that of water (Goldbaum et al., 1996). Although other antibody±protein antigen (Kelley et al., 1992) and antibody±carbohydrate antigen (Sigurskjold et al., 1991) interactions also appear enthalpically driven, this may not be the general rule for antibody±antigen associations in particular or for protein±protein interactions in general (see below). In cases in which the structures of the unbound antibody and/or antigen are known, a comparison of the free and complexed structures has revealed small changes in conformation that probably serve to optimize the ®t between the proteins. Structural rearrangements in the Ê ) concerted antibody on antigen binding are restricted to (1) small (< 3 A movements in the CDR loops; (2) adjustments in side-chain positions; and (3) slight shifts in the relative orientation of the VL and VH domains, equivalent to rotations of < 38. Conformational changes of similar magnitude occur in the antigen. Somewhat larger conformational changes, Ê , have been described for including CDR loop displacements up to 7 A certain complexes involving DNA or peptide antigens (Wilson and Stan®eld, 1993; Tormo et al., 1994). Structural ¯exibility also appears to play an important role in the af®nity maturation process whereby relatively low-af®nity and -speci®city germline antibodies become mature antibodies with high antigen af®nity and speci®city (see below).

III. ANTIBODY CROSS-REACTIVITY AND MOLECULAR MIMICRY Although antibodies commonly have high af®nity and speci®city for a single antigen, it is not at all uncommon for antibodies to cross-react with many, structurally similar, yet distinct, antigenic molecules. In some cases, antibodies can bind better to antigens not used in challenging the immune system than to the original immunogen, a phenomenon known as heteroclitic binding. The monoclonal antibody (mAb) D11.15, which was raised against HEL, interacts with higher af®nity with several other avian lysozymes. In order to understand the structural basis for cross-reactivity in the D11.15±lysozyme system, Chitarra et al. (1993)

126

ERIC J. SUNDBERG AND ROY A. MARIUZZA

A

Ser100

Asp101

PHL JEL

Ser100

Asp103

Asp101 Val102 His103

VH

VL

B PHL JEL

Lys113 Asn113 Tyr57

VH

C

VL

Trp92 Ser93

Ser93

91O

HEL Gln121

Tyr32

Tyr32 92O

93N 121Oε1

Trp92

121Nε2

ε2

121N

TEL His121

FIG. 1. Cross-reactivity of antibodies. (A) Interaction of FvD11.15 (VH domain in blue, VL domain in green) with PHL (yellow) and JEL (red). The left panel is a closeup view of the encircled region in the right panel, highlighting the relative displacement of the 100±104 loop region between PHL and JEL resulting in a steric clash between JEL residues Val-102 and His-103 with the FvD11.15 VH domain. (B) Interaction between FvD11.15 [same color scheme as in (A) ] and PHL (yellow) and HEL (red). The left panel is a close-up view of the encircled region in the right panel and highlights the productive interactions that are made between FvD11.15 VH Tyr-57 and PHL Lys-113 (four hydrogen bonds, indicated by dotted lines). Conversely, productive interactions between FvD11.15 VH Tyr-57 and HEL Asn-113 are largely absent (one hydrogen bond, not shown for clarity) and is likely the reason for the binding af®nity discrepancy between the two antigens. (C) Hydrogen bonding between FvD1.3

MOLECULAR RECOGNITION IN ANTIBODY±ANTIGEN COMPLEXES

127

tested FvD11.15 binding to eight different avian lysozymes, all of which had high af®nity for the antibody, and two of which, pheasant egg-white lysozyme (PHL) and guinea fowl egg-white lysozyme (GEL), bound FvD11.15 with higher af®nity than HEL. The authors also determined the crystal structures of PHL, GEL, Japanese quail egg-white lysozyme ( JEL), and the FvD11.15±PHL complex. The af®nity of JEL for FvD11.15 is slightly lower than that of HEL (1:5 109 M 1 versus 4:0 109 M 1 ), which is likely derived from two amino acid differences in the loop region from residues 100 to 104 that forms part of the epitope in the FvD11.15±PHL complex. Whereas residue 102 is a Gly in HEL, it is a Val in JEL, while residue 103 is an Asn in HEL and a His in JEL. The result of these changes is a displacement of the 100±104 Ê into a conformation that would likely clash sterically loop region by 7.5 A with the VH CDR3 loop of FvD11.15 (Fig. 1A). Conversely, two amino acid differences between PHL and HEL, at residues 113 (Asn in HEL, Lys in PHL) and 121 (Gln in HEL, Asn in PHL), confer higher af®nity to the FvD11.15±PHL complex relative to the complex with HEL by two orders of magnitude. The crystal structure of FvD11.15±PHL reveals that the major structural difference in these two complexes is that Lys-113 in PHL makes several nonpolar contacts with VH Tyr-57 (Fig. 1B). Another anti-HEL antibody, D1.3, binds only its immunogen and one other avian lysozyme, bobwhite quail egg-white lysozyme, with high af®nity. Much of the sequence variability between the eight lysozymes tested occurs at HEL residue Gln-121. For the highly cross-reactive D11.15, lysozyme residue 121 is located at the periphery of the antigenic epitope. Conversely, for the highly speci®c D1.3, this residue is located centrally to the binding interface and acts as a hot spot in binding for the D1.3±HEL complex (Dall'Acqua et al., 1996). One of the avian lysozymes that binds poorly to D1.3, turkey egg-white lysozyme (TEL), has been investigated structurally (Braden et al., 1996b). HEL and TEL differ only at residue 121, which is a His in TEL, with a concomitant decrease in af®nity by two orders of magnitude, due primarily to a reduction in the on-rate of the interaction. Whereas Gln-121 of HEL makes two hydrogen bonds to VL domain main-chain atoms,

residues VL Tyr-32, Phe-91, Trp-92, and Ser-93 with HEL Gln-121 (left) and TEL His121 (right), the only amino acid difference between these two antigens. HEL Gln-121 makes three hydrogen bonds (indicated by dotted lines) to the main-chain nitrogen atom of Ser-93, the main-chain oxygen atom of Phe-91, and the phenyl ring of Tyr-32. All three of these hydrogen bonds are lost in the FvD1.3±TEL complex; however, a peptide ¯ip between FvD1.3 residues Trp-92 and Ser-93 results in a new hydrogen bond between the TEL His-121 side chain and the main chain oxygen atom of Trp-92.

128

ERIC J. SUNDBERG AND ROY A. MARIUZZA

TEL His-121 makes only one hydrogen bond to the antibody light chain and induces a peptide ¯ip between residues VL W92 and VL S93, a conformational change that is likely responsible for the slower on-rate of the interaction (Fig. 1C). Anti-idiotopic antibodies (Poljak, 1994; Pan et al., 1995) recognize an antigenic determinant that is unique to an antibody or group of antibodies, or idiotope. An idiotope is de®ned functionally by the interaction of an anti-idiotopic antibody (Ab2) with an antibody (Ab1) bearing the idiotope. Conventional Ab2 antibodies recognize idiotopes outside of the antibody combining site paratope, while internal image Ab2 antibodies are able to mimic the molecular surface encountered by Ab1, thereby mimicking stereochemically the antigen speci®c for Ab1. Numerous efforts have been made to use these molecular mimics as therapeutics, similar to vaccines. Several structural studies have been performed detailing a diverse range of idiotope±anti-idiotope interactions. The ®rst crystallographic analysis was of a complex between FabD1.3 and its idiotypic antibody FabE225 (Bentley et al., 1990). These Fab fragments interact primarily through their CDR loops (14 CDR residues from FabE225 and 10 CDR residues from FabD1.3 make intermolecular contacts), with a minor contribution through framwork residues (1 residue from FabE225, 3 residues from FabD1.3). Although there is a partial overlap of the FabD1.3 idiotope for FabE225 and paratope for HEL, with 7 of 13 residues in common between the two binding sites, there is no molecular mimicry at the atomic level, as the common contacts are stereochemically distinct in the two complexes. Another anti-idiotypic antibody system exploited for structural studies is that which centers around angiotensin II (AII), an octapeptide hormone. A mAb (Ab1) raised against AII was used to obtain anti-Ab1 polyclonal antibodies (Ab2s), which were, in turn, used to obtain an anti-anti-Ab1 mAb (Ab3) (Budisavljevic et al., 1988). In this system, Ab1 binds speci®cally to AII, Ab2s to Ab1, Ab3 to Ab2s, and, presumably due to molecular mimicry, Ab3 to the original antigen, AII. The crystal structure of the Ab3±AII complex (Garcia et al., 1992a) and sequence comparison of the binding determinants of Ab1 and Ab3 (Garcia et al., 1992b) have shown that the original antibody and its anti-anti-idiotypic antibody are superimposable and contact residues of Ab3 are highly conserved in the Ab1 sequence, even though the two antibodies are derived from distinct germline genes. Thus, although the structures of Ab2s could not be determined due to their molecular diversity, these polyclonal antibodies would appear to satisfy the stereochemical requirements for molecular mimicry at the atomic level, thereby producing an internal image of the antigen.

129

MOLECULAR RECOGNITION IN ANTIBODY±ANTIGEN COMPLEXES

We have studied the binding of the D1.3 antibody to two structurally distinct ligands: its cognate antigen, HEL, and the anti-idiotypic antibody E5.2. The crystal structure of the complex formed by FvD1.3 with Ê (Bhat et HEL has been determined to a nominal resolution of 1.8 A al., 1994) (Fig. 2A). In addition, the structure of the complex between the Fv fragments of D1.3 and the anti-D1.3 antibody E5.2 is known to Ê resolution (Fields et al., 1995; Braden et al., 1996a) (Fig. 2B). 1.9-A Surprisingly, it was found that D1.3 contacts HEL and E5.2 through essentially the same set of combining site residues (and most of the same atoms) (Table I). Thus, of the 18 D1.3 residues that contact E5.2 and the 17 that contact HEL, 14 are in contact with both E5.2 and HEL. These 14 D1.3 residues make up 75% of the total contact area with E5.2 and 87% of that with HEL. Furthermore, the positions of

A

B HEL

FvE5.2 VH

VL L1*

L3* H2*

H1*

L2*

H3*

H2

L2 L1

L3 H3

L1 L3

VH

VL

FvD1.3

H2

L2

H1

H1 H3

VL

VH

FvD1.3

FIG. 2. Structure of antigen±antibody complexes. (A) Ribbon diagram of the FvD1.3±HEL complex. Colors are as follows: HEL (yellow), D1.3 VL domain (green), and D1.3 VH domain (blue). Residues of HEL and D1.3 involved in interactions in the antigen±antibody interface are cyan and red, respectively. Heavy (H) and light (L) chain CDRs 1±3 are numbered. (B) Diagram of the FvD1.3±FvE5.2 complex. D1.3 VL domain (green), D1.3 VH domain (blue), E5.2 VL domain (yellow), and E5.2 VH domain (gray). Residues of D1.3 and E5.2 in contact in the structure are red and cyan, respectively. D1.3 and E5.2 heavy (H) and light (L) chain CDRs are labeled 1±3, with an asterisk denoting CDRs from E5.2.

TABLE I Intermolecular Contacts in the FvD1.3±FvE5.2 and FvD1.3±HEL Complexesa D1.3

E5.2 e1

L1 30 C

L3 93 C

L1 32 Cz

H3 100b Cd

L1 Tyr 32b L1 32 OH z

HEL

g2

D1.3 H1 30 O

121 Ne2 H3 100b Cd

E5.2

HEL

H3 98 N

D1.3 H2 52 C

E5.2 z3

118 C

H3 100 Ca

118 N

H3 97 Cg2

H3 99 O

119 Ca

H3 97 C

D1.3 H3 100 C

H3 97 Cb

E5.2 g

HEL e1

H1 33 C

H1 33 Ne2

a

24 N 27 Nd2

d1

e2

L2 49 C

H2 54 N

H2 54 Nd2

22 Ca

H1 31 C

H3 98 Cb

117 Ca

L2 50 Ce2

H2 58 Ne2

18 Cg

H1 31 O

H3 98 Cb

116 Cg

L2 50 Cz

H2 58 Ne2

18 Cg

H1 32 N

H3 98 Cb

119 Cg

18 Od1

H1 32 Ca

H3 98 Cd1

119 Od1

27 Nd2 24 Cb

130 H2 58 Ne2

18 Od2

H1 32 C

18 Od1 L2 50 Cd2

H1 32 Cz

119 Cb

H1 33 N

119 Cg

H2 52 Cb

119 Nd2

H2 52 Cg

L2 53 Og1

119 Nd2

L3 91 O

121 Ne2 H3 100b Cz

121 Cd

H3 100b NZ2

121 Ne2

H3 98 O

117 O

H3 98 Cb

H2 53 Ca

H3 98 Cb

117 O

H3 98 O

117 O

H2 52 Cd2 H2 52 Ce2

24 N 22 O 23 C 23 Ca

H3 100 Od2

H1 33 Ce1

H3 101 N

H3 98 OH

H3 101 Cg

H3 98 OH

24 Og

Wat

H2 53 C

H3 97 Cg2

H3 101 Cd2

H3 98 OH

121 Cb

H2 54 Cb

H3 99 Cd2

H3 101 Ce2

H3 98 OH

121 Cb

L2 49 OH

H3 101 Ce1

H2 54 Cg

H3 100 Ca

120 Cg2 119 Ca

H3 100 N

H3 100 Cb

121 Cd L3 92 Ce3

H2 53 N

116 Ce

H2 52 Od1

119 Cb

H3 98 Cb

H3 98 Cd1

H1 33 N

119 Ca

H2 52 CZ2

H3 98 Cd1

18 Cg

H3 100 O

119 Cg

117 Ca

18 Od2 L2 50 OH

119 C

b

L2 49 OH

19 Cb

H2 52 C

z2

24 Cb 24 Og

119 Cb

H3 97 C

d2

H3 98 C

HEL e1

H2 54 Od2

H3 100 N

Wat

H3 101 OH

H3 100a O

119 Od1

H3 100 Ca

L2 49 OH

Wat

120 N

H3 100 Cb

L2 49 Cz

118 Cb

H3 100b Ca

119 C

L2 49 Ce1

118 Cg2

H3 100b Cb

121 N

H3 100 Ca H3 100 Cg

119 Cb H2 56 Cb

H3 100 Cd

119 C

(continued)

TABLE IÐContinued D1.3 L3 92 Cz3

L3 92 CZ2

E5.2

HEL

D1.3

H3 100b Cz

121 Cd

H2 52 Ce3

HEL

D1.3

E5.2

H3 98 Cd1

118 C

H3 100 Oe1

L3 92 O

H3 98 Ce1

119 Ca

H3 100 Ne3

H3 100bNZ2

H3 99 C

119 N

L3 93 Cg2 L3 92 O

L3 92 Cz2

131

L3 92 Oc L3 93 N L3 93 Ca L3 93 Cb

E5.2

125 Ca 125 Cg

H2 56 Cg

H3 100 Cd

H3 99 O H3 100 Ca

H2 56 Nd2

H3 100 Oe1 H3 100 Oe1

H2 58 Cg

H3 100 Ne2

125 Cg

H2 52 Cd1

H3 100 Cg

125 Cd

H2 52 Ne1

H3 100 Cg

H3 100bNZ2 121 Oe1 121 Oe1 125 Cd 125 Ne

H2 58 Od1

H3 100 Ne2

H3 98 Cg H3 98 Cd

H3 98 OH H3 98 Ce1 H3 98 OH H3 98 OH H3 98 Cz H3 98 Ce1 H1 30 O H2 53 Cb

H3 98 Oe1

H3 99 NZ1

HEL

D1.3

E5.2

HEL 121 Cb 119 Ca

H3 101 Cz

121 Cb

H3 102 NZ2

119 Od1 22 O

Wat

102 O 102 O

Ê ) less than or equal to C±C, 4.1; C±N, 3.8; C±O, 3.7; N±N, 3.4; N±O, 3.4; and Intermolecular contacts de®ned by atom pair distances (in A O±O, 3.3. Atom pairs in the FvD1.3± FvE5.2 complex that mimic hydrogen bonds in the FvD1.3±HEL complex are in boldface type. bHydrogen bond interaction with the phenyl ring of Tyr-32. c Mimicry of FvE5.2 residue H3 100b for HEL 121 is maintained through a peptide ¯ip at FvD1.3 VL Trp-92. a

132

ERIC J. SUNDBERG AND ROY A. MARIUZZA

the atoms of E5.2 that contact D1.3 are close to those of HEL that contact D1.3, and 6 of the 12 hydrogen bonds in the D1.3±E5.2 interface are structurally equivalent to hydrogen bonds in the D1.3±HEL interface. Thus, E5.2 mimics HEL in its binding interactions with D1.3.

IV. THERMODYNAMIC MAPPING OF ANTIGEN±ANTIBODY INTERFACES In contrast to the wealth of structural information on antigen±antibody and other protein±protein interfaces, the available data on the thermodynamics of the association reactions are far more limited. Indeed, our current view of the energetics of protein±protein association is based largely on detailed mutagenesis and binding studies of only a few complexes (Bogan and Thorn, 1998). In this review, we will focus on the D1.3±HEL and D1.3±E5.2 complexes, since these represent the most extensively studied models for antigen±antibody recognition at the present time. To evaluate the relative contribution of individual residues to stabilization in the D1.3±HEL and D1.3±E5.2 complexes, we have performed alanine-scanning mutagenesis in the D1.3 combining site. In total, 16 single alanine substitutions were introduced and their effects on af®nity for HEL and for E5.2 were measured using surface plasmon resonance (SPR) detection, ¯uorescence quench titration, or sedimentation equilibrium (Dall'Acqua et al., 1996). Mutagenesis of D1.3 residues in contact with HEL in the crystal structure of the FvD1.3±HEL complex (Dall'Acqua et al., 1996) revealed that residues in VL CDR1 and VH CDR3 contribute more to binding than residues in VL CDR2, VL CDR3, VH CDR1, and VH CDR2. By far the greatest reductions in af®nity (DGmutant DGwild-type > 2:5 kcal/ mol) occurred on substituting three residues: VL W92, VH D100, and VH Y101. Signi®cant effects (1.0 to 2.0 kcal/mol) were also seen for substitutions at positions VL Y32 and VH E98, even though the latter is not involved in direct contacts with HEL. Mutations at 9 other contact positions (VL H30, VL Y49, VL Y50, VL S93, VH Y32, VH W52, VH D54, VH D58, andVH R99) had little or no effect (< 1:0 kcal/mol). Therefore, the binding of HEL by D1.3 is largely mediated by only 5 of the 14 residues tested. For the interaction of D1.3 with E5.2, af®nity measurements showed that VH CDR2, VH CDR3, and VL CDR1 of D1.3 are more important for binding E5.2 than VH CDR1, VL CDR2, and VL CDR3. Overall, D1.3 VH residues appear to contribute more to the free energy of binding than VL residues, since the most destabilizing alanine substitutions (> 2:5 kcal/ mol) are located in VH CDR2 (W52A and D54A) and VH CDR3 (E98A, D100A, and Y101A). Signi®cant effects (1.0 to 2.0 kcal/mol) were also

MOLECULAR RECOGNITION IN ANTIBODY±ANTIGEN COMPLEXES

133

observed for the following contact residues: H30 and Y32 in VL CDR1, Y49 in VL CDR2, Y32 in VH CDR1, N56 and D58 in VH CDR2, and R99 in VH CDR3. Mutations at positions VL Y50, VL W92, and VH T30 had little or no effect (< 1:0 kcal/mol). Thus, of the 15 contact residues tested, 12 make signi®cant contributions to binding E5.2. On the basis of extensive mutational analysis of the complex between human growth hormone and its receptor, Wells and colleagues proposed that the formation of speci®c protein±protein complexes is mediated by only a few productive interactions which dominate the energetics of association (see Clackson and Wells, 1995; Wells and de Vos, 1996; Clackson et al., 1998). In agreement with this idea, our analysis of the FvD1.3±HEL interaction revealed that only a small subset of the total combining site residues of D1.3 appears to account for a large proportion of the binding energy; most residues (9 of 14) make little or no apparent net contribution (< 1:0 kcal/mol). This contrasts with the interaction of D1.3 with E5.2 in which nearly all the contacting residues play a demonstrable role in binding ligand (> 1:0 kcal/mol), even though a number of hot spots (DDG > 2:5 kcal/mol) are clearly present. Therefore, stabilization of the D1.3±E5.2 complex is achieved by the accumulation of many productive interactions of varying strengths over the entire interface between the two proteins. The functional surfaces of D1.3 involved in binding HEL and E5.2 mapped onto its three-dimensional structure are shown in Fig. 3A and 3B, respectively. With the exception of VL W92, which lies at the periphery, the residues of D1.3 most important for binding HEL (VH Y101, VH D100, VL Y32, and VH E98) are located in a contiguous patch at the center of the combining site. Residues at the periphery make only minor contributions to the binding energy. A similar pattern is observed for the FvD1.3±FvE5.2 complex, with the most important residues (VL Y32, VH W52, VH D54, VH E98, VH D100, and VH Y101) forming a central band of key contacts. For the most part, however, the hot spots for the two interactions do not overlap. For instance, alanine substitution at position VL W92 of D1.3 produces a 100-fold decrease in af®nity for HEL but does not appreciably affect binding to E5.2. Conversely, the VH W52A substitution decreases af®nity for E5.2 1000-fold but has virtually no effect on binding to HEL. Only substitutions VH D100A and VH Y101A greatly affect the binding to both HEL and E5.2. We therefore conclude that a single set of contact residues on D1.3 binds HEL and FvE5.2 in energetically different ways. Thus, although D1.3 recognizes these two proteins in ways that are structurally very similar, this similarity extends only partially to the functional epitopes. To probe the relative contribution to binding of HEL residues in contact with D1.3 in the crystal structure of the FvD1.3±HEL complex,

134

ERIC J. SUNDBERG AND ROY A. MARIUZZA

FIG. 3. Energetic maps of antigen±antibody interfaces. (A) Space-®lling model of the surface of D1.3 (left) in contact with HEL and of the surface of HEL (right) in contact with D1.3. The two proteins are oriented such that they may be docked by folding the page along a vertical axis between the components. Residues are colorcoded according to the loss of binding free energy upon alanine substitution: red, >4 kcal/mol; yellow, 2±4 kcal/mol; green, 1±2 kcal/mol; blue, <1 kcal/mol. VL and VH residues are labeled in white and VL residues are indicated by an asterisk. (B) Model of the surface of D1.3 (left) in contact with E5.2 and of the surface of E5.2 (right) in contact with D1.3. Residues are colored and labeled as in (A).

12 nonglycine HEL residues were individually mutated to alanine and their af®nities for wild-type D1.3 measured (Dall'Acqua et al., 1998). Signi®cant decreases in binding (DDG > 1 kcal/mol) were observed for

MOLECULAR RECOGNITION IN ANTIBODY±ANTIGEN COMPLEXES

135

substitutions at only four contact positions: Gln-121, Ile-124, Arg-125, and Asp-119. The most destabilizing mutation was at position Gln-121 (DDG 2:9 kcal=mol). In the wild-type structure, Gln-121 penetrates a hydrophobic pocket where it is surrounded by the aromatic side chains of VL Tyr-32, VL Trp-92 and VH Tyr-101. It also makes two hydrogen bonds to the backbone of VL CDR3 (Gln-121HEL N2-O VL Phe-91 and Gln-121HEL Oe1-N VL Ser-93) and is nearly completely buried upon complex formation (Bhat et al., 1994). Mutations at the remaining eight contact positions (Asp-18, Asn-19, Tyr-23, Ser-24, Lys-116, Thr-118, Val120, and Leu-129) had little or no effect (DDG < 1 kcal/mol). Therefore, for both the D1.3 and HEL sides of this interface, only small subsets of the total contacting residues appear to account for a large portion of the binding energy. As shown in Fig. 3A, the residues of HEL most important for binding D1.3 (Asp-119, Gln-121, Ile-124, and Arg-125) form a contiguous patch located at the periphery of the surface contacted by the antibody (Dall'Acqua et al., 1998). Hot spot residues on the D1.3 side of the interface generally correspond to hot spot positions on the HEL side. For example, HEL hot spot residues Gln-121 (DDG 2:9 kcal/mol) and Arg-125 (1.8 kcal/mol) contact D1.3 hot spot residue VL Trp-92 (3.3 kcal/mol); in addition, Gln-121HEL contacts VL Tyr-32 (1.7 kcal/mol) and VH Tyr-101 (> 4:0 kcal/mol). Similarly, functionally less important D1.3 and HEL residues tend to be juxtaposed in the antigen±antibody interface: Asp-18HEL (DDG 0:3 kcal=mol) and Thr-118 (0.8 kcal/mol) interact with D1.3 VL Tyr-50 (0.5 kcal/mol) and VH Trp-52 (0.9 kcal/mol), respectively. To investigate the apparent contribution of E5.2 residues to stabilization of the D1.3±E5.2 complex, we introduced single alanine substitutions at 9 of 21 positions in the combining site of E5.2 involved in contacts with D1.3 and measured the af®nity of the mutants for wild-type D1.3 (Goldman et al., 1997). The most destabilizing substitutions are located at positions VH Y98 and VH R100b (DDG > 4:0 kcal/mol). Substitutions at the other 7 positions tested (VL Y49, VH K30, VH H33, VH D52, VH N54, VH I97, andVH Q100) also resulted in signi®cant effects on binding (1.2 to 2.8 kcal/mol). When the residues of D1.3 and E5.2 important in complex stabilization were mapped onto the three-dimensional structure of each Fv fragment, we observed that hot spot positions on the E5.2 side of the interface generally corresponded to hot spots on the D1.3 side (Fig. 3B), as in the FvD1.3±HEL interface (Fig. 3A). This complementarity of functional epitopes is in agreement with the observation of Clackson and Wells (1995), who found that energetically critical regions on human growth hormone match those on its corresponding receptor. In their case, however, the functional epitopes pack together to form a hydrophobic core surrounded by hydrophilic residues, with substantial reductions

136

ERIC J. SUNDBERG AND ROY A. MARIUZZA

in af®nity occurring only on substitution of the nonpolar residues. In contrast, our analysis of the D1.3±E5.2 and D1.3±HEL systems shows that both polar (e.g., D1.3 residues VH D54, VH E98 and VH D100) and nonpolar residues (e.g., D1.3 residues VL W92 and VH W52) play a prominent role in complex stabilization and that there is not a clear segregation of polar residues at the periphery of the interface and of nonpolar residues at the core (Fig. 3). On the basis of these studies, two broad categories of protein±protein interfaces may be de®ned: (1) those in which ligand binding is mediated by a small subset of contact residues and (2) those in which the free energy of binding arises from many productive interactions distributed over the entire protein±protein interface. In addition, each of these categories may be further subdivided into (1) those that resemble cross sections through folded proteins in which hydrophobic residues are in the interior and hydrophilic residues at the periphery and in which productive binding is mediated largely by the former and (2) those in which polar and nonpolar residues are evenly distributed throughout the interface and in which both residue types make comparable contributions to complex stabilization.

V. DISSECTION OF BINDING ENERGETICS IN ANTIGEN±ANTIBODY INTERFACES USING DOUBLE-MUTANT CYCLES A. The FvD1.3±FvE5.2 Complex Except in favorable cases, the strength of an interaction between two amino acid residues in a protein or protein±protein complex cannot be measured by simply mutating one of them because not only is the interaction of interest disrupted but other interactions, both direct and indirect, are disrupted as well (Ackers and Smith, 1985; Fersht, 1988). Thus, comparing the binding of a wild-type protein with that of a mutant in which a side chain has been truncated gives an apparent binding energy that is generally greater than the incremental binding energy attributable to that side chain. A more sophisticated approach to dissecting the energetics of pairwise interactions makes use of double-mutant cycles (Ackers and Smith, 1985; Serrano et al., 1990). We have used this method to analyze speci®c interactions responsible for stabilization of the FvD1.3± FvE5.2 and FvD1.3±HEL complexes (Goldman et al., 1997; Dall'Acqua et al., 1998). In the case of the FvD1.3±FvE5.2 complex, FvD1.3 residue A and FvE5.2 residue B were mutated (i.e., A ! A 0 , B ! B0 ) separately and together to construct the cycle:

MOLECULAR RECOGNITION IN ANTIBODY±ANTIGEN COMPLEXES

137

FvD1:3(A)FvE5:2(B0 ) ! FvD1:3(A 0 )FvE5:2(B0 ) " " FvD1:3(A)FvE5:2(B) ! FvD1:3(A 0 )FvE5:2(B): If the effects of the mutations are independent, the change in free energy for the double mutant is the sum of those for the two single mutations. But if there is an interaction between the mutated residues, the change in free energy for the double mutant is different from the sum of the two single mutants. Assuming that the mutations do not introduce major structural changes in the complex, the differences in free energy on association can provide a quantitative estimate of the energy of interaction of the two side chains (A and B) based on the premise that the interactions of A and B with the rest of the interface will cancel out. The coupling, or interaction, energy between A and B (DDGint ) is given by DDGint DDGAB!A0 B0

DDGAB!A0 B

DDGAB!AB0 ,

where each of the DDG terms is determined by measuring the difference in the free energy of binding of the Fv fragments in the initial and ®nal states of the corresponding transition. If DDGint equals zero, the effects of the two mutations are independent of each other and the two residues are not coupled. A DDGint other than zero indicates that residues A and B are coupled, either directly or indirectly. Thus, double-mutant cycles were constructed for 11 amino acid pairs in the interface between D1.3 and E5.2 (Goldman et al., 1997). Of these 11 pairs, 7 had interacting side chains as judged from the crystal structure (Fields et al., 1995; Braden et al., 1996a), whereas 4 were not expected to show coupling. As shown in Table IIA, the interaction between D1.3 VH E98 and E5.2 VH Y98 has the largest coupling energy, 4.3 kcal/mol. These two residues form a buried hydrogen bond in the crystal structure; in addition, ®ve van der Waals contacts are expected to be lost in the double mutant. Thus, the measured coupling energy re¯ects the combination of a number of different atomic interactions, including hydrophobic contacts. While our value is consistent with the energies found by Fersht et al. (1985) for breaking one hydrogen bond between an enzyme and a substrate (0.5 to 4.5 kcal/mol, with the highest energies when the bond is between a charged±neutral pair), it is important to note that only part of the 4.3 kcal/mol should be attributable to the hydrogen bond alone. Residues interacting through more solvent-exposed hydrogen bonds (D1:3VH D54 and E5:2VL Y49, D1:3VH D58 and E5:2VH Q100, and D1:3VL Y49 and E5:2VH N54) have signi®cantly lower coupling energies (1.7, 1.7, and 1.6 kcal/mol, respectively). This may re¯ect, in part, a higher dielectric constant in the more peripheral regions of this protein±protein interface. In addition to the solvated hydrogen bonds, residues D1:3VH D54 and E5:2VH Q100 make three van der Waals contacts, while the D1:3VL Y49 E5:2VH N54 and

TABLE II Coupling Energies between Amino Acid Pairs as Measured by Double-Mutant Cycles A. In the FvD1.3±FvE5.2 Complexa D1.3

E5.2

DDGint (kcal/mol)

Number and type of interactions lost in double mutants (as determined from crystal structures)

138

VH E98

VH Y98

4.3

VH D54

VL Y49

1.7

1 buried H bond and 5 van der Waals contacts 1 solvated H bond and 3 van der Waals contacts

VH D58

VH Q100

1.7

1 solvated H bond and 1 van der Waals contact

VL Y49

VH N54

1.6

1 solvated H bond and 1 van der Waals contact

VH W52

VH Q100

1.6

3 van der Waals contacts

VL Y32

VH R100b

1.5

2 van der Waals contacts

VH N56 VL Y49

VH Q100 VH H33

1.3 1.0

3 van der Waals contacts No direct contacts

VH D100

VH N54

0.5

No direct contacts

VH N56

VH N54

0.8

No direct contacts (far apart)

VH H33

0.6

No direct contacts (far apart)

B. In the FvD1.3±HEL Complexa D1.3 VL Y32

HEL

DDGint (kcal/mol)

Number and type of interactions lost in double mutants (as determined from crystal structures)

Q121

2.0

I124

0.0

Ê apart) No direct contacts (4.1 A

VL Y50

D18

0.4

D119

0.3

1 H bond (solvent-exposed) and 7 van der Waals contacts Ê apart) No direct contacts (14 A

VL W92

Q121 R125

2.7 1.7

I124

0.7

L129

0.2

1 H bond

3 van der Waals contacts 3 van der Waals contacts Ê apart) No direct contacts (4.1 A Ê apart) No direct contacts (6.5 A

139

VH Y32

K116

0.2

1 van der Waals contact and 1 long H bond*

VH W52

D119

0.3

3 van der Waals contacts

VH D54

T118

0.6

1 van der Waals contact

VH D100

S24

0.3

1 H bond (in water channel) and 1 van der Waals contact

VH Y101

D119 V120

0.1 0.0

1 H bond (partially buried) and 1 van der Waals contact 1 van der Waals contact

a

Coupling energies are de®ned as DDGint DDGAB!A 0 B DDGAB!AB0 DDGAB!A0 B0 , where DDGAB!A 0 B is the change in binding free energy (relative to wild type) on mutation of the D1.3 residue to alanine (A 0 ), DDGAB!AB0 is the change in binding free energy on mutation of E5.2 or HEL residue B to alanine (B0 ), and DDGAB!A 0 B0 is the change in binding energy on mutation of both residues A Ê ) less than or equal to the following: C±C, 4.1; and B to alanine. Intermolecular contacts were de®ned by atom pair distances in (A C±N, 3.8; C±O, 3.7; N±N, 3.4; N±O, 3.4; O±O, 3.3. The one exception is the hydrogen bond labeled with an asterisk between D1.3 Ê. VH Y32 and HEL K116 with a N±O distance of 3.5 A

140

ERIC J. SUNDBERG AND ROY A. MARIUZZA

D1:3VH D58 E5:2VL Y49 interactions include one additional van der Waals contact each. These results show the importance of local environment, in addition to bond geometry, in determining the relative strengths of different types of hydrogen bonds. Residues that interact only through van der Waals interactions tend to show slightly lower coupling energies (Table IIA). The three pairs tested (D1:3VH W52 E5:2VH Q100, D1:3VL Y32 E5:2VH R100b, and D1:3VH N56 E5:2VH Q100) make between two and six van der Waals contacts between their side chains that are expected to be lost on alanine substitution. The highest coupling energy (1.6 kcal/mol) is observed between D1.3 residue VH W52 and E5.2 residue VH Q100. Alanine substitution should result in the loss of three side chain±side chain contacts, while a number of main chain±main chain interactions should be preserved. A coupling energy of 1.3 kcal/mol was measured between D1:3VH N56 and E5:2VH Q100, in which three side chain±side chain van der Waals contacts should again be lost in the double mutant, and a coupling energy of 1.5 kcal/mol was measured between D1:3VL Y32 and E5:2VH R100b, in which only two such contacts should be lost. On this basis, we can calculate that each van der Waals contact contributes from 0.4 to 0.7 kcal/mol to complex stabilization. However, these values should be considered only upper limits since they almost certainly include signi®cant contributions from the hydrophobic effect (see below). This contribution may be particularly large for the D1:3VH W52 E5:2VH Q100 interaction as it involves a tryptophan residue. Double-mutant cycles were also constructed for four residue pairs not involved in direct interactions in the crystal structure: D1:3VL Y49± E5:2VH H33, D1:3VH D100 E5:2VH N54, D1:3VH N56 E5:2VH N54, and D1:3VH N56 E5:2VH H33 (Table IIA). The ®rst two pairs are in proximity in the interface, but do not form any direct contacts, while the second Ê ). The coupling energies for these two are physically far apart (20 to 30 A interactions vary from 0.5 and 1.0 kcal/mol. These values are signi®cantly different from zero based on the experimental error of our measurements (0:3 kcal=mol). Small-magnitude energetic coupling between amino acid residues separated by large distances has been observed in several systems (Green and Shortle, 1993; LiCata and Ackers, 1995), including barnase±barstar (Schreiber and Fersht, 1995). These couplings are generally < 1 kcal/mol between sites in proteins that cannot be in direct contact. There are several possible explanations for these results. The assumption that the mutations do not have signi®cant effects on protein conformation may not be valid. In addition, the mutations may introduce solvent rearrangements in the Fv±Fv interface, as described for complexes between mutants of FvD1.3 and HEL (Fields et al., 1996;

MOLECULAR RECOGNITION IN ANTIBODY±ANTIGEN COMPLEXES

141

Dall'Acqua et al., 1998; Sundberg et al., 2000). Such changes, though localized, may result in global perturbations in electrostatic ®elds or vibrational modes within the interface. With the exception of weak interactions between certain noncontacting residues, the relative strengths of coupling energies in the FvD1.3± FvE5.2 interface are broadly consistent with expectations based on the three-dimensional structure. Thus, the highest coupling energy (4.3 kcal/ mol) was measured for a charged±neutral pair forming a buried hydrogen bond. For residues interacting through solvent-exposed hydrogen bonds, coupling energies were approximately 1.7 kcal/mol, regardless of whether a neutral±neutral or charged±neutral pair was involved. Interactions formed by solvent-mediated hydrogen bonds were energetically neutral, while coupling energies of about 1.5 kcal/mol were measured for residues engaged only in van der Waals contacts. However, in contrast to the D1.3±E5.2 complex, double-mutant cycle analysis of the D1.3±HEL complex revealed that most residue pairs in direct contact in the crystal structure exhibited no signi®cant energetic coupling. B. The FvD1.3±HEL Complex As for the D1.3±E5.2 complex, double-mutant cycles were constructed for amino acid pairs in the D1.3±HEL interface in order to measure interaction energies (DDGint ) at the contact surfaces between these two proteins (Dall'Acqua et al., 1998) (Table IIB). Of the 14 pairs tested, 10 have interacting side chains according to the crystal structure, 2 do not Ê ), and 2 are far apart (6±15 form direct contacts but are in proximity (4 A Ê A). Surprisingly, only 3 of the 10 residue pairs in direct contact in the crystal structure show coupling energies of greater than 1.0 kcal/mol: VL Tyr-32 Gln-121HEL (2.0 kcal/mol), VL Trp-92 Gln-121HEL (2.7 kcal/ mol), and VL Trp-92 Arg-125HEL (1.7 kcal/mol). Indeed, with the exception of the residue pair VH Asp-54 Thr-118HEL (DDGint 0:6 kcal/ mol), none of the remaining 7 residue pairs have coupling energies exceeding the estimated experimental error of 0:3 kcal=mol: VL Tyr-50 Asp-18HEL ( 0:4 kcal=mol), VH Tyr-32 Lys-116HEL (0.2 kcal/ mol), VH Trp-52 Asp-119HEL ( 0:3 kcal=mol), VH Asp-100 Ser-24HEL (0.3 kcal/mol), VH Tyr-101 Asp-119HEL ( 0:1 kcal=mol), and VH Tyr-101 Val-120HEL (0.0 kcal/mol). In fact, in terms of their coupling energies, these residue pairs are indistinguishable from the 4 pairs that do not form direct contacts: VL Tyr-32 Ile-124HEL (DDGint 0:0 kcal=mol), VL Tyr-50 Asp-119HEL (0.3 kcal/mol), VL Trp-92 Ile-124HEL (0.7 kcal/mol), and VL Trp-92 Leu-129HEL (0.2 kcal/mol). Therefore, the simple fact that two residues make direct contacts in a protein±protein interface does not

142

ERIC J. SUNDBERG AND ROY A. MARIUZZA

necessarily indicate that a net productive interaction exists between them. Rather, the majority of such contacts may be energetically neutral, as in the present case. These ®ndings differ markedly from those for the D1.3±E5.2 (Goldman et al., 1997) and barnase±barstar (Schreiber and Fersht, 1995) comÊ of one another showed plexes in which nearly all residues within 4 A signi®cant coupling (> 1:0 kcal=mol). With one exception, none of the hydrogen bonds we examined in the D1.3±HEL interface make signi®cant net contributions to complex stabilization. These include hydrogen bonds which, on the basis of donor±acceptor distance and relative orientation of the interacting groups, would be predicted to be strong, as well as those expected to be weak. For example, VH Tyr-101 is situated near the center of the D1.3±HEL interface and makes a short Ê ) hydrogen bond with good geometry to HEL residue Asp-119 (2.7 A (VH Try-101 OZ-Ogl Asp-119HEL ); in addition, this hydrogen bond involves a charged±neutral pair and is buried in the interface. We therefore expected a DDGint of 3±4 kcal/mol, based on measurements in other systems (Serrano et al., 1990; Goldman et al., 1997). However, the observed DDGint for the VH Tyr-101 Asp-119HEL residue pair is 0:1 kcal=mol, which is effectively zero within experimental error (Table IIB). Similarly, VL Tyr-50 and VH Asp-100 make ``strong'' hydrogen bonds with HEL Asp-18 and Ser-24, respectively: VL Tyr-50 Ê ) and VH Asp-100 Og2-Og Ser-24HEL (2.8 A Ê ). OZ-Og2 Asp-18HEL (2.7 A Although these two bonds are solvated and might therefore be expected to be weaker than the buried VH Tyr-101 OZ-Og1 Asp-119HEL hydrogen bond, the DDGint values for residue pairs making similar bonds in the D1.3±E5.2 interface are about 1.5 kcal/mol. Nevertheless, there is no apparent interaction between the hydrogen-bonded VL Try-50 Asp18HEL and VH Asp-100 Ser-24HEL residue pairs in the D1.3±HEL complex (Table IIB). The only hydrogen-bonded residues in this protein± protein interface that show signi®cant coupling are Gln-121HEL and VL Tyr-32 (DDGint 2:0 kcal=mol). This interaction consists of a hydrogen bond between the Ne2 group of Gln-121 acting as a hydrogen bond donor and the center of the phenyl ring of VL Tyr-32 acting as a hydrogen bond acceptor. Hydrogen bonds with aromatic rings as proton acceptors are expected to contribute approximately 3 kcal/mol of stabilizing free energy to molecular associations (Levitt and Perutz, 1988), in good agreement with our measured value of 2.0 kcal/mol. The reasons for the observed differences between actual and expected coupling energies for most of the hydrogen-bonded residue pairs we examined can be understood in terms of an analysis of the theory of double-mutant cycles by Serrano et al. (1990). If A and B are two hydrogen-bonded residues and the mutations A ! A 0 and B ! B0 are

MOLECULAR RECOGNITION IN ANTIBODY±ANTIGEN COMPLEXES

143

nondisruptive (i.e., neither A nor any other protein residue moves on mutating B and vice versa), the coupling energy between A and B is given by DDGint GA=B

DGA=solv

DGB=solv ,

where DGA=solv is the solvation energy of A, DGB=solv is the solvation energy of B, and GA=B is energy of the hydrogen bond between A and B. If mutation of A allows water to bind to B, and vice versa, DGA=solv and DGB=solv are the energies of the hydrogen bonds with water. This equation then measures the free energy change for the exchange reaction A±H2 O H2 O B ! A B H2 O H2 O in which A and B exchange their interactions with water to interact with each other. If there is no access of water on mutation of A or B, DGA=solv DGB=solv is zero and DDGint GA=B . For the interaction of VL Tyr-50 with HEL Asp-18 (DDGint 0:4 kcal/mol), high-resolution X-ray crystallographic analysis of the single-mutant complexes has clearly shown that, in the mutants, the hydrogen-bonding potentials of the VL Tyr-50 and Asp-18HEL side chains are satis®ed by interface solvent molecules and that the mutations are nondisruptive (Fields et al., 1996; Dall'Acqua et al., 1998). It is therefore likely that, at least in this case, DDGint is a measure of the free energy change of the exchange reaction, rather than of the intrinsic energy of the VL Tyr-50 OZ-Od2 Asp-18HEL hydrogen bond. Since this DDGint is effectively zero, the strength of this particular protein±protein hydrogen bond is comparable to that of the water±protein hydrogen bonds it replaces and the bond makes no net contribution to complex stabilization. The observed lack of coupling between other hydrogen-bonded residue pairs in the D1.3±HEL interface may be similarly explained. This does not account, however, for the ®nding that hydrogen-bonded residue pairs in the D1.3±E5.2 interface showed coupling energies of 1±4 kcal/mol (Goldman et al., 1997). One possibility is that there is no access (or restricted access) of solvent on mutation of residues A or B in the D1.3±E5.2 complex such that DDGint more closely re¯ects GA=B . This is unlikely in the case of solvated hydrogen bonds, but probably applies to buried hydrogen bonds. Alternatively, hydrogen bond strengths may be highly dependent on local environment (for example, on the dielectric constant at the site of mutation), such that even bonds of similar lengths and geometry may possess very different energies. X-ray crystallographic studies of D1.3±E5.2 mutant complexes will be required to distinguish between these possibilities for individual hydrogen-bonded residue pairs. Regardless of the physical interpretation of our DDGint measurements in terms of the actual strength of individual bonds in the D1.3±HEL and D1.3±E5.2 interfaces, however, the fact remains that a DDGint equal to zero for any two target residues indicates that there is no net interaction between them.

144

ERIC J. SUNDBERG AND ROY A. MARIUZZA

The interaction of Gln-121HEL with VL Trp-92 has the largest coupling energy (DDGint 2:7 kcal/mol). Since this interaction involves only three van der Waals contacts, this value almost certainly includes a large contribution from the hydrophobic effect arising from burial of the tryptophan side chain (see below). Similar considerations probably apply to the interaction of VL Trp-92 with Arg-125HEL (DDGint 1:7 kcal/mol). Of the four residue pairs not involved in direct contacts in the crystal Ê ; VL Tyr-50 Asp-119HEL , 14 A Ê; structure (VL Tyr-32 Ile-124HEL , 4.1 A Ê Ê VL Trp-92 Ile-124HEL , 4.1 A; and VL Trp-92 Leu-129HEL , 6.5 A), only VL Trp-92 and Ile-124HEL show signi®cant coupling (0.7 kcal/mol). We do not ®nd examples of long-range interactions as in the D1.3±E5.2 (Goldman et al., 1997) and barnase±barstar interfaces, although a more extensive survey of noncontacting residues would likely reveal these. These results demonstrate that considerable caution should be exercised when attempting to estimate the strengths of speci®c interactions in an antigen±antibody (or other protein±protein) interface on the basis of three-dimensional structures alone. Although recent computational methods for predicting the strengths of these interactions appear promising (Shoichet and Kuntz, 1996; Wallqvist and Covell, 1996; Chong et al., 1999; Burnett et al., 2000), information on the relative contribution of individual residues to complex stabilization can be reliably obtained at the present time only through actual af®nity measurements of site-directed mutants of the interacting species. For example, a modeling analysis of the D1.3±HEL interface based on pairwise surface preferences predicted that the largest changes in binding free energy should arise from the mutation of HEL residues Asn-19 and Gln-121 (Covell and Wallqvist, 1997). While we did indeed ®nd Gln-121 to be functionally the most important residue among those tested (DDGint 2:9 kcal/mol), we also found Asn-19 to make essentially no net contribution to complex stabilization (0.3 kcal/mol) (Table IIA).

VI. ACCOMMODATION OF MUTATIONS IN ANTIGEN±ANTIBODY INTERFACES Alanine-scanning mutagenesis of the D1.3±HEL interface has demonstrated that it is remarkably tolerant to mutations that, on the basis of the three-dimensional structure of the wild-type complex, might be expected to have pronounced effects on af®nity (Dall'Acqua et al., 1998). For example, truncation of HEL residue Asp-18 to alanine should result in the loss of a direct hydrogen bond to the side chain of D1.3 VL Tyr-50 (Asp-18HEL Od2-OZ D1:3 VL Tyr-50), as well as the loss of seven van der Waals contacts to this residue. Nevertheless, the af®nity of HEL D18A for D1.3 (4:5 107 M 1 ) is nearly identical to that of the wild-type

MOLECULAR RECOGNITION IN ANTIBODY±ANTIGEN COMPLEXES

145

(8:0 107 M 1 ), corresponding to a DDGint of only 0.3 kcal/mol. To understand how this mutation is nearly completely accommodated in the D1.3±HEL interface, we determined the crystal structure of the Ê resolution (Dall'Acqua et al., FvD1.3±HEL D18A complex at 1.5-A 1998). The mutation does not affect the overall structure of the complex and conformational changes at the site of the mutation are also small. The major difference in the structure of the FvD1.3±HEL D18A complex is a rearrangement of solvent such that three additional water molecules are stably incorporated in the interface at the site of the mutation (Fig. 4A). These bound waters (designated WATa, WATb, and WATc) form several hydrogen bonds with protein atoms and with other water molecules, some of which mimic hydrogen bonds made by Asp-18HEL. In the wild-type structure (Fig. 4B), one of the carbonyl oxygens of Asp-18HEL forms two hydrogen bonds with WAT1 and WAT2, the latter of which is hydrogen-bonded to the hydroxy group of VL Tyr-32. In the mutant, WATa serves as a substitute for Asp-18HEL by partially occupying some of the volume taken up by the side chain of this residue and by making hydrogen bonds with WAT1 and WAT2, thereby preserving the interface water network. In addition, WATa forms a hydrogen bond with the main-chain nitrogen of Leu-25HEL , which further anchors it in the interface. The other carbonyl oxygen of Asp-18HEL forms a direct hydrogen bond with the hydroxy group of VL Tyr-50 in the wild-type structure. In the mutant, VL Tyr-50 makes hydrogen bonds with WATb and WATc, which, like WATa, are positioned to help ®ll the cavity created by the mutation and form part of the rearranged solvent network bridging antigen and antibody. Thus, the loss of complementarity in the D1.3± HEL interface resulting from replacement of Asp-18HEL by alanine is compensated for by the stable inclusion of additional water molecules and by local rearrangements in solvent structure, rather than by adjustments in the conformation of the protein. Similar mechanisms may explain the tolerance of the D1.3±HEL interface to mutations at other solvent-accessible sites on the antigen, such as Ser-24HEL and Lys-116HEL , which also make multiple contacts with D1.3. We also observed solvent rearrangements, including the incorporation of additional interface waters, in X-ray crystallographic studies of other site-directed mutants of FvD1.3 complexed with wild-type HEL, including VL Y50S, VH Y32A, and VL W92D (Ysern et al., 1994; Fields et al., 1996). In the FvD1.3 VL Y50S HEL complex, for example, two additional waters occupy some of the volume taken up by the VL Tyr-50 side chain in the wild-type structure. In these cases, however, the mutations are only partially compensated for by the solvent rearrangements, since the VL Y50S, VH Y32A, and VL W92D mutant fragments bind HEL with 10 , 4 , and 1000-fold lower af®nities, respectively, than the

146

ERIC J. SUNDBERG AND ROY A. MARIUZZA

FIG. 4. Solvent rearrangement in an antigen±antibody interface induced by a mutation in the antigen. (A) Schematic representation of the FvD1.3±HEL Asp-18 ! Ala complex in the vicinity of the mutation. (B) Schematic showing the same region in the wild-type FvD1.3±HEL complex. Water molecules present in both structures are labeled WAT1±4. WATa, WATb, and WATc are additional waters in the FvD1.3±HEL Asp-18 ! Ala interface.

original antibody. Thus, there appears to be a wide range in the extent to which solvent rearrangements in a protein±protein interface can accommodate mutations, which probably depends on the nature of the local environment. In this respect, it is interesting to note that VL Tyr-50 and Asp-18HEL are juxtaposed in the complex structure (Fig. 4B). This suggests that these two residues de®ne a site in the interface that, perhaps because of its peripheral location, is particularly suitable for the stable

MOLECULAR RECOGNITION IN ANTIBODY±ANTIGEN COMPLEXES

147

incorporation of new waters to occupy cavities or channels created by side-chain truncations. In other cases, seemingly conservative mutations have been found to greatly affect antigen±antibody binding. For example, substitution of lysine for arginine at HEL position 68 in the HyHEL±5/HEL interface produced a 1000-fold decrease in af®nity (Chacko et al., 1995). Comparison of the crystal structures of the mutant and wild-type complexes revealed only small rearrangements in the vicinity of the mutation, with little change in buried surface area. A bound water molecule replaces the two nitrogens of the guanidinium group of the arginine and partially compensates for the loss of two salt bridges between HEL Arg-68 and HyHEL-5 VH Glu-50 (Chacko et al., 1995). The net consequence of the mutation is a loss of hydrogen bonding, since the side-chain amino group of the lysine cannot form the same bonding arrangement as the guanidinium group of the arginine. We have recently taken advantage of the relative accommodation of mutations of the hot spot VL Trp-92 residue in D1.3 to estimate the magnitude of the hydrophobic effect in an antigen±antibody interface (Sundberg et al., 2000). By replacing VL Trp-92 with residues bearing increasingly smaller side chains and determining the crystal structures and thermodynamic parameters of binding for each of the resulting mutant FvD1.3±HEL complexes, we have demonstrated a correlation between the binding free energy and the apolar surface area that corresÊ 2 . This value is in excellent agreement with ponds to 21 cal mol 1 A transfer free energy values for small hydrophobic solutes (Chothia, 1976; Reynolds et al., 1977; Hermann, 1977; Eisenberg and McLachlan, 1986; Ooi et al., 1987) and is lower than the hydrophobic stabilization energy for folding (Yutani et al., 1987; Matsumura et al., 1988; Kellis et al., 1989; Shortle et al., 1990; Eriksson et al., 1992; Takano et al., 1997; Xu et al., 1998) and theoretical estimates of the interfacial free energy of protein±protein interactions (Sharp et al., 1991; Nicholls et al., 1991). Furthermore, changes in binding free energy are derived almost entirely from changes in the solvent entropy, demonstrating that the exclusion of solvent from the molecular interface at position VL 92 is the predominant energetic factor in the formation of this protein complex. Notably, residues at position VL 92 are partially solvent exposed at the periphery of the interface. This may contribute to the agreement between the magnitude of the effective hydrophobicity measured for this protein±protein interaction and transfer free energy values of hydrophobic solutes. A similar mutational analysis of a hydrophobic hot spot residue in the center of a protein±protein binding interface may yield hydrophobicity values closer to those for protein folding stabilization due to cavity formation at the interface or values closer to those measured for the partially

148

ERIC J. SUNDBERG AND ROY A. MARIUZZA

solvent-exposed residue VL 92 due to accommodation by rearrangement and addition of interfacial water molecules.

VII. FUNCTIONAL ROLES FOR PROTEIN PLASTICITY IN ANTIGEN RECOGNTION The function of the immune system is dependent on the recognition of essentially any antigenic material, yet the structural diversity of antigens greatly outweighs the genetic diversity encoded by immune system genes. Thus, molecular recognition of diverse antigens is accomplished by producing antibodies with high af®nity and speci®city for almost any antigen in two ways. First, recombination and imprecise joining of V, D, and J antibody gene segments result in germline antibodies of relatively low af®nity and speci®city (Tonegawa, 1983). Second, somatic hypermutation of antibody V regions and selective expansion of clones on the basis of antibody af®nity produce mature antibodies that are high in both af®nity and speci®city (Rajewsky, 1996). A wide range of conformational changes have been observed in antibodies related to their movement through this process of af®nity maturation as well as during interactions as mature antibodies with speci®c antigens. While there appears to be an af®nity ceiling for antibodies in vivo (Roost et al., 1995; Batista and Neuberger, 1998), production by in vitro evolution of Fv fragments with antigen af®nities well above this af®nity threshold (Boder et al., 2000) has implicated further conformational re®nement as the basis for extending improvements in antigen af®nity. A. Af®nity Maturation While V±D±J gene rearrangement focuses molecular diversity at the center of the antibody combining site, somatic hypermutation generally spreads this diversity to regions at the periphery of the binding site (Tomlinson et al., 1996). Somatic mutation is primarily a point mutation process in gene regions that are highly conserved in the primary repertoire, resulting in codon insertions or deletions (Tomlinson et al., 1996), as opposed to junctional diversity in the primary repertoire, which produces CDR loops of different lengths (Chothia et al., 1992; Tomlinson et al., 1995). A number of structural studies comparing germline and mature antibodies bound to the same antigen have helped to elucidate the functional role of mutations distal to the antigen-binding interface in antigen af®nity and speci®city. Both the mature Fab48G7 and its germline counterpart, Fab48G7g, bind a nitrophenyl phosphonate transition-state analog, but

MOLECULAR RECOGNITION IN ANTIBODY±ANTIGEN COMPLEXES

149

with a 30,000-fold difference in af®nity, primarily due to a decrease in the dissociation rate (Wedemayer et al., 1997a). The sequence differences between the Fabs are limited to nine somatic hypermutations, six in VH Ê from the bound hapten. Crystal structures of and three in VL , up to 15 A the unliganded germline Fab48G7g and its complex with hapten (Wedemayer et al., 1997a) reveal large conformational changes induced upon antigen binding. Conversely, crystal structures of the mature Fab48G7 (Patten et al., 1996; Wedemayer et al., 1997b) in its free and hapten-bound forms exhibit very few conformational changes upon complex formation. Relative rotations of the VL and VH domains of only 0.448 were found for the mature Fab structures, while the germline Fab structures had a relative VL =VH domain rotation of 4.68 between free and bound species. Importantly, the conformational changes induced upon binding antigen by Fab48G7g are later observed in the mature Fab structure even in the absence of antigen. It appears, at least in the case of the Fab48G7 system, that the af®nity maturation process is driven in large part by a mechanism of preorganizing the antibody combining site into a conformation that is favorable for binding its hapten antigen. By introducing the forward and back mutations in the germline and mature Fabs by site-directed mutagenesis and measuring binding af®nity for antigen by SPR techniques, Yang and Schultz (1999) have ef®ciently deconstructed the effects of the nine somatic hypermutations on the af®nity maturation pathway of Fab48G7. For each of the individual mutations, the effect on binding was either positive or neutral. In this system, the additive changes in af®nity of the individual somatic mutations were not equal to the overall change in af®nity between germline and mature Fabs. By making double mutations, the authors determined that a high degree of cooperativity existed between mutations, not only between individually neutral mutations but also between even the two most positive individual mutations. Cooperativity between somatic hypermutations, however, does not appear to be a required mechanism for af®nity maturation. For Fab39-A11, which catalyzes a Diels±Alder reaction, only two somatic mutations exist between the germline and mature counterparts, only one of which contributes the majority of binding af®nity to mature Fab (Romesberg et al., 1998). Another catalytic antibody, AZ-28, which catalyzes an oxy-Cope rearrangement, has six somatic mutations, ®ve of which contribute to differences in af®nity between germline and mature antibodies in a strictly additive way (Ulrich et al., 1997). The number and cooperativity of somatic hypermutations appear to be dependent on the af®nity differences between the germline and mature antibodies. While the af®nity discrepancy between Fab48G7 and Fab48G7g is 30,000-fold (Wedemayer et al., 1997a), FabAZ-28, with only ®ve signi®cant somatic mutations, has an antigen af®nity only 40-

150

ERIC J. SUNDBERG AND ROY A. MARIUZZA

fold greater than its germline counterpart (Ulrich et al., 1997). Furthermore, Fab39-A11 and Fab39-A11g, with only one signi®cant amino acid difference, both bind nine haptens, for most of which the difference in af®nity is within an order of magnitude (Romesberg et al., 1998). If one considers that mature antibodies must break a minimum af®nity threshold for antigen binding through a limited number of somatic mutations to be functional in vivo, then it follows that the number of somatic mutations will increase as the difference in af®nities between germline and mature antibodies becomes larger and cooperativity between the somatic mutations will be utilized in cases where the af®nity maturation process must overcome extreme germline±mature af®nity discrepancies. It is important to note that all of these experiments probing the effects of individual somatic mutations on af®nity maturation have been performed with hapten binding. It remains to be seen whether these mechanisms for af®nity maturation hold true for larger antigens, such as peptides, nucleic acids, carbohydrates, and proteins. Some of the energetic factors involved in the preorganization of mature antibodies through somatic hypermutation of germline antibodies have been elucidated recently using SPR techniques utilized effectively for investigating the association of T cell receptors and their speci®c peptide±major histocompatibility complex-binding partners (Willcox et al., 1999; Boniface et al., 1999) in which different binding characteristics at various temperatures of the same complex provide information relative to the enthalpic and entropic contributions to the interaction. Using a model synthetic 40-mer peptide, PS1CT1, Manivel et al. (2000) generated panels of early primary and secondary response mAbs and the af®nity of each mAb for PS1CT3 was determined at two temperatures. While the effects of temperature on the dissociation step of the interaction were similar for mAbs in both panels, opposite temperature effects on association were observed for each panel of mAbs. For primary mAbs, complex association was enthalpically highly favorable but entropically unfavorable, while dissociation was enthalpically unfavorable and entropically favorable. The equilibrium binding for primary mAbs was enthalpically driven with a large entropic cost of complex formation, resulting in relatively low af®nity. Conversely, in secondary mAbs, association was enthalpically unfavorable but the entropic costs had been reduced dramatically. Because the dissociation step of the reaction was similar to that for primary mAbs, equilibrium binding in the secondary mAbs was essentially independent of enthalpy effects, and, instead, was driven by entropic changes. Thus, the relative high af®nity of the secondary mAbs is derived exclusively from the nearly complete abolishment of any entropic costs of complex association in comparison to the primary mAbs. While these experiments seem to con®rm the idea of antibody af®nity

MOLECULAR RECOGNITION IN ANTIBODY±ANTIGEN COMPLEXES

151

maturation through paratope preorganization, it is intriguing to note that the increased af®nity in the Fab48G7 system derives nearly entirely from a decrease in the dissociation phase of the reaction (Wedemayer et al., 1997a). Although similar experiments examining enthalpy and entropy effects on antigen binding to germline and mature Fab48G7 have not been performed, it is likely that these types of experiments would reveal that the complex is stabilized due to a large entropic barrier to dissociation in the mature versus germline Fab. More experimental evidence is needed to conclude whether stabilization of mature antibodies relative to their germline counterparts either through reducing entropic barriers to association or by increasing entropic barriers to dissociation, or some combination thereof, will emerge as the appropriate model for af®nity maturation of antibodies. B. Induced Fit Although the af®nity maturation of antibodies seems to be controlled predominantly by structural preorganization of the antibody paratope for its speci®c antigen, a number of conformational changes have been observed in mature antibodies on binding antigen. An expanding database of unliganded and antigen-bound structures of the same antibody has allowed the identi®cation of some trends in the induced ®t mechanism utilized for antigen recognition. In general, it appears that for smaller antigens, notably peptides and DNA, antibody plasticity is more pronounced than for protein antigens, although associations involving the latter commonly involve a nominal degree of molecular ¯exibility and cannot necessarily be classi®ed as typical ``lock-and-key'' interactions. Atomic movements induced by antigen binding fall into a number of distinct categories, from gross domain movements to speci®c ¯uctuations in side-chain rotamer positions. Flexibility in the junction between the V and C superdomains, as de®ned by the Fab elbow angle, has been observed with essentially equal frequency in both liganded and unliganded antibody crystal structures. Furthermore, two otherwise identical Fabs residing in the same asymmetric unit of a crystal have exhibited different elbow angles, revealing that changes in elbow angles are as likely to be the result of crystallographic artifacts as they are of speci®c antigen binding (Rini et al., 1993; Tormo et al., 1994). Thus, while elbow angle changes induced on antigen binding were once argued to be a key component of B cell receptor signaling, the lack of dependence of elbow angle changes on antigen binding has essentially invalidated this theory. It is also likely that, as a general rule, elbow angle diversity is irrelevant to antigen recognition.

152

ERIC J. SUNDBERG AND ROY A. MARIUZZA

For most structural comparisons of antibodies in their antigen-bound and free forms, some degree of rearrangement of the two V domains from the heavy and light chains has been observed. One of the most striking examples of this phenomenon is for the anti-HIV-1 gp120 peptide Fab50.1 (Stan®eld et al., 1993) in which a relative reorientation of the VL and VH domains of 16.38 was observed on binding of the peptide antigen. Other anti-peptide antibodies have revealed more modest VL =VH domain rearrangements, including the antiviral capsid protein VP2 peptide Fab8F5, for which there was a 3.58 rotation (Tormo et al., 1994), and the anti-in¯uenza virus hemagglutinin peptide Fab17/9 in which VL =VH relative domain rotations on complexation were not signi®cantly larger than for the C domains (Rini et al., 1992). An antibody speci®c for a single-stranded DNA fragment, FabBV04-01, has displayed a relatively large V domain rearrangement of 7.58 on binding its antigen Ê , relative VL =VH domain rota(Herron et al., 1991). Another large, 8.5-A tion has been observed for Fab13B5 (Berthet-Colominas et al., 1999; Monaco-Malbet et al., 2000), which forms a complex with a helix±turn± helix motif in the HIV-1 capsid protein p24. Arguably, this epitope may be considered more as a continuous peptide than a protein epitope as the helix±turn±helix extends from the antigen surface and the interaction is characterized by a buried surface area similar to other peptide±antibody complexes. A smaller magnitude V domain rearrangement has been observed also for FvD1.3 on binding its protein antigen, HEL (Bhat et al., 1994). In another anti-HEL antibody, FabHyHEL-63, VL =VH domain relative rotations in the antigen-bound complex were not signi®cantly different than those between two different crystal forms of the unbound Fab (Li et al., 2000). The relative V domain rotation on antigen binding also seems to bear some correlation to the buried surface area between these domains (Stan®eld et al., 1993), a feature that is independent of antigen speci®city. Two types of backbone movements within the antibody combining site have commonly been observed upon antibody±antigen complex formationÐconcerted movements of multiple residue segments of CDR loops and more heterogeneous rearrangements of CDR residues. On binding antigen, heavy chain CDR loops in the anti-peptide Fab8F5 undergo essentially rigid-body movements in which the unliganded loop conformations are conserved, while changes in the main-chain conformation of the light chain are not signi®cant (Tormo et al., 1994) Ê , occurs (Fig. 5A). The largest backbone displacement, greater than 7 A for the VH CDR3 residue Tyr-102. The culmination of concerted heavy chain CDR movements toward the light chain reduces the volume of the antigen-binding site by some 3% relative to the unbound Fab8F5. Other examples of segments of CDR loops moving en masse

153

MOLECULAR RECOGNITION IN ANTIBODY±ANTIGEN COMPLEXES

VH Tyr102

A

B VH Tyr102

Lys157

VH Asn100a

Tyr105

Ala106

Ser101

VH Gly100b

Asp104

VH Arg95 VH Arg94

Fab8F5

C

Fab17/9

D FabHyHEL63 VH CDR3

FabHyHEL63 VH CDR2

FIG. 5. Conformational changes induced by antigen binding. (A) Concerted movement of the Fab8F5 CDR H3 loop induced on binding its peptide antigen. The unbound Fab structure is in green, the bound Fab structure is in blue, and the peptide antigen is in yellow. VH Ser-101 in the bound form makes two hydrogen bonds to LysÊ between the unbound 157 of the peptide and VH Tyr-102 is displaced by more than 7 A and bound Fab8F5 molecules. (B) Atomic rearrangement of the CDR H3 loop of the anti-peptide Fab17/9. The color scheme is the same as in (A). Side chains of residues in contact between the antibody and antigen in the bound complex are shown. Contrary to many anti-peptide antibodies, anti-protein antibodies generally exhibit relatively small conformational changes on binding antigen as shown in (C) and (D) for the antiHEL antibody FabHyHEL63. (C) Superposition of the CDR H3 loops of FabHyHEL63 bound to HEL (blue) and three different unbound forms: solved in the C2 space group (green); one molecule from the asymmetric unit of the free antibody solved in the P1 space group (red); and the second molecule of the asymmetric unit of the free antibody solved in the P1 space group (yellow). (D) Superposition of the CDR H2 loops of FabHyHEL63 bound to HEL (blue) and three different unbound forms. The color scheme is the same as in (C).

toward antigen have been observed (Stan®eld et al., 1990). In Fab17/9, a signi®cant rearrangement of the VH CDR3 loop is induced by binding of its peptide antigen (Fig. 5B), for which the largest backbone changes are Ê (Rini et al., 1992). Restructuring of CDR loop regions from both the 5A

154

ERIC J. SUNDBERG AND ROY A. MARIUZZA

heavy and light chains of the anti-DNA antibody FabBV04-01 have also been observed (Herron et al., 1991). Induced CDR loop movements upon antigen binding seem to be less extreme for anti-protein antibodies. Ê (Li et Generally, these are small, concerted displacements of less than 3 A al., 2000; Bhat et al., 1994; Braden et al., 1994; Mylvaganum et al., 1998; Prasad et al., 1993) (Fig. 5C). Amino acid side-chain rearrangements have been observed in all structures of antigen-bound antibodies relative to their unbound forms. Indeed, side-chain rearrangements are nearly ubiquitous in all types of protein±protein interactions and can range from speci®c rotamer choices to form productive interactions with ligand to relatively inconsequential movements associated predominantly with the avoidance of steric hindrance in the newly formed complex. It is dif®cult to predict the energetic value of these side-chain movements solely from structural data without mutagenesis and binding analyses. It is unlikely that side-chain movements in antibody±antigen interactions play a greater role in binding than in other macromolecular associations. Molecular ¯exibility is not limited to the antibody side of the reaction, as a number of structural studies have shown varying degrees of protein plasticity for antigens on binding. HEL can be crystallized in several space groups (Ramanadham et al., 1990; Harrata, 1994; Kurinov and Harrison, 1995). Comparison of the structures reveals signi®cant ¯exibility of several loops at the molecular surface, including a number of Ca atom Ê between HEL molecules from different displacements greater than 3 A space groups. Between crystal structures of HEL bound to different antibodies, some main-chain movements become more pronounced. Relative to the D1.3±HEL complex (Bhat et al., 1994), Gly-102 and AsnÊ in complexes with the anti-HEL 103 of HEL are displaced some 8 A antibodies HyHEL-10 (Padlan et al., 1989) and D11.15 (Chitarra et al., 1993). In the HyHEL-63±HEL complex (Li et al., 2000), residues 99 to Ê 104 of this same loop region of HEL have an rms deviation of 6.8 A relative to their positions in HEL complexed with D1.3 (Bhat et al., 1994). Molecular movement in this HyHEL-63 complex is highlighted by a peptide ¯ip at residue Asp-101, which allows the formation of ®ve hydrogen bonds between this residue and the antibody. A number of other, less extreme examples of antigen plasticity have been documented for HEL on antibody binding (Davies and Cohen, 1996), but the excessive ¯exibility of this particular loop region may contribute to the characteristics of this surface that make it an especially favorable epitope, as evidenced by the large number of anti-HEL antibodies that share this region as part of their recognition surfaces. Smaller conformational changes are seen in the HIV-1 capsid protein p24 on binding Fab13B5 (Berthet-Colominas et al., 1999; Monaco-Malbet et al., 2000). Localized to the turn portion of

MOLECULAR RECOGNITION IN ANTIBODY±ANTIGEN COMPLEXES

155

the helix±turn±helix motif, the ¯exibility of the antigen is highlighted by Ê displacement of the carbonyl oxygen of Pro-207, which points in a a 4-A direction opposite from its unbound form to adapt to the molecular environment of the antibody. Due to the high degree of ¯exibility of uncomplexed peptides (Dyson and Wright, 1995), the Fab13B5±p24 complex, with its continuous peptide-like epitope, may present the best current measure for the role of peptide antigen plasticity in antibody recognition. C. Beyond the Af®nity Ceiling Most mature antibodies have af®nities for their speci®c antigens in the range of 107 108 M 1 and it has been proposed (Foote and Eisen, 1995) that, due to diffusion rates and the residence time required for antibody internalization controlling on- and off-rates, an af®nity ceiling exists for antibody±antigen interactions of approximately 1010 M 1. Antibodies with antigen af®nities above this threshold, presumably, would not be further advantaged over their lower af®nity counterparts in the antibody selection process in vivo. Batista and Neuberger (1998), in a study involving mutated lysozyme antigens with varying af®nities for the lysozyme-speci®c B cell receptor, have effectively demonstrated that this af®nity ceiling does exist in vivo. These authors showed that a minimum af®nity of 106 M 1 and a half-life of 1 s were required for detectable B cell triggering, that the antigen concentration required for signaling and its af®nity were inversely correlated, and that this concentration/af®nity correlation reached a plateau for af®nities beyond 1010 M 1 . Nevertheless, a variant of Fab4-4-20 has been produced in vitro by yeast display that has femtomolar af®nity for the hapten ¯uorescein (Boder et al., 2000), well beyond the in vivo af®nity ceiling. This antibody contains 10 mutations, 9 of which are in the heavy V chain, 6 of which are in the CDR H3 loop, and only one of which is in contact with the antigen as modeled using the Fab4-4-20±¯uorescein structure (Herron et al., 1994). Phage display af®nity maturation of another anti-hapten antibody, Fab17E8 (Arkin and Wells, 1998), has also demonstrated the importance of mutations in residues beyond the zone of antigen-contacting residues in improving the af®nity of mature antibodies. It is likely that these af®nity-evolved antibodies that bind antigen above the in vivo af®nity ceiling accomplish this feat by a similar structural preorganization of the antibody combining site as by somatic hypermutation, although experiments showing such have yet to be performed. Thus, it seems that antibodies might indeed have the intrinsic ability in vivo to break the af®nity ceiling through further rounds of somatic hypermutation, but do not do so only because it would be super¯uous to the functioning of the immune system.

156

ERIC J. SUNDBERG AND ROY A. MARIUZZA

VIII. CONCLUSIONS Beyond de®ning their associations for purely immunological purposes, structural studies of antibody±antigen complexes have been used as models for protein±protein interactions. While some of the basic rules of protein±protein interactions seem to be reasonably established through studies of antibody±antigen complexes and other protein±protein associations, we remain far from understanding all of the factors that govern these reactions. With our relatively small database of information pertaining to antibody±antigen interactions it is easy to ®nd con¯icting examples for almost any of the factors that we believe may in¯uence these associations, while it is exceedingly dif®cult to detect trends, let alone rules, for molecular recognition between proteins. Thus, although much progress has been made in de®ning factors that are important in antibody±antigen interactions, we continue to ®nd ourselves in a situation in which extreme caution should be exercised in extending the speci®c to the general. We are beginning to see evidence for correlations between surface complementarity and af®nity, although this relationship is confounded by the role of water molecules in the interface, some of which are simply displaced on binding while others remain to increase the shape and chemical complementarity of the interface. These water molecules may also ®ll holes in the interface created by mutations to surface residues, complicating our understanding of how important hydrogen-bonding interactions may be in a particular complex. The use of double-mutant cycles has provided more accurate assessments of the energetic importance of some of these hydrogen bonds. Recent experimental results have only just allowed us to estimate the magnitude of the hydrophobic effect at the protein interface, but these measurements from mutagenesis studies of a residue located at a partially solvent-exposed position in the interface may or may not extend to the center of the protein±protein interface. While hot spots for binding are more commonly located in this region, some complexes have hot spots throughout the extent of the contact area. Structural plasticity appears to play myriad roles in antigen recognition by antibodies. The af®nity maturation process that produces mature antibodies with high antigen af®nity and speci®city seems to be governed, in large part, by entropic control of the association through somatic hypermutations distal to the antigen-contacting residues. Rigidi®cation of the combining site is by no means exhaustive though, as most mature antibodies exhibit conformational changes upon binding antigen. This conformational diversity continues to make antibody structure and antigen docking predictions dif®cult. Currently, even the effects of a single amino acid substitution on af®nity and speci®city of antibody±antigen complexes

MOLECULAR RECOGNITION IN ANTIBODY±ANTIGEN COMPLEXES

157

often cannot be predicted with con®dence. Somatic-like mutations can confer much higher af®nity to an antibody±antigen interaction, presumably by further restriction of the paratope to a more restricted array of conformations. The importance of residues outside of the binding interface markedly increases the complexity of protein±protein interactions. Only by utilizing protein engineering techniques combined with structural, thermodynamic, and kinetic analyses to probe distinct factors controlling complex formation in future studies will we be able to bring the picture of protein±protein interaction into sharper focus.

REFERENCES Ackers, G. K., and Smith, F. R. (1985). Annu. Rev. Biochem. 54, 597±629. Amit, A. G., Mariuzza, R. A., Phillips, S. E., and Poljak, R. J. (1986). Nature 233, 747±753. Amzel, L. M., and Poljak, R. J. (1979). Annu. Rev. Biochem. 48, 961±997. Arkin, M. R., and Wells, J. A. (1998). J. Mol. Biol. 284, 1083±1094. Ban, N., Escobar, C., Garcia, R., Hasel, K., Day, J., Greenwood, A., and McPherson, A. (1994). Proc. Natl. Acad. Sci. USA 91, 1604±1608. Batista, F. D., and Neuberger, M. S. (1998). Immunity 8, 751±759. Bentley, G. A., Boulot, G., Riottot, M. M., and Poljak, R. J. (1990). Nature 348, 254±257. Berthet-Colominas, C., Monaco, S., Novelli, A., Sibai, G., Mallet, F., and Cusack, S. (1999). EMBO J. 18, 1124±1136. Bhat, T. N., Bentley, G. A., Boulot, G., Greene, M. I., Tello, D., Dall'Acqua, W., Souchon, H., Schwarz, F. P., Mariuzza, R. A., and Poljak, R. J. (1994). Proc. Natl. Acad. Sci. USA 91, 1089±1093. Bizebard, T., Gigant, B., Rigolet, P., Rasmussen, B., Diat, O., Bosecke, P., Wharton, S. A., Skehel, J. J., and Knossow, M. (1995). Nature 376, 92±94. Boder, E. T., Midelfort, K. S., and Wittrup, K. D. (2000). Proc. Natl. Acad. Sci. USA 97, 10701±10705. Bogan, A. A., and Thorn, K. S. (1998). J. Mol. Biol. 280, 1±9. Boniface, J. J., Reich, Z., Lyons, D. S., and Davis, M. M. (1999). Proc. Natl. Acad. Sci. USA 96, 11446±11451. Bossart-Whitaker, P., Chang, C. Y., Novotny, J., Benjamin, D. C., and Sheriff, S. (1995). J. Mol. Biol. 253, 559±575. Braden, B. C., Goldman, E. R., Mariuzza, R. A., and Pojak, R. J. (1998). Immunol, Rev. 163, 45±57. Braden, B. C., Souchon, H., Eisele, J. L., Bentley, G. A., Bhat, T. N., Navaza, J., and Poljak, R. J. (1994). J. Mol. Biol. 243, 767±781. Braden, B. C., Fields, B. A., and Poljak, R. J. (1995). J. Mol. Recogn. 8, 317±325. Braden, B. C., Fields, B. A., Ysern, X., Dall'Acqua, W., Goldbaum, F. A., Poljak, R. J., and Mariuzza, R. A. (1996a). J. Mol. Biol. 264, 137±151. Braden, B. C., Fields, B. A., Ysern, X., Goldbaum, F. A., Dall'Acqua, W., Schwarz, F. P., Poljak, R. J., and Mariuzza, R. A. (1996b). J. Mol. Biol. 257, 889±894. Budisavljevic, M., Geniteau-Legendre, M., Baudouin, B., Pontillon, F., Verroust, P. J., and Ronco, P. M. (1988). J. Immunol. 140, 3059±3065. Burnett, J. C., Kellogg, G. E., and Abraham, D. J. (2000). Biochemistry 39, 1622±1633. Chacko, S., Silverton, E., Kam-Morgan, L., Smith-Gill, S., Cohen, G., and Davies, D. (1995). J. Mol. Biol. 245, 261±274.

158

ERIC J. SUNDBERG AND ROY A. MARIUZZA

Chitarra, V., Alzari, P. M., Bentley, G. A., Bhat, T. N., Eisele, J. L., Houdusse, A., Lescar, J., Souchon, H., and Poljak, R. J. (1993). Proc. Natl. Acad. Sci. USA 90, 7711±7715. Chong, L. T., Duan, Y., Wang, L., Massova, I., and Kollman, P. A. (1999). Proc. Natl. Acad. Sci. USA 96, 14330±14335. Chothia, C. (1976). J. Mol. Biol. 105, 1±14. Chothia, C., and Janin, J. (1975). Nature 256, 705±708. Chothia, C., Lesk, A. M., Tramontano, A., Levitt, M., Smith-Gill, S. J., Air, G., Sheriff, S., Padlan, E. A., Davies, D., Tulip, W. R., et al. (1989). Nature 342, 877±883. Chothia, C., Lesk, A. M., Gherardi, E., Tomlinson, I. M., Walter, G., Marks, J. D., Llewelyn, M. B., and Winter, G. (1992). J. Mol. Biol. 227, 799±817. Clackson, T., and Wells, J. A. (1995). Science 267, 383±386. Clackson, T., Ultsch, M. H., Wells, J. A., and de Vos, A. M. (1998). J. Mol. Biol. 277, 1111±1128. Colman, P. M., Laver, W. G., Varghese, J. N., Baker, A. T., Tulloch, P. A., Air, G. M., and Webster, R. G. (1987). Nature 326, 358±363. Covell, D. G., and Wallqvist, A. (1997). J. Mol. Biol. 269, 281±297. Dall'Acqua, W., Goldman, E. R., Eisenstein, E., and Mariuzza, R. A. (1996). Biochemistry 35, 9667±9676. Dall'Acqua, W., Goldman, E. R., Lin, W., Teng, C., Tsuchiya, D., Li, H., Ysern, X., Braden, B. C., Li, Y., Smith-Gill, S. J., and Mariuzza, R. A. (1998). Biochemistry 37, 7981±7991. Davies, D. R., and Cohen, G. H. (1996). Proc. Natl. Acad. Sci. USA 93, 7±12. Decanniere, K., Desmyter, A., Lauwereys, M., Ghahroudi, M. A., Muyldermans, S., and Wyns, L. (1999). Structure Fold. Des. 7, 361±370. Dyson, H. J., and Wright, P. E. (1995). FASEB J. 9, 37±42. Eisenberg, D., and McLachlan, A. D. (1986). Nature 319, 199±203. Eriksson, A. E., Baase, W. A., Zhang, X.-J., Heinz, D. W., Blaber, M., Baldwin, E. P., and Matthews, B. W. (1992). Science 255, 178±183. Evans, S. V., Rose, D. R., To, R., Young, N. M., and Bundle, D. R. (1994). J. Mol. Biol. 241, 691±705. Fersht, A. R. (1988). Biochemistry 27, 1577±1580. Fersht, A. R., Shi, J. P., Knill-Jones, J., Lowe, D. M., Wilkinson, A. J., Blow, D. M., Brick, P., Carter, P., Waye, M. M., and Winter, G. (1985). Nature 314, 235±238. Fields, B. A., Goldbaum, F. A., Ysern, X., Poljak, R. J., and Mariuzza, R. A. (1995). Nature 374, 739±742. Fields, B. A., Goldbaum, F. A., Dall'Acqua, W., Malchiodi, E. L., Cauerhff, A., Schwarz, F. P., Ysern, X., Poljak, R. J., and Mariuzza, R. A. (1996). Biochemistry 35, 15494±15503. Foote, J., and Eisen, H. N. (1995). Proc. Natl. Acad. Sci. USA 92, 1254±1256. Garcia, K. C., Ronco, P. M., Verroust, P. J., Brunger, A. T., and Amzel, L. M. (1992a). Science 257, 502±507. Garcia, K. C., Desiderio, S. V., Ronco, P. M., Verroust, P. J., and Amzel, L. M. (1992b). Science 257, 528±531. Goldbaum, F. A., Schwarz, F. P., Eisenstein, E., Cauerhoff, A., Mariuzza, R. A., and Poljak, R. J. (1996). J. Mol. Recogn. 9, 6±12. Goldman, E. R., Dall'Acqua, W., Braden, B. C., and Mariuzza, R. A. (1997). Biochemistry 36, 49±56. Green, S. M., and Shortle, D. (1993). Biochemistry 32, 10131±10139. Hamers-Casterman, C., Atarhouch, T., Muyldermans, S., Robinson, G., Hamers, C., Songa, E. B., Bendahman, N., and Hamers, R. (1993). Nature 363, 446±448. Harrata, K. (1994). Acta Crystallogr. Sect. D 50, 250±257. Hermann, R. B. (1977). Proc. Natl. Acad. Sci. USA 74, 4144±4148. Herron, J. N., He, X. M., Ballard, D. W., Blier, P. R., Pace, P. E., Bothwell, A. L., Voss, E. W., Jr., and Edmundson, A. B. (1991). Proteins 11, 159±175.

MOLECULAR RECOGNITION IN ANTIBODY±ANTIGEN COMPLEXES

159

Herron, J. N., Terry, A. H., Johnston, S., He, X. M., Guddat, L. W., Voss, E. W., Jr., and Edmundson, A. B. (1994). Biophys. J. 67, 2167±2183. Housset, D., Mazza, G., Gregoire, C., Piras, C., Malissen, B., and Fontecilla-Camps, J. C. (1997). EMBO J. 16, 4205±4216. Huang, M., Syed, R., Stura, E. A., Stone, M. J., Stefanko, R. S., Ruf, W., Edgington, T. S., and Wilson, I. A. (1998). J. Mol. Biol. 275, 873±894. Janin, J. (1999). Structure Fold. Des. 7, R277±279. Janin, J., and Chothia, C. (1990). J. Biol. Chem. 265, 16027±16030. Jones, S., and Thornton, J. M. (1996). Proc. Natl. Acad. Sci. USA 93, 13±20. Kabat, E. A., Wu, T. T., and Bilofsky, H. (1977). J. Biol. Chem. 252, 6609±6616. Kelley, R. F., O'Connell, M. P., Carter, P., Presta, L., Eigenbrot, C., Covarrubias, M., Snedecor, B., Bourell, J. H., and Vetterlein, D. (1992). Biochemistry 31, 5434±5441. Kellis, J. T., Jr., Nyberg, K., and Fersht, A. R. (1989). Biochemistry 28, 4914±4922. Kondo, H., Shiroishi, M., Matsushima, M., Tsumoto, K., and Kumagai, I. (1999). J. Biol. Chem. 274, 27623±27631. Kurinov, I. V., and Harrison, R. W. (1995). Acta Crystallogr. Sect. D 51, 98±109. Lawrence, M. C., and Colman, P. M. (1993). J. Mol. Biol. 234, 946±950. Lea, S., and Stuart, D. (1995). FASEB J. 9, 87±93. Levitt, M., and Perutz, M. F. (1988). J. Mol. Biol. 201, 751±754. Li, Y., Li, H., Smith-Gill, S. J., and Mariuzza, R. A. (2000). Biochemistry 39, 6296±6309. LiCata, V. J., and Ackers, G. K. (1995). Biochemistry 34, 3133±3139. Lo Conte, L., Chothia, C., and Janin, J. (1999). J. Mol. Biol. 285, 2177±2198. MacCallum, R. M., Martin, A. C., and Thornton, J. M. (1996). J. Mol. Biol. 262, 732±745. Malby, R. L., Tulip, W. R., Harley, V. R., McKimm-Breschkin, J. L., Laver, W. G., Webster, R. G., and Colman, P. M. (1994). Structure 2, 733±746. Manivel, V., Sahoo, N. C., Salunke, D. M., and Rao, K. V. (2000). Immunity 13, 611±620. Matsumura, M., Becktel, W. J., and Matthews, B. W. (1988). Nature 334, 406±410. Monaco-Malbet, S., Berthet-Colominas, C., Novelli, A., Battai, N., Piga, N., Cheynet, V., Mallet, F., and Cusack, S. (2000). Structure Fold. Des. 8, 1069±1077. Muller, Y. A., Chen, Y., Christinger, H. W., Li, B., Cunningham, B. C., Lowman, H. B., and de Vos, A. M. (1998). Structure 6, 1153±1167. Mylvaganam, S. E., Paterson, Y., and Getzoff, E. D. (1998). J. Mol. Biol. 281, 301±322. Nicholls, A., Sharp, K. A., and Honig, B. (1991). Proteins Struct. Funct. Genet. 11, 281±296. Ooi, T., Oobatake, M., Nemethy, G., and Sheraga, H. A. (1987). Proc. Natl. Acad. Sci. USA 84, 3086±3090. Padlan, E. A. (1990). Proteins 7, 112±124. Padlan, E. A., Silverton, E. W., Sheriff, S., Cohen, G. H., Smith-Gill, S. J., and Davies, D. R. (1989). Proc. Natl. Acad. Sci. USA 86, 5938±5942. Pan, Y., Yuhasz, S. C., and Amzel, L. M. (1995). FASEB J. 9, 43±49. Patten, P. A., Gray, N. S., Yang, P. L., Marks, C. B., Wedemayer, G. J., Boniface, J. J., Stevens, R. C., and Schultz, P. G. (1996). Science 271, 1086±1091. Poljak, R. J. (1994). Proc. Natl. Acad. Sci. USA 91, 1599±1600. Prasad, L., Sharma, S., Vandonselaar, M., Quail, J. W., Lee, J. S., Waygood, E. B., Wilson, K. S., Dauter, Z., and Delbaere, L. T. (1993). J. Biol. Chem. 268, 10705±10708. Rajewsky, K. (1996). Nature 381, 751±758. Ramanadham, M., Sieker, L. C., and Jensen, L. H. (1990). Acta Crystallogr. Sect. B 46, 63±69. Reynolds, J. A., Gilbert, D. B., and Tanford, C. (1977). Proc. Natl. Acad. Sci. USA 74, 2925±2929. Rini, J. M., Schulze-Gahmen, U., and Wilson, I. A. (1992). Science 255, 959±965. Rini, J. M., Stan®eld, R. L., Stura, E. A., Salinas, P. A., Profy, A. T., and Wilson, I. A. (1993). Proc. Natl. Acad. Sci. USA 90, 6325±6329. Romesberg, F. E., Spiller, B., Schultz, P. G., and Stevens, R. C. (1998). Science 279, 1929±1933.

160

ERIC J. SUNDBERG AND ROY A. MARIUZZA

Roost, H. P., Bachmann, M. F., Haag, A., Kalinke, U., Pliska, V., Hengartner, H., and Zinkernagel, R. M. (1995). Proc. Natl. Acad. Sci. USA 92, 1257±1261. Schreiber, G., and Fersht, A. R. (1995). J. Mol. Biol. 248, 478±486. Serrano, L., Horovitz, A., Avron, B., Bycroft, M., and Fersht, A. R. (1990). Biochemistry 29, 9343±9352. Sharp, K. A., Nicholls, A., Fine, R. F., and Honig, B. (1991). Science 252, 106±109. Sheriff, S., Silverton, E. W., Padlan, E. A., Cohen, G. H., Smith-Gill, S. J., Finzel, B. C., and Davies, D. R. (1987). Proc. Natl. Acad. Sci. USA 84, 8075±8079. Shoichet, B. K., and Kuntz, I. D. (1996). Chem. Biol. 3, 151±156. Shortle, D., Stites, W. E., and Meeker, A. K. (1990). Biochemistry 29, 8033±8041. Sigurskjold, B. W., Altman, E., and Bundle, D. R. (1991). Eur. J. Biochem. 197, 239±246. Stan®eld, R. L., Fieser, T. M., Lerner, R. A., and Wilson, I. A. (1990). Science 248, 712±719. Stan®eld, R. L., Takimoto-Kamimura, M., Rini, J. M., Profy, A. T., and Wilson, I. A. (1993). Structure 1, 83±93. Sundberg, E. J., and Mariuzza, R. A. (2000). Structure Fold. Des. 8, R137±142. Sundberg, E. J., Urrutia, M., Braden, B. C., Isern, J., Tsuchiya, D., Fields, B. A., Malchiodi, E. L., Tormo, J., Schwarz, F. P., and Mariuzza, R. A. (2000). Biochemistry 39, 15375±15387. Takano, K., Ogasahara, Kaneda, H., Yamagata, Y., Fujii, S., Kanaya, E., Kikuchi, M., Oobatake, M., and Yutani, K. (1995). J. Mol. Biol. 254, 62±76. Takano, Y., Yamagata, Y., Fujii, S., and Yutani, K. (1997). Biochemistry 36, 688±698. Tomlinson, I. M., Cox, J. P., Gherardi, E., Lesk, A. M., and Chothia, C. (1995). EMBO J. 14, 4628±4638. Tomlinson, I. M., Walter, G., Jones, P. T., Dear, P. H., Sonnhammer, E. L., and Winter, G. (1996). J. Mol. Biol. 256, 813±817. Tonegawa, S. (1983). Nature 302, 575±581. Tormo, J., Blaas, D., Parry, N. R., Rowlands, D., Stuart, D., and Fita, I. (1994). EMBO J. 13, 2247±2256. Tulip, W. R., Varghese, J. N., Laver, W. G., Webster, R. G., and Colman, P. M. (1992). J. Mol. Biol. 227, 122±148. Ulrich, H. D., Mundorff, E., Santarsiero, B. D., Driggers, E. M., Stevens, R. C., and Schultz, P. G. (1997). Nature 389, 271±275. Wallqvist, A., and Covell, D. G. (1996). Proteins 25, 403±419. Wang, J., Lim, K., Smolyar, A., Teng, M., Liu, J., Tse, A. G., Liu, J., Hussey, R. E., Chishti, Y., Thomson, C. T., Sweet, R. M., Nathenson, S. G., Chang, H. C., Sacchettini, J. C., and Reinherz, E. L. (1998). EMBO J. 17, 10±26. Webster, D. M., Henry, A. H., and Rees, A. R. (1994). Curr. Opin. Struct. Biol. 4, 123±129. Wedemayer, G. J., Patten, P. A., Wang, L. H., Schultz, P. G., and Stevens, R. C. (1997a). Science 276, 1665±1669. Wedemayer, G. J., Wang, L. H., Patten, P. A., Schultz, P. G., and Stevens, R. C. (1997b). J. Mol. Biol. 268, 390±400. Wells, J. A., and de Vos, A. M. (1996). Annu. Rev. Biochem. 65, 609±634. Willcox, B. E., Gao, G. F., Wyer, J. R., Ladbury, J. E., Bell, J. I., Jakobsen, B. K., and van der Merwe, P. A. (1999). Immunity 10, 357±365. Wilson, I. A., and Stan®eld, R. L. (1993). Curr. Opin. Struct. Biol. 3, 113±118. Wilson, I. A., and Stan®eld, R. L. (1994). Curr. Opin. Struct. Biol. 4, 857±867. Xu, J., Baase, W. A., Baldwin, E., and Matthews, B. W. (1998). Protein Sci. 7, 158±177. Yang, P. L., and Schultz, P. G. (1999). J. Mol. Biol. 294, 1191±1201. Ysern, X., Fields, B. A., Bhat, T. N., Goldbaum, F. A., Dall'Acqua, W., Schwarz, F. P., Poljak, R. J., and Mariuzza, R. A. (1994). J. Mol. Biol. 238, 496±500. Yutani, K., Ogasahara, K., Tsujiya, T., and Sugino, Y. (1987). Proc. Natl. Acad. Sci. USA 84, 4441±4444.

MOLECULAR RECOGNITION BY SH2 DOMAINS BY J. MICHAEL BRADSHAW AND GABRIEL WAKSMAN Department of Biochemistry and Molecular Biophysics, Washington University School of Medicine, St. Louis, Missouri 63110

I. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . II. Single SH2 Domain Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A. Fold and Architecture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B. pTyr Recognition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . C. Speci®city Determining Interactions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . III. Functional Analysis of Single SH2 Domain Binding . . . . . . . . . . . . . . . . . . . . . . A. Determinants of pTyr Recognition by SH2 Domains . . . . . . . . . . . . . . . . . . B. Determinants of Speci®c SH2 Domain Recognition . . . . . . . . . . . . . . . . . . . C. Implications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . IV. Structure and Function of SH2 Domains in the Context of Other Protein Modules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A. Structures. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B. Solution Studies of Multiple Protein Modules . . . . . . . . . . . . . . . . . . . . . . . . V. Unusual SH2 Domain Interactions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A. The SAP SH2 Domain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B. The Cbl SH2 Domain. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . C. SH2 Domain±Phospholipid Interactions? . . . . . . . . . . . . . . . . . . . . . . . . . . . VI. SH2 Domains as Drug Targets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A. Dif®culties in Targeting Pharmaceuticals to SH2 Domains. . . . . . . . . . . . . . B. Src SH2 Domain Binding Inhibitors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . C. Grb2 SH2 Domain Binding Inhibitors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . VII. Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

161 164 164 165 166 172 172 176 183 184 185 192 197 198 201 201 202 202 203 203 204 205

I. INTRODUCTION Src homology 2 (SH2) domains are 100-amino-acid protein modules found in many proteins involved in tyrosine kinase signaling cascades (Pawson and Gish, 1992; Pawson and Schlessinger, 1993). Their function is to bind tyrosine phosphorylated sequences in speci®c target proteins (Koch et al., 1991). A general illustration of the biological role of SH2 domains in signaling pathways is shown in Fig. 1. Binding of an extracellular ligand to its receptor induces dimerization of the receptor, which brings the intracellular tyrosine kinase domains of the receptor molecules into proximity. This allows the kinase domains to phosphorylate speci®c tyrosine residues in the receptor and nearby proteins. These phosphorylated tyrosines then become docking sites for SH2 domains, serving to recruit SH2 domain-containing proteins to the cell membrane and/ or alter their enzymatic activity; in either case, SH2 domain binding facilitates propagation of the signal. Hence, binding of an SH2 domain 161 ADVANCES IN PROTEIN CHEMISTRY, Vol. 61

Copyright 2003, Elsevier Science (USA). All rights reserved. 0065-3233/03 $35.00

162

J. MICHAEL BRADSHAW AND GABRIEL WAKSMAN

FIG. 1. Schematic model of signal transduction involving SH2 domains. Extracellular growth factors (purple), the extracellular domains of receptor tyrosine kinases (yellow), intracellular tyrosine kinase domains (green), and SH2 domains (red) are shown. Sites of tyrosine phosphorylation are depicted by the letter P

to its particular tyrosine phosphorylated sequence links receptor activation to downstream signaling both to the nucleus to regulate gene expression and throughout the cytoplasm of the cell. Originally identi®ed as regions of homology between the src and fps genes (Sadowski et al., 1986), SH2 domains have now been identi®ed in over 100 different protein sequences in humans (Rose et al., 1999). Evolutionarily, they are found in both simple and complex metazoans, but not in unicellular organisms, underscoring the concept that they have evolved for intercellular communication processes (Brown and Cooper, 1996; Kawata et al., 1997). Due to their role in many signaling pathways, SH2 domains have received much attention as potential targets of pharmaceuticals (Brugge, 1993; Sawyer, 1998). Indeed, it should be noted that inappropriate signaling through tyrosine kinases has been linked to many pathologic conditions (i.e., oncogenesis, autoimmune disease, asthma, allergies). Hence, there is a signi®cant effort aimed at attempting to discover ligands with high af®nity and speci®city for particular SH2 domains.

MOLECULAR RECOGNITION BY SH2 DOMAINS

163

It is appropriate that this volume devoted to protein±protein interactions contains a chapter on SH2 domains since the study of these domains has signi®cantly broadened the way the scienti®c community views these interactions. In fact, two particularly important current ideas regarding the role of protein±protein interactions in biology emerged from the discovery and subsequent functional characterization of SH2 domains (Hunter, 2000). One of these is that many intracellular proteins contain small structural units which exist for the sole purpose of facilitating protein±protein interactions. These units are independently folded and can be expressed, puri®ed, and studied in vitro. Examples include SH2, SH3, WW, FHA, SAM, LIM, PX, EH, EVH1, and PDZ domains (Pawson and Scott, 1997). Each of these modules, including SH2 domains, has been found to function by engaging a speci®c sequence motif in the target and hence localizing the domain-containing protein to a particular site within the cell. A second idea about protein±protein interactions which has arisen from the study of SH2 domains is that protein±protein interactions within the cell can be regulated (i.e., turned on and off) by extracellular stimuli. In the case of SH2 domains, this occurs due to selective binding of the SH2 domain to the tyrosine phosphorylated, as opposed to unphosphorylated, state of the target. Hence, the study of SH2 domains revealed for the ®rst time that reversible protein±protein interactions are a component of the signaling ``switch'' which induces many cellular processes such as mitosis, differentiation, and apoptosis. Two questions regarding the interaction between SH2 domains and their targets have been paramount in understanding how these domains function. The ®rst deals with the molecular basis of switch [or phosphotyrosine (pTyr) ] recognition: How do SH2 domains bind selectively to the tyrosine phosphorylated state of targets? The second involves the speci®city of SH2 domain binding: How do particular SH2 domains seem to recognize and bind only certain tyrosine phosphorylated sequences? In this chapter, we describe the current understanding of how SH2 domains accomplish both of these tasks. It will become apparent that SH2 domains have two different regions devoted to the two functions described above (Kuriyan and Cowburn, 1997). The ®rst, the so-called ``pTyr binding cavity,'' is a ``switch sensing region'' that distinguishes whether or not the target is tyrosine phosphorylated. The second region, the so-called ``speci®city determining region,'' evaluates the identity of the residues ¯anking the pTyr, hence providing the basis of SH2 domain binding speci®city. One of the key recent ®ndings regarding SH2 domains has been that the two regions described above are not equally effective at performing their functions (Bradshaw et al., 1999, 2000). While the pTyr binding pocket is highly adept at binding only the tyro-

164

J. MICHAEL BRADSHAW AND GABRIEL WAKSMAN

sine phosphorylated state of targets (Bradshaw et al., 1999), the speci®city determining region of SH2 domains is only modestly discriminating when it comes to binding target sequences (Bradshaw et al., 2000). This is a somewhat puzzling observation: low speci®city levels in SH2 domain binding would result in some degree of cross-talk among signaling pathways. Until recently, this was believed to be detrimental to the proper functioning of the cell and its ability to respond speci®cally to environmental stimuli. However, recent evidence has shown that signal transduction pathways are interconnected to form networks (reviewed in Pawson and Saxton, 1999). Therefore, in this context, it may not be so surprising that single SH2 domains should display only modest degrees of selectivity.

II. SINGLE SH2 DOMAIN STRUCTURE The importance of SH2 domains in tyrosine kinase signaling initially became evident with the discovery that this module was common to many proteins which bind the activated, tyrosine phosphorylated state of receptors and other cellular proteins (Escobedo et al., 1991; Koch et al., 1991; Matsuda et al., 1990; Mayer and Hanafusa, 1990a, b). SH2 domains were con®rmed to be critical for signaling by experiments which demonstrated that these domains are both necessary and suf®cient for binding of cellular proteins to tyrosine phosphorylated sequences (Anderson et al., 1990; Moran et al., 1990) and also by studies which showed that introducing phenylalanine in place of tyrosine in potential SH2 domain targets abrogates both SH2 domain attachment and downstream signaling events (Fantl et al., 1992, 1993). A further critical early advance in the study of SH2 domain recognition came with the realization that SH2 domain±target interactions could be studied in vitro using short tyrosine phosphorylated peptides based on regions of the cellular targets of SH2 domains (Mayer et al., 1991, 1992). This discovery opened the door for the detailed structural investigations of SH2 domains that are the subject of this chapter. A. Fold and Architecture The ®rst SH2 domain structures to be solved were those of the Src tyrosine kinase (Waksman et al., 1992), the Abl tyrosine kinase (Overduin et al., 1992), and the N-terminal SH2 domain of the p85 subunit of the PI 30 -kinase (N-p85 SH2 domain) (Booker et al., 1992). These structures revealed the architecture of SH2 domains. The SH2 domain fold is relatively simple: it consists of a central antiparallel b-sheet ¯anked by

MOLECULAR RECOGNITION BY SH2 DOMAINS

165

two a-helices (Fig. 2A). A common nomenclature for SH2 domains has been established such that the strands are labeled bA through bG, with the helices referred to as aA and aB (Eck et al., 1993). In the Src SH2 domain, the module begins with the bA strand, followed by the aA helix which packs against the strands bB, bC, and bD that form the core of the central b-sheet of the domain. The bD strand is followed by a b-meander formed from an extension of the bD strand (referred to as bD0 ) as well as strands bE and bF. This element then leads to the second helix, aB. A loop (referred to as the BG loop) connects the aB helix to the bG strand. Strand bG forms the ®nal strand of the central b-sheet. The modular nature of the SH2 domain fold is illustrated by the fact that the N- and C-termini of the SH2 domain are found in close proximity to one another. The architecture of SH2 domains is remarkably conserved: there has been little variation observed in the SH2 domain fold in the multiple SH2 domain structures which have now been determined (reviewed in Kuriyan and Cowburn 1997). The structure of SH2 domains can be conveniently divided into two functional regions: that involved in coordination of the pTyr of the target and that which contacts the residues C-terminal to the pTyr (Fig. 2A). The pTyr binding cavity is found in the N-terminal half of the protein between the central b-sheet and helix aA, while the speci®city determining region is found primarily in the C-terminal half of the protein between the central b-sheet and helix aB (Fig. 2A). Each of these regions is described in detail below. B. pTyr Recognition The structure of the Src SH2 domain in complex with two low-af®nity tyrosyl phosphopeptides elucidated the structural basis for pTyr recognition by SH2 domains (Waksman et al., 1992). The pTyr is stabilized by a dense network of hydrogen bonds and ionic interactions contributed by SH2 domain residues forming a deep cavity, the pTyr binding pocket (Fig. 2B). Most noteworthy, a universally conserved arginine residue (Arg bB5) at the center and base of the cavity makes a bidentate ionic interaction with two oxygens of the phosphate group. Arg bB5 is nearly completely solvent inaccessible in the free form of the protein. In the bound form, the ionic interaction which Arg bB5 makes with the phosphate is also entirely removed from solvent. Two other positively charged residues are located in the pTyr binding pocket of the Src SH2 domain: Arg aA2 and Lys bD6 (Fig. 2B). Arg aA2 interacts with the phosphate group and also makes an unusual amino± aromatic interaction with the phenol ring of the pTyr. This type of

166

J. MICHAEL BRADSHAW AND GABRIEL WAKSMAN

interaction arises from favorable interactions between the imido group of the arginine and the p electrons of the aromatic ring; it has been estimated that the strength of this interaction is close to that of a conventional hydrogen bond between uncharged groups (1 to 2 kcal/mol). The other positively charged residue, Lys bD6, is located on the other side of the pTyr's phenol ring from Arg aA2 so that the two residues together form a clamp around the pTyr. The pTyr binding pocket of the Src SH2 domain demonstrates several other interesting features. A histidine residue (His bD4) which had been previously hypothesized to be important for pTyr recognition forms a wall of the binding cavity but does not make any direct interactions with the pTyr. However, two hydrogen bond donating residues (Ser bB7, Thr BC2) are observed to hydrogen bond directly with the phosphate. Finally, a free cysteine residue (Cys bC3) is located in close proximity to both Arg bB5 and the phosphate group; this residue is unique to the Src SH2 domain, with most other SH2 domains having a serine, threonine, valine, or alanine at this position. The conserved nature of the residues in the Src SH2 domain participating in pTyr binding suggested that many features of pTyr recognition would be conserved among SH2 domains: the structures of many other SH2 domains have revealed that this is indeed the case (Kuriyan and Cowburn, 1997). For example, the ionic interaction between Arg bB5 and the phosphate is a universally conserved feature of SH2 domain recognition. However, the amino±aromatic interaction involving Arg aA2 is not conserved; while it is present in many SH2 domain structures [for example, that of the Src family tyrosine kinase Lck (Eck et al., 1993) and of the C-terminal SH2 domain of phospholipase Cg (C-PLCg SH2 domain (Pascal et al., 1994) ], it is absent in others [the N-p85 SH2 domain, for example (Nolte et al., 1996) ]. Furthermore, the Nterminal SH2 domain of the SHP-2 phosphatase (N-SHP-2 SH2 domain) lacks Arg aA2 altogether (Lee et al., 1994). Hence, the structural data suggest that the amino±aromatic interaction involving Arg aA2 is not essential for pTyr recognition, an observation which has been con®rmed by mutational studies (see Section III). C. Speci®city Determining Interactions While the interactions at the pTyr binding pocket are generally similar for all SH2 domains, those which involve residues other than pTyr are not. These interactions help to determine the speci®city of SH2 domain±target recognition. Comparison of the many SH2 domain±phosphopeptide structures has revealed that different SH2 domains use somewhat different mechanisms to engage their respective targets.

MOLECULAR RECOGNITION BY SH2 DOMAINS

167

1. The ``Two-Pronged Plug±Two-Holed Socket'' Mode of Binding: Src Family SH2 Domains and the N-Terminal and C-Terminal p85 SH2 Domains Information about the speci®city of SH2 domain binding came initially from the structures of the Src and Lck SH2 domains in complex with tyrosyl phosphopeptides with the sequence Glu, Glu, Ile (EEI) C-terminal to the pTyr (pY) (Eck et al., 1993; Waksman et al., 1993). This pYEEI sequence had previously been identi®ed to be speci®c for the Src family of SH2 domains, which includes Lck (see Section III). The Src and Lck structures are very similar to one another. Although the peptides employed in both studies contained residues outside the pYEEI motif, the structures indicate that only these four residues contact the SH2 domain. The peptide binds the SH2 domains perpendicular to the central b-sheet in an extended conformation (Fig. 2A). The pTyr makes contacts very similar to those observed for the structure of the Src SH2 domain in complex with low-af®nity tyrosyl phosphopeptides (see above). However, in addition, the EEI motif makes speci®c contact with other residues of the protein. The interface between the Src and Lck SH2 domains and the pYEEI Ê 2 of surface area buried peptide is quite complementary (1066 A upon binding in the Src system, for example). In both the Src and Lck structures, the glutamate residue one position C-terminal to the pTyr (1 Glu) interacts along the surface of the SH2 domain with a positively charged residue, Lys bD3 (Fig. 2C). However, this interaction is longÊ ). In the Src SH2 domain structure, the glutamate two ranged (4.5 A positions C-terminal to the pTyr (2 Glu) also interacts with a positively charged residue, Arg bD0 1, of the SH2 domain. This interaction is also long-ranged (Fig. 2C) and, in one of the three molecules which the asymmetric unit of the crystal contains, it is mediated by a water molecule network. In the Lck SH2 domain structure, the 2 Glu interacts with a symmetry-related molecule. The most striking interaction between the Src family SH2 domains and the EEI motif is the burial of the isoleucine 3 residues C-terminal to the pTyr (3 Ile) in a hydrophobic binding cavity, referred to as the hydrophobic (or 3) binding pocket (Figs. 2C and 3C). This pocket is formed by several residues including Tyr bD5, Leu BG4, Ile bE4, Thr EF1, and Tyr aB9. The extensive interactions between the 3 Ile of the peptide and its binding pocket prompted the twopronged plug±two-holed socket model of high-af®nity SH2 domain binding which suggested that a pTyr and a large hydrophobic residue at the 3 position of the peptide must occupy the pTyr and 3 position binding cavities, respectively, for high-af®nity binding (Waksman et al., 1993).

168

J. MICHAEL BRADSHAW AND GABRIEL WAKSMAN

pTyr-binding Region

CD

A

βD βC

αA

βB βG BG +1Glu +3Ile Specificity Determining Region

pTyr

+2Glu EF

αB

BC βF βD'

βE

DE

B αA βD

βB SerβC5 ArgαA2 βC

ArgβB5

HisβD4 βA

CysβC3

AA

GluBC1

pTyr

ThrBC2 SerβB7 LysβD6

βD'

C

ThrBC3 BC

βD

LysβD3 βG TyrβD5 BG LeuBG4 TyrαB9 AspBG2

ArgEF3

+1Glu

+3Ile +2Glu ThrEF1 IleβE4 αB EF

ArgbD'1

βF FB

βE

βD'

DE

FIG. 2. Structures of SH2 domains. (A) Stereo ribbon diagram of an SH2 domain. The structure shown is the SH2 domain of the Src kinase complexed with the high-

MOLECULAR RECOGNITION BY SH2 DOMAINS

169

The structure of the N- and C-terminal SH2 domains of the p85 subunit of PI-30 kinase (N- and C-p85 SH2 domains) in complex with high-af®nity tyrosyl phosphopeptides revealed that these SH2 domains also demonstrate a binding mode consistent with the two-pronged plug± two-holed socket mechanism of binding (Breeze et al., 1996; Nolte et al., 1996). PI 30 -kinase is an enzyme which speci®cally phosphorylates the 30 position of phosphatidylinositols. It becomes activated on engagement of its SH2 domains to particular pTyrs on various cell surface receptors. In the p85 SH2 domain structures, the peptides also bind across the front face of the SH2 domain with only the 3 amino acids C-terminal to the pTyr making signi®cant interactions with the protein. Furthermore, the residue at the 3 position (a methionine in both peptides) also occupies a deep hydrophobic binding pocket. The structures of the p85 SH2 domains also revealed signi®cant differences in binding compared to the Src family SH2 domains. It had previously been determined that the p85 SH2 domains bound speci®cally to the sequence Val/Met-X-Met (V/M-X-M) following the pTyr (see Section III). The structures of the p85 SH2 domains revealed the molecular basis of this preference. First, the 3 position binding pocket of the p85 SH2 domains is deeper and narrower than that of Src, explaining the preference for Met, rather than Ile, at the 3 position (Fig. 2D). Second, a shallow hydrophobic pocket is present in the p85 but not the Src SH2 domains to accommodate a hydrophobic residue at the 1 position (Fig. 2D). Other interesting features of the N-p85 SH2 domain structure include an amphipathic helix at its N-terminus which is absent in other SH2 domains (Nolte et al., 1996). Also, in its unbound form, the N-p85 SH2 domain has a residue (Tyr BG5) occupying its 3 position binding pocket; this residue is displaced on peptide binding. af®nity pYEEI tyrosyl phosphopeptide (Waksman et al., 1993). The a-helices, b-sheets, and loops of the SH2 domain are depicted as cyan ribbons, green arrows, and single orange lines, respectively. The notation used to label secondary strand elements is that of Eck et al. (1993) and Waksman et al. (1993). The pTyr, 1 Glu, 2 Glu, and 3 Ile of the pYEEI peptide are depicted as black solid bonds. The pTyr binding pocket of the SH2 domain, as well as the speci®city determining region of the protein, is also identi®ed. (B) Stereo diagram of Src SH2 domain's pTyr binding pocket. Colorcoding and labeling are as in A. The pTyr and the SH2 domain residues of the pTyr binding cavity are shown in ball-and-stick representation. Atoms are colorcoded gray, black, red, blue, yellow, and magenta for carbon atoms in the SH2 domain, carbon atoms in the pTyr, oxygen atoms, nitrogen atoms, sulfur atoms, and phosphorus atoms, respectively. (C) Stereo diagram of the speci®city determining region of the Src SH2 domain. Color-coding and labeling are as in A and B. Shown are the 1 Glu, 2 Glu, and 3 Ile of the pYEEI peptide and the speci®city determining residues of the Src SH2 domain. Note the broad, hydrophobic binding pocket which is formed for the 3 Ile of the peptide (also depicted in Fig. 3C).

170

J. MICHAEL BRADSHAW AND GABRIEL WAKSMAN

βD LeuBG9 LysβD3

D

AsnBG6 +1Val CysβD5

BG LeuBG2

+3Met TyrαB9

+2Pro

PheβE4 αB

AlaEF1 EF βD'

βE DE

LysBG5

E

βD

LysβD3 TyrBG3

ArgBG4

+6Asp +5Pro BG +3Pro

CysβD5

+2Ile LeuβE4 EF

+4Leu

AsnEF2

+1Ile

αB

βD'

βE βF FB DE

GlnβD3

F

βD

BG

PheβD5

+Val

TrpEF1 αB EF

+2Asn +3Val LeuβE4 LysβD6 LeuβD'1

βF

βE

DE

βD'

MOLECULAR RECOGNITION BY SH2 DOMAINS

171

2. The ``Open Groove'' Mode of Binding: The SHP-2 Phosphatase N-Terminal SH2 Domain and the C-Terminal SH2 Domain of Phospholipase Cg The two-pronged plug±two-holed socket model is not applicable to all SH2 domain±phosphopeptide interactions. For example, several SH2 domains make contacts with their peptide targets beyond the 3 position of the peptide: these include the N-terminal SH2 domain of the SHP-2 phosphatase (N-SHP-2 SH2 domain) and the C-terminal SH2 domain of PLCg (C-PLCg SH2 domain). The structure of the N-SHP-2 SH2 domain in complex with peptides based on the platelet±derived growth factor receptor (PDGFR) and the adapter protein IRS1 revealed that the interaction between the protein and the peptide is extended to the residue 5 positions C-terminal to the pTyr (Lee et al., 1994). In these structures, the 3 position binding pocket is ``opened'' in order to create such an extended interface. The interactions between the peptide and the protein are primarily hydrophobic. For example, a Phe at the 5 position of the IRS1 peptide binds between the EF and BG loops and interacts with the Ile at the 3 position (Lee et al., 1994). Other than this extended interface, the interactions in this structure are reminiscent of those observed in the Src SH2 domain structure: the peptide binds in an extended conformation perpendicular to the central b-sheet. The nuclear magnetic resonance (NMR) structure of the C-PLCg SH2 domain in complex with a tyrosyl phosphopeptide from the PDGFR has also revealed that an extended groove forms the binding interface (Fig. 2E (Pascal et al., 1994). Here, the groove is also relatively hydrophobic in nature. This presumably allows the SH2 domain to bind the hydrophobic residues of the ligand C-terminal to the pTyr (1 Ile, 2 (D) Stereo diagram of the speci®city determining region of the C-p85 SH2 domain (Breeze et al., 1996). Color-coding and labeling are as in C. The three residues Cterminal to the pTyr of a high-af®nity binding peptide (1 Val, 2 Pro, 3 Met) as well as important residues of the C-p85 SH2 domain are shown (in ball-and-stick representation and color-coded in black and gray, respectively). Note the narrow but deep hydrophobic binding pocket formed for the 3 Met of the peptide. (E) Stereo diagram of the speci®city determining region of the C-PLCg SH2 domain (Pascal et al., 1994). The six residues C-terminal to the pTyr of the peptide and the protein residues involved in contacts with this region of the peptide are shown (in ball-and-stick representation and color-coded in black and gray, respectively). Note the open groove adopted by the protein in order to contact residues of the peptide beyond the 3 position. (F) Stereo diagram of the speci®city determining region of the Grb2 SH2 domain (Rahuel et al., 1996). The three residues C-terminal to the pTyr in the peptide (1 Val, 2 Asn, 3 Val) and the important residues of the C-terminal half of the SH2 domain are shown. Note the location of Trp EF1 occluding the 3 binding pocket and the orientation of the peptide which forms a b-turn. This ®gure was generated using the program Ribbons (Carson, 1997).

172

J. MICHAEL BRADSHAW AND GABRIEL WAKSMAN

Ile, 3 Pro, 4 Leu, 5 Pro, 6 Asp). Hence, like the two±pronged plug±two-holed socket type of interaction, the open groove binding mechanism appears to be a common way for SH2 domains to bind their targets. 3. The ``b-Turn'' Mode of Binding: The Grb2 SH2 Domain The structure of the Grb2 SH2 domain in complex with a high-af®nity tyrosyl phosphopeptide revealed an entirely new mode of speci®c SH2 domain interaction (Rahuel et al., 1996). It had previously been demonstrated that the Grb2 SH2 domain is unusual among SH2 domains in showing a strong selection preference for an asparagine (Asn) residue at the 2 position of the peptide (see Section III). In the Grb2 SH2 domain structure, the cavity which normally forms the 3 binding pocket is occluded due to the fact that a Trp residue (instead of Thr in Src) is present at the EF1 position of the protein (Fig. 2F). This prevents a hydrophobic residue at the 3 position of the peptide from occupying this cavity. To bind with high af®nity, the peptide must adopt a new binding conformation: it forms a b-turn at the 2 position. This interaction is stablized due to the interactions between the 2 Asn and the protein: the Od1 and Nd2 atoms of the Asn side chain both make hydrogen bonds with backbone atoms of the SH2 domain. Hence, the Grb2 SH2 domain shows that an entirely different peptide binding orientation can be used by SH2 domains to establish speci®c interactions with their targets.

III. FUNCTIONAL ANALYSIS OF SINGLE SH2 DOMAIN BINDING A. Determinants of pTyr Recognition by SH2 Domains Signaling cascades involving SH2 domain-containing proteins are initiated when SH2 domains engage sites of tyrosine phosphorylation on receptors and cellular proteins. Hence, the integrity of signaling pathways critically depends on the speci®c recognition of the phosphorylated state of tyrosine residues by SH2 domains. The question of how SH2 domains accomplish this task is the subject of this section. Nearly all SH2 domains require tyrosine phosphorylation of their targets for recognition (for the exception, see Section V). This has been established from numerous studies which have compared the binding of phosphorylated and dephosphorylated peptides to SH2 domains (Bibbins et al., 1993; Domchek et al., 1992; Gilmer et al., 1994; Lemmon and Ladbury, 1994; Mayer et al., 1992; Piccione et al., 1993). These binding experiments have established that high-af®nity tyrosyl

MOLECULAR RECOGNITION BY SH2 DOMAINS

173

phosphopeptide binding to SH2 domains occurs with a dissociation constant (Kd ) in the range of 100 nM to 1 mM, depending on the SH2 domain (Ladbury et al., 1995). In these initial experiments the af®nities of dephosphorylated peptides were too weak to be measured, demonstrating that phosphorylation of tyrosine plays a crucial role in binding. A quantitative estimate of the importance of tyrosine phosphorylation in SH2 domain recognition was determined from a study which measured the af®nity of a dephosphorylated peptide for the SH2 domain of the Src kinase (Bradshaw et al., 1999). In this work, titration calorimetry was used to establish the af®nity of the Src SH2 domain for its consensus high-af®nity peptide (i.e., a peptide containing the EEI motif C-terminal to the pTyr; also referred to as ``the pYEEI peptide''), while a competitive binding approach was employed to determine the af®nity of the dephosphorylated form of the same peptide. Binding of the dephosphorylated peptide was found to occur with a very weak af®nity of 2 mM, a value which is over 4 orders of magnitude weaker than the binding of the corresponding phosphorylated peptide (Bradshaw et al., 1999). Hence, SH2 domains are very ef®cient at binding only the phosphorylated state of tyrosine residues. Further experiments have shown that SH2 domains bind with high af®nity only to sequences that are phosphorylated on tyrosine residues. For instance, it has been shown that a phosphoserine-containing peptide binds the Src SH2 domain just as weakly as a dephosphorylated tyrosinecontaining peptide (Bradshaw et al., 1999). Furthermore, 3-phosphohistidine has also been demonstrated to be a very poor substitute for pTyr in SH2 domain recognition (Senderowicz et al., 1997). Hence, the positioning of the phosphate deep within the pTyr binding cavity is critical for high-af®nity recognition. Investigations which have probed the binding of just the amino acid pTyr (or close variants thereof) to SH2 domains have underscored the dominant role played by just this single amino acid in SH2 domain binding. It was initially qualitatively observed that pTyr weakly associates with SH2 domains (Lemmon and Ladbury, 1994; Mayer et al., 1992). Competition binding experiments then revealed that pTyr-like mimetics have measurable af®nity for the Src and N-p85 SH2 domains at a Kd of 1 mM (Burke et al., 1995). These ®ndings have been con®rmed by direct calorimetric titrations to measure binding of pTyr alone to the Src SH2 domain, which indicated that pTyr binds with an af®nity of 333 mM at 258C (Bradshaw et al., 1999). The pTyr of tyrosyl phosphopeptides is a negatively charged amino acid whose phosphate group carries a 2 charge at neutral pH. Experiments monitoring the pH dependence of SH2 domain binding have probed the importance of the phosphate's charge and have shown that

174

J. MICHAEL BRADSHAW AND GABRIEL WAKSMAN

the 2 form of this group is essential for high-af®nity recognition. For example, calorimetric studies of Src SH2 domain binding to the pYEEI phosphopeptide have indicated that the phosphate group undergoes a shift in pKa from 6.1 in the unbound form to 4.4 in the bound form (Bradshaw and Waksman, 1998). This pKa shift re¯ects the fact that the 2 form of the phosphate group binds almost 2 orders of magnitude more tightly than the 1 form. pH titration studies of the C-PLCg SH2 domain using NMR have indicated that a similar shift in phosphate pKa of about 2pKa units (from 6.1 to 4.0) occurs upon peptide binding to this SH2 domain (Singer and Forman-Kay, 1997). These results have shown that the highly basic nature of the pTyr binding cavity necessitates the presence of a divalently charged phosphate anion for high-af®nity binding. Studies of SH2 domain mutants have established which side chains are required for coordination of the pTyr. It has been determined that Arg bB5, the universally conserved residue which forms the base of the pTyr binding cavity, is by far the most important residue for recognition. Mutation of this residue in vivo abrogates SH2 domain attachment to pTyr-containing targets (Bibbins et al., 1993). In vitro, an Arg bB5 to Ala mutation in the Src SH2 domain showed a 500-fold decrease in tyrosyl phosphopeptide af®nity compared to wild type (Bradshaw et al., 1999). Mutation of Arg bB5 in other SH2 domains results in either an insoluble protein (Lemmon and Ladbury, 1994) or the complete elimination of detectable binding to tyrosine phosphorylated targets (Mayer et al., 1992). While Arg bB5 is the most critical residue for target recognition, other residues certainly contribute to the binding of pTyr. A detailed mutational investigation of the importance of residues in the pTyr binding cavity has been performed using the Src SH2 domain (Bradshaw Ê et al., 1999). Here, the importance of the 10 residues located within 5 A of the pTyr was investigated by mutating each residue individually to alanine (Fig. 3A). In addition to establishing the importance of Arg bB5 (as described above), it was found that the other positively charged residues of the cavity (Lys bD6, Arg aA2) as well as the groups which hydrogen bonded directly with the phosphate (Ser bB7, Thr BC2) contribute moderately to pTyr recognition (5- to 10-fold increase in Kd on mutation). However, elimination of several highly conserved residues which might have also been expected to be important (His bD4, Ser bC5) caused little effect on binding (less than 2-fold increase in Kd ). Hence, only a limited subset of pTyr binding pocket residues cooperate with Arg bB5 in coordinating the pTyr. This mutational study also revealed that loss of Cys bC3, a residue unique to the SH2 domain of

MOLECULAR RECOGNITION BY SH2 DOMAINS

175

FIG. 3. Energetic mapping of Src SH2 domain recognition. (A) The speci®city determining region. Shown is a molecular surface diagram of the Src SH2 domain obtained from the Src SH2 domain±p YEEI peptide crystal structure (Waksman et al., 1993) that was generated using the program GRASP (Nicholls et al., 1991). Residues are color-coded based on the change in the DG of pYEEI peptide binding compared to the wild-type pYEEI peptide interaction when that residue is changed to alanine, except for Tyr bD5 where the mutation is to Ile (DDG ). The color scheme is shown on the right. (B) The pTyr binding pocket. Shown is the same molecular surface as shown in A. Residues are color-coded based on the change in the DG of pYEEI peptide binding compared to the wild-type pYEEI peptide interaction when that residue is changed to alanine (DDG ). The color scheme is the same as in A and in shown on the left. (C) Electrostatic potential map of the Src SH2 domain surface calculated and displayed using the program GRASP (Nicholls et al., 1991). Shown is the solvent accessible surface of the Src SH2 domain. The surface is colored according to the local electrostatic potential; blue represents the most positive regions and red the most negative regions, with linear interpolation for values in between. The residues of the pYEEI peptide, as well as several charged residues of the SH2 domain, are depicted.

Src which has been successfully used to target pharmaceuticals (see Section VI), enhanced peptide binding af®nity by 4- to 8-fold (Bradshaw et al., 1999).

176

J. MICHAEL BRADSHAW AND GABRIEL WAKSMAN

The structural studies of SH2 domains together with the mutational investigations of the pTyr binding pocket suggested that electrostatic interactions play a central role in SH2 domain recognition. Another method which has been employed to demonstrate the importance of these interactions has been to examine the dependence of binding on salt concentration and type. Two SH2 domain systems have been studied: the tandem SH2 domain of the Syk kinase which binds to doubly phosphorylated peptides containing immunoreceptor tyrosine activation motifs (ITAMs; see Section IV) and the Src SH2 domain (Grucza et al., 2000). Here it was found that the association of SH2 domains and tyrosyl phosphopeptides is highly dependent on both salt concentration and type. For example, the slope of the log K vs log [NaCl] plot for binding of the Syk tandem SH2 domain to a doubly phosphorylated ITAM (dpITAM) peptide based on the Fcg receptor (an immune receptor to IgG) was 3:1, indicating that this interaction is one of the more salt± sensitive protein±protein interactions thus far studied. Studies of the salt dependence of pTyr binding alone and of monophosphorylated ITAM binding showed that the interaction between the pTyrs and their binding pockets is responsible for the majority of the salt dependence of binding (Grucza et al., 2000). These results provide further con®rmation of the critical role of electrostatic interactions in the pTyr binding pocket for high-af®nity SH2 domain binding. The experiments described in this section all indicate that SH2 domains are exquisitely selective in recognizing the pTyr residues in their targets; they bind very weakly (or not at all) to targets containing dephosphorylated tyrosines, other phosphorylated amino acids, and even pTyr with only a 1 (instead of a 2) net charge. This strict selection for tyrosine phosphorylation arises primarily from the large energetic bonus obtained upon formation of the buried ionic salt bridge between Arg bB5 and the phosphate. Hence, nature has ef®ciently designed a system where SH2 domains will bind tyrosine residues when they are phosphorylated upon receptor activation, but will not inappropriately bind a target (and hence inappropriately activate signal transduction) in the absence of a tyrosine phosphorylation signal. B. Determinants of Speci®c SH2 Domain Recognition The ability to bind tyrosine phosphorylated sequences is common to all SH2 domains. However, all SH2 domains do not bind equally well to all pTyr-containing sequences. Rather, SH2 domains show selectivity in the particular tyrosine phosphorylated sequence that they will engage. In this section, we describe studies that have uncovered the principles underlying SH2 domain binding speci®city.

MOLECULAR RECOGNITION BY SH2 DOMAINS

177

1. Initial Investigations of SH2 Domain Speci®city: Phosphopeptide Library Selection It was identi®ed early in the study of SH2 domains that these modules interact with only a subset of potential pTyr residues within the cell and that the sequence context of the pTyr was an important determinant of SH2 domain binding. For instance, the p85 subunit of PI-30 kinase (which contains two SH2 domains) was found to engage pTyrs in both the polyoma middle T antigen and the PDGFR that were followed by the sequence motif Met/Val-X-Met, where X is any amino acid (Auger et al., 1992; Cantley et al., 1991; Escobedo et al., 1991), suggesting that this motif is essential for the SH2 domains of PI-30 kinase to bind these receptors. Furthermore, it was found from mutagenesis of individual tyrosines within the PDGFR that speci®c SH2 domains (those of Ras GAP, PLCg, PI-30 kinase, etc.) associate with distinct pTyrs on that receptor, with the basis of selection presumably the amino acids ¯anking the pTyr (Fantl et al., 1992, 1993; Kazlauskas et al., 1992). A landmark investigation into the basis of speci®c SH2 domain recognition was the use of phosphopeptide library selection to identify the most preferred binding sequence of a series of different SH2 domains (Songyang et al., 1993, 1994). In these studies, the three positions Cterminal to the pTyr in tyrosyl phosphopeptides were randomized in order to identify the amino acid sequence which preferentially binds to a series of SH2 domains. It was found that different groups of SH2 domains preferentially select different motifs at the three positions tested. For example, all members of the Src family of SH2 domains tested bound preferentially to a peptide where pTyr was followed immediately by a Glu-Glu-Ile motif (EEI motif). A different group of SH2 domains selected a binding motif with a hydrophobic-X-hydrophobic pattern, where X is any amino acid. For example, the N-p85 SH2 domain selected the motif Val/Met-X-Met. The SH2 domain of Grb2 showed a markedly different selection pattern from either of these groups since it demonstrated a strong preference for asparagine at the 2 position C-terminal to the pTyr. Other SH2 domains demonstrated still different selection patterns. These phosphopeptide selection studies contributed greatly to our knowledge of the speci®city of SH2 domain interactions. In addition, the many subsequent crystallographic and NMR studies of SH2 domains (described in Section II) revealed many of the structural principles which determine the selection preference of SH2 domains. However, these studies were not suf®cient to fully elucidate the process of speci®c target recognition by SH2 domains: solution-based biophysical investigations of the energetics of SH2 domain binding have also been integral components of framing the current understanding of SH2 domain binding and

178

J. MICHAEL BRADSHAW AND GABRIEL WAKSMAN

the speci®city of SH2 domain interactions. Some of these investigations are described below. 2. The Src SH2 Domain Investigations of the binding of the Src SH2 domain to different tyrosyl phosphopeptides have con®rmed that a Glu-Glu-Ile (EEI) motif following the pTyr is the preferred binding sequence for high-af®nity binding, although the level of selectivity displayed for this sequence is modest (2 orders of magnitude in af®nity or less). For instance, the binding of the high-af®nity pYEEI peptide has been compared to that of two of the nonspeci®c tyrosyl phosphopeptides for which Src SH2 domain±peptide complex structures were also available (Bradshaw et al., 1998). One nonspeci®c peptide was based on a region of the PDGFR not thought to be the site of Src's interaction with that receptor, while the other nonspeci®c peptide was based on the C-terminal tail of the Src kinase. It was discovered that the pYEEI peptide bound the Src SH2 domain with an af®nity of about 200 nM (Kd ). However, the pYEEI peptide bound only about 25-fold more favorably than the PDGFR peptide and also only about 200-fold more favorably than the Ctail peptide. These ®ndings indicated that large differences in the sequence surrounding the pTyr cause only modest changes in the af®nity of Src SH2 domain±tyrosyl phosphopeptide interaction. In addition, it has been determined that introducing single substitutions at the 1, 2, or 3 positions C-terminal to the pTyr of the pYEEI peptide causes only small effects on binding (Bradshaw and Waksman, 1999). In this study, a substitution strategy was employed that involved making ®rst conservative (Glu to Gln for example) and then progressively more dramatic (Glu to Ala for example) mutations at each peptide position. Analysis of peptide binding revealed that conservative substitutions at each peptide position induce only very minor, 3-fold or less, decreases in af®nity compared to the wild-type pYEEI peptide. It was also found that the effect of alanine substitutions is generally small; the pYAEI, pYEAI, and pYEEA peptides bound only 2-, 6-, and 10-fold, respectively, more weakly than the pYEEI peptide (Bradshaw and Waksman, 1999). These modest mutational effects, especially in the 3 position of the peptide, were surprising based on the previous phosphopeptide library selection studies (Songyang et al., 1993, 1994) and the structural investigations of Src SH2 domain binding (Waksman et al., 1992, 1993). For instance, since the structure of the Src SH2 domain showed a distinct hydrophobic cavity that looked perfectly formed to ®t an Ile at the peptide's 3 position, the occupancy of this cavity by Ile was thought to be crucial for high-af®nity binding (Waksman et al., 1993): however, the thermodynamic analysis revealed that this is not the case.

MOLECULAR RECOGNITION BY SH2 DOMAINS

179

A frequently observed theme in studies of molecular recognition is that the effects of mutations are complementary: energetically less important residues from one binding partner interact with less important residues from the other binding partner (Clackson and Wells, 1995; Dall'Acqua et al., 1998; Goldman et al., 1997). A mutational investigation of the Src SH2 domain itself has probed the importance of the residues in the SH2 domain responsible for coordinating the EEI motif (the so-called speci®city determining residues) and determined that the majority of these residues make only small energetic contributions to binding (Bradshaw et al., 2000). In this investigation, individual mutation to alanine of six SH2 domain residues (Arg bD0 1, Leu BG4, Ile bE4, Thr EF1, Arg EF3, and Asp BG2) caused a 3-fold or less reduction in binding af®nity (Fig. 3B). However, mutation of Tyr bD5, the residue which forms a platform for the peptide backbone and contributes part of the hydrophobic 3 binding cavity as well as interaction with the Cb of the 1 Glu, induced a greater than 10-fold loss in af®nity on mutation. Furthermore, mutation of Lys bD3, the residue which coordinates the 1 Glu of the peptide, caused a 7-fold reduction in af®nity. The magnitude of binding af®nity lost on introducing a mutation within a protein±protein interface can be affected by the presence or absence of other residues in the same interface due to the web of interactions between binding subsites on proteins. A way of probing the relevance of these interactions is to construct double-mutant cycles (Ackers and Smith, 1985; Carter et al., 1984). For the Src SH2 domain±pYEEI peptide interaction, several double-mutant studies have probed the interactions between residues in the speci®city determining region of the interface (Bradshaw et al., 1999, 2000; Bradshaw and Waksman, 1999). In these studies, it has been frequently found that the coupling between residues is small, indicating that the residues in this region bind for the most part independent of one another. However, signi®cant coupling has been observed for one interaction, that between the 1 Glu of the peptide and Lys bD3 of the protein (Bradshaw et al., 2000). This coupling is due to a speci®city switch at the 1 position of the peptide from negatively charged to neutral or positively charged when Lys bD3 is removed (see below). 3. Other Src Family SH2 Domains Binding studies using other Src family member SH2 domains have revealed that these domains have binding properties similar to those of the Src SH2 domain. For instance, several different investigations have determined that Src family SH2 domains bind selectively to the pYEEI peptide, but the level of speci®city is modest: (1) competition binding assays using both the Lck and Src SH2 domains and radioactive peptides

180

J. MICHAEL BRADSHAW AND GABRIEL WAKSMAN

have indicated that these SH2 domains have a 1 to 2 order of magnitude binding preference for the pYEEI peptide compared to other tyrosyl phosphopeptides derived from potential SH2 domain attachment sites in the cell (Payne et al., 1993); (2) studies using the SH2 domain of Fyn (also a Src family kinase) have found that the pYEEI peptide binds this SH2 domain only about 40-fold more tightly than a peptide based on the Fyn C-terminus (sequence: pYQPG) (Ladbury et al., 1996); and (3) comparison of Lck SH2 domain binding to the pYEEI peptide and a peptide based on its C-terminus (sequence: pYQPQ) using a ¯uorescence-based binding technique yielded only an 11-fold difference in relative af®nity between peptides (Cousins-Wasti et al., 1996) (see also Section IV for the role of the C-terminus in regulating Src family kinases). Several other investigations have dissected the pYEEI motif to establish which residues are most crucial for binding to Src family SH2 domains (Bradshaw and Waksman, 1999; Cousins-Wasti et al., 1996; Gilmer et al., 1994; Morelock et al., 1995; Payne et al., 1994). In studies with the Lck SH2 domain, it was found that individual substitution of 1 Glu, 2 Glu, and 3 Ile to Gly caused from 10-fold to greater than 100-fold losses in af®nity (Cousins-Wasti et al., 1996; Morelock et al., 1995). Here it was observed that the largest loss in af®nity (>100-fold) occurred on loss of the 1 Glu, which was surprising given that the salt bridge made between this residue and the protein is solvent exposed and longranged. A titration calorimetry study performed with the Src SH2 domain which examined the binding of peptides with both alanine and glycine substituted in place of the 1 Glu (pYAEI and pYGEI peptides) helped clarify the importance of the 1 position residue for binding (Bradshaw and Waksman, 1999). Here the Gly substitution caused a relatively large loss in af®nity (50-fold) compared to the pYEEI peptide while the Ala mutation resulted in a much smaller change (2-fold); these ®ndings indicate that an interaction involving the Cb of the 1 Glu (likely its van der Waals contact with Tyr bD5) is energetically important for binding to Src family SH2 domains (Bradshaw and Waksman, 1999). 4. The N-p85 SH2 Domain The N-p85 SH2 domain is another SH2 domain whose binding speci®city has been investigated in detail. This SH2 domain has a selection preference for a Met at the position 3 C-terminal to the pTyr (Songyang et al., 1993). Studies aimed at testing the stringency of such a sequence requirement have yielded contradictory results. A scrambled peptide sequence binds the N-p85 SH2 domain only 60-fold more weakly than the consensus high-af®nity pYMXM sequence (Piccione et al., 1993). However, placement of glycine in place of either the 1 Met or the 3

MOLECULAR RECOGNITION BY SH2 DOMAINS

181

Met of a high-af®nity pYMDM peptide decreased N-p85 SH2 domain binding by about 1000-fold (Gu È nther et al., 1996). Studies have also probed the structural changes that occur in the N-p85 SH2 domain upon peptide binding. For instance, NMR experiments have shown synergistic interplay between the two important binding regions of the SH2 domain (the pTyr binding pocket and the speci®city determining region); this has been demonstrated by the fact that introducing mutations in the C-terminal segments of peptides bound to the N-p85 SH2 domain perturbs the chemical shift of residues in the pTyr binding cavity (Gu È nther et al., 1996). Initial spectroscopic studies of N-p85 SH2 domain binding also suggested that a signi®cant conformational change might occur upon binding to tyrosyl phosphopeptides (Panayotou et al., 1992). However, the structures of the bound and unbound forms of the N-p85 SH2 domain indicate that this is not the case (Nolte et al., 1996): it may be that the spectroscopic signal interpreted as a sign of a conformational change re¯ected only the conformational rearrangement of a tyrosine residue (Tyr BG5) upon ligand binding (Nolte et al., 1996). 5. The N-SHP-2 and C-PLCg SH2 Domains Structural interpretations of which residues are important for binding are not always straighforward. This has been demonstrated in studies of the C-PLCg and N-SHP-2 SH2 domains. The structure of the C-PLCg SH2 domain in complex with a tyrosyl phosphopeptide shows a peptide binding interface which is extended to the residue 5 positions Cterminal to the pTyr (Pascal et al., 1994). A similar extended binding interface is seen in the N-SHP-2 SH2 domain (Lee et al., 1994). It might have been expected that the interactions between residues throughout this extended interface would be critical for high-af®nity binding. However, binding experiments using a series of truncated peptides have indicated that only the residue immediately C-terminal to the pTyr is required to maintain the majority of binding energy of the C-PLCg SH2 domain binding interaction (Kay et al., 1998). Hence, the majority of interactions seen in the structure contribute little to binding. It was further observed that the side chains which are of less importance for binding show greater mobility in the bound state than the residues of the pTyr binding cavity which are essential for binding (Kay et al., 1996, 1998). For the most part, the speci®city of SH2 domain binding is determined by the residues of the target C-terminal to the pTyr. However, for some SH2 domains, the residues N-terminal to the pTyr can be important. For example, it has been determined that the N-SHP-2 SH2 domain has a requirement for a hydrophobic residue (Val, Ile, or Leu) at the position 2 to the pTyr (Burshtyn et al., 1997; Huyer et al., 1995). This unusual

182

J. MICHAEL BRADSHAW AND GABRIEL WAKSMAN

requirement has been traced to the presence of a Gly at the aA2 position of the SH2 domain, a position that an Arg occupies in 80% of known SH2 domains (Huyer and Ramachandran, 1998). Apparently, the absence of Arg aA2, a residue which forms one wall of the pTyr binding cavity, necessitates that a hydrophobic residue be present at the 2 residue of the ligand to ensure proper formation of the pTyr binding cavity. 6. The Grb2 SH2 Domain The Grb2 SH2 domain is relatively unique in having a strong requirement for a residue (Asn) at the position 2 C-terminal to the pTyr (Songyang et al., 1993). A thermodynamic and mutational analysis of Grb2 SH2 domain binding has underscored the importance of this 2 Asn for high-af®nity tyrosyl phosphopeptide binding (McNemar et al., 1997). Here it was demonstrated that substitution of the 2 Asn to Ala in a high-af®nity Grb2 tyrosyl phosphopeptide decreased the binding af®nity by 3 orders of magnitude. Alanine mutation of the other residues had less than an order of magnitude effect on binding. 7. Switching SH2 Domain Speci®city The fact that different SH2 domains prefer to bind different consensus sequences has suggested that SH2 domains might be easily engineered to bind particular targets. As a starting point for the design of SH2 domains with novel properties, several different attempts have been made to switch the speci®city of one SH2 domain to resemble another (Bradshaw et al., 2000; Marengere et al., 1994; Songyang et al., 1995). A few of these attempts have been successful. For instance, it has been found that mutating the EF1 residue of the Src SH2 domain (a Thr) to the corresponding residue in the SH2 domain of Grb2 (a Trp) switches the speci®city of the Src SH2 domain to resemble that of Grb2 (Marengere et al., 1994). This modi®ed Src SH2 domain is able to substitute for the Grb2 SH2 domain in activation of the Ras signaling pathway in vivo. It has also been reported that mutating a residue of the N-p85 SH2 domain (Ile bD5) to the corresponding residue of the Src SH2 domain (a Tyr) results in a mutant whose phosphopeptide selection pattern resembles that of the Src SH2 domain (Songyang et al., 1995). However, introducing an Ile into the bD5 position of the Src SH2 domain does not give the Src SH2 domain the ability to bind with high af®nity to the N-p85 SH2 domain consensus binding sequence (Bradshaw et al., 2000). Hence, although there appear to be a few cases where the speci®city of SH2 domains can successfully be altered by point mutations, in general it is likely that engineering the speci®city of most SH2 domains will be a complicated process.

MOLECULAR RECOGNITION BY SH2 DOMAINS

183

Such complexity was recently illustrated with an attempt at changing the selectivity of the Src SH2 domain for the 1 position residue of the peptide target (G. Waksman and O. Lubman, unpublished results). In the Src SH2 domain±pYEEI peptide complex structure, Lys bD3 lies within Ê of the charged tip of the 1 Glu residue of the peptide. When Lys 4.5 A bD3 was mutated to Ala, the resulting SH2 domain displayed a signi®cantly decreased af®nity for the EEI peptide but, surprisingly, showed an increase in af®nity for a peptide containing an Arg at the 1 position. This result could be interpreted in light of the fact that beyond the bD3 position lie two aspartate residues (Asp bC8 and Asp CD2 in Fig. 3C) which, in the presence of a Lys at position bD3, cannot in¯uence binding (Lys bD3 is located between the 1 position of the peptide and these aspartate residues). However, when Lys is changed to Ala, the negative electrostatic potential contributed by these two aspartate residues acts as a repulsive force when a Glu is presented at the 1 position of the peptide (hence the decreased af®nity for the EEI motif) and as a magnet when an Arg is presented at the same position (hence the increased af®nity for a REI motif). This example demonstrates that single substitutions of amino acids at a protein±protein interface can have complex effects such as involving residues in binding which, in the wild-type interaction, play no discernible role.

C. Implications The ®ndings described above indicate that interactions between the pTyr in the target and its binding pocket in the SH2 domain are those most crucial for the molecular recognition of SH2 domains; the interactions between other residues of the target and the speci®city determining residues of the SH2 domain appear to be quite secondary in importance in terms of SH2 domain binding af®nity. Furthermore, the level of speci®city exhibited by SH2 domains, particularly those of the Src family, is modest. These results have profound implications for the mechanism by which SH2 domains speci®cally engage their targets within cells. In particular, they suggest that initial models of SH2 domain recognition, where single SH2 domains are able to selectively recognize their particular binding sites due only to a distinct set of interactions between the SH2 domain and the target, may be overly simplistic. Rather, a single SH2 domain alone may not contain all the required elements for selective recognition of targets. How, if the interactions involving the speci®city determining residues are weak, can SH2 domains be targeted to particular pTyr-containing

184

J. MICHAEL BRADSHAW AND GABRIEL WAKSMAN

sites within the cell? For single SH2 domain-containing proteins, speci®c recruitment to larger signaling complexes may aid in speci®c target recognition. Most of these proteins have several other domains or motifs in addition to their SH2 domain which facilitate targeting of the protein to a macromolecular assembly of signaling proteins (Pawson and Scott, 1997). For example, N-myristoylation of the Src kinaseÐand hence its association with the cell membraneÐmay facilitate the interaction of the Src SH2 domain with the most membrane proximal pTyr of the bPDGFR (Mori et al., 1993). Another possibility stems from the fact that many SH2 domaincontaining proteins do not contain just a single SH2 domain, but rather SH2 domains located in tandem (Zap-70, Syk, PI-30 kinase, PLCg, etc.). For these proteins, the concerted binding of multiple domains likely enhances overall speci®city. This would be achieved in tandem SH2 domain binding for two reasons. First, the binding of two SH2 domains may be positively coupled, which would result in an increase in binding speci®city. Second, improper targeting to tyrosine phosphorylated sites may be reduced by the topological constraints imposed by the strict spacing between pTyrs. Hence, although SH2 domains appear to have evolved only low levels of speci®city, speci®c targeting of SH2 domaincontaining proteins to signaling complexes may be achieved through cooperative interactions between these domains and other partners in the signaling complex. Although this is an attractive notion, it needs to be examined experimentally. In the next section, we describe studies which examine how single SH2 domains interact with other protein modules, including other SH2 domains.

IV. STRUCTURE AND FUNCTION OF SH2 DOMAINS IN THE CONTEXT OF OTHER PROTEIN MODULES Initial investigations of SH2 domains focused on understanding how these domains functioned in isolation. As described in Sections II and III, SH2 domains were expressed, puri®ed, and studied in vitro alone in order to understand their structure and mechanism of binding. While this reductionist approach has allowed extensive characterization of SH2 domains, studies of isolated SH2 domains have not addressed how SH2 domains communicate with other protein domains in order to determine the biological function of SH2 domain-containing proteins. In this section, we describe both structural (Section IV,A) and solutionbased biophysical (Section IV,B) investigations which have probed the mechanisms by which SH2 domains function within the context of other domains or of full-length proteins.

MOLECULAR RECOGNITION BY SH2 DOMAINS

185

A. Structures 1. The Tandem SH2 Domains A signi®cant number of SH2 domain-containing proteins have SH2 domains located in tandem. One of the best studied families of tandem SH2 domain-containing proteins is the Syk family of nonreceptor tyrosine kinases. This family consists of two proteins often found in immune cells: Syk and Zap-70. On immune cell activation, these proteins are recruited via their tandem SH2 domains to speci®c sites of tyrosine phosphorylation on immune receptors (T-cell receptor, B-cell receptor, etc.) known as immunoreceptor tyrosine activation motifs (ITAMs). These are sequences containing 2 tyrosine residues spaced 10±11 residues apart, with either a Leu or an Ile 3 residue C-terminal to each Tyr [sequence: Yxx(L=I)x7=8 Yxx(L/I)]. Binding of Syk or Zap-70 to doubly phosphorylated ITAM sequences activates the kinase and hence promotes downstream signaling. While apparently similar in many ways, Syk and Zap-70 demonstrate different patterns of expression; Syk is widely expressed in a variety of immune (and some nonimmune) cells, while Zap-70 is restricted to signaling primarily in T cells and natural killer cells (Bubeck-Wardenburg et al., 1999). The crystal structure of the tandem SH2 domain of Zap-70 in complex with a dpITAM peptide derived from the T-cell receptor has revealed the molecular basis of Zap-70±dpITAM recognition (Hatada et al., 1995). In this structure, the two SH2 domains of Zap-70 are juxtaposed such that both peptide binding interfaces are oriented in the same direction; hence, the tandem SH2 domain forms a contiguous binding site for the dpITAM peptide (Fig. 4A). A 65-residue connector links the two SH2 domains, and this sequence forms an a-helical coiled coil which is located on the opposite side of the tandem SH2 domain from the peptide binding site. The 19-residue T-cell receptor dpITAM peptide binds across the front face of the tandem SH2 domain with the N-terminal pTyr contacting the C-terminal SH2 domain and the C-terminal pTyr binding the N-terminal SH2 domain (a con®guration termed ``antiparallel''; Fig. 4A). The contacts between each individual SH2 domain and the corresponding segment of dpITAM peptide are reminiscent of the Src SH2 domain±pYEEI peptide interaction in that the pTyr and the 3 position of the peptide are inserted in positively charged and hydrophobic binding cavities, respectively. However, this structure lacks the solvent-exposed salt bridges emanating from the 1 and 2 peptide positions which are found in the Src SH2 domain structure (Waksman et al., 1993). Another feature of the Zap-70±T-cell receptor peptide interaction which is particularly intriguing is that the pTyr binding pocket for the

186

J. MICHAEL BRADSHAW AND GABRIEL WAKSMAN

A

dpIAMPhosphopeptide C-pTyr N-pTyr

Zap-70C-SH2 Zap-70N-SH2 Linker

B Grb2C-SH3

Grb2SH2

Grb2N-SH3

FIG. 4. Structures of SH2 domains in the context of other protein modules. These structures demonstrate the multiple ways SH2 domains regulate protein function. (A) The structure of the tandem SH2 domains of Zap-70 in complex with a T-cell receptor dpITAM tyrosyl phosphopeptide. The N-and C-terminal SH2 domains are shown in red and rose, respectively, while the coiled-coil linker between the SH2 domains is shown in gray. The T-cell receptor dpITAM peptide (in ribbon representation) and the two pTyrs (in ball-and-stick representation) are shown in orange. (B) The structure of Grb2. The SH2 domain is shown in red, with the N-terminal and C-terminal SH3 domains shown in gray. Very few contacts link the SH2 domain with the SH3 domains, indicating that the domains function largely independently. (C) The structure of the full-length Src kinase. The SH2 domain is shown in red, with the SH3 domain, kinase domain, and linker region shown in gray. The C-terminal tail of Src (in ribbon representation), which is a ligand for the SH2 domain, and the pTyr (in ball-and-stick representation) are shown in orange. Binding of the C-terminal tail of Src to the SH2 domain helps transmit an allosteric signal which down-regulates the kinase. (D) The structure of the STAT dimer bound to DNA. The SH2 domains of the two STAT molecules (molecule 1 and molecule 2) are shown in red and rose, respectively, while the other domains of the molecule (the coiled-coil domain, the DNA binding domain, and the linker domain) are shown in gray. The C-terminal tails of the two molecules are shown in orange and gold, respectively, with the pTyr of each

MOLECULAR RECOGNITION BY SH2 DOMAINS

C

187

C-terminaltail pTyr

Kinase

SrcSH2

Domain

SrcSH3

D

pTyr#1 C-tail#1 C-tail#2 pTyr#2 StatSH2#2

StatSH2#1

Linker Domain#2

Linker Domain#1

Coiled-coil Domain#2

Coiled-coil Domain#1 DNA DNABinding DNABinding Domain#2 Domain#1

represented in ball-and-stick representation. The double-stranded DNA is shown in ball-and-stick representation and is color-coded in gray. The structure illustrates how SH2 domain-mediated dimerization of STAT molecules properly orients the protein for DNA binding. This ®gure was generated using the program Ribbons (Carson, 1997).

C-terminal pTyr is formed by residues not just from the N-terminal SH2 domain, but also from the C-terminal SH2 domain. Hence, the Nterminal SH2 domain is, by itself, incomplete and can be fully functional only in the context of the C-terminal SH2 domain. This imposes a strict spacing requirement on ITAM peptides for binding; it is likely that only peptides with the pTyr residues located about 10 to 11 residues apart can ef®ciently interact with the Zap-70 tandem SH2 domain. This strict spacing requirement likely provides another level of speci®city in Zap70 tandem SH2 domain±target interactions in addition to the inherent speci®city conferred by the sequence context of the peptide. The crystal structure of the tandem SH2 domain of the Syk kinase has also been solved in complex with a dpITAM peptide (Fu È tterer et al., 1998). This structure is similar to that of Zap-70: (1) the peptide binding sites of each SH2 domain are spatially colinear so that a single dpITAM peptide bridges the two SH2 domains; (2) an a-helical coiled coil links the two SH2 domains; (3) the peptide and tandem SH2 domains bind

188

J. MICHAEL BRADSHAW AND GABRIEL WAKSMAN

antiparallel to one another; and (4) the nonpolar residues in the dpITAM 3 residues C-terminal to the pTyrs (Leu/Ile) occupy hydrophobic binding cavities in the SH2 domains. The crystals of the Syk tandem SH2 domain contained six copies of the liganded protein within the asymmetric unit. These six ``snapshots'' all differ from one another with respect to the relative orientation of the Nand C-terminal SH2 domains, with the two extreme orientations Ê translation (Fu differing by 188 in rotation and a 2.0-A È tterer et al., 1998). These results indicate that the Syk tandem SH2 domains can demonstrate considerable conformational ¯exibility, even when bound to dually phosphorylated peptides. A signi®cant difference between the Syk and Zap-70 tandem SH2 domain structures is the architecture of the C-terminal pTyr binding site. Unlike in Zap-70, this pTyr binding site in Syk is nearly completely self-contained within the N-terminal SH2 domain. Only a single residue (Lys bF1) from the C-terminal SH2 domain interacts with the C-terminal pTyr; however, this interaction occurs in only three of the six snapshots within the asymmetric unit and even in these complexes the electron density for this residue is poorly de®ned, suggesting that even this single interaction does not contribute signi®cantly to binding. The conformational ¯exibility exhibited by the Syk tandem SH2 domain together with the structural independence of its C-terminal pTyr binding site suggests that Syk has the ¯exibility to accommodate dually phosphorylated targets of various lengths, which may explain the relative ubiquity of Syk in signaling pathways. The structures of the related tandem SH2 domains of Zap-70 and Syk differ substantially from that of another protein, the SHP-2 phosphatase (Eck et al., 1996). In this structure, the binding interfaces contributed by each SH2 domain do not align. Moreover, the peptide binding sites are Ê ). As a result, the tandem SH2 domain separated by a large distance (40 A contains two distinct binding sites for tyrosyl phosphorylated peptides. Indeed, each SH2 domain is complexed with a singly phosphorylated peptide. This structure also suggested a ®xed relative orientation of the two SH2 domains stabilized by a disul®de bond; however, the subsequent structure of the full-length SHP-2 phosphatase (described below) shows an entirely different orientation between the SH2 domains (Hof et al., 1998), suggesting that the orientation seen in the tandem SH2 domain structure might not be functionally relevant. 2. The SH2±SH3 Domain Constructs In addition to being located in tandem, SH2 domains are also often found in proteins in conjunction with other protein modules known as Src homology 3 (SH3) domains (Pawson, 1995; Pawson and Schlessinger,

MOLECULAR RECOGNITION BY SH2 DOMAINS

189

1993; Pawson and Scott, 1997). These modules are 60 amino acids in length and form a ®ve-stranded b-sandwhich structure Musacchio et al., 1992; Yu et al., 1992; see also the chapter by Musacchio, this volume); their function is to bind proline-rich sequences in cellular proteins and hence facilitate the assembly of signaling complexes (Koch et al., 1991). A key question for the many proteins which contain both SH2 and SH3 domains is whether these domains cooperate in binding. Information regarding this question has been obtained from the structures of SH2± SH3 domain constructs, with the relevant data being the extent of molecular contacts and orientation between the SH2 and the SH3 domains. Grb2 is an SH2 and SH3 domain-containing adapter protein which has long been known to couple activation of receptor tyrosine kinases to signaling in the Ras pathway; it binds activated receptors through its SH2 domain and the Ras guanine nucleotide exchange factor Son of Sevenless (Sos) constitutively through its SH3 domains. The crystal structure of Grb2 revealed the molecular arrangement of its one SH2 domain and two SH3 domains (Maignan et al., 1995). All three protein modules display their canonical fold (Fig. 4B). However, very few intramolecular contacts link the domains to one another. Furthermore, all SH2 and SH3 domain binding sites are accessible in the unliganded state of the protein. Hence, Grb2 appears to be a ¯exible adapter with independently functioning SH2 and SH3 domains. These observations suggest that binding of the Grb2 SH2 domain to a tyrosine phosphorylated receptor does not regulate Ras through a conformational change in Grb2; rather it is more likely that SH2 domain binding serves simply to localize the Grb2±Sos complex to its proper cellular location. Structures have also been determined for the SH2±SH3 domain segments of the nonreceptor tyrosine kinases Lck and Abl (Eck et al., 1994; Nam et al., 1996). Like Grb2, the Lck SH2 and SH3 domains were found to have few intramolecular domain contacts (Eck et al., 1994), indicating ¯exibility with respect to their orientation. Unlike both the Grb2 and Lck SH2±SH3 domain segments, the Abl SH2±SH3 domain construct Ê 2 ) interface between the SH2 and SH3 domains showed a modest (700 A (Nam et al., 1996). However, NMR experiments have indicated that the domain orientation of the Abl SH2±SH3 domain construct is not ®xed (see below). Taken together, the structural data concerning SH2±SH3 domain constructs suggest a primarily independent function for the SH2 and SH3 domains and indicate that SH2 and SH3 domains do not have a single preferred relative orientation. 3. The Full-Length Src Family Kinases While the structures of SH2±SH3 domain constructs described above provide some hint at ways in which SH2 domains might interact with

190

J. MICHAEL BRADSHAW AND GABRIEL WAKSMAN

other domains, a better understanding of SH2 domain interactions requires examination of entire SH2 domain-containing proteins. In that respect, the structures of several Src family kinases have greatly illuminated the mechanism by which the SH2 domain of these proteins communicates with the kinase domain in order to regulate enzyme activity. These structures were those for human Src (Xu et al., 1997), chicken Src (Williams et al., 1997), and the Src family member Hck (hematopoietic cell kinase) (Sicheri et al., 1997). The Src family kinases are a closely related family of nonreceptor tyrosine kinases which are involved in signaling pathways that control the growth and differentiation of cells. All family members share the same domain structure: they contain (from N-terminal to C-terminal) a membrane localization motif, a unique region, an SH3 domain, an SH2 domain, a kinase domain, and a short C-terminal tail (Brown and Cooper, 1996). Prior to the crystallographic analysis, many experiments probed the mechanism of action of the Src kinase, and the accumulated data led to the following model of activation (see Cooper and Howell, 1993). In the inactive state, Src is phosphorylated on Tyr-527 (Src numbering) in the C-terminal tail region. This site binds the SH2 domain through an intramolecular interaction, which places the kinase into a repressed, or ``closed,'' conformation. Dephosphorylation of pTyr-527, or displacement of the C-terminal tail by competitive binding of the SH2 domain to another target, ``opens'' the repressed conformation, allowing kinase activation. A higher level of activation can be achieved by phosphorylation of Tyr-416 within the catalytic domain. The structures of Src family kinases, each in their closed conformations, have provided a much better understanding of how these proteins operate (Sicheri and Kuriyan, 1997). As expected from previous biochemical results, the SH2 domain is bound intramolecularly to pTyr-527 (Fig. 4C). Unexpectedly, the SH3 domain binds to an internal polyproline type II helix which links the SH2 domain and the catalytic domain (Fig. 4C). However, neither the SH2 domain nor the SH3 domain is located near the kinase domain active site; both are located on the distal face of the catalytic domain, leaving the kinase relatively unimpeded even though it is in an inactive state. In fact, the inactive kinase domains of Src and Hck demonstrate only one signi®cant structural difference from the previously determined ``active state'' of the kinase domains of Lck and protein kinase A (Johnson et al., 1996; Yamaguchi and Hendrickson, 1996): an alpha helix (aC) near the active site is displaced in orientation, disrupting a salt bridge between the conserved Glu-310 and Lys-295 residues presumed to be essential for ATP binding and hydrolysis. These structures suggest the following model of Src kinase inhibition. Engagement of the phosphorylated C-terminal tail to the SH2 domain, in

MOLECULAR RECOGNITION BY SH2 DOMAINS

191

conjunction with the SH3 domain's binding of the internal polyproline II helix, results in the displacement of the aC helix and a remodeling of the ATP binding site, causing inhibition of the kinase activity (Gon¯oni et al., 1999; Moare® et al., 1997). Hence, these structures demonstrate a mechanism whereby SH2 domain±target recognition induces a long-ranged allosteric effect to alter protein function. 4. The SHP-2 Phosphatase and the STAT Transcriptional Activators The structures of Src family kinases indicate that binding of Src's phosphorylated tail to its SH2 domain transmits a conformational change which perturbs enzyme activity. Is this mechanism common to other families of SH2 domain-containing proteins or do different SH2 domain-containing proteins use other means of communication? Recent structures of other full-length SH2 domain-containing proteins, the tandem SH2 domain-containing SHP-2 phosphatase (Hof et al., 1998) and two different signal transducers and activators of transcription (STATs) proteins bound to DNA (Becker et al., 1998; Chen et al., 1998), suggest that there may be multiple ways to couple SH2 domain ligation to a particular molecular function. Higher eukaryotes contain two SH2 domain-containing phosphatases, SHP-1 and SHP-2, which are highly related in structure (60% sequence identity) (Bartford and Neel, 1998). These proteins contain two SH2 domains (the N-terminal and C-terminal SH2 domains, respectively) followed by a phosphatase domain. SHP-1 is largely restricted in expression to hematopoetic cells, where it serves to negatively regulate several different signaling pathways. In contrast, SHP-2 is widely expressed. Studies have demonstrated that SHP-2 is essentially inactive under basal conditions, but becomes active either in a construct which lacks its SH2 domains or on SH2 domain ligation, suggesting that the SH2 domains play a negative regulatory role in the phosphatase activity (Bartford and Neel, 1998). The crystal structure of the full-length SHP-2 phosphatase revealed the molecular mechanism of this inhibition (Hof et al., 1998). In this structure, the phosphatase is directly inhibited due to the insertion of a section of the N-terminal SH2 domain (the D'E loop and the bD and bE strands) within the phosphatase active site. This intramolecular binding slightly distorts the fold of the SH2 domain, resulting in an SH2 domain conformation which appears to be less favorable for binding tyrosyl phosphopeptides. This was the ®rst structural identi®cation of a signi®cant conformational change occurring within an SH2 domain. However, it is consistent with previous studies which have identi®ed the N-terminal SH2 domain of SHP phosphatases as being crucial for phosphatase inhibition (Pei et al., 1996). While the N-terminal SH2 domain appears

192

J. MICHAEL BRADSHAW AND GABRIEL WAKSMAN

important for regulation, the C-terminal SH2 domain makes no direct contact with either the phosphatase domain or the other SH2 domain, suggesting that it plays a lesser role in phosphatase activation. The crystal structures of the proteins STAT-1 (Chen et al., 1998) and STAT-3 (Becker et al., 1998) bound to DNA illustrate, at atomic resolution, another function of SH2 domain±target recognition: homodimerization. The STATs are a family of SH2 domain-containing molecules which directly respond to receptor activation by transducing to the nucleus and activating transcription (Darnell, 1997). Hence, they contain both an SH2 domain and a DNA binding region. STAT activation is induced by tyrosine phosphorylation in a tail region just C-terminal to the SH2 domain, which in turn induces dimerization of STAT molecules. The crystal structure of the two STATs reveals the molecular basis of this SH2 domain-mediated dimerization (Fig. 4D). In the structures, each STAT SH2 domain interacts with the phosphorylated C-terminal tail of the other partner in the dimer (Fig. 4D) and dimerization of the SH2 domains exactly positions the DNA binding segments of the STATs for sequence-speci®c DNA recognition. Do the currently available structures of full-length SH2 domaincontaining proteins (Src family kinases, SHP-2, STATs) reveal common mechanisms of SH2 domain-mediated regulation? When compared to the mechanism of SH2 domain-mediated inhibition of the Src kinases, the mechanism of inhibition of the SHP-2 phosphatase is completely different. In one case, the Src kinases use intramolecular SH2 domain ligation to transmit a subtle allosteric effect to the active site; in the other case, the SHP-2 phosphatase is directly inhibited by steric hindrance of the active site by the SH2 domain. The mechanism of STAT dimerization is different again from each of these cases in that it is a change in quaternary, not tertiary, structure which determines the biological effect. Hence, nature has apparently devised a plethora of different mechanisms by which the signal of SH2 domain ligation can be transmitted. B. Solution Studies of Multiple Protein Modules The structures of SH2 domains in the context of other protein modules provided much information on how these modules function within the context of a larger protein. However, the availibility of structures is just the ®rst step of many which are needed to understand how macromolecules function. In this section, we describe solution studies of SH2 domains placed in the context of either larger protein fragments or full-length proteins that have provided information about the various mechanisms by which SH2 domains regulate protein function.

MOLECULAR RECOGNITION BY SH2 DOMAINS

193

1. Syk and Zap-70 Tandem SH2 Domains a. Conformational Flexibility. The structure of the tandem SH2 domain of Syk in complex with a dpITAM phosphopeptide suggested that signi®cant changes in the conformation and/or orientation of the tandem SH2 domain might accompany dpITAM binding (Fu È tterer et al., 1998). This hypothesis has been supported by investigations which have characterized the energetics of Syk tandem SH2 domain±dpITAM binding (Grucza et al., 1999). Evidence for signi®cant ¯exibility of the Syk tandem SH2 domain in solution came from calorimetric studies of the interaction between the Syk tandem SH2 domain and a dpITAM peptide based on a sequence of the CD3-e chain of the T-cell receptor (CD3±e dpITAM). In this study, it was shown that the enthalpy of binding demonstrates an unusual nonlinear dependence on temperature (Grucza et al., 1999). This result was interpreted using a model where dpITAM binding is coupled to a temperature-dependent conformational equilibrium in the protein. In this model, higher temperatures induce a transition of the protein to a higher entropy form, whereas binding of the dpITAM peptide causes the peptide to revert to a lower entropy form. Kinetic and spectroscopic evaluation of the Syk tandem SH2 domain in both the free and bound forms further supported the notion that the free form of Syk populates different conformations at different temperatures (Grucza et al., 1999). The conformational ¯exibility shown by the Syk tandem SH2 domain could have been due to either large changes in tertiary structure or local changes in domain orientation. However, investigation of both the nearand far-UV CD spectra of the Syk tandem SH2 domain as a function of temperature indicated that there was little to no temperature dependence of these parameters. Along with the modest ¯exibility in domain orientation seen in the Syk tandem SH2 domain structure (Fu È tterer et al., 1998), these results suggest that the Syk tandem SH2 domain conformational transition involves changes in domain orientation, rather than major secondary or tertiary structural rearrangements. What are the functional consequences of conformational ¯exibility for the Syk tandem SH2 domain? It is possible that this ¯exibility may be required to facilitate binding of Syk to a variety of different receptors with different spaced dpITAMs. For instance, in at least two targets for Syk [the FcgRIIA receptor (a class II Fc receptor) and the colony-stimulating factor receptor] the spacing between pTyrs is longer than usual (14±15 residues instead of 10±11) (Chacko et al., 1996; Corey et al., 1994), suggesting that the ability of the Syk tandem SH2 domain to adopt different conformations is required for the biological action of Syk. In vitro binding experiments have indeed shown that the Syk tandem SH2

194

J. MICHAEL BRADSHAW AND GABRIEL WAKSMAN

domain can bind with high af®nity to a dpITAM peptide based on the FcgRIIA receptor which contains 5 more residues between the pTyrs than the CD3-e dpITAM (Grucza et al., 2000). Does the Zap-70 tandem SH2 domain also exhibit signi®cant conformational ¯exibility? The orientation of the SH2 domains, as well as the inter-SH2 linker, in the Zap-70 structure suggests that a change in conformation of this module might accompany binding (Hatada et al., 1995). Indeed, a change in the Zap-70 tandem SH2 domain conformation upon dpITAM binding has been characterized using sedimentation velocity analytical ultracentrifugation (Labadia et al., 1997). Here a reduction in the frictional coef®cient of the Zap-70 tandem SH2 domain is seen on dpITAM binding, suggesting that the complex adopts a more compact structure than the unligated protein. Furthermore, a study which has used antibodies directed against the inter-SH2 domain region of Zap-70 has also detected conformational changes in the tandem SH2 domain on binding to the z-chain of the T-cell receptor (Grazioli et al., 1998). Here, the inter-SH2 domain region was far more accessible in the tandem SH2 domain than the full-length Zap-70 molecule; furthermore, the accessibility of the inter-SH2 domain region was signi®cantly reduced on dpITAM ligation. Finally, a thermodynamic study which has compared and contrasted the recognition properties of the tandem SH2 domains of Zap-70 and the p85 subunit of PI-30 /kinase has also suggested that a conformational change occurs on binding of the Zap-70 tandem SH2 domain (O'Brien et al., 2000). b. Tandem SH2 Domain Speci®city. An essential question about SH2 domain recognition is the speci®city of interaction with targets. Single SH2 domains have a surprisingly low level of speci®city, binding nonspeci®c tyrosine phosphorylated targets only 50- to 200-fold more weakly than speci®c targets (Bradshaw et al., 1998). It is at present not clear whether tandem SH2 domains demonstrate higher levels of sequence speci®city. Grucza et al. (2000) have shown that Syk tandem SH2 domain binding to the noncognate CD3-e dpITAM peptide is only 20-fold weaker than binding to the biologically relevant (and similarly spaced) dpITAM derived from the g chain of the FceRI receptor (FceRI dpITAM peptide), a class I Fc receptor to IgE. Also, mutations at the 1, 2, or 3 positions C-terminal to both the N- and C-terminal pTyrs do not dramatically affect binding to the Syk tandem SH2 domain (G. Waksman and R. A. Grucza, unpublished results). Similarly, modest differences in af®nity were observed when binding of the Syk and Zap-70 tandem SH2 domains to various dpITAM-based peptides was tested (Ottinger et al., 1998). However, when doubly phosphorylated and similarly spaced peptides derived from the PDGFR were tested,

MOLECULAR RECOGNITION BY SH2 DOMAINS

195

binding was found to be 3±4 orders of magnitude weaker (Ottinger et al., 1998). Therefore, it would appear from this example that SH2 domains in tandem may exhibit higher levels of sequence speci®city than single SH2 domains. The inter-pTyr spacing seems to contribute only modestly to peptide binding speci®city. For example, Syk binding of the FceRI dpITAM peptide which contains 10 residues between the pTyrs is as tight as that of a dpITAM peptide based on the FcgRIIA receptor, a class II Fc receptor to IgG, which contains an additional 5 residues in the interpTyr region (Grucza et al., 2000). Consistent with these results, Ottinger et al. (1998) showed that a peptide based on the sequence of SHPS (a substrate for the SHP-2 phosphatase) which contains 13 additional residues in the inter-pTyr region binds with only a 10- to 20-fold weaker af®nity to both Syk and Zap-70 tandem SH2 domains than a peptide containing only 10 residues in the inter-pTyr region (Ottinger et al., 1998). Depending on the degree of receptor activation, ITAM sequences within the cell can be unphosphorylated, singly phosphorylated (mpITAM), or doubly phosphorylated. Hence, an additional issue regarding the speci®city of tandem SH2 domain binding is the preferential binding of dually phosphorylated, as opposed to singly phosphorylated or unphosphorylated, ITAM sequences. It has been experimentally observed that the Syk tandem SH2 domain displays dramatic differences in af®nity between mpITAMs and dpITAMs (Grucza et al., 2000). For instance, the Syk tandem SH2 domain binds to the FceRI ITAM with an af®nity of about 2 nM, whereas af®nities are about 1 and 10mM, respectively for the corresponding ITAM with the N-terminal or C-terminal phosphate removed (Grucza et al., 2000); hence loss of one phosphate of the ITAM causes an approximately 1000-fold loss in af®nity. This difference is comparable to that observed for the binding of the Src SH2 domain to a phosphorylated versus dephosphorylated peptide (Bradshaw et al., 1999). Hence, tandem SH2 domains are highly speci®c for doubly phosphorylated sequences. 2. SH2±SH3 Domain Constructs SH2±SH3 domain-containing adapter proteins such as Grb2 or individual SH2±SH3 domain constructs such as those of Abl, Lck, Src, and Itk have been used as models to study the communication between SH2 domains and other protein modules. In most cases, it has been shown that only a small degree of interaction occurs between SH2 and SH3 domains. The structure of Grb2 suggested that the SH2 and SH3 domains bind to their targets independent of one another (Maignan et al., 1995). Solution

196

J. MICHAEL BRADSHAW AND GABRIEL WAKSMAN

studies have tested this hypothesis by investigating how ligation of either the SH2 domain or the SH3 domain affects the binding af®nity of the other domain. A set of titration calorimetric experiments indicated that ligation of the SH3 domains by a polyproline-containing peptide caused no experimentally discernible effect on binding of a tyrosyl phosphopeptide to the SH2 domain (Lemmon et al., 1994), supporting the hypothesis that the domains function independently. Similar results have been obtained using ¯uorescence spectroscopy (Cussac et al., 1994). However, experiments which used the entire Sos protein, rather than polyproline-containing peptides, to ligate the Grb2 SH3 domains did demonstrate that ligation of the SH3 domains enhanced the af®nity of the SH2 domain for a tyrosyl phosphopeptide (Chook et al., 1996). However, the enhancement in af®nity is relatively small (three- to eight-fold). Hence, binding of targets to either the SH2 or the SH3 domains of Grb2 appears to have little effect on the binding properties of the other domains. The SH2 and SH3 domain-containing segments of the Abl tyrosine kinase have also been the target of several investigations. Similar to Grb2, it has been determined that the ligation of either the SH2 or the SH3 domain has no effect on the other domain's binding af®nity; hence the two sites of ligand binding are independent (Gosser et al., 1995). It has also been determined that the relative orientation between the SH2 and SH3 domains is not ®xed. This has been established using bivalent ligands for the Abl SH2±SH3 domain construct (Cowburn et al., 1995; Xu et al., 1999) and also by using 15 N NMR relaxation experiments (Fushman et al., 1999). Hence, the Abl SH2±SH3 domain construct appears to have the ¯exibility needed to adopt different conformations, which is likely required for proper Abl function. SH2±SH3 domain constructs other than Abl have been studied using NMR. Heteronuclear NMR studies using the Src SH2±SH3 domain construct have demonstrated that only very modest chemical shift changes occur in the SH3 domain on peptide ligation of the SH2 domain, indicating only modest structural changes in the SH3 domain on SH2 domain ligation (Tessari et al., 1997). However, it has been observed in the SH2 and SH3 domains of the Itk kinase that a section of the Itk SH2 domain interacts with the SH3 domain binding site through a motif which does not contain proline residues; this interaction has been proposed to be important in regulating Itk (A. H. Andreotti, unpublished results). In addition to NMR, hydrogen exchange and mass spectrometry have been used to investigate the functional regulation between SH2 and SH3 domains. In the case of Hck, it was found that the SH3 domain of the SH2±SH3 domain construct undergoes more rapid unfolding than the SH3 domain expressed alone, suggesting that the Hck SH2 domain in some manner destabilizes the SH3 domain (Engen et al., 1999).

MOLECULAR RECOGNITION BY SH2 DOMAINS

197

3. The Role of the SH2 Domain in the Context of the Full-Length Src Kinase Src is kept in an inactive, or closed, state through an intramolecular interaction between its SH2 domain and pTyr-527 within its C-terminal tail. Displacement of this interaction via binding of the SH2 domain to other targets is known to activate the kinase. Several studies have examined how effectively exogenous tyrosyl phosphopeptides can compete with the intramolecular interaction between the tyrosine phosphorylated C-terminal tail and the SH2 domain. It has been found that incubation of the Src family member Hck with a pYEEI phosphopeptide causes a greater than twofold activation of the kinase activity, con®rming that displacement of the tail by a higher af®nity ligand somewhat activates Src family kinases (Moare® et al., 1997). Signi®cant activation of kinase activity has also been noted when a high-af®nity SH2 domain binding peptide is incubated with the Src kinase (Weijland et al., 1997). The ability of the pYEEI peptide to activate Hck is abrogated when the Gln-Pro-Gly sequence immediately C-terminal to pTyr-527 in the tail is mutated to Glu-Glu-Ile, con®rming that the sequence Gln-Pro-Gly Cterminal to Tyr-527 has evolved to have low af®nity in order to allow kinase activation (Porter et al., 2000). These results demonstrate the important role played by the SH2 domain in regulating the activation of the Src kinase. Unregulated activation of Src often results in oncogenic transformation of cells (Jove and Hanafusa, 1987). Since the intramolecular interaction between the Src SH2 domain and pTyr-527 suppresses kinase activity, the importance of the SH2 domain in Src function can be assessed by examining the effect of mutations in the SH2 domain on the ability of Src to transform cells. Indeed, mutations in the SH2 domain of Src, including a mutation at Arg bB5, result in oncogenic transformation (Hirai and Varmus, 1990a,b), presumably by abolishing the ability of Src to form a closed conformation. These results parallel the ®nding that mutation of Tyr-527 in Src to Phe also activates Src and allows Src to transform cells. Mutations in the Src SH2 domain have also been shown to repress Src's ability to be inhibited by the Csk kinase (which phosphorylates Tyr-527 in Src) (Superti-Furga et al., 1993), con®rming that the SH2 domain is required for inhibition of Src.

V. UNUSUAL SH2 DOMAIN INTERACTIONS SH2 domains share many common properties of sequence, structure, and target recognition. Each binds the pTyr residue in its targets in a relatively similar mode, with each protein recognizing its speci®c target

198

J. MICHAEL BRADSHAW AND GABRIEL WAKSMAN

based in large part on the interactions between the C-terminal sequences in the target and residues in the C-terminal half of the SH2 domain. Furthermore, phosphorylation of the target is essential for recognition. These binding characteristics (and others) at ®rst appeared to be general for all SH2 domains. However, as is often the case, as the level of knowledge of a subject grows, exceptions to these general rules of SH2 domain recognition have arisen. These unusual SH2 domains are described below. A. The SAP SH2 Domain Given that many proteins in tyrosine kinase signaling pathways are potentially oncogenic, it might have been expected that mutations within SH2 domains would be a frequent cause of human disease. However, no mutations in an SH2 domain were associated with disease processes until recently when it was discovered that mutations in the SH2 domain of SAP [or SLAM (signaling lymphocyte activation molecule)-associated protein, also known as SH2D1A or DSHP] cause X-linked lymphoproliferative disease (Coffey et al., 1998; Nichols et al., 1998; Sayos et al., 1998). This disorder is characterized by an extreme sensitivity to Epstein±Barr virus, with infection resulting in uncontrolled B-cell proliferation. The average age of onset in affected young boys is 2.5 years, with 75% mortality by age 10 and 100% mortality by age 40 (Coffey et al., 1998). SAP functions in immune cell signaling, with its expression restricted primarily to T cells (Sayos et al., 1998). It interacts with the protein SLAM, a 70-kDa glycoprotein expressed on many lymphocyte cell surfaces. SLAM is a selfligand type of receptor that operates in bidirectional signaling between B and T cells, perhaps functioning in the switching of the helper T-cell phenotype. In addition to being the ®rst SH2 domain directly associated with a human disorder, SAP has several other properties which distinguish it from other SH2 domain-containing proteins. One of these is that SAP itself contains only an SH2 domain and a small (26-amino-acid) Cterminal tail of unknown function. All other SH2 domain-containing proteins contain either a domain with enzymatic activity or other small protein modules which facilitate protein±protein interactions. The lack of any additional domains suggests that SAP's function might be to act as a binding inhibitor. Indeed, the binding of SAP to SLAM blocks recruitment of the negative signaling regulator SHP-2 phosphatase to SLAM, which results in the stimulation of downstream signaling events (Sayos et al., 1998). The most unique feature of the SAP SH2 domain is that it can bind targets in a phosphate-independent manner. The ®rst hint at this non-

MOLECULAR RECOGNITION BY SH2 DOMAINS

199

canonical behavior was that SAP±SLAM binding was detected in a yeast two-hybrid screen, an interaction which would not have occurred if tyrosine phosphorylation was required (Sayos et al., 1998). A peptide surrounding Tyr-281 of SLAM has been found to immunoprecipitate SAP independent of whether Tyr-281 is phosphorylated, unphosphorylated, or mutated to phenylalanine. Fluorescence polarization binding studies have con®rmed that tyrosine phosphorylation of the SLAM peptide is not absolutely required for SAP SH2 domain binding, although it does somewhat increase binding af®nity (Li et al., 1999). This investigation has also identi®ed that SAP binding requires residues both Nterminal and C-terminal to Tyr-281 in its SLAM target (Li et al., 1999). The requirement for a speci®c sequence N-terminal to the pTyr is unusual for SH2 domains: such a property is shared only with the Nterminal SH2 domain of SHP-2 (see Section III). These studies have suggested that, in SAP, concerted binding of speci®c sequences both Nand C-terminal to Tyr-281 can largely overcome the requirement of tyrosine phosphorylation. The structures of SAP in its unliganded form, as well as bound to both tyrosine phosphorylated and unphosphorylated peptides, have been recently solved by X-ray crystallography (Figs. 5A and 5B) (Poy et al., 1999). The SAP SH2 domain exhibits the same general architecture as all SH2 domains: a central b-sheet ¯anked by two a-helices. Furthermore, as is typically seen, the SLAM peptide binds across the front face of the SH2 domain perpendicular to the central b-sheet. However, atypically, the residues in the peptide N-terminal to Tyr-281 make speci®c contacts with SAP: the hydrophobic 1 Ile and 3 Leu make contact with residues in the bD strand, while the 2 Thr forms hydrogen bonds with a glutamate on the protein. These interactions explain the requirement of the SAP SH2 domain for residues N-terminal to the pTyr (Li et al., 1999). The interaction between the pTyr in the peptide and Arg bB5 in the protein is reminiscent of that seen in many SH2 domain structures (Kuriyan and Cowburn, 1997). In the dephosphorylated peptide structure, water molecules take the place of the absent phosphate oxygens; these waters hydrogen bond to both Arg bB5 and the 2 Thr in the peptide. The presence of these waters suggests that the SAP SH2 domain does not form a pTyr binding cavity which is as solvent inaccessible as those found in other SH2 domains. SAP is truly a rogue SH2 domain in many ways, particularly because it binds so effectively to non-tyrosine-phosphorylated targets. The biological signi®cance of such a tyrosine-independent mode of SH2 domain binding is unknown, but it may serve to block either the phosphorylation or the binding of other proteins to speci®c cellular sites.

200

J. MICHAEL BRADSHAW AND GABRIEL WAKSMAN

A

AB

TyrβD1

LeuαA12

-3Leu βD

TyrβD3

GluαA6 αA

-2Thr ArgαA2 ArgβA5

-1lle βB pTyr

GluBC1 SerβB7

βC

ArgβD6

ThrBC2

B

AB

TyrβD1

BC

LeuαA12

-3Leu βD

TyrβD3

GluαA6 αA

-2Thr

ArgαA2 ArgβA5

-1lle βB Tyr

GluBC1

βC

ArgβD6

SerβB7 ThrBC2

C

CblSH2

BC

Phosphopeptide

pTyr

4helixbundle EFhand

FIG. 5. The structures of unusual SH2 domain-containing proteins. (A) Stereo diagram of the N-terminal region of SAP in complex with a tyrosine phosphorylated peptide from SLAM. Color-coding of secondary structures and residues is as in Fig. 2. Note that the SAP SH2 domain makes extensive contacts with the residues of the peptide N-terminal region to the pTyr, an unusual feature for SH2 domains. (B) Stereo diagram

MOLECULAR RECOGNITION BY SH2 DOMAINS

201

B. The Cbl SH2 Domain SH2 domains are normally identi®ed easily in protein sequences due to the conservation of a number of sequence motifs among SH2 domains. However, the ability to de®nitively recognize SH2 domains in protein sequences has been called into question by the recent structures of the Nterminal region of the protein Cbl, both in the apo form and in complex with a tyrosine phosphorylated peptide (Meng et al., 1999). Cbl is an adapter molecule which is involved in the negative regulation of a number of signaling pathways (Liu and Altman, 1998). In particular, it has been shown to interact with a tyrosine phosphorylated sequence in Zap-70. Although tyrosine phosphorylation was known to be essential for this interaction, the basis of recognition was unknown since Cbl showed no recognizable sequence homology to any known pTyr binding motifs. The structures of Cbl revealed that this protein does in fact contain an SH2 domain, albeit one which lacks strands bD0 1, bE, and bF and a key loop, the BG loop (Fig. 5C). Overall, the Cbl SH2 domain shares only 11% sequence homology with other SH2 domains. Despite low sequence identity, Cbl does retain the key Arg residue for phosphate recognition (Arg bB5), and peptides bind the SH2 domain in the canonical fashion, i.e., perpendicular to its central b-sheet. The Cbl structure also contains two other well-known protein domains, a four-helix bundle and an EF-hand, which appear also to aid in the recognition of targets (Meng et al., 1999). The identi®cation of an SH2 domain which is so divergent in sequence from other SH2 domains calls into question the evolutionary history of SH2 domains and suggests that SH2 domainmediated signaling may have evolved earlier than previously thought (Kawata et al., 1997).

C. SH2 Domain±Phospholipid Interactions? It has been suggested that SH2 domains may have another biological function apart from recognition of tyrosine phosphorylated protein sequences: binding to phospholipids such as phosphatidylinositols (Rameh of the N-terminal region of SAP in complex with an unphosphorylated, tyrosinecontaining peptide from SLAM. Secondary structure elements and residues are labeled as in A. (C) The structure of the N-terminal segment of Cbl in a complex with a singly phosphorylated tyrosyl phosphopeptide derived from the ZAP-70 kinase. Color-coding is as in Fig. 4 with the SH2 domain shown in red and the other domains (the EF-hand and the 4-helix bundle) shown in gray. The Zap-70 tyrosyl phosphopeptide is shown in orange, with the pTyr represented in ball-and-stick representation. Note that the strands bD0 1, bE, and bF are absent from the Cbl SH2 domain. This ®gure was generated using the program Ribbons (Carson, 1997).

202

J. MICHAEL BRADSHAW AND GABRIEL WAKSMAN

et al., 1995). Binding of phospholipids is a function which had been reserved for other protein modules, such as pleckstrin homology domains (Bottomley et al., 1998). However, the evidence for a biologically relevant interaction between SH2 domains and phosphatidylinositols is scarce. It has been observed that the SH2 domains of the p85 subunit of PI 30 -kinase bind to phosphatidylinositol-containing lipid vesicles (Rameh et al., 1995). However, a separate study concluded that these interactions do not occur with high af®nity or speci®city, and hence they are likely not of physiological relevance (Surdo et al., 1999).

VI. SH2 DOMAINS AS DRUG TARGETS As discussed in the Introduction (Section I), the central role of SH2 domains in many tyrosine kinase signaling pathways makes them at least potentially attractive targets of pharmaceuticals, and hence there have been signi®cant efforts aimed at discovering SH2 domain binding inhibitors (Brugge, 1993; Sawyer, 1998). The structural and biophysical investigations of SH2 domain recognition described here have provided essential insights into SH2 domain binding which may educate the efforts of targeting pharmaceutical compounds to speci®c SH2 domains. A. Dif®culties in Targeting Pharmaceuticals to SH2 Domains Many of the studies of SH2 domain binding described here have revealed signi®cant obstacles which complicate the discovery of SH2 domain-targeted pharmaceuticals. For instance, mutational studies have indicated that the SH2 domain binding mechanism is such that a charged interaction in the pTyr binding pocket, that between Arg bB5 and the phosphate of the pTyr, is most crucial for binding. Hence, charged ligands would be optimal to inhibit SH2 domain±target interactions. Unfortunately, highly charged ligands frequently have dif®culty penetrating the cell membrane (Charifson et al., 1997), and hence charged interactions are dif®cult to target with pharmaceuticals. Furthermore, the primacy of the Arg bB5±phosphate interaction in SH2 domain binding also suggests that targeting particular pharmaceuticals solely to one particular SH2 domain or class of SH2 domain will be dif®cult. This will especially be true if the Arg bB5±phosphate interaction is as crucial for binding in other SH2 domains as it is in the Src SH2 domain (which is likely, given the universal conservation of Arg bB5) (Bradshaw et al., 1999). Despite these dif®culties, signi®cant progress has been made in discovering pharmaceuticals which target particular SH2 domains. For instance, a dif®culty originally associated with tyrosine phosphorylated

MOLECULAR RECOGNITION BY SH2 DOMAINS

203

inhibitors was their susceptibility to cellular phosphatases; this problem has been overcome through the design of nonhydrolyzable phosphonopeptides which bind with af®nity comparable to that of tyrosyl phosphopeptides (Domchek et al., 1992). Furthermore, molecules have recently been found which appear to show promise for inhibiting the Src and Grb2 SH2 domains. B. Src SH2 Domain Binding Inhibitors As described earlier, the Src SH2 domain has a cysteine residue (Cys bC3) located within its pTyr binding cavity which is unique to Src; other SH2 domains have a serine, threonine, valine, or alanine at this position (Rose et al., 1999). Cysteine residues such as Cys bC3 are potentially good targets for pharmaceuticals due to their nucleophilic character: these residues can in certain cases be induced to form covalent bonds with aldehyde groups on pharmaceutical targets, as has been shown for cysteine proteases (Rich, 1986). In particular, Cys bC3 might be particularly reactive due to the fact that the positively charged environment of the pTyr binding cavity lowers its pKa (Bradshaw and Waksman, 1998). It has indeed been shown using X-ray crystallography and titration microcalorimetry that Cys bC3 can form a covalent bond with aldehyde groups on pharmaceutical-like compounds (Alligood et al., 1998; Charifson et al., 1997), indicating that it might be possible to selectively target the Src SH2 domain with speci®c compounds in vivo. This possibility has been further advanced by studies which have examined the effect of a Cys bC3 targeting compound in living cells. It has long been established that mice lacking the src gene demonstrate osteopetrosis (a condition of thickened bones) (Soriano et al., 1991), indicating that Src plays a role in bone formation and suggesting that Src might be a useful target for treatment of osteoporosis. A compound that targets Cys bC3 of the Src SH2 domain has been shown to preferentially bind to the Src SH2 domain in vitro (Violette et al., 2000). This compound inhibits Src SH2 domain±phosphopeptide interactions within cells and also diminishes the resorptive activity of rabbit osteoclasts (Violette et al., 2000). Hence, the strategy of targeting pharmaceuticals to Cys bC3 of the Src SH2 domain appears promising for the treatment of osteoporosis. C. Grb2 SH2 Domain Binding Inhibitors The SH2 domain of Grb2 is perhaps the most selective SH2 domain in terms of target recognition: there is a strict requirement for Asn 2 residues C-terminal to the pTyr (McNemar et al., 1997; Songyang et al.,

204

J. MICHAEL BRADSHAW AND GABRIEL WAKSMAN

1993). Furthermore, only in the Grb2 SH2 domain structure does the peptide bind in a b-turn conformation to accommodate interactions between the Asn and the protein. These ®ndings may partially explain why non-tyrosine-phosphorylated peptidomimetics have been successfully designed and reported for the Grb2 SH2 domain (Oligino et al., 1997), but not any other SH2 domain. These nonphosphorylated peptide inhibitors are very desirable given that the negatively charged phosphate group of phosphopeptides is a detriment to the cell permeability of pharmaceutical compounds. For instance, a cyclic, 11-residue nonphosphorylated peptide that contains the crucial Asn has been shown to bind speci®cally and with micromolar af®nity to the Grb2 SH2 domain (Oligino et al., 1997). Although it is not entirely clear how this peptide binds, a Glu residue 2 residues N-terminal to the tyrosine may be important (Long et al., 1999). This compound may be a starting point for the design of other nonphosphorylated inhibitors of the Grb2 SH2 domain with improved pharmacologic properties. Other noncyclic inhibitors of the Grb2 SH2 domain have also been found. Studies with these compounds have highlighted how Grb2 couples activation of the Ras signaling pathway to cellular transformation (Gay et al., 1999a). Furthermore, Grb2 inhibitors have also demonstrated that this protein plays a role in the motility of tumor cells (Gay et al., 1999b). These results have highlighted how inhibition of the Grb2 SH2 domain could be an effective anticancer treatment.

VII. SUMMARY In this chapter, we have described the biophysical investigations which have dissected the mechanisms of SH2 domain function. Due to nearly a decade and a half of investigation on SH2 domains, much about their binding mechanism has been characterized. SH2 domains have been found to have a positively charged binding cavity, largely conserved between different SH2 domains, which coordinates binding of the pTyr in the target. The ionic interactions between this pocket and the pTyr, in particular, between Arg bB5 and the phosphate, provide the majority of the binding energy stabilizing SH2 domain±target interactions. The speci®city in SH2 domain±target interactions emanates most often from the interactions between the residues C-terminal to the pTyr in the target and the speci®city determining residues in the C-terminal half of the SH2 domain. However, the interactions in the speci®city determining region of SH2 domains are weak, and hence single SH2 domains show only a modest level of speci®city for tyrosine phosphorylated targets. Greater speci®city in SH2 domain-containing protein±tyrosine phosphorylated

MOLECULAR RECOGNITION BY SH2 DOMAINS

205

target interactions can be achieved by placing SH2 domains in tandem (as is often found) or possibly through speci®c localization of SH2 domaincontaining proteins within the cell. Although a relatively good understanding of how SH2 domains function in isolation has been obtained, the ways in which SH2 domain binding is coupled to allosteric transmission of signals in larger SH2 domain-containing proteins are still not clear. Hence, the future should bring further investigations of the mechanisms by which SH2 domain ligation alters the enzymatic activity and cellular localization of SH2 domain-containing proteins.

ACKNOWLEDGMENT This work was funded by NIH Grant GM60231.

REFERENCES Ackers, G. K., and Smith, F. R. (1985). Annu. Rev. Biochem. 54, 597±629. Alligood, K. J., Charifson, P. S., Crosby, R., Consler, T. G., Feldman, P. L., Gampe, R. T., Gilmer, T. M., Jordan, S. R., Milstead, M. W., Mohr, C., Peel, M. R., Rocque, W., Rodriguez, M., Rusnak, D. W., Shewchuk, L. M., and Sternbach, D. D. (1998). Bioorg. Med. Chem. Lett. 8, 1189±1194. Anderson, D., Koch, C. A., Grey, L., Ellis, C., Moran, M. F., and Pawson, T. (1990). Science 250, 979±982. Auger, K. R., Carpenter, C. L., Shoelson, S. E., Piwnica, W. H., and Cantley, L. C. (1992). J. Biol. Chem. 267, 5408±5415. Bartford, D., and Neel, B. G. (1998). Structure 6, 249±254. Becker, S., Groner, B., and Mu È ller, C. W. (1998). Nature 394, 145±151. Bibbins, K. B., Boeuf, H., and Varmus, H. E. (1993). Mol. Cell. Biol. 13, 7278±7287. Booker, G. W., Breeze, A. L., Downing, A. K., Panayotou, G., Gout, I., Water®eld, M. D., and Campbell, I. D. (1992). Nature 358, 684±687. Bottomley, M. J., Salim, K., and Panayotou, G. (1998). Biochim. Biophys. Acta 1436, 165±183. Bradshaw, J. M., Grucza, R. A., Ladbury, J. E., and Waksman, G. (1998). Biochemistry 37, 9083±9090. Bradshaw, J. M., Mitaxov, V., and Waksman, G. (1999). J. Mol. Biol. 293, 971±985. Bradshaw, J. M., Mitaxov, V., and Waksman, G. (2000). J. Mol. Biol. 299, 549±563. Bradshaw, J. M., and Waksman, G. (1998). Biochemistry 37, 15400±15407. Bradshaw, J. M., and Waksman, G. (1999). Biochemistry 38, 5147±5154. Breeze, A. L., Kara, B. V., Barratt, D. G., Anderson, M., Smith, J. C., Luke, R. W., Best, J. R., and Cartlidge, S. A. (1996). EMBO J. 15, 3579±3589. Brown, M. T., and Cooper, J. A. (1996). Biochem. Biophys. Acta 1287, 121±149. Brugge, J. S. (1993). Science 260, 918±919. Bubeck-Wardenburg, J., Wong, J., Futterer, K., Pappu, R., Fu, C., Waksman, G., and Chan, A. C. (1999). Prog. Biophys. Mol. Biol. 71, 373±392. Burke, T. R., Jr., Barchi, J. J., Jr., George, C., Wolf, G., Shoelson, S. E., and Yan, X. (1995). J. Med. Chem. 38, 1386±1396. Burshtyn, D. N., Yang, W., Yi, T., and Long, E. O. (1997). J. Biol. Chem. 272, 13066±13072.

206

J. MICHAEL BRADSHAW AND GABRIEL WAKSMAN

Cantley, L. C., Auger, K. R., Carpenter, C., Duckworth, B., Graziani, A., Kapeller, R., and Soltoff, S. (1991). Cell 64, 281±302. Carson, M. (1997). Methods Enzymol. 277, 493±505. Carter, P. J., Winter, G., Wilkinson, A. J., and Fersht, A. R. (1984). Cell 38, 835±840. Chacko, G. W., Brandt, J. T., Coggeshall, K. M., and Anderson, C. L. (1996). J. Biol. Chem. 271, 10775±10781. Charifson, P. S., Shewchuk, L. M., Rocque, W., Hummel, C. W., Jordan, S. R., Mohr, C., Pacofsky, G. J., Peel, M. R., Rodriguez, M., Sternbach, D. D., and Consler, T. G. (1997). Biochemistry 36, 6283±6293. Chen, X., Vinkemeier, U., Zhao, Y., Jeruzalmi, D., Darnell, J. E., Jr., and Kuriyan, J. (1998). Cell 93, 827±839. Chook, Y. M., Gish, G. D., Kay, C. M., Pai, E. F., and Pawson, T. (1996). J. Biol. Chem. 271, 30472±30478. Clackson, T., and Wells, J. A. (1995). Science 267, 383±386. Coffey, A. J., Brooksbank, R. A., Brandau, O., Oohashi, T., Howell, G. R., Bye, J. M., Cahn, A. P., Durham, J. D., Heath, P., Wray, P., Pavitt, R., Wilkinson, J., Leversha, M., Huckel, E., Shaw-Smith, C. J., Dunham, A., Rhodes, S., Schuster, V., Porta, G., Yin, L., Sera®ni, P., Sylla, B., Zollo, M., Franco, B., Bolino, A., Seri, M., Lanyi, A., Davis, J. R., Webster, D., Harris, A., Lenoir, G., de St Basile, G., Jones, A., Behoradsky, B. H., Achatz, H., Murken, J., Fassler, R., Sumegi, J., Romeo, G., Vaudin, M., Ross, M. T., Meindl, A., and Bentley, D. R. (1998). Nat. Genet. 20, 129±135. Cooper, J. A., and Howell, B. (1993). Cell 73, 1051±1054. Corey, S. J., Burkhardt, A. L., Bolen, J. B., Geahlen, R. L., Tkatch, L. S., and Tweardy, D. J. (1994). Proc. Natl. Acad. Sci. USA 91, 4683±4687. Cousins-Wasti, R., Ingraham, R. H., Morelock, M. M., and Grygon, C. A. (1996). Biochemistry 35, 16746±16752. Cowburn, D., Zheng, J., Xu, Q., and Baranay, G. (1995). J. Biol. Chem. 270, 26738±26741. Cussac, D., Frech, M., and Chardin, P. (1994). EMBO J. 13, 4011±4021. Dall'Acqua, W., Goldman, E. R., Lin, W., Teng, C., Tsuchiya, D., Li, H., Ysern, X., Braden, B. C., Li, Y., Smith-Gill, S. J., and Mariuzza, R. A. (1998). Biochemistry 37, 7981±7991. Darnell, J. E. Jr. (1997). Science 277, 1630±1635. Domchek, S. M., Auger, K. R., Chatterjee, S., Burke, T. R., Jr., and Shoelson, S. E. (1992). Biochemistry 31, 9865±9870. Eck, M. J., Shoelson, S. E., and Harrison, S. C. (1993). Nature 362, 87±91. Eck, M. J., Atwell, S. K., Shoelson, S. E., and Harrison, S. C. (1994). Nature 368, 764±769. Eck, M. J., Pluskey, S., Trub, T., Harrison, S. C., and Shoelson, S. E. (1996). Nature 379, 277±280. Engen, J. R., Smithgall, T. E., Gmeiner, W. H., and Smith, D. L. (1999). J. Mol. Biol. 287, 645±656. Escobedo, J. A., Kaplan, D. R., Kavanaugh, W. M., Turck, C. W., and Williams, L. T. (1991). Mol. Cell. Biol. 11, 1125±1132. Fantl, W. J., Escobedo, J. A., Martin, G. A., Turck, C. W., Rosario, M., McCormick, F., and Williams, L. T. (1992). Cell 69, 413±423. Fantl, W. J., Johnson, D. E., and Williams, L. T. (1993). Annu. Rev. Biochem. 62, 453±481. Fushman, D., Xu, R., and Cowburn, D. (1999). Biochemistry 38, 10225±10230. Fu È tterer, K., Wong, J., Grucza, R. A., Chan, A. C., and Waksman, G. (1998). J. Mol. Biol. 281, 523±537. Gay, B., Suarez, S., Caravatti, G., Furet, P., Meyer, T., and Schoepfer, J. (1999a). Int. J. Cancer 83, 235±241. Gay, B., Suarez, S., Weber, C., Raheul, J., Fabbro, D., Furet, P., Caravatti, G., and Schoepfer, J. (1999b). J. Biol. Chem. 274, 23311±23315.

MOLECULAR RECOGNITION BY SH2 DOMAINS

207

Gilmer, T., Rodriguez, M., Jordan, S., Crosby, R., Alligood, K., Green, M., Kimery, M., Wagner, C., Kinder, D., Charifson, P., Hassel, A. M., Willard, D., Luther, M., Rusnak, D., Sternbach, D. D., Mehrotra, M., Peel, D., Shampine, L., Davis, R., Robbins, J., Patel, I. R., Kassel, D., Burkhart, W., Moyer, M., Bradshaw, T., and Berman, J. (1994). J. Biol. Chem. 269, 31711±31719. Goldman, E. R., Dall'Acqua, W., Braden, B. C., and Mariuzza, R. A. (1997). Biochemistry 36, 49±56. Gon¯oni, S., Frischknecht, F., Way, M., and Superti-Furga, G. (1999). Nat. Struct. Biol. 6, 760±764. Gosser, Y. Q., Zheng, J., Overduin, M., Mayer, B. J., and Cowburn, D. (1995). Structure 3, 1075±1086. Grazioli, L., Germain, V., Weiss, A., and Acuto, O. (1998). J. Biol. Chem. 273, 8916±8921. Grucza, R. A., Bradshaw, J. M., Mitaxov, V., and Waksman, G. (2000), Biochemistry 39, 10072±10081. Grucza, R. A., Fu È tterer, K., Chan, A. C., and Waksman, G. (1999). Biochemistry 38, 5024±5033. Gu È nther, U. L., Liu, Y. S., Sanford, D., Bachovlin, W. W., and Schaffhausen, B. (1996). Biochemistry 35, 15570±15581. Hatada, M. H., Lu, X., Laird, E. R., Green, J., Morgenstern, J. P., Lou, M., Marr, C. S., Phillips, T. B., Ram, M. K., Theriault, K., Zoller, M. J., and Karas, J. L. (1995). Nature 377, 32±38. Hirai, H., and Varmus, H. E. (1990a). Proc. Natl. Acad. Sci. USA 87, 8592±8596. Hirai, H., and Varmus, H. E. (1990b). Mol. Cell. Biol. 10, 1307±1318. Hof, P., Pluskey, S., Dhe-Paganon, S., Eck, M. J., and Shoelson, S. E. (1998). Cell 92, 441±450. Hunter, T. (2000). Cell 100, 113±127. Huyer, G., Li, Z. M., Adam, M., Huckle, W. R., and Ramachandran, C. (1995). Biochemistry 34, 1040±1049. Huyer, G., and Ramachandran, C. (1998). Biochemistry 37, 2741±2747. Johnson, L. N., Noble, N. E. M., and Owen, D. J. (1996). Cell 85, 149±158. Jove, R., and Hanafusa, H. (1987). Annu. Rev. Cell Biol. 3, 31±56. Kawata, T., Shevchenko, A., Fukuzawa, M., Jermyn, K. A., Totty, N. F., Zhukovskaya, N. V., Sterling, A. E., Mann, M., and Williams, J. G. (1997). Cell 89, 909±916. Kay, L. E., Muhandiram, D. R., Farrow, N. A., Aubin, Y., and Forman-Kay, J. D. (1996). Biochemistry 35, 361±368. Kay, L. E., Muhandiram, D. R., Wolf, G., Shoelson, S. E., and Forman-Kay, J. D. (1998). Nat. Struct. Biol. 5, 156±163. Kazlauskas, A., Kashishian, A., Cooper, J. A., and Valius, M. (1992). Mol. Cell. Biol. 12, 2534±2544. Koch, C. A., Anderson, D., Moran, M. F., Ellis, C., and Pawson, T. (1991). Science 252, 668±674. Kuriyan, J., and Cowburn, D. (1997). Annu. Rev. Biophys. Biomol. Struct. 26, 259±288. Labadia, M. E., Jakes, S., Grygon, C. A., Greenwood, D. J., Schembri-King, J., Lukas, S. M., Warren, T. C., and Ingraham, R. H. (1997). Arch. Biochem. Biophys. 342, 117±125. Ladbury, J. E., Hensmann, M., Panayotou, G., and Campbell, I. D. (1996). Biochemistry 35, 11062±11069. Ladbury, J. E., Lemmon, M. A., Zhou, M., Green, J., Bot®eld, J., and Schlessinger, J. (1995). Proc. Natl. Acad. Sci. USA 92, 3199±3203. Lee, C. H., Kominos, D., Jacques, S., Margolis, B., Schlessinger, J., Shoelson, S. E., and Kuriyan, J. (1994). Structure 2, 423±438.

208

J. MICHAEL BRADSHAW AND GABRIEL WAKSMAN

Lemmon, M. A., and Ladbury, J. E. (1994). Biochemistry 33, 5070±5076. Lemmon, M. A., Ladbury, J. E., Mandiyan, V., Zhou, M., and Schlessinger, J. (1994). J. Biol. Chem. 269, 31653±31658. Li, S.-C., Gish, G., Yang, D., Coffey, A. J., Forman-Kay, J. D., Ernberg, I., Kay, L. E., and Pawson, T. (1999). Curr. Biol. 9, 1355±1362. Liu, Y. C., and Altman, A. (1998). Cell Signal. 10, 377±385. Long, Y. Q., Yao, Z. J., Voigt, J. H., Lung, F. D., Luo, J. H., Burke, T. R., Jr., King, C. R., Yang, D., and Roller, P. P. (1999). Biochem. Biophys. Res. Commun. 264, 902±908. Maignan, S., Guilloteau, J. P., Fromage, N., Arnoux, B., Becquart, J., and Ducruix, A. (1995). Science 268, 291±293. Marengere, L. E., Songyang, Z., Gish, C. D., Schaffer, M. D., Parsons, J. T., Stern, M. J., Cantley, L. C., and Pawson, T. (1994). Nature 369, 502±505. Matsuda, M., Mayer, B. J., Fukui, Y., and Hanafusa, H. (1990). Science 248, 1537±1539. Mayer, B. J., and Hanafusa, H. (1990a). J. Virol. 64, 3581±3589. Mayer, B. J., and Hanfusa, H. (1990b). Proc. Natl. Acad. Sci. USA 87, 2638±2642. Mayer, B. J., Jackson, P. K., and Baltimore, D. (1991). Proc. Natl. Acad. Sci. USA 88, 627±631. Mayer, B. J., Jackson, P. K., Van Etten, R. A., and Baltimore, D. (1992). Mol. Cell. Biol. 12, 609±618. McNemar, C., Snow, M. E., Windsor, W. T., Prongay, A., Mui, P., Zhang, R., Durkin, J., Le, H. V., and Weber, P. C. (1997). Biochemistry 36, 10006±10014. Meng, W., Sawasdikosoi, S., Burakoff, S. J., and Eck, M. J. (1999). Nature 398, 84±90. Moare®, I., LaFevre-Bernt, M., Sicheri, F., Huse, M., Lee, C.-H., Kuriyan, J., and Miller, W. T. (1997). Nature 385, 650±653. Moran, M. F., Koch, C. A., Anderson, D., Ellis, C., England, L., Martin, G. S., and Pawson, T. (1990). Proc. Natl. Acad. Sci. USA 87, 8622±8626. Morelock, M. M., Ingraham, R. H., Betageri, R., and Jakes, S. (1995). J. Med. Chem. 38, 1309±1318. Mori, S., Ro Ènnstrand, L., Yokote, K., Engstro Èm, A., Courtneidge, S. A., Claesson-Welsh, L., and Heldin, C.-H. (1993). EMBO J. 12, 2257±2264. Musacchio, A., Noble, M., Pauptit, R., Wierenga, R., and Saraste, M. (1992). Nature 359, 851±855. Nam, H.-J., Haser, W. G., Roberts, T. M., and Frederick, C. A. (1996). Structure 4, 1105± 1114. Nicholls, A., Sharp, K. A., and Honig, B. (1991). Protein Struct. Funct. Genet. 11, 281±296. Nichols, K. E., Harkin, D. P., Levitz, S., Krainer, M., Kolquist, K. A., Genovese, C., Bernhard, A., Ferguson, M., Zuo, L., Snyder, E., Buckler, A. J., Wise, C., Ashley, J., Lovett, M., Valentine, M. B., Look, A. T., Gerald, W., Housman, D. E., and Haber, D. A. (1998). Proc. Natl. Acad. Sci. USA 95, 13765±13770. Nolte, R. T., Eck, M. J., Schlessinger, J., Shoelson, S. E., and Harrison, S. C. (1996). Nat. Struct. Biol. 3, 364±373. O'Brien, R., Rugman, P., Renzoni, D., Layton, M., Handa, R., Hilyard, K., Water®eld, M. D., Driscoll, P. C., and Ladbury, J. E. (2000). Protein Sci. 9, 570±579. Oligino, L., Lung, F. T., Sastry, L., Bigelow, J., Cao, T., Curran, M., Burke, T. R., Jr., Wang, S., Krag, D., Roller, P. P., and King, C. R. (1997). J. Biol. Chem. 272, 29046±29052. Ottinger, E. A., Bot®eld, M. C., and Shoelson, S. E. (1998). J. Biol. Chem. 273, 729±735. Overduin, M., Rios, C. B., Mayer, B. J., Baltimore, D., and Cowburn, D. (1992). Cell 70, 697±704. Panayotou, G., Bax, B., Gout, I., Federwisch, M., Wroblowski, B., Dhand, R., Fry, M. J., Blundell, T. L., Wollmer, A., and Water®eld, M. D. (1992). EMBO J. 11, 4261±4272. Pascal, S., Singer, A. U., Gish, G., Yamazaki, T., Shoelson, S. E., Pawson, T., Kay, L. E., and Forman-Kay, J. D. (1994). Cell 77, 461±472.

MOLECULAR RECOGNITION BY SH2 DOMAINS

209

Pawson, T. (1995). Nature 373, 573±580. Pawson, T., and Gish, G. D. (1992). Cell 71, 359±362. Pawson, T., and Saxton, T. M. (1999). Cell 97, 675±678. Pawson, T., and Schlessinger, J. (1993). Curr. Biol. 3, 434±442. Pawson, T., and Scott, J. D. (1997). Science 278, 2075±2080. Payne, G., Shoelson, S. E., Gish, G. D., Pawson, T., and Walsh, C. T. (1993). Proc. Natl. Acad. Sci. USA 90, 4902±4906. Payne, G., Stolz, L. A., Pei, D., Band, H., Shoelson, S. E., and Walsh, C. T. (1994). Chem. Biol. 1, 99±105. Pei, D., Wang, J., and Walsh, C. T. (1996). Proc. Natl. Acad. Sci. USA 93, 1141±1145. Piccione, E., Case, R. D., Domchek, S. M., Hu, P., Chaudhuri, M., Backer, J. M., Schlessinger, J., and Shoelson, S. E. (1993). Biochemistry 32, 3197±3202. Porter, M., Schindler, T., Kuriyan, J., and Miller, W. T. (2000). J. Biol. Chem. 275, 2721±2726. Poy, F., Yaffe, M. B., Sayos, J., Saxena, K., Morra, M., Sumegi, J., Cantley, L. C., Terhorst, C., and Eck, M. J. (1999). Mol. Cell 4, 555±561. Rahuel, J., Gay, B., Erdmann, D., Strauss, A., Garcia-Echeverria, C., Furet, P., Caravatti, G., Fretz, H., Schoepfer, J., and Grutter, M. G. (1996). Nat. Struct. Biol. 3, 586±589. Rameh, L. E., Chen, C., and Cantley, L. C. (1995). Cell 83, 821±830. Rich, D. H. (1986). ``Proteinase Inhibitors.'' Elsevier, Amsterdam. Rose, T., Rose, J. M., and Waksman, G. (1999). www.biochem.wustl.edu/~sh2domains. Sadowski, I., Stone, J. C., and Pawson, T. (1986). Mol. Cell. Biol. 6, 4396±4408. Sawyer, T. K. (1998). Biopolymers 47, 243±261. Sayos, J., Wu, C., Morra, M., Wang, N., Zhang, X., Allen, D., van Schaik, S., Notaranglo, L., Geha, R., Roncarolo, M. G., Oettgen, H., De Vries, J. E., and Ternost, C. (1998). Nature 395, 462±469. Senderowicz, L., Wang, J.-X., Wang, L.-Y., Yoshizawa, S., Kavanaugh, W. M., and Turck, C. W. (1997). Biochemistry 36, 10538±10544. Sicheri, F., and Kuriyan, J. (1997). Curr. Opin. Struct. Biol. 7, 777±785. Sicheri, F., Moare®, I., and Kuryan, J. (1997). Nature 385, 602±609. Singer, A. U., and Forman-Kay, J. D. (1997). Protein Sci. 6, 1910±1919. Songyang, Z., Gish, G., Mbamalu, G., Pawson, T., and Cantley, L. C. (1995). J. Biol. Chem. 270, 26029±26032. Songyang, Z., Shoelson, S. E., Chaudhuri, M., Gish, G., Pawson, T., Haser, W. G., King, F., Roberts, T., Ratnofski, S., Lechleider, R. J., Neel, B. G., Birge, R. B., Fajardo, J. E., Chou, M. M., Hanafusa, H., Schaffhausen, B., and Cantley, L. C. (1993). Cell 72, 767±778. Songyang, Z., Shoelson, S. E., McGlade, J., Olivier, P., Pawson, T., Bustelo, X. R., Barbacid, M., Sabe, H., Hanafusa, H., Yi, T., Ren, R., Baltimore, D., Ratnofski, S., Feldman, R. A., and Cantley, L. C. (1994). Mol. Cell. Biol. 14, 2777±2785. Soriano, P., Montgomery, C., Geske, R., and Bradley, A. (1991). Cell 64, 693±702. Superti-Furga, G., Fumagalli, S., Koegl, M., Courtneidge, S., and Draetta, G. (1993). EMBO J. 12, 2625±2634. Surdo, P. L., Bottomley, M. J., Arcaro, A., Siegal, G., Panayotou, G., Sanakar, A., Gaffney, P. R. J., Riley, A. M., Potter, B. V. L., Water®eld, M. D., and Driscoll, P. C. (1999). J. Biol. Chem. 274, 15678±15685. Tessari, M., Gentile, L. N., Taylor, S. J., Shalloway, D. I., Nicholson, L. K., and Vuister, G. W. (1997). Biochemistry 36, 14561±14571. Violette, S. M., Shakespeare, W. C., Barlett, C., Guan, W., Smith, J. A., Rickles, R. J., Bohacek, R. S., Holt, D. A., Baron, R., and Sawyer, T. K. (2000). Chem. Biol. 7, 225±235. Waksman, G., Kominos, D., Robertson, S. R., Pant, N., Baltimore, D., Birge, R. B., Cowburn, D., Hanafusa, H., Mayer, B. J., Overduin, M., Resh, M. D., Rios, C. B., Silverman, L., and Kuriyan, J. (1992). Nature 358, 646±653.

210

J. MICHAEL BRADSHAW AND GABRIEL WAKSMAN

Waksman, G., Shoelson, S. E., Pant, N., Cowburn, D., and Kuriyan, J. (1993). Cell 72, 779±790. Weijland, A., Williams, J. C., Neubauer, G., Courtneidge, S. A., Wierenga, R. K., and Superti-Furga, G. (1997). Proc. Natl. Acad. Sci. USA 94, 3590±3595. Williams, J. C., Weijland, A., Gon¯oni, S., Thompson, A., Courtneidge, S. A., Superti-Furga, G., and Wierenga, R. K. (1997). J. Mol. Biol. 274, 757±775. Xu, Q., Zheng, J., Xu, R., Barany, G., and Cowburn, D. (1999). Biochemistry 38, 3491±3497. Xu, W., Harrison, S. C., and Eck, M. J. (1997). Nature 385, 595±602. Yamaguchi, H., and Hendrickson, W. A. (1996). Nature 384, 484±489. Yu, H., Rosen, M. K., Shin, T. B., Seidel-Dugan, C., Brugge, J. S., and Schreiber, S. L. (1992). Science 258, 1665±1668.

HOW SH3 DOMAINS RECOGNIZE PROLINE BY ANDREA MUSACCHIO Department of Experimental Oncology, European Institute of Oncology, 20141 Milan, Italy

I. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . II. Proline and Polyproline Type II Helices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . III. The SH3 Domain: A Model System to Understand Interactions Mediated by Proline-Rich PPII Helices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A. General Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B. Occurrence of SH3 Domains . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . C. Structure of SH3 Domains . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . D. Ligand Speci®city of SH3 Domains: An Outline. . . . . . . . . . . . . . . . . . . . . . . E. The SH3 Ligand-Binding Site . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . F. Lim's Model of SH3±Peptide Interactions. . . . . . . . . . . . . . . . . . . . . . . . . . . . G. Detailed Analysis of Selected SH3±Peptide Interactions . . . . . . . . . . . . . . . . H. The Interaction of HIV Nef with SH3 Domains . . . . . . . . . . . . . . . . . . . . . . I. Macromolecular Assemblies Containing SH3 Domains . . . . . . . . . . . . . . . . . J. Interaction of SH3 Domains with Sequences That Do Not Conform to a Proline-Rich Consensus. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . K. Regulation of SH3 Binding to Substrates . . . . . . . . . . . . . . . . . . . . . . . . . . . . L. Inhibition of SH3±Mediated Interactions . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

211 212 215 215 215 216 221 230 232 235 243 245 249 250 252 255

I. INTRODUCTION It became clear in the early 1990s that short proline-rich segments in proteins are directly involved in molecular recognition events such as signal transduction, vesicular traf®cking, and cytoskeletal dynamics. In this chapter, we focus on the key structural features of the interaction of proline-rich regions with target proteins or protein modules. In particular, we address the mechanisms of protein recognition mediated by nonrepetitive proline-rich regions adopting a polyproline helix type II (PPII helix) conformation. Cellular targets of this type of proline-rich region, the Src-homology 3 (SH3) domains and pro®lin, were identi®ed in the early 1990s. These discoveries were a prelude to the identi®cation of several other proteins or protein domains involved in the speci®c, direct recognition of proline-rich regions, such as the WW and EVH1 domains (reviewed in references 1, 2). It has now become clear that the network of cellular interactions involving PPII helices is astonishingly complex. Molecules such as the Wiskott±Aldrich Syndrome protein (WASP), the Son-of-sevenless (Sos) protein, and dynamin, which contain extended proline-rich regions, are able to engage in simultaneous or exclusive interactions with a variety of different targets. How these interactions are regulated and how they are used to gain control over the catalytic output 211 ADVANCES IN PROTEIN CHEMISTRY, Vol. 61

Copyright 2003, Elsevier Science (USA). All rights reserved. 0065-3233/03 $35.00

212

ANDREA MUSACCHIO

delivered by these proteins are only now beginning to be understood. The interaction of SH3 domains with biologically relevant proline-rich targets has been intensely investigated, and a general model explaining the structural basis of speci®city in this system is now available. In this chapter, we will focus on the structural basis of speci®c and selective recognition of proline-rich regions by the SH3 domain using appropriate examples available in the structure database.

II. PROLINE AND POLYPROLINE TYPE II HELICES A proline residue in a polypeptide has its f torsion angle restricted to 65 (158). The d carbon of the proline ring interacts with the preceding residue, hindering that residue from adopting helical c angles (3). In this chapter we will be especially concerned with nonrepetitive prolinerich regions. These segments occur in many prokaryotic and eukaryotic proteins and appear to be involved in the regulation of important cellular phenomena (1, 4, 5). In most instances, they are expected to form a secondary structural element, known as the poly-L -proline helix of type II, an all-trans left-handed helix, with f and c angles centered around 78 and 146 , respectively (6). The PPII helix is characterized by perfect three-fold rotational symmetry, and the extension of one turn is Ê . As a way of simplifying our view of PPII helices, we can about 9 A represent them as triangular prisms (Fig. 1). This representation is quite useful and will be used at several stages in this paper. It reminds us that PPII helices are threefold symmetric along the helical axis. Also, it

FIG. 1. The polyproline type II helix is an all-trans left-handed helix, whose structure can be represented as a triangular prism. The three consecutive side chains that account for one turn of the PPII helix occupy a different edge of the prism.

HOW SH3 DOMAINS RECOGNIZE PROLINE

213

suggests the presence of twofold symmetry perpendicular to the helical axis. We will show in a later section that this is not generally true, but that it becomes roughly correct when proline is included at certain positions in the helix. In fact, the pseudo-twofold symmetry implied by the presence of proline in a PPII helix is a key structural feature to understanding the binding of proline-rich peptides to SH3 domains. The carbonyl and amide groups of the polypeptide main chain in the PPII conformation are solvent-exposed and available for hydrogen bonding. Very few side chain interactions can occur within the PPII helix, due to its extended conformation. The most frequent side chain rotamers are characterized by x1 trans ( 120 > x1 > 120 ) and gauche (0 > x1 > 120 ) angles (7). Intrachain hydrogen bonding, however, is possible, with Gln being the most frequent ``capping'' residue. Surveys of the structural data bank have shown that PPII helices occur quite commonly within globular proteins (7±9). Within a group of 274 nonhomologous, high-resolution polypeptide chains from proteins of known structure, more than half were found to contain at least one segment of the PPII helix of chain length greater than 3 (7). According to this survey, PPII represents 2% of the total residues in this database, and no chains were found in which PPII constitutes more than 12% of the secondary structure of a globular protein. The threefold symmetry of the PPII helix and its high surface exposure result in a correlation between the polarity of residue pairs spaced i and i 3, indicating that PPII helices tend to be amphipathic (7). Despite the relatively low content, PPII helices tend to be very well conserved in divergent members of homologous proteins, indicating that they are structurally or functionally important elements of a protein's secondary structure (10). Little is known concerning the overall structural organization of long proline-rich regions (PRR), such as those found in the C-terminal tail of RNA polymerase II, in dynamin, or in Sos. An intrinsic limitation in this assessment is the dif®culty of obtaining structural information on long proline-rich fragments. Globular proteins accessible for high-resolution structural analysis may not be representative of the actual fraction of residues in PPI conformation, and the overall content of this secondary structural element may be larger. Indeed, Sreerama and Woody estimated that 10% of all protein residues might adopt the PPI conformation (9). The long C-terminal PRR of RNA polymerase II, a binding target of a WW domain characterized by tandem copies of the heptapeptide sequence YSPTSPS (11), is predominantly unordered in water (12). Similarly, short proline-rich peptides corresponding to known SH3 ligands exhibit little or no secondary structure prior to binding their cognate protein target (13±17). Under these conditions, the association of a

214

ANDREA MUSACCHIO

proline-rich peptide with the SH3 domain indicates unfavorable binding entropy, likely resulting from a loss of rotational freedom on formation of the PPII helix (16). Thus, the completion of the folding process for polyproline helices may be triggered by the interaction with one or multiple ligands. This may be especially true for sequences that on binding the ligand adopt this conformation even if they do not contain proline residues, such as the peptides bound to the class II major histocompatibility complex (18, 19). Partial stabilization of the PPII conformation may also occur through the interaction with folded segments within the same polypeptide chain, as shown by the Src homology 2 (SH2)kinase linker region in Src family kinases (20±22). 1 of the total proline residues in the globular protein Although only 10 data set were in PPII conformation, proline largely predominates in PPII helices (7). The only other signi®cantly favored residue is Gln. Gly is particularly disfavored. Apolar residues are also disfavored, probably because PPII helices tend to be on the surface of proteins and are thus markedly solvent-exposed. The sequences recognized by SH3 or WW domains, however, contain a signi®cant fraction of apolar residues. There is some confusion on the issue of whether proline is essential to maintain the PPII conformation. We have already reported examples of peptides adopting this conformation in the absence of proline residues. As already observed, proline restricts the conformational freedom of the preceding residue in the polypeptide chain. Simulation of PPII formation by peptides with decreasing proline content suggests that the presence of proline has a profound effect on the formation of the PPII conformation by ``steric interaction'' directed toward the N-terminus of the peptide (23). However, the effect is local, and it does not extend beyond the preceding residue, as shown by simulations with sequences containing several nonproline (and nonglycine) residues followed by a stretch of prolines (23). This suggests that the folding process leading to the formation of PPII helices in segments containing several residues other than proline bene®ts from interactions with the body of the protein to which that helix belongs and/or with a protein ligand. When considering the details of PPII helix binding to SH3 or WW domains, however, it is important to emphasize that the special role of proline can be explained by its direct role in ligand recognition, rather than by increased structural stability. The proline ring (or more speci®cally, its Cd atom) is to be regarded as the main creator of ligand speci®city in this system, and its substitution has a deleterious effect on binding that derives mainly from the loss of van der Waals contacts, rather than from the destabilization of the PPII conformation. The interaction of SH3 domains with proline-rich peptides relies on the unique feature of the proline side chain, N-alkyl substitution.

HOW SH3 DOMAINS RECOGNIZE PROLINE

215

III. THE SH3 DOMAIN: A MODEL SYSTEM TO UNDERSTAND INTERACTIONS MEDIATED BY PROLINE-RICH PPII HELICES A. General Remarks We report the main features of SH3 domains and attempt an explanation of the structural basis of their ability to interact with proline-rich substrates. Initially, we will limit our discussion to interactions that can be collectively regarded as involving proline-rich substrates containing an XPXXP motif. We will present a model that explains the role played by proline in the interaction with substrates. This will serve as an introduction to the Nef±Hck-SH3 interaction, which has added to our structural knowledge the understanding that interactions outside the XPXXP binding region are also important for SH3 binding. We will also present the structurally well-characterized Src family tyrosine kinases to illustrate the function of an SH3 domain within an intramolecular autoregulatory switch. Finally, we will discuss examples of noncanonical interactions depending on sequences other than proline-rich segments.

B. Occurrence of SH3 Domains SH3 domains are 60-residue protein modules found in proteins generally involved in the regulation of dynamic processes occurring at the plasma membrane, such as the organization of the cytoskeleton, the transduction of extracellular signals, and the internalization of membrane receptors. In order to maintain a focus on the general structural aspects of SH3 interaction with its substrates, we deliberately chose to avoid detailed explanations of the biological signi®cance of the pathways in which SH3 domains are involved. Several excellent recent reviews have focused on this aspect of SH3 (24±28). A survey of DNA and protein sequence databases using the SMART domain analysis tool (http://smart.embl-heidelberg.de/) (29) retrieved 1013 occurrences of the domain from a total of 781 proteins or open reading frames in the nonredundant database (July 2000). Of these, 25 (with a total of 29 domains) were from the budding yeast Saccharomyces cerevisiae, 37 from the fruit ¯y Drosophila melanogaster, and 64 from the nematode Caenorhabditis elegans. Two hundred seventy human SH3containing proteins are identi®ed. Among these, 53 contain more than one SH3 domain, and the total number of occurrences of the domain is 375. These SH3 domains are often ¯anked by other protein- or lipidbinding modules. For instance, 54 SH3-containing proteins also contain one or more SH2 domains, and 39 contain pleckstrin homology (PH) domains (27 of which did not contain an SH2 domain). Thus SH2, PH, and

216

ANDREA MUSACCHIO

the SH3 domain itself represent the most frequent classes of noncatalytic domains accompanying SH3 domains. Several SH3-containing proteins do not contain a catalytic domain and are exclusively made up of domains involved in protein±protein interactions. These molecules go by the name of ``adapter proteins,'' and act as molecular connectors to activate certain signaling pathways under the appropriate physiological conditions. A well-known example of a molecular adapter is the Grb2 protein, which links activated transmembrane tyrosine kinase receptors to the upregulation of the MAP kinase cascade regulated by the Ras protooncogene (reviewed in reference 30). In addition to noncatalytic modules, SH3 domains are often found in the proximity of catalytic domains. Six human serine±threonine kinases contain an SH3 domain, while there are 30 instances of SH3-containing human tyrosine kinases. The most frequent catalytic domain to be found within SH3-containing proteins is the guanylate kinase domain (31 occurrences). A large fraction of these proteins are subunits of membrane ion channels or proteins localized at tight junctions. A large fraction of SH3-containing proteins are associated with domains regulating the activity of GTP-binding proteins: there are 21 SH3 domains associated with guanine-nucleotide exchange factor domains speci®c for the Rho family of small GTP-binding proteins and 11 with GTPase-activating protein domains speci®c for the Ras or Rho families. So far as we can tell, SH3 domains are always ¯anked by other domains. The only exception to this statement is represented by an S. cerevisiae open reading frame (Accession No. Q12065m) of 103 amino acids that is almost exclusively made up of an SH3 domain. The SH3 domain does not occupy a ®xed topological position, but its modular nature should not be taken as an indication that its function is exercised in isolation from the other domains within the protein body. We will present striking examples of ``domain cooperation'' in which the SH3 domain utilizes the presence of other domains in cis to carry out its function. This is not an uncommon theme in multidomain proteins. C. Structure of SH3 Domains The almost concomitant report of the crystal and solution structures from several unliganded SH3 domains (31±36) revealed a completely conserved fold: a compact b-barrel, formed by two anti-parallel, threestranded b-sheets (Fig. 2A). A structure-based alignment of a subset of SH3 domains is shown in Fig. 2B. The structure starts at the poorly conserved b1 strand. After this strand, the polypeptide chain engages in the formation of the so-called RT loop, connecting strands bA and bB (see below for an explanation of this nomenclature). One of the sequence

HOW SH3 DOMAINS RECOGNIZE PROLINE

217

FIG. 2. (A) Ribbon model of the Src-SH3 domain, in two different orientations related by a 908 rotation. On the right-hand side of the ®gure, residues belonging to the ligand-binding site are shown, together with the neighboring sequence, so that its location can be found in the alignment presented in (B). The ampersand above the alignment indicates a residue that is critical for ligand binding and orientation, whose function is discussed in the text. The asterisks indicate residues whose side chain is shown in (A). The ®gure was obtained with the program Ribbons (235).

trademarks of SH3 domain, the so-called ALYDY motif, is located in this b-hairpin-like loop. After hydrogen bonding with the bA strand with its ®rst part, the bB strand bends quite abruptly, thus taking part in the second b-sheet. The loops connecting strands bB and bC and strands bC and bD have been named the n-Src loop and the distal loop, respectively. After the latter loop, the SH3 polypeptide chain is characterized by some of its most conserved features, such as the WW dipeptide, situated in the bD strand, and the PXXY motif, located in the 310 helix connecting the bD and bE strands. The sequence in the bE strand tends to be poorly conserved. The original nomenclature for SH3 domains was speci®cally designed to describe the effects of mutations on Src activity (36), thus offering an immediate reference to the possible biological relevance of

218

ANDREA MUSACCHIO

certain segments of the domain. The RT loop received its name because mutations affecting residues R95 and T96 in the Src kinase SH3 domain are characterized by deregulated kinase activity and partial transformation ability. The name n-Src loop is justi®ed by the presence of insertions within this loop, which are found in the neuronal form of the Src tyrosine kinase and in other SH3 domains. The distal loop received its name because it occupies a position of the SH3 domain opposite to that of the SH3 ligand-binding site, as will be described below. The adoption of a nomenclature that bears no explicit reference to the topology is limiting. Throughout this chapter, we will refer to the SH3 loops (RT, n-Src, and distal) using the original nomenclature, but will add a reference to the strands and refer to these loops as RTAB , n-SrcBC , and distalCD loop. The SH3 fold has been observed in other protein domains with distinct function (relevant information can be found in reference 37). Recently, a group of bacterial proteins bearing signi®cant sequence similarity to the SH3 domain was discovered (38). Whether these prokaryotic domains share the SH3 fold and are involved in PPII recognition is currently unknown. The simplicity of the fold, the great wealth of structural information, and the ease of its biochemical handling have made the SH3 domain a very-well-characterized model system for the study of proteinfolding mechanisms (see for instance references 39±46), but these studies will not be reviewed here. A list of SH3 structures present in the protein database in July 2000 is presented in Table I. All these structures show virtually identical topological and structural features regardless of ligand binding and the presence of neighboring domains (47). Some exceptions, however, have been reported. The p85 subunit of phosphatidylinositol 3-kinase (PI3K) contains an SH3 domain whose native fold is similar to that of other SH3 domains (34, 35). The domain partially unfolds at low pH and slowly assembles into amyloid ®brils (48). The structure of amyloid ®bers Ê formed by PI3K-SH3 has been determined at a resolution of 25 A using cryo-EM and image processing techniques (49). X-Ray ®ber diffraction studies indicate that the proto®lament cores of b-amyloids contain a cross-b-scaffold and that the b-strands are arranged perpendicular to the ®ber axis, while b-sheets are parallel to it (50). In the case of the PI3KSH3 domain, the EM reconstruction shows a double helix of two proto®lament pairs wound around a hollow core, with a distribution of helical Ê and an axial subunit repeat of crossover repeats between 545 and 660 A Ê about 27 A. A tentative ®tting of the native state of the PI3K-SH3 domain into a single-particle 3D reconstruction of segments with similar helical crossovers fails to produce a convincing match. Instead, the native SH3 domain is expected to unfold to adopt a longer, thinner shape in the Ê proto®laments can accommodate amyloid form (48, 49). The 20 40 A

219

HOW SH3 DOMAINS RECOGNIZE PROLINE

TABLE I Experimentally Determined Three-Dimensional Structures of SH3 Domains SH3

pdb ID

Comment

Experimental technique Reference

Unliganded structures Abl

1ABQ

Unliganded

X-ray

Amphiphysin 2 1BB9

Unliganded

X-ray

86 99

Btk

1AWX

Unliganded

NMR

222

Csk

1CSK

Unliganded

X-ray

223

Eps8

1AQJ

Unliganded

X-ray

51

Fyn C-Grb2

1SHF 1GFC

Unliganded Unliganded

X-ray NMR

36 224

N-Grb2

1AZE

Unliganded (mutant)

NMR

88

Hck

1BU1

Unliganded

X-ray

121

Nebulin

1NEB

Unliganded

NMR

225

PI3K p85a

1PHT

Unliganded

X-ray

226

PI3K p85a

1PKS

Unliganded

NMR

34

PI3K p85a

1PNJ

Unliganded

NMR

35

PLCg a-Spectrin

1HSQ 1SHG

Unliganded Unliganded

NMR X-ray

33 32

a-Spectrin

1TUD

Unliganded (circular permutant)

NMR

227

a-Spectrin

1AEY

Unliganded

NMR

228

a-Spectrin

1BK2

Unliganded (mutant)

X-ray

229

Src

1SRL

Unliganded

NMR

31 230

Unliganded structures including other ``cis'' domains Ab1

1AWO

SH3±SH2 moiety

NMR

Ab1

2ABL

SH3±SH2 moiety

X-ray

231

Grb2

1GRI

Entire protein

X-ray

108

Hck

1AD5

Shows mechanism of intramolecular regulation of tyrosine kinases

X-ray

20

Lck

1LCK

SH3±SH2 moiety

X-ray

232

Src

1FMK

X-ray

21

Src

2PTK

Shows mechanism of intramolecular regulation of tyrosine kinases Shows mechanism of intramolecular regulation of tyrosine kinases

X-ray

22

Complexes with peptides, peptoids, or nonpeptide elements Ab1 1ABO Complexed with 3BP1 peptide

X-ray

86

Ab1

1BBZ

Complexed with p41 peptide

X-ray

106

Btk

1QLY

Complexed with Cb1 peptide

NMR

233 (continued)

220

ANDREA MUSACCHIO

TABLE IÐContinued SH3

pdb ID

Comment

Experimental technique

Reference

c-Crk

1B07

Complexed with peptoid inhibitor

X-ray

90

c-Crk

1CKA

Complexed with C3G peptide

X-ray

98

c-Crk

1CKB

Complexed with SOS peptide

X-ray

98

Fyn

1AZG

Complexed with PI3K peptide

NMR

15

Fyn

1NYF

Complexed with PI3K peptide

NMR

16

Fyn

1FYN

Complexed with 3BP2 peptide

X-ray

86

N-Grb2

1GBR

Complexed with SOS peptide

NMR

113

Hck

4HCK

Complexed with RasGAP peptide

NMR

234

Itk

1AWJ

Intramolecular complex

NMR

153

Sem5

1SEM

Complexed with Sos peptide

X-ray

85

Sem5

2SEM

Complexed with peptoid inhibitor

X-ray

90

Sem5

3SEM

X-ray

90

Src

1NLO

Complexed with peptoid inhibitor Complexed with nonpeptide elements

NMR

71

Src

1NLP

Complexed with nonpeptide elements

NMR

71

Src

1PRM

Complexed with PLR1 peptide

NMR

84

Src

1QWE

Complexed with peptide, showing interactions outside P-rich core

NMR

100

Src

1QWF

Complexed with peptide, showing interactions outside P-rich core

NMR

100

Src

1RLQ

Complexed with PLR2 peptide

NMR

84

Complexes involving globular, folded structures Fyn

1AVZ

Complex with Nef

X-ray

122

Fyn R!I 53BP2

1EFN 1YCS

Complex with Nef Complexed to the p53 protein

X-ray X-ray

101 102

HOW SH3 DOMAINS RECOGNIZE PROLINE

221

only one pair of ¯at b-sheets stacked against one another, with very little interstrand twist (49). Another exception is represented by Eps8-SH3. The crystal structure of this domain revealed an intertwined dimer, obtained by a ``strand exchange'' mechanism (51). The two anti-parallel b-sheets of the Eps8SH3 domain belong to different polypeptide chains. Residues that in the monomeric form of the SH3 domain correspond to the n-Src loop assume an extended conformation and function as a connecting segment to allow the exchange between monomers. These rearrangements, however, result in half-dimers that structurally resemble monomeric SH3 domains and can be superimposed with minimal root mean square deviation to monomeric domains. In solution, there appears to be an equilibrium between a monomeric and a dimeric form of the domain (52). Interestingly, the monomer, not the dimer, is the species active in peptide binding, indicating that dimerization is not the feature accounting for the different speci®city of this domain (52). More recently, the crystal structure of a monomeric form of the domain was also reported and found to be similar to that of monomeric SH3 domains (53). D. Ligand Speci®city of SH3 Domains: An Outline The ®rst evidence that SH3 domains recognize proline-rich ligands came from the work of Baltimore and collaborators, who identi®ed two proteins (3BP1 and 3BP2, for SH3-binding proteins 1 and 2) as ligands of the Abl-SH3 domain (54, 55). Studies with deletion mutants established that the SH3 ligand-binding site coincided with proline-rich stretches in 3BP1 and 3BP2. Similar sites were found in other proteins in the sequence database, including the formins (54). Over the years, there has been a remarkable accumulation of data on SH3 domain-binding speci®city. On the one hand, this search has relied on large-scale approaches based on phage display and synthetic peptide library technology (56±71). Parallel approaches focused on the identi®cation of biologically relevant SH3-mediated interaction, applying techniques such as the yeast twohybrid screen and an array of protein±protein interaction assays. A spectacular endeavor was the concomitant realization by several groups that the interaction of Grb2 with Sos is mediated by the Grb2-SH3 domains (72±80). A full account of the current list of reported SH3 ligands has become an almost unattainable task by a single human being (speci®cally, the author). A database of SH3 ligands has been recently reported (81). An accessory, partly overlapping, and by no means complete list of interactions is presented in Table II. Two methods for the prediction of proline-rich ligands of SH3 domain in the sequence

TABLE II SH3 Ligandsa SH3 Src

Ligand

Consensus

Class

KD (mM)

Reference

Phage library consensus

LXXRPLPXBP

I

63

Phage library consensus Phage library consensus

XXXRPLPPLPXP RPLPPLP

I I

68 64

Individual library peptide

RSSRPLPPIP

I

19.5

68

Individual library peptide

RPLPPLP

I

3.7

100 100

222

Individual library peptide

VSLARRPLPPLP

I

0.45

Individual library peptide

VPLARRPLPPLP

I

0.54

Individual peptide

APPLPPR

II

22.0

68, 100

Individual peptide

APPLPPRRNRPRL

II

1.2

68, 100

Hs Cdc42 GAP Hs hnRNP K

TAPKPMPPRPPL SRARNLPLPPPP

I I

Mm p62

TVTRGVPPPPTV

I

Hs PI3K

RPPRPLPVAPGS

I

Kv1.5

RPLPPLP RPLPSPP

I I

p130Cas Clone 2b

100

194 195 195 b

196 197 198

ARPLPPPPPA

I

199

Clone 10a

RPLPALP

I

199

Hs Dynamin Hs Dynamin

GGAPPVPSRPG GPPPQVPSRPN

II II

182 182

Hs Dynamin

RAPPGVPSRSG

II

182

Hs hnRNP K

PLPPPPPPRGG

II

195 (continued)

TABLE IIÐContinued SH3

223

Fyn, Lyn, Hck Hck

Ligand

Consensus

Class

KD (mM)

Reference

Hs PI3K p85

QPAPALPPKPP

II

196

Hs Shb

GGPPPGPGRRG

II

200

Hs Shb

TKSPPQPPRPD

II

200

Synapsin Clone 9h

GPAGPTR PPPPPPPLPPR

II II

201 199

Nef (NL4-3)

Full-length

II

Hs Paxillin

AVPPPVPPPPS

?

202

Hs AFAP-110

PPDNGPPPLPTS

?

203

Hs AFAP-110

PPQMPLPEIPQQ

?

203

Hs Shc p52

VRKQMLPPPPCP

?

195

D-peptide

RCLSGLRLGLVPCA

?

Btk Clone 10a

KKPLPPTPEE RPLPALP

I I

11.4

64.0

121

189 204 199

Nef (NL4-3; T to R)

PVRPQVPLRP

II

Nef (NL4-3; T to R)

Full-length

II

0.25

Nef (NL4-3)

Full-length

II

0.6

121

Nef (NL4-3)

D1

57

II

1.5

121

57

II

15.0

Fyn

Nef (NL4-3)

D1

Fyn (R!I)

Nef (NL4-3; T to R)

Full-length

II

Fyn Itk/Tsk

PI3K p85 Phage library consensus

PPRPLPVAPGSSKT GWYXKXPPPIP

I I

90.0

0.38 31.0

120 120

121 120 15 152 (continued)

TABLE IIÐContinued SH3 Cap C

PI3K p85

Ligand

Consensus

Class

Phage library consensus

XPXPPXRXSSL

II

Hs C-Cb1

PPPPPARHSLI

II

Hs Sos

GPPVPPRQSTS

II

Phage library consensus Phage library consensus

XXXRPLPPLPXP RXLPPRPXX

I I

KD (mM)

Reference 57

68 65

Individual library peptide (RLP1)

RKLPPRPSK

I

9.1

65

Individual library peptide

RPLPPLP

I

20.0

100

224

Individual library peptide

VSLARRPLPPLP

I

12.0

100

Individual library peptide

APPLPPR

II

24.0

100

Individual library peptide

APPLPPRRNRPRL RPLPSPP

II I

3.8

100 205

PPPVIAPRPEHTKSVYTR BXXRPLPXLP

? I

0.024

206 63

PIX Yes

p130CAS PAK Phage library consensus Hs Yap65

PVKQPPPLAPQS

?

Ab1

Phage library consensus

PPXZXPPPBP

I

63

Phage library consensus

PPPYPPPPBPXX

I

68

207

Library peptide

PPPYPPPPIP

I

2.0

68

3BP-1

APTMPPPLPP

I

34.0

14, 54

3BP-1

APTMPPPLPPVPP

I

66.0

68

3BP-1 3BP-2

APTMPPPLPP PPAYPPPPVP

I I

37.0 5.0

65 54

p40

APTYSPPPPP

I

0.4

87

p41

APSYSPPPPP

I

1.5

87 (continued)

TABLE IIÐContinued SH3

Ligand

Consensus

Class

KD (mM)

Reference

N-Methyl-D-aspartic acid receptor NR2D subunuit ATM

Proline-rich fragment

?

DPAPNPPHFP

I

209

ST5

LPPSPTPAAP

I

210

208

ST5

KPSNGLPPSP

I

210

ST5

PPLPSTPAPP

I

210 199

225

Clone 9h

PPPPPPPLPPR

I

Cortactin

Phage library consensus

PPBPXKPXWL

II

63

P53bp2

Phage library consensus

RPXBPBRSXP

II

63

Crk N

Phage library consensus Hs Ab1

XPBLPBK QAPELPTKTR

II II

63 211

Hs Ab1

VSPLLPRKER

II

Hs C3G

PPPALPPKKR

II

1.9

211 212

Hs C3G mutant

PPPALPPRKR

II

17.2

212

Hs C3G

TPPALPEKKR

II

115

Hs C3G

KPPPLPEKKN

II

115

Hs C3G

PPPALPPKQR

II

Sos Eps15

PPPVPPRRRR PPALPPKV

II II

115 5.2

117 212

Eps15R

TPALPPKK

II

Arg

LPILPSKTR

II

212 213

Arg

SPALPRKQR

II

213 (continued)

TABLE IIÐContinued SH3 Nck 2

Ligand

Consensus

Class

KD (mM)

Reference

aPAK

DKPPAPPMRNTST

II

173, 214, 215

bPAK

EKPPAPPLRMNSN

II

173

NIK

EVPPRVPVRTTSR

I

173

PTP-PEST Synaptojanin

DSPPPKPPRTRSC WTPPLPPPRSRSS

I I

173 173

ARG

RNAPTPPKRSSSF

I

173

WIP

N/D

?

216

PSTPIP

WASP

PPPPGRGGPPPPPPPAT

?

217

PLCg

Phage library consensus

XPPVPPRPXXTL

II

63

Dynamin

APPVPSRPGASP

II

182

226

Grb2 N

Dynamin

PPQVPSRPNRNR

II

182

c-Cbl Phage library consensus

LPPVPPRLDLLP ZDXPLPXLP

II I

218 63

Hs c-Cbl

PQRRPLPCTPGD

I

109

Hs c-Cbl

WLPRPIPKVPVS

I

109

Hs Dynamin

GGAPPVPSRPG

II

182

Hs Dynamin

GPPPQVPSRPN

II

182

Hs Dynamin

RAPPGVPSRSG

II

182

Hs Ab1

LQAPELPTKTR

II

211

Hs Ab1 Hs Ab1

AVSPLLPRKER KTAPTPPKRSS

II II

211 211

Hs c-Cbl

ASLPPVPPRLD

II

109

Hs SOS1 (1)

PVPPPVPPRRR

II

73 (continued)

TABLE IIÐContinued SH3

Ligand

Consensus

Class

KD (mM)

Reference

Hs SOS1 (1)

PVPPPVPPRRR

II

3.5

117

Hs SOS1 (1)

PVPPPVPPRRR

II

5.7

110

Hs SOS1 (2)

DSPPAIPPRQP

II

73

Hs SOS1 (3) Hs SOS1 (4)

ESPPLLPPREP IAGPPVPPRQS

II II

73 73

227

C3G

PPPALPPKKR

II

142.0

117

Grb2 C

Hs SOS1 (1)

PVPPPVPPRRR

II

39.0

110

N-WASP

PPPPPXR

II

Endophilin

Peptide library consensus

XPRPPXP

?

56

Amphiphysin

Peptide library consensus

B/ZXRPXP

?

p67phox C p130Cas

Dynamin p47phox p125FAK

VPSRPNR QPAVPPRP APPKPSR

II II II

56 59

a

B, aliphatic; Z, aromatic; , basic. af®nity of this interaction was reported in references 98 and 117. c This value refers to the af®nity of the SH3 domain for the entire proline-rich region of dynamin. bThe

219

0.19c b

220 221

228

ANDREA MUSACCHIO

database were presented recently (81, 82). On the other hand, several techniques have been used for the isolation of SH3 ligands of PPII helices. An interesting large-scale approach has been recently proposed by Cesareni and collaborators (83). A signi®cant fraction of the sequences reported in Table II were derived from screening chemical or phage-displayed peptide libraries, but several others were identi®ed through protein±protein interaction screens, and the signi®cance of the SH3 interaction in vivo was validated through appropriate biological assays. This effort has established beyond reasonable doubt that SH3 domains speci®cally recognize linear target sequences containing two XP moieties (where X can be any amino acidÐbut see restrictions belowÐand P is proline), with a scaffolding residue (often a proline) separating them. Thus, the most general statement concerning the speci®city of SH3 domains is that they recognize XP-X-XP motifs in target proteins (where the dashes are used just to emphasize that XP dipeptides are involved in binding, as explained in Section III, F). This motif is often referred to as the ``core'' of SH3 ligands. Flanking residues are essential for speci®c and selective recognition. The ®rst three-dimensional structures of complexes between SH3 domains and proline-rich peptides (65, 84±86) clari®ed that the core region (containing the XPXXP motif) of ligand peptides adopts a lefthanded PPII conformation. Because this conformation is precisely characterized by three residues per helical turn, each P in the XP motifs faces the same edge of the helix, precisely at the distance of one turn (very Ê ); so would the Xs preceding the P in the XP motif, but they close to 9.1 A will be located on the neighboring edge of the PPII helix. The structures showed that these XP moieties occupy the two edges of the PPII helix that make contact with the SH3 domain, as diagrammed in Fig. 3. The residue X between the two XP motifs occupies the edge that points away from the SH3 surface; the presence of proline at this site in several ligands identi®ed by phage display and other techniques suggests that this residue may be important in stabilizing the structure of the PPII helix. However, the introduction of leucine in the 3BP-1 peptide has been reported to increase its af®nity for the Abl-SH3 domain (87). Two seminal papers (84, 85) also established that SH3 domains utilize one or more residues outside the proline-rich core to position the peptide in one of two possible orientations relative to the SH3 surface, named plus and minus (ligands conforming to one or the other binding mode were classi®ed as class I and class II, respectively). For instance, peptide library screenings show that the Src-SH3 domain selects peptides containing the consensus sequences R-X-XP-X-XP (class I) or XP-X-XP-X-R (84).

HOW SH3 DOMAINS RECOGNIZE PROLINE

229

FIG. 3. Schematic view of class I and class II (top and bottom) peptides binding to the SH3 domain. An inversion of the direction of the polypeptide chain results in the repositioning of the proline residues relative to the SH3 surface. The ``compass'' residue (R in this example) occupies a similar position on the SH3 surface regardless of peptide orientation. The ®gure was reproduced from reference (24).

The arginine residue within these sequences establishes orientationindependent interactions with the same residues in the RT loop of the SH3 domain, thus determining the orientation of the peptide. Residues other than arginine play a similar role in other SH3-substrate interactions. In summary, the minimal binding element of canonical SH3 ligands is a XP-X-XP core region followed by residues located either upstream (class I) or downstream (class II) from this proline-rich core and acting as determinants of peptide orientation. The latter residues are herewith de®ned as ``compass residues.''

230

ANDREA MUSACCHIO

E. The SH3 Ligand-Binding Site The region of the SH3 domain involved in the interaction with the proline-rich core of target sequences is a surface patch formed by the side chains of a few well-conserved residues scattered along the primary structure (31, 32, 65, 86). Critical to the composition of this site are the residues indicated in the alignment presented in Fig. 2. The most common trademarks are the two aromatic residues in the ALYDY motif, the ®rst W of the characteristic WW dipeptide motif, and the 310 helix formed by the PXXY motif. As shown in Fig. 2, these residues are aligned to form a surface patch, quite hydrophobic in character, in which the aromatic side chains are stacked against each other. Being formed by the most highly conserved residues in the family, only subtle differences emerge from the comparison of the ligand binding sites of different SH3 domains. However, the site is ¯anked by the RTAB and the n-SrcBC loops, which are variable in sequence and impart a good degree of selectivity to SH3-mediated interactions (see Section III,G). Residues in these loops are probably better targets for mutational analysis than the conserved residues in the SH3 ligand-binding site. The interaction of Grb2 with Sos is disrupted by mutations of the conserved P residue in the PXXY motif. Vidal and collaborators have found that the introduction of this mutation in the Grb2-N-SH3 domain determines its unfolding (88). Studies with a truncated form of the domain have also revealed the inability to reach the folded state (89). A surface representation of the SH3 domain shows a rather shallow surface, with the most prominent feature being the side chain of the conserved W, oriented with the indole nitrogen in position to donate a hydrogen bond to a carbonyl group on the substrate peptide (Fig. 4). The hydroxyl group of the tyrosine residue in the C-terminal PXXY motif is also surface-exposed and in a position to donate a hydrogen bond to a peptide carbonyl. The SH3-binding site is characterized by the presence of two XP-binding pockets and the neighboring compass pocket, a valley surrounded by the RT and n-Src loops. Some degree of geometric complementarity is immediately evident from the view presented in Fig. 4. The two XP-binding pockets are located at a distance compatible with the arrangement of two such motifs in PPII ligands. Each of the two XPbinding pockets may host the XP dipeptide in opposite orientations, a property deriving from the pseudo-symmetry of these motifs in PPII helices. It is convenient to describe the ligand-binding site on the SH3 domain as being formed by ``slots,'' rather than pockets. The P0 , and P3 slots face the RT loop (AB), while the P 1 and the P2 slots face the n-Src loop. The P 1 , P0 , P2 , and P3 slots are involved in binding the two XP moieties of the substrate in either orientation. The P 2 and P1 slots are

HOW SH3 DOMAINS RECOGNIZE PROLINE

231

FIG. 4. (Top) Ab1-SH3 bound to the 3BP1 peptide, a class I ligand. The essential proline residues are located on slots 0 and 3 (shown schematically in the bottom part of the ®gure), whereas residues occupying the 1 and 2 slots can be changed into alanine without substantial alterations of the binding af®nity (54). The compass pocket hosts the side chain of proline (at position 2 of the peptide) and of methionine (at position 4). The ®gure was created with the program GRASP (236).

purely ideal: they correspond to peptide residues that point away from the SH3-binding site. Numbering refers explicitly to regions of the SH3 ligand-binding site, and each slot is colinear with a position in the peptide ligand (Fig. 4). This is convenient, as it provides a common reference for peptides that bind in opposite orientations. The colinearity of the SH3 ligand-binding site and the ligand is perfect for the regions of the peptide that adopt a PPII helix. Things become more complicated for the regions that do not adopt a PPII conformation and that are generally found to interact with residues in the compass pocket between the RT and n-Src loops. The name P 3 for this pocket re¯ects the original observation that positively charged residues three positions upstream from the XPXXP motif (class I) are important for orientation. However, there are SH3 domains, such as that from Abl, for which this scheme becomes unsatisfactory, as this domain seems to require the presence of a proline residue

232

ANDREA MUSACCHIO

®ve residues upstream from the ®rst proline in the XPXXP motif, rather than that of a charged residue three positions upstream. Thus, we ®nd the de®nition of compass pocket more satisfactory. F. Lim's Model of SH3±Peptide Interactions From the arguments presented above, it follows that the observed speci®city of SH3 domains toward regularly spaced prolines must arise from an orientation-dependent property of proline-containing PPII ligands. What determines the absolute requirements for proline at one of the two edges of the triangular prism contacting the SH3 domain? Lim and collaborators proposed a convincing answer to this question (85, 90). To explain their proposal (henceforth, Lim's model), we start by displaying a cross section of a polyalanine chain in a PPII helix conformation (Fig. 5A). As we proceed from the N- to the C-terminal end for one turn of the helix, we distinguish three cases: (a) The residue packs against a slot with the Ca atom directed toward the protein surface relative to the neighboring main chain nitrogen. This situation is identi®ed as internal packing. (b) The residue packs against a pocket with the a-carbon oriented away from the protein surface relative to the neighboring main chain nitrogen. This mode of interaction has been de®ned external packing. (c) The residue does not make contact with the SH3 surface, as its side chain points away from it. These three modes account for the interactions of a full turn of the PPII helix with its SH3 substrate. Moving from N to C, the modes of interaction follow one another in the order: internal packing, external packing, no packing. It goes without saying that this is an intrinsic property of PPII helices, which remains true irrespective of the orientation of the helix. According to this design, when an SH3 target is bound to the SH3 domain, the X position in the XP motif coincides with sites of internal packing, while P residues occur at the edge of the prism that established an external packing interaction with the SH3 surface. Why is proline preferred over other residues at sites of external packing? Figure 5B represents the same view as in the previous ®gure, but new side chains have now been introduced to complete the argument. Figure 5B shows that the N-alkyl substitution characteristic of proline accounts for its selection at sites of external packing. The Ca and N atoms in the proline ring are almost structurally equivalent. If a nonproline residue is introduced in the helix, the alpha carbon and backbone nitrogen are no longer structurally equivalent. A nonalanine side chain () adopting the favorable x1 torsion angles of 608 or 1808 will be oriented into the binding pocket at an internal packing site, allowing it to make the same, or better, van der Waals and hydrophobic interactions as a proline. At an

HOW SH3 DOMAINS RECOGNIZE PROLINE

233

FIG. 5. Properties of the PPII helix. The main chain of a polyalanine (AAA) chain in the PPII conformation is shown in (A). Going from N to C, we encounter an internal packing site, which is followed by an external packing site. The following side chain is directed away from the SH3 surface (see text for a de®nition of internal and external). The arrows represent the direction of the N to Ca vectors. In (B), (C), and (D) different side chains are added. The ®gure clari®es that the presence of proline is required at external packing sites to maximize the interaction with the SH3 surface.

external packing site, however, a nonproline side chain will be oriented away from the protein surface and will be unable to ®ll the binding pocket. A large decrease in stability will occur on substitution of proline at an external packing site, as only proline, with its N-alkyl substitution, can mimic at external sites the packing interactions made by nonproline side chains at an internal site. As the terms ``internal'' and ``external'' refer to speci®c positions along a PPII helix oriented against the SH3 surface, it is clear that an inversion of the chain direction will result in the reorientation of the external and internal sites relative to the SH3 surface. If a peptide belongs to class I,

234

ANDREA MUSACCHIO

proline is absolutely required at the P0 slot. After a chain reversal (a class II peptide), the preference at the same slot will often address a different residue, generally a hydrophobic residue, but often also positively charged residues, and sometimes even glycine (see Table II). Conversely, the preference at the P 1 slot will go to proline when the peptide is classi®ed as class II, but it will often go to a different residue in the class I option. Why is the preference of the SH3 domain for proline at external packing sites maintained regardless of the particular orientation of the peptide? This property clearly indicates some degree of pseudosymmetry within the SH3 ligand-binding site. In our description of Lim's model, we have so far assumed that both the ligand- and the SH3-binding site can be effectively regarded as being two-fold symmetric. Indeed, pseudo-twofold symmetry in the SH3 ligand-binding site can be regarded as a by-product of the binding of a roughly two-fold symmetric substrate composed of XP motifs in a PPII conformation. The assumption that SH3-binding sites effectively behave as twofold symmetric may not be completely accurate. Certain SH3 domains, such as those of Src or PI3K, are capable of binding peptide ligands belonging to both orientation classes with similar af®nity (65, 84), but the actual degree of internal symmetry may vary from domain to domain, according to the particular sequence surrounding the SH3 ligand-binding site. Certain SH3 domains may display a preference for one particular orientation (AblSH3, for instance). We suspect that some SH3 domains might have engineered an escape from this ``built-in'' two-fold symmetry of the ligand-binding site by establishing speci®c interactions with the side chains of X residues in the XP motifs, in particular with side chain atoms other than Cb , which cannot be mimicked by proline. In summary, Lim's model states that the essential feature recognized by SH3 domains (and WW domains) is an irregular backbone substitution pattern made of N-substituted residues placed at key positions along an otherwise normal Ca-substituted peptide scaffold. The implication is that the key feature to explain the presence of proline at external binding sites is not its unusual side chain. Rather, what is essential is N-alkyl substitution. Lim and collaborators demonstrated this point experimentally, by showing that substitution of proline with sarcosine (N-methyl glycine) at external packing sites results in peptides with similar or enhanced af®nity for the SH3 domain (90; reviewed in reference 91). Simulation studies arrived at similar conclusions (92). Thus, proline is required at external binding sites because it is the only naturally available N-substituted residue. XP motifs are recognized as essentially symmetric units by the SH3 ligand-binding site. Using this mode of recognition, SH3 domains can achieve a high level of discrimination against nonproline sequences while maintaining relatively low binding af®nities, which may be advantageous

HOW SH3 DOMAINS RECOGNIZE PROLINE

235

for the biological function of this recognition module. Lim's model also applies to the binding of WW domains to proline-rich ligands, but may not be seen as a general model for the interaction of all proline-rich regions with target proteins. Lock-and-key mechanisms based on the speci®c recognition of the proline side chain may play a role in the interaction of such regions with other proteins or protein domains (90, 91, 93±97). G. Detailed Analysis of Selected SH3±Ligand Interactions Most SH3 ligands are characterized by the presence of the XP-X-XP sequence and of compass residues, and Lim's model provides an explanation of this observed speci®city. In addition to determining peptide orientation, residues ¯anking the XP-X-XP core can play a role in determining the selectivity of proline-rich regions for SH3 domains. For instance, the c-Crk N-terminal SH3 (N-SH3) domain binds in a highly selective manner to a region of the C3G protein characterized by the presence of an XP-X-XP-X-K motif (98). In this class II peptide, lysine acts as an orienting residue rather than the more usual arginine, and the interaction confers exquisite speci®city toward the very acidic RT loop of Crk-N-SH3. The n-Src loop may also contribute signi®cantly to the creation of selectivity in SH3-mediated interactions (see, for instance, references 56, 99). Large-scale peptide searches combined with ELISAs have con®rmed that selectivity of SH3 interactions may be the rule rather than the exception (see, for instance, reference 63). In addition to the problem of how selectivity of SH3 binding is generated, a second important issue is represented by the fact that the interactions of SH3 domains with the minimal binding element are usually low af®nity (see Table II). The problem of whether the information contained within the minimal binding site fully accounts for the interaction potential of SH3 domains has been addressed (see, for instance, references 63, 67, 100). By screening peptide libraries containing the minimal binding element within extended ¯anking regions, the importance of regions beyond the minimal binding element in SH3 recognition to achieve high-af®nity binding has been recognized. For instance, the peptides VSLARRPLPPLP (class I) and APPLPPRNRPRL (class II) bind to the Src-SH3 domain with KD of 450 nM and 1.2 mM, respectively (100). In comparison, the af®nity of the APPLPPR minimal class II-binding element for the Src SH3 domain is 22 mM (84). The determination of the solution structures of SH3 complexes with these extended peptides revealed the presence of signi®cant additional interactions with the SH3binding site (100). In particular, the ¯anking regions bind in the compass pocket, which among the SH3-binding pockets is the one showing the least

236

ANDREA MUSACCHIO

degree of sequence conservation. Thus, additional contacts with the compass pocket may increase the selectivity and af®nity of the SH3-ligand interaction. The importance of residues outside the minimal binding element was also observed in the case of the interaction of the HIV protein Nef with the SH3 domain of the Hck Src family kinase, another highaf®nity SH3-ligand interaction (101). In this case, the relevant residues are located on an exposed hydrophobic surface patch, which hosts the side chain of an SH3 residue located in the RT loop (as detailed in Section III, H). The interaction with p53 requires the cooperation of the SH3 domain and ankyrin repeats of 53BP2 and does not necessitate a proline-rich target on p53, but rather an exposed patch comprising the L2 and L3 loops (102). This example is interesting, as it suggests that the identi®cation of certain SH3 targets might have been hindered by screening approaches designed with isolated SH3 domains, rather than by the proper combination of cooperating domains. In this respect, it is striking that the 53BP2-SH3 domain detects proline-rich peptides in peptide library screens when used in isolation (63). Another interesting example of domain cooperation is offered by the interaction of the Btk TH and SH3 domains to the alpha subunit of the trimeric GTP-binding protein Gq (103). What is the fraction of high-af®nity interactions of SH3 domains with non-proline-rich sequences? The available literature indicates that the application of ``quasi-whole protein'' methods, such as the yeast twohybrid screen, very often resulted in the isolation of proline-rich targets. However, the possibility cannot be ruled out that more often than not target validation was carried out only for the ``common wisdom'' isolates, namely, those containing proline-rich sequences. Thus, the possibility cannot be ruled out that the number of interactions not mediated by proline-rich regions may be currently underestimated. We start our review of SH3-mediated interactions by presenting structures of SH3 complexes with short peptide ligands in class I and class II orientations. These interactions are low af®nity, with KD usually between Ê 2 on 0.4 and 50 mM and buried interaction surfaces on the order of 400 A each of the binding partners. 1. Class I Interactions a. Src-SH3 and p85-SH3. Concerning to substrate orientation, Src± SH3 is a good example of a promiscuous domain. Peptide library screens using this domain were crucial for the discovery that SH3 domains are capable of binding ligands in opposite orientations (31, 84). Src±SH3 appears to be an extremely versatile binding module, as shown by the list of interacting peptides reported in Table II. This promiscuity may re¯ect multiple activation mechanisms for this kinase. Similarly promiscuous habits have been uncovered for the SH3 domain of the p85 subunit

HOW SH3 DOMAINS RECOGNIZE PROLINE

237

of PI3K, a lipid kinase (65, 68, 100). Also in this case, it appears that peptides of two different classes can be isolated from library screens. The solution structure of the complex of the p85-SH3 domain with the RLP1 class I peptide (RKLPPRPSK) (65) was the ®rst structure of a SH3peptide complex ever to be described (although the atomic coordinates have never been deposited in the structure data bank). In this structure, the XPXXP moiety of the peptide folded in the PPII helix conformation as explained in Section III,C, the proline side chains interacted with the P0 and P3 slots, and their conversion into alanine resulted in a signi®cant reduction in binding af®nity. The structure also showed that the selection of arginine on the peptide substrate was the consequence of a speci®c interaction with a conserved aspartic acid in the RT loop. Mutation of this residue into asparagine resulted in a 50-fold reduction in binding af®nity, while the mutation of other negatively charged residues in the surroundings left the af®nity unaffected (65). Precisely the same interactions were observed in the complex of the Src±SH3 domain with a class I peptide, which was also isolated through a peptide library approach (84) (Fig. 6B). Rickles and collaborators showed that the Src and p85 SH3 domains display a relatively weak selectivity for residues ¯anking the core region of the peptide (68). The interaction with elements preceding (class I) or following (class II) the core results in a remarkable increase in the binding af®nity of peptides for these domains (see Table II). Structural analyses revealed that the ¯anking regions deeply insert into the compass-and-speci®city pocket, where they establish extensive van der Waals contacts, thus increasing considerably the SH3 surface buried on peptide binding (100). It has been suggested that class I and II peptides may bind with different selectivity to the Src-SH3 domain (104). Class I peptides isolated from library screens bind to the Src-SH3 domain selectively and cannot recognize other domains with signi®cant af®nity. Instead, class II peptides isolated in the same screen also bound several other SH3 domains and could be used for the systematic isolation of SH3 domains (104). b. Abl±SH3. Abl, the product of the ABL1 protooncogene, encodes a cytoplasmic and nuclear protein implicated in processes of cell differentiation, cell division, cell adhesion, and stress response (105). The ligand speci®city of the Abl-SH3 domain was the ®rst to be reported (54). Although there is a clearly discernible XPXXP motif in the sequences recognized by Abl-SH3 in target proteins, there is a clear lack of charged residues (Table II). This has been con®rmed by peptide library screening (63, 68). Thus, Abl-SH3 does not use the same code used by other SH3 domains to orient the peptide with respect to the SH3 scaffold, and indeed this domain does not signi®cantly interact with charged peptides

238

ANDREA MUSACCHIO

FIG. 6. Ribbon diagram and surface representation of different SH3 peptide complexes. In (A) Ab1-SH3 is bound to the p41 peptide described in the text. This is a class I interaction. In (B) and (C) the Grb2±Sos and Crk±C3G interactions are depicted, respectively. The sequence of SH3 residues involved in the interaction can be gathered from the comparison of this illustration with Fig. 2.

HOW SH3 DOMAINS RECOGNIZE PROLINE

239

(14). Two high-resolution crystal structures of complexes with cognate (3BP1) and optimized peptide ligands have revealed the structural basis for the peculiar speci®city of this domain (86, 106). The crystal structure of the Abl-SH3-3BP1 peptide (APTMPPPLPP) complex was determined Ê resolution by X-ray crystallography (86). The peptide sits on the at 2.0-A SH3 surface in the class I orientation, with the N-terminus near the compass pocket (Fig. 4A). The bulky side chain of methionine is in a position similar to that of arginine in the Src class I complexes revealed by NMR and is in van der Waals contact with residues in the RT loop. Indeed, it can be seen in Fig. 2 that the sequence features of the AblSH3 RT loop are different from those of other SH3 domains in that a threonine is present instead of the more common Asp or Glu. The presence of acidic residues at this position of the alignment correlates with the ability of SH3 domains to interact with positively charged compass residues. Indeed, substitution of threonine with aspartate in the Abl domain results in a mutated Abl-SH3 domain with binding properties similar to those of Src (107), indicating that this position plays a key role in determining the peculiar selectivity of Abl-SH3. Another striking feature of this complex is the deep interaction of the residues Ala-1 and Pro-2 in the compass pocket. The presence of Pro at position 2 of the peptide appears to be an absolute requirement for the interaction with the Abl-SH3 domain (Table II). Indeed, this was the ®rst visualization of the extensive role played by the compass pocket in peptide recognition. Furthermore, the structure revealed the hydrogen-bonding pattern described in Section III, E, which may be common to the vast majority of SH3-peptide interactions. The structure of the 3BP1 complex inspired the design of high-af®nity ligands for Abl-SH3 (87, 106). Two peptides, p40 and p41, showed enhanced af®nity and selectivity for the Abl-SH3 domain compared to the original 3BP-1 and 3BP-2 peptide, which showed instead relatively poor selectivity (14). The sequences of these peptides (APTYSPPPPP and APSYSPPPPP, respectively) are very similar to that of the 3BP-1 peptide, but the presence of tyrosine at the position occupied by methionine in 3BP-1 plays a pivotal role in increasing the af®nity and selectivity of the interaction. The structure of the complex of Abl-SH3 with the p41 peptide (Fig. 6A) revealed that the hydroxyl oxygen of the tyrosine is involved in the formation of two hydrogen bonds with the side chains of Ser-12 and Asp-14 in the RT loop (106). Due to the presence of relatively small side chains, and especially of glycine at position 13, there is a large surface pocket in the region of the Abl-SH3 domain (the pocket that accommodates Met or Tyr from the peptide). Conversely, this pocket is not present in other SH3 domains, such as that of Fyn, explaining why the selectivity of the p40 and p41 peptides toward Abl is increased by the

240

ANDREA MUSACCHIO

introduction of tyrosine in place of methionine, which has a smaller side chain. The Abl-SH3-p41 peptide complex also shows how the peptide side chain of Ser-5 forms an intramolecular hydrogen bond with the main chain carbonyl group of Pro-6. Intramolecular hydrogen-bonding interactions between side chains and the main chain occur relatively commonly in PPII helices (7), with the Gln side chain being the most favored one in forming this type of interaction. To our knowledge, the example of an intramolecular hydrogen bond in PPII involving the serine side chain is unique. The introduction of serine in place of proline at position 5 results in a threefold increase in af®nity for Abl-SH3 and a decrease of equivalent magnitude in the af®nity toward Fyn-SH3 (87). Comparison of the 3BP1 and p41 peptides shows that the formation of the intramolecular hydrogen bond results in the introduction of a deviation from ideal PPII main chain dihedral angles. In the case of Abl-SH3, the stabilization of this main chain conformation favors the optimal ®t of the side chain of Tyr-4 against the RT loop. In Fyn-SH3, the same main chain conformation would result in a severe clash of the tyrosine side chain with the RT loop. The reason that the p40 and p41 peptides bind to Abl-SH3 with different af®nities is not clear. The only difference between these peptides is the presence of Ser in p41 at a position in which the p40 and 3BP-1 peptides contain a threonine residue. The side chains of serine and threonine at this position point away from the SH3 ligand-binding site and do not make contact with SH3 groups. Thus, the different af®nities of these peptides cannot be simply explained by differential interactions with the SH3 domain. Rather, they may be due to small differences in the solvation of the peptide or in electrostatic interactions. In summary, the special selectivity of the AblSH3 domain can be explained by the presence of a threonine for an acidic residue substitution in the RT loop and by the extensive interaction of a proline residue at position 2 of the peptide with the compass pocket. For reasons that are not easily understood, the domain appears to be selective for peptides in the class I orientation. 2. Class II Interactions a. Grb2-SH3. Grb, mammalian homologue of the C. elegans Sem5 protein, is a typical adapter protein containing two SH3 domains, separated by an SH2 domain, and no catalytic moieties. The crystal structure of Grb2 showed that there are no contacts between the SH2 domain and either SH3 domain and that the interface between the SH3 domains is less than ideal to ensure a stable intramolecular interaction, suggesting the possibility that the relative position of the domains may be variable (108). The binding sites of each of the domains in the assembly shown by the crystal structure are exposed and available for interaction with target proteins.

HOW SH3 DOMAINS RECOGNIZE PROLINE

241

The Grb2-SH3 domains bind the C-terminal PRR of the Sos protein (residues 1000±1333). A detailed analysis of the interaction of Grb2 showed that there are six XP-X-XP motifs in this region of Sos (70). Four of these (Sos2, Sos3, Sos4, and Sos6) conform to class II peptides, and the N-SH3 domain of Grb2 readily binds to these sequences. The same domain is unable to bind to the Sos1 and Sos5 sequences. These sequences do not conform to arginine-oriented class II ligands, showing that the N-SH3 domain may recognize the orientation of substrates selectively. However, the interaction of Grb2 with c-Cbl, which utilizes class I peptides, seems to question this statement (109). Under the conditions used by Simon and Schreiber, the C-SH3 domain did not interact with any of the peptides (70). The KD of the interaction of Grb2-N-SH3 with the Sos2 peptide is 5.7 mM, while C-SH3 binds with lower af®nity, 39 mM (110). An independent estimate suggested a KD of around 25 nM, but the use of full-length protein in this experiment suggests that this particularly low value may re¯ect the concomitant binding of the C-terminal SH3 domain to the peptide immobilized on the BioSensor chip (78). There are several structures of the Sos2 peptide bound to the N- or C-SH3 domain of Grb2 (85, 111±113). In all these cases, the peptide is bound in the class II orientation. Detailed analysis of the complex of the Sem5-N-SH3 domain with the PPPVPPRRR Sos-derived peptide, the ®rst structure of a class II complex to be determined, led Lim and collaborators to propose the general model for the interaction of SH3 domains with proline-rich regions described in Section III, F (85). This crystallographic complex shows that the side chain of the compass residue (arginine) forms a salt bridge with an acidic residue (Glu) in the RT loop (Fig. 6B). This interaction is equivalent to that involving the compass residue of class I peptides on binding to PI3K- and Src-SH3. The same ®nding was described for the class II structure of the Src-SH3 domain (84). Thus, the positively charged compass residues in class I and class II peptides interact with the same residue on the SH3 domain. b. Crk-SH3. The product of the cellular form of the Crk protooncogene (c-Crk) consists of an SH2 domain followed by two divergent SH3 domains (for a recent review, see references 114). Crk interacts with C3G, a guanine-nucleotide exchange factor, using its N-SH3 domain (115, 116). The C3G sequence targeted by the Crk N-SH3 domain (PPPALPPKKR) conforms to the requirements for a class II peptide (98). There are two XP motifs (in bold face type), followed by a positively charged compass residue. The most intriguing aspect of this sequence is the presence of lysine rather than the more common arginine as a compass residue. The C3G peptide binds poorly to the N-terminal SH3

242

ANDREA MUSACCHIO

domain of Grb2 (KD 142 mM), while the interaction with Crk has high af®nity (KD 1:9 mM). In contrast, a class II peptide with a sequence derived from Sos (PPPVPPRRRR), and containing arginine instead of lysine at the equivalent position, binds the N-SH3 domains of Crk and Grb2 with similar af®nities (KD 5:2 and 3.5 mM, respectively). Substitution of lysine with arginine in the C3G peptide leads to a nine-fold reduction in the af®nity for Crk-N-SH3 (KD of 17.2 mM) and increased af®nity for the N-SH3 of Grb2 (KD of 27.9 mM) (98). Thus, the presence of lysine confers exquisite selectivity to the interaction of Crk with C3G, while the presence of arginine at the same position results in promiscuous binding habits. Interestingly, amino acid substitutions in the viral form of Crk (v-Crk) render the domain insensitive to the presence of arginine or lysine at this site (98, 117). The crystal structure of the complex between c-Crk N-terminal SH3 and the C3G peptide (Fig. 6C) revealed the structural basis of the role played by lysine in this complex (98). There are four acidic residues in the Crk-N-SH3 domain RT loop. Three of these (D147, E149, and D150) approach one another closely and present three carboxylate oxygen atoms that form a nearly equilateral triangle, Ê with the lysine nitrogen at the geometric center and at a distance of 1.0 A from the plane of the oxygen atoms. The binding is thus remarkably well suited to interaction with lysine, since each of the three protons in the sp3 hybridized amino group can be donated to oxygen atoms. Kuriyan and collaborators showed that this represented a rare example in the protein database. The structure of the Sos peptide bound to the same N-Crk-SH3 domain explains why arginine is less suitable than lysine at this position (98). In this structure, the arginine side chain assumes a less extended conformation if compared to that from lysine. The side chain of E149 is not engaged in the interaction. The key residues are D150 and, to a lesser extent, D147. In fact, E149 is replaced by glycine in the v-Crk-SH3 domain. The latter domain binds the C3G peptide with KD 25:5 mM, a substantially larger value with respect to that of the c-Crk domain, and binds to mutant peptide containing arginine instead of lysine with similar af®nity, demonstrating a concomitant loss of discrimination. In summary, the presence of three acidic residues in the RT loop allows for the formation of a lysine-speci®c interaction that is unique to c-Crk and its close relatives, explaining the selectivity of this domain. Arginine can form two or three hydrogen bonds with only one or two acidic residues in the RT loop. This is not true for lysine, which needs three carboxylate groups to fully satisfy its hydrogen-bonding potential. As shown in Table II, all interactions typical of the c-Crk-N-SH3 domain are characterized by the presence of lysine downstream from the conserved proline-rich core. Thus, it appears that this domain selects class II ligands.

HOW SH3 DOMAINS RECOGNIZE PROLINE

243

H. The Interaction of HIV Nef with SH3 Domains The ability of Nef to bind SH3 domains is necessary for the optimal spread of HIV-1 infection in primary cell cultures, suggesting that the virus has evolved to exploit SH3-mediated cellular processes to enhance its replication (118, 119). Nef binds to SH3 domains in a selective manner. The Hck- and Lyn-SH3 domains are recognized ef®ciently, but not those of Fyn or other less related domains (118, 120). The KD for the association of Nef and Hck-SH3 is 250 nM, as measured by Surface plasmon resonance, or 188 nM, as measured by isothermal titration calorimetry (ITC). In contrast, the af®nity of Nef for Fyn-SH3 is characterized by a KD larger than 20 mM. Other measurements using ITC reported a KD of 600 nM for the association of the Hck-SH3 domain with full-length Nef from the NL43 strain and of 11.4 mM for the interaction of the latter with Src-SH3 (121). A deletion mutant of Nef lacking the ®rst 57 amino acids interacted with Hck-SH3 with a KD of 1.5 mM (121), suggesting that the N-terminal region of Nef may contribute additional stability to the complex. The discrepancy in the af®nity values reported above might derive from a threonine to arginine mutation that was introduced to mimic the sequence of most nef alleles obtained directly from patients (120). The affected residue, at position 71, represents the ®rst X in the XPXXP motif of Nef and thus interacts directly with the SH3 domain. The Nef SH3 region involved in SH3 binding (PVRPQVPLRP) conforms to a class II SH3-binding motif. Substitution of the second and third prolines in this motif (in boldface type) results in the loss of binding (120). A very interesting observation is that the KD for the interaction of an isolated synthetic peptide encompassing the Nef proline-rich sequence with the Hck-SH3 domain was estimated to be 90 mM, roughly 300 times weaker than that obtained with the full-length protein (120). Furthermore, comparison of this value with that obtained for the interaction of Fyn-SH3 with the same peptide (KD 202 mM) indicates a dramatic loss in selectivity (120). Thus, interactions measured with full-length proteins reveal elements of speci®city and selectivity that may not be reproduced by the interaction of SH3 domains with synthetic peptides. Studies aiming at understanding the basis of the differential af®nity of the Hck- and Fyn-SH3 domains for Nef concentrated on the possible role of the RT loop in which many sequence differences between Src family members reside. A single amino acid substitution (R ! I)in the RT loop of the Fyn-SH3 domain results in a KD of 380 nM, an af®nity even larger than that measured with constructs in which the entire Hck-SH3 RT loop had been grafted onto Fyn-SH3 (120). The crystal structures of intact Nef complexes with a Fyn-SH3 R!I mutant (101) and with wild-type Fyn SH3 (122) provided the framework to rationalize these results (for a

244

ANDREA MUSACCHIO

FIG. 7. The interaction of Nef with the SH3 domain (A) and a close-up view of the site of interaction (B). The side chain of an isoleucine residue in the RT loop important for binding has been omitted to improve the clarity of the diagram. This residue binds in an exposed hydrophobic groove formed by the two helices of Nef that contact the SH3 domain.

commentary, see references 123). The complex reported by Kuriyan and collaborators (Fig. 7) comprises a protease-resilient fragment of Nef (amino acids 58±203) (101). As expected from its agreement with class II sequences, the Nef PPII helix interacts with the SH3 domain in the minus orientation, and the interface with the SH3 domain is strikingly similar to that seen in the structures of similarly oriented peptides bound to the SH3 domains of Sem5 and Crk (described above). The selectivity of the interaction of Nef with the Hck-SH3 domain is based on the interaction that the RT loop of the SH3 domain can form by extending over the surface of Nef. The RT loop is a remarkably variable and ¯exible structure, which in Src kinases can be differentially stabilized by networks of hydrogen-bonding interactions, according to its particular sequence (121). These differences may at least in part account for the selectivity of the Nef±SH3 interaction by stabilizing a particular conformation of the loop necessary for an optimal interaction with Nef. The character of this interaction is hydrophobic: the side chain of Ile-96 of the Fyn-SH3 R!I mutant inserts into an exposed, remarkably hydrophobic crevice between the aA and aB helices of Nef. A correlation exists between the character of residues contributing to this speci®city pocket and the ability of HIV-1, HIV-2, or SIV to interact with different Src family SH3 domains (124). The importance of this interaction is con®rmed by the selection of RT-loop mutants of the Hck-SH3 domain that bind to Nef with af®nities up to 40-fold higher than those of parental

HOW SH3 DOMAINS RECOGNIZE PROLINE

245

Hck-SH3 (125). However, structure determination of the complex between wild type Fyn-SH3 and Nef (122) revealed that the aliphatic moiety of the arginine side chain accommodates neatly within the same crevice. This led to propose that the loss in binding free energy observed for the wild-type domain may be explained by partial desolvation and entropic penalties due to the rearrangement of the arginine side chain (122). These authors pointed out that Nef binding triggers a critical rearrangement of the hydrogen-bonding pattern that characterizes the unliganded Fyn-SH3 RT loop (36), allowing an induced ®t of the RT loop onto the Nef surface (122). Critical to this rearrangement is the formation of an ion pair between Asp-100 on the SH3 domain and Arg-77 in the PPII portion of the ligand. As explained above, this interaction is typical of both class I and class II interactions that use positive residues as compass residues, suggesting that one of the consequences of this type of interaction is the destabilization of the unliganded conformation of the RT loop. The total accessible surface on Nef and on the SH3 domain buried on Ê 2 , with approximately formation of the complex is approximately 1200 A 2 Ê 600 A buried on each partner (101, 122). In comparison, if the PPII helix of Nef is considered, alone, the total buried surface area is approximately Ê 2 , with roughly half of it buried on the SH3 domain. The latter values 780 A are similar to those reported for other SH3-peptide interactions (an example is found in reference 86), and the increase observed in the Nef± SH3 complex structure is due to the additional interactions occurring outside the proline-rich core (described above). The total accessible area of the Nef PPII helix, considered separately from the rest of the molecule, is Ê 2 . In the Nef±SH3 complex, the SH3 domain buries one-third of the 1200 A surface, and another third is buried by packing interactions with the rest of the Nef protein. The main interaction of the PPII helix with other elements of Nef involves the C-terminal region of helix aB and the following loop. The side chain of Q73 in the PPII helical segment protrudes away from the SH3 surface and is hydrogen bonded to the backbone of aB. Three other hydrogen bond interactions are likely to be important for stabilizing the position of R77, a critical component of the Nef±SH3 interface (Fig. 7). A separate NMR study of the Nef core showed that the XPXXP motif of Nef forms a PPII helix even in the absence of SH3 substrate (126).

I. Macromolecular Assemblies Containing SH3 Domains 1. Src as an Example of an Intramolecular Switch Regulated by SH3 Domain Interactions The tyrosine kinase product of the Src protooncogene can be regarded as a molecular switch translating into catalytic activity the regulatory

246

ANDREA MUSACCHIO

information received via intramolecular interactions. It has long been known that mutations affecting the C-terminal domain of Src, a region containing the conserved tyrosine residue Y527, result in kinase activation and transformation (reviewed in references 127, 128). It also became clear that other regions within the molecule may be important for the regulation of the catalytic activity. Amino acid substitutions affecting residues Arg-95 and Thr-96 in the RT loop of the SH3 domain are activating and partially transforming (129, 130). Using elegant approaches exploiting the heterologous expression of Src kinases in yeast cells, it was established that the SH3 domain plays a central role in the regulation of Src activity (131±134). These experiments de®ned a model of regulation according to which the phosphorylation of Y527 by the Csk tyrosine kinase leads to repression of kinase activity via an intramolecular mechanism involving the Src SH2 domain, a well-characterized phosphotyrosine-binding module. The SH3 domain would contribute to establishing a repressed state via some other type of intramolecular interaction. The determination of the three-dimensional structure of the inactive conformations of Hck (an Src family member) and Src ®nally delivered the structural basis of Src regulation (20±22, 135). For comments and reviews see references 136±142. As expected, the SH2 domain was found to be involved in an intramolecular interaction with the phosphorylated Y527 (Fig. 8). The apparent role of this interaction is that of appropriately positioning a linker segment located in the primary structure between the SH2 domain and the small lobe of the kinase domain. One of the most interesting revelations proposed by these structures was that the linker adopts a PPII conformation. In this conformation, it is sandwiched between the SH3 ligand-binding site and the small lobe of the kinase, thus contributing to the interaction of these fragments, which appears to be critical for the repression of the catalytic activity of the molecule (143). The interaction between the SH3 and the small lobe is further buttressed by the side chains of residues in the SH3 RT loop, such as Arg-95 and Thr-96, two residues with activation and transforming capacity. The ®rst of these residues is also involved in the formation of contacts with the peptide linker (20±22). The Src linker segment emerges from the SH2 domain in an extended conformation, which is retained until residue G254. Residues 249±253 within this region (KPQTQ) form a left-handed PPII helix and bind to the SH3 domain in the characteristic class II orientation. The side chain of P250, the only proline residue in this peptide, inserts in the P2 slot on the SH3 surface, precisely as expected for a class II peptide binding to this domain. P 1 , the other slot normally involved in binding proline residues in the minus orientation, is occupied instead by Q253, which is

HOW SH3 DOMAINS RECOGNIZE PROLINE

247

FIG. 8. (A) Interaction of p53 with 53BP2. The SH3 domain of 53BP2 is ¯anked by ankyrin repeats, and both moieties participate in the interaction with p53. The L3 loop of p53 is the main site of contact in the interaction. (B) The structure of Src. The SH3 domain binds to the linker region located between the SH2 domain and the kinase (in the primary sequence). The linker adopts a PPII conformation. The interaction is important in stabilizing the closed (inactive) conformation of the kinase. Also critical to inhibition is the interaction of the SH2 domain with the phosphorylated Tyr-527 near the C-terminus.

unable to make the optimal contacts with the SH3 surface at this external binding site of the PPII helix. After this residue, the polypeptide chain adopts a more contorted conformation and creates contacts between the SH3 and the catalytic domains that appear to be crucial to couple SH3linker binding to regulation of catalysis. In Hck, a somewhat longer fragment (encompassing residues P244±W254) is in the PPII conformation. The actual sequence of the linker (KPQKPWEK) reveals the presence of a PXXP motif. Only the ®rst proline is conserved in the linker region of Src family members, and similar to the case of Src, it is involved in interactions occurring at the P2 slot. Although the presence of a second proline residue in this region suggests a potentially stabler interaction, the side chains of Lys at the P3 and P0 slots are not positioned to take full advantage of the hydrophobic character of the surroundings. Thus, it appears that in the case of both Src and Hck the interactions of the PPII linker with the SH3-binding surface may be less than optimal. This may be important to achieve full kinase activation by competition mechanisms triggered by the interaction of the SH3 domain with optimal ligands (see for instance, references 144±146). While the mutation of the conserved proline residue in the Src linker results in the lack of Csk

248

ANDREA MUSACCHIO

down-regulation (despite phosphorylation of Y527), full conservation of the linker's primary structure is not strictly necessary (147). For instance, grafting the Lck linker onto Src results in an active kinase whose activity can be down-regulated by Csk (147). Two residues that appear to be key in coupling SH3 engagement to down-regulation of the catalytic domain, L255 and W260, lie immediately after the SH3-binding site on the linker (147±149). W260 is completely conserved in Src family kinases, while position 255 tends to contain large hydrophobic or aromatic residues. Interestingly, the stability of the interaction of the SH3 domain with the small lobe of the kinase is subject to regulation by the catalytic domain. In the inactive state, the critical aC-helix in the catalytic domain is positioned such that the formation of the Glu-310±Lys-295 salt bridge is precluded. At the same time, Y416 in the kinase activation loop is unphosphorylated, and the SH2 and SH3 domains are unavailable for interactions with other proteins. Phosphorylation of Y416 and mutations in the loop preceding the aC-helix activate Src and increase the accessibility of the SH3 domain for ligands. Thus, once active, Src family kinases become less prone to regulation, indicating a positive feedback loop for their activity (150). The domain organization of Src family kinases is found in other protein tyrosine kinases. The N-terminal region of Abl is similar to that of Src kinases, and it has been reported that the Abl-SH3 domain plays a negative regulatory role in the activity of the kinase. The linker between the SH2 and kinase domains in Abl contains the same KP motif found in Src. Mutations in these residues, or in other residues that may couple SH3 domain activity to catalytic repression, cause deregulation of the kinase activity (151). 2. Other Examples Screening of phage display peptide libraries revealed that Itk, a member of the Tek family of protein-tyrosine kinases, shows a preference for class I ligand KXXPXXP containing a lysine as the compass residue (152). A motif conforming to this consensus sequence precedes the SH3 domain of Itk. This prompted an analysis aimed at determining whether the sequence in Itk could act as an inter- or intramolecular ligand for the Itk-SH3 domain (153). No intermolecular interaction could be detected, but it became clear that an intramolecular association is possible. The proline-rich target binds in a class I orientation, and it is separated from the N-terminal region of the SH3 domain by a disordered loop. The af®nity of the interaction is quite low and apparently not suf®cient to prevent the binding of cellular ligands of the Itk-SH3 domain such as Grb2 and Sam68. However, the interaction may acquire a different signi®cance within the full-length protein (153).

HOW SH3 DOMAINS RECOGNIZE PROLINE

249

The overall structural organization of p85, the regulatory subunit of PI3K, has not yet been elucidated. However, there is good biochemical and functional evidence for the existence of intermolecular interactions involving the PI3K-SH3 domain and neighboring proline-rich regions and Bcr-homology (BH domains) (68, 154). The PI3K-SH3 domain recognizes the ®rst PRR (PRR1) of PI3K (154). In sedimentation equilibrium-analytical ultracentrifugation, experiments on the full-length p85 protein indicated the presence of dimers, whose formation was completely disrupted by an excess of a peptide corresponding to the p85 PRR1 sequence (154). A number of interaction experiments described by these authors clearly indicate the participation of the BH domain in dimerization, a hypothesis that was also proposed after crystal structure determination of this domain (155). Autoregulatory mechanisms, at least in part similar to those regulating Src-kinase activity, may occur within the membrane-associated guanylate kinases family. An interaction between the PSD-95 guanylate kinase (GK) and SH3 domains was ®rst identi®ed by a two-hybrid screen and was later con®rmed using a biochemical approach (156, 157). Mutation of the conserved typtophan residue in the SH3 domain to phenylalanine did not result in abolition of GK binding by the SH3 domain. Interestingly, there is no XP-X-XP motif in the GK domain, and there is a minimal requirement for binding by the SH3 domain that seems to correspond to a protein domain, rather than to a linear peptide (156). A point mutation in the SH3 domain that disrupts SH3±GK binding in vitro was identi®ed by its strong phenotype (158). Similar phenotypes are induced by short deletions that have the same disruptive effect on the interaction between the two domains (158). Similar autoregulatory switches have been described in other members of the family (159, 160). J. Interaction of SH3 Domains with Sequences That Do Not Conform to a Proline-Rich Consensus A number of linear SH3-binding sequences that do not conform to the XP-X-XP motif have been described (for instance, see references 52, 56, 161±164). High-resolution structural information on the interaction of these unusual peptides with the relevant SH3 domains has not been obtained yet. It is possible, but not yet proved, that the Pex13p-SH3 domain binds the non-proline-rich region of the Pex5p protein in a region different from that involved in PPII recognition (164). Phage display and other binding assays have been used to establish that the sequence speci®city of the Eps8-SH3 domain differs from the classical XP-X-XP model (52). Rather, this domain recognizes sequences containing a PXXDY motif. These sequences were isolated from peptide libraries,

250

ANDREA MUSACCHIO

but are also present in proteins known to interact with eps8 in vivo, such as E3B1 (165). Two striking features characterizing this domain, namely, the fact that it is able to form intertwined dimers (see above) and the presence of isoleucine in place of tyrosine within the C-terminal PXXY motif, have been shown to be irrelevant for the unusual ligand speci®city of this domain (52). A similar substitution (to leucine) is present within the 53BP2-SH3 domain, which recognizes proline-rich motifs conforming to classical class I consensus in phage display experiments (63). The p53 tumor suppressor induces cell cycle arrest or apoptosis on detection of DNA damage (166). 53BP2 was originally identi®ed for its ability to interact with p53 in a yeast two-hybrid screen (167). The 53BP2 Cterminal domain (residues 291 to 519) contains four ankyrin repeats (AR) and an SH3 domain. The crystal structure of this region complexed to the p53 DNA-binding domain (102) is a vivid example of an SH3 interaction that does not involve binding to a PPII helix. The structure indicated that the last AR and the SH3 domain are directly involved in binding the p53 DNA-binding domain (168), with a predominance of SH3-mediated interactions (Fig. 8). Although the SH3 ligand-binding site is directly involved in the interaction, the details of binding differ substantially from those observed in canonical SH3 binding to PPI helices. The interaction is con®ned to the p53 L3 loop, a rigid hairpin structure that is stabilized by multiple intramolecular hydrogen bonds and the coordination of a Zn atom (169). Two segments from this loop make contact with the SH3 domain. Some of these contacts resemble those observed for other SH3-peptide interactions, while others, in particular those with the RT loop, are novel. A key feature of the interaction is the fact that the side chain of Met-243 in p53 binds to the peptidebinding groove of the SH3 domain in a way that is made possible by the substitution of the very-well-conserved tyrosine residue in the C-terminal SH3 motif PXXY with leucine. The resulting cavity hosts the side chain of Met243. Despite this fact, the presence of leucine rather than tyrosine does not substantially affect the ability of the 53BP2-SH3 domain to interact with proline-rich regions (63). Another structurally characterized example of an SH3 interaction that does not involve canonical PPII motifs, the complex of the Vav and Grb2 SH3 domains, was reported recently (170). K. Regulation of SH3 Binding to Substrates To our knowledge, there is no evidence that proline isomerization may act as a regulatory mechanism in the interaction of PPII helices with their targets. Instead, there are examples of regulation of SH3 activity by phosphorylation. The Btk tyrosine kinase autophosphorylates on

HOW SH3 DOMAINS RECOGNIZE PROLINE

251

Tyr-223, a residue in the SH3 domain ligand-binding site, and mutation of Tyr-223 to Phe dramatically increases the transforming potential of Btk (171). Interestingly, in vitro experiments indicate that phosphorylation of Btk-SH3 may differentially modulate the af®nity of this domain for different targets (172). Two examples can be used to show how phosphorylation in the proximity of the target proline-rich sequence may in¯uence SH3 binding. Binding of the second SH3 domain of Nck to several targets is affected by phosphorylation of the target sequence at a serine residue three residues downstream from the arginine in the class II target motif (173). In a similar way, the interaction between Sos and Grb2 is modulated by the phosphorylation of the Sos C-terminal tail by mitogen-activated protein kinase (174, 175). Recently, a novel mechanism of regulation by arginine methylation has been described for the interaction of Fyn-SH3 with Sam68 (176). The proline-rich motifs in this protein are ¯anked by RG repeats that are targets for arginine N-methyltransferase. Interestingly, binding by WW domains to the Sam68 P-rich region is not affected by the modi®cation, suggesting a mechanism of differential regulation of binding. RG motifs are present in the proximity of proline-rich sites in other proteins, such as the dynamin tail (RRAPAVPPARPGSRG). We have shown above that Src is activated on disengagement from an intramolecular interaction triggered by phosphorylation of Tyr-527 and maintained by the SH3 domain. On the other hand, the phosphorylation of p47phox leads to an intramolecular inhibition and prepares the protein for interaction with p22phox and activation of the oxidase activity (177). A most interesting feature of proline-rich proteins is the existence of several overlapping or closely spaced putative binding sites for different SH3 domains. It is now beginning to be understood that these arrays of binding sites work to create the proper combination of proteins, one appropriate for that particular physiological state of the cell. A striking example of this is the switch in Sos speci®city observed for Sos in Ras to Rac signaling (165). WASP and dynamin are also able to bind to several different targets, each of which appears to be capable of triggering different effects (reviewed in references 178±180). There are at least ®ve potential SH3-binding sites in the carboxy-terminal tail of dynamin, and binding assays coupled to deletion analysis indicate that SH3 domains from different proteins bind at different sites (181). The N-Grb2and amphiphysin SH3-domains, for instance, recognize overlapping class II proline-rich sites in the dynamin C-terminal tail (PPQVPSRPNR and PPQVPSRPNR, respectively), and the presence of Grb2 prevents amphiphysin binding (59). Other Grb2 sites are present in this area of the molecule (182), but deletion analysis con®rms the absolute relevance of the interaction with the sequence reported above. In vivo, the

252

ANDREA MUSACCHIO

interaction between the SH3 domain of amphiphysin and dynamin plays an essential role in vesicle recycling, as microinjection of the domain, or of the dynamin proline-rich peptide, impairs synaptic vesicle endocytosis (183). It has been shown that SH3 binding stimulates the GTPase activity of dynamin. The implications of these results are that this may happen in several ways, as each SH3 binds to different regions. L. Inhibition of SH3-Mediated Interactions The de®nition ``Molecular Velcro'' for SH3 domains (184) is appropriate, as it conveys the idea that these moieties act as molecular adhesives. One of the main unsolved puzzles in the ®eld is whether selective target recognition is a property of SH3 domain interactions in vivo and what proportion of these is regulated. The combination of a relatively small interaction interface with the scarcity of hydrogen bonds may explain the highly promiscuous character of SH3 interactions (185). While it is not unlikely that low af®nity is the desired property of the particular brand of Velcro licensed by SH3 domains, conclusions drawn from the use of peptides to evaluate the strength and selectivity of an SH3-mediated interaction may suffer from a fundamental limitation. ``Real'' interactions of SH3 domains with targets, such as those involving Nef, p53, or Sos, indicate higher af®nities than those usually attributed to interactions of Prich peptides with single SH3 domains. For instance, Nef potently activates the Src-family kinase Hck by a mechanism of SH3 displacement, but an Sos-derived peptide binding to Hck-SH3 with a KD of 5 mM does not (186). Moreover, instances can be expected in which the cellular concentration of an SH3 target may be suf®ciently high to make it impossible to avoid affecting other SH3-mediated pathways. The playground in which SH3 domains play their sticky games is one that references human health (for a review see references 187). Is it possible to obtain small synthetic compounds counteracting some of these cellular interactions? Even if the nature of the SH3 ligand-binding site makes it unlikely that these domains be identi®ed as primary drug targets, the availability of compounds selectively targeting speci®c cellular pathways would be a tremendous tool for studying the function of the SH3 domains in vivo. Ideally, the interactions of these compounds with SH3 domains should be characterized by high af®nity and selectivity. Several strategies have been explored to obtain high-af®nity and -selectivity ligands for SH3 domains. Cell-penetrating peptides that bind with high af®nity and selectivity to the ®rst SH3 domain of the adapter protein Crkl have been described (188). Approaches based on phage display and solid-phase peptide synthesis have been described in previous sections. In most instances, the libraries utilized in these efforts were biased by the

HOW SH3 DOMAINS RECOGNIZE PROLINE

253

introduction of a proline-rich scaffold, while positions at which proline is not absolutely required were randomized. These approaches have been particularly useful for the de®nition of the binding speci®city of SH3 domains, but with a few exceptions they failed to provide ligands with improved af®nity and selectivity in comparison to natural ligands. Kim and collaborators devised a modi®ed version of phage display. These authors proposed a method, mirror-image phage display, ultimately to deliver peptide ligands of the SH3 domain containing the D-enantiomer of the constituent amino acids (189). In this method, the D-enantiomer of a protein is prepared by chemical synthesis and used to isolate L -peptide ligands that interact with it from a phage display library. The selection process is carried out in an achiral solvent (water), and the assumption is made that the interaction between the L -peptide and the D-protein is unlikely to require any chiral cofactors. Consequently, the D-enantiomeric form of the isolated L -peptide ligands should interact with the protein of the natural, L -amino acid con®guration. When the phage display library was selected using the D-enantiomeric SH3 domain of Src, a set of ligands was isolated bearing no sequence similarity to the P-rich peptide. These peptides are characterized by a combination of conserved Leu and Gly residues and a conserved Arg or Lys. These charged residues are located in the middle of the sequence, rather than near the end, as in most SH3 (L -) ligands. All ligands contain a pair of Cys residues, suggesting that they may be circular. (D-Peptide ligands do not contain proline simply because you cannot ®t the mirror image of the PPII helix within the pockets of the SH3-binding site. The mirror image of the SH3 ligandbinding site would be needed to bind proline-rich D-peptides.) A Dpeptide was created based on the sequence of the L -enantiomer and bound to the L -form of Src-SH3 with a KD of 63 mM. The reduced form of this peptide bound with more than 10-fold reduced af®nity, suggesting that disul®de formation is required for ef®cient binding. NMR chemical shift experiments indicate that the D-peptide occupies only part of the SH3 ligand-binding site normally contacted by the PPI helix, suggesting that better ligands may be obtained with the help of chemical design (189). Schreiber and collaborators proposed a split-pool combinatorial chemistry approach based on the combination of three consecutive members of a library of 32 monomers with the C-terminal portion of an ideal Src class I proline-rich moiety (PLPPLP) (190, 191). The ligands , whose general structure is thus Cap±monomer 1±monomer 2±monomer 3± PLPPLP, were selected using the Src-SH3 domain. The tightest binding member in the set of isolated molecules interacted with the Src-SH3 with a KD of 3.4 mM. A solution structure of this complex revealed the structural basis of interaction (71). As expected, the ligand binds as a class I

254

ANDREA MUSACCHIO

ligand, with the proline-rich moiety hosted between the P 1 and P3 slots. The compass pocket is occupied by the stacking ring interaction between an SH3 tyrosine and monomer 29 from the library and other areas of Ê 2 of exposed hydrophobic contact with the RT loop. Indeed, over 600 A SH3 surface are buried on ligand binding, a large value by SH3 standards. The af®nities of these interactions are not higher than those charÊ 2 on the SH3 acteristic of a proline-rich peptide burying less than 400 A domain. If needed, this is further evidence of the dif®culties one encounters when elaborating rational explanations for the structural basis of binding af®nity. Attempts to apply similar split-pool techniques for the isolation of molecules mimicking the proline-rich portion of the peptide have also been reported (192, 193). Yet a different approach was proposed by Cadus Pharmaceuticals (187). The most successful approach thus far exploited N-substituted amino acids (peptoids) to derive high-af®nity and selective binders (90, 93; reviewed in references 91). The inspiring idea behind this approach is that N-alkyl substitution is the chemical principle guiding the interaction of proline-rich regions with SH3 domains. By maintaining an N-substituted backbone pattern, peptoids function as proline mimetics, while suf®cient chemical diversity can be introduced by modifying the side chain of the different N-substituted amino acids (Fig. 9). The usefulness of this approach was originally demonstrated by individually replacing residues in the proline-rich core of several natural SH3 and WW ligands by either alanine or sarcosine (N-methyl glycine). As already explained in Section III, F, alanine mutants of proline residues involved in external

FIG. 9. (A) N-substituted peptides (peptoids) at external binding sites mimic the function of proline in the interaction of PPII helics with SH3 and WW domains. Z is Nmethyl glycine. In (B) the equivalent peptide containing a proline in the same position is shown.

HOW SH3 DOMAINS RECOGNIZE PROLINE

255

packing show reduced af®nity, whereas ligands containing sarcosine at these sites bind their target domains with af®nities comparable to that of wild-type peptides, showing the requirement for an N-substituted residue. Conversely, Ca substitution is needed at the position preceding the Nsubstituted residue (90). These observations indicate that a strategy to obtain SH3 superbinders may consist of the introduction of a repertoire of N-substituted side chains at external packing sites. Because proline is the only natural N-substituted amino acid, chemical synthesis has proved the only way to obtain such high-af®nity ligands. Lim and collaborators reported the usefulness of this approach in tailoring high-af®nity ligands. For instance, substitution of an N-(S)-phenylethyl group at site P 1 in the peptide YEVPPPVPPRRR (the P 1 proline is shown in bold face type) results in a peptoid that binds to the Grb2-N-SH3 domain with a KD of 40 nM and acts as a potent competitive inhibitor of the interaction of the Grb2-SH3 domain with a Sos peptide fusion protein (90). The peptoid also shows largely enhanced selectivity toward Grb2-N-SH3. More recently, the same authors showed that the entire XPXXP core can be replaced by nonproline residues and in particular that each ``external site'' P can be mutated into a peptoid. Several high-af®nity, selective compounds for the SH3 domains of Grb2, Crk, and Src were identi®ed (93).

ACKNOWLEDGMENTS I apologize to all those whose work has not been appropriately reviewed due to space limitations. The literature on the SH3 domain has become immense, and some degree of selection was absolutely required. Special thanks are due to Gianni Cesareni, Gianluca Cestra, Manuela Helmer Citterich, Trevor Creamer, Stephan Feller, Brian Kay, Christine Kinnon, Wendell Lim, Michael Rosen, and Giulio Superti-Furga for discussions and the sharing of unpublished information and other material. A.M. is a Scholar of the Italian Foundation for Cancer Research and an EMBO Young Investigator. This work is dedicated to the memory of Matti Saraste.

REFERENCES 1. Kay, B. K., Williamson, M. P., and Sudol, M. (2000). The importance of being proline: The interaction of proline-rich motifs in signaling proteins with their cognate domains. FASEB J. 14, 231±241. 2. Sudol, M. (1998). From Src homology domains to other signaling modules: Proposal of the `protein recognition code.' Oncogene 17, 1469±1474. 3. MacArthur, M. W., and Thornton, J. M. (1991). In¯uence of proline residues on protein conformation. J. Mol. Biol. 218, 397±412.

256

ANDREA MUSACCHIO

4. Williamson, M. P. (1994). The structure and function of proline-rich regions in proteins. Biochem. J. 297, 249±260. 5. Siligardi, G., and Drake, A. F. (1995). The importance of extended conformations and, in particular, the PII conformation for the molecular recognition of peptides. Biopolymers 37, 281±292. 6. Cowan, P. M., and McGavin, S. (1955). Structure of poly-L -proline. Nature 176, 470±478. 7. Stapley, B. J., and Creamer, T. P. (1995). A survey of left-handed polyproline II helices. Protein Sci. 8, 587±595. 8. Adzhubei, A. A., and Sternberg, M. J. (1993). Left-handed polyproline II helices commonly occur in globular proteins. J. Mol. Biol. 229, 472±493. 9. Sreerama, N., and Woody, R. W. (1994). Poly(pro)II helices in globular proteins: Identi®cation and circular dichroic analysis. Biochemistry 33, 10022±10025. 10. Adzhubei, A. A., and Sternberg, M. J. (1994). Conservation of polyproline II helices in homologous proteins: Implications for structure prediction by model building. Protein Sci. 3, 2395±2410. 11. Chang, A., Cheang, S., Espanel, X., and Sudol, M. (2000). Rsp5 WW domains interact directly with the carboxyl-terminal domain of RNA polymerase II. J. Biol. Chem. 275, 20562±20571. 12. Bienkiewicz, E. A., Moon Woody, A., and Woody, R. W. (2000). Conformation of the RNA polymerase II C-terminal domain: Circular dichroism of long and short fragments. J. Mol. Biol. 297, 119±133. 13. Bousquet, J. A., Garbay, C., Roques, B. P., and Mely, Y. (2000). Circular dichroic investigation of the native and non-native conformational states of the growth factor receptor-binding protein 2 N-terminal src homology domain 3: Effect of binding to a proline-rich peptide from guanine nucleotide exchange factor. Biochemistry 39, 7722±7735. 14. Viguera, A. R., Arrondo, J. L., Musacchio, A., Saraste, M., and Serrano, L. (1994). Characterization of the interaction of natural proline-rich peptides with ®ve different SH3 domains. Biochemistry 33, 10925±10933. 15. Renzoni, D. A., Pugh, D. J., Siligardi, G., Das, P., Morton, C. J., Rossi, C., Water®eld, M. D., Campbell, I. D., and Ladbury, J. E. (1996). Structural and thermodynamic characterization of the interaction of the SH3 domain from Fyn with the proline-rich binding site on the p85 subunit of PI3-kinase. Biochemistry 35, 15646±15653. 16. Morton, C. J., Pugh, D. J., Brown, E. L., Kahmann, J. D., Renzoni, D. A., and Campbell, I. D. (1996). Solution structure and peptide binding of the SH3 domain from human Fyn. Structure 4, 705±714. 17. Dong, J., Misselwitz, R., Wel¯e, H., and Westermann, P. (2000). Expression and puri®cation of dynamin II domains and initial studies on structure and function. Protein Expr. Purif. 20, 314±323. 18. Jardetzky, T. S., Brown, J. H., Gorga, J. C., Stern, L. J., Urban, R. G., Strominger, J. L., and Wiley, D. C. (1996). Crystallographic analysis of endogenous peptides associated with HLA-DR1 suggests a common, polyproline II-like conformation for bound peptides. Proc. Natl. Acad. Sci. USA 93, 734±738. 19. Stern, L. J., Brown, J. H., Jardetzky, T. S., Gorga, J. C., Urban, R. G., Strominger, J. L., and Wiley, D. C. (1994). Crystal structure of the human class II MHC protein HLA-DR1 complexed with an in¯uenza virus peptide. Nature 368, 215±221. 20. Sicheri, F., Moare®, I., and Kuriyan, J. (1997). Crystal structure of the Src family tyrosine kinase Hck. Nature 385, 602±609. 21. Xu, W., Harrison, S. C., and Eck, M. J. (1997). Three-dimensional structure of the tyrosine kinase c-Src. Nature 385, 595±602.

HOW SH3 DOMAINS RECOGNIZE PROLINE

257

22. Williams, J. C., Weijland, A., Gon¯oni, S., Thompson, A., Courtneidge, S. A., SupertiÊ crystal structure of the inactivated Furga, G., and Wierenga, R. K. (1997). The 2.35 A form of chicken Src: A dynamic molecule with multiple regulatory interactions. J. Mol. Biol. 274, 757±775. 23. Creamer, T. P. (1998). Left-handed polyproline II helix formation is (very) locally driven. Proteins 33, 218±226. 24. Mayer, B. J. (2001). SH3 domains: Complexity in moderation. J. Cell Sci. 114, 1253± 1263. 25. Schlessinger, J. (1994). SH2/SH3 signaling proteins. Curr. Opin. Genet. Dev. 4, 25±30. 26. Cohen, G. B., Ren, R., and Baltimore, D. (1995). Modular binding domains in signal transduction proteins. Cell 80, 237±248. 27. Kay, B. K., Williamson, M. P., and Sudol, M. (2000). The importance of being proline: the interaction of proline-rich motifs in signaling proteins with their cognate domains. FASEB J. 14, 231±241. 28. Pawson, T., and Nash, P. (2000). Protein±protein interactions de®ne speci®city in signal transduction. Genes Dev. 14, 1027±1047. 29. Schultz, J., Copley, R. R., Doerks, T., Ponting, C. P., and Bork, P. (2000). SMART: A Webbased tool for the study of genetically mobile domains. Nucleic Acids Res. 28, 231±234. 30. Schlessinger, J. (1993). How receptor tyrosine kinases activate Ras. Trends Biochem Sci. 18, 273±275. 31. Yu, H., Rosen, M. K., Shin, T. B., Seidel-Dugan, C., Brugge, J. S., and Schreiber, S. L. (1992). Solution structure of the SH3 domain of Src and identi®cation of its ligandbinding site. Science 258, 1665±1668. 32. Musacchio, A., Noble, M., Pauptit, R., Wierenga, R., and Saraste, M. (1992). Crystal structure of a Src-homology 3 (SH3) domain. Nature 359, 851±855. 33. Kohda, D., Hatanaka, H., Odaka, M., Mandiyan, V., Ullrich, A., Schlessinger, J., and Inagaki, F. (1993). Solution structure of the SH3 domain of phospholipase C-gamma. Cell 72, 953±960. 34. Koyama, S., Yu, H., Dalgarno, D. C., Shin, T. B., Zydowsky, L. D., and Schreiber, S. L. (1993). Structure of the PI3K SH3 domain and analysis of the SH3 family. Cell 72, 945±952. 35. Booker, G. W., Gout, I., Downing, A. K., Driscoll, P. C., Boyd, J., Water®eld, M. D., and Campbell, I. D. (1993). Solution structure and ligand-binding site of the SH3 domain of the p85 alpha subunit of phosphatidylinositol 3-kinase. Cell 73, 813±822. 36. Noble, M. E., Musacchio, A., Saraste, M., Courtneidge, S. A., and Wierenga, R. K. (1993). Crystal structure of the SH3 domain in human Fyn: Comparison of the threedimensional structures of SH3 domains in tyrosine kinases and spectrin. EMBO J. 12, 2617±2624. 37. Musacchio, A., Wilmanns, M., and Saraste, M. (1994). Structure and function of the SH3 domain. Prog. Biophys. Mol. Biol. 61, 283±297. 38. Whisstock, J. C., and Lesk, A. M. (1999). SH3 domains in prokaryotes. Trends Biochem. Sci. 24, 132±133. 39. Kortemme, T., Kelly, M. J., Kay, L. E., Forman-Kay, J., and Serrano, L. (2000). Similarities between the spectrin SH3 domain denatured state and its folding transition state. J. Mol. Biol. 297, 1217±1229. 40. Li, L., Mirny, L. A., and Shakhnovich, E. I. (2000). Kinetics, thermodynamics and evolution of non-native interactions in a protein folding nucleus. Nat. Struct. Biol. 7, 336±342. 41. Fersht, A. R. (2000). Transition-state structure as a unifying basis in protein-folding mechanisms: Contact order, chain topology, stability, and the extended nucleus mechanism. Proc. Natl. Acad. Sci. USA 97, 1525±1529.

258

ANDREA MUSACCHIO

42. Gu, H., Doshi, N., Kim, D. E., Simons, K. T., Santiago, J. V., Nauli, S., and Baker, D. (1999). Robustness of protein folding kinetics to surface hydrophobic substitutions. Protein Sci. 8, 2734±2741. 43. Riddle, D. S., Grantcharova, V. P., Santiago, J. V., Alm, E., Ruczinski, I., and Baker, D. (1999). Experiment and theory highlight role of native state topology in SH3 folding. Nat. Struct. Biol. 6, 1016±1024. 44. Martinez, J. C., and Serrano, L. (1999). The folding transition state between SH3 domains is conformationally restricted and evolutionarily conserved. Nat. Struct. Biol. 6, 1010±1016. 45. Tsai, J., Levitt, M., and Baker, D. (1999). Hierarchy of structure loss in MD simulations of src SH3 domain unfolding. J. Mol. Biol. 291, 215±225. 46. Sadqi, M., Casares, S., Abril, M. A., Lopez-Mayorga, O., Conejero-Lara, F., and Freire, E. (1999). The native state conformational ensemble of the SH3 domain from alphaspectrin. Biochemistry 38, 8899±8906. 47. Larson, S. M., and Davidson, A. R. (2000). The identi®cation of conserved interactions within the SH3 domain by alignment of sequences and structures. Protein Sci. 9, 2170±2180. 48. Guijarro, J. I., Sunde, M., Jones, J. A., Campbell, I. D., and Dobson, C. M. (1998). Amyloid ®bril formation by an SH3 domain. Proc. Natl. Acad. Sci. USA 95, 4224±4228. 49. Jimenez, J. L., Guijarro, J. I., Orlova, E., Zurdo, J., Dobson, C. M., Sunde, M., and Saibil, H. R. (1999). Cryo-electron microscopy structure of an SH3 amyloid ®bril and model of the molecular packing. EMBO J. 18, 815±821. 50. Sunde, M., and Blake, C. C. (1998). From the globular to the ®brous state: Protein structure and structural conversion in amyloid formation. Q. Rev. Biophys. 31, 1±39. 51. Kishan, K. V., Scita, G., Wong, W. T., Di Fiore, P. P., and Newcomer, M. E. (1997). The SH3 domain of Eps8 exists as a novel intertwined dimer. Nat. Struct. Biol. 4, 739±743. 52. Mongiovi, A. M., Romano, P. R., Panni, S., Mendoza, M., Wong, W. T., Musacchio, A., Cesareni, G., and Di Fiore, P. P. (1999). A novel peptide±SH3 interaction. EMBO J. 18, 5300±5309. 53. Kishan, K. V., Newcomer, M. E., Rhodes, T. H., and Guilliot, S. D. (2001). Effect of pH and salt bridges on structural assembly: Molecular structures of the monomer and intertwined dimer of the Eps8 SH3 domain. Protein Sci. 10, 1046±1055. 54. Ren, R., Mayer, B. J., Cicchetti, P., and Baltimore, D. (1993). Identi®cation of a tenamino acid proline-rich SH3 binding site. Science 259, 1157±1161. 55. Cicchetti, P., Mayer, B. J., Thiel, G., and Baltimore, D. (1992). Identi®cation of a protein that binds to the SH3 region of Abl and is similar to Bcr and GAP-rho. Science 257, 803±806. 56. Cestra, G., Castagnoli, L., Dente, L., Minenkova, O., Petrelli, A., Migone, N., Hoffmuller, U., Schneider-Mergener, J., and Cesareni, G. (1999). The SH3 domains of endophilin and amphiphysin bind to the proline-rich region of synaptojanin 1 at distinct sites that display an unconventional binding speci®city. J. Biol. Chem. 274, 32001±32007. 57. Kurakin, A., Hoffman, N. G., and Kay, B. K. (1998). Molecular recognition properties of the C-terminal Sh3 domain of the Cbl associated protein, Cap. J. Peptide. Res. 52, 331±337. 58. Sparks, A. B., Rider, J. E., and Kay, B. K. (1998). Mapping the speci®city of SH3 domains with phage-displayed random-peptide libraries. Methods Mol. Biol. 84, 87±103. 59. Grabs, D., Slepnev, V. I., Songyang, Z., David, C., Lynch, M., Cantley, L. C., and De Camilli, P. (1997). The SH3 domain of amphiphysin binds the proline-rich domain of dynamin at a single site that de®nes a new SH3 binding consensus sequence. J. Biol. Chem. 272, 13419±13425. 60. Yamabhai, M., and Kay, B. K. (1997). Examining the speci®city of Src homology 3 domain±ligand interactions with alkaline phosphatase fusion proteins. Anal. Biochem. 247, 143±151.

HOW SH3 DOMAINS RECOGNIZE PROLINE

259

61. Hoffman, N. G., Sparks, A. B., Carter, J. M., and Kay, B. K. (1996). Binding properties of SH3 peptide ligands identi®ed from phage-displayed random peptide libraries. Mol. Divers. 2, 5±12. 62. Schmitz, R., Baumann, G., and Gram, H. (1996). Catalytic speci®city of phosphotyrosine kinases Blk, Lyn, c-Src and Syk as assessed by phage display. J. Mol. Biol. 260, 664±677. 63. Sparks, A. B., Rider, J. E., Hoffman, N. G., Fowlkes, D. M., Quillam, L. A., and Kay, B. K. (1996). Distinct ligand preferences of Src homology 3 domains from Src, Yes, Abl, Cortactin, p53bp2, PLCgamma, Crk, and Grb2. Proc. Natl. Acad. Sci. USA 93, 1540±1544. 64. Sparks, A. B., Quilliam, L. A., Thorn, J. M., Der, C. J., and Kay, B. K. (1994). Identi®cation and characterization of Src SH3 ligands from phage-displayed random peptide libraries. J. Biol. Chem. 269, 23853±23856. 65. Yu, H., Chen, J. K., Feng, S., Dalgarno, D. C., Brauer, A. W., and Schreiber, S. L. (1994). Structural basis for the binding of proline-rich peptides to SH3 domains. Cell 76, 933±945. 66. De Leo, F. R., Ulman, K. V., Davis, A. R., Jutila, K. L., and Quinn, M. T. (1996). Assembly of the human neutrophil NADPH oxidase involves binding of p67phox and ¯avocytochrome b to a common functional domain in p47phox. J. Biol. Chem. 271, 17013±17020. 67. Rickles, R. J., Bot®eld, M. C., Zhou, X. M., Henry, P. A., Brugge, J. S., and Zoller, M. J. (1995). Phage display selection of ligand residues important for Src homology 3 domain binding speci®city. Proc. Natl. Acad. Sci. USA 92, 10909±10913. 68. Rickles, R. J., Bot®eld, M. C., Weng, Z., Taylor, J. A., Green, O. M., Brugge, J. S., and Zoller, M. J. (1994). Identi®cation of Src, Fyn, Lyn, PI3K and Abl SH3 domain ligands using phage display libraries. EMBO J. 13, 5598±5604. 69. Cheadle, C., Ivashchenko, Y., South, V., Searfoss, G. H., French, S., Howk, R., Ricca, G. A., and Jaye, M. (1994). Identi®cation of a Src SH3 domain binding motif by screening a random phage display library. J. Biol. Chem. 269, 24034±24039. 70. Simon, J. A., and Schreiber, S. L. (1995). Grb2 SH3 binding to peptides from Sos: Evaluation of a general model for SH3±ligand interactions. Chem. Biol. 2, 53±60. 71. Feng, S., Kapoor, T. M., Shirai, F., Combs, A. P., and Schreiber, S. L. (1996). Molecular basis for the binding of SH3 ligands with non-peptide elements identi®ed by combinatorial synthesis. Chem. Biol. 3, 661±670. 72. Egan, S. E., Giddings, B. W., Brooks, M. W., Buday, L., Sizeland, A. M., and Weinberg, R. A. (1993). Association of Sos Ras exchange protein with Grb2 is implicated in tyrosine kinase signal transduction and transformation. Nature 363, 45±51. 73. Rozakis-Adcock, M., Fernley, R., Wade, J., Pawson, T., and Bowtell, D. (1993). The SH2 and SH3 domains of mammalian Grb2 couple the EGF receptor to the Ras activator mSos1. Nature 363, 83±85. 74. Buday, L., and Downward, J. (1993). Epidermal growth factor regulates p21ras through the formation of a complex of receptor, Grb2 adapter protein, and Sos nucleotide exchange factor. Cell 73, 611±620. 75. Skolnik, E. Y., Batzer, A., Li, N., Lee, C. H., Lowenstein, E., Mohammadi, M., Margolis, B., and Schlessinger, J. (1993). The function of GRB2 in linking the insulin receptor to Ras signaling pathways. Science 260, 1953±1955. 76. Baltensperger, K., Kozma, L. M., Cherniack, A. D., Klarlund, J. K., Chawla, A., Banerjee, U., and Czech, M. P. (1993). Binding of the Ras activator son of sevenless to insulin receptor substrate-1 signaling complexes. Science 260, 1950±1952. 77. Chardin, P., Camonis, J. H., Gale, N. W., van Aelst, L., Schlessinger, J., Wigler, M. H., and Bar-Sagi, D. (1993). Human Sos1: A guanine nucleotide exchange factor for Ras that binds to GRB2. Science 260, 1338±1343.

260

ANDREA MUSACCHIO

78. Li, N., Batzer, A., Daly, R., Yajnik, V., Skolnik, E., Chardin, P., Bar-Sagi, D., Margolis, B., and Schlessinger, J. (1993). Guanine-nucleotide-releasing factor hSos1 binds to Grb2 and links receptor tyrosine kinases to Ras signalling. Nature 363, 85±88. 79. Gale, N. W., Kaplan, S., Lowenstein, E. J., Schlessinger, J., and Bar-Sagi, D. (1993). Grb2 mediates the EGF-dependent activation of guanine nucleotide exchange on Ras. Nature 363, 88±92. 80. Olivier, J. P., Raabe, T., Henkemeyer, M., Dickson, B., Mbamalu, G., Margolis, B., Schlessinger, J., Hafen, E., and Pawson, T. (1993). A Drosophila SH2±SH3 adaptor protein implicated in coupling the sevenless tyrosine kinase to an activator of Ras guanine nucleotide exchange, Sos. Cell 73, 179±191. 81. Brannetti, B., Via, A., Cestra, G., Cesareni, G., and Citterich, M. H. (2000). SH3-SPOT: An algorithm to predict preferred ligands to different members of the SH3 gene family. J. Mol. Biol. 298, 313±328. 82. Yaffe, M. B., Leparc, G. G., Lai, J., Obata, T., Volinia, S., and Cantley, L. C. (2001). A motif-based pro®le scanning approach for genome-wide prediction of signaling pathways. Nat. Biotechnol. 19, 348±353. 83. Zucconi, A., Dente, L., Santonico, E., Castagnoli, L., and Cesareni, G. (2001). Selection of ligands by panning of domain libraries displayed on phage lambda reveals new potential partners of synaptojanin 1. J. Mol. Biol. 307, 1329±1339. 84. Feng, S., Chen, J. K., Yu, H., Simon, J. A., and Schreiber, S. L. (1994). Two binding orientations for peptides to the Src SH3 domain: Development of a general model for SH3±ligand interactions. Science 266, 1241±1247. 85. Lim, W. A., Richards, F. M., and Fox, R. O. (1994). Structural determinants of peptidebinding orientation and of sequence speci®city in SH3 domains. Nature 372, 375±379. 86. Musacchio, A., Saraste, M., and Wilmanns, M. (1994). High-resolution crystal structures of tyrosine kinase SH3 domains complexed with proline-rich peptides. Nat. Struct. Biol. 1, 546±551. 87. Pisabarro, M. T., and Serrano, L. (1996). Rational design of speci®c high-af®nity peptide ligands for the Abl-SH3 domain. Biochemistry 35, 10634±10640. 88. Vidal, M., Goudreau, N., Cornille, F., Cussac, D., Gincel, E., and Garbay, C. (1999). Molecular and cellular analysis of Grb2 SH3 domain mutants: Interaction with Sos and dynamin. J. Mol. Biol. 290, 717±730. 89. Chen, Y. J., Lin, S. C., Tzeng, S. R., Patel, H. V., Lyu, P. C., and Cheng, J. W. (1996). Stability and folding of the SH3 domain of Bruton's tyrosine kinase. Proteins 26, 465±471. 90. Nguyen, J. T., Turck, C. W., Cohen, F. E., Zuckermann, R. N., and Lim, W. A. (1998). Exploiting the basis of proline recognition by SH3 and WW domains: Design of Nsubstituted inhibitors. Science 282, 2088±2092. 91. Aghazadeh, B., and Rosen, M. K. (1999). Ligand recognition by SH3 and WW domains: The role of N-alkylation in PPII helices. Chem. Biol. 6, R241±R246. 92. Kolafa, J., Perram, J. W., and Bywater, R. P. (2000). Essential motions and energetic contributions of individual residues in a peptide bound to an SH3 domain. Biophys. J. 79, 646±655. 93. Nguyen, J. T., Porter, M., Amoui, M., Miller, W. T., Zuckermann, R. N., and Lim, W. A. (2000). Improving SH3 domain ligand selectivity using a non-natural scaffold. Chem. Biol. 7, 463±473. 94. Fedorov, A. A., Fedorov, E., Gertler, F., and Almo, S. C. (1999). Structure of EVH1, a novel proline-rich ligand-binding module involved in cytoskeletal dynamics and neural function. Nat. Struct. Biol. 6, 661±665. 95. Macias, M. J., Hyvonen, M., Baraldi, E., Schultz, J., Sudol, M., Saraste, M., and Oschkinat, H. (1996). Structure of the WW domain of a kinase-associated protein complexed with a proline-rich peptide. Nature 382, 646±649.

HOW SH3 DOMAINS RECOGNIZE PROLINE

261

96. Mahoney, N. M., Rozwarski, D. A., Fedorov, E., Fedorov, A. A., and Almo, S. C. (1999). Pro®lin binds proline-rich ligands in two distinct amide backbone orientations. Nat. Struct. Biol. 6, 666±671. 97. Huang, X., Poy, F., Zhang, R., Joachimiak, A., Sudol, M., and Eck, M. J. (2000). Structure of a WW domain containing fragment of dystrophin in complex with betadystroglycan. Nat. Struct. Biol. 7, 634±638. 98. Wu, X., Knudsen, B., Feller, S. M., Zheng, J., Sali, A., Cowburn, D., Hanafusa, H., and Kuriyan, J. (1995). Structural basis for the speci®c interaction of lysine-containing proline-rich peptides with the N-terminal SH3 domain of c-Crk. Structure 3, 215±226. 99. Owen, D. J., Wigge, P., Vallis, Y., Moore, J. D., Evans, P. R., and McMahon, H. T. (1998). Crystal structure of the amphiphysin-2 SH3 domain and its role in the prevention of dynamin ring formation. EMBO J. 17, 5273±5285. 100. Feng, S., Kasahara, C., Rickles, R. J., and Schreiber, S. L. (1995). Speci®c interactions outside the proline-rich core of two classes of Src homology 3 ligands. Proc. Natl. Acad. Sci. USA 92, 12408±12415. 101. Lee, C. H., Saksela, K., Mirza, U. A., Chait, B. T., and Kuriyan, J. (1996). Crystal structure of the conserved core of HIV-1 Nef complexed with a Src family SH3 domain. Cell 85, 931±942. 102. Gorina, S., and Pavletich, N. P. (1996). Structure of the p53 tumor suppressor bound to the ankyrin and SH3 domains of 53BP2. Science 274, 1001±1005. 103. Ma, Y. C., and Huang, X. Y. (1998). Identi®cation of the binding site for Gqalpha on its effector Bruton's tyrosine kinase. Proc. Natl. Acad. Sci. USA 95, 12197±12201. 104. Sparks, A. B., Hoffman, N. G., McConnell, S. J., Fowlkes, D. M., and Kay, B. K. (1996). Cloning of ligand targets: Systematic isolation of SH3 domain-containing proteins. Nat. Biotechnol. 14, 741±744. 105. Van Etten, R. A. (1999). Cycling, stressed-out and nervous: Cellular functions of c-Abl. Trends Cell Biol. 9, 179±186. 106. Pisabarro, M. T., Serrano, L., and Wilmanns, M. (1998). Crystal structure of the ablSH3 domain complexed with a designed high-af®nity peptide ligand: Implications for SH3±ligand interactions. J. Mol. Biol. 281, 513±521. 107. Weng, Z., Rickles, R. J., Feng, S., Richard, S., Shaw, A. S., Schreiber, S. L., and Brugge, J. S. (1995). Structure±function analysis of SH3 domains: SH3 binding speci®city altered by single amino acid substitutions. Mol. Cell. Biol. 15, 5627±5634. 108. Maignan, S., Guilloteau, J. P., Fromage, N., Arnoux, B., Becquart, J., and Ducruix, A. (1995). Crystal structure of the mammalian Grb2 adaptor. Science 268, 291±293. 109. Odai, H., Sasaki, K., Iwamatsu, A., Hanazono, Y., Tanaka, T., Mitani, K., Yazaki, Y., and Hirai, H. (1995). The proto-oncogene product c-Cbl becomes tyrosine phosphorylated by stimulation with GM-CSF or Epo and constitutively binds to the SH3 domain of Grb2/Ash in human hematopoietic cells. J. Biol. Chem. 270, 10800±10805. 110. Chen, J. K., Lane, W. S., Brauer, A., Tanaka, A., and Schreiber, S. L. (1993). Biased combinatorial libraries: Novel ligands for the SH3 domain of phosphatidylinositol 3kinase. J. Am. Chem. Soc. 115, 12591±12592. 111. Goudreau, N., Cornille, F., Duchesne, M., Parker, F., Tocque, B., Garbay, C., and Roques, B. P. (1994). NMR structure of the N-terminal SH3 domain of GRB2 and its complex with a proline-rich peptide from Sos. Nat. Struct. Biol. 1, 898±907. 112. Terasawa, H., Kohda, D., Hatanaka, H., Tsuchiya, S., Ogura, K., Nagata, K., Ishii, S., Mandiyan, V., Ullrich, A., Schlessinger, J., et al. (1994). Structure of the N-terminal SH3 domain of GRB2 complexed with a peptide from the guanine nucleotide releasing factor Sos. Nat. Struct. Biol. 1, 891±897. 113. Wittekind, M., Mapelli, C., Farmer, B. T., II Suen, K. L., Goldfarb, V., Tsao, J., Lavoie, T., Barbacid, M., Meyers, C. A., and Mueller, L. (1994). Orientation of peptide

262

114.

115.

116.

117.

118.

119. 120.

121.

122.

123. 124.

125.

126.

127. 128. 129.

ANDREA MUSACCHIO

fragments from Sos proteins bound to the N-terminal SH3 domain of Grb2 determined by NMR spectroscopy. Biochemistry 33, 13531±13539. Feller, S. M., Posern, G., Voss, J., Kardinal, C., Sakkab, D., Zheng, J., and Knudsen, B. S. (1998). Physiological signals and oncogenesis mediated through Crk family adapter proteins. J. Cell Physiol. 177, 535±552. Knudsen, B. S., Feller, S. M., and Hanafusa, H. (1994). Four proline-rich sequences of the guanine-nucleotide exchange factor C3G bind with unique speci®city to the ®rst Src homology 3 domain of Crk. J. Biol. Chem. 269, 32781±32787. Tanaka, S., Morishita, T., Hashimoto, Y., Hattori, S., Nakamura, S., Shibuya, M., Matuoka, K., Takenawa, T., Kurata, T., Nagashima, K., et al. (1994). C3G, a guanine nucleotide-releasing protein expressed ubiquitously, binds to the Src homology 3 domains of CRK and GRB2/ASH proteins. Proc. Natl. Acad. Sci. USA 91, 3443±3447. Knudsen, B. S., Zheng, J., Feller, S. M., Mayer, J. P., Burrell, S. K., Cowburn, D., and Hanafusa, H. (1995). Af®nity and speci®city requirements for the ®rst Src homology 3 domain of the Crk proteins. EMBO J. 14, 2191±2198. Saksela, K., Cheng, G., and Baltimore, D. (1995). Proline-rich (PxxP) motifs in HIV1 Nef bind to SH3 domains of a subset of Src kinases and are required for the enhanced growth of Nef viruses but not for down-regulation of CD4. EMBO J. 14, 484±491. Cullen, B. R. (1998). HIV-1 auxiliary proteins: Making connections in a dying cell. Cell 93, 685±692. Lee, C. H., Leung, B., Lemmon, M. A., Zheng, J., Cowburn, D., Kuriyan, J., and Saksela, K. (1995). A single amino acid in the SH3 domain of Hck determines its high af®nity and speci®city in binding to HIV-1 Nef protein. EMBO J. 14, 5006±5015. Arold, S., O'Brien, R., Franken, P., Strub, M. P., Hoh, F., Dumas, C., and Ladbury, J. E. (1998). RT loop ¯exibility enhances the speci®city of Src family SH3 domains for HIV1 Nef. Biochemistry 37, 14683±14691. Arold, S., Franken, P., Strub, M. P., Hoh, F., Benichou, S., Benarous, R., and Dumas, C. (1997). The crystal structure of HIV-1 Nef protein bound to the Fyn kinase SH3 domain suggests a role for this complex in altered T cell receptor signaling. Structure 5, 1361±1372. Lim, W. A. (1996). Reading between the lines: SH3 recognition of an intact protein. Structure 4, 657±659. Collette, Y., Arold, S., Picard, C., Janvier, K., Benichou, S., Benarous, R., Olive, D., and Dumas, C. (2000). HIV-2 and SIV nef proteins target different Src family SH3 domains than does HIV-1 Nef because of a triple amino acid substitution. J. Biol. Chem. 275, 4171±4176. Hiipakka, M., Poikonen, K., and Saksela, K. (1999). SH3 domains with high af®nity and engineered ligand speci®cities targeted to HIV-1 Nef. J. Mol. Biol. 293, 1097± 1106. Grzesiek, S., Bax, A., Clore, G. M., Gronenborn, A. M., Hu, J. S., Kaufman, J., Palmer, I., Stahl, S. J., and Wing®eld, P. T. (1996). The solution structure of HIV-1 Nef reveals an unexpected fold and permits delineation of the binding surface for the SH3 domain of Hck tyrosine protein kinase. Nat. Struct. Biol. 3, 340±345. Laird, A. D., and Shalloway, D. (1997). Oncoprotein signalling and mitosis. Cell Signal. 9, 249±255. Brown, M. T., and Cooper, J. A. (1996). Regulation, substrates and functions of src. Biochim. Biophys. Acta. 1287, 121±149. Kato, J. Y., Takeya, T., Grandori, C., Iba, H., Levy, J. B., and Hanafusa, H. (1986). Amino acid substitutions suf®cient to convert the nontransforming p60c-src protein to a transforming protein. Mol. Cell. Biol. 6, 4155±4160.

HOW SH3 DOMAINS RECOGNIZE PROLINE

263

130. Potts, W. M., Reynolds, A. B., Lansing, T. J., and Parsons, J. T. (1988). Activation of pp60c-src transforming potential by mutations altering the structure of an amino terminal domain containing residues 90±95. Oncogene Res. 3, 343±355. 131. Murphy, S. M., Bergman, M., and Morgan, D. O. (1993). Suppression of c-Src activity by C-terminal Src kinase involves the c-Src SH2 and SH3 domains: Analysis with Saccharomyces cerevisiae. Mol. Cell. Biol. 13, 5290±5300. 132. Superti-Furga, G., Fumagalli, S., Koegl, M., Courtneidge, S. A., and Draetta, G. (1993). Csk inhibition of c-Src activity requires both the SH2 and SH3 domains of Src. EMBO J. 12, 2625±2634. 133. Okada, M., Howell, B. W., Broome, M. A., and Cooper, J. A. (1993). Deletion of the SH3 domain of Src interferes with regulation by the phosphorylated carboxyl-terminal tyrosine. J. Biol. Chem. 268, 18070±18075. 134. Erpel, T., Superti-Furga, G., and Courtneidge, S. A. (1995). Mutational analysis of the Src SH3 domain: the same residues of the ligand binding surface are important for intra-and intermolecular interactions. EMBO J. 14, 963±975. 135. Xu, W., Doshi, A., Lei, M., Eck, M. J., and Harrison, S. C. (1999). Crystal structures of c-Src reveal features of its autoinhibitory mechanism. Mol. Cell 3, 629±638. 136. Superti-Furga, G., and Gon¯oni, S. (1997). A crystal milestone: The structure of regulated Src. Bioessays 19, 447±450. 137. Sicheri, F., and Kuriyan, J. (1997). Structures of Src-family tyrosine kinases. Curr. Opin. Struct. Biol. 7, 777±785. 138. Pawson, T. (1997). New impressions of Src and Hck. Nature 385, 582±585. 139. Nguyen, J. T., and Lim, W. A. (1997). How Src exercises self-restraint. Nat. Struct. Biol. 4, 256±260. 140. Mayer, B. J. (1997). Signal transduction: Clamping down on Src activity. Curr. Biol. 7, R295±R298. 141. Shalloway, D., and Taylor, S. J. (1997). Src: More than the sum of its parts. Trends Cell Biol. 7, 215±217. 142. Williams, J. C., Wierenga, R. K., and Saraste, M. (1998). Insights into Src kinase functions: Structural comparisons. Trends Biochem. Sci. 23, 179±184. 143. Young, M. A., Gon¯oni, S., Superti-Furga, G., Roux, B., and Kuriyan, J. (2001). Dynamic coupling between the SH2 and SH3 domains of c-Src and Hck underlies their inactivation by C-terminal tyrosine phosphorylation. Cell 105, 115± 126. 144. Pellicena, P., and Miller, W. T. (2001). Processive phosphorylation of p130Cas by Src depends on SH3±polyproline interactions. J. Biol. Chem. 276, 28190±28196. 145. Scott, M. P., and Miller, W. T. (2000). A peptide model system for processive phosphorylation by Src family kinases. Biochemistry 39, 14531±14537. 146. Moare®, I., LaFevre-Bernt, M., Sicheri, F., Huse, M., Lee, C. H., Kuriyan, J., and Miller, W. T. (1997). Activation of the Src-family tyrosine kinase Hck by SH3 domain displacement. Nature 385, 650±653. 147. Gon¯oni, S., Williams, J. C., Hattula, K., Weijland, A., Wierenga, R. K., and SupertiFurga, G. (1997). The role of the linker between the SH2 domain and catalytic domain in the regulation and function of Src. EMBO J. 16, 7261±7271. 148. Gon¯oni, S., Frischknecht, F., Way, M., and Superti-Furga, G. (1999). Leucine 255 of Src couples intramolecular interactions to inhibition of catalysis. Nat. Struct. Biol. 6, 760±764. 149. LaFevre-Bernt, M., Sicheri, F., Pico, A., Porter, M., Kuriyan, J., and Miller, W. T. (1998). Intramolecular regulatory interactions in the Src family kinase Hck probed by mutagenesis of a conserved tryptophan residue. J. Biol. Chem. 273, 32129±32134.

264

ANDREA MUSACCHIO

150. Gon¯oni, S., Weijland, A., Kretzschmar, J., and Superti-Furga, G. (2000). Crosstalk between the catalytic and regulatory domains allows bidirectional regulation of Src. Nat. Struct. Biol. 7, 281±286. 151. Barila, D., and Superti-Furga, G. (1998). An intramolecular SH3-domain interaction regulates c-Abl activity. Nat. Genet. 18, 280±282. 152. Bunnell, S. C., Henry, P. A., Kolluri, R., Kirchhausen, T., Rickles, R. J., and Berg, L. J. (1996). Identi®cation of Itk/Tsk Src homology 3 domain ligands. J. Biol. Chem. 271, 25646±25656. 153. Andreotti, A. H., Bunnell, S. C., Feng, S., Berg, L. J., and Schreiber, S. L. (1997). Regulatory intramolecular association in a tyrosine kinase of the Tec family. Nature 385, 93±97. 154. Harpur, A. G., Layton, M. J., Das, P., Bottomley, M. J., Panayotou, G., Driscoll, P. C., and Water®eld, M. D. (1999). Intermolecular interactions of the p85alpha regulatory subunit of phosphatidylinositol 3-kinase. J. Biol. Chem. 274, 12323±12332. 155. Musacchio, A., Cantley, L. C., and Harrison, S. C. (1996). Crystal structure of the breakpoint cluster region-homology domain from phosphoinositide 3-kinase p85 alpha subunit. Proc. Natl. Acad. Sci. USA 93, 14373±14378. 156. McGee, A. W., and Bredt, D. S. (1999). Identi®cation of an intramolecular interaction between the SH3 and guanylate kinase domains of PSD-95. J. Biol. Chem. 274, 17431± 17436. 157. Shin, H., Hsueh, Y. P., Yang, F. C., Kim, E., and Sheng, M. (2000). An intramolecular interaction between Src homology 3 domain and guanylate kinase-like domain required for channel clustering by postsynaptic density-95/SAP90. J. Neurosci. 20, 3580±3587. 158. Woods, D. F., Hough, C., Peel, D., Callaini, G., and Bryant, P. J. (1996). Dlg protein is required for junction structure, cell polarity, and proliferation control in Drosophila epithelia. J. Cell Biol. 134, 1469±1482. 159. Wu, H., Reissner, C., Kuhlendahl, S., Coblentz, B., Reuver, S., Kindler, S., Gundel®nger, E. D., and Garner, C. C. (2000). Intramolecular interactions regulate SAP97 binding to GKAP. EMBO J. 19, 5740±5751. 160. Nix, S. L., Chishti, A. H., Anderson, J. M., and Walther, Z. (2000). hCASK and hDlg associate in epithelia, and their src homology 3 and guanylate kinase domains participate in both intramolecular and intermolecular interactions. J. Biol. Chem. 275, 41192±41200. 161. Kato, M., Miyazawa, K., and Kitamura, N. (2000). A deubiquitinating enzyme UBPY interacts with the Src homology 3 domain of Hrs-binding protein via a novel binding motif PX(V/I)(D/N)RXXKP. J. Biol. Chem. 275, 37481±37487. 162. Kang, H., Freund, C., Duke-Cohan, J. S., Musacchio, A., Wagner, G., and Rudd, C. E. (2000). SH3 domain recognition of a proline-independent tyrosine-based RKxxYxxY motif in immune cell adaptor SKAP55. EMBO J. 19, 2889±2899. 163. Urquhart, A. J., Kennedy, D., Gould, S. J., and Crane, D. I. (2000). Interaction of Pex5p, the type 1 peroxisome targeting signal receptor, with the peroxisomal membrane proteins Pex14p and Pex13p. J. Biol. Chem. 275, 4127±4136. 164. Barnett, P., Bottger, G., Klein, A. T., Tabak, H. F., and Distel, B. (2000). The peroxisomal membrane protein Pex13p shows a novel mode of SH3 interaction. EMBO J. 19, 6382±6391. 165. Scita, G., Nordstrom, J., Carbone, R., Tenca, P., Giardina, G., Gutkind, S., Bjarnegard, M., Betsholtz, C., and Di Fiore, P. P. (1999). EPS8 and E3B1 transduce signals from Ras to Rac. Nature 401, 290±293. 166. Kaelin, W. G., Jr. (1999). The p53 gene family. Oncogene 18, 7701±7705. 167. Iwabuchi, K., Bartel, P. L., Li, B., Marraccino, R., and Fields, S. (1994). Two cellular proteins that bind to wild-type but not mutant p53. Proc. Natl. Acad. Sci. USA 91, 6098± 6102.

HOW SH3 DOMAINS RECOGNIZE PROLINE

265

168. Cho, Y., Gorina, S., Jeffrey, P. D., and Pavletich, N. P. (1994). Crystal structure of a p53 tumor suppressor-DNA complex: Understanding tumorigenic mutations. Science 265, 346±355. 169. Cho, Y., Gorina, S., Jeffrey, P. D., and Pavletich, N. P. (1994). Crystal structure of a p53 tumor suppressor±DNA complex: Understanding tumorigenic mutations. Science 265, 346±355. 170. Nishida, M., Nagata, K., Hachimori, Y., Horiuchi, M., Ogura, K., Mandiyan, V., Schlessinger, J., and Inagaki, F. (2001). Novel recognition mode between Vav and Grb2 SH3 domains. EMBO J. 20, 2995±3007. 171. Park, H., Wahl, M. I., Afar, D. E., Turck, C. W., Rawlings, D. J., Tam, C., Scharenberg, A. M., Kinet, J. P., and Witte, O. N. (1996). Regulation of Btk function by a major autophosphorylation site within the SH3 domain. Immunity 4, 515±525. 172. Morrogh, L. M., Hinshelwood, S., Costello, P., Cory, G. O., and Kinnon, C. (1999). The SH3 domain of Bruton's tyrosine kinase displays altered ligand binding properties when auto-phosphorylated in vitro. Eur. J. Immunol. 29, 2269±2279. 173. Zhao, Z., Manser, E., and Lim, L. (2000). Interaction between PAK and nck: A template for nck targets and role of PAK autophosphorylation. Mol. Cell. Biol. 20, 3906±3917. 174. Rozakis-Adcock, M., van der Geer, P., Mbamalu, G., and Pawson, T. (1995). MAP kinase phosphorylation of mSos1 promotes dissociation of mSos1±Shc and mSos1± EGF receptor complexes. Oncogene 11, 1417±1426. 175. Corbalan-Garcia, S., Yang, S. S., Degenhardt, K. R., and Bar-Sagi, D. (1996). Identi®cation of the mitogen-activated protein kinase phosphorylation sites on human Sos1 that regulate interaction with Grb2. Mol. Cell. Biol. 16, 5674±5682. 176. Bedford, M. T., Frankel, A., Yaffe, M. B., Clarke, S., Leder, P., and Richard, S. (2000). Arginine methylation inhibits the binding of proline-rich ligands to src homology 3, but not WW, domains. J. Biol. Chem. 275, 16030±16036. 177. Huang, J., and Kleinberg, M. E. (1999). Activation of the phagocyte NADPH oxidase protein p47(phox). Phosphorylation controls SH3 domain-dependent binding to p22(phox). J. Biol. Chem. 274, 19731±19737. 178. Hinshaw, J. E. (2000). Dynamin and its role in membrane ®ssion. Annu. Rev. Cell Dev. Biol. 16, 483±519. 179. Nonoyama, S., and Ochs, H. D. (1998). Characterization of the Wiskott±Aldrich syndrome protein and its role in the disease. Curr. Opin. Immunol. 10, 407±412. 180. Sever, S., Damke, H., and Schmid, S. L. (2000). Garrotes, springs, ratchets, and whips: Putting dynamin models to the test. Traf®c 1, 385±392. 181. Okamoto, P. M., Herskovits, J. S., and Vallee, R. B. (1997). Role of the basic, prolinerich region of dynamin in Src homology 3 domain binding and endocytosis. J. Biol. Chem. 272, 11629±11635. 182. Gout, I., Dhand, R., Hiles, I. D., Fry, M. J., Panayotou, G., Das, P., Truong, O., Totty, N. F., Hsuan, J., Booker, G. W., et al. (1993). The GTPase dynamin binds to and is activated by a subset of SH3 domains. Cell 75, 25±36. 183. Shupliakov, O., Low, P., Grabs, D., Gad, H., Chen, H., David, C., Takei, K., De Camilli, P., and Brodin, L. (1997). Synaptic vesicle endocytosis impaired by disruption of dynamin±SH3 domain interactions. Science 276, 259±263. 184. Morton, C. J., and Campbell, I. D. (1994). SH3 domains: Molecular `Velcro.' Curr. Biol. 4, 615±617. 185. Kuriyan, J., and Cowburn, D. (1997). Modular peptide recognition domains in eukaryotic signaling. Annu. Rev. Biophys. Biomol. Struct. 26, 259±288. 186. Moare®, I., LaFevre-Bernt, M., Sicheri, F., Huse, M., Lee, C. H., Kuriyan, J., and Miller, W. T. (1997). Activation of the Src-family tyrosine kinase Hck by SH3 domain displacement. Nature 385, 650±653.

266

ANDREA MUSACCHIO

187. Dalgarno, D. C., Bot®eld, M. C., and Rickles, R. J. (1997). SH3 domains and drug design: Ligands, structure, and biological function. Biopolymers 43, 383±400. 188. Kardinal, C., Konkol, B., Schulz, A., Posern, G., Lin, H., Adermann, K., Eulitz, M., Estrov, Z., Talpaz, M., Arlinghaus, R. B., and Feller, S. M. (2000). Cell-penetrating SH3 domain blocker peptides inhibit proliferation of primary blast cells from CML patients. FASEB J. 14, 1529±1538. 189. Schumacher, T. N., Mayr, L. M., Minor, D. L., Jr., Milhollen, M. A., Burgess, M. W., and Kim, P. S. (1996). Identi®cation of D-peptide ligands through mirror-image phage display. Science 271, 1854±1857. 190. Combs, A. P., Kapoort, T. M., Feng, S., Chen, J. K., Daude-Snow, L. F., and Schreiber, S. L. (1996). Protein structure-based combinatorial chemistry: Discovery of non-peptide binding elements to Src SH3 domain. J. Am. Chem. Soc. 118, 287±288. 191. Feng, S., and Schreiber, S. L. (1997). Enantiomeric binding elements interacting at the same site of an SH3 protein receptor. J. Am. Chem. Soc. 119, 10873±10874. 192. Morken, J. P., Kapoor, T. M., Feng, S., Shirai, F., and Schreiber, S. L. (1998). Exploring the leucine±proline binding pocket of the Src SH3 domain using structure-based, split-pool synthesis and af®nity-based selection. J. Am. Chem. Soc. 120, 30± 36. 193. Kapoor, T. M., Andreotti, A. H., and Schreiber, S. L. (1998). Exploring the speci®city pockets of two homologous SH3 domains using structure-based, split-pool synthesis and af®nity-based selection. J. Am. Chem. Soc. 120, 23±29. 194. Barfod, E. T., Zheng, Y., Kuang, W. J., Hart, M. J., Evans, T., Cerione, R. A., and Ashkenazi, A. (1993). Cloning and expression of a human CDC42 GTPase-activating protein reveals a functional SH3±binding domain. J. Biol. Chem. 268, 26059±26062. 195. Weng, Z., Thomas, S. M., Rickles, R. J., Taylor, J. A., Brauer, A. W., Seidel-Dugan, C., Michael, W. M., Dreyfuss, G., and Brugge, J. S. (1994). Identi®cation of Src, Fyn, and Lyn SH3-binding proteins: Implications for a function of SH3 domains. Mol. Cell. Biol. 14, 4509±4521. 196. Kapeller, R., Prasad, K. V., Janssen, O., Hou, W., Schaffhausen, B. S., Rudd, C. E., and Cantley, L. C. (1994). Identi®cation of two SH3-binding motifs in the regulatory subunit of phosphatidylinositol 3±kinase. J. Biol. Chem. 269, 1927±1933. 197. Holmes, T. C., Fadool, D. A., Ren, R., and Levitan, I. B. (1996). Association of Src tyrosine kinase with a human potassium channel mediated by SH3 domain. Science 274, 2089±2091. 198. Nakamoto, T., Sakai, R., Ozawa, K., Yazaki, Y., and Hirai, H. (1996). Direct binding of C-terminal region of p130Cas to SH2 and SH3 domains of Src kinase. J. Biol. Chem. 271, 8959±8965. 199. Alexandropoulos, K., Cheng, G., and Baltimore, D. (1995). Proline-rich sequences that bind to Src homology 3 domains with individual speci®cities. Proc. Natl. Acad. Sci. USA 92, 3110±3114. 200. Karlsson, T., Songyang, Z., Landgren, E., Lavergne, C., Di Fiore, P. P., Ana®, M., Pawson, T., Cantley, L. C., Claesson-Welsh, L., and Welsh, M. (1995). Molecular interactions of the Src homology 2 domain protein Shb with phosphotyrosine residues, tyrosine kinase receptors and Src homology 3 domain proteins. Oncogene 10, 1475± 1483. 201. Onofri, F., Giovedi, S., Vaccaro, P., Czernik, A. J., Valtorta, F., De Camilli, P., Greengard, P., and Benfenati, F. (1997). Synapsin I interacts with c-Src and stimulates its tyrosine kinase activity. Proc. Natl. Acad. Sci. USA 94, 12168±12173. 202. Weng, Z., Taylor, J. A., Turner, C. E., Brugge, J. S., and Seidel-Dugan, C. (1993). Detection of Src homology 3-binding proteins, including paxillin, in normal and v-Srctransformed Balb/c 3T3 cells. J. Biol. Chem. 268, 14956±14963.

HOW SH3 DOMAINS RECOGNIZE PROLINE

267

203. Flynn, D. C., Leu, T. H., Reynolds, A. B., and Parsons, J. T. (1993). Identi®cation and sequence analysis of cDNAs encoding a 110-kilodalton actin ®lament-associated pp60src substrate. Mol. Cell. Biol. 13, 7892±7900. 204. Cheng, G., Ye, Z. S., and Baltimore, D. (1994). Binding of Bruton's tyrosine kinase to Fyn, Lyn, or Hck through a Src homology 3 domain-mediated interaction. Proc. Natl. Acad. Sci. USA 91, 8152±8155. 205. Li, E., Stupack, D. G., Brown, S. L., Klemke, R., Schlaepfer, D. D., and Nemerow, G. R. (2000). Association of p130CAS with phosphatidylinositol-3±OH kinase mediates adenovirus cell entry. J. Biol. Chem. 275, 14729±14735. 206. Manser, E., Loo, T. H., Koh, C. G., Zhao, Z. S., Chen, X. Q., Tan, L., Tan, I., Leung, T., and Lim, L. (1998). PAK kinases are directly coupled to the PIX family of nucleotide exchange factors. Mol. Cell 1, 183±192. 207. Sudol, M. (1994). Yes-associated protein (YAP65) is a proline-rich phosphoprotein that binds to the SH3 domain of the Yes proto-oncogene product. Oncogene 9, 2145±2152. 208. Glover, R. T., Angiolieri, M., Kelly, S., Monaghan, D. T., Wang, J. Y., Smithgall, T. E., and Buller, A. L. (2000). Interaction of the N-methyl-D-aspartic acid receptor NR2D subunit with the c-Abl tyrosine kinase. J. Biol. Chem. 275, 12725±12729. 209. Shafman, T., Khanna, K. K., Kedar, P., Spring, K., Kozlov, S., Yen, T., Hobson, K., Gatei, M., Zhang, N., Watters, D., Egerton, M., Shiloh, Y., Kharbanda, S., Kufe, D., and Lavin, M. F. (1997). Interaction between ATM protein and c-Abl in response to DNA damage . Nature 387, 520±523. 210. Majidi, M., Hubbs, A. E., and Lichy, J. H. (1998). Activation of extracellular signal-regulated kinase 2 by a novel Abl-binding protein, ST5. J. Biol. Chem. 273, 16608±16614. 211. Ren, R., Ye, Z. S., and Baltimore, D. (1994). Abl protein-tyrosine kinase selects the Crk adapter as a substrate using SH3-binding sites. Genes Dev. 8, 783±794. 212. Schumacher, C., Knudsen, B. S., Ohuchi, T., Di Fiore, P. P., Glassman, R. H., and Hanafusa, H. (1995). The SH3 domain of Crk binds speci®cally to a conserved prolinerich motif in Eps15 and Eps15R. J. Biol. Chem. 270, 15341±15347. 213. Wang, B., Mysliwiec, T., Feller, S. M., Knudsen, B., Hanafusa, H., and Kruh, G. D. (1996). Proline-rich sequences mediate the interaction of the Arg protein tyrosine kinase with Crk. Oncogene 13, 1379±1385. 214. Bokoch, G. M., Wang, Y., Bohl, B. P., Sells, M. A., Quilliam, L. A., and Knaus, U. G. (1996). Interaction of the Nck adapter protein with p21-activated kinase (PAK1). J. Biol. Chem. 271, 25746±25749. 215. Galisteo, M. L., Chernoff, J., Su, Y. C., Skolnik, E. Y., and Schlessinger, J. (1996). The adaptor protein Nck links receptor tyrosine kinases with the serine±threonine kinase Pak1. J. Biol. Chem. 271, 20997±21000. 216. Anton, I. M., Lu, W., Mayer, B. J., Ramesh, N., and Geha, R. S. (1998). The Wiskott± Aldrich syndrome protein-interacting protein (WIP) binds to the adaptor protein Nck. J. Biol. Chem. 273, 20992±20995. 217. Wu, Y., Dowbenko, D., and Lasky, L. A. (1998). PSTPIP 2, a second tyrosine phosphorylated, cytoskeletal-associated protein that binds a PEST-type protein-tyrosine phosphatase. J. Biol. Chem. 273, 30487±30496. 218. Rivero-Lezcano, O. M., Sameshima, J. H., Marcilla, A., and Robbins, K. C. (1994). Physical association between Src homology 3 elements and the protein product of the c-cbl proto-oncogene. J. Biol. Chem. 269, 17363±17366. 219. Carlier, M. F., Nioche, P., Broutin-L'Hermite, I., Boujemaa, R., Le Clainche, C., Egile, C., Garbay, C., Ducruix, A., Sansonetti, P. J., and Pantaloni, D. (2000). GRB2 links signalling to actin assembly by enhancing interaction of neural Wiskott±Aldrich syn-

268

220.

221.

222.

223.

224.

225.

226. 227.

228.

229.

230.

231.

232. 233.

234.

235. 236.

ANDREA MUSACCHIO

drome protein (N-Wasp) with actin-related-protein (ARP2/3) complex. J. Biol. Chem 275, 21946±21952. Finan, P., Shimizu, Y., Gout, I., Hsuan, J., Truong, O., Butcher, C., Bennett, P., Water®eld, M. D., and Kellie, S. (1994). An SH3 domain and proline-rich sequence mediate an interaction between two components of the phagocyte NADPH oxidase complex. J. Biol. Chem. 269, 13752±13755. Harte, M. T., Hildebrand, J. D., Burnham, M. R., Bouton, A. H., and Parsons, J. T. (1996). p130Cas, a substrate associated with v-Src and v-Crk, localizes to focal adhesions and binds to focal adhesion kinase. J. Biol. Chem. 271, 13649±13655. Hansson, H., Mattsson, P. T., Allard, P., Haapaniemi, P., Vihinen, M., Smith, C. I., and Hard, T. (1998). Solution structure of the SH3 domain from Bruton's tyrosine kinase. Biochemistry 37, 2912±2924. Borchert, T. V., Mathieu, M., Zeelen, J. P., Courtneidge, S. A., and Wierenga, R. K. (1994). The crystal structure of human CskSH3: Structural diversity near the RT-Src and n-Src loop. FEBS Lett. 341, 79±85. Kohda, D., Terasawa, H., Ichikawa, S., Ogura, K., Hatanaka, H., Mandiyan, V., Ullrich, A., Schlessinger, J., and Inagaki, F. (1994). Solution structure and ligandbinding site of the carboxy-terminal SH3 domain of GRB2. Structure 2, 1029±1040. Politou, A. S., Millevoi, S., Gautel, M., Kolmerer, B., and Pastore, A. (1998). SH3 in muscles: Solution structure of the SH3 domain from nebulin. J. Mol. Biol. 276, 189±202. Liang, J., Chen, J. K., Schreiber, S. T., and Clardy, J. (1996). Crystal structure of P13K SH3 domain at 2.0 angstroms resolution. J. Mol. Biol. 257, 632±643. Viguera, A. R., Blanco, F. J., and Serrano, L. (1995). The order of secondary structure elements does not determine the structure of a protein but does affect its folding kinetics. J. Mol. Biol. 247, 670±681. Blanco, F. J., Ortiz, A. R., and Serrano, L. (1997). 1 H and 15 N NMR assignment and solution structure of the SH3 domain of spectrin: Comparison of unre®ned and re®ned structure sets with the crystal structure. J. Biomol. NMR 9, 347±357. Martinez, J. C., Pisabarro, M. T., and Serrano, L. (1998). Obligatory steps in protein folding and the conformational diversity of the transition state. Nat. Struct. Biol. 5, 721±729. Gosser, Y. Q., Zheng, J., Overduin, M., Mayer, B. J., and Cowburn, D. (1995). The solution structure of Abl SH3, and its relationship to SH2 in the SH(32) construct. Structure 3, 1075±1086. Nam, H. J., Haser, W. G., Roberts, T. M., and Frederick, C. A. (1996). Intramolecular interactions of the regulatory domains of the Bcr-Abl kinase reveal a novel control mechanism. Structure 4, 1105±1114. Eck, M. J., Atwell, S. K., Shoelson, S. E., and Harrison, S. C. (1994). Structure of the regulatory domains of the Src-family tyrosine kinase Lck. Nature 368, 764±769. Tzeng, S. R., Lou, Y. C., Pai, M. T., Jain, M. L., and Cheng, J. W. (2000). Solution structure of the human BTK SH3 domain complexed with a proline-rich peptide from p120cbl. J. Biomol. NMR 16, 303±312. Horita, D. A., Baldisseri, D. M., Zhang, W., Altieri, A. S., Smithgall, T. E., Gmeiner, W. H., and Byrd, R. A. (1998). Solution structure of the human Hck SH3 domain and identi®cation of its ligand binding site. J. Mol. Biol. 278, 253±265. Carson, M. (1991). Ribbons 2.0. J. Appl. Crystallogr. 24, 958±961. Nicholls, A., Sharp, K. A., and Honig, B. (1991). Protein folding and association: Insights from the interfacial and thermodynamic properties of hydrocarbons. Proteins 11, 281±296.

STRUCTURAL BIOLOGY OF eIF4F: mRNA RECOGNITION AND PREPARATION IN EUKARYOTIC TRANSLATION INITIATION BY JOSEPH MARCOTRIGIANO*,² ,1 AND STEPHEN K. BURLEY*,³ ,2 *Laboratory of Molecular Biophysics,² Laboratory of Virology and Infectious Disease, and Howard Hughes Medical Institute, The Rockefeller University, New York, New York 10021

³

I. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . II. Structural Features of Eukaryotic mRNAs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A. 50 ,7-Methyl-G mRNA Cap . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B. Translation Start Site . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . C. 30 Poly(A) Tail . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . III. General Mechanisms of Cellular, Cap-Dependent Translation Initiation. . . . . . A. Recognition of 50 ,7-Methyl-G Cap Structure . . . . . . . . . . . . . . . . . . . . . . . . . . B. Regulation of eIF4F by Molecular Mimicry . . . . . . . . . . . . . . . . . . . . . . . . . . IV. HEAT Repeats within eIF4G Direct Assembly of Translation Machinery . . . . . V. Preparation of the mRNA 50 -UTR for Small Ribosomal Subunit Binding . . . . VI. Conclusion and Perspectives. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

269 271 271 273 274 274 276 279 283 288 293 294

I. INTRODUCTION Protein synthesis is an intricate, multistep biochemical process, which has been conserved across all three kingdoms of life. Common features of translation initiation include preparation and recruitment of the small ribosomal subunit to the mRNA start site, anticodon docking of the aminoacylated methionyl initiator tRNA to the start codon, joining of the large ribosomal subunit, and enzyme-catalyzed formation of the ®rst peptide bond. Thereafter, the ribosome translocates in a 50 to 30 direction along the mRNA to permit catalytic addition of the next amino acid to the growing polypeptide chain. On reaching the termination codon the large and small ribosomal subunits dissociate from the mRNA and are then recycled to permit another round of translation initiation. Despite evident similarities in mRNA translation, the molecular mechanisms underlying ribosomal recruitment and start site selection differ substantially between bacteria and eukaryotes. In most prokaryotes protein synthesis and mRNA synthesis occur simultaneously, whereas eukaryotes have decoupled gene expression by localizing transcription 1 Present address: Laboratory of Virology and Infectious Disease, Centre for the Study of Hepatitis C, Rockfeller University, New York, New York 10021. 2 Present address: Structural Genomix, Inc., San Diego, California 92121.

269 ADVANCES IN PROTEIN CHEMISTRY, Vol. 61

Copyright 2003, Elsevier Science (USA). All rights reserved. 0065-3233/03 $35.00

270

JOSEPH MARCOTRIGIANO AND STEPHEN K. BURLEY

to the nucleus and translation to the cytoplasm. Spatiotemporal separation has added extra levels of complexity to the eukaryotic system, creating opportunities for myriad strategies of translational regulation. Prokaryotic mRNAs contain a Shine±Dalgarno sequence located 7±10 nucleotides upstream of the translation start site. The 16S rRNA, a component of the small ribosomal subunit, forms Watson±Crick basepairs with the Shine± Dalgarno sequence to position the ribosome appropriately (Steitz and Jakes, 1975). In contrast, the overwhelming majority of eukaryotic mRNAs possess a 50 cap structure and a 30 polyadenylic acid [poly(A) ] tail that synergize in stimulating translation initiation (Sachs, 2000). Eukaryotic 40S ribosomal subunits do not bind directly to mRNA, but instead rely on various accessory proteins, known as eukaryotic translation initiation factors, or eIFs, that recognize various structural features of the mRNA and/or other proteins and recruit the small subunit to the 50 -untranslated region (50 -UTR). Once bound to the mRNA, the eukaryotic small ribosomal subunit is thought to scan along the 50 -UTR until it encounters the proper translation start site, where subunit joining occurs to give a fully assembled ribosome±mRNA complex competent for protein synthesis. Like transcription, translation provides important entry points for the regulation of eukaryotic gene expression. Control of gene expression at the level of translation offers several advantages over transcriptional regulation, such as a much more rapid response to environmental stimuli, localization of protein synthesis to speci®c subcellular regions, and creation of protein gradients within the cell. Biologically signi®cant examples of translational control of gene expression include embryonic development in model organisms (e.g., ¯y and worm) and mammalian reticulocyte maturation. Invertebrate oocytes are largely quiescent with very low levels of gene expression. After fertilization, the rate of protein synthesis increases with little or no change in mRNA synthesis. Early in development there continues to be little or no transcription, and morphogenetic changes (including cell division) are controlled by differential translation of select maternal mRNAs. Human red blood cells provide another example of a transcriptionally silent cell. During the ®nal steps of erythrocyte maturation, mammalian reticulocytes become enucleated yet continue to synthesize proteins (particularly a and b globins) in the absence of transcription. Translational control in eukaryotes has been observed at all three mechanistic stages (initiation, elongation, and termination) of this complicated process. Not surprisingly, however, rate-limiting steps are thought to occur primarily during the initiation phase. For example, entry into and transit through the G1 phase of the cell cycle are correlated with increased rates of protein synthesis that have been attributed to upregulation of translation initiation (Sonenberg, 1996). This chapter

STRUCTURAL BIOLOGY OF eIF4F

271

provides a detailed account of the results of structural and biochemical studies of some of the earliest steps in translation initiation in eukaryotes, which involve recognition and preparation of the mRNA by eIF4F. We begin with a description of eukaryotic mRNA structure and then discuss our current structural and biochemical understanding of how mRNAs are recognized and prepared for small ribosomal subunit binding during both cap-dependent and cap-independent initiation of protein synthesis.

II. STRUCTURAL FEATURES OF EUKARYOTIC mRNAS A. 50 ,7-Methyl-G mRNA Cap All cellular, eukaryotic mRNAs (excluding organellar mRNAs) contain a cap structure at their 50 termini. Capping of RNA polymerase II (pol II) transcripts follows synthesis of the ®rst 25±30 nucleotides (Coppola et al., 1983; Hagler and Shuman, 1992). The capping machinery initially removes the g-phosphate from the 50 end of the pre-mRNA and catalyzes the addition of GMP. Subsequently, this terminal guanosine is methylated at the seventh position of the base by guanine-7-methyltransferase, using S-adenosylmethionine as the methyl group donor (Pillutla et al., 1998). The ®nal product is a guanosine, methylated at position 7 of the base, connected by a 50 to 50 triphosphate linkage to the ®rst nucleotide of the message (7methyl-GpppN, where N is any nucleotide). Alkylation at the seventh position produces a delocalized positive charge on the double-ring system. A schematic diagram of the 50 ,7-methyl-G mRNA cap structure is shown in Fig. 1. The 50 to 50 linkage differs from the usual 50 to 30 linkage of nucleic acids and produces a polynucleotide lacking an exposed 50 hydroxyl group. mRNAs in higher eukaryotes (excluding fungi and plants) can undergo additional methylation events, involving the 20 hydroxyl groups of the ribose sugars within the ®rst few bases. The following nomenclature has been established to describe these additional chemical modi®cations: 7-methyl-GpppN (cap 0), 7-methyl-GpppNm (cap 1), 7methyl-GpppNmpNm (cap 2), m7GpppNmpNmpNm (cap 3), and 7-methyl-GpppNmpNmpNmpNmpN (cap 4) (where m is a methyl group at the 20 position). At present, there is no de®nitive evidence that these additional methylation events are of biological importance. There are, however, some data to suggest that cap-speci®c 20 -O-methylation (cap 1 formation) is correlated with enhanced translation initiation during development (Muthukrishnan et al., 1976; Kuge et al., 1998). Given that cap structures are found at the 50 ends of all cellular eukaryotic mRNAs, it is not surprising that they are functionally important in translation initiation. Initial studies using reovirus and vesicular

272

JOSEPH MARCOTRIGIANO AND STEPHEN K. BURLEY

FIG. 1. 50 ,7-Methyl-G mRNA cap structure. Arrows depict the locations of additional methyl groups in cap 1 and cap 2 structures.

stomatitis virus mRNAs demonstrated a relationship between mRNA methylation and stimulation of protein synthesis (Both et al., 1975). Reovirus mRNAs containing 7-methyl-GpppN on the 50 terminus were preferentially translated over RNAs bearing GpppG and ppG 50 termini in extracts prepared from wheat germ and mouse L cells (Muthukrishnan et al., 1975). Nuclease digestion experiments demonstrated that the small ribosomal subunit protects only capped mRNAs, suggesting that discrimination of methylated versus unmethylated caps occurs during the recruitment stage of translation initiation (Both et al., 1975). These ®ndings established the 50 ,7-methyl-G cap as a critical mRNA feature for translation initiation in eukaryotes. The biochemical function of the cap structure is correlated with the presence of a delocalized positive charge on the base. Inhibition studies using cap analogues demonstrated that 7-methyl-, 7-ethyl-, and 7-benzylGDP, but not GDP, abrogated binding of mRNA to the ribosome (Adams et al., 1978). Moreover, synthetic mRNAs with ethyl, benzyl, and 2phenylethyl substituents at position 7 can support translation initiation in vitro (Darzynkiewicz et al., 1989). Reduction of 7-methyl-GDP to 8hydro-7-methyl-GDP, which removes the positive charge without affecting the 7-substituent, reverses inhibition (Adams et al., 1978). Although the only natural modi®cation at the seventh position is methylation, the delocalized positive charge (a consequence of any form of alkylation at position 7) contributes to mRNA recognition during translation initiation.

STRUCTURAL BIOLOGY OF eIF4F

273

B. Translation Start Site The small ribosomal subunit uses a unique aminoacylated initiator tRNA Met to recognize the translation start site (AUG). Selection of the decoding start site on a majority of eukaryotic mRNAs follows a two-step mechanism in which the 40S ribosomal subunit binds to the 50 -UTR in the vicinity of the cap structure and then traverses the mRNA to the proper start site. This strategy is commonly referred to as scanning (Kozak and Shatkin, 1978). Separation of the start site from the cap structure, RNA secondary structural elements within the 50 -UTR, and ¯anking sequences around the AUG condon all in¯uence translation ef®ciency (Kozak, 1999). Regions of high sequence complementarity in the vicinity of the cap structure or the translation start site could mask these elements from the eIFs or the ribosomal subunits. In addition, start sites too close to the cap may not be properly loaded into the decoding region of the ribosome or may be blocked by initiation factors, while the presence of an extremely long 50 -UTR increases the probability that the small ribosomal subunit will dissociate from the mRNA during scanning. Sequence motifs ¯anking the AUG can also modulate the ef®ciency of translation start site recognition during scanning. In vertebrates, initiation sites usually contain purines at position 3 and 4 (1 corresponds to the A of the AUG) (Kozak, 1987). Scanning 40S ribosomal subunits will bypass an AUG in a ``poor'' context and initiate at a downstream AUG in a more favorable context. This phenomenon is known as leaky scanning and usually results in two translation products with either different N-termini or different amino acid sequences (Kozak, 1987). A second strategy for start site selection utilized by a few eukaryotic mRNAs embodies features of prokaryotic translation initiation in which the ribosome and/or the initiation machinery is recruited to an internal ribosomal entry site or IRES. As in the prokaryotic system, the small ribosomal subunit can be recruited to an IRES, which lies close to a bona ®de translation start site entailing a minimal scanning requirement. In contrast to the relatively simple Shine±Dalgarno rRNA-binding sites, IRESs are much longer in sequence, are predicted to contain long stretches of complementarity, and are not thought to recruit the ribosome through RNA±rRNA interactions ( Jackson, 2000). The best studied of the IRESs come from the Picornaviridae, which have single-stranded, positive-sense RNA genomes bearing a covalently attached viral protein (VPg) at the 50 end instead of a 7-methyl-G cap. IRES-dependent initiation also occurs with several other viral mRNAs (including hepatitis A, hepatitis C, and some pestiviruses) and with a subset of cellular mRNAs.

274

JOSEPH MARCOTRIGIANO AND STEPHEN K. BURLEY

C. 30 Poly(A) Tail Ribonuclease digestion experiments of polysomal mRNA (i.e., mRNA bound to multiple active ribosomes) identi®ed uninterrupted stretches of poly(A) located at the 30 end of the mRNA (Lim and Canellakis, 1970; Darnell et al., 1971; Edmonds et al., 1971; Lee et al., 1971). These poly(A) tails are found at the 30 ends of most eukaryotic and viral mRNAs. Poly(A) addition occurs via endonucleolytic cleavage of heteronuclear RNA followed by addition of adenine nucleotides (Colgan and Manley, 1997). Poly(A) tails range widely in size (depending on the organism) from 200± 250 adenylate residues in mammals to 60±80 adenylate residues in yeast ( Jacobson, 1996). The length of the poly(A) tail is thought to be correlated with the processivity of the poly(A) polymerase (Sachs and Wahle, 1993). Early attempts to compare translation of poly(A) and poly(A) mRNAs (e.g., histone mRNAs are not polyadenylated) failed to document any differences, probably because of the use of cytoplasmic extracts with suboptimal reinitiation capacity. Subsequent studies, however, established a stimulatory role for poly(A) tails during translation. In 1976, Doel and Carey demonstrated that native ovalbumin mRNA had increases in polysome fraction, rate of peptide elongation, and number of rounds of translation per mRNA compared to its deadenylated form (Doel and Carey, 1976). Addition of exogenous poly(A) RNA inhibits translation of poly(A) mRNA and not poly(A) mRNA ( Jacobson and Favreau, 1983), similar to the way that cap analogues competitively inhibit the translation of capped mRNAs. Several studies documented a positive correlation between the length of the poly(A) tail and translation activity ( Jacobson and Favreau, 1983; Munroe and Jacobson, 1990). Finally, a direct effect of poly(A) tail length on translation initiation has also been observed during early development (Richter, 1996). In the oocyte, several maternal mRNAs have short poly(A) tails and are translationally dormant. Following meiotic maturation or fertilization, these mRNAs become polyadenylated and are then recruited into polysomes.

III. GENERAL MECHANISMS OF CELLULAR, CAP-DEPENDENT TRANSLATION INITIATION Translation initiation of a eukaryotic, cellular mRNA is a complicated process requiring various protein factors to assemble the ribosome at the proper start site (Fig. 2). There are several different models of ribosomal recruitment to the mRNA. The model presented below explains translation of most cellular mRNAs [for alternative mechanistic

275

STRUCTURAL BIOLOGY OF eIF4F

eIF2 eIF3

MET GTP

2 4B

4E 7mG

AUG

eIF1A

40S

4A

4G

1 ATP

IA eIF3

MET GTP

40S

3

43S preinitiation ribosomal complex 4B

4E 7mG

AUG

4A

4G

IA

eIF3

METGTP

40S

ATP

4

eIF1 ATP+Pi

4B

4E 7mG

AUG

4A

4G

eIF3

IA

METGTP 40S

eIF5

5

60S ribosomal subunit

Pi, eIF2.GDP

60S 4B

4E 7mG

AUG

4A

4G

MET 40S

FIG. 2. Mechanism of cap-dependent translation initiation in eukaryotes. Step 1: eIF3, eIF1A, the ternary complex of eIF2, GTP, and aminoacylated initiator tRNA Met bind the 40S ribosomal subunit forming a 43S preinitiation complex. Step 2: eIF4F (eIF4E, eIF4G, and eIF4A) recognize the 50 ,7-methyl-G mRNA cap structure. eIF4A in concert with eIF4B unwinds secondary structural elements in the 50 -UTR in an ATP-dependent manner. Step 3: eIF4G recruits the 43S preinitiation complex to the 50 -UTR forming a 48S preinitiation complex. Step 4: In the presence of eIF1 and eIF1A, the 43S preinitiation complex scans (50 to 30 ) along the mRNA to the translation start site (AUG). Step 5: eIF5 hydrolyzes GTP to dissociate eIF2, eIF3, and eIF1A, allowing for 60S ribosomal subunit joining. Protein synthesis begins with the aminoacylated initiator tRNA Met in the P-site. [Adapted from Merrick and Hershey (1996).]

276

JOSEPH MARCOTRIGIANO AND STEPHEN K. BURLEY

models, see Merrick and Hershey (1996) ]. First, the 40S ribosomal subunit binds eIF3, eIF1A, a ternary complex of eIF2, aminoacylated initiator tRNA Met, and GTP, producing a 43S ribosomal preinitiation complex. eIF4F recognizes the 50 ,7-methyl-G mRNA cap structure and prepares the message for recruitment of the 43S ribosomal preinitiation complex. In higher eukaryotes, eIF4F is a heterotrimer consisting of eIF4E, eIF4G, and eIF4A. eIF4E is a 25-kDa protein that recognizes the mRNA cap. eIF4A is an ATP-dependent, RNA helicase, which in concert with another initiation factor, eIF4B, is thought to unwind secondary structural features in the 50 -UTR. eIF4G acts as a molecular bridge between eIF4E and eIF4A and as a platform for assembly of the translation machinery. The 43S ribosomal preinitiation complex is recruited to the 50 -UTR through interactions between eIF3 and eIF4G (Lamphear et al., 1995; Morino et al., 2000). In the presence of eIF1 and eIF1A, the small ribosomal subunit scans the 50 -UTR in a 50 to 30 direction until it encounters a translation start site in the appropriate context (Kozak, 1989; Pestova et al., 2000). Once the start codon has been engaged, eIF5 stimulates GTP hydrolysis, which leads to the dissociation of eIF2±GDP and eIF3, leaving the aminoacylated initiator tRNA Met in the P site. eIF5B (also known as eukaryotic IF2) completes the formation of the ribosome by facilitating 60S ribosomal subunit joining (Pestova et al., 2000; Roll-Mecak et al., 2000), yielding an 80S ribosomal complex competent for translation initiation. A. Recognition of the 50 ,7-Methyl-G Cap Structure eIF4E is the least abundant of the general translation initiation factors and is considered to be the factor limiting recruitment of the ribosome to the translation start site in most circumstances. eIF4E was ®rst identi®ed as a 25-kDa protein that cross-linked to free 20 ,30 -cis-diols within the 50 ,7methyl-G cap (Sonenberg et al., 1978) and was subsequently puri®ed by af®nity chromatography with 7-methyl-GDP±sepharose (Sonenberg et al., 1979). Cytoplasmic extracts treated with 7-methyl-GDP af®nity resins display dramatic reductions in cap-dependent protein synthesis, which can be restored by addition of recombinant eIF4E (Svitkin et al., 1996). The three-dimensional structures of murine and Saccharomyces cerevisiae eIF4E bound to a 50 mRNA cap analogue, 7-methyl-GDP, were determined by X-ray crystallography and NMR, respectively (Marcotrigiano et al., 1997; Matsuo et al., 1997). The phylogenetically conserved, cap-binding protein consists of a single a=b domain shaped like a cupped hand. Secondary structural elements include three long a-helices (H1±H3), one short a-helix (H4), and a curved, eight-stranded anti-parallel b-sheet (1±8) (Fig. 3A). The cap analogue was found in a narrow ligand-binding cleft

STRUCTURAL BIOLOGY OF eIF4F

277

A H4

B E103

W102 W56

c7 W166

R112

R157

FIG. 3. X-ray structure of the eIF4E/7±methyl-GDP binary complex. (A) Ribbon drawing of the concave cap-binding surface of eIF4E. 7-Methyl-GDP (ball-and-stick) is located in the cap-binding slot. a-Helices are labeled H1±H4 and b-strands are labeled 1±8, with the N- and C-termini labeled with N and C, respectively. The 50 -untranslated region of the mRNA would presumably project down and left of the entrance of the cap-binding slot, overlying a-helix H3 and b-strands S5, S6, and S4. (B) Ribbon drawing of 7-methyl-GDP (ball-and-stick) in the cap-binding slot of eIF4E, showing select residues (ball-and-stick) involved in 50 mRNA cap recognition. The methyl group at the seventh position of the base is labeled with C7. This view is identical to that shown in A. [Adapted from Marcotrigiano et al. (1997).]

278

JOSEPH MARCOTRIGIANO AND STEPHEN K. BURLEY

(cap-binding slot) consisting of the concave surface of the b-sheet, the short a-helix (H4), and the loop between strands S1 and S2. The capbinding slot is closed at one end by the loop connecting strands S3 and S4 and is open at the other end (Figs. 3A and 3B). The base and both phosphate groups (a and b) of 7-methyl-GDP bind to the protein in an extended conformation. The mouse and yeast eIF4E/7±methyl-GDP binary complex structures display identical protein±ligand interactions. The alkylated base is sandwiched between the side chains of two invariant tryptophans, Trp-56 and Trp-102, located on the S1±S2 and S3±S4 loops (for simplicity, murine eIF4E residue numbering will be used throughout). This mode of indole side chain±base interaction was correctly predicted from the results of small-molecule crystallographic work (Ishida et al., 1983, 1988, 1991) and ¯uorescence studies with eIF4E (Ueda et al., 1991). Energetically favorÊ interplanar spacing, akin to able base±aromatic side chain stacking (3.5-A base±base stacking in nucleic acids) arises from interactions between the highest occupied molecular orbital (HOMO) of the indole group of the tryptophan side chain and the lowest unoccupied molecular orbital (LUMO) of the base. The delocalized positive charge on the base, caused by methylation at position 7, lowers the energy of the LUMO of the base, thereby enhancing interactions with the HOMO of the indole group (Ishida et al., 1988). The side chain of a third invariant tryptophan (Trp-166) is located in the same plane as the akylated base and makes a van der Waals contact with the N7-methyl group. eIF4E also embodies another feature of the structure of double-stranded nucleic acids by serving as a molecular mimic of cytosine. O-6 of the base is oriented toward the b-sheet and forms a hydrogen bond with the backbone amino group of Trp-102. N1 and N2 emerge from between the two tryptophan side chains forming hydrogen bonds with the carboxylate oxygen atoms of invariant Glu-103. This hydrogen-bonding network between eIF4E and the guanine is identical to that seen in a Watson± Crick G-C basepair (donor plus two acceptors). The ribose and disphosphate moieties extend away from the methylated purine toward the entrance to the cap-binding slot, terminating shortly before the end of the b-sheet (Fig. 3B). The plane of the ribose group lies almost perpendicular to the plane of the alkylated base. eIF4E makes only one contact with C-10 of the ribose, while the O-20 and O-30 hydroxyl groups are directed out to the solvent. Several positively charged residues (Arg-112, Arg-157, and Lys-162) line the cap-binding slot, presumably providing some degree of neutralization of the negatively charged phosphate groups. The structure of an unrelated cap-binding protein (VP39) has been determined in its apo form and in the presence of various cap analogues

STRUCTURAL BIOLOGY OF eIF4F

279

(Hodel et al., 1996, 1997, 1998). VP39 contributes to maturation of both ends of the vaccinia viral mRNA by catalyzing methylation of 20 hydroxyl groups on ribose sugars in the vicinity of the cap structure (giving rise to higher order cap structures) and by acting as a processivity factor for the viral poly(A) polymerase. VP39 consists of a single a=b domain, composed of a mixed seven-stranded, twisted b-sheet surrounded by parallel a-helices. The structure resembles other methytransferases, with the capbinding slot located in a groove on a ¯attened surface of the molecule. Although the polypeptide chain fold of VP39 differs from that of eIF4E, the mechanism of cap recognition by alkylated base±aromatic side chain stacking and Watson±Crick-type hydrogen bonding is very similar. In the case of VP39, 7-methyl-G is stacked between the side chains of tyrosine and phenylalanine residues, while hydrogen bonds are formed with the side chains of two acidic residues and one bridging water molecule. Although eIF4E and VP39 do not share any structural similarities, they do exploit very similar cap recognition strategies. Thus, 7-methyl-G recognition appears to have arisen via convergent evolution and may be utilized by other cap-binding proteins (Rom et al., 1998). B. Regulation of eIF4F by Molecular Mimicry Because of its critical role in assembling the translation initiation machinery (Fig. 2), the eIF4F heterotrimer represents an important target for regulation of protein synthesis. Formation of an active eIF4F multiprotein complex is suppressed by eIF4E-binding proteins, such as mammalian 4E-BP1, 4E-BP2, and 4E-BP3 (Gingras et al., 1999b) and yeast p20 (Altmann et al., 1997). These negative regulators of protein synthesis compete with eIF4G for binding to eIF4E, thereby inhibiting assembly of the eIF4F heterotrimer (Haghighat et al., 1995; Mader et al., 1995; Altmann et al., 1997). Sequence analyses of the 4E-BPs and the eIF4Gs reveal two conserved protein families that appear unrelated except for the presence of a common Tyr-X-X-X-X-Leu-F segment (where X is variable and F is Leu, Met, or Phe; Fig. 4A) (Mader et al., 1995; Altmann et al., 1997). Treatment of cells with mitogens or growth factors leads to phosphorylation of the 4E-BPs. Hyperphosphorylated 4E-BP1 no longer forms a stable complex with eIF4E, permitting assembly of the eIF4F complex. Insulin as well as other hormones, mitogens, and growth factors increases protein synthesis, at least, in part, by relieving the repressive effect of 4E-BP1 (Lin et al., 1994; Pause et al., 1994). The signal transduction pathway mediating 4E-BP1 phosphorylation has been examined in detail. Both phosphatidylinositol 3-kinase c-Akt and FKBP12-rapamycin-associated protein/mammalian target of rapamycin (FRAP/mTOR) appear to play important roles (Gingras et al., 1999b).

280

JOSEPH MARCOTRIGIANO AND STEPHEN K. BURLEY

FRAP/mTOR phosphorylates 4E-BP1 both in vivo and in vitro, and expression of a rapamycin-resistant form of FRAP/mTOR inhibits 4E-BP1 phosphorylation. 4E-BP1 appears to be phosphorylated at

B

C

N eIF4GII peptide

N eIF4E

N eIF4E

C eIF4GII peptide

N 4E-BP1 peptide

C eIF4E C 4E-BP1 peptide

C eIF4E

FIG. 4. X-ray structures of two eIF4E/7-methyl-GDP/eIF4E-recognition motif ternary complexes. (A) Sequence alignments of the eIF4E-recognition motifs of mammalian 4E-BP1 (Pause et al., 1994), 4E-BP2 (Pause et al., 1994), 4E-BP3 (Poulin et al., 1998), mammalian eIF4GI (Imataka et al., 1998), eIF4GII (Gradi et al., 1998), S. cerevisiae TIF4631 (Goyer et al., 1993), S. cerevisiae TIF4632 (Goyer et al., 1993), S. cerevisiae CAF20 (p20) (Altmann et al., 1997), and wheat germ p82 (Allen et al., 1992). Individual and consensus sequences are numbered with respect to the invariant tyrosine (position 0). Y, invariant tyrosine; L, invariant leucine; F, conserved hydrophobic residue. (B and C) Ribbon drawings of the eIF4E/7-methyl-GDP (ball-and-stick representation)/oligopeptide ternary complex structures viewed perpendicular to the b-sheet of eIF4E showing the eIF4GII oligopeptide (B) and the 4E-BP1 peptide (C) forming similar L-shaped structures that interact with the N-terminus and a-helices (H1 and H2) on the convex dorsal surface of eIF4E. [Adapted from Marcotrigiano et al. (1999).]

STRUCTURAL BIOLOGY OF eIF4F

281

multiple sites, and it is not known which site(s) is responsible for release of eIF4E (Gingras et al., 1999a). A series of biochemical and biophysical studies has demonstrated that oligopeptides encompassing the Tyr-X-X-X-X-Leu-F segments of the 4EBPs and eIF4Gs suf®ce for eIF4E binding and inhibit cap-dependent translation in vitro (Fletcher et al., 1998; Marcotrigiano et al., 1999). NMR studies showed that a small number of residues from full-length, recombinant 4E-BP1 interact with eIF4E, and the 4E-BP1 peptide produced chemical shifts similar to those of the full-length protein (Fletcher et al., 1998). Moreover, the equilibrium dissociation constants for comparable 4E-BP1, eIF4GI, and eIF4GII fragments are within an order of magnitude of the experimental value obtained with full-length 4E-BP1, suggesting that these oligopeptides support most if not all of the energetically signi®cant interactions with eIF4E (Marcotrigiano et al., 1999). The structures of eIF4E plus 7-methyl-GDP interacting with similar fragments of mammalian eIF4GII (active ternary complex) and 4E-BP1 (inhibited ternary complex) have been determined by X-ray crystallography (Figs. 4B and 4C) (Marcotrigiano et al., 1999). These Tyr-X-X-X-XLeu-F segments serve as eIF4E-recognition motifs that form structurally similar eIF4E/7-methyl-GDP/oligopeptide ternary complexes. Both eIF4E-recognition motifs (eIF4GII, Fig. 4B; 4E-BP1, Fig. 4C) adopt an L-shaped extended chain/a-helical structure, which interacts with the convex dorsal surface of eIF4E. The ®rst ®ve residues of the peptides run anti-parallel to the axis of a-helix H2, with the remaining residues forming a short a-helix running perpendicular to and across the surfaces Ê away from the of a-helices H1 and H2. The peptides bind about 35 A cap-binding slot and do not show any direct interactions with either the cap or the cap-binding slot. The structure of the eIF4E/7-methyl-GDP binary complex is unaffected by binding of the 4E-BP1 or eIF4GII peptides. Detailed analyses of cap recognition revealed no signi®cant differences among structures of the murine and yeast eIF4E/7-methylGDP binary complexes (Marcotrigiano et al., 1997; Matsuo et al., 1997) and the active and inhibited ternary complex structures depicted in Figs. 4B and 4C. The eIF4GII and 4E-BP1 oligopeptides adopt similar random coil/ahelical structures and make similar intra- and intermolecular interactions when bound to the convex dorsal surface of eIF4E (Figs. 4B and 4C). The invariant Tyr (0) (residues are identi®ed by their position relative to the invariant Tyr, position 0 in Fig. 4A) of the eIF4E-recognition motif is located on the extended chain portion of the oligopeptide N-terminal to the a-helix. The conserved Leu (5) and F (6) residues give rise to the ®rst turn of the short a-helix. The invariant Tyr (0), Leu (5), and F (6) make

282

JOSEPH MARCOTRIGIANO AND STEPHEN K. BURLEY

multiple contacts with the surface of eIF4E, involving His-37, Pro-38, Val-69, Trp-73, Leu-131, and Leu-135 (bold face type denotes conservation among all known eIF4Es). All eIF4E residues making contacts with Tyr (0), Leu (5), or F (6) are invariant among all published eIF4E sequences except for Leu-135, which is a methionine in Xenopus laevis and wheat germ. The aliphatic portion of conserved hydrophilic side chains at position 9 makes several interactions with Trp-73 and serves as another important anchor to the dorsal surface of eIF4E. Of all the residues on the surface of eIF4E, Trp73 appears to be the most important since it makes several interactions with positions 5, 6, and 9, and substitutions of Trp-73 in yeast and mouse eIF4E completely abolish 4E-BP and eIF4G binding (Ptushkina et al., 1998; Pyronnet et al., 1999). NMR titration studies demonstrated large chemical shift changes for residues 32±50 and 62±79 of yeast eIF4E following addition of full-length 4E-BP2, con®rming that the X-ray structures of the inhibited ternary complex revealed most if not all of the interactions between eIF4E and 4E-BPs (Matsuo et al., 1997). A majority of the contacts between the 4E-BP1 and eIF4GII oligopeptides and eIF4E involve amino acids common to all eIF4Gs and 4E-BPs. In fact, both peptides interact with the same conserved, solvent-accessible surface feature on the convex dorsum of eIF4E using similar intra- and intermolecular contacts. Since the eIF4Gs and the 4E-BPs share no signi®cant sequence homology outside their eIF4Erecognition motifs, the similarity of the ternary complex crystal structures suggests that molecular mimicry of the eIF4Gs by the 4E-BPs arose via convergent evolution. NMR and circular dichroism studies showed that the full-length 4EBPs and the eIF4E-recognition motif oligopeptides are random coil in the absence of eIF4E (Fletcher et al., 1998; Marcotrigiano et al., 1999). These ®ndings suggest that the eIF4E-recognition motifs are unstructured in the absence of eIF4E and undergo a target-directed disorder-toorder transition on binding eIF4E. The same phenomenon has been observed in the regulation of pol II transcription by activation domains that become a-helical on binding to their targets (Kussie et al., 1996; Radhakrishnan et al., 1997; Uesugi et al., 1997). It may be that this molecular recognition strategy is particularly suited to biological control systems dependent on formation of transient protein±protein interactions, because exploitation of a target-directed disorder-to-order structural transition could increase binding speci®city without a concomitant increase in af®nity. We cannot, however, exclude the possibility that the eIF4E-recognition motifs are ordered within a larger portion of eIF4G, in the phosphorylated form of 4E-BP, or when bound to other regulatory factors.

STRUCTURAL BIOLOGY OF eIF4F

283

IV. HEAT REPEATS WITHIN eIF4G DIRECT ASSEMBLY OF TRANSLATION MACHINERY eIF4G is a modular adapter protein (Fig. 5) that plays a pivotal role in coordinating the assembly of translation factors and the small ribosomal subunit during eukaryotic protein synthesis (Hentze, 1997; Morley et al., 1997; Gingras et al., 1999b). This component of the translation machinery

88%

33%

36%

40%

23%

FIG. 5. Alignment of eIF4Gs and eIF4G-related proteins. Schematic alignment of eIF4Gs and eIF4G-related proteins from human eIF4GII (Gradi et al., 1998), human eIF4GI (Gradi et al., 1998), S. cerevisiae TIF4631 (Goyer et al., 1993), S. cerevisiae TIF4632 (Goyer et al., 1993), p97 (Imataka et al., 1997), Paip1 (Craig et al., 1998), and mammalian 4E-BP1 (Pause et al., 1994) highlighting conserved protein-binding regions. The middle domain of eIF4G is represented with black shading, with amino acid sequence identity between the middle domain of eIF4GII and related proteins. [Adapted from Marcotrigiano et al. (2001).]

284

JOSEPH MARCOTRIGIANO AND STEPHEN K. BURLEY

contains a poly(A)-binding protein (PABP) interacting motif (Tarun and Sachs, 1996; Imataka et al., 1998), a Tyr-X-X-X-X-Leu-F eIF4E-recognition motif (Mader et al., 1995), and a phylogenetically conserved ``middle'' segment responsible for binding eIF4A, the 43S preinitiation complex, and certain picornaviral RNA IRESs (Lamphear et al., 1995; Pestova et al., 1996a, b; Imataka and Sonenberg, 1997). The mammalian eIF4Gs (eIF4GI and eIF4GII) possess an eIF3-binding site, an additional C-terminal region that contains a second eIF4A-binding site, and a segment that interacts with a physiologic eIF4E kinase, Mnk-1 (Lamphear et al., 1995; Imataka and Sonenberg, 1997; Pyronnet et al., 1999). The carboxyterminal region is absent in yeast and is not absolutely required for in vitro translation with cytoplasmic extracts from higher eukaryotes. Ribosome-binding experiments with b-globin mRNA have demonstrated that the eIF4E-recognition motif and the middle portion of eIF4G constitute a minimal eIF4G core that supports cap-dependent translation initiation in vitro (Morino et al., 2000). Moreover, a chimeric protein consisting of the RNA-binding region of iron response protein IRP-1 fused to the middle domain of eIF4G can direct cap-independent translation of a cistron bearing an upstream iron-responsive element (De Gregorio et al., 1999). Mammalian eIF4Gs have some distant relatives that act as translational regulators, including p97 (Imataka et al., 1997) [also known as NAT1 (Yamanaka et al., 1997), DAP-5 (Levy-Strumpf et al., 1997), and eIF4G2 (Shaughnessy et al., 1997) ] and the PABP interacting protein-1 (Piap1) (Craig et al., 1998). p97 is similar to the C-terminal two-thirds of eIF4G and can bind both eIF3 and eIF4A, but not eIF4E or PABP. In vitro and in vivo studies have shown that p97 inhibits protein synthesis (Imataka and Sonenberg, 1997). p97 has been postulated to function as a negative regulator by sequestering eIF3 and eIF4A into inactive complexes. The more distant eIF4G relative, Paip1, contains a region similar to the middle domain of eIF4G and a C-terminal PABP-binding site. Overexpression of Paip1 has been shown to increase the rate of translation initiation (Craig et al., 1998). Modulation of eIF4G levels or activity can also have pronounced effects on translation initiation. During infection by certain picornaviruses (all genera except Cardiovirus and Hepatovirus) eIF4G undergoes speci®c proteolytic cleavage separating the eIF4E-recognition motif and the PABP-binding site from the eIF4A/IRES-binding sites (Fig. 5). In contrast, cardioviruses (including encephalomyocarditis virus, EMCV) inhibit cellular mRNA translation by dephosphorylating 4E-BP1 (Gingras et al., 1996), thereby disrupting the eIF4E/eIF4G complex depicted in Fig. 4B. Silencing of cap-dependent translation during picornaviral infections permits exploitation of the host cell translation machinery for synthesis of viral proteins.

STRUCTURAL BIOLOGY OF eIF4F

285

Picornaviral IRESs have been grouped into two different subcategories (type I, enterovirus and rhinovirus; and type II, cardiovirus and aphthovirus) based on primary sequence and predicted secondary structure ( Jackson and Kaminski, 1995). Reconstitution assays have shown that cap-independent translation from the EMCV (type II) IRES utilizes a subset of the canonical translation initiation factors required for capdependent protein synthesis (Pestova et al., 1996a). Further biochemical and mechanistic studies of type II IRESs have demonstrated that the middle region of human eIF4GI recognizes a structured element within the IRES, which is located immediately upstream of the translation start site (Ohlmann et al., 1996; Pestova et al., 1996b; Kolupaeva et al., 1998; Pilipenko et al., 2000). Together with eIF4A, eIF4G and the IRES recruit the 43S ribosomal complex to the viral RNA (Lomakin et al., 2000). Subsequent binding of the 60S ribosomal subunit permits translation of the viral polyprotein. The three-dimensional structure of the middle domain of human eIF4GII is illustrated schematically in Fig. 6 (Marcotrigiano et al., 2001). The domain consists of 10 a±helices with an overall crescent shape. The polypeptide chain forms a right-handed solenoid, with its superhelical axis perpendicular to the cylindrical axes of the a±helices. The molecular crescent is generated by ®ve repeating pairs of anti-parallel a±helices, stacked one repeat on the other. The 10 a±helices are arranged in the other 1a-1b-2a-2b-3a-3b-4a-4b-5a-5b (numbers denote each pair and the boldface lowercase letters refer to individual a±helices within each pair). This repeating pattern gives rise to a double layer of a±helices with the convex and concave surfaces formed by a and b a±helices, respectively. The intrarepeat and interrepeat loops segregate to opposite surfaces of the crescent-shaped molecule (Fig. 6). For simplicity, we will refer to the molecular surfaces of the eIF4GII crescent as concave (b a±helices; right in Fig. 6) and concave (a a±helices; left in Fig. 6). Adjacent layers of a-helical pairs are stabilized by salt bridges and van der Waals interactions, which span the length of the polypeptide chain giving rise to an extended hydrophobic core. a-Helices within repeats 1± 2 and 3±5 are arranged in a parallel stack with a rotation of less than 258 between neighboring repeats about the solenoid axis. The crescent shape arises from a right-handed 508 rotation of repeat 3 relative to repeat 2. The 2b±3a and 3a±3b loops located at the site of this large rotation do not have a well-de®ned electron density. There is no evidence that the relative orientations of these two subregions are likely to change upon higher order complex formation since the buried contact surface area between Ê 2 . It is, however, possible that individual each helical pair is 510±730 A repeats will shift on binding other factors or RNA. The structure of the middle domain unexpectedly revealed that the eIF4Gs are members of the HEAT repeat family of proteins [named for

286

JOSEPH MARCOTRIGIANO AND STEPHEN K. BURLEY

N

1a 1b 2a

2b 3a 3b

4b 4a C

5b 5a

FIG. 6. Structure of the middle domain of eIF4G. Ribbon drawing of the conserved central region of human eIF4GII viewed along the axes of the a-helices. HEAT repeats are labeled 1±5 and individual a-helices within each repeat are labeled a and b. The convex and concave surfaces are formed by a and b helices, respectively. The intraand interrepeat loops are located on the background and foreground of the ®gure, respectively. [Adapted from Marcotrigiano et al. (2001).]

Huntingtin, elongation factor 3, a subunit of protein phosphatase 2A (PP2A), and target of rapamycin]. HEAT repeat proteins participate in a wide variety of cellular processes that are dependent on assembling large multiprotein complexes (Andrade and Bork, 1995). A single HEAT repeat consists of two anti-parallel a±helices of varying length (designated a and b) that occur in tandem arrays repeated 3 to 22 times, with an average of 14 repeats per member (Andrade and Bork, 1995). Comparison of eIF4G with two other HEAT repeat family members [PR65/A subunit of PP2A (Groves et al., 1999) and importin b (Chook and Blobel, 1999; Cingolani et al., 1999) ] demonstrated less than 20% amino acid identity but high structural similarity with Z-scores greater than 9.0. The

STRUCTURAL BIOLOGY OF eIF4F

287

most signi®cant difference between the middle domain of eIF4GII and these canonical HEAT repeat proteins involves the linearity of the ahelices. A majority of the a-helices within PR65/A and importin b are bent, whereas the a-helices of eIF4GII are predominantly straight. We believe that the architecture of the HEAT repeat is compatible with a large number of quite different protein sequences, which would in turn generate a tremendous diversity in terms of the surface properties and, hence, biochemical and biological functions of these structurally related proteins. Given that HEAT repeats do not contain even one invariant or highly conserved residue (Andrade and Bork, 1995), it may be that the observed structural similarities between the middle domain of eIF4GII and PR65/A and importin b do not re¯ect an evolutionary relationship. The energetic constraints imposed by ``knobs in holes'' close-packing of pairs of anti-parallel and parallel a-helices (Bowie, 1997) could have led to HEAT repeat structures via convergent evolution. The three-dimensional structure of the eIF4A/IRES-binding domain of eIF4GII provided some insight into the higher order complex formation during cap-dependent and cap-independent translation. Results of sitedirected mutagenesis studies of human and yeast eIF4Gs (Imataka and Sonenberg, 1997; Neff and Sachs, 1999; Lomakin et al., 2000; Morino et al., 2000; Marcotrigiano et al., 2001) and human p97 (Imataka and Sonenberg, 1997) identi®ed potential eIF4A-binding sites and the EMCV IRES-binding sites on the molecular surface of eIF4GII. Mutations affecting EMCV IRES and eIF4A binding map to the entire length of the intrarepeat face and the interrepeat face overlying repeats 1, 2, and 3, respectively. The location of the putative eIF4A- and EMCV IRESbinding sites on different, adjacent surfaces is consistent with previous biochemical observations that eIF4G can simultaneously bind eIF4A and EMCV IRES. eIF4A has been shown to enhance binding of eIF4G to the EMCV IRES (Pestova et al., 1996b; Lomakin et al., 2000). In addition, mutations at the intersection of these two surfaces can disrupt binding to both the EMCV IRES and eIF4A (Marcotrigiano et al., 2001). The IRES-binding surface contains both positively charged and hydrophobic features that could contribute, respectively, to phosphate neutralization and base recognition during IRES binding. The putative eIF4A-binding site is highly polar, suggesting that this protein±protein interaction involves hydrogen bonds and/or salt bridges. Earlier biochemical studies of eIF4F demonstrated disruption of the eIF4A±eIF4G interaction during anion-exchange chromatography (Etchison and Milburn, 1987), which is consistent with largely polar protein±protein interactions.

288

JOSEPH MARCOTRIGIANO AND STEPHEN K. BURLEY

V. PREPARATION OF THE mRNA 50 -UTR FOR SMALL RIBOSOMAL SUBUNIT BINDING Secondary structural features within the 50 -UTR, particularly those near the cap structure, interfere with binding and/or scanning of the small ribosomal subunit. The remaining component of the eIF4F heterotrimer, eIF4A, is an ATP-dependent, RNA helicase thought to unwind mRNA secondary structural elements in preparation of ribosome binding. eIF4A cannot recognize the 50 -UTR directly and is brought to the mRNA as part of the eIF4F complex, which is 20-fold more active in RNA unwinding than eIF4A itself (Rozen et al., 1990). Several lines of evidence suggest that multiple copies of eIF4A may bind to a single mRNA. Protein synthesis in vitro requires excess eIF4A in addition to the eIF4A subunit of the eIF4F complex (Conroy et al., 1990), and pools of free eIF4A readily exchange in and out of the eIF4F heterotrimer (Yoder-Hill et al., 1993). Mammalian eIF4Gs contain two separate eIF4A-binding sites that can bind independent molecules of eIF4A (Korneeva et al., 2001). Finally, eIF4A is six times more abundant than either eIF4G or eIF4E (Duncan et al., 1987). These ®ndings suggest that the eIF4E/eIF4G heterodimer can recruit multiple copies of eIF4A to the 50 -UTR. Given that eIF4A can unwind only small stretches of duplex RNA, multiple copies may be necessary to unwind large 50 -UTRs with complicated secondary structures. There is also some evidence that eIF4A contributes to aspects of translation initiation in addition to preparation of the mRNA for small ribosomal subunit binding. eIF4A-depleted yeast cell extracts cannot translate mRNAs containing short 50 -UTRs devoid of secondary structural features, even if supplemented with a helicase-de®cient point mutant of eIF4A (Blum et al., 1992). Finally, cap-independent translation from a type II IRES requires eIF4A (Pestova et al., 1996a), although it is not thought that RNA unwinding per se is necessary for IRES function. eIF4A is an RNA-dependent ATPase with bidirectional helicase activity, which is essential for yeast cell viability (Blum et al., 1989). Mammals possess three isoforms of eIF4A, termed eIF4AI, eIF4AII, and eIF4AIII. eIF4AI and eIF4AII appear to be very similar in both amino acid sequence (89% identity) and biochemical function, although they are markedly different in terms of tissue distribution (Nielsen and Trachsel, 1988) and developmental regulation (Morgan and Sargent, 1997). eIF4AIII shares about 60% identity with the other two isoforms, but does not support translation initiation in vitro (Li et al., 1999). Yeast require an unrelated RNA helicase, Ded1, for translation initiation in addition to eIF4A (Chuang et al., 1997). Disruption of the Ded1 gene can be complemented by PL10, a mouse homologue of Ded1 that is required for

STRUCTURAL BIOLOGY OF eIF4F

289

spermatogenesis. It seems likely, therefore, that eIF4A is not the only RNA helicase of importance in eukaryotic translation initiation. eIF4A is a prototypic member of the Asp-Glu-X-Asp/His or DEXD/Hbox family of RNA and DNA helicases, which participate in diverse biological processes including translation, pre-mRNA splicing, ribosome biogenesis, and development. Helicases have been divided into several different superfamilies (SF) based on sequence comparisons (Gorbalenya and Koonin, 1993). Most helicases fall into SF-1 and SF-2, although overall amino acid sequence similarity and the relative arrangement of conserved motifs suggest that all SF members evolved from a common ancestor. All members share seven conserved polypeptide chain segments (motifs I, Ia, II, III, IV, V, and VI; Fig. 7). The eIF4As contain two additional conserved regions located between motifs Ia and II (GG and TPGR; Fig. 7A). Mutational analyses of mammalian and yeast eIF4As provided some insight into the biochemical roles of regions I, II, III, and VI (Schmid and Linder, 1991; Blum et al., 1992; Pause and Sonenberg, 1992; Pause et al., 1993). Mutations in motif I, the Walker ATPase Amotif, disrupt both ATP binding and helicase activity. Motif II contains the Asp-Glu-X-Asp/His motif, the Walker APTase B-motif, from which the enzyme gets its name. Mutations in this region disrupt ATP hydrolysis and RNA helicase activity without affecting ATP binding, suggesting that these amino acids coordinate the essential Mg2 divalent cation during the hydrolysis reaction. Motif III (SAT) is important for RNA unwinding, but not for ATP binding, hydrolysis, or RNA binding. Finally, mutations within His-Arg-Ile-Gly-Arg-X-X-Arg (Motif VI) reduce RNA binding and abolish RNA helicase activity. Kinetic and thermodynamic studies of ATP hydrolysis by eIF4A suggest that RNA binding is modulated by the presence of the g-phosphate (Lorsch and Herschlag, 1998a,b). In addition, RNA cross-linking and limited proteolysis experiments demonstrated that the g-phosphate induces conformational changes in eIF4A (Pause et al., 1993). Experimental electron density maps obtained from crystals of fulllength yeast eIF4A revealed only the N-terminal domain with an apparently disordered C-terminal domain ( Johnson and McKay, 1999). Three-dimensional structures of the N- and C-terminal domains of yeast eIF4A were determined independently by X-ray crystallography (Benz et al., 1999; Johnson and McKay, 1999; Caruthers et al., 2000). The structures of the individual domains permitted elucidation of the structure of the fulllength protein, and a three-dimensional model of an alternate structure of the full-length protein has been calculated (Caruthers et al., 2000). The N-terminal ATPase domain of eIF4A (Fig. 7B) consists of an a/b domain with nine a-helices, surrounding an eight-stranded b-sheet

290

JOSEPH MARCOTRIGIANO AND STEPHEN K. BURLEY

B

C

C D

FIG. 7. X-ray structures of yeast eIF4A. (A) Schematic representation of yeast eIF4A showing the orientation and numbering of conserved helicase motifs. (B and C) Ribbon drawings of the N- and C-terminal domains of eIF4A, respectively, viewed perpendicular to the b-sheet. a-Helices are labeled H1±H15 and b-strands are labeled 1±14, with the N- and C-termini of each domain labeled N and C, respectively. The locations of select helicase motifs are also indicated. (D) Ribbon drawing of the experimental structure of full-length eIF4A. The N- and C-terminal domains are labeled N and C, respectively, and the 11-residue linker is shaded dark.

(Benz et al., 1999; Johnson and McKay, 1999). Seven of the eight bstrands (S2±S8) form a parallel b-sheet with the short N-terminal b-strand, S1, anti-parallel to the remainder. With the exception of a-

STRUCTURAL BIOLOGY OF eIF4F

291

helix H2 the axes of all a-helices lie anti-parallel to the b-sheet. a-Helices H1±H5 and H6±H9 are packed against opposite sides of the b-sheet. The N- and C-termini of the domain are located on adjacent b-strands (S1 and S8). Six of the nine conserved eIF4A regions fall within the Nterminal domain (Figs. 7A and 7B). All of the conserved motifs are located in loops connecting a b-strand with an a-helix, and they map to the upper edge of the b-sheet (Fig. 7B). Motif I (Walker ATPase A-motif) is located on the S2±H4 loop and near motifs II and III. Motif II (Walker ATPase B-motif) contains the Asp-Glu-X-Asp/His sequence and occurs on the S6±H8 loop. Between motifs I and II is the S7±H9 loop that contains motif II (SAT motif). On the opposite side of the Asp-Glu-X-Asp/His-box, motif Ia occurs on loop S3±H5. The GG and TPRG motifs, which are not conserved among all helicases, are found on the S5±H7 and S4±H6 loops, respectively. The N-terminal domain of yeast eIF4A is structurally similar to RecA, two DNA helicases (Pcr and Rep), and the RNA helicase NS3 from hepatitis C virus. Despite differences in amino acid sequence and substrate speci®city, these domains share a common fold involved in ATP hydrolysis (Bird et al., 1998). The domains have a core structure composed of ®ve parallel b-strands connected by a-helices within an a/b motif. In eIF4A, the helicase core consists of b-strands S2, S3, and S5± S7 and a-helices H4, H5, and H7±H9. In most extant helicase structures, polypeptide regions ¯anking the core have either additional secondary structure elements or entire domains that are speci®c to the biochemical functions to these related molecules. The core domain establishes a scaffold upon which residues involved in ATP binding and hydrolysis are located (motifs I, Ia, II, and III). Structures of the N-terminal domain of eIF4A were also determined in the presence of ADP and ATP (Benz et al., 1999). Both nucleotides bind in an extended conformation on the upper edge of the b-sheet. The adenine base ®ts into a pocket formed by a-helices H2 and H3 and the intervening loop. The benzene ring of Phe-41 (yeast eIF4A residue numbering) at the C-terminus of a-helix H2 makes a p p staking interaction with the base. The carbonyl oxygen of Glu-43 and the side chain of Gln-48 form pseudo-Watson±Crick A±T hydrogen bonds (donor plus acceptor) with N6 and N7 of the adenine. No interactions were detected between eIF4A and the ribose of either ADP or ATP. The phosphates of ADP and ATP make salt bridges with the side chain of Lys-71 (an invariant amino acid in motif I) and hydrogen bonds with the backbone nitrogens of Gly-68±Thr-72 of motif I. A Mg2 cation coordinates the b-phosphate of ADP and bridges the b- and g-phosphate groups of ATP. In addition, the Mg2 is coordinated by Thr-72, Asp-169 (the ®rst residue of Asp-Glu-X-Asp/His in motif II), and water molecules.

292

JOSEPH MARCOTRIGIANO AND STEPHEN K. BURLEY

The C-terminal domain of yeast eIF4A (Fig. 7C) is composed of a seven-stranded, parallel b-sheet, S9±S15, surrounded by ®ve a-helices, H10±H15 (Caruthers et al., 2000). All a-helices lie anti-parallel to the direction of the b-strands, with a-helices H10±H11 and H12±H15, respectively, packing against opposite faces of the b-sheet. The overall topology of the C-terminal domain is similar to that of the equivalent domains in other helicases for which structural information is available, permitting de®nition of a C-terminal core composed of six of the bstrands and a-helices H10, H11, H13, and H14, with conserved motifs IV±VI located in loop regions on the upper edge of the b-sheet. This Cterminal core resembles the N-terminal core described earlier, which suggests that eIF4A and other helicases may have arisen from a dimeric ancestral protein that underwent gene duplication and gene fusion, yielding a single polypeptide chain with two similar domains. The C-terminal domain structure was used as a source of phase information to enable completion of the structure determination for crystalline full-length eIF4A ( Johnson and McKay, 1999; Caruthers et al., 2000). The molecule has an overall ``dumbbell'' shape with the N- and C-terminal domains connected by an extended, 11-amino-acid linker Ê in length (Fig. 7D). McKay and co-workers have approximately 18 A suggested that eIF4A has several conformations in solution, with their particular crystallization conditions favoring the observed extended conformation (see Caruthers et al., 2000). In the full-length eIF4A crystal lattice, the conserved helicase motifs within the N- and C-terminal domains are spatially separate. Assuming that ef®cient energy transfer from ATP hydrolysis to RNA unwinding could be achieved only by close apposition of the conserved motifs, the enzyme must undergo a reorganization of the two domains into a compact structure on binding nucleotide and/or RNA. In addition, protease cleavage experiments demonstrated that the proteolytically sensitive linker region (Fig. 7D) and loops are protected from digestion on binding nucleotide and RNA (Lorsch and Herschlag, 1998b), which could be explained by domain rearrangement. A three-dimensional model of a more compact alternate structure of eIF4A has been calculated with the aid of other helicase structures as modeling templates (Caruthers et al., 2000). The calculated model juxtaposes motifs I and II of the N-terminal domain with motifs V and VI of the C-terminal domain, resulting in a compact structure [see Fig. 3 in Caruthers et al. (2000) ]. Several lines of evidence provide indirect support for this calculated model: (i) The termini of the domains are separÊ , which could be accommodated by the linker ated by about 18±20 A region. (ii) Apposition of the two domains creates a basic surface feature that could serve as a binding site for single-stranded RNA, consisting of Arg-98 and Arg-148 of the N-terminal domain and Arg-269, Arg-270,

STRUCTURAL BIOLOGY OF eIF4F

293

and Arg-298 of the C-terminal domain. (iii) The spatial proximity of motifs involved in ATP binding and hydrolysis to motifs involved in RNA binding and unwinding suggests that energy transfer could be accomplished by conformational changes in the protein.

VI. CONCLUSION AND PERSPECTIVES Early appreciation of the importance of eIF4F in mRNA recognition and preparation encouraged biologists to think that this large heterotrimeric assembly would play a central regulatory role in eukaryotic translation initiation. The enormous body of work summarized above clearly documents that these expectations have been realized. Threedimensional structures of its smallest subunit (eIF4E) recognizing a 50 ,7-methyl-G mRNA cap analogue and of an active eIF4E/7methyl-GDP/eIF4G ternary complex elucidated novel protein±RNA interactions and explained, at least in part, how eIF4G plays its bridging role within the heterotrimer. The structure of an inhibitory eIF4E/7methyl-GDP/4E-BP1 ternary complex demonstrated that the eIF4Ebinding proteins function as molecular mimics of eIF4G, underscoring the regulatory signi®cance of its bridging function. The structure of the middle domain of eIF4G permitted identi®cation of adjacent binding sites for both eIF4A and a type II picornaviral IRES and revealed an unexpected architectural relationship between the translation machinery and nuclear import factors. Finally, various structures of eIF4A provided intriguing insights into the behavior of this conformationally heterogeneous DEXD/H-box, ATP-dependent RNA helicase. While the complexity of eIF4F and eukaryotic translation poses a formidable barrier to our hopes of developing a detailed understanding of translational regulatory mechanisms, the progress reviewed here is decidedly encouraging. High-resolution structural data provide a basis for designing a program of rational mutagenesis to con®rm the importance of intermolecular interactions identi®ed by biochemical and genetic analyses. In the immediate term, structural work can be combined with biophysical and biochemical studies of the kinetic behavior of eIF4F (and its interacting components) during mRNA recognition and preparation. Structural genomics efforts (Burley et al., 1999) will almost certainly contribute individual domain structures of yet more complicated translation factors (e.g., eIF3). Finally, and perhaps most important, dramatic recent progress toward a structural understanding of both the bacterial ribosomes (Cate et al., 1999; Ban et al., 2000; Frank and Agrawal, 2000; Schluenzen et al., 2000; Wimberly et al., 2000) and yeast ribosomes (Verschoor et al., 1998) offers the eventual promise of being able to

294

JOSEPH MARCOTRIGIANO AND STEPHEN K. BURLEY

study interactions between eIF4F and the 43S ribosome preinitiation complex.

ACKNOWLEDGMENTS We are grateful to Drs. L. Bellsolell, R. C. Deo, T. E. Dever, A.-C. Gingras, C. M. Groft, C. U. T. Hellen, A. Hinnebusch, W. C. Merrick, T. Pestova, A. Roll-Mecak, and N. Sonenberg and the many other translation factor biologists who provided us with reprints and access to unpublished material. We apologize for failing to cite work that may have been relevant, but excluded because of space limitations. We thank Ms. T. B. Niven for editorial assistance. This work was supported by grants from the Human Frontiers Science Program and the National Institute of General Medical Sciences (S.K.B.) and by David Rockefeller Alumni and Burroughs Wellcome Interfaces Grant Fellowships ( J.M.). S.K.B. is an Investigator of the Howard Hughes Medical Institute.

REFERENCES Adams, B. L., Morgan, M., Muthukrishnan, S., Hecht, S. M., and Shatkin, A. J. (1978). J. Biol. Chem. 253, 2589±2595. Allen, M. L., Metz, A. M., Timmer, R. T., Rhoads, R. E., and Browning, K. S. (1992). J. Biol. Chem. 267, 23232±23236. Altmann, M., Schmitz, N., Berset, C., and Trachsel, H. (1997). EMBO J. 16, 1114±1121. Andrade, M. A., and Bork, P. (1995). Nat. Genet. 11, 115±116. Ban, N., Nissen, P., Hansen, J., Moore, P. B., and Steitz, T. A. (2000). Science 289, 905±290. Benz, J., Trachsel, H., and Baumann, U. (1999). Structure 7, 671±679. Bird, L. E., Subramanya, H. S., and Wigley, D. B. (1998). Curr. Opin. Struct. Biol. 8, 14±18. Blum, S., Mueller, M., Schmid, S. R., Linder, P., and Trachsel, H. (1989). Proc. Natl. Acad. Sci. USA 86, 6043±6046. Blum, S., Schmid, S. R., Pause, A., Buser, P., Linder, P., Sonenberg, N., and Trachsel, H. (1992). Proc. Natl. Acad. Sci. USA 89, 7664±7668. Both, G. W., Lavi, S., and Shatkin, A. J. (1975). Cell 4, 173±180. Bowie, J. U. (1997). Nat. Struct. Biol. 4, 915±917. Burley, S. K., Almo, S. C., Bonanno, J. B., Capel, M., Chance, M. R., Gaasterland, T., Lin, D., Sali, A., Studier, F. W., and Swaminathan, S. (1999). Nat. Genet. 23, 151±157. Caruthers, J. M., Johnson, E. R., and McKay, D. B. (2000). Proc. Natl. Acad. Sci. USA 97, 13080±13085. Cate, J. H., Yusupov, M. M., Yusupova, G. Z., Earnest, T. N., and Noller, H. F. (1999). Science 285, 2095±2104. Chook, Y. M., and Blobel, G. (1999). Nature 399, 230±237. Chuang, R. Y., Weaver, P. L., Liu, Z., and Chang, T. H. (1997). Science 275, 1468±1471. Cingolani, G., Petosa, C., Weis, K., and Muller, C. W. (1999). Nature 399, 221±229. Colgan, D. F., and Manley, J. L. (1997). Genes Dev. 11, 2755±2766. Conroy, S. C., Dever, T. E., Owens, C. L., and Merrick, W. C. (1990). Arch. Biochem. Biophys. 282, 363±371. Coppola, J. A., Field, A. S., and Luse, D. S. (1983). Proc. Natl. Acad. Sci. USA 80, 1251±1255. Craig, A. W., Haghighat, A., Yu, A. T., and Sonenberg, N. (1998). Nature 392, 520±523. Darnell, J. E., Wall, R., and Tushinski, R. J. (1971). Proc. Natl. Acad. Sci. USA 68, 1321±1325.

STRUCTURAL BIOLOGY OF eIF4F

295

Darzynkiewicz, E., Stepinski, J., Ekiel, I., Goyer, C., Sonenberg, N., Temeriusz, A., Jin, Y., Eijuwade, T., Haber, D., and Tahara, S. M. (1989). Biochemistry 28, 4771±4778. De Gregorio, E., Preiss, T., and Hentze, M. W. (1999). EMBO J. 18, 4865±4874. Doel, M. T., and Carey, N. H. (1976). Cell 8, 51±58. Duncan, R., Milburn, S. C., and Hershey, J. W. (1987). J. Biol. Chem. 262, 380±388. Edmonds, M., Vaughan, M., Jr., and Nakazato, H. (1971). Proc. Natl. Acad. Sci. USA 68, 1336±1340. Etchison, D., and Milburn, S. (1987). Mol. Cell. Biochem. 76, 15±25. Fletcher, C. M., McGuire, A. M., Gingras, A.-C., Li, H., Matsuo, H., Sonenberg, N., and Wagner, G. (1998). Biochemistry 37, 9±15. Frank, J., and Agrawal, R. K. (2000). Nature 406, 318±322. Gingras, A.-C., Gygi, S. P., Raught, B., Polakiwicz, R., Abraham, R. T., Abersold, R., and Sonenberg, N. (1999a), Genes Dev. 13, 1422±1437. Gingras, A.-C., Raught, B., and Sonenberg, N. (1999b). Annu. Rev. Biochem. 68, 913±963. Gingras, A.-C., Svitkin, Y., Belsham, G. J., Pause, A., and Sonenberg, N. (1996). Proc. Natl. Acad. Sci. USA 93, 5578±5583. Gorbalenya, A. E., and Koonin, E. V. (1993). Curr. Opin. Struct. Biol. 3, 419±429. Goyer, C., Altman, M., Lee, H. S., Blanc, A., Deshmukh, M., Woolford, J. L., Trachsel, H. S., and Sonenberg, N. (1993). Mol. Cell. Biol. 13, 4860±4874. Gradi, A., Imataka, H., Svitkin, Y. V., Rom, E., Raught, B., Morino, S., and Sonenberg, N. (1998). Mol. Cell. Biol. 18, 334±342. Groves, M. R., Hanlon, N., Turowski, P., Hemmings, B. A., and Barford, D. (1999). Cell 96, 99±110. Haghighat, A., Mader, S., Pause, A., and Sonenberg, N. (1995). EMBO J. 14, 5701±5709. Hagler, J., and Shuman, S. (1992). Science 255, 983±986. Hentze, M. W. (1997). Science 275, 500±501. Hodel, A. E., Gershon, P. D., and Quiocho, F. A. (1998). Mol. Cell 1, 443±447. Hodel, A. E., Gershon, P. D., Shi, X., and Quiocho, F. A. (1996). Cell 85, 247±256. Hodel, A. E., Gershon, P. D., Shi, X. N., Wang, S. M., and Quiocho, F. A. (1997). Nat. Struct. Biol. 4, 350±354. Imataka, H., Gradi, A., and Sonenberg, N. (1998). EMBO J. 17, 7480±7489. Imataka, H., Olsen, H. S., and Sonenberg, N. (1997). EMBO J. 16, 817±825. Imataka, H., and Sonenberg, N. (1997). Mol. Cell. Biol. 17, 6940±6947. Ishida, T., Doi, M., Ueda, H., Inoue, M., and Sheldrick, G. M. (1988). J. Am. Chem. Soc. 110, 2286±2294. Ishida, T., Iyo, H., Ueda, H., Doi, M., Inoue, M., Nishimura, S., and Kitamura, K. (1991). J. Chem. Soc. Perkins Trans. I, 1847±1853. Ishida, T., Shibata, M., Fujii, K., and Inoue, M. (1983). Biochemistry 22, 3571±3581. Jackson, R. J. (2000). In ``Translational Control of Gene Expression'' (N. Sonenberg, J. W. B. Hershey, and M. B. Mathews, Eds.), pp. 127±183. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY. Jackson, R. J., and Kaminski, A. (1995). RNA 1, 985±1000. Jacobson, A. (1996). In ``Translational Control'' ( J. W. B. Hershey, M. B. Mathews, and N. Sonenberg, Eds.), pp. 451±480. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY. Jacobson, A., and Favreau, M. (1983). Nucleic Acids Res. 11, 6353±6368. Johnson, E. R., and McKay, D. B. (1999). RNA 5, 1526±1534. Kolupaeva, V. G., Pestova, T. V., Hellen, C. U., and Shatsky, I. N. (1998). J. Biol. Chem. 273, 18599±18604. Korneeva, N. L., Lamphear, B. J., Hennigan, F. L., Merrick, W. C., and Rhoads, R. E. (2001). J. Biol. Chem. 276, 2872±2879.

296

JOSEPH MARCOTRIGIANO AND STEPHEN K. BURLEY

Kozak, M. (1987). Nucleic Acids Res. 15, 8125±8148. Kozak, M. (1989). J. Cell. Biol. 108, 229±241. Kozak, M. (1999). Gene 234, 187±208. Kozak, M., and Shatkin, A. J. (1978). Cell 13, 201±212. Kuge, H., Brownlee, G. G., Gershon, P. D., and Richter, J. D. (1998). Nucleic Acids Res. 26, 3208±3214. Kussie, P. H., Gorina, S., Marechal, V., Elenbaas, B., Moreau, J., Levine, A. J., and Pavletich, N. P. (1996). Science 274, 948±953. Lamphear, B. J., Kirchweger, R., Skern, T., and Rhoads, R. E. (1995). J. Biol. Chem. 270, 21975±21983. Lee, S. Y., Mendecki, J., and Brawerman, G. (1971). Proc. Natl. Acad. Sci. USA 68, 1331±1335. Levy-Strumpf, N., Deiss, L. P., Berissi, H., and Kimchi, A. (1997). Mol. Cell. Biol. 17, 1615±1625. Li, Q., Imataka, H., Morino, S., Rogers, G. W., Richter-Cook, N. J., Merrick, W. C., and Sonenberg, N. (1999). Mol. Cell. Biol. 19, 7336±7346. Lim, L., and Canellakis, E. S. (1970). Nature 227, 710±712. Lin, T.-A., Kong, X., Haystead, T. A. J., Pause, A., Belsham, G., Sonenberg, N., and Lawrence, J. C. (1994). Science 266, 653±656. Lomakin, I. B., Hellen, C. U., and Pestova, T. V. (2000). Mol. Cell. Biol. 20, 6019±6029. Lorsch, J. R., and Herschlag, D. (1998a). Biochemistry 37, 2180±2193. Lorsch, J. R., and Herschlag, D. (1998b). Biochemistry 37, 2194±2206. Mader, S., Lee, H., Pause, A., and Sonenberg, N. (1995). Mol. Cell. Biol. 15, 4990±4997. Marcotrigiano, J., Gingras, A.-C., Sonenberg, N., and Burley, S. K. (1997). Cell 89, 951±961. Marcotrigiano, J., Gingras, A.-C., Sonenberg, N., and Burley, S. K. (1999). Mol. Cell 3, 707±716. Marcotrigiano, J., Lomakin, I. B., Sonenberg, N., Pestova, T. V., Hellen, C. U. T., and Burley, S. K. (2001). Mol. Cell 7, 193±203. Matsuo, H., Li, H., McGuire, A. M., Fletcher, C. M., Gingras, A.-C., Sonenberg, N., and Wagner, G. (1997). Nat. Struct. Biol. 4, 717±724. Merrick, W. C., and Hershey, J. W. B. (1996). In ``Translational Control'' ( J. W. B. Hershey, M. B. Mathews, and N. Sonenberg, Eds.), pp. 31±69. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY. Morgan, R., and Sargent, M. G. (1997). Development 124, 2751±2760. Morino, S., Imataka, H., Svitkin, Y. V., Pestova, T. V., and Sonenberg, N. (2000). Mol. Cell. Biol. 20, 468±477. Morley, S. J., Curtis, P. S., and Pain, V. M. (1997). RNA 3, 1085±1104. Munroe, D., and Jacobson, A. (1990). Mol. Cell. Biol. 10, 3441±3455. Muthukrishnan, S., Both, G. W., Furuichi, Y., and Shatkin, A. J. (1975). Nature 255, 33±37. Muthukrishnan, S., Morgan, M., Banerjee, A. K., and Shatkin, A. J. (1976). Biochemistry 15, 5761±5768. Neff, C. L., and Sachs, A. B. (1999). Mol. Cell. Biol. 19, 5557±5564. Nielsen, P. J., and Trachsel, H. (1988). EMBO J. 7, 2097±2105. Ohlmann, T., Rau, M., Pain, V. M., and Morley, S. J. (1996). EMBO J. 15, 1371±1382. Pause, A., Belsham, G. J., Gingras, A.-C., Donze, O., Lin, T.-A., Lawrence, J. C., and Sonenberg, N. (1994). Nature 371, 762±767. Pause, A., Methot, N., and Sonenberg, N. (1993). Mol. Cell. Biol. 13, 6789±6798. Pause, A., and Sonenberg, N. (1992). EMBO J. 11, 2643±2654. Pestova, T. V., Hellen, C. U. T., and Shatsky, I. N. (1996a). Mol. Cell. Biol. 16, 6859±6869. Pestova, T. V., Lomakin, I. B., Lee, J. H., Choi, S. K., Dever, T. E., and Hellen, C. U. T. (2000). Nature 403, 332±335. Pestova, T. V., Shatsky, I. N., and Hellen, C. U. (1996b). Mol. Cell. Biol. 16, 6870±6878.

STRUCTURAL BIOLOGY OF eIF4F

297

Pilipenko, E. V., Pestova, T. V., Kolupaeva, V. G., Khitrina, E. V., Poperechnaya, A. N., Agol, V. I., and Hellen, C. U. T. (2000). Genes Dev. 14, 2028±2045. Pillutla, R. C., Yue, Z., Maldonado, E., and Shatkin, A. J. (1998). J. Biol. Chem. 273, 21443±21446. Poulin, F., Gingras, A.-C., Olsen, H., Chevalier, S., and Sonenberg, N. (1998). J. Biol. Chem. 273, 14002±14007. Ptushkina, M., van der Haar, T., Vasilescu, S., Birkenhager, R., and McCarthy, J. E. (1998). EMBO J. 17, 4798±4808. Pyronnet, S., Imataka, H., Gingras, A.-C., Fukunaga, R., Hunter, T., and Sonenberg, N. (1999). EMBO J. 18, 270±279. Radhakrishnan, I., Perez-Alvarado, G. C., Parker, D., Dyson, H. J., Montminy, M. R., and Wright, P. E. (1997). Cell 91, 741±752. Richter, J. D. (1996). In ``Translational Control'' ( J. W. B. Hershey, M. B. Mathews, and N. Sonenberg, Eds.), pp. 481±503. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY. Roll-Mecak, A., Cao, C., Dever, T. E., and Burley, S. K. (2000). Cell 103, 781±792. Rom, E., Kim, H. C., Gingras, A. C., Marcotrigiano, J., Favre, D., Olsen, H., Burley, S. K., and Sonenberg, N. (1998). J. Biol. Chem. 273, 13104±13109. Rozen, F., Edery, I., Meerovitch, K., Dever, T. E., Merrick, W. C., and Sonenberg, N. (1990). Mol. Cell. Biol. 10, 1134±1144. Sachs, A. (2000). In ``Translational Control of Gene Expression'' (N. Sonenberg, J. W. B. Hershey, and M. B. Mathews, Eds.), pp. 447±467. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY. Sachs, A. B., and Wahle, E. (1993). J. Biol. Chem. 268, 22955±22958. Schluenzen, F., Tocilj, A., Zarivach, R., Harms, J., Gluehmann, M., Janell, D., Bashan, A., Bartels, H., Agmon, I., Franceschi, F., and Yonath, A. (2000). Cell 102, 615±623. Schmid, S. R., and Linder, P. (1991). Mol. Cell. Biol. 11, 3463±3471. Shaughnessy, J. D., Jr., Jenkins, N. A., and Copeland, N. G. (1997). Genomics 39, 192±197. Sonenberg, N. (1996). In ``Translational Control'' ( J. W. B. Hershey, M. B. Mathews, and N. Sonenberg, Eds.), pp. 245±269. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY. Sonenberg, N., Morgan, M. A., Merrick, W. C., and Shatkin, A. J. (1978). Proc. Natl. Acad. Sci. USA 75, 4843±4847. Sonenberg, N., Rupprecht, M. A., Merrick, W. C., and Shatkin, A. J. (1979). Proc. Natl. Acad. Sci. USA 76, 4345±4349. Steitz, J. A., and Jakes, K. (1975). Proc. Natl. Acad. Sci. USA 72, 4734±4738. Svitkin, Y. V., Ovchinnikov, L. P., Dreyfuss, G., and Sonenberg, N. (1996). EMBO J. 15, 7147±7155. Tarun, S. Z., Jr., and Sachs, A. B. (1996). EMBO J. 15, 7168±7177. Ueda, H., Maruyama, H., Doi, M., Inoue, M., Ishida, T., Morioka, H., Tanaka, T., Nishikawa, S., and Uesugi, S. (1991). J. Biochem. 109, 882±889. Uesugi, M., Nyanguile, O., Lu, H., Levine, A. J., and Verdine, G. L. (1997). Science 277, 1310±1313. Verschoor, A., Warner, J. R., Srivastava, S., Grassucci, R. A., and Frank, J. (1998). Nucleic Acids Res. 26, 655±661. Wimberly, B. T., Brodersen, D. E., Clemons, W. M., Morgan-Warren, R. J., Carter, A. P., Vonrhein, C., Hartsch, T., and Ramakrishnan, V. (2000). Nature 407, 327±339. Yamanaka, S., Poksay, K. S., Arnold, K. S., and Innerarity, T. L. (1997). Genes Dev. 11, 321±333. Yoder-Hill, J., Pause, A., Sonenberg, N., and Merrick, W. C. (1993). J. Biol. Chem. 268, 5566±5573.

This Page Intentionally Left Blank

AUTHOR INDEX

A Abagyan, R., 26, 36, 44, 59, 86 Abersold, R., 279, 280, 283 Abraham, R. T., 279, 280, 283 Abril, M. A., 218 Achatz, H., 198 Ackermann, F., 58 Ackers, G. K., 138, 140, 179 Acuto, O., 194 Adam, M., 181 Adams, B. L., 272 Adams, M. D., 1 Adams, M. J., 77, 82 Adamska, I., 2 Aderem, A., 91 Adermann, K., 252 Adzhubei, A. A., 213 Afar, D. E., 251 A¯alo, C., 56, 59 Aghazadeh, B., 234, 235, 254 Agmon, I., 293 Agol, V. I., 285 Agrawal, R. K., 293 Aguayo, C., 65 Aguinaldo, A. M., 93 Ahringer, J., 96 Ailey, B., 83 Air, G. M., 122, 123 Aitchison, J. D., 2 Ajuh, P., 2 Alagona, G., 54 Albeck, S., 46 Alberts, I., 19, 21 Alexandropoulos, K., 219, 220 Allard, P., 226 Alleaume, A.-M., 96 Allen, D., 198, 199 Allen, M. L., 280 Alligood, K., 172, 180, 203 Alm, E., 218

Almo, S. C., 235, 293 Altamirano, M. M., 65 Altieri, A. S., 227 Altman, A., 201 Altman, E., 125 Altmann, M., 279, 280, 283 Altschul, S. F., 77 Alzari, P. M., 34, 122, 125, 154 Amadei, A., 65 Amit, A. G., 122 Amoui, M., 235, 254 Amzel, L. M., 120, 124, 128 Ana®, M., 220 Anderson, C. L., 193 Anderson, D., 161, 164, 189 Anderson, J. M., 249 Anderson, M., 169, 171 Andrade, M. A., 80, 114, 286, 287 Andreev, J., 86 Andreotti, A. H., 227, 254 Andrews, L. C., 56 Angiolieri, M., 221 Anton, I. M., 223 Antosiewicz, J., 60 Aravind, L., 80, 91, 92, 100 Arcaro, A., 202 Ardelt, W., 26 Argos, P., 12 Arkin, M. R., 155 Arlinghaus, R. B., 252 Arnold, K. S., 284 Arnoux, B., 189, 195, 226, 240 Arold, S., 220, 226, 227, 243, 244, 245 Aronheim, A., 86, 105 Arrondo, J. L., 213, 221, 239 Artymiuk, P. J., 55, 63 Ashkenazi, A., 219 Ashley, J., 198 Atwell, S. K., 189, 226 Aubin, Y., 181 Auger, K. R., 172, 177, 203 299

300

AUTHOR INDEX

Avron, B., 138, 142

B Baase, W. A., 147 Bachar, O., 54 Bachmann, M. F., 148 Bachovlin, W. W., 181 Backer, J. M., 172, 180 Bader, G. D., 3 Bae, H., 37 Bagheri, B., 60 Bahler, M., 86 Bairoch, A., 78, 82 Baker, A. T., 122 Baker, D., 218 Baldauf, S. L., 93 Baldisseri, D. M., 227 Baldwin, E. P., 147 Ball, C. A., 100 Ballard, D. W., 152, 154 Baltensperger, K., 225 Baltimore, D., 164, 165, 172, 173, 174, 177, 178, 215, 219, 220, 221, 222, 223, 225, 231, 237, 243 Ban, N., 122, 293 Band, H., 180 Banerjee, A. K., 271 Banerjee, U., 225 Banyai, L., 76, 78 Barabasi, A. L., 5, 6 Baraldi, E., 235 Barany, G., 196 Barbacid, M., 177, 178, 227, 241 Barchi, J. J., Jr., 173 Barfod, E. T., 219 Barford, D., 286 Barila, D., 248 Barlett, C., 203 Barnett, P., 249 Baron, M. K., 3, 10, 76 Baron, R., 203 Barratt, D. G., 169, 171 Bar-Sagi, D., 87, 225, 241, 251 Bartel, P. L., 250 Bartels, H., 293 Bartels, K., 29 Bartford, D., 191 Bashan, A., 293 Bateman, A., 93, 114, 115 Batista, F. D., 148, 155

Battaglia, V., 3, 104 Battai, N., 152, 154 Batzer, A., 225, 241 Baudouin, B., 128 Baumann, G., 225 Baumann, U., 289, 290, 291 Bax, A., 245 Bax, B., 181 Beamer, L. J., 27 Becker, S., 191, 192 Becktel, W. J., 147 Becquart, J., 189, 195, 226, 240 Bedford, M. T., 251 Behoradsky, B. H., 198 Bell, J. A., 12 Bell, J. I., 150 Belsham, G. J., 279, 280, 283, 284 Benarous, R., 227, 243, 244, 245 Benefenati, F., 220 Benichou, S., 227, 243, 244, 245 Benjamin, D. C., 122 Benjamin, D. R., 115 Bennett, P., 224 Bentley, D. R., 198 Bentley, G. A., 34, 122, 124, 125, 128, 131, 137, 152, 154 Benz, J., 289, 290, 291 Berendsen, H. J., 65 Berg, L. J., 220, 227 Berg, O. G., 45, 47 Bergman, M., 246 Berissi, H., 284 Berman, H. M., 10, 17, 31, 38, 39, 40 Berman, J., 172, 180 Bernhard, A., 198 Berset, C., 279, 280 Berthet-Colominas, C., 152, 154 Best, J. R., 169, 171 Betageri, R., 180 Betsholtz, C., 250, 251 Bhat, T. N., 34, 122, 124, 125, 131, 137, 145, 152, 154 Bibbins, K. B., 172, 174 Bieganowski, P., 88 Bienkiewicz, E. A., 213 Bigelow, J., 204 Bilofsky, H., 121 Birck, C., 35 Bird, L. E., 291 Birge, R. B., 164, 165, 177, 178, 180, 182, 203

AUTHOR INDEX

Birkenhager, R., 282 Birney, E., 93, 114, 115 Bizebard, T., 122 Bjarnegard, M., 250, 251 Blaas, D., 125, 151, 152 Blaber, M., 147 Black, D. L., 91 Blackburn, J. M., 65 Blake, C. C., 218 Blanc, A., 280, 283 Blanco, F. J., 226 Blaney, J. M., 50 Blier, P. R., 152, 154 Blobel, G., 286 Block, C., 86 Blom, N. S., 56 Blomberg, N., 86 Blow, D. M., 139 Blum, S., 288, 289 Blundell, T. L., 181 Bode, W., 29, 31, 36 Boder, E. T., 148, 155 Bodian, D. L., 52, 54 Bodner, S., 92 Boeuf, H., 172, 174 Bogan, A. A., 10, 120, 121, 122, 132 Bohacek, R. S., 203 Bohl, B. P., 222 Bohm, A., 26, 38 Bokoch, G. M., 222 Bolen, J. B., 193 Bolino, A., 198 Bollschweiler, C., 36 Bolotin-Fukuhara, M., 93 Bonanno, J. B., 293 Boniface, J. J., 149, 150 Booker, G. W., 164, 216, 218, 219, 223, 226, 251 Borchert, T. V., 226 Bork, P., 76, 77, 78, 80, 92, 93, 96, 114, 215, 286, 287 Bosecke, P., 122 Bossart-Whitaker, P., 122 Bot®eld, J., 173 Bot®eld, M. C., 194, 195, 219, 221, 225, 235, 237, 249, 252, 254 Both, G. W., 272 Bothwell, A. L., 152, 154 Botstein, D., 100 Bottger, G., 249 Bottomley, M. J., 202, 249

301

Boujemaa, R., 224 Boulot, G., 122, 124, 125, 128, 131, 137, 152, 154 Bourell, J. H., 125 Bourne, H. R., 86 Bousquet, J. A., 213 Bouton, A. H., 224 Bouveret, E., 109 Bowie, J. U., 287 Bowtell, D., 223, 224, 225 Boyd, J., 216, 218, 226 Braden, B. C., 26, 122, 124, 127, 131, 137, 138, 139, 141, 142, 143, 144, 145, 147, 154, 179 Bradley, A., 203 Bradshaw, J. M., 163, 164, 173, 174, 175, 176, 178, 179, 180, 182, 194, 195, 202, 203 Bradshaw, T., 172, 180 Bragado-Nilsson, E., 109 Brandau, O., 198 Brandt, J. T., 193 Brannetti, B., 225, 228 Brasch, M. A., 3 Brauer, A. W., 219, 220, 221, 223, 224, 225, 228, 230, 234, 237, 241 Brawerman, G., 274 Bray, J. E., 83 Bredt, D. S., 249 Breeze, A. L., 164, 169, 171 Brehm, M., 96 Brenner, C., 88 Brenner, S. E., 4, 83 Brick, P., 40, 139 Brickmann, J., 52 Bricogne, G., 34 Briggs, J. M., 47, 60 Broder, Y. C., 105 Brodersen, D. E., 293 Brodin, L., 252 Brooks, C. L. III, 44 Brooks, M. W., 225 Brooksbank, R. A., 198 Broome, M. A., 246 Broutin-L'Hermite, I., 224 Brown, E. L., 213, 214, 227 Brown, J. H., 214 Brown, M., 92 Brown, M. T., 162, 190, 246 Brown, S. L., 221 Browning, K. S., 280

302

AUTHOR INDEX

Brownlee, G. G., 271 Brugge, J. S., 162, 189, 202, 216, 219, 220, 221, 225, 226, 230, 235, 236, 237, 239, 249 Brunger, A. T., 124, 128 Bryant, P. J., 249 Bryant, S. H., 12 Bryngelson, J. D., 48 Bubeck-Wardenburg, J., 185 Buchberg, A. M., 88 Bucher, P., 82 Buckle, A. M., 34 Buckler, A. J., 198 Buday, L., 225 Budisavljevic, M., 128 Buller, A. L., 221 Bundle, D. R., 122, 125 Bunnell, S. C., 220, 227 Burakoff, S. J., 201 Burgess, M. W., 220, 253 Burke, T. R., Jr., 172, 173, 203, 204 Burkhardt, A. L., 193 Burkhart, W., 172, 180 Burley, S. K., 28, 31, 40, 276, 277, 279, 280, 281, 282, 283, 285, 286, 287, 293 Burnham, M. R., 224 Burrell, S. K., 222, 223, 224, 242 Burshtyn, D. N., 181 Buser, P., 288, 289 Bustelo, X. R., 177, 178 Butcher, C., 224 Bycroft, M., 138, 142 Bye, J. M., 198 Byrd, R. A., 227 Bywater, R. P., 234

C C. elegans Sequencing Consortium, 2, 92 Caffrey, D. R., 94 Cagney, G., 3, 104 Cahn, A. P., 198 Callaini, G., 249 Callebaut, I., 80 Calvio, C., 2 Camacho, C. J., 46, 63, 65 Camonis, J. H., 225 Campbell, I. D., 76, 164, 180, 213, 214, 216, 218, 220, 226, 227, 252 Campiglio, M., 88 Canellakis, E. S., 274

Cantley, L. C., 177, 178, 180, 182, 199, 201, 202, 203, 219, 220, 224, 225, 228, 249, 251 Cao, C., 276 Cao, T., 204 Capecchi, M. R., 92 Capel, M., 293 Caravatti, G., 171, 172, 204 Carbone, R., 250, 251 Cardozo, T., 86 Carey, N. H., 274 Carlier, M. F., 224 Carmack, E., 2, 107 Carpenter, C. L., 177 Carson, M., 171, 187, 201, 217 Carter, A. P., 293 Carter, J. M., 225 Carter, P. J., 45, 125, 139, 179 Cartlidge, S. A., 169, 171 Carugo, O., 12 Caruthers, J. M., 289, 292 Casares, S., 218 Casari, G., 86 Case, D. A., 54 Case, R. D., 172, 180 Caspary, F., 109 Cassin, E., 96 Castagnoli, L., 224, 225, 228, 235, 249 Cate, J. H., 293 Cauerhff, A., 140, 143, 145 Cerione, R. A., 219 Cesareni, G., 224, 225, 228, 235, 249, 250 Cestra, G., 224, 225, 228, 235, 249 Chacko, G. W., 193 Chacko, S., 147 Chait, B. T., 2, 227, 236, 243, 244, 245 Chan, A. C., 176, 185, 187, 188, 193, 194, 195 Chance, M. R., 293 Chang, A., 213 Chang, C. Y., 122 Chang, H. C., 122 Chapon, C., 102 Chardin, P., 196, 225, 241 Charifson, P. S., 172, 180, 202, 203 Chatterjee, S., 172, 203 Chaudhuri, M., 172, 177, 178, 180, 182, 203 Chawla, A., 225 Cheadle, C., 225 Cheang, S., 213 Chemama, Y., 104

AUTHOR INDEX

Chen, C., 201, 202 Chen, H., 28, 252 Chen, J. K., 221, 223, 224, 225, 226, 227, 228, 230, 234, 235, 236, 237, 241, 253 Chen, R., 52, 55, 61, 63 Chen, X., 191, 192 Chen, X. Q., 221 Chen, Y., 122, 124 Chen, Y. J., 230 Cheng, G., 219, 220, 243 Cheng, J. W., 226, 230 Cher®ls, J., 26, 36, 52, 53, 59 Cherniack, A. D., 225 Chernoff, J., 222 Cherry, J. M., 100 Chervitz, S. A., 100 Chevalier, S., 280 Cheynet, V., 152, 154 Chia, W., 95 Chiba, T., 3, 104 Chijiwa, T., 85 Chinardet, N., 35 Chishti, A. H., 249 Chishti, Y., 122 Chitarra, V., 34, 122, 125, 154 Cho, Y., 250 Choi, S. K., 276 Chong, L. T., 144 Chook, Y. M., 196, 286 Chothia, C., 4, 10, 11, 13, 14, 17, 18, 19, 20, 21, 22, 23, 27, 29, 30, 31, 32, 33, 34, 35, 36, 75, 83, 120, 121, 122, 123, 147, 148 Chou, M. M., 177, 178, 180, 182, 203 Chowdhry, B. Z., 41 Christinger, H. W., 122, 124 Chuang, R. Y., 288 Cicchetti, P., 221, 225, 231, 237 Cingolani, G., 286 Citterich, M. H., 225, 228 Clackson, T., 45, 133, 137, 179 Claesson-Welsh, L., 184, 220 Clardy, J., 226 Clarke, S., 251 Clemons, W. M., 293 Clore, G. M., 245 Coblentz, B., 249 Coffey, A. J., 198, 199 Coggeshall, K. M., 193 Cohen, F. E., 86, 227, 232, 234, 235, 254, 255 Cohen, G. B., 215 Cohen, G. H., 29, 43, 49, 121, 122, 147, 154

303

Colgan, D. F., 274 Collette, Y., 244 Colman, P. M., 16, 21, 122, 123 Combs, A. P., 225, 227, 253 Conejero-Lara, F., 218 Conery, J. S., 92 Conklin, E., 109 Connolly, M. L., 16, 51, 52 Conover, D., 104 Conroy, S. C., 288 Consler, T. G., 202, 203 Cooper, J. A., 162, 177, 190, 246 Copeland, N. G., 284 Copley, R. R., 77, 78, 80, 93, 96, 114, 215 Coppola, J. A., 271 Corbalan-Garcia, S., 251 Corey, S. J., 193 Cornette, J. L., 63 Cornille, F., 226, 230, 241 Corpet, F., 81 Cory, G. O., 251 Cossart, P., 80 Costello, P., 251 Coulson, A., 96 Courtneidge, S. A., 184, 190, 197, 214, 216, 217, 226, 245, 246 Cousins-Wasti, R., 180 Coutavas, E., 112 Covarrubias, M., 125 Covell, D. G., 144 Cowan, P. M., 212 Cowburn, D., 163, 164, 165, 166, 167, 169, 175, 178, 185, 196, 199, 220, 222, 223, 224, 226, 227, 235, 241, 242, 243, 252 Cox, J. P., 148 Craig, A. W., 283, 284 Cramer, W. A., 109 Crane, D. I., 249 Creamer, T. P., 213, 214, 240 Croce, C. M., 88 Crosby, R., 172, 180, 203 Cullen, B. R., 243 Cunningham, A. M., 86 Cunningham, B. C., 122, 124 Curran, M., 204 Curtis, P. S., 283 Cusack, S., 152, 154 Cussac, D., 196, 226, 230 Czech, M. P., 225 Czernik, A. J., 220

304

AUTHOR INDEX

D Daaka, Y., 37 Dahiyat, B. I., 65 Dalgarno, D. C., 216, 218, 221, 225, 226, 228, 230, 234, 237, 252, 254 Dall'Acqua, W., 26, 124, 125, 127, 131, 132, 137, 138, 139, 141, 142, 143, 144, 145, 152, 154, 179 Daly, R., 225, 241 Damelin, M., 114 Damke, H., 251 Darnell, J. E., 274 Darnell, J. E., Jr., 179, 191, 192 Darzynkiewicz, E., 272 Das, P., 213, 219, 220, 223, 227, 249, 251 Dasgupta, S., 12 Daude-Snow, L. F., 253 Dauter, Z., 122, 154 David, C., 224, 225, 251, 252 Davidson, A. R., 218 Davies, D. R., 29, 43, 49, 121, 122, 123, 147, 154 Davies, K., 92 Davis, A. R., 225 Davis, J. R., 198 Davis, M. E., 60, 62 Davis, M. M., 150 Davis, R., 172, 180 Day, J., 122 Dear, P. H., 148 De Camilli, P., 220, 224, 225, 251, 252 Deconinck, N., 92 Degenhardt, K. R., 251 De Gregorio, E., 284 Dehoux, P., 80 Deisenhofer, J., 29, 31, 36 Deiss, L. P., 284 Dekel, I., 89 Delbaere, L. T., 122, 154 De Leo, F. R., 225 DeLisi, C., 46, 63, 65 De Lotto, R., 95 Dente, L., 224, 225, 228, 235, 249 Der, C. J., 219, 225 De Reuse, H., 3, 104 Desai, S., 109 Deshimaru, M., 85 Deshmukh, M., 280, 283 Desiderio, S. V., 124, 128 de St. Basile, G., 198

D'Eustachio, P., 112 Deutscher, M., 106 Dever, T. E., 276, 288 de Vos, A. M., 30, 122, 124, 133 De Vries, J. E., 198, 199 Dhand, R., 181, 219, 223, 251 Dhe-Paganon, S., 188, 191 Diat, O., 122 Dickson, B., 225 Di Fiore, P. P., 220, 222, 225, 226, 249, 250, 251 Distel, B., 249 Dixit, V. M., 91 Dixon, J. S., 59 Dobson, C. M., 115, 218, 225 Dodson, E. J., 34 Dodson, G. G., 34 Doel, M. T., 274 Doerks, T., 77, 78, 80, 215 Doi, M., 278 Dolinski, K., 100 Domchek, S. M., 172, 180, 203 Donaldson, I., 3 Dong, J., 213 Donze, O., 279, 280, 283 Doolittle, R. F., 78 Doolittle, W. F., 93 Doshi, A., 246 Doshi, N., 218 Dove, S. L., 105 Dowbenko, D., 223 Downing, A. K., 164, 216, 218, 226 Downward, J., 225 Doye, V., 102 Draetta, G., 197, 246 Draganescu, A., 88 Dragenescu, A., 88 Drake, A. F., 212 Dreyfuss, G., 219, 220, 276 Driggers, E. M., 149, 150 Driscoll, P. C., 194, 202, 216, 218, 226, 249 Druck, T., 88 Duan, Y., 144 Duchesne, M., 241 Duckworth, B., 177 Ducruix, A., 189, 195, 224, 226, 240 Duke-Cohen, J. S., 249 Dumas, C., 220, 226, 227, 243, 244, 245 Duncan, R., 288 Dunham, A., 198 Duperon, J., 96

305

AUTHOR INDEX

Duquerroy, S., 52, 53 Durbin, R., 93 Durham, J. D., 198 Durkin, J., 182, 203 Dwight, S. S., 100 Dyson, H. J., 155, 282

E Earnest, T. N., 293 Echeverri, C., 96 Eck, M. J., 165, 166, 167, 169, 181, 188, 189, 190, 191, 199, 201, 214, 226, 235, 246 Eddy, S. R., 93 Edery, I., 288 Edgington, T. S., 122 Edmonds, M., 274 Edmundson, A. B., 152, 154, 155 Egan, S. E., 225 Egerton, M., 222 Egile, C., 224 Eigenbrot, C., 125 Eijuwade, T., 272 Eisele, J. L., 34, 122, 125, 154 Eisen, H. N., 155 Eisenberg, D., 3, 5, 10, 82, 87, 88, 147 Eisenstein, E., 127, 132 Eisenstein, M., 26, 36, 56, 59 Ekiel, I., 272 Elcock, A. H., 47, 60, 61, 62, 63 Elenbaas, B., 282 Ellis, C., 161, 164, 189 Eng, J., 2, 107 Engen, J. R., 196 England, L., 164 Engstro Èm, A., 184 Enright, A. J., 5, 87 Erdmann, D., 171, 172 Erickson, H. P., 46, 62 Eriksson, A. E., 147 Ermak, D. L., 60 Ernberg, I., 199 Erpel, T., 246 Escobar, C., 122 Escobedo, J. A., 164, 177 Espanel, X., 86, 213 Esposito, G., 47 Estrov, Z., 252 Etchison, D., 287 Eulitz, M., 252 Evans, C. A., 1

Evans, P. R., 226, 235 Evans, S. V., 122 Evans, T., 219

F Fabbro, D., 204 Fadool, D. A., 219 Fajardo, J. E., 177, 178, 180, 182, 203 Falquet, L., 82 Fantl, W. J., 164, 177 Farmer, B. T. II, 227, 241 Farrow, N. A., 181 Fassler, R., 198 Favre, D., 279 Favreau, M., 274 Federwisch, M., 181 Fedorov, A. A., 235 Fedorov, E., 235 Feldman, P. L., 203 Feldman, R. A., 177, 178 Feller, S. M., 222, 223, 224, 227, 235, 241, 242 Feng, S., 219, 221, 225, 227, 228, 230, 234, 235, 236, 237, 239, 241, 253, 254 Ferguson, K. M., 86 Ferguson, M., 198 Fernley, R., 223, 224, 225 Ferrin, T. E., 50 Fersht, A. R., 34, 42, 43, 45, 46, 62, 65, 138, 139, 140, 142, 147, 179, 218 Field, A. S., 271 Fields, B. A., 26, 122, 124, 127, 131, 139, 140, 141, 143, 145, 147 Fields, S., 1, 2, 3, 5, 9, 103, 104, 250 Fieser, T. M., 124, 153 Finan, P., 224 Fine, R. F., 147 Finkelstein, A. V., 44 Finzel, B. C., 43, 49, 122 Fischer, D., 54, 55, 58 Fisher, H. F., 41 Fisher, R., 92 Fita, I., 125, 151, 152 Fitch, W. M., 91 Flannery, B. P., 56 Fletcher, C. M., 276, 281, 282 Flohrs, K., 96 Flynn, D. C., 220 Fogolari, F., 47 Fontecilla-Camps, J. C., 33, 122

306

AUTHOR INDEX

Foote, J., 155 Ford, C. E., 37 Ford, G. C., 77, 82 Forman-Kay, J. D., 166, 171, 174, 181, 199, 218 Fowlkes, D. M., 219, 221, 222, 223, 225, 235, 236, 237, 250 Fox, R. O., 227, 228, 232, 241 Franceschi, F., 293 Franco, B., 198 Frank, J., 293 Frankel, A., 251 Franken, P., 220, 226, 227, 243, 244, 245 Fraser, S. G., 96 Frech, M., 196 Frederick, C. A., 189, 226 Freire, E., 218 French, S., 225 Fretz, H., 171, 172 Freund, C., 249 Friedman, R. A., 47 Friedrich, T., 36 Friesem, A. A., 56, 59 Frisch, C., 34, 42, 43 Frischknecht, F., 191, 248 Friso, G., 2 Froloff, N., 44 Frolow, F., 47 Fromage, N., 189, 195, 226, 240 Frontali, L., 93 Fry, M. J., 181, 219, 223, 251 Fu, C., 185 Fucini, P., 115 Fujii, S., 147 Fujinaga, M., 26 Fukata, M., 111 Fukui, Y., 164 Fukumaki, Y., 85 Fukunaga, R., 282, 284 Fukuzawa, M., 162, 201 Fumagalli, S., 197, 246 Furet, P., 171, 172, 204 Furuichi, Y., 272 Fushman, D., 196 Fu È tterer, K., 176, 185, 187, 188, 193, 194, 195

G Gaasterland, T., 293 Gabb, H. A., 56, 57, 58

Gabdoulline, R. R., 46, 60, 62 Gad, H., 252 Gaffney, P. R. J., 202 Gale, N. W., 225 Galisson, F., 102 Galisteo, M. L., 222 Gamble, T. R., 35 Gampe, R. T., 203 Gao, G. F., 150 Garbay, C., 213, 224, 226, 230, 241 Garcia, K. C., 124, 128 Garcia, R., 122 Garcia-Echeverria, C., 171, 172 Gardiner, E. J., 55, 63 Garey, J. R., 93 Garner, C. C., 249 Garvik, B. M., 2, 107 Gatei, M., 222 Gautel, M., 226 Gavin, A. C. 2, 6 Gay, B., 171, 172, 204 Geahlen, R. L., 193 Geha, R., 198, 199, 223 Geniteau-Legendre, M., 128 Genovese, C., 198 Gentile, L. N., 196 George, C., 173 Gerald, W., 198 Germain, V., 194 Gershon, P. D., 271, 279 Gerstein, M., 17, 83 Gertler, F., 235 Geske, R., 203 Getzoff, E. D., 122, 124, 154 Gherardi, E., 148 Ghio, C., 54 Giardina, G., 250, 251 Gibson, T. J., 80 Giddings, B. W., 225 Giesing, H. A., 44 Gigant, B., 122 Gilbert, D., 5, 147 Gilbert, W., 78 Gill, D. S., 42, 43 Gill, S. J., 44 Gillis, J. M., 92 Gilmer, T. M., 172, 180, 203 Gilson, M. K., 38, 54, 60, 62, 63 Gincel, E., 226, 230 Gingras, A.-C., 276, 277, 279, 280, 281, 282, 283, 284

307

AUTHOR INDEX

Giot, L., 3, 104 Giovedi, S., 220 Gish, C. D., 182 Gish, G. D., 86, 161, 166, 171, 177, 178, 180, 181, 182, 196, 199, 203 Gish, W., 77 Glassman, R. H., 222 Glover, R. T., 221 Gluemann, M., 293 Gmeiner, W. H., 196, 227 Go, M., 76 Goddsell, D. S., 26 Godwin, B., 104 Goesson, A., 96 Goldbaum, F. A., 26, 122, 124, 127, 131, 139, 140, 143, 145 Goldfarb, V., 227, 241 Goldman, E. R., 127, 132, 137, 138, 139, 141, 142, 143, 144, 145, 179 Go Ènczy, P., 96 Gon¯oni, S., 190, 191, 214, 226, 246, 248 Goodsell, D. S., 12 Gorbalenya, A. E., 289 Gorga, J. C., 214 Gorina, S., 227, 236, 250, 282 Gosser, Y. Q., 191, 196, 226 Goudreau, N., 226, 230, 241 Gould, S. J., 249 Gout, I., 164, 181, 216, 218, 219, 223, 224, 226, 251 Gouzy, J., 81 Goyer, C., 272, 280, 283 Grabs, D., 224, 225, 251, 252 Gradi, A., 280, 283, 284 Gram, H., 225 Grandi, P., 102 Grandori, C., 246 Grantcharova, V. P., 218 Grassucci, R. A., 293 Gray, N. S., 149 Graziani, A., 177 Grazioli, L., 194 Greaves, S., 102 Green, J., 173, 185, 194 Green, M., 172, 180 Green, O. M., 219, 221, 225, 237, 249 Green, S. M., 140 Greene, M. I., 124, 125, 131, 137, 152, 154 Greengard, P., 220 Greenwood, A., 122 Greenwood, D. J., 194

Gregoire, C., 33, 122 Grey, L., 164 Gribskov, M., 82 Grishin, N. V., 85 Gronenborn, A. M., 245 Groner, B., 191, 192 Grossman, S. R., 27 Grosveld, G., 92 Groves, M. R., 286 Grucza, R. A., 176, 178, 187, 188, 193, 194, 195 Grutter, M. G., 171, 172 Grygon, C. A., 180, 194 Grzesiek, S., 245 Gu, H., 218 Guan, W., 203 Guddat, L. W., 155 Guijarro, J. I., 218, 225 Guillet, V., 34, 43 Guilliot, S. D., 225 Guilloteau, J. P., 189, 195, 226, 240 Gunde®nger, E. D., 249 Gu È nther, U. L., 181 Gutkind, S., 250, 251 Guzikevich-Guerstein, G., 47 Gygi, S. P., 279, 280, 283

H Haag, A., 148 Haapaniemi, P., 226 Haber, D. A., 198, 272 Hachimori, Y., 250 Hafen, E., 225 Haghighat, A., 279, 283, 284 Halay, E. D., 28 Halperin, I., 55, 63 Hamlin, R., 40 Hamm, H. E., 26, 38 Hanafusa, H., 164, 165, 177, 178, 180, 182, 197, 203, 222, 223, 224, 227, 235, 241, 242, 246 Hanazono, Y., 223, 241 Handa, R., 194 Hanlon, N., 286 Hannak, E., 96 Hannenhalli, S. S., 86 Hansen, J., 293 Hansson, H., 226 Hard, T., 226 Harkin, D. P., 198

308

AUTHOR INDEX

Harley, V. R., 122 Harlow, E., 107 Harms, J., 293 Harpaz, Y., 17 Harpur, A. G., 249 Harrata, K., 154 Harris, A., 198 Harris, M. A., 100 Harrison, A. P., 83 Harrison, R. W., 56, 154 Harrison, S. C., 165, 166, 167, 169, 181, 188, 189, 190, 196, 214, 226, 246, 249 Hart, M. J., 219 Harte, M. T., 224 Hartl, F. U., 115 Hartley, J. L., 3 Hartley, R. W., 34 Hartsch, T., 293 Hasel, K., 122 Haser, W. G., 177, 178, 180, 182, 189, 203, 226 Hashimoto, Y., 241 Hassel, A. M., 172, 180 Hatada, M. H., 185, 194 Hatanaka, H., 216, 226, 241 Hattori, M., 3, 104 Hattori, S., 85, 241 Hattula, K., 248 Haystead, T. A. J., 279 Hazbun, T. R., 3 He, X. M., 152, 154, 155 Heath, P., 198 Hecht, J. L., 47 Hecht, S. M., 272 Hegde, R. S., 27 Heger, A., 81 Hegler, J., 270 Heiden, W., 52 Heinz, D. W., 147 Heldin, C.-H., 184 Hellen, C. U., 276, 283, 284, 285, 286, 287, 288 Hemmings, B. A., 286 Hendrickson, W. A., 34, 190 Hendsch, Z. S., 44, 45 Hengartner, H., 148 Henkemeyer, M., 225 Hennigan, F. L., 288 Henry, A. H., 124 Henry, P. A., 220, 225, 235

Hensmann, M., 180 Hentze, M. W., 270, 284 Hermann, R. B., 147 Herrmann, G., 58 Herron, J. N., 152, 154, 155 Herschlag, D., 289, 292 Hershey, J. W., 275, 276, 288 Hershkovits, J. S., 251 Hery, S., 65 Hibbits, K. A., 42, 43 Hiipakka, M., 245 Hildebrand, J. D., 224 Hiles, I. D., 219, 223, 251 Hill, C. P., 35 Hilvert, D., 65 Hilyard, K., 194 Hinshaw, J. E., 251 Hinshelwood, S., 251 Hirai, H., 197, 219, 223, 241 Hjertaas, K., 2 Ho, Y., 2, 6 Hobson, K., 222 Hochschild, A., 105 Hodawadekar, S. C., 88 Hodel, A. E., 279 Hof, P., 188, 191 Hoffken, W., 36 Hoffman, A., 28 Hoffman, N. G., 219, 221, 222, 223, 225, 235, 236, 237, 250 Hoffmuller, U., 224, 225, 235, 249 Hofmann, K., 82 Hogue, C. W., 3 Hoh, F., 220, 226, 227, 243, 244, 245 Holm, L., 81 Holmes, T. C., 219 Holt, D. A., 203 Holt, R. A., 1 Honig, B., 44, 45, 47, 54, 147, 175, 231 Horita, D. A., 227 Horiuchi, M., 250 Horovitz, A., 138, 142 Hou, W., 219, 220 Houdusse, A., 34, 122, 125, 154 Hough, C., 249 Houseweart, M., 35 Housman, D. E., 198 Housset, D., 33, 122 Howe, K. L., 93 Howell, B., 190 Howell, B. W., 246

309

AUTHOR INDEX

Howell, G. R., 198 Howk, R., 225 Hseuh, Y. P., 249 Hsuan, J., 219, 223, 224, 251 Hu, J. S., 245 Hu, P., 172, 180 Huang, J., 88, 251 Huang, M., 122 Huang, X. Y., 235, 236 Hubbard, T. J., 4, 83 Hubbs, A. E., 222 Huber, R., 29, 31, 36 Huckel, E., 198 Huckle, W. R., 181 Huebner, K., 88 Hummel, C. W., 202, 203 Hunter, T., 163, 282, 284 Hurley, J. H., 80 Hurt, E. C., 102 Huse, M., 164, 191, 197, 247, 252 Hussey, R. E., 122 Huyer, G., 181, 182 Hyman, A. A., 96 Hyvonen, M., 235

I Iba, H., 246 Ichikawa, S., 226 Ihle, J. N., 92 Ilin, A., 60 Iliopoulos, I., 5, 87 Imataka, H., 276, 280, 282, 283, 284, 287, 288 Inagaki, F., 216, 226, 250 Ingraham, R. H., 180, 194 Innerarity, T. L., 284 Inoue, M., 278 International Human Genome Sequencing Consortium, 1 Isakoff, S. J., 86 Isern, J., 141, 147 Ishida, T., 278 Ishii, S., 241 Ito, T., 3, 104 Ivashchenko, Y., 225 Iwabuchi, K., 250 Iwamatsu, A., 111, 223, 241 Iyengar, R., 37 Iyer, G. H., 12 Iyo, H., 278

J Jack, A., 34 Jackson, P. K., 172, 173, 174 Jackson, R. J., 273, 285 Jackson, R. M., 56, 57, 58 Jacobson, A., 274 Jacques, S., 166, 171, 181 Jain, M. L., 226 Jakes, K., 270 Jakes, S., 180, 194 Jakobsen, B. K., 150 James, M. N., 26 Janell, D., 293 Janin, J., 4, 10, 11, 12, 13, 14, 15, 17, 18, 19, 20, 21, 22, 23, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 40, 43, 44, 46, 48, 49, 50, 51, 52, 53, 59, 75, 120, 121, 122, 123 Janssen, O., 219, 220 Janvier, K., 244 Jardetzky, T. S., 214 Jaye, M., 225 Jeffrey, P. D., 250 Jenkins, N. A., 284 Jensen, L. H., 154 Jeong, H., 5, 6 Jermyn, K. A., 162, 201 Jeruzalmi, D., 191, 192 Jiang, F., 50, 56 Jimenez, J. L., 218, 225 Jin, Y., 272 Joachimiak, A., 47, 235 Jockel, J., 86 Johnson, C. M., 34, 42, 43 Johnson, D. E., 164, 177 Johnson, E. R., 289, 290, 292 Johnson, J. E., 44 Johnson, L. N., 190 Johnson, S., 155 Johnsson, N., 2, 105 Johnston, M., 104 Johnston, S. A., 108 Jones, A., 198 Jones, D. T., 4 Jones, J. A., 218 Jones, P. T., 148 Jones, S., 10, 11, 12, 16, 17, 22, 23, 31, 37, 38, 39, 40, 120, 122, 123, 124 Jones, S. J. M., 96 Jones, T., 89 Jordan, S. R., 172, 180, 202, 203

310

AUTHOR INDEX

Joung, J. K., 105 Jove, R., 197 Judson, R. S., 3, 104 Juenemann, R., 115 Jutila, K. L., 225

K Kabat, E. A., 121 Kaelin, W. G., Jr., 250 Kahmann, J. D., 213, 214, 227 Kahn, D., 81 Kaibuchi, K., 111 Kalb¯eisch, T., 104 Kalhammer, G., 86 Kalinke, U., 148 Kalume, D. E., 2 Kamath, R. S., 96 Kaminski, A., 285 Kam-Morgan, L., 147 Kang, H., 249 Kapeller, R., 177, 219, 220 Kaplan, D. R., 164, 177 Kaplan, S., 225 Kapoor, T. M., 225, 227, 253, 254 Kara, B. V., 169, 171 Karas, J. L., 185, 194 Kardinal, C., 241, 252 Karlin, S., 77 Karlsson, T., 220 Karplus, K., 82 Kasahara, C., 219, 221, 227, 235, 237 Kashishian, A., 177 Kassel, D., 172, 180 Katchalski-Katzir, E., 26, 36, 56, 59 Kato, J. Y., 246 Kato, M., 249 Katz, S., 105 Katzav, S., 89 Kaufman, J., 245 Kavanaugh, W. M., 164, 173, 177 Kawata, T., 162, 201 Kay, B. K., 211, 212, 215, 219, 221, 222, 223, 225, 235, 236, 237, 250 Kay, C. M., 196 Kay, L. E., 166, 171, 181, 199, 218 Kazlauskas, A., 177 Kedar, P., 222 Keil, P., 110 Kelley, R. F., 125 Kellie, S., 224

Kellis, J. T., Jr., 147 Kelly, M. J., 218 Kelly, S., 221 Kemp, G. J., 58, 59 Kennedy, D., 249 Khan, D., 92 Khanna, K. K., 222 Kharbanda, S., 222 Khitrina, E. V., 285 Kim, D. E., 218 Kim, E., 249 Kim, H. C., 279 Kim, M., 36 Kim, P. S., 220, 253 Kim, S. H., 50, 56 Kimber, M. S., 86 Kimchi, A., 284 Kimery, M., 172, 180 Kimura, S. R., 63 Kinder, D., 172, 180 Kindler, S., 249 Kinet, J. P., 251 King, A., 2 King, C. R., 204 King, F., 177, 178, 180, 182, 203 Kinnon, C., 251 Kirchhausen, T., 220 Kirchhoff, P. D., 62 Kirchweger, R., 276, 284 Kirkham, M., 96 Kishan, K. V., 225, 226 Kitamura, K., 278 Kitamura, N., 249 Klarlund, J. K., 225 Klein, A. T., 249 Kleinberg, M. E., 251 Klemke, R., 221 Kline, C., 40 Knaus, U. G., 222 Knight, J. R., 3, 104 Knill-Jones, J., 139 Knossow, M., 122 Knudsen, B. S., 222, 223, 224, 227, 235, 241, 242 Kobayashi, K., 111 Koch, C. A., 161, 164, 189 Koegl, M., 197, 246 Koekoek, R., 77, 82 Koh, C. G., 221 Kohda, D., 216, 226, 241 Kolafa, J., 234

AUTHOR INDEX

Kollman, P. A., 54, 144 Kolluri, R., 220 Kolmerer, B., 226 Kolquist, K. A., 198 Kolupaeva, V. G., 285 Kominos, D., 164, 165, 166, 171, 178, 181 Kondo, H., 122, 124 Kong, X., 279 Konkol, B., 252 Koonin, E. V., 76, 80, 91, 92, 100, 289 Korneeva, N. L., 288 Korostensky, C., 2 Kortemme, T., 218 Kossiakoff, A. A., 30 Kourinov, I. V., 56 Koyama, S., 216, 218, 226 Kozak, M., 273, 276 Kozlov, S., 222 Kozma, L. M., 225 Krag, D., 204 Krainer, M., 198 Kraut, J., 19, 35 Kretzschmar, J., 248 Kreychman, J., 83 Kruh, G. D., 222 Kuang, W. J., 219 Kufe, D., 222 Kuge, H., 271 Kuhlendahl, S., 249 Kukla, D., 29 Kumagai, I., 122, 124 Kumar, A., 44 Kuntz, I. D., 26, 36, 50, 52, 54, 59, 144 Kurakin, A., 225 Kurata, T., 241 Kurinov, I. V., 154 Kuriyan, J., 87, 163, 164, 165, 166, 167, 169, 171, 175, 178, 181, 185, 190, 191, 192, 197, 199, 214, 220, 226, 227, 235, 236, 241, 242, 243, 244, 245, 246, 247, 248, 252 Kuroda, S., 111 Kussie, P. H., 282 Kwong, P. D., 34 Kyrpides, N. C., 5, 87

L Labadia, M. E., 194 Labigne, A., 104

311

Ladbury, J. E., 41, 47, 150, 172, 173, 174, 178, 180, 194, 196, 213, 220, 226, 227, 243, 244 LaFevre-Bernt, M., 191, 197, 248, 252 Lai, J., 228 Laimins, L. A., 27 Laird, A. D., 246 Laird, E. R., 185, 194 Lake, J. A., 93 Lambright, D. G., 26, 38 Lamond, A., 2 Lamphear, B. J., 276, 284, 288 Lande, R., 93 Lander, 80, 82, 92, 93 Landgren, E., 220 Lane, D., 107 Lane, W. S., 223, 224, 241 Langridge, R., 50 Lansing, T. J., 246 Lanyi, A., 198 Lappe, M., 5 Lapthorn, A., 34, 43 Larsen, T. A., 26 Larson, S. M., 218 Laskowski, M., Jr., 26 Laskowski, R. A., 17, 21 Lasky, L. A., 223 Laver, W. G., 122 Lavergne, C., 220 Lavi, S., 272 Lavin, M. F., 222 Lavoie, T., 227, 241 Lawrence, C. E., 12 Lawrence, J. C., 279, 280, 283 Lawrence, M. C., 16, 21, 123 Layton, M. J., 194, 249 Lazdunski, M., 33, 35 Le, H. V., 182, 203 Lea, S., 121 Lechleider, R. J., 177, 178, 180, 182, 203 Lechner, M., 105 Le Clainche, C., 224 Leder, P., 251 Lee, B., 11 Lee, C. H., 164, 166, 171, 181, 191, 197, 220, 225, 227, 236, 243, 244, 245, 247, 252 Lee, D., 83 Lee, H. S., 279, 280, 283, 284 Lee, J. H., 276 Lee, J. S., 122, 154

312

AUTHOR INDEX

Lee, R. H., 52 Lee, S. Y., 274 LeFevre-Bernt, M., 247 Legrain, P., 102, 104, 105 Lei, M., 246 Leidel, S., 96 Lemmon, M. A., 86, 172, 173, 174, 196, 220, 243 Lenoir, G., 198 Lentz, P. J., 77, 82 Lenzen, G., 3, 104 Leparc, G. G., 228 Lerner, R. A., 124, 153 Lescar, J., 34, 122, 125, 154 Lesk, A. M., 21, 22, 123, 148, 218 Leu, T. H., 220 Leung, B., 220, 243 Leung, T., 221 Leversha, M., 198 Levine, A. J., 282 Levitan, I. B., 219 Levitt, M., 51, 123, 142, 218 Levitz, S., 198 Levy, J. B., 246 Levy-Strumpf, N., 284 Li, B., 122, 124, 250 Li, E., 220 Li, H., 36, 122, 124, 137, 138, 141, 143, 144, 145, 152, 154, 179, 276, 281, 282 Li, L., 218 Li, N., 225, 241 Li, P. W., 1 Li, Q., 288 Li, S.-C., 199 Li, Y., 104, 122, 124, 137, 138, 141, 143, 144, 145, 152, 154, 179 Li, Z., 86, 181 Liang, C., 103 Liang, J., 226 LiCata, V. J., 140 Lichtarge, O., 86 Lichy, J. H., 222 Lim, K., 122 Lim, L., 221, 222, 223, 251, 274 Lim, W. A., 227, 228, 232, 234, 235, 241, 244, 246, 254, 255 Lin, D., 293 Lin, H., 252 Lin, S. C., 230 Lin, S. L., 54, 55, 58 Lin, T.-A., 279, 280, 283

Lin, W., 137, 138, 141, 143, 144, 145, 179 Linder, P., 288, 289 Linford, L. S., 93 Link, A. J., 2, 107 Linssen, A. B., 65 Lipman, D. J., 77 Liu, J. H., 36, 122 Liu, Y. C., 201 Liu, Y. S., 181 Liu, Z., 288 Llewelyn, M. B., 148 Lockshon, D., 3, 104 Lo Conte, L., 10, 13, 14, 18, 19, 20, 22, 23, 27, 30, 31, 32, 33, 34, 35, 36, 83, 120, 121, 122, 123 Lomakin, I. B., 276, 283, 285, 286, 287 Long, E. O., 181 Long, Y. Q., 204 Loo, T. H., 221 Look, A. T., 198 Lopez-Mayorga, O., 218 Lorsch, J. R., 289, 292 Lou, M., 185, 194 Lou, Y. C., 226 Lovett, M., 198 Low, P., 252 Lowe, D. M., 139 Lowenstein, E. J., 225 Lowman, H. B., 122, 124 Lu, H., 282 Lu, W., 223 Lu, X., 3, 185, 194 Lukas, S. M., 194 Luke, R. W., 169, 171 Lung, F. D., 204 Lung, F. T., 204 Luo, J. H., 204 Luse, D. S., 271 Luther, M., 172, 180 Luty, B. A., 60, 62 Lynch, M., 92, 224, 225, 251 Lyons, D. S., 150 Lyu, P. C., 230

M Ma, Y. C., 236 MacArthur, M. W., 212 MacBeath, G., 113 MacCullum, R. M., 123, 124 Macias, M. J., 235

AUTHOR INDEX

Madden, T. L., 77 Mader, S., 279, 284 Madura, J. D., 60, 62 Mahoney, N. M., 235 Maignan, S., 189, 195, 226, 240 Majidi, M., 222 Makarova, K. S., 85 Makhatadze, G. I., 42, 44 Malby, R. L., 122 Malchiodi, E. L., 122, 124, 131, 139, 140, 141, 143, 145, 147 Maldonado, E., 271 Malissen, B., 33, 122 Mallet, F., 152, 154 Mandiyan, V., 216, 226, 241, 250 Manivel, V., 150 Manley, J. L., 274 Manley, N. R., 92 Mann, M., 2, 109, 162, 201 Manser, E., 221, 222, 223, 251 Mans®eld, T. A., 3, 104 Mapelli, C., 227, 241 Marcilla, A., 223 Marcotrigiano, J., 276, 277, 279, 280, 281, 282, 283, 285, 286, 287 Marcotte, E. M., 3, 5, 10, 87, 88 Marechal, V., 282 Marengere, L. E., 182 Margolis, B., 166, 171, 181, 225, 241 Mariuzza, R. A., 26, 36, 120, 122, 124, 125, 127, 131, 132, 137, 138, 139, 140, 141, 142, 143, 144, 145, 147, 152, 154, 179 Marks, C. B., 149 Marks, J. D., 148 Maroun, M., 105 Marquart, M., 31, 36 Marr, C. S., 185, 194 Marraccino, R., 250 Marsolier, M., 105 Martin, A. C., 123, 124 Martin, C., 96 Martin, G. A., 164, 177 Martin, G. S., 164 Martinez, J. C., 218, 226 Martinez-Campos, M., 96 Marugasu-Oei, B., 95 Maruizza, R. A., 141, 147 Maruyama, H., 278 Mason, S. P., 5, 6 Massova, I., 144 Matheiu, M., 226

313

Matsuda, M., 164 Matsumura, M., 147 Matsuo, H., 276, 281, 282 Matsushima, M., 122, 124 Matsuura, Y., 111 Matthews, B. W., 147 Mattson, G., 109 Mattsson, P. T., 226 Matuoka, K., 241 Mauguen, Y., 34, 43 Mayer, B. J., 164, 165, 172, 173, 174, 178, 196, 215, 221, 223, 225, 226, 229, 231, 237, 246 Mayer, J. P., 222, 223, 224, 242 Mayo, S. L., 65 Mayr, L. M., 220, 253 Mazo, A., 88 Mazza, G., 33, 122 Mbamalu, G., 182, 225, 251 McCammon, J. A., 38, 47, 60, 61, 62, 63, 64 McCarthy, J. E., 282 McConnell, S. J., 237 McCormick, F., 164, 177 McDonald, I. K., 23, 25 McGavin, S., 212 McGee, A. W., 249 McGlade, J., 177, 178 McGuire, A. M., 276, 281, 282 McKay, C., 92 McKay, D. B., 289, 290, 292 McKimm-Breschkin, J. L., 122 McLachlan, A. D., 82, 147 McMahon, H. T., 226, 235 McNemar, C., 182, 203 McPherson, A., 122 McPherson, A., Jr., 77, 82 Meeker, A. K., 147 Meerovitch, K., 288 Mehrotra, M., 172, 180 Meindl, A., 198 Melamud, E., 11 Melcher, K., 108 Mely, Y., 213 Mendecki, J., 274 Mendoza, M., 225, 249, 250 Meng, W., 201 Merrick, W. C., 275, 276, 288 Merschak, P., 105 Methot, N., 289 Metz, A. M., 280 Meyer, M., 58, 59

314

AUTHOR INDEX

Meyer, T., 204 Meyers, C. A., 227, 241 Michael, W. M., 219, 220 Midelfort, K. S., 148, 155 Migone, N., 224, 225, 235, 249 Milburn, S. C., 287, 288 Milhollen, M. A., 220, 253 Miller, S., 21, 22 Miller, W. T., 77, 164, 191, 197, 235, 247, 248, 252, 254 Millevoi, S., 226 Milligan, R. A., 64 Milpetz, F., 77 Milstead, M. W., 203 Minenkova, O., 224, 225, 235, 249 Minor, D. L., Jr., 220, 253 Mirny, L. A., 218 Mirza, U. A., 227, 236, 243, 244, 245 Misra, V. K., 47 Misselwitz, R., 213 Mitani, K., 223, 241 Mitaxov, V., 163, 164, 173, 174, 175, 176, 179, 182, 194, 195, 202 Miyazawa, K., 249 Mize, G. J., 2, 107 Moare®, I., 190, 191, 197, 214, 226, 247, 252 Moeri, N., 82 Mohammadi, M., 225 Mohr, C., 202, 203 Mohr, S., 100 Monaco, S., 152, 154 Monaco-Malbet, S., 152, 154 Monaghan, D. T., 221 Mongiovi, A. M., 225, 249, 250 Montgomery, C., 203 Montminy, M. R., 282 Moon Woody, A., 213 Moore, J. D., 226, 235 Moore, P. B., 293 Moran, M. F., 161, 164, 189 Moras, D., 77 Moreau, J., 282 Morelock, M. M., 180 Morgan, D. O., 246 Morgan, M., 271, 272, 276 Morgan, R., 288 Morgan-Warren, R. J., 293 Morgensen, S., 109 Morgenstern, J. P., 185, 194 Mori, S., 184 Morino, S., 276, 280, 283, 284, 287, 288

Morioka, H., 278 Morishita, T., 241 Morken, J. P., 254 Morley, S. J., 283, 285 Morra, M., 198, 199 Morris, D. R., 2, 107 Morrogh, L. M., 251 Mortin, M. A., 89 Morton, C. J., 213, 214, 220, 227, 252 Morton, R. T., 44 Mott, R., 77 Moult, J., 11 Moura, J. J. 55, 66 Mourey, L., 35 Moyer, M., 172, 180 Mueller, L., 227, 241 Mueller, M., 288 Muhandiram, D. R., 181 Mui, P., 182, 203 Mu È ller, C. W., 191, 192, 286 Muller, Y. A., 122, 124 Mundorff, E., 149, 150 Munroe, D., 274 Mural, R. J., 1 Murken, J., 198 Murphy, S. M., 246 Murzin, A. G., 83 Musacchio, A., 189, 213, 216, 217, 218, 221, 225, 226, 227, 228, 230, 236, 239, 245, 249, 250 Muthukrishnan, S., 271, 272 Myers, E. W., 1 Mylvaganam, S. E., 122, 124, 154 Mysliwiec, T., 222

N Nachman, J., 86 Nadassy, K., 10, 14, 15, 19, 20, 21, 22, 23, 25, 28, 29, 31, 33, 40 Nagase, T., 111 Nagashima, K., 241 Nagata, K., 241, 250 Nakamoto, T., 219 Nakamura, S., 241 Nakamura, T., 111 Nakashima, K., 85 Nakazato, H., 274 Nam, H.-J., 189, 226 Narayan, D., 3 Narayan, V., 104

315

AUTHOR INDEX

Nash, P., 215 Nathenson, S. G., 122 Nauli, S., 218 Navazza, J., 122, 154 Needleman, S. B., 77 Neel, B. G., 177, 178, 180, 182, 191, 203 Neff, C. L., 287 Nehrbass, U., 102 Nemerow, G. R., 221 Nemethy, G., 147 Neubauer, G., 2, 197 Neuberger, M. S., 148, 155 Neupert, W., 110 Newcomer, M. E., 225, 226 Ng, H. L., 5, 87, 88 Nguyen, J. T., 227, 232, 234, 235, 246, 254, 255 Nicholls, A., 44, 147, 175, 231 Nichols, K. E., 198 Nicholson, L. K., 196 Nielander, G., 109 Nielsen, P. J., 288 Nierhaus, K. H., 115 Nikolov, D. B., 28, 40 Nilges, M., 86 Nilsson, F., 2 Nimnual, A. S., 87 Nioche, P., 224 Nishida, M., 250 Nishikawa, S., 278 Nishimura, S., 278 Nissen, P., 293 Niwa, M., 85 Nix, S. L., 249 Noble, M., 189, 216, 230, 236 Noble, M. E., 216, 217, 226, 245 Noble, N. E. M., 190 Nobuhisa, I., 85 Noel, J. P., 38 Noller, H. F., 293 Nolte, R. T., 166, 169, 181 Nomura, N., 111 Nonoyama, S., 251 Nordstrom, J., 250, 251 Norel, R., 44, 45, 54 Norman, D. G., 76 Northrup, S. H., 46, 62 Notaranglo, L., 198, 199 Novelli, A., 152, 154 Novotny, J., 122

Nussinov, R., 23, 45, 54, 55, 58 Nussinov, Z., 55, 63 Nyanguile, O., 282 Nyberg, K., 147

O Oatley, S. J., 50 Obata, T., 228 O'Brien, R., 194, 220, 226, 243, 244 Ochs, H. D., 251 O'Connell, M. P., 125 Odai, H., 223, 241 Odaka, M., 216, 226 Oegema, J., 96 Oegema, K., 96 Oettgen, H., 198, 199 Ogasahara, K., 147 Ogata, K., 65 Ogawa, T., 85 Ogura, K., 226, 241, 250 Ohlmann, T., 285 Ohno, M., 85 Ohuchi, T., 222 Okada, M., 246 Okamoto, P. M., 251 Oligino, L., 204 Olive, D., 244 Olivier, J. P., 225 Olivier, P., 177, 178 Ollis, D. L., 40 Olsen, H., 279, 280 Olsen, K. W., 77 Olson, A. J., 12, 26 Oltvai, Z. N., 5, 6 O'Neill, L. A., 94 Onofri, F., 220 Oobatake, M., 44, 147 Oohashi, T., 198 Ooi, T., 44, 147 Oppenheim, J. D., 112 Orengo, C. A., 4, 83 Orlova, E., 218, 225 Ortiz, A. R., 226 Oschkinat, H., 235 Ottinger, E. A., 194, 195 Ouellette, B. F., 3 Ouzounis, C. A., 4, 5, 87 Ovchinnikov, L. P., 276 Overduin, M., 164, 165, 178, 196, 226 Owen, D. J., 190, 226, 235

316

AUTHOR INDEX

Owens, C. L., 288 Ozawa, K., 219 Ozawa, R., 3, 104 Ozlu, N., 96

P Pabo, C. O., 27 Pace, H. C., 88 Pace, P. E., 152, 154 Pacofsky, G. J., 202, 203 Padlan, E. A., 43, 49, 121, 122, 123, 154 Pai, E. F., 86, 196 Pai, M. T., 226 Pain, V. M., 283, 285 Palma, P. N., 55, 66 Palmer, I., 245 Pan, Y., 128 Panayotou, G., 164, 180, 181, 202, 219, 223, 249, 251 Panni, S., 225, 249, 250 Pant, N., 164, 165, 167, 169, 175, 178, 185 Pantaloni, D., 224 Pappu, R., 185 Park, H., 251 Park, J., 5 Parker, D., 282 Parker, F., 241 Parmley, S. F., 105 Parry, N. R., 125, 151, 152 Parsons, J. T., 182, 220, 224, 246 Pascal, S., 166, 171, 181 Pastore, A., 226 Patel, H. V., 230 Patel, I. R., 172, 180 Paterson, Y., 122, 124, 154 Patten, P. A., 149, 151 Patthy, L., 76, 78 Pauptit, R., 189, 216, 230, 236 Pause, A., 279, 280, 283, 284, 288, 289 Pavitt, R., 198 Pavletich, N. P., 227, 236, 250, 282 Pawson, T., 3, 86, 161, 162, 163, 164, 166, 171, 177, 178, 180, 181, 182, 184, 188, 189, 196, 199, 203, 215, 220, 223, 224, 225, 246, 251 Payne, G., 180 Pearl, F. M., 83 Pearson, W. R., 77 Peel, D., 172, 180, 249 Peel, M. R., 202, 203

Pei, D., 180, 191 Pekarsky, Y., 88 Pellegrini, M., 5, 87, 88 Pelletier, H., 19, 35 Pellicena, P., 247 Peltier, J. B., 2 Perez-Alvarado, G. C., 282 Perram, J. W., 234 Perutz, M. F., 142 Pestova, T. V., 276, 283, 284, 285, 286, 287, 288 Petel, F., 3, 104 Petosa, C., 286 Petrelli, A., 224, 225, 235, 249 Pfanner, N., 110 Phelps, S., 92 Phillips, S. E., 122 Phillips, T. B., 185, 194 Phizicky, E. M., 9 Picard, C., 244 Piccione, E., 172, 180 Pichler, S., 96 Pico, A., 248 Piga, N., 152, 154 Pilipenko, E. V., 285 Pillutla, R. C., 271 Piras, C., 33, 122 Pisabarro, M. T., 221, 226, 228, 239, 240 Piwnica, W. H., 177 Pliska, V., 148 Pluskey, S., 188, 191 Pochart, P., 3, 104 Poikonen, K., 245 Poksay, K. S., 284 Polakiwicz, R., 279, 280, 283 Politou, A. S., 226 Poljak, R. J., 26, 34, 120, 122, 124, 125, 127, 128, 131, 137, 139, 140, 143, 145, 152, 154 Pontillon, F., 128 Ponting, C. P., 77, 78, 80, 85, 88, 92, 114, 215 Poperechnaya, A. N., 285 Porta, G., 198 Porter, M., 197, 235, 248, 254 Portman, J. J., 47 Posch, S., 58 Posern, G., 241, 252 Post, C. B., 44 Potter, B. V. L., 202 Potts, W. M., 246 Poulin, F., 280

AUTHOR INDEX

Poy, F., 199, 235 Prasad, K. V., 219, 220 Prasad, L., 122, 154 Preiss, T., 284 Press, W. H., 56 Presta, L., 125 Privalov, P. L., 42, 44 Profeta, S., Jr., 54 Profy, A. T., 151, 152 Prongay, A., 182, 203 Ptushkina, M., 282 Pugh, D. J., 213, 214, 220, 227 Puig, O., 109 Pyronnet, S., 282, 284

Q Quail, J. W., 122, 154 Quilliam, L. A., 219, 221, 222, 223, 225, 235, 236, 237, 250 Quinn, D. M., 62 Quinn, M. T., 225 Quiocho, F. A., 279 Qureshi-Emili, A., 104

R Raabe, T., 225 Rabinovich, D., 47 Radhakrishnan, I., 282 Radic, Z., 62 Raff, R. A., 93 Raheul, J., 171, 172, 204 Rain, J. C., 3, 104 Rajewsky, K., 148 Ram, M. K., 185, 194 Ramachandran, C., 181, 182 Ramakrishnan, V., 293 Ramanadham, M., 154 Rameh, L. E., 201, 202 Ramesh, N., 223 Rao, K. V., 150 Rappsilber, J., 2 Rasmussen, B., 122 Rassow, J., 110 Ratnofski, S., 177, 178, 180, 182, 203 Rau, M., 285 Raught, B., 279, 280, 283 Ravichandran, S., 60 Rawlings, D. J., 251 Read, R. J., 26

317

Reddy, V. S., 44 Redl, B., 105 Rees, A. R., 124 Reich, Z., 150 Reinherz, E. L., 36, 122 Reissner, C., 249 Ren, M., 112 Ren, R., 177, 178, 215, 219, 221, 222, 223, 225, 231, 237 Renzoni, D. A., 194, 213, 214, 220, 227 Resh, M. D., 164, 165, 178 Reuveny, E., 37 Reuver, S., 249 Reverdy, C., 3, 104 Reynolds, A. B., 220, 246 Reynolds, J. A., 147 Rhoads, R. E., 276, 280, 284, 288 Rhodes, S., 198 Rhodes, T. H., 225 Ricca, G. A., 225 Rice, D. W., 3, 5, 10, 87, 88 Rich, D. H., 203 Richard, S., 239, 251 Richards, F. M., 11, 17, 227, 228, 232, 241 Richter, I. D., 271 Richter, J. D., 274 Richter-Cook, N. J., 288 Rickles, R. J., 203, 219, 220, 221, 225, 227, 235, 237, 239, 249, 252, 254 Riddle, D. S., 218 Rider, J. E., 219, 221, 222, 223, 225, 235, 236, 237, 250 Rigaut, G., 109 Rigolet, P., 122 Riley, A. M., 202 Rinaldi, T., 93 Rini, J. M., 124, 151, 152, 153 Rios, C. B., 164, 165, 178 Riottot, M. M., 122, 128 Ritchie, D. W., 58, 59 Rivera, M. C., 93 Rivero-Lezcano, O. M., 223 Robbins, J., 172, 180 Robbins, K. C., 223 Roberts, T. M., 177, 178, 180, 182, 189, 203, 226 Robertson, S. R., 164, 165, 178 Robinson, C. V., 115 Robinson, J., 34 Rocque, W., 202, 203

318

AUTHOR INDEX

Rodier, F., 12, 13 Rodrigues, V., 95 Rodriguez, M., 172, 180, 202, 203 Roeder, R. G., 28 Roepstorff, P., 2 Roger, A. J., 93 Rogers, G. W., 288 Roller, P. P., 204 Roll-Mecak, A., 276 Rom, E., 279, 280, 283 Romano, P. R., 225, 249, 250 Romeo, G., 198 Romesberg, F. E., 149, 150 Roncarolo, M. G., 198, 199 Ronco, P. M., 124, 128 Ro Ènnstrand, L., 184 Roost, H. P., 148 Roques, B. P., 213, 241 Rosal, R., 37 Rosario, M., 164, 177 Rose, D. R., 122 Rose, G. D., 4, 52 Rose, J. M., 162, 203 Rose, T., 162, 203 Rosen, M. K., 189, 216, 226, 230, 234, 235, 236, 254 Ross, M. T., 198 Rossi, C., 213, 220, 227 Rossmann, M. G., 77, 82 Rostom, A. A., 115 Rothberg, J., 104 Rothman, J. H., 88 Rout, M. P., 2 Roux, B., 246 Rowlands, D., 125, 151, 152 Rozakis-Adcock, M., 223, 224, 225, 251 Rozen, F., 288 Rozwarski, D. A., 235 Rubin, G. M., 92, 94, 100 Ruczinski, I., 218 Rudd, C. E., 219, 220, 249 Ruf, W., 122 Rugman, P., 194 Rupprecht, M. A., 276 Rush, M. G., 112 Rusnak, D. W., 172, 180, 203 Russek, N., 89 Russell, R. B., 80, 85, 86 Rutz, B., 109 Rydel, T. J., 36

S Sabe, H., 177, 178 Sacchettini, J. C., 122 Sachs, A. B., 270, 274, 284, 287 Sadowski, I., 162 Sadqi, M., 218 Sagerer, G., 58 Sahoo, N. C., 150 Saibil, H. R., 218, 225 Sakai, R., 219 Sakaki, Y., 3, 104 Sakkab, D., 241 Saksela, K., 220, 227, 236, 243, 244, 245 Sali, A., 227, 235, 241, 242, 293 Salim, K., 202 Salinas, P. A., 151 Salunke, D. M., 150 Salwinski, L., 3, 10 Samama, J. P., 35 Sameshima, J. H., 223 Sanakar, A., 202 Sander, C., 11, 86 Sanford, D., 181 Sanson, B., 102 Sansonetti, P. J., 224 Santarsiero, B. D., 149, 150 Santiago, J. V., 218 Santonico, E., 228 Saraste, M., 189, 213, 216, 217, 218, 221, 226, 227, 228, 230, 235, 236, 239, 245, 246 Sargent, M. G., 288 Sasaki, K., 223, 241 Sastry, L., 204 Savage, M. D., 109 Sawasdikosoi, S., 201 Sawyer, T. K., 162, 202, 203 Saxena, K., 199 Saxton, T. M., 164 Sayos, J., 198, 199 Schachter, V., 3, 104 Schaffer, A. A., 77 Schaffer, M. D., 182 Schaffhausen, B., 177, 178, 180, 181, 182, 203, 219, 220 Schagger, H., 109 Schapira, M., 44 Scharenberg, A. M., 251 Schembri-King, J., 194 Scherer, P. E., 80

AUTHOR INDEX

Schevitz, R. W., 77, 82 Schieltz, D. M., 2, 107 Schindler, T., 197 Schlaepfer, D. D., 221 Schlenkrich, M., 52 Schlessinger, J., 161, 166, 169, 171, 172, 173, 180, 181, 188, 215, 216, 222, 225, 226, 229, 241, 250 Schlossmann, J., 110 Schluenzen, F., 293 Schmid, S. L., 251 Schmid, S. R., 288, 289 Schmitz, F., 86 Schmitz, N., 279, 280 Schmitz, R., 225 Schneider-Mergerner, J., 224, 225, 235, 249 Schoepfer, J., 171, 172, 204 Schomburg, D., 58, 59 Schreiber, G., 34, 42, 43, 46, 62, 140, 142 Schreiber, S. L., 113, 189, 216, 218, 219, 221, 223, 224, 225, 226, 227, 228, 230, 234, 235, 236, 237, 239, 241, 253, 254 Schreiber, S. T., 226 Schuetz, E., 92 Schultz, J., 77, 78, 80, 92, 114, 215, 235 Schultz, P. G., 149, 150, 151 Schultz-Gahman, U., 124, 152, 153 Schulz, A., 252 Schumacher, C., 222 Schumacher, T. N., 220, 253 Schuster, V., 198 Schwager, P., 29 Schwarz, F. P., 122, 124, 125, 127, 131, 137, 139, 140, 141, 143, 145, 147, 152, 154 Schwehm, J. M., 9 Schwikowski, B., 3, 5 Scita, G., 225, 226, 250, 251 Scott, J. D., 163, 184, 189 Scott, L. R., 60 Scott, M. P., 247 Searfoss, G. H., 225 Sedkov, Y., 88 Seidel-Dugan, C., 189, 216, 219, 220, 226, 230, 236 Selig, L., 3, 104 Sells, M. A., 222 Selzer, T., 46 Senderowicz, L., 173 Sentenac, A., 105 Sept, D., 60, 61, 63 Sera®ni, P., 198

319

SeÂraphin, B., 109 Seri, M., 198 Serrano, L., 138, 142, 213, 218, 221, 226, 228, 239, 240 Servant, F., 81 Sever, S., 251 Shafman, T., 222 Shakespeare, W. C., 203 Shakhnovich, E. I., 218 Shakked, Z., 47 Shalloway, D., 196, 246 Shampine, L., 172, 180 Shapiro, L., 80 Shariv, I., 56, 59 Sharma, S., 122, 154 Sharp, K. A., 47, 147, 175, 231 Shatkin, A. J., 271, 272, 273, 276 Shatsky, I. N., 276, 284, 285, 287, 288 Shaughnessy, J. D., Jr., 284 Shaw, A. S., 239 Shaw-Smith, C. J., 198 Sheinerman, F. B., 44, 45 Sheldrick, G. M., 278 Sheng, M., 249 Sheraga, H. A., 147 Sheriff, S., 43, 49, 121, 122, 123, 154 Sherlock, G., 100 Shevchenko, A., 109, 162, 201 Shewchuk, L. M., 202, 203 Shi, J., 139 Shi, X. N., 279 Shibata, M., 278 Shibuya, M., 241 Shields, D. C., 94 Shiloh, Y., 222 Shimizu, Y., 224 Shimohigashi, Y., 85 Shin, H., 249 Shin, T. B., 189, 216, 218, 226, 230, 236 Shirai, F., 225, 227, 253, 254 Shiroishi, M., 122, 124 Shoelson, S. E., 165, 166, 167, 169, 171, 172, 173, 175, 177, 178, 180, 181, 182, 185, 188, 189, 191, 194, 195, 196, 203, 226 Shoemaker, B. A., 47 Shoichet, B. K., 26, 36, 52, 54, 59, 144 Shortle, D., 140, 147 Shrinivasan, M., 3 Shuman, S., 270 Shupliakov, O., 252 Sibai, G., 152, 154

320

AUTHOR INDEX

Sicheri, F., 190, 191, 197, 214, 226, 246, 247, 248, 252 Siegal, G., 202 Sieker, L. C., 154 Sielecki, A. R., 26 Sigler, P. B., 26, 27, 38, 47 Sigurskjold, B. W., 125 Siligardi, G., 212, 213, 220, 227 Silltoe, I., 83 Silver, P. A., 114 Silverman, L., 164, 165, 178 Silverton, E. W., 43, 49, 121, 122, 147, 154 Simon, J. A., 225, 227, 228, 234, 235, 236, 237, 241 Simon, S., 3, 104 Simons, K. T., 218 Singer, A. U., 166, 171, 174, 181 Singh, N., 41 Singh, U. C., 54 Siparashvili, Z., 88 Sizeland, A. M., 225 Sjolander, K., 86 Skehel, J. J., 122 Skekter, L. R., 37 Skern, T., 276, 284 Skiba, N. P., 26, 37, 38 Skolnik, E. Y., 86, 222, 225, 241 Sleeman, J., 2 Slepnev, V. I., 224, 225, 251 Smiley, I. E., 77, 82 Smith, C. I., 226 Smith, C. L., 95 Smith, D. L., 196 Smith, F. R., 138, 179 Smith, G. P., 105 Smith, G. R., 60, 65, 67 Smith, H. O., 1 Smith, J. A., 203 Smith, J. C., 169, 171 Smith, T., 100 Smith, T. F., 77 Smithgall, T. E., 196, 221, 227 Smith-Gill, S. J., 43, 49, 121, 122, 123, 124, 137, 138, 141, 143, 144, 145, 147, 152, 154, 179 Smoluchowski, M. V., 62 Smolyar, A., 36, 122 Snedecor, B., 125 Snow, M. E., 182, 203 Snyder, E., 198 Sodroski, J., 34

Sohrmann, M., 96 Soisson, S. M., 87 Sollner, T., 110 Soltoff, S., 177 Sondek, J., 26, 38 Sonenberg, N., 270, 272, 276, 277, 279, 280, 281, 282, 283, 284, 285, 286, 287, 288, 289 Song, O., 2, 103 Songyang, Z., 177, 178, 180, 182, 203, 220, 224, 225, 251 Sonnhammer, E. L., 93, 148 Sordella, R., 3 Soriano, P., 203 Souchon, H., 34, 122, 124, 125, 131, 137, 152, 154 South, V., 225 Sparks, A. B., 219, 221, 222, 223, 225, 235, 236, 237, 250 Spiller, B., 149, 150 Spring, K., 222 Sreerama, N., 213 Srere, P. A., 9 Srinivasan, M., 104 Srivastava, S., 293 Stagljar, I., 2 Stahl, S. J., 245 Stan®eld, R. L., 122, 124, 125, 151, 152, 153 Stapley, B. J., 213, 214, 240 Stefanko, R. S., 122 Steigemann, W., 29 Steitz, J. A., 270 Steitz, T. A., 40, 293 Stepinski, J., 272 Sterling, N. V., 162, 201 Stern, L. J., 214 Stern, M. J., 182 Sternbach, D. D., 172, 180, 202, 203 Sternberg, M. J., 56, 57, 58, 60, 65, 67, 213 Stevens, R. C., 149, 150, 151 Stillman, B., 103 Stites, W. E., 9, 147 Stolz, L. A., 180 Stone, J. C., 162 Stone, M. J., 122 Strauss, A., 171, 172 Stravopodis, D., 92 Strominger, J. L., 214 Strub, M. P., 220, 226, 227, 243, 244, 245 Strynadka, N. C., 26, 36, 59 Stuart, D., 121, 125, 151, 152

AUTHOR INDEX

Stubbs, M. T., 36 Studier, F. W., 293 Stupack, D. G., 221 Superti-Furga, G., 2, 6 Stura, E. A., 122, 151 Sturtevant, J. M., 47 Su, Y. C., 222 Suarez, S., 204 Subramanya, H. S., 291 Sudol, M., 86, 211, 212, 213, 215, 221, 235 Suen, K. L., 227, 241 Sugino, Y., 147 Sumegi, J., 198, 199 Sun, Z. Y., 36 Sundberg, E. J., 120, 141, 147 Sunde, M., 218, 225 Sundquist, W. I., 35 Superti-Furga, G., 190, 191, 197, 214, 226, 246, 248 Suprapto, A., 2 Surdo, P. L., 202 Sutton, G. G., 1 Svitkin, Y. V., 276, 280, 283, 284, 287 Swaf®eld, J. C., 108 Swaminathan, S., 293 Sweet, R. M., 122 Sweet, R. W., 34 Syed, R., 122 Sygusch, J., 56 Sylla, B., 198

T Tabak, H. F., 249 Tahara, S. M., 272 Takano, Y., 147 Takei, K., 252 Takenawa, T., 241 Takeya, T., 246 Takimoto-Kamimura, M., 152 Talbot, J., 60 Talpaz, M., 252 Tam, C., 251 Tan, K., 36 Tan, L., 221 Tanaka, A., 223, 224, 241 Tanaka, S., 241 Tanaka, T., 223, 241, 278 Tanford, C., 147 Tarun, S. Z., Jr., 284

321

Taylor, J. A., 219, 220, 221, 225, 237, 249 Taylor, P., 62 Taylor, R., 17 Taylor, S. J., 196, 246 Teglund, S., 92 te Heesen, S., 2 Teichmann, S. A., 5 Tello, D., 124, 125, 131, 137, 152, 154 Temeriusz, A., 272 Temple, G. F., 3 Tenca, P., 250, 251 Teng, C., 137, 138, 141, 143, 144, 145, 179 Teng, M., 122 Terasawa, H., 226, 241 Terhorst, C., 198, 199 Terry, A. H., 155 Tessari, M., 196 Teukolsky, S. A., 56 Theriault, K., 185, 194 Thiel, G., 225 Thierry-Mieg, N., 3 Thomas, S. M., 219, 220 Thompson, A., 190, 214, 226, 246 Thompson, M. J., 5, 87 Thomson, C. T., 122 Thorn, J. M., 219, 225 Thorn, K. S., 10, 120, 121, 122, 132 Thornton, J. M., 4, 10, 11, 12, 16, 17, 22, 23, 25, 31, 37, 38, 39, 40, 83, 120, 122, 123, 124, 212 Tidor, B., 44, 45 Tillib, S., 88 Timmer, R. T., 280 Tinsley, J., 92 Tkatch, L. S., 193 To, R., 122 Tocilj, A., 293 Tocque, B., 241 Todd, A. E., 83 Tomas-Oliveira, I., 19, 21 Tomlinson, I. M., 148 Tonegawa, S., 148 Tordai, H., 78 Tormo, J., 125, 141, 147, 151, 152 Totrov, M., 26, 36, 44, 59 Totty, N. F., 162, 201, 219, 223, 251 Tovchigrechko, A., 64 Trachsel, H., 279, 280, 283, 288, 289, 290, 291 Tramontano, A., 123

322

AUTHOR INDEX

Trexler, M., 76 Trub, T., 188, 196 Truong, O., 219, 223, 224, 251 Tsai, C. J., 23, 45 Tsai, J., 17, 218 Tsao, J., 227, 241 Tse, A. G., 122 Tsoka, S., 4 Tsuchiya, D., 137, 138, 141, 143, 144, 145, 147, 179 Tsuchiya, S., 241 Tsujihita, Y., 80 Tsujiya, T., 147 Tsumoto, K., 122, 124 Tulinsky, A., 36 Tulip, W. R., 122, 123 Tulloch, P. A., 122 Turbeville, J. M., 93 Turck, C. W., 164, 173, 177, 227, 232, 234, 235, 251, 254, 255 Turner, C. E., 220 Turowski, P., 286 Tushinaki, R. J., 274 Tweardy, D. J., 193 Tyers, M., 2, 6 Tzeng, S. R., 226, 230

U Ueda, H., 278 Uesugi, M., 282 Uesugi, S., 278 Uetz, P., 3, 5, 104 Ulevitch, R. J., 91 Ullrich, A., 216, 226, 241 Ulman, K. V., 225 Ulrich, H. D., 149, 150 Ultsch, M., 30 Ultsch, M. H., 133 Urban, R. G., 214 Urquhart, A. J., 249 Urrutia, M., 141, 147 Uy, M., 87

V Vaccaro, P., 220 Vajda, S., 46, 63, 65 Vajdos, F. F., 35 Vakser, I. A., 56, 58, 59, 64 Valencia, A., 86

Valentine, M. B., 198 Vali, Z., 76 Valius, M., 177 Vallee, R. B., 251 Vallis, Y., 226, 235 Valtorta, F., 220 van Aelst, L., 225 van de Locht, A., 36 van der Geer, P., 251 van der Haar, T., 282 van der Merwe, P. A., 150 van Deursen, J. M., 92 Vandonselaar, M., 122, 154 Van Etten, R. A., 172, 173, 174, 237 van Helden, J., 5 van Heyningen, P., 10, 17, 31, 38, 39, 40 van Schaik, S., 198, 199 van Wijk, K. J., 2 Varadi, A., 76 Varghese, J. N., 122 Varmus, H. E., 172, 174, 197 Varshavsky, A., 105 Vasilescu, S., 282 Vasmatzis, G., 63 Vaudin, M., 198 Vaughan, M., Jr., 274 Venter, J. C., 1 Verdine, G. L., 282 Verroust, P. J., 124, 128 Verschoor, A., 293 Vetterlein, D., 125 Vetterling, W. T., 56 Via, A., 225, 228 Vidal, M., 3, 105, 226, 230 Viglino, P., 47 Viguera, A. R., 213, 221, 226, 239 Vihinen, M., 226 Vijayadamodar, G., 104 Vincent, J. P., 33, 35, 102 Vinkemeier, U., 191, 192 Violette, S. M., 203 Vitkup, D., 11 Voight, J. H., 204 Volinia, S., 228 von Hippel, P. H., 45, 47 von Jagow, G., 109 Vonrhein, C., 293 Voss, E. W., Jr., 152, 154, 155 Voss, J., 241 Vuister, G. W., 196

AUTHOR INDEX

W Wade, J., 223, 224, 225 Wade, R. C., 46, 60, 62 Wagner, C., 172, 180 Wagner, G., 36, 249, 276, 281, 282 Wahl, M. I., 251 Wahle, E., 274 Waksman, G., 162, 163, 164, 165, 167, 169, 173, 174, 175, 176, 178, 179, 180, 182, 185, 187, 188, 193, 194, 195, 202, 203 Walhout, A. J., 3 Wall, R., 274 Wallquist, A., 144 Walsh, C. T., 180, 191 Walter, G., 148 Walter, J., 31, 36 Walther, Z., 249 Wang, B., 222 Wang, D., 92 Wang, H., 52, 54 Wang, J., 122, 191 Wang, J. H., 36 Wang, J.-X., 173 Wang, J. Y., 221 Wang, L., 144 Wang, L. H., 149, 151 Wang, L.-Y., 173 Wang, N., 198, 199 Wang, S., 204 Wang, S. M., 279 Wang, Y., 222 Warner, J. R., 293 Warren, T. C., 194 Washburn, M. P., 2 Water®eld, M. D., 164, 181, 194, 202, 213, 216, 218, 220, 224, 226, 227, 249 Waterman, M. S., 77 Watson, M., 2 Watters, D., 222 Way, M., 191, 248 Waye, M. M., 139 Waygood, E. B., 122, 154 Weaver, P. L., 288 Weber, C., 204 Weber, P. C., 182, 203 Webster, D. M., 124, 198 Webster, R. G., 122 Wedemayer, G. J., 149, 151 Weijland, A., 190, 197, 214, 226, 246, 248

323

Weinberg, R. A., 225 Weiner, P., 54 Weiner, S. J., 54 Weinreich, M., 103 Weis, K., 286 Weiss, A., 194 Welch, M., 35 Wel¯e, H., 213 Wells, C. A., 64 Wells, J. A., 45, 133, 137, 155, 179 Welsh, M., 220 Weng, G., 37 Weng, S., 100 Weng, Z., 46, 52, 55, 61, 63, 65, 219, 220, 221, 225, 237, 239, 249 Wenk-Siefert, I., 93 Wermuth, P., 88 Werner, M. H., 31 Wernisch, L., 5, 65 Westermann, P., 213 Wetlaufer, D. B., 4 Wharton, S. A., 122 Whisstock, J. C., 218 White, P., 102 Wiedmann, M., 110 Wierenga, R. K., 189, 190, 197, 214, 216, 217, 226, 230, 236, 245, 246, 248 Wigge, P., 226, 235 Wigler, M. H., 225 Wigley, D. B., 291 Wiley, D. C., 214 Wilkinson, A. J., 45, 139, 179 Wilkinson, J., 198 Willard, D., 172, 180 Willcox, B. E., 150 Williams, J. C., 190, 214, 226, 246, 248 Williams, J. G., 162, 197, 201 Williams, L. T., 164, 177 Williamson, M. P., 211, 212, 215 Willson, R. C., 42, 43 Wilm, M., 106, 109 Wilmanns, M., 218, 226, 227, 228, 230, 239, 245 Wilson, C. A., 83 Wilson, I. A., 122, 124, 125, 151, 152, 153 Wilson, K. S., 122, 154 Wilson, P., 58, 59 Wimberly, B. T., 293 Wimmer, C., 102 Windemuth, A., 44

324

AUTHOR INDEX

Windsor, W. T., 182, 203 Wing®eld, P. T., 245 Winter, G., 45, 139, 148, 179 Wise, C., 198 Witte, O. N., 251 Wittekind, M., 227, 241 Wittrup, K. D., 148, 155 Wodak, S. J., 4, 5, 10, 14, 15, 19, 20, 21, 22, 23, 25, 28, 29, 31, 33, 40, 50, 51, 52, 53, 65 Wojcik, J., 3, 104 Wojnar, P., 105 Wolf, G., 173, 181 Wolfson, H. J., 54, 55 Wolfson, H. L., 54, 58 Wollmer, A., 181 Wolters, D., 2 Wolting, C., 3 Wolynes, P. G., 47, 48 Wonacott, A. J., 77, 82 Wong, J., 185, 187, 188, 193 Wong, W. T., 225, 226, 249, 250 Woods, D. F., 249 Woody, R. W., 213 Woolford, J. L., 280, 283 Worthylake, D. K., 35 Wray, P., 198 Wriggers, W., 64 Wright, J. G., 47 Wright, P. E., 155, 282 Wroblowski, B., 181 Wu, C., 198, 199 Wu, H., 249 Wu, T. T., 121 Wu, X., 227, 235, 241, 242 Wu, Y., 223 Wunsch, C. D., 77 Wyatt, R., 34 Wyer, J. R., 150

X Xenarios, I., 3, 10 Xiao, L., 45 Xu, D., 23, 45 Xu, J., 147 Xu, Q., 196 Xu, R., 196 Xu, W., 190, 214, 226, 246 Xuong, N. G., 40

Y Yaffe, M. B., 199, 228, 251 Yajnik, V., 225, 241 Yamabhai, M., 225 Yamagata, Y., 147 Yamaguchi, H., 190 Yamanaka, S., 284 Yamashina, I., 85 Yamazaki, T., 166, 171, 181 Yan, X., 173 Yandell, M., 1 Yang, A. S., 47 Yang, C. S., 37 Yang, D., 199, 204 Yang, F. C., 249 Yang, M., 104 Yang, P. L., 149 Yang, S. S., 251 Yang, W., 181 Yang, X., 95 Yao, Z. J., 204 Yates, J. R., 2 Yazaki, Y., 219, 223, 241 Ye, Z. S., 220, 222, 223 Yeates, T. O., 5, 87, 88 Yen, T., 222 Yi, T., 177, 178, 181 Yin, L., 198 Yoder-Hill, J., 288 Yokote, K., 184 Yonath, A., 293 Yoo, S., 35 Yoshida, M., 3, 104 Yoshida-Kubomura, N., 111 Yoshizawa, S., 173 Young, M. A., 246 Young, N. M., 122 Ysern, X., 26, 36, 122, 124, 127, 131, 137, 138, 139, 140, 141, 143, 144, 145, 179 Yu, A. T., 283, 284 Yu, H., 189, 216, 218, 221, 225, 226, 227, 228, 230, 234, 235, 236, 237, 241 Yue, Z., 271 Yuhasz, S. C., 128 Yusupov, M. M., 293 Yusupova, G. Z., 293 Yutani, K., 147

AUTHOR INDEX

Z Zachmann, C. D., 52 Zarivach, R., 293 Zeelen, J. P., 226 Zhang, C., 63 Zhang, J., 77 Zhang, N., 222 Zhang, R., 182, 203, 235 Zhang, T. H., 288 Zhang, W., 227 Zhang, X., 198, 199 Zhang, X.-J., 147 Zhang, Z., 77 Zhao, Y., 2, 191, 192 Zhao, Z. S., 221, 222, 223, 251 Zheng, J., 196, 220, 222, 223, 224, 226, 227, 235, 241, 242, 243

325

Zheng, Y., 219 Zhou, M., 173 Zhou, X. M., 225, 235 Zhukovskaya, N. V., 162, 201 Zimmerman, F., 26, 36, 59 Zinkernagel, R. M., 148 Zipperlen, P., 96 Zoller, M. J., 185, 194, 219, 221, 225, 235, 237, 249 Zollo, M., 198 Zucconi, A., 228 Zuckermann, R. N., 227, 232, 234, 235, 254, 255 Zuo, L., 198 Zurdo, J., 218, 225 Zydowsky, L. D., 216, 218, 226

This Page Intentionally Left Blank

SUBJECT INDEX

A ABl1 protoonocogene-SH3 interactions, 237, 239±240 Af®nity ceiling, in antigen recognition, 155 low, protein-protein classi®cation, 35±36 maturation, in antigen recognition, 148±151 puri®cation, protein-protein complexes, 108±109 selection, protein-protein complexes, 110±112 Agr ûB5, in SH2 target recognition, 174 Antibody -antigen complexes, 120±121 -antigen interfaces binding energetics, dissection FvD1.3-Fve5.2 complex, 138±141 FvD1.3-HEL complex, 141±144 mutations, accommodation, 144±148 structures, 121±125 thermodynamic mapping, 132±134, 137±138 characterization, 120 cross-reactivity, 125±132 molecular mimicry, 125±132 Antigen -antibody complexes, 120±121 -antibody interfaces binding energetics, dissection FvD1.3-Fve5.2 complex, 138±141 FvD1.3-HEL complex, 141±144 mutations, accommodation, 144±148 structures, 121±125 thermodynamic mapping, 132±134, 137±138 recognition, role of protein plasticity af®nity ceiling, beyond, 155

af®nity maturation, 148±151 induced ®t, 151±155 Atomic packing macromolecular surfaces, 16±17 protein-DNA interfaces, 19 protein-protein interfaces, 17±19 shape vs., 21 ATP binding, mutations, 289±291

B Barnase-barstar interface elestrostatic association, 46 energetics, 42±44 size, 34 Binding energetics in antigen-antibody interface FvD1.3-FvE5.2 complex, 138±141 FvD1.3-HEL complex, 141±144 BLAST, 82 Brownian dynamics models, 60±63

C Cbl molecule, 201 CDD. see Conserved Domain Database CDRs. see Complementarity determining regions CLIP domain, function, 95 Complementarity determining regions function, 120 loops concerted movements, 152±154 function, 122±123, 125 Conserved Domain Database, 82 Crk protoonocogene-SH3 interactions, 241±242 Crystal contacts, 12±14 Cysteine residue ûC3, 203

327

328

SUBJECT INDEX

D Ded1 gene, 288±289 DNA backbone, 38±40 -protein complexes classi®cation, 36±40 interface area, 14±15 -protein interfaces chemical composition, 24 conformational changes, 29±31 dry, 25±29 energetics, 46±47 packing at, 19 polar interactions, 24±25 wet, 25±29 Docking procedures assessing, 59 description, 50±51 by FFT, 56±59 geometric hashing, 54, 56 shape complementarity, 52, 54 simpli®ed protein models, 51±52 Domains (protein) classi®cation methods, 81±85 context zooming in, 86 zooming out, 86±89 functions, 85±86 genome-wide analysis comparative, 93±95 function evolution, 93±95 in gene prediction, 89, 91 orthology, 91±92 paralogy, 91±92 historical approaches, 77±78 identi®cation, 75±77 interacting, characterization, 114±115 modern approaches, 79±80 Double stranded RNA binding motif, 93 Drug targets Src homology 2 domains as Grb2 binding inhibitors, 203±204 Src binding inhibitors, 203 testing dif®culties, 202 DSHP. see Signaling lymphocyte activation molecule-associated protein DSRM. see Double stranded RNA binding motif

E eIF4E characterization, 276 protein-ligand interactions, 278 structure, 276±277 eIF4 f in ADP, 291 alternative structure, 292 in ATP, 289±291 characterization, 288±289 family, 289 HEAT repeats, 283±287 50 -UTR recognition, 288 regulation by molecular mimicry, 279±282 translation initiation and, 288 Ermak-McCammon algorithm, 60

F Fab39-A11, 149±150 FabAZ-28, 149 Fab48G7, 148±149, 151 Fast Fourier Transform, 56±59 FFT. see Fast Fourier Transform FHA. see Fork-headed domains Fork-headed domains identi®cation, 87 Fyn protein, 251

G GEFs. see Guanine nucleotide exchange factors Geometric complementarity macromolecular recognition protein-protein interfaces atomic packing, 17±19 Voronoi volumes, 17±19 shape vs. atomic packing, 21 shape vs. packing, 21 macromolecular surfaces, evaluation, 16±17 G proteins, components, 36±37 Grb2 protein binding inhibitors, 203±204 characterization, 189 SH3 interactions, 240±241 Sos binding sites, 230

SUBJECT INDEX

Sos speci®city, 225 structure, 195±196 Guanine nucleotide exchange factors, 87

H HEAT repeat proteins, 283±287 Hidden Markov models, 82 Hirudin, 36, 38 HMMs. see Hidden Markov models Human Immunode®ciency viruse-1, 243 HyHe15, energetics, 42±44

I Immunoreceptor tyrosine activation motifs cellular sequences, 195 dpITAM, characterization, 185±188 dpITAM, in Sky Zap tandem SH2 domains, 193±194 in SH2 target recognition, 176 Internal ribosomal entry site -eIF4G binding, 287 function, 273 IRES. see Internal ribosomal entry site ITAMs. see Immunoreceptor tyrosine activation motifs Itk protein, 248

L Lck protein, 189 Lim's model, 232±235

M Macromolecular recognition atomic packing, evaluation, 16±17 DNA-protein associations, conformational changes, 31 DNA-protein complexes chemical composition, 24 classi®cation, 36±40 DNA-protein interfaces atomic packing, 19, 21 dry, 25±29 energetics, 46±47 interface areas, 14±15 polar interactions, 24±25 shape complementarity, 21

329

speci®city, statistical mechanics of, 47±50 wet, 25±29 geometric complementarity atomic packing, shape vs., 21 evaluation, 16±17 interface areas crystal contacts, 12±14 description, 11 evaluation, 11±12 protein-protein associations conformational changes, 29±31 simulations, 59±64 protein-protein complexes classi®cation low-af®nity, 35±36 small, 35±36 thermodynamic parameters, 40±41 protein-protein interactions docking procedures assessing, 59 description, 50±51 by FFT, 56±59 geometric hashing, 54, 56 shape complementarity, 52, 54 simpli®ed protein models, 51±52 hydrophobic/polar, 42±44 polar/hydrophobic, 42±44 protein-protein interfaces association kinetics, 45±46 atomic packing, 17±19 chemical composition, 21±23 classi®cation, 32±35 dry, 25±29 electrostatics, 44±46 interface areas, 14±15 polar interactions, 21±23 shape complementarity, 21 speci®city, statistical mechanics of, 47±50 Voronoi volumes, 17±19 wet, 25±29 Modules, term usage, 75±76 mRNA Shine-Dalgarno sequence, 270 small ribosomal subunit binding description, 269±271 50 -UTR preparation, 288±293 structural features 30 poly(A) tail, 274 50 ,7-methyl-G cap, 271±272

330

SUBJECT INDEX

mRNA (continued) translation sites, 273 translation Cap-dependent, initiation mechanisms eIF4 f regulation, 279±282 models, 274, 276 50 ,7 methyl-G cap structure, 276±279 characterization, 270±271 Mutations antigen-antibody interfaces, 144±148 double cycles, in binding energetic dissection FvD1.3-FvE5.2 complex, 138±141 FvD1.3-HEL complex, 141±144

N Nef-SH3 interactions, 243±245

O Ornithodorin-thrombin interface, 36 Orthology, 91±92

P p85, 249 Pancreatic trypsin inhibitor system conformational changes, 29±31 size, 36±38 Paralogy, 91±92 PDGFR peptide, 178±179 PDZ domains, identi®cation, 86±87 Pex5p protein, 249 PFAM, 82±83 PH. see Pleckstrin homology Phospholipid SH2 interactions, 201±202 Phosphotyrosine-binding domain, identi®cation, 86±87 Phosphotyrosine (pTry) recognition binding pocket, 202 discovery, 163 library selection, 177±178 in SH2 domain, 165±166 SH2 speci®city C-PLCg domain, 181±181 Grb2 domain, 182 implications, 183±184 N-p85 SH2 domain, 180±181 N-SHP-2 domain, 181±181 Src Sh2 domains, 178±180

switching speci®city, 182±183 P13K, 249 Plasticity protein, in antigen recognition af®nity ceiling, beyond, 155 induced ®t, 151±155 Pleckstrin homology, 86±87 Polyproline type II helices characterization, 212 SH3 interactions binding site, 230±232 Lim's model, 232±235 structure, 213±214 PPII. see Polyproline type II helices Proline characterization, 212±213 SH3 interaction ligand-binding sites, 230±232 ligand speci®city, 225±228 Lim's model, 232±235 speci®city, 225±228 structures of complex, 228±229 structure, 213±214 PROSITE, 82 Proteins. see also speci®c proteins -DNA complexes classi®cation, 36±38 interface area, 14±15 -DNA interfaces atomic packing, 19 chemical composition, 24 conformational changes, 29±31 dry, 25±29 polar interactions, 24±25 speci®city, statistical mechanics of, 47±50 wet, 25±29 domain families classi®cation methods, 81±85 context zooming in, 86 zooming out, 86±89 functions, 85±86 genome-wide analysis comparative, 93±95 function evolution, 93±95 in gene prediction, 89, 91 orthology, 91±92 paralogy, 91±92 historical approaches, 77±78 identi®cation, 75±77

331

SUBJECT INDEX

modern approaches, 79±80 modules, description, 4 plasticity, in antigen recognition af®nity ceiling, beyond, 155 af®nity maturation, 148±151 induced ®t, 151±155 -protein associations, simulations, 59±64 -protein complexes classi®cation large interfaces, 36±38 low-af®nity, 35±36 -protein interactions complexity, 100 docking procedures assessing, 59 description, 50±51 by FFT, 56±59 geometric hashing, 54, 56 shape complementarity, 52, 54 simpli®ed protein models, 51±52 identi®cation, biochemical af®nity puri®cation, 108±109 af®nity selection, 110±112 co-immunoprecipition, 107, 114 cross-linking, 109±110 far-western, 112 native gel analysis, 109 protein arrays, 112±113 puri®cation, 105±107 identi®cation, biological phage display, 105 two-hybrid analyses, 103±105 identi®cation, genetic dosage effects, 102±103 suppressor analyses, 100±102 synthetic phenotypes, 100±102 subunits/domains, characterization, 114±115 thermodynamic parameters, 40±41 validation of, 113±114 -protein interfaces association kinetics, 45±46 atomic packing, 17±19 broad categories, de®nition, 138 chemical composition, 21±23 classi®cation, standard size, 32±35 conformational changes, 29±31 dry, 25±29 electrostatics, 44±45, 44±46 hydrophobic/polar, 42±44

macromolecular recognition, Voronoi volumes, 17±19 polar/hydrophobic, 42±44 polar interactions, 23±24 speci®city, statistical mechanics of, 47±50 wet, 25±29 synthesis, characterization, 269 Protein-tyrosine kinases, 248 PTB. see Phosphotyrosine-binding domain p53 tumor suppressor, 250 pYAEI peptide, 178±179 pYEEA peptide, 178±180 pYEEI peptide, 178±179

R Random Energy Model, 48 Ras proteins, 189, 251

S Sam68 protein, 251 SANT domains, 93 SAP, 198±199 SH2D1A. see Signaling lymphocyte activation molecule-associated protein Shine-Dalgarno sequence, 270 Signaling lymphocyte activation moleculeassociated protein, 198±199 Signal transducers and activators of transcription, 191±192 Simple Modular Architecture Research Tool, 82±85 SLAM. see Signaling lymphocyte activation molecule-associated protein SMART. see Simple Modular Architecture Research Tool Son of Sevenless (Sos) Grb2 interactions binding, sequences, 215 -Grb2 protein, binding sites, 230 -Grb2 protein, speci®city, 225 in Ras, binding sites, 251 Speci®city determining region, 163±164 Src family kinases characterization, 190 as intramolecular switch, 245±247 SH2 interactions, 197 SH3 interactions, 236±237 structure, 190±191

332

SUBJECT INDEX

Src homology 2 domains characterization, 161 as drug target Grb2 binding inhibitors, 203±204 Src binding inhibitors, 203 testing dif®culties, 202 function, 161±162 identi®cation, 86±87, 162 in¯uence on protein-protein studies, 163±164 with other protein modules Cbl SH2 domain, 201 full-length Src family kinases solution studies, 197 structures, 189±191 phospholipid interactions, 201±202 SAP SH2 domain, 198±199 SHP-2 phosphatase, structures, 191±192 SH2-SH3 constructs, structures, 188±189 SH2-SH3 domain constructs, solution studies, 195±196 STAT transcriptional activators, structures, 191±192 Syk and Zap-70 tandem domains, solution studies, 193±195 tandem SH2 domains, structures, 185±188 pTry recognition determents, 172±176 speci®city C-PLCg domain, 181±181 Grb2 domain, 182 implications, 183±184 library selection, 177±178 N-p85 SH2 domain, 180±181 N-SHP-2 domain, 181±181 Src SH2 domains, 178±180 switching speci®city, 182±183 structure, 165±166 structure architecture, 164±165 fold, 164±165 pTry recognition, 165±166 speci®city determining interactions open groove mode of binding, 169, 171±172 2-pronged plug-2-hold socket, 167, 169

b-turn mode of binding, 172 Src homology 3 domains Eps8, 225, 249±250 ligand interactions Abl-SH3, 237 characterization, 235±236 Crkl-SH3, 241±242 p85-SH3, 236±237 Src-SH3, 236±237 ligand speci®city, 225±229 macromolecular assemblies, 245±249 mediated interactions, inhibition, 252±255 -Nef interactions, 243±245 nonproline interactions, 249±250 occurrence, 215±2165 P13K, 218 PPII interaction ligand binding sites, 230±232 Lim's model, 232±235 proline interaction discovery, 211 ligand speci®city, 225±228 Lim's model, 232±235 structure, 216±218, 225 substrate binding, regulation, 250±252 STATs. see Signal transducers and activators of transcription Syk-Zap 70 tandem domains conformational ¯exibility, 193±194 SH2 speci®city, 194±195 structure, 185±188

T Thermodynamic mapping, 132±134, 137±138 Thrombin-ornithodorin interface, classi®cation, 36 TIR domain, in gene prediction, 91 Trypsin-like serine protease, function, 94±95 Trypsin-PTI system conformational changes, 29±31 size, 36±38 Tyr-223 protein, 251 Tyr-527 protein, 251 Tyrosine phosphorylation, in SH2 target recognition, 172±174

333

SUBJECT INDEX

V Voronoi volumes, protein-protein interfaces, 17±19 VP39 protein, structure, 278±279

X X-linked Lymphoproliferative disease, 198 XP motifs

binding pockets, 230 location, 228

Z Zap 70-Syk tandem domains conformational ¯exibility, 193±194 SH2 speci®city, 194±195 structure, 185±188

90051

9 780120 342617

ISBN 0-12-034261-8

Protein Modules and Protein-Protein Interactions, Volume 61 (Advances in Protein Chemistry)

Advances in Protein Chemistry, Volume 59: Protein Folding in the Cell (Advances in Protein Chemistry)

Protein Protein Interactions

Computational Protein-Protein Interactions

Advances in Protein Chemistry, Volume 34

Advances in Protein Chemistry, Volume 36

Advances in Protein Chemistry, Volume 39

Advances in Protein Chemistry Volume 41

Advances in Protein Chemistry, Volume 43

Advances in Protein Chemistry, Volume 32

Advances in Protein Chemistry Volume 41

Advances in Protein Chemistry, Volume 38

Advances in Protein Chemistry, Volume 57: Prion Proteins (Advances in Protein Chemistry)

Protein-protein interactions in plant biology

Drug Discovery and Design (Advances in Protein Chemistry, Vol 56) (Advances in Protein Chemistry)

Proteomics and Protein-Protein Interactions: Biology, Chemistry, Bioinformatics, and Drug Design (Protein Reviews)

Proteomics and Protein-Protein Interactions: Biology, Chemistry, Bioinformatics, and Drug Design (Protein Reviews)

Advances in Protein Chemistry, Volume 69: DNA Repair and Replication

Advances in Protein Chemistry, Volume 45: Lipoproteins, Apolipoproteins, and Lipases

Protein-Solvent Interactions

Membrane Protein-Cytoskeleton Interactions

Protein-Solvent Interactions

Advances in Protein Chemistry: v. 16

Advances in Protein Chemistry: v. 17

Advances in Protein Chemistry: v. 7

Protein Structure and Diseases (Advances in Protein Chemistry and Structural Biology, Volume 83)

Advances in Protein Chemistry: Copper Containing Molecules (Advances in Protein Chemistry)

Advances in Protein Chemistry: v. 10

Advances in Protein Chemistry: v. 17

Advances in Protein Chemistry: v. 16

Advances in Protein Chemistry: v. 20

Protein Modules and Protein-Protein Interactions, Volume 61 (Advances in Protein Chemistry)

Advances in Protein Chemistry, Volume 59: Protein Folding in the Cell (Advances in Protein Chemistry)

Protein Protein Interactions

Computational Protein-Protein Interactions

Advances in Protein Chemistry, Volume 34

Advances in Protein Chemistry, Volume 36

Advances in Protein Chemistry, Volume 39

Advances in Protein Chemistry Volume 41

Advances in Protein Chemistry, Volume 43

Advances in Protein Chemistry, Volume 32

Advances in Protein Chemistry Volume 41

Advances in Protein Chemistry, Volume 38

Advances in Protein Chemistry, Volume 57: Prion Proteins (Advances in Protein Chemistry)

Protein-protein interactions in plant biology

Drug Discovery and Design (Advances in Protein Chemistry, Vol 56) (Advances in Protein Chemistry)

Proteomics and Protein-Protein Interactions: Biology, Chemistry, Bioinformatics, and Drug Design (Protein Reviews)

Proteomics and Protein-Protein Interactions: Biology, Chemistry, Bioinformatics, and Drug Design (Protein Reviews)

Advances in Protein Chemistry, Volume 69: DNA Repair and Replication

Advances in Protein Chemistry, Volume 45: Lipoproteins, Apolipoproteins, and Lipases

Protein-Solvent Interactions

Membrane Protein-Cytoskeleton Interactions

Protein-Solvent Interactions

Advances in Protein Chemistry: v. 16

Advances in Protein Chemistry: v. 17

Advances in Protein Chemistry: v. 7

Protein Structure and Diseases (Advances in Protein Chemistry and Structural Biology, Volume 83)

Advances in Protein Chemistry: Copper Containing Molecules (Advances in Protein Chemistry)

Advances in Protein Chemistry: v. 10

Advances in Protein Chemistry: v. 17

Advances in Protein Chemistry: v. 16

Advances in Protein Chemistry: v. 20

Recommend Documents