Methods in enzymology 483 - Cryo-EM, Part C: Analyses, Interpretation, and Case Studies - PDF Free Download (2024)

METHODS IN ENZYMOLOGY Editors-in-Chief

JOHN N. ABELSON AND MELVIN I. SIMON Division of Biology California Institute of Technology Pasadena, California Founding Editors

SIDNEY P. COLOWICK AND NATHAN O. KAPLAN

Academic Press is an imprint of Elsevier 525 B Street, Suite 1900, San Diego, CA 92101-4495, USA 30 Corporate Drive, Suite 400, Burlington, MA 01803, USA 32 Jamestown Road, London NW1 7BY, UK First edition 2010 Copyright # 2010, Elsevier Inc. All Rights Reserved. No part of this publication may be reproduced, stored in a retrieval system or transmitted in any form or by any means electronic, mechanical, photocopying, recording or otherwise without the prior written permission of the publisher Permissions may be sought directly from Elsevier’s Science & Technology Rights Department in Oxford, UK: phone (+44) (0) 1865 843830; fax (+44) (0) 1865 853333; email: permissions@ elsevier.com. Alternatively you can submit your request online by visiting the Elsevier web site at http://elsevier.com/locate/permissions, and selecting Obtaining permission to use Elsevier material Notice No responsibility is assumed by the publisher for any injury and/or damage to persons or property as a matter of products liability, negligence or otherwise, or from any use or operation of any methods, products, instructions or ideas contained in the material herein. Because of rapid advances in the medical sciences, in particular, independent verification of diagnoses and drug dosages should be made For information on all Academic Press publications visit our website at elsevierdirect.com ISBN: 978-0-12-384993-9 ISSN: 0076-6879 Printed and bound in United States of America 10 11 12 10 9 8 7 6 5 4 3 2 1

CONTRIBUTORS

Francisco Asturias Department of Cell Biology, The Scripps Research Institute, La Jolla, California, USA Mariah R. Baker National Center for Macromolecular Imaging, Verna and Marrs McLean Department of Biochemistry and Molecular Biology, Baylor College of Medicine, Houston, Texas, USA Matthew L. Baker National Center for Macromolecular Imaging, Verna and Marrs McLean Department of Biochemistry and Molecular Biology, Baylor College of Medicine, Houston, Texas, USA Martin Beck European Molecular Biology Laboratory, Heidelberg, Germany Edward J. Brignole Department of Cell Biology, The Scripps Research Institute, La Jolla, California, USA Bridget Carragher National Resource for Automated Molecular Microscopy, Department of Cell Biology, The Scripps Research Institute, La Jolla, California, USA Anchi Cheng National Resource for Automated Molecular Microscopy, Department of Cell Biology, The Scripps Research Institute, La Jolla, California, USA Frank DiMaio Department of Biochemistry, University of Washington, Seattle, Washington, USA Kenneth H. Downing Life Sciences Division, Donner Laboratory, Lawrence Berkeley National Laboratory, Berkeley, California, USA Nadav Elad Department of Life Sciences, and The National Institute for Biotechnology in the Negev, Ben Gurion University of the Negev, Beer-Sheva, Israel

xi

xii

Contributors

Andreas Engel Department of Pharmacology, Case Western Reserve University, Cleveland, Ohio, USA ¨rster Friedrich Fo Max-Planck Institute of Biochemistry, Department of Structural Biology, Martinsried, Germany Lauren Fisher The Scripps Research Institute, La Jolla, California, USA Yoshinori Fujiyoshi Department of Biophysics, Kyoto University, Oiwake, Kitashirakawa, Sakyo-ku, Kyoto, Japan Jan Giesebrecht Institut fu¨r medizinische Physik und Biophysik, Charite´, Universita¨tsmedizin Berlin, Berlin, Germany John Paul Glaves Department of Biochemistry, School of Molecular and Systems Medicine, and National Institute for Nanotechnology, University of Alberta, Edmonton, Alberta, Canada Bong-Gyoon Han Life Sciences Division, Lawrence Berkeley National Laboratory, University of California, Berkeley, California, USA Dorit Hanein Sanford-Burnham Medical Research Institute, La Jolla, California, USA Amber Herold National Resource for Automated Molecular Microscopy, Department of Cell Biology, The Scripps Research Institute, La Jolla, California, USA Richard K. Hite Department of Cell Biology, Harvard Medical School, Boston, Massachusetts, USA Eric Hou National Resource for Automated Molecular Microscopy, Department of Cell Biology, The Scripps Research Institute, La Jolla, California, USA Corey F. Hryc National Center for Macromolecular Imaging, Verna and Marrs McLean Department of Biochemistry and Molecular Biology, Baylor College of Medicine, Houston, Texas, USA

Contributors

xiii

Christopher Irving National Resource for Automated Molecular Microscopy, Department of Cell Biology, The Scripps Research Institute, La Jolla, California, USA Erica L. Jacovetty National Resource for Automated Molecular Microscopy, Department of Cell Biology, The Scripps Research Institute, La Jolla, California, USA Pick-Wei Lau National Resource for Automated Molecular Microscopy, Department of Cell Biology, The Scripps Research Institute, La Jolla, California, USA Catherine L. Lawson Department of Chemistry and Chemical Biology and Research Collaboratory for Structural Bioinformatics, Rutgers, The State University of New Jersey, USA Jun Liu Department of Pathology and Laboratory Medicine, University of Texas Medical School at Houston, Houston, Texas, USA Justus Loerke Institut fu¨r medizinische Physik und Biophysik, Charite´, Universita¨tsmedizin Berlin, Berlin, Germany Dmitry Lyumkis National Resource for Automated Molecular Microscopy, Department of Cell Biology, The Scripps Research Institute, La Jolla, California, USA Asaf Mader Department of Life Sciences, and The National Institute for Biotechnology in the Negev, Ben Gurion University of the Negev, Beer-Sheva, Israel Ohad Medalia Department of Life Sciences, and The National Institute for Biotechnology in the Negev, Ben Gurion University of the Negev, Beer-Sheva, Israel Arne Moeller National Resource for Automated Molecular Microscopy, Department of Cell Biology, The Scripps Research Institute, La Jolla, California, USA Anke M. Mulder National Resource for Automated Molecular Microscopy, Department of Cell Biology, The Scripps Research Institute, La Jolla, California, USA Eva Nogales Life Sciences Division, Donner Laboratory, and Life Sciences Division, Lawrence Berkeley National Laboratory, and Department of Molecular and Cell Biology, Howard Hughes Medical Institute, UC Berkeley, Berkeley, California, USA

xiv

Contributors

Clinton S. Potter National Resource for Automated Molecular Microscopy, Department of Cell Biology, The Scripps Research Institute, La Jolla, California, USA James Pulokas National Resource for Automated Molecular Microscopy, Department of Cell Biology, The Scripps Research Institute, La Jolla, California, USA Joel D. Quispe National Resource for Automated Molecular Microscopy, Department of Cell Biology, The Scripps Research Institute, La Jolla, California, USA Andreas D. Schenk Department of Cell Biology, Harvard Medical School, Boston, Massachusetts, USA Christian M. T. Spahn Institut fu¨r medizinische Physik und Biophysik, Charite´, Universita¨tsmedizin Berlin, Berlin, Germany Elizabeth Villa Max-Planck Institute of Biochemistry, Department of Structural Biology, Martinsried, Germany Niels Volkmann Sanford-Burnham Medical Research Institute, La Jolla, California, USA Neil R. Voss National Resource for Automated Molecular Microscopy, Department of Cell Biology, The Scripps Research Institute, La Jolla, California, USA Thomas Walz Department of Cell Biology, and Howard Hughes Medical Institute, Harvard Medical School, Boston, Massachusetts, USA Andrew Ward The Scripps Research Institute, La Jolla, California, USA Hanspeter Winkler Institute of Molecular Biophysics, Florida State University, Tallahassee, Florida, USA Elizabeth R. Wright Department of Pediatrics, Emory University School of Medicine, Atlanta, Georgia, USA Howard S. Young Department of Biochemistry, School of Molecular and Systems Medicine, and National Institute for Nanotechnology, University of Alberta, Edmonton, Alberta, Canada

PREFACE

In this, the fifty-fourth year of Methods in Enzymology, we celebrate the discovery and initial characterization of thousands of individual enzymes, the sequencing of hundreds of whole genomes, and the structure determination of tens of thousands of proteins. In this context, the architectures of multienyzme/multiprotein complexes and their arrangement within cells have now come to the fore. A uniquely powerful method in this field is electron cryomicroscopy (cryo-EM), which in its broadest sense, is all those techniques that image cold samples in the electron microscope. Cryo-EM allows individual enzymes and proteins, macromolecular complexes, assemblies, cells, and even tissues to be observed in a “frozen-hydrated,” nearnative state free from the artifacts of fixation, dehydration, plastic-embedding, or staining typically used in traditional forms of EM (Chapter 3, Vol. 481). This series of volumes is therefore dedicated to a description of the instruments, samples, protocols, and analyses that belong to the growing field of cryo-EM. The material could have been organized well by two schemes. The first is by the symmetry of the sample. Because the fundamental limitation in cryo-EM is radiation damage (Chapter 15, Vol. 481), a defining characteristic of each method is whether and how low-dose images of identical copies of the specimen can be averaged. In the most favorable case, large numbers of identical copies of the specimen of interest, like a single protein, can be purified and crystallized within thin “two-dimensional” crystals (Chapter 1, Vol. 481). In this case, truly atomic resolution reconstructions have been obtained through averaging very low dose images of millions of copies of the specimen (Chapter 11, Vol. 481; Chapter 4, Vol. 482; Chapters 5 and 6, Vol. 483). The next most favorable case is helical crystals, which present a range of views of the specimen within a single image (Chapter 2, Vol. 481 and Chapter 7, Vol. 483) and can also deliver atomically interpretable reconstructions, although through quite different data collection protocols and reconstruction mathematics (Chapters 5 and 6, Vol. 482). At an intermediate level of (60-fold) symmetry, icosahedral viruses have their own set of optimal imaging and reconstruction protocols, and are just now also reaching atomic interpretability (Chapters 7 and 14, Vol. 482). Less symmetric particles, such as many multienyzme/multiprotein complexes, invite yet another set of challenges and methods (Chapters 3, 5, and 6, Vol. 481; Chapters 8–10, Vol. 482). Many are conformationally heterogeneous, requiring that images of different particles xv

xvi

Preface

be first classified and then averaged (Chapters 10 and 12, Vol. 482; Chapters 8 and 9, Vol. 483). Heterogeneity and the precision to which these images can be aligned have limited most such reconstructions to “sub-nanometer” resolution, where the folds of proteins are clear but not much more (Chapter 1, Vol. 483). Finally, the most challenging samples are those which are simply unique (Chapter 8,Vol. 481), eliminating any chance of improving the clarity of reconstructions through averaging. For these, tomographic methods are required (Chapter 12, Vol. 481; Chapter 13, Vol. 482), and only nanometer resolutions can be obtained (Chapters 10–13, Vol. 483). But instead of organizing topics according to symmetry, following a wonderful historical perspective by David DeRosier (Historical Perspective,Vol. 481), I chose to order the topics in experimental sequence: Sample preparation and data collection/microscopy (Vol. 481); 3-D reconstruction (Vol. 482); and analyses and interpretation, including case studies (Vol. 483). This organization emphasizes how the relatedness of the mathematics (Chapter 1, Vol. 482), instrumentation (Chapters 10 and 14, Vol. 482), and methods (Chapter 15, Vol. 482; Chapter 9, Vol. 481) underlying all cryoEM approaches allow practictioners to easily move between them. It further highlights how in a growing number of recent cases, the methods are being mixed (Chapter 13, Vol. 481), for instance, through the application of “single particle-like” approaches to “unbend” and average 2-D and helical crystals (Chapter 6, Vol. 482), but also average subvolumes within tomograms. Moreover, different samples are always more-or-less well-behaved, so the actual resolution achieved may be less than theoretically possible for a particular symmetry, or to the opposite effect; extensively known constraints may allow a more specific interpretation than usual for a given resolution (Chapters 2-4 and 6, Vol. 483). Nevertheless, within each section, the articles are ordered as much as possible according to the symmetry of the sample as described above (i.e. methods for preparing samples proceed from 2-D and helical crystals to sectioning of high-pressure-frozen tissues; Chapter 8, Vol. 481). The cryo-EM beginner with a new sample must then first recognize its symmetry and then identify the relevant chapters within each volume. As a final note, our field has not yet reached a consensus on the placement of the prefix “cryo” and other details of the names of cryo-EM techniques. Thus, “cryo-electron microscopy” (CEM), “electron cryomicroscopy” (ECM), and “cryo-EM” should all be considered synonyms here. Likewise, “single particle reconstruction” (SPR) and “single particle analysis” (SPA) refer to a single technique, as do “cryo-electron tomography” (CET), “electron cryo-tomography” (ECT), and cryo-electron microscope tomography (cEMT). GRANT J. JENSEN

METHODS IN ENZYMOLOGY

VOLUME I. Preparation and Assay of Enzymes Edited by SIDNEY P. COLOWICK AND NATHAN O. KAPLAN VOLUME II. Preparation and Assay of Enzymes Edited by SIDNEY P. COLOWICK AND NATHAN O. KAPLAN VOLUME III. Preparation and Assay of Substrates Edited by SIDNEY P. COLOWICK AND NATHAN O. KAPLAN VOLUME IV. Special Techniques for the Enzymologist Edited by SIDNEY P. COLOWICK AND NATHAN O. KAPLAN VOLUME V. Preparation and Assay of Enzymes Edited by SIDNEY P. COLOWICK AND NATHAN O. KAPLAN VOLUME VI. Preparation and Assay of Enzymes (Continued) Preparation and Assay of Substrates Special Techniques Edited by SIDNEY P. COLOWICK AND NATHAN O. KAPLAN VOLUME VII. Cumulative Subject Index Edited by SIDNEY P. COLOWICK AND NATHAN O. KAPLAN VOLUME VIII. Complex Carbohydrates Edited by ELIZABETH F. NEUFELD AND VICTOR GINSBURG VOLUME IX. Carbohydrate Metabolism Edited by WILLIS A. WOOD VOLUME X. Oxidation and Phosphorylation Edited by RONALD W. ESTABROOK AND MAYNARD E. PULLMAN VOLUME XI. Enzyme Structure Edited by C. H. W. HIRS VOLUME XII. Nucleic Acids (Parts A and B) Edited by LAWRENCE GROSSMAN AND KIVIE MOLDAVE VOLUME XIII. Citric Acid Cycle Edited by J. M. LOWENSTEIN VOLUME XIV. Lipids Edited by J. M. LOWENSTEIN VOLUME XV. Steroids and Terpenoids Edited by RAYMOND B. CLAYTON xvii

xviii

Methods in Enzymology

VOLUME XVI. Fast Reactions Edited by KENNETH KUSTIN VOLUME XVII. Metabolism of Amino Acids and Amines (Parts A and B) Edited by HERBERT TABOR AND CELIA WHITE TABOR VOLUME XVIII. Vitamins and Coenzymes (Parts A, B, and C) Edited by DONALD B. MCCORMICK AND LEMUEL D. WRIGHT VOLUME XIX. Proteolytic Enzymes Edited by GERTRUDE E. PERLMANN AND LASZLO LORAND VOLUME XX. Nucleic Acids and Protein Synthesis (Part C) Edited by KIVIE MOLDAVE AND LAWRENCE GROSSMAN VOLUME XXI. Nucleic Acids (Part D) Edited by LAWRENCE GROSSMAN AND KIVIE MOLDAVE VOLUME XXII. Enzyme Purification and Related Techniques Edited by WILLIAM B. JAKOBY VOLUME XXIII. Photosynthesis (Part A) Edited by ANTHONY SAN PIETRO VOLUME XXIV. Photosynthesis and Nitrogen Fixation (Part B) Edited by ANTHONY SAN PIETRO VOLUME XXV. Enzyme Structure (Part B) Edited by C. H. W. HIRS AND SERGE N. TIMASHEFF VOLUME XXVI. Enzyme Structure (Part C) Edited by C. H. W. HIRS AND SERGE N. TIMASHEFF VOLUME XXVII. Enzyme Structure (Part D) Edited by C. H. W. HIRS AND SERGE N. TIMASHEFF VOLUME XXVIII. Complex Carbohydrates (Part B) Edited by VICTOR GINSBURG VOLUME XXIX. Nucleic Acids and Protein Synthesis (Part E) Edited by LAWRENCE GROSSMAN AND KIVIE MOLDAVE VOLUME XXX. Nucleic Acids and Protein Synthesis (Part F) Edited by KIVIE MOLDAVE AND LAWRENCE GROSSMAN VOLUME XXXI. Biomembranes (Part A) Edited by SIDNEY FLEISCHER AND LESTER PACKER VOLUME XXXII. Biomembranes (Part B) Edited by SIDNEY FLEISCHER AND LESTER PACKER VOLUME XXXIII. Cumulative Subject Index Volumes I-XXX Edited by MARTHA G. DENNIS AND EDWARD A. DENNIS VOLUME XXXIV. Affinity Techniques (Enzyme Purification: Part B) Edited by WILLIAM B. JAKOBY AND MEIR WILCHEK

Methods in Enzymology

VOLUME XXXV. Lipids (Part B) Edited by JOHN M. LOWENSTEIN VOLUME XXXVI. Hormone Action (Part A: Steroid Hormones) Edited by BERT W. O’MALLEY AND JOEL G. HARDMAN VOLUME XXXVII. Hormone Action (Part B: Peptide Hormones) Edited by BERT W. O’MALLEY AND JOEL G. HARDMAN VOLUME XXXVIII. Hormone Action (Part C: Cyclic Nucleotides) Edited by JOEL G. HARDMAN AND BERT W. O’MALLEY VOLUME XXXIX. Hormone Action (Part D: Isolated Cells, Tissues, and Organ Systems) Edited by JOEL G. HARDMAN AND BERT W. O’MALLEY VOLUME XL. Hormone Action (Part E: Nuclear Structure and Function) Edited by BERT W. O’MALLEY AND JOEL G. HARDMAN VOLUME XLI. Carbohydrate Metabolism (Part B) Edited by W. A. WOOD VOLUME XLII. Carbohydrate Metabolism (Part C) Edited by W. A. WOOD VOLUME XLIII. Antibiotics Edited by JOHN H. HASH VOLUME XLIV. Immobilized Enzymes Edited by KLAUS MOSBACH VOLUME XLV. Proteolytic Enzymes (Part B) Edited by LASZLO LORAND VOLUME XLVI. Affinity Labeling Edited by WILLIAM B. JAKOBY AND MEIR WILCHEK VOLUME XLVII. Enzyme Structure (Part E) Edited by C. H. W. HIRS AND SERGE N. TIMASHEFF VOLUME XLVIII. Enzyme Structure (Part F) Edited by C. H. W. HIRS AND SERGE N. TIMASHEFF VOLUME XLIX. Enzyme Structure (Part G) Edited by C. H. W. HIRS AND SERGE N. TIMASHEFF VOLUME L. Complex Carbohydrates (Part C) Edited by VICTOR GINSBURG VOLUME LI. Purine and Pyrimidine Nucleotide Metabolism Edited by PATRICIA A. HOFFEE AND MARY ELLEN JONES VOLUME LII. Biomembranes (Part C: Biological Oxidations) Edited by SIDNEY FLEISCHER AND LESTER PACKER

xix

xx

Methods in Enzymology

VOLUME LIII. Biomembranes (Part D: Biological Oxidations) Edited by SIDNEY FLEISCHER AND LESTER PACKER VOLUME LIV. Biomembranes (Part E: Biological Oxidations) Edited by SIDNEY FLEISCHER AND LESTER PACKER VOLUME LV. Biomembranes (Part F: Bioenergetics) Edited by SIDNEY FLEISCHER AND LESTER PACKER VOLUME LVI. Biomembranes (Part G: Bioenergetics) Edited by SIDNEY FLEISCHER AND LESTER PACKER VOLUME LVII. Bioluminescence and Chemiluminescence Edited by MARLENE A. DELUCA VOLUME LVIII. Cell Culture Edited by WILLIAM B. JAKOBY AND IRA PASTAN VOLUME LIX. Nucleic Acids and Protein Synthesis (Part G) Edited by KIVIE MOLDAVE AND LAWRENCE GROSSMAN VOLUME LX. Nucleic Acids and Protein Synthesis (Part H) Edited by KIVIE MOLDAVE AND LAWRENCE GROSSMAN VOLUME 61. Enzyme Structure (Part H) Edited by C. H. W. HIRS AND SERGE N. TIMASHEFF VOLUME 62. Vitamins and Coenzymes (Part D) Edited by DONALD B. MCCORMICK AND LEMUEL D. WRIGHT VOLUME 63. Enzyme Kinetics and Mechanism (Part A: Initial Rate and Inhibitor Methods) Edited by DANIEL L. PURICH VOLUME 64. Enzyme Kinetics and Mechanism (Part B: Isotopic Probes and Complex Enzyme Systems) Edited by DANIEL L. PURICH VOLUME 65. Nucleic Acids (Part I) Edited by LAWRENCE GROSSMAN AND KIVIE MOLDAVE VOLUME 66. Vitamins and Coenzymes (Part E) Edited by DONALD B. MCCORMICK AND LEMUEL D. WRIGHT VOLUME 67. Vitamins and Coenzymes (Part F) Edited by DONALD B. MCCORMICK AND LEMUEL D. WRIGHT VOLUME 68. Recombinant DNA Edited by RAY WU VOLUME 69. Photosynthesis and Nitrogen Fixation (Part C) Edited by ANTHONY SAN PIETRO VOLUME 70. Immunochemical Techniques (Part A) Edited by HELEN VAN VUNAKIS AND JOHN J. LANGONE

Methods in Enzymology

xxi

VOLUME 71. Lipids (Part C) Edited by JOHN M. LOWENSTEIN VOLUME 72. Lipids (Part D) Edited by JOHN M. LOWENSTEIN VOLUME 73. Immunochemical Techniques (Part B) Edited by JOHN J. LANGONE AND HELEN VAN VUNAKIS VOLUME 74. Immunochemical Techniques (Part C) Edited by JOHN J. LANGONE AND HELEN VAN VUNAKIS VOLUME 75. Cumulative Subject Index Volumes XXXI, XXXII, XXXIV–LX Edited by EDWARD A. DENNIS AND MARTHA G. DENNIS VOLUME 76. Hemoglobins Edited by ERALDO ANTONINI, LUIGI ROSSI-BERNARDI, AND EMILIA CHIANCONE VOLUME 77. Detoxication and Drug Metabolism Edited by WILLIAM B. JAKOBY VOLUME 78. Interferons (Part A) Edited by SIDNEY PESTKA VOLUME 79. Interferons (Part B) Edited by SIDNEY PESTKA VOLUME 80. Proteolytic Enzymes (Part C) Edited by LASZLO LORAND VOLUME 81. Biomembranes (Part H: Visual Pigments and Purple Membranes, I) Edited by LESTER PACKER VOLUME 82. Structural and Contractile Proteins (Part A: Extracellular Matrix) Edited by LEON W. CUNNINGHAM AND DIXIE W. FREDERIKSEN VOLUME 83. Complex Carbohydrates (Part D) Edited by VICTOR GINSBURG VOLUME 84. Immunochemical Techniques (Part D: Selected Immunoassays) Edited by JOHN J. LANGONE AND HELEN VAN VUNAKIS VOLUME 85. Structural and Contractile Proteins (Part B: The Contractile Apparatus and the Cytoskeleton) Edited by DIXIE W. FREDERIKSEN AND LEON W. CUNNINGHAM VOLUME 86. Prostaglandins and Arachidonate Metabolites Edited by WILLIAM E. M. LANDS AND WILLIAM L. SMITH VOLUME 87. Enzyme Kinetics and Mechanism (Part C: Intermediates, Stereo-chemistry, and Rate Studies) Edited by DANIEL L. PURICH VOLUME 88. Biomembranes (Part I: Visual Pigments and Purple Membranes, II) Edited by LESTER PACKER

xxii

Methods in Enzymology

VOLUME 89. Carbohydrate Metabolism (Part D) Edited by WILLIS A. WOOD VOLUME 90. Carbohydrate Metabolism (Part E) Edited by WILLIS A. WOOD VOLUME 91. Enzyme Structure (Part I) Edited by C. H. W. HIRS AND SERGE N. TIMASHEFF VOLUME 92. Immunochemical Techniques (Part E: Monoclonal Antibodies and General Immunoassay Methods) Edited by JOHN J. LANGONE AND HELEN VAN VUNAKIS VOLUME 93. Immunochemical Techniques (Part F: Conventional Antibodies, Fc Receptors, and Cytotoxicity) Edited by JOHN J. LANGONE AND HELEN VAN VUNAKIS VOLUME 94. Polyamines Edited by HERBERT TABOR AND CELIA WHITE TABOR VOLUME 95. Cumulative Subject Index Volumes 61–74, 76–80 Edited by EDWARD A. DENNIS AND MARTHA G. DENNIS VOLUME 96. Biomembranes [Part J: Membrane Biogenesis: Assembly and Targeting (General Methods; Eukaryotes)] Edited by SIDNEY FLEISCHER AND BECCA FLEISCHER VOLUME 97. Biomembranes [Part K: Membrane Biogenesis: Assembly and Targeting (Prokaryotes, Mitochondria, and Chloroplasts)] Edited by SIDNEY FLEISCHER AND BECCA FLEISCHER VOLUME 98. Biomembranes (Part L: Membrane Biogenesis: Processing and Recycling) Edited by SIDNEY FLEISCHER AND BECCA FLEISCHER VOLUME 99. Hormone Action (Part F: Protein Kinases) Edited by JACKIE D. CORBIN AND JOEL G. HARDMAN VOLUME 100. Recombinant DNA (Part B) Edited by RAY WU, LAWRENCE GROSSMAN, AND KIVIE MOLDAVE VOLUME 101. Recombinant DNA (Part C) Edited by RAY WU, LAWRENCE GROSSMAN, AND KIVIE MOLDAVE VOLUME 102. Hormone Action (Part G: Calmodulin and Calcium-Binding Proteins) Edited by ANTHONY R. MEANS AND BERT W. O’MALLEY VOLUME 103. Hormone Action (Part H: Neuroendocrine Peptides) Edited by P. MICHAEL CONN VOLUME 104. Enzyme Purification and Related Techniques (Part C) Edited by WILLIAM B. JAKOBY

Methods in Enzymology

xxiii

VOLUME 105. Oxygen Radicals in Biological Systems Edited by LESTER PACKER VOLUME 106. Posttranslational Modifications (Part A) Edited by FINN WOLD AND KIVIE MOLDAVE VOLUME 107. Posttranslational Modifications (Part B) Edited by FINN WOLD AND KIVIE MOLDAVE VOLUME 108. Immunochemical Techniques (Part G: Separation and Characterization of Lymphoid Cells) Edited by GIOVANNI DI SABATO, JOHN J. LANGONE, AND HELEN VAN VUNAKIS VOLUME 109. Hormone Action (Part I: Peptide Hormones) Edited by LUTZ BIRNBAUMER AND BERT W. O’MALLEY VOLUME 110. Steroids and Isoprenoids (Part A) Edited by JOHN H. LAW AND HANS C. RILLING VOLUME 111. Steroids and Isoprenoids (Part B) Edited by JOHN H. LAW AND HANS C. RILLING VOLUME 112. Drug and Enzyme Targeting (Part A) Edited by KENNETH J. WIDDER AND RALPH GREEN VOLUME 113. Glutamate, Glutamine, Glutathione, and Related Compounds Edited by ALTON MEISTER VOLUME 114. Diffraction Methods for Biological Macromolecules (Part A) Edited by HAROLD W. WYCKOFF, C. H. W. HIRS, AND SERGE N. TIMASHEFF VOLUME 115. Diffraction Methods for Biological Macromolecules (Part B) Edited by HAROLD W. WYCKOFF, C. H. W. HIRS, AND SERGE N. TIMASHEFF VOLUME 116. Immunochemical Techniques (Part H: Effectors and Mediators of Lymphoid Cell Functions) Edited by GIOVANNI DI SABATO, JOHN J. LANGONE, AND HELEN VAN VUNAKIS VOLUME 117. Enzyme Structure (Part J) Edited by C. H. W. HIRS AND SERGE N. TIMASHEFF VOLUME 118. Plant Molecular Biology Edited by ARTHUR WEISSBACH AND HERBERT WEISSBACH VOLUME 119. Interferons (Part C) Edited by SIDNEY PESTKA VOLUME 120. Cumulative Subject Index Volumes 81–94, 96–101 VOLUME 121. Immunochemical Techniques (Part I: Hybridoma Technology and Monoclonal Antibodies) Edited by JOHN J. LANGONE AND HELEN VAN VUNAKIS VOLUME 122. Vitamins and Coenzymes (Part G) Edited by FRANK CHYTIL AND DONALD B. MCCORMICK

xxiv

Methods in Enzymology

VOLUME 123. Vitamins and Coenzymes (Part H) Edited by FRANK CHYTIL AND DONALD B. MCCORMICK VOLUME 124. Hormone Action (Part J: Neuroendocrine Peptides) Edited by P. MICHAEL CONN VOLUME 125. Biomembranes (Part M: Transport in Bacteria, Mitochondria, and Chloroplasts: General Approaches and Transport Systems) Edited by SIDNEY FLEISCHER AND BECCA FLEISCHER VOLUME 126. Biomembranes (Part N: Transport in Bacteria, Mitochondria, and Chloroplasts: Protonmotive Force) Edited by SIDNEY FLEISCHER AND BECCA FLEISCHER VOLUME 127. Biomembranes (Part O: Protons and Water: Structure and Translocation) Edited by LESTER PACKER VOLUME 128. Plasma Lipoproteins (Part A: Preparation, Structure, and Molecular Biology) Edited by JERE P. SEGREST AND JOHN J. ALBERS VOLUME 129. Plasma Lipoproteins (Part B: Characterization, Cell Biology, and Metabolism) Edited by JOHN J. ALBERS AND JERE P. SEGREST VOLUME 130. Enzyme Structure (Part K) Edited by C. H. W. HIRS AND SERGE N. TIMASHEFF VOLUME 131. Enzyme Structure (Part L) Edited by C. H. W. HIRS AND SERGE N. TIMASHEFF VOLUME 132. Immunochemical Techniques (Part J: Phagocytosis and Cell-Mediated Cytotoxicity) Edited by GIOVANNI DI SABATO AND JOHANNES EVERSE VOLUME 133. Bioluminescence and Chemiluminescence (Part B) Edited by MARLENE DELUCA AND WILLIAM D. MCELROY VOLUME 134. Structural and Contractile Proteins (Part C: The Contractile Apparatus and the Cytoskeleton) Edited by RICHARD B. VALLEE VOLUME 135. Immobilized Enzymes and Cells (Part B) Edited by KLAUS MOSBACH VOLUME 136. Immobilized Enzymes and Cells (Part C) Edited by KLAUS MOSBACH VOLUME 137. Immobilized Enzymes and Cells (Part D) Edited by KLAUS MOSBACH VOLUME 138. Complex Carbohydrates (Part E) Edited by VICTOR GINSBURG

Methods in Enzymology

xxv

VOLUME 139. Cellular Regulators (Part A: Calcium- and Calmodulin-Binding Proteins) Edited by ANTHONY R. MEANS AND P. MICHAEL CONN VOLUME 140. Cumulative Subject Index Volumes 102–119, 121–134 VOLUME 141. Cellular Regulators (Part B: Calcium and Lipids) Edited by P. MICHAEL CONN AND ANTHONY R. MEANS VOLUME 142. Metabolism of Aromatic Amino Acids and Amines Edited by SEYMOUR KAUFMAN VOLUME 143. Sulfur and Sulfur Amino Acids Edited by WILLIAM B. JAKOBY AND OWEN GRIFFITH VOLUME 144. Structural and Contractile Proteins (Part D: Extracellular Matrix) Edited by LEON W. CUNNINGHAM VOLUME 145. Structural and Contractile Proteins (Part E: Extracellular Matrix) Edited by LEON W. CUNNINGHAM VOLUME 146. Peptide Growth Factors (Part A) Edited by DAVID BARNES AND DAVID A. SIRBASKU VOLUME 147. Peptide Growth Factors (Part B) Edited by DAVID BARNES AND DAVID A. SIRBASKU VOLUME 148. Plant Cell Membranes Edited by LESTER PACKER AND ROLAND DOUCE VOLUME 149. Drug and Enzyme Targeting (Part B) Edited by RALPH GREEN AND KENNETH J. WIDDER VOLUME 150. Immunochemical Techniques (Part K: In Vitro Models of B and T Cell Functions and Lymphoid Cell Receptors) Edited by GIOVANNI DI SABATO VOLUME 151. Molecular Genetics of Mammalian Cells Edited by MICHAEL M. GOTTESMAN VOLUME 152. Guide to Molecular Cloning Techniques Edited by SHELBY L. BERGER AND ALAN R. KIMMEL VOLUME 153. Recombinant DNA (Part D) Edited by RAY WU AND LAWRENCE GROSSMAN VOLUME 154. Recombinant DNA (Part E) Edited by RAY WU AND LAWRENCE GROSSMAN VOLUME 155. Recombinant DNA (Part F) Edited by RAY WU VOLUME 156. Biomembranes (Part P: ATP-Driven Pumps and Related Transport: The Na, K-Pump) Edited by SIDNEY FLEISCHER AND BECCA FLEISCHER

xxvi

Methods in Enzymology

VOLUME 157. Biomembranes (Part Q: ATP-Driven Pumps and Related Transport: Calcium, Proton, and Potassium Pumps) Edited by SIDNEY FLEISCHER AND BECCA FLEISCHER VOLUME 158. Metalloproteins (Part A) Edited by JAMES F. RIORDAN AND BERT L. VALLEE VOLUME 159. Initiation and Termination of Cyclic Nucleotide Action Edited by JACKIE D. CORBIN AND ROGER A. JOHNSON VOLUME 160. Biomass (Part A: Cellulose and Hemicellulose) Edited by WILLIS A. WOOD AND SCOTT T. KELLOGG VOLUME 161. Biomass (Part B: Lignin, Pectin, and Chitin) Edited by WILLIS A. WOOD AND SCOTT T. KELLOGG VOLUME 162. Immunochemical Techniques (Part L: Chemotaxis and Inflammation) Edited by GIOVANNI DI SABATO VOLUME 163. Immunochemical Techniques (Part M: Chemotaxis and Inflammation) Edited by GIOVANNI DI SABATO VOLUME 164. Ribosomes Edited by HARRY F. NOLLER, JR., AND KIVIE MOLDAVE VOLUME 165. Microbial Toxins: Tools for Enzymology Edited by SIDNEY HARSHMAN VOLUME 166. Branched-Chain Amino Acids Edited by ROBERT HARRIS AND JOHN R. SOKATCH VOLUME 167. Cyanobacteria Edited by LESTER PACKER AND ALEXANDER N. GLAZER VOLUME 168. Hormone Action (Part K: Neuroendocrine Peptides) Edited by P. MICHAEL CONN VOLUME 169. Platelets: Receptors, Adhesion, Secretion (Part A) Edited by JACEK HAWIGER VOLUME 170. Nucleosomes Edited by PAUL M. WASSARMAN AND ROGER D. KORNBERG VOLUME 171. Biomembranes (Part R: Transport Theory: Cells and Model Membranes) Edited by SIDNEY FLEISCHER AND BECCA FLEISCHER VOLUME 172. Biomembranes (Part S: Transport: Membrane Isolation and Characterization) Edited by SIDNEY FLEISCHER AND BECCA FLEISCHER

Methods in Enzymology

xxvii

VOLUME 173. Biomembranes [Part T: Cellular and Subcellular Transport: Eukaryotic (Nonepithelial) Cells] Edited by SIDNEY FLEISCHER AND BECCA FLEISCHER VOLUME 174. Biomembranes [Part U: Cellular and Subcellular Transport: Eukaryotic (Nonepithelial) Cells] Edited by SIDNEY FLEISCHER AND BECCA FLEISCHER VOLUME 175. Cumulative Subject Index Volumes 135–139, 141–167 VOLUME 176. Nuclear Magnetic Resonance (Part A: Spectral Techniques and Dynamics) Edited by NORMAN J. OPPENHEIMER AND THOMAS L. JAMES VOLUME 177. Nuclear Magnetic Resonance (Part B: Structure and Mechanism) Edited by NORMAN J. OPPENHEIMER AND THOMAS L. JAMES VOLUME 178. Antibodies, Antigens, and Molecular Mimicry Edited by JOHN J. LANGONE VOLUME 179. Complex Carbohydrates (Part F) Edited by VICTOR GINSBURG VOLUME 180. RNA Processing (Part A: General Methods) Edited by JAMES E. DAHLBERG AND JOHN N. ABELSON VOLUME 181. RNA Processing (Part B: Specific Methods) Edited by JAMES E. DAHLBERG AND JOHN N. ABELSON VOLUME 182. Guide to Protein Purification Edited by MURRAY P. DEUTSCHER VOLUME 183. Molecular Evolution: Computer Analysis of Protein and Nucleic Acid Sequences Edited by RUSSELL F. DOOLITTLE VOLUME 184. Avidin-Biotin Technology Edited by MEIR WILCHEK AND EDWARD A. BAYER VOLUME 185. Gene Expression Technology Edited by DAVID V. GOEDDEL VOLUME 186. Oxygen Radicals in Biological Systems (Part B: Oxygen Radicals and Antioxidants) Edited by LESTER PACKER AND ALEXANDER N. GLAZER VOLUME 187. Arachidonate Related Lipid Mediators Edited by ROBERT C. MURPHY AND FRANK A. FITZPATRICK VOLUME 188. Hydrocarbons and Methylotrophy Edited by MARY E. LIDSTROM VOLUME 189. Retinoids (Part A: Molecular and Metabolic Aspects) Edited by LESTER PACKER

xxviii

Methods in Enzymology

VOLUME 190. Retinoids (Part B: Cell Differentiation and Clinical Applications) Edited by LESTER PACKER VOLUME 191. Biomembranes (Part V: Cellular and Subcellular Transport: Epithelial Cells) Edited by SIDNEY FLEISCHER AND BECCA FLEISCHER VOLUME 192. Biomembranes (Part W: Cellular and Subcellular Transport: Epithelial Cells) Edited by SIDNEY FLEISCHER AND BECCA FLEISCHER VOLUME 193. Mass Spectrometry Edited by JAMES A. MCCLOSKEY VOLUME 194. Guide to Yeast Genetics and Molecular Biology Edited by CHRISTINE GUTHRIE AND GERALD R. FINK VOLUME 195. Adenylyl Cyclase, G Proteins, and Guanylyl Cyclase Edited by ROGER A. JOHNSON AND JACKIE D. CORBIN VOLUME 196. Molecular Motors and the Cytoskeleton Edited by RICHARD B. VALLEE VOLUME 197. Phospholipases Edited by EDWARD A. DENNIS VOLUME 198. Peptide Growth Factors (Part C) Edited by DAVID BARNES, J. P. MATHER, AND GORDON H. SATO VOLUME 199. Cumulative Subject Index Volumes 168–174, 176–194 VOLUME 200. Protein Phosphorylation (Part A: Protein Kinases: Assays, Purification, Antibodies, Functional Analysis, Cloning, and Expression) Edited by TONY HUNTER AND BARTHOLOMEW M. SEFTON VOLUME 201. Protein Phosphorylation (Part B: Analysis of Protein Phosphorylation, Protein Kinase Inhibitors, and Protein Phosphatases) Edited by TONY HUNTER AND BARTHOLOMEW M. SEFTON VOLUME 202. Molecular Design and Modeling: Concepts and Applications (Part A: Proteins, Peptides, and Enzymes) Edited by JOHN J. LANGONE VOLUME 203. Molecular Design and Modeling: Concepts and Applications (Part B: Antibodies and Antigens, Nucleic Acids, Polysaccharides, and Drugs) Edited by JOHN J. LANGONE VOLUME 204. Bacterial Genetic Systems Edited by JEFFREY H. MILLER VOLUME 205. Metallobiochemistry (Part B: Metallothionein and Related Molecules) Edited by JAMES F. RIORDAN AND BERT L. VALLEE

Methods in Enzymology

xxix

VOLUME 206. Cytochrome P450 Edited by MICHAEL R. WATERMAN AND ERIC F. JOHNSON VOLUME 207. Ion Channels Edited by BERNARDO RUDY AND LINDA E. IVERSON VOLUME 208. Protein–DNA Interactions Edited by ROBERT T. SAUER VOLUME 209. Phospholipid Biosynthesis Edited by EDWARD A. DENNIS AND DENNIS E. VANCE VOLUME 210. Numerical Computer Methods Edited by LUDWIG BRAND AND MICHAEL L. JOHNSON VOLUME 211. DNA Structures (Part A: Synthesis and Physical Analysis of DNA) Edited by DAVID M. J. LILLEY AND JAMES E. DAHLBERG VOLUME 212. DNA Structures (Part B: Chemical and Electrophoretic Analysis of DNA) Edited by DAVID M. J. LILLEY AND JAMES E. DAHLBERG VOLUME 213. Carotenoids (Part A: Chemistry, Separation, Quantitation, and Antioxidation) Edited by LESTER PACKER VOLUME 214. Carotenoids (Part B: Metabolism, Genetics, and Biosynthesis) Edited by LESTER PACKER VOLUME 215. Platelets: Receptors, Adhesion, Secretion (Part B) Edited by JACEK J. HAWIGER VOLUME 216. Recombinant DNA (Part G) Edited by RAY WU VOLUME 217. Recombinant DNA (Part H) Edited by RAY WU VOLUME 218. Recombinant DNA (Part I) Edited by RAY WU VOLUME 219. Reconstitution of Intracellular Transport Edited by JAMES E. ROTHMAN VOLUME 220. Membrane Fusion Techniques (Part A) Edited by NEJAT DU¨ZGU¨NES¸ VOLUME 221. Membrane Fusion Techniques (Part B) Edited by NEJAT DU¨ZGU¨NES¸ VOLUME 222. Proteolytic Enzymes in Coagulation, Fibrinolysis, and Complement Activation (Part A: Mammalian Blood Coagulation Factors and Inhibitors) Edited by LASZLO LORAND AND KENNETH G. MANN

xxx

Methods in Enzymology

VOLUME 223. Proteolytic Enzymes in Coagulation, Fibrinolysis, and Complement Activation (Part B: Complement Activation, Fibrinolysis, and Nonmammalian Blood Coagulation Factors) Edited by LASZLO LORAND AND KENNETH G. MANN VOLUME 224. Molecular Evolution: Producing the Biochemical Data Edited by ELIZABETH ANNE ZIMMER, THOMAS J. WHITE, REBECCA L. CANN, AND ALLAN C. WILSON VOLUME 225. Guide to Techniques in Mouse Development Edited by PAUL M. WASSARMAN AND MELVIN L. DEPAMPHILIS VOLUME 226. Metallobiochemistry (Part C: Spectroscopic and Physical Methods for Probing Metal Ion Environments in Metalloenzymes and Metalloproteins) Edited by JAMES F. RIORDAN AND BERT L. VALLEE VOLUME 227. Metallobiochemistry (Part D: Physical and Spectroscopic Methods for Probing Metal Ion Environments in Metalloproteins) Edited by JAMES F. RIORDAN AND BERT L. VALLEE VOLUME 228. Aqueous Two-Phase Systems Edited by HARRY WALTER AND GO¨TE JOHANSSON VOLUME 229. Cumulative Subject Index Volumes 195–198, 200–227 VOLUME 230. Guide to Techniques in Glycobiology Edited by WILLIAM J. LENNARZ AND GERALD W. HART VOLUME 231. Hemoglobins (Part B: Biochemical and Analytical Methods) Edited by JOHANNES EVERSE, KIM D. VANDEGRIFF, AND ROBERT M. WINSLOW VOLUME 232. Hemoglobins (Part C: Biophysical Methods) Edited by JOHANNES EVERSE, KIM D. VANDEGRIFF, AND ROBERT M. WINSLOW VOLUME 233. Oxygen Radicals in Biological Systems (Part C) Edited by LESTER PACKER VOLUME 234. Oxygen Radicals in Biological Systems (Part D) Edited by LESTER PACKER VOLUME 235. Bacterial Pathogenesis (Part A: Identification and Regulation of Virulence Factors) Edited by VIRGINIA L. CLARK AND PATRIK M. BAVOIL VOLUME 236. Bacterial Pathogenesis (Part B: Integration of Pathogenic Bacteria with Host Cells) Edited by VIRGINIA L. CLARK AND PATRIK M. BAVOIL VOLUME 237. Heterotrimeric G Proteins Edited by RAVI IYENGAR VOLUME 238. Heterotrimeric G-Protein Effectors Edited by RAVI IYENGAR

Methods in Enzymology

xxxi

VOLUME 239. Nuclear Magnetic Resonance (Part C) Edited by THOMAS L. JAMES AND NORMAN J. OPPENHEIMER VOLUME 240. Numerical Computer Methods (Part B) Edited by MICHAEL L. JOHNSON AND LUDWIG BRAND VOLUME 241. Retroviral Proteases Edited by LAWRENCE C. KUO AND JULES A. SHAFER VOLUME 242. Neoglycoconjugates (Part A) Edited by Y. C. LEE AND REIKO T. LEE VOLUME 243. Inorganic Microbial Sulfur Metabolism Edited by HARRY D. PECK, JR., AND JEAN LEGALL VOLUME 244. Proteolytic Enzymes: Serine and Cysteine Peptidases Edited by ALAN J. BARRETT VOLUME 245. Extracellular Matrix Components Edited by E. RUOSLAHTI AND E. ENGVALL VOLUME 246. Biochemical Spectroscopy Edited by KENNETH SAUER VOLUME 247. Neoglycoconjugates (Part B: Biomedical Applications) Edited by Y. C. LEE AND REIKO T. LEE VOLUME 248. Proteolytic Enzymes: Aspartic and Metallo Peptidases Edited by ALAN J. BARRETT VOLUME 249. Enzyme Kinetics and Mechanism (Part D: Developments in Enzyme Dynamics) Edited by DANIEL L. PURICH VOLUME 250. Lipid Modifications of Proteins Edited by PATRICK J. CASEY AND JANICE E. BUSS VOLUME 251. Biothiols (Part A: Monothiols and Dithiols, Protein Thiols, and Thiyl Radicals) Edited by LESTER PACKER VOLUME 252. Biothiols (Part B: Glutathione and Thioredoxin; Thiols in Signal Transduction and Gene Regulation) Edited by LESTER PACKER VOLUME 253. Adhesion of Microbial Pathogens Edited by RON J. DOYLE AND ITZHAK OFEK VOLUME 254. Oncogene Techniques Edited by PETER K. VOGT AND INDER M. VERMA VOLUME 255. Small GTPases and Their Regulators (Part A: Ras Family) Edited by W. E. BALCH, CHANNING J. DER, AND ALAN HALL

xxxii

Methods in Enzymology

VOLUME 256. Small GTPases and Their Regulators (Part B: Rho Family) Edited by W. E. BALCH, CHANNING J. DER, AND ALAN HALL VOLUME 257. Small GTPases and Their Regulators (Part C: Proteins Involved in Transport) Edited by W. E. BALCH, CHANNING J. DER, AND ALAN HALL VOLUME 258. Redox-Active Amino Acids in Biology Edited by JUDITH P. KLINMAN VOLUME 259. Energetics of Biological Macromolecules Edited by MICHAEL L. JOHNSON AND GARY K. ACKERS VOLUME 260. Mitochondrial Biogenesis and Genetics (Part A) Edited by GIUSEPPE M. ATTARDI AND ANNE CHOMYN VOLUME 261. Nuclear Magnetic Resonance and Nucleic Acids Edited by THOMAS L. JAMES VOLUME 262. DNA Replication Edited by JUDITH L. CAMPBELL VOLUME 263. Plasma Lipoproteins (Part C: Quantitation) Edited by WILLIAM A. BRADLEY, SANDRA H. GIANTURCO, AND JERE P. SEGREST VOLUME 264. Mitochondrial Biogenesis and Genetics (Part B) Edited by GIUSEPPE M. ATTARDI AND ANNE CHOMYN VOLUME 265. Cumulative Subject Index Volumes 228, 230–262 VOLUME 266. Computer Methods for Macromolecular Sequence Analysis Edited by RUSSELL F. DOOLITTLE VOLUME 267. Combinatorial Chemistry Edited by JOHN N. ABELSON VOLUME 268. Nitric Oxide (Part A: Sources and Detection of NO; NO Synthase) Edited by LESTER PACKER VOLUME 269. Nitric Oxide (Part B: Physiological and Pathological Processes) Edited by LESTER PACKER VOLUME 270. High Resolution Separation and Analysis of Biological Macromolecules (Part A: Fundamentals) Edited by BARRY L. KARGER AND WILLIAM S. HANCOCK VOLUME 271. High Resolution Separation and Analysis of Biological Macromolecules (Part B: Applications) Edited by BARRY L. KARGER AND WILLIAM S. HANCOCK VOLUME 272. Cytochrome P450 (Part B) Edited by ERIC F. JOHNSON AND MICHAEL R. WATERMAN VOLUME 273. RNA Polymerase and Associated Factors (Part A) Edited by SANKAR ADHYA

Methods in Enzymology

xxxiii

VOLUME 274. RNA Polymerase and Associated Factors (Part B) Edited by SANKAR ADHYA VOLUME 275. Viral Polymerases and Related Proteins Edited by LAWRENCE C. KUO, DAVID B. OLSEN, AND STEVEN S. CARROLL VOLUME 276. Macromolecular Crystallography (Part A) Edited by CHARLES W. CARTER, JR., AND ROBERT M. SWEET VOLUME 277. Macromolecular Crystallography (Part B) Edited by CHARLES W. CARTER, JR., AND ROBERT M. SWEET VOLUME 278. Fluorescence Spectroscopy Edited by LUDWIG BRAND AND MICHAEL L. JOHNSON VOLUME 279. Vitamins and Coenzymes (Part I) Edited by DONALD B. MCCORMICK, JOHN W. SUTTIE, AND CONRAD WAGNER VOLUME 280. Vitamins and Coenzymes (Part J) Edited by DONALD B. MCCORMICK, JOHN W. SUTTIE, AND CONRAD WAGNER VOLUME 281. Vitamins and Coenzymes (Part K) Edited by DONALD B. MCCORMICK, JOHN W. SUTTIE, AND CONRAD WAGNER VOLUME 282. Vitamins and Coenzymes (Part L) Edited by DONALD B. MCCORMICK, JOHN W. SUTTIE, AND CONRAD WAGNER VOLUME 283. Cell Cycle Control Edited by WILLIAM G. DUNPHY VOLUME 284. Lipases (Part A: Biotechnology) Edited by BYRON RUBIN AND EDWARD A. DENNIS VOLUME 285. Cumulative Subject Index Volumes 263, 264, 266–284, 286–289 VOLUME 286. Lipases (Part B: Enzyme Characterization and Utilization) Edited by BYRON RUBIN AND EDWARD A. DENNIS VOLUME 287. Chemokines Edited by RICHARD HORUK VOLUME 288. Chemokine Receptors Edited by RICHARD HORUK VOLUME 289. Solid Phase Peptide Synthesis Edited by GREGG B. FIELDS VOLUME 290. Molecular Chaperones Edited by GEORGE H. LORIMER AND THOMAS BALDWIN VOLUME 291. Caged Compounds Edited by GERARD MARRIOTT VOLUME 292. ABC Transporters: Biochemical, Cellular, and Molecular Aspects Edited by SURESH V. AMBUDKAR AND MICHAEL M. GOTTESMAN

xxxiv

Methods in Enzymology

VOLUME 293. Ion Channels (Part B) Edited by P. MICHAEL CONN VOLUME 294. Ion Channels (Part C) Edited by P. MICHAEL CONN VOLUME 295. Energetics of Biological Macromolecules (Part B) Edited by GARY K. ACKERS AND MICHAEL L. JOHNSON VOLUME 296. Neurotransmitter Transporters Edited by SUSAN G. AMARA VOLUME 297. Photosynthesis: Molecular Biology of Energy Capture Edited by LEE MCINTOSH VOLUME 298. Molecular Motors and the Cytoskeleton (Part B) Edited by RICHARD B. VALLEE VOLUME 299. Oxidants and Antioxidants (Part A) Edited by LESTER PACKER VOLUME 300. Oxidants and Antioxidants (Part B) Edited by LESTER PACKER VOLUME 301. Nitric Oxide: Biological and Antioxidant Activities (Part C) Edited by LESTER PACKER VOLUME 302. Green Fluorescent Protein Edited by P. MICHAEL CONN VOLUME 303. cDNA Preparation and Display Edited by SHERMAN M. WEISSMAN VOLUME 304. Chromatin Edited by PAUL M. WASSARMAN AND ALAN P. WOLFFE VOLUME 305. Bioluminescence and Chemiluminescence (Part C) Edited by THOMAS O. BALDWIN AND MIRIAM M. ZIEGLER VOLUME 306. Expression of Recombinant Genes in Eukaryotic Systems Edited by JOSEPH C. GLORIOSO AND MARTIN C. SCHMIDT VOLUME 307. Confocal Microscopy Edited by P. MICHAEL CONN VOLUME 308. Enzyme Kinetics and Mechanism (Part E: Energetics of Enzyme Catalysis) Edited by DANIEL L. PURICH AND VERN L. SCHRAMM VOLUME 309. Amyloid, Prions, and Other Protein Aggregates Edited by RONALD WETZEL VOLUME 310. Biofilms Edited by RON J. DOYLE

Methods in Enzymology

xxxv

VOLUME 311. Sphingolipid Metabolism and Cell Signaling (Part A) Edited by ALFRED H. MERRILL, JR., AND YUSUF A. HANNUN VOLUME 312. Sphingolipid Metabolism and Cell Signaling (Part B) Edited by ALFRED H. MERRILL, JR., AND YUSUF A. HANNUN VOLUME 313. Antisense Technology (Part A: General Methods, Methods of Delivery, and RNA Studies) Edited by M. IAN PHILLIPS VOLUME 314. Antisense Technology (Part B: Applications) Edited by M. IAN PHILLIPS VOLUME 315. Vertebrate Phototransduction and the Visual Cycle (Part A) Edited by KRZYSZTOF PALCZEWSKI VOLUME 316. Vertebrate Phototransduction and the Visual Cycle (Part B) Edited by KRZYSZTOF PALCZEWSKI VOLUME 317. RNA–Ligand Interactions (Part A: Structural Biology Methods) Edited by DANIEL W. CELANDER AND JOHN N. ABELSON VOLUME 318. RNA–Ligand Interactions (Part B: Molecular Biology Methods) Edited by DANIEL W. CELANDER AND JOHN N. ABELSON VOLUME 319. Singlet Oxygen, UV-A, and Ozone Edited by LESTER PACKER AND HELMUT SIES VOLUME 320. Cumulative Subject Index Volumes 290–319 VOLUME 321. Numerical Computer Methods (Part C) Edited by MICHAEL L. JOHNSON AND LUDWIG BRAND VOLUME 322. Apoptosis Edited by JOHN C. REED VOLUME 323. Energetics of Biological Macromolecules (Part C) Edited by MICHAEL L. JOHNSON AND GARY K. ACKERS VOLUME 324. Branched-Chain Amino Acids (Part B) Edited by ROBERT A. HARRIS AND JOHN R. SOKATCH VOLUME 325. Regulators and Effectors of Small GTPases (Part D: Rho Family) Edited by W. E. BALCH, CHANNING J. DER, AND ALAN HALL VOLUME 326. Applications of Chimeric Genes and Hybrid Proteins (Part A: Gene Expression and Protein Purification) Edited by JEREMY THORNER, SCOTT D. EMR, AND JOHN N. ABELSON VOLUME 327. Applications of Chimeric Genes and Hybrid Proteins (Part B: Cell Biology and Physiology) Edited by JEREMY THORNER, SCOTT D. EMR, AND JOHN N. ABELSON

xxxvi

Methods in Enzymology

VOLUME 328. Applications of Chimeric Genes and Hybrid Proteins (Part C: Protein–Protein Interactions and Genomics) Edited by JEREMY THORNER, SCOTT D. EMR, AND JOHN N. ABELSON VOLUME 329. Regulators and Effectors of Small GTPases (Part E: GTPases Involved in Vesicular Traffic) Edited by W. E. BALCH, CHANNING J. DER, AND ALAN HALL VOLUME 330. Hyperthermophilic Enzymes (Part A) Edited by MICHAEL W. W. ADAMS AND ROBERT M. KELLY VOLUME 331. Hyperthermophilic Enzymes (Part B) Edited by MICHAEL W. W. ADAMS AND ROBERT M. KELLY VOLUME 332. Regulators and Effectors of Small GTPases (Part F: Ras Family I) Edited by W. E. BALCH, CHANNING J. DER, AND ALAN HALL VOLUME 333. Regulators and Effectors of Small GTPases (Part G: Ras Family II) Edited by W. E. BALCH, CHANNING J. DER, AND ALAN HALL VOLUME 334. Hyperthermophilic Enzymes (Part C) Edited by MICHAEL W. W. ADAMS AND ROBERT M. KELLY VOLUME 335. Flavonoids and Other Polyphenols Edited by LESTER PACKER VOLUME 336. Microbial Growth in Biofilms (Part A: Developmental and Molecular Biological Aspects) Edited by RON J. DOYLE VOLUME 337. Microbial Growth in Biofilms (Part B: Special Environments and Physicochemical Aspects) Edited by RON J. DOYLE VOLUME 338. Nuclear Magnetic Resonance of Biological Macromolecules (Part A) Edited by THOMAS L. JAMES, VOLKER DO¨TSCH, AND ULI SCHMITZ VOLUME 339. Nuclear Magnetic Resonance of Biological Macromolecules (Part B) Edited by THOMAS L. JAMES, VOLKER DO¨TSCH, AND ULI SCHMITZ VOLUME 340. Drug–Nucleic Acid Interactions Edited by JONATHAN B. CHAIRES AND MICHAEL J. WARING VOLUME 341. Ribonucleases (Part A) Edited by ALLEN W. NICHOLSON VOLUME 342. Ribonucleases (Part B) Edited by ALLEN W. NICHOLSON VOLUME 343. G Protein Pathways (Part A: Receptors) Edited by RAVI IYENGAR AND JOHN D. HILDEBRANDT VOLUME 344. G Protein Pathways (Part B: G Proteins and Their Regulators) Edited by RAVI IYENGAR AND JOHN D. HILDEBRANDT

Methods in Enzymology

xxxvii

VOLUME 345. G Protein Pathways (Part C: Effector Mechanisms) Edited by RAVI IYENGAR AND JOHN D. HILDEBRANDT VOLUME 346. Gene Therapy Methods Edited by M. IAN PHILLIPS VOLUME 347. Protein Sensors and Reactive Oxygen Species (Part A: Selenoproteins and Thioredoxin) Edited by HELMUT SIES AND LESTER PACKER VOLUME 348. Protein Sensors and Reactive Oxygen Species (Part B: Thiol Enzymes and Proteins) Edited by HELMUT SIES AND LESTER PACKER VOLUME 349. Superoxide Dismutase Edited by LESTER PACKER VOLUME 350. Guide to Yeast Genetics and Molecular and Cell Biology (Part B) Edited by CHRISTINE GUTHRIE AND GERALD R. FINK VOLUME 351. Guide to Yeast Genetics and Molecular and Cell Biology (Part C) Edited by CHRISTINE GUTHRIE AND GERALD R. FINK VOLUME 352. Redox Cell Biology and Genetics (Part A) Edited by CHANDAN K. SEN AND LESTER PACKER VOLUME 353. Redox Cell Biology and Genetics (Part B) Edited by CHANDAN K. SEN AND LESTER PACKER VOLUME 354. Enzyme Kinetics and Mechanisms (Part F: Detection and Characterization of Enzyme Reaction Intermediates) Edited by DANIEL L. PURICH VOLUME 355. Cumulative Subject Index Volumes 321–354 VOLUME 356. Laser Capture Microscopy and Microdissection Edited by P. MICHAEL CONN VOLUME 357. Cytochrome P450, Part C Edited by ERIC F. JOHNSON AND MICHAEL R. WATERMAN VOLUME 358. Bacterial Pathogenesis (Part C: Identification, Regulation, and Function of Virulence Factors) Edited by VIRGINIA L. CLARK AND PATRIK M. BAVOIL VOLUME 359. Nitric Oxide (Part D) Edited by ENRIQUE CADENAS AND LESTER PACKER VOLUME 360. Biophotonics (Part A) Edited by GERARD MARRIOTT AND IAN PARKER VOLUME 361. Biophotonics (Part B) Edited by GERARD MARRIOTT AND IAN PARKER

xxxviii

Methods in Enzymology

VOLUME 362. Recognition of Carbohydrates in Biological Systems (Part A) Edited by YUAN C. LEE AND REIKO T. LEE VOLUME 363. Recognition of Carbohydrates in Biological Systems (Part B) Edited by YUAN C. LEE AND REIKO T. LEE VOLUME 364. Nuclear Receptors Edited by DAVID W. RUSSELL AND DAVID J. MANGELSDORF VOLUME 365. Differentiation of Embryonic Stem Cells Edited by PAUL M. WASSAUMAN AND GORDON M. KELLER VOLUME 366. Protein Phosphatases Edited by SUSANNE KLUMPP AND JOSEF KRIEGLSTEIN VOLUME 367. Liposomes (Part A) Edited by NEJAT DU¨ZGU¨NES¸ VOLUME 368. Macromolecular Crystallography (Part C) Edited by CHARLES W. CARTER, JR., AND ROBERT M. SWEET VOLUME 369. Combinational Chemistry (Part B) Edited by GUILLERMO A. MORALES AND BARRY A. BUNIN VOLUME 370. RNA Polymerases and Associated Factors (Part C) Edited by SANKAR L. ADHYA AND SUSAN GARGES VOLUME 371. RNA Polymerases and Associated Factors (Part D) Edited by SANKAR L. ADHYA AND SUSAN GARGES VOLUME 372. Liposomes (Part B) Edited by NEJAT DU¨ZGU¨NES¸ VOLUME 373. Liposomes (Part C) Edited by NEJAT DU¨ZGU¨NES¸ VOLUME 374. Macromolecular Crystallography (Part D) Edited by CHARLES W. CARTER, JR., AND ROBERT W. SWEET VOLUME 375. Chromatin and Chromatin Remodeling Enzymes (Part A) Edited by C. DAVID ALLIS AND CARL WU VOLUME 376. Chromatin and Chromatin Remodeling Enzymes (Part B) Edited by C. DAVID ALLIS AND CARL WU VOLUME 377. Chromatin and Chromatin Remodeling Enzymes (Part C) Edited by C. DAVID ALLIS AND CARL WU VOLUME 378. Quinones and Quinone Enzymes (Part A) Edited by HELMUT SIES AND LESTER PACKER VOLUME 379. Energetics of Biological Macromolecules (Part D) Edited by JO M. HOLT, MICHAEL L. JOHNSON, AND GARY K. ACKERS VOLUME 380. Energetics of Biological Macromolecules (Part E) Edited by JO M. HOLT, MICHAEL L. JOHNSON, AND GARY K. ACKERS

Methods in Enzymology

VOLUME 381. Oxygen Sensing Edited by CHANDAN K. SEN AND GREGG L. SEMENZA VOLUME 382. Quinones and Quinone Enzymes (Part B) Edited by HELMUT SIES AND LESTER PACKER VOLUME 383. Numerical Computer Methods (Part D) Edited by LUDWIG BRAND AND MICHAEL L. JOHNSON VOLUME 384. Numerical Computer Methods (Part E) Edited by LUDWIG BRAND AND MICHAEL L. JOHNSON VOLUME 385. Imaging in Biological Research (Part A) Edited by P. MICHAEL CONN VOLUME 386. Imaging in Biological Research (Part B) Edited by P. MICHAEL CONN VOLUME 387. Liposomes (Part D) Edited by NEJAT DU¨ZGU¨NES¸ VOLUME 388. Protein Engineering Edited by DAN E. ROBERTSON AND JOSEPH P. NOEL VOLUME 389. Regulators of G-Protein Signaling (Part A) Edited by DAVID P. SIDEROVSKI VOLUME 390. Regulators of G-Protein Signaling (Part B) Edited by DAVID P. SIDEROVSKI VOLUME 391. Liposomes (Part E) Edited by NEJAT DU¨ZGU¨NES¸ VOLUME 392. RNA Interference Edited by ENGELKE ROSSI VOLUME 393. Circadian Rhythms Edited by MICHAEL W. YOUNG VOLUME 394. Nuclear Magnetic Resonance of Biological Macromolecules (Part C) Edited by THOMAS L. JAMES VOLUME 395. Producing the Biochemical Data (Part B) Edited by ELIZABETH A. ZIMMER AND ERIC H. ROALSON VOLUME 396. Nitric Oxide (Part E) Edited by LESTER PACKER AND ENRIQUE CADENAS VOLUME 397. Environmental Microbiology Edited by JARED R. LEADBETTER VOLUME 398. Ubiquitin and Protein Degradation (Part A) Edited by RAYMOND J. DESHAIES

xxxix

xl

Methods in Enzymology

VOLUME 399. Ubiquitin and Protein Degradation (Part B) Edited by RAYMOND J. DESHAIES VOLUME 400. Phase II Conjugation Enzymes and Transport Systems Edited by HELMUT SIES AND LESTER PACKER VOLUME 401. Glutathione Transferases and Gamma Glutamyl Transpeptidases Edited by HELMUT SIES AND LESTER PACKER VOLUME 402. Biological Mass Spectrometry Edited by A. L. BURLINGAME VOLUME 403. GTPases Regulating Membrane Targeting and Fusion Edited by WILLIAM E. BALCH, CHANNING J. DER, AND ALAN HALL VOLUME 404. GTPases Regulating Membrane Dynamics Edited by WILLIAM E. BALCH, CHANNING J. DER, AND ALAN HALL VOLUME 405. Mass Spectrometry: Modified Proteins and Glycoconjugates Edited by A. L. BURLINGAME VOLUME 406. Regulators and Effectors of Small GTPases: Rho Family Edited by WILLIAM E. BALCH, CHANNING J. DER, AND ALAN HALL VOLUME 407. Regulators and Effectors of Small GTPases: Ras Family Edited by WILLIAM E. BALCH, CHANNING J. DER, AND ALAN HALL VOLUME 408. DNA Repair (Part A) Edited by JUDITH L. CAMPBELL AND PAUL MODRICH VOLUME 409. DNA Repair (Part B) Edited by JUDITH L. CAMPBELL AND PAUL MODRICH VOLUME 410. DNA Microarrays (Part A: Array Platforms and Web-Bench Protocols) Edited by ALAN KIMMEL AND BRIAN OLIVER VOLUME 411. DNA Microarrays (Part B: Databases and Statistics) Edited by ALAN KIMMEL AND BRIAN OLIVER VOLUME 412. Amyloid, Prions, and Other Protein Aggregates (Part B) Edited by INDU KHETERPAL AND RONALD WETZEL VOLUME 413. Amyloid, Prions, and Other Protein Aggregates (Part C) Edited by INDU KHETERPAL AND RONALD WETZEL VOLUME 414. Measuring Biological Responses with Automated Microscopy Edited by JAMES INGLESE VOLUME 415. Glycobiology Edited by MINORU FUKUDA VOLUME 416. Glycomics Edited by MINORU FUKUDA

Methods in Enzymology

xli

VOLUME 417. Functional Glycomics Edited by MINORU FUKUDA VOLUME 418. Embryonic Stem Cells Edited by IRINA KLIMANSKAYA AND ROBERT LANZA VOLUME 419. Adult Stem Cells Edited by IRINA KLIMANSKAYA AND ROBERT LANZA VOLUME 420. Stem Cell Tools and Other Experimental Protocols Edited by IRINA KLIMANSKAYA AND ROBERT LANZA VOLUME 421. Advanced Bacterial Genetics: Use of Transposons and Phage for Genomic Engineering Edited by KELLY T. HUGHES VOLUME 422. Two-Component Signaling Systems, Part A Edited by MELVIN I. SIMON, BRIAN R. CRANE, AND ALEXANDRINE CRANE VOLUME 423. Two-Component Signaling Systems, Part B Edited by MELVIN I. SIMON, BRIAN R. CRANE, AND ALEXANDRINE CRANE VOLUME 424. RNA Editing Edited by JONATHA M. GOTT VOLUME 425. RNA Modification Edited by JONATHA M. GOTT VOLUME 426. Integrins Edited by DAVID CHERESH VOLUME 427. MicroRNA Methods Edited by JOHN J. ROSSI VOLUME 428. Osmosensing and Osmosignaling Edited by HELMUT SIES AND DIETER HAUSSINGER VOLUME 429. Translation Initiation: Extract Systems and Molecular Genetics Edited by JON LORSCH VOLUME 430. Translation Initiation: Reconstituted Systems and Biophysical Methods Edited by JON LORSCH VOLUME 431. Translation Initiation: Cell Biology, High-Throughput and Chemical-Based Approaches Edited by JON LORSCH VOLUME 432. Lipidomics and Bioactive Lipids: Mass-Spectrometry–Based Lipid Analysis Edited by H. ALEX BROWN

xlii

Methods in Enzymology

VOLUME 433. Lipidomics and Bioactive Lipids: Specialized Analytical Methods and Lipids in Disease Edited by H. ALEX BROWN VOLUME 434. Lipidomics and Bioactive Lipids: Lipids and Cell Signaling Edited by H. ALEX BROWN VOLUME 435. Oxygen Biology and Hypoxia Edited by HELMUT SIES AND BERNHARD BRU¨NE VOLUME 436. Globins and Other Nitric Oxide-Reactive Protiens (Part A) Edited by ROBERT K. POOLE VOLUME 437. Globins and Other Nitric Oxide-Reactive Protiens (Part B) Edited by ROBERT K. POOLE VOLUME 438. Small GTPases in Disease (Part A) Edited by WILLIAM E. BALCH, CHANNING J. DER, AND ALAN HALL VOLUME 439. Small GTPases in Disease (Part B) Edited by WILLIAM E. BALCH, CHANNING J. DER, AND ALAN HALL VOLUME 440. Nitric Oxide, Part F Oxidative and Nitrosative Stress in Redox Regulation of Cell Signaling Edited by ENRIQUE CADENAS AND LESTER PACKER VOLUME 441. Nitric Oxide, Part G Oxidative and Nitrosative Stress in Redox Regulation of Cell Signaling Edited by ENRIQUE CADENAS AND LESTER PACKER VOLUME 442. Programmed Cell Death, General Principles for Studying Cell Death (Part A) Edited by ROYA KHOSRAVI-FAR, ZAHRA ZAKERI, RICHARD A. LOCKSHIN, AND MAURO PIACENTINI VOLUME 443. Angiogenesis: In Vitro Systems Edited by DAVID A. CHERESH VOLUME 444. Angiogenesis: In Vivo Systems (Part A) Edited by DAVID A. CHERESH VOLUME 445. Angiogenesis: In Vivo Systems (Part B) Edited by DAVID A. CHERESH VOLUME 446. Programmed Cell Death, The Biology and Therapeutic Implications of Cell Death (Part B) Edited by ROYA KHOSRAVI-FAR, ZAHRA ZAKERI, RICHARD A. LOCKSHIN, AND MAURO PIACENTINI VOLUME 447. RNA Turnover in Bacteria, Archaea and Organelles Edited by LYNNE E. MAQUAT AND CECILIA M. ARRAIANO

Methods in Enzymology

xliii

VOLUME 448. RNA Turnover in Eukaryotes: Nucleases, Pathways and Analysis of mRNA Decay Edited by LYNNE E. MAQUAT AND MEGERDITCH KILEDJIAN VOLUME 449. RNA Turnover in Eukaryotes: Analysis of Specialized and Quality Control RNA Decay Pathways Edited by LYNNE E. MAQUAT AND MEGERDITCH KILEDJIAN VOLUME 450. Fluorescence Spectroscopy Edited by LUDWIG BRAND AND MICHAEL L. JOHNSON VOLUME 451. Autophagy: Lower Eukaryotes and Non-Mammalian Systems (Part A) Edited by DANIEL J. KLIONSKY VOLUME 452. Autophagy in Mammalian Systems (Part B) Edited by DANIEL J. KLIONSKY VOLUME 453. Autophagy in Disease and Clinical Applications (Part C) Edited by DANIEL J. KLIONSKY VOLUME 454. Computer Methods (Part A) Edited by MICHAEL L. JOHNSON AND LUDWIG BRAND VOLUME 455. Biothermodynamics (Part A) Edited by MICHAEL L. JOHNSON, JO M. HOLT, AND GARY K. ACKERS (RETIRED) VOLUME 456. Mitochondrial Function, Part A: Mitochondrial Electron Transport Complexes and Reactive Oxygen Species Edited by WILLIAM S. ALLISON AND IMMO E. SCHEFFLER VOLUME 457. Mitochondrial Function, Part B: Mitochondrial Protein Kinases, Protein Phosphatases and Mitochondrial Diseases Edited by WILLIAM S. ALLISON AND ANNE N. MURPHY VOLUME 458. Complex Enzymes in Microbial Natural Product Biosynthesis, Part A: Overview Articles and Peptides Edited by DAVID A. HOPWOOD VOLUME 459. Complex Enzymes in Microbial Natural Product Biosynthesis, Part B: Polyketides, Aminocoumarins and Carbohydrates Edited by DAVID A. HOPWOOD VOLUME 460. Chemokines, Part A Edited by TRACY M. HANDEL AND DAMON J. HAMEL VOLUME 461. Chemokines, Part B Edited by TRACY M. HANDEL AND DAMON J. HAMEL VOLUME 462. Non-Natural Amino Acids Edited by TOM W. MUIR AND JOHN N. ABELSON VOLUME 463. Guide to Protein Purification, 2nd Edition Edited by RICHARD R. BURGESS AND MURRAY P. DEUTSCHER

xliv

Methods in Enzymology

VOLUME 464. Liposomes, Part F Edited by NEJAT DU¨ZGU¨NES¸ VOLUME 465. Liposomes, Part G Edited by NEJAT DU¨ZGU¨NES¸ VOLUME 466. Biothermodynamics, Part B Edited by MICHAEL L. JOHNSON, GARY K. ACKERS, AND JO M. HOLT VOLUME 467. Computer Methods Part B Edited by MICHAEL L. JOHNSON AND LUDWIG BRAND VOLUME 468. Biophysical, Chemical, and Functional Probes of RNA Structure, Interactions and Folding: Part A Edited by DANIEL HERSCHLAG VOLUME 469. Biophysical, Chemical, and Functional Probes of RNA Structure, Interactions and Folding: Part B Edited by DANIEL HERSCHLAG VOLUME 470. Guide to Yeast Genetics: Functional Genomics, Proteomics, and Other Systems Analysis, 2nd Edition Edited by GERALD FINK, JONATHAN WEISSMAN, AND CHRISTINE GUTHRIE VOLUME 471. Two-Component Signaling Systems, Part C Edited by MELVIN I. SIMON, BRIAN R. CRANE, AND ALEXANDRINE CRANE VOLUME 472. Single Molecule Tools, Part A: Fluorescence Based Approaches Edited by NILS WALTER VOLUME 473. Thiol Redox Transitions in Cell Signaling, Part A Chemistry and Biochemistry of Low Molecular Weight and Protein Thiols Edited by ENRIQUE CADENAS AND LESTER PACKER VOLUME 474. Thiol Redox Transitions in Cell Signaling, Part B Cellular Localization and Signaling Edited by ENRIQUE CADENAS AND LESTER PACKER VOLUME 475. Single Molecule Tools, Part B: Super-Resolution, Particle Tracking, Multiparameter, and Force Based Methods Edited by NILS G. WALTER VOLUME 476. Guide to Techniques in Mouse Development, Part A Mice, Embryos, and Cells, 2nd Edition Edited by PAUL M. WASSARMAN AND PHILIPPE M. SORIANO VOLUME 477. Guide to Techniques in Mouse Development, Part B Mouse Molecular Genetics, 2nd Edition Edited by PAUL M. WASSARMAN AND PHILIPPE SORIANO VOLUME 478. Glycomics, Part A Edited by MINORU FUKUDA

Methods in Enzymology

VOLUME 479. Functional Glycomics Edited by MINORU FUKUDA VOLUME 480. Glycobiology Edited by MINORU FUKUDA VOLUME 481. Cryo-EM, Part A Sample Preparation and Data Collection Edited by GRANT J. JENSEN VOLUME 482. Cryo-EM, Part B 3-D Reconstruction Edited by GRANT J. JENSEN VOLUME 483. Cryo-EM, Part C Analyses, Interpretation, and Case Studies Edited by GRANT J. JENSEN

xlv

C H A P T E R

O N E

Analyses of Subnanometer Resolution Cryo-EM Density Maps Matthew L. Baker,* Mariah R. Baker,* Corey F. Hryc,* and Frank DiMaio† Contents 2 3 4 4 7 7 8 8 9 10 11 11 13 13 14 14 14 16 16 16 17 17 17 18 18 18 19

1. Introduction 2. Features in a Subnanometer Resolution Density Map 3. Tools: Analyzing a Subnanometer Resolution Density Map 3.1. Fitting atomic models 3.2. Constrained modeling with cryo-EM density 3.3. Extracting protein subunits from a density map 3.4. Secondary structure identification 3.5. De novo modeling 4. Protocol: From Density Map to Atomic Model 4.1. Segmentation 4.2. Identifying secondary structure elements 4.3. Secondary structure annotation 4.4. Structural homologues from sequence 4.5. Identifying structural homologues from SSEs 4.6. Fitting atomic models in cryo-EM density maps 4.7. Predicting SSEs from sequence 4.8. SSE correspondence 4.9. Ca placement 4.10. Assigning Ca positions in helices 4.11. Assigning Ca positions in sheets and loops 4.12. Fixing an atomic model 4.13. Ca optimization 4.14. Building a macromolecular model 4.15. Map rescaling 4.16. Ca to atomic model 4.17. Model optimization 4.18. Monitoring model quality

* National Center for Macromolecular Imaging, Verna and Marrs McLean Department of Biochemistry and Molecular Biology, Baylor College of Medicine, Houston, Texas, USA Department of Biochemistry, University of Washington, Seattle, Washington, USA

{

Methods in Enzymology, Volume 483 ISSN 0076-6879, DOI: 10.1016/S0076-6879(10)83001-0

#

2010 Elsevier Inc. All rights reserved.

1

2

Matthew L. Baker et al.

4.19. Ca to atomic model optimization with Rosetta 4.20. Model optimization with Rosetta 5. Case Study: Mm-cpn 6. Discussion Acknowledgments References

19 21 21 24 25 25

Abstract Today, electron cryomicroscopy (cryo-EM) can routinely achieve subnanometer resolutions of complex macromolecular assemblies. From a density map, one can extract key structural and functional information using a variety of computational analysis tools. At subnanometer resolution, these tools make it possible to isolate individual subunits, identify secondary structures, and accurately fit atomic models. With several cryo-EM studies achieving resolutions beyond 5 A˚, computational modeling and feature recognition tools have been employed to construct backbone and atomic models of the protein components directly from a density map. In this chapter, we describe several common classes of computational tools that can be used to analyze and model subnanometer resolution reconstructions from cryo-EM. A general protocol for analyzing subnanometer resolution density maps is presented along with a full description of steps used in analyzing the 4.3 A˚ resolution structure of Mm-cpn.

1. Introduction Electron microscopy has played an increasingly important role in understanding the structure and function of macromolecular assemblies that contribute to numerous biological processes. It offers an advantage over other structural techniques, like X-ray crystallography, by imaging macromolecular assemblies in near-native conditions (Baumeister and Steven, 2000; Frank, 2002). Even at non-atomic resolutions, three-dimensional (3D) reconstructions (volumetric density maps) from electron microscopy can describe the size, shape, and composition of a macromolecular assembly. In 1975, the first subnanometer resolution electron microscopy data was derived from regularly arrayed 2D crystals of bacteriorhodopsin by Henderson and Unwin (1975). At 7 A˚ resolution, the seven transmembrane helices of bacteriorhodopsin were clearly visible. Fifteen years later, Henderson et al. (1990) reported the first atomic model constructed directly from an electron microscopy density map. Another milestone was achieved in 2005, as water molecules were clearly resolved in the 1.9 A˚ resolution structure of aquaporin-0 (Gonen et al., 2005). In single-particle electron cryomicroscopy (cryo-EM), a macromolecular assembly does not need to form a regular array. Rather, images of randomly orientated particles are processed to generate a 3D density map

Cryo-M Map Analysis

3

(Cong and Ludtke, 2010). Early 3D reconstructions achieved relatively low resolutions due to several factors including small data sets, sample heterogeneity, and technical limitations of the microscopes. In 1997, a significant milestone was achieved in single-particle cryo-EM; two reconstructions of ˚ resoluthe Hepatitis B virus obtained subnanometer resolutions (9 and 7.4 A tion; Bo¨ttcher et al., 1997; Conway et al., 1997). From these density maps, it was possible for the first time to clearly identify the core capsid protein and visualize rod-like structures corresponding to a-helices. Using the connectivity of the helices, the overall fold of the 183 amino acid core capsid protein was proposed. Several years later, a reconstruction of the rice dwarf virus ˚ resolution) was the first cryo-EM reconstruction to clearly resolve (6.8 A b-sheets as flat planes of density (Zhou et al., 2001). Like the density maps from the Hepatitis B reconstructions, this resolution allowed for the description of the protein fold, though no models were constructed. Subsequent atomic models confirmed the proposed folds for both Hepatitis B (Wynne et al., 1999) and rice dwarf virus (Nakagawa et al., 2003). Today, technical advances in computing, specimen preparation and data acquisition have made it possible to routinely achieve subnanometer resolutions on a wide variety of specimens with single-particle cryo-EM (Chiu et al., 2005). A decade after the first subnanometer resolution cryo-EM structures, another milestone was reached; the first near-atomic resolution structures were produced by single-particle cryo-EM. In reconstructions of rotavirus ˚ ; Zhang et al., 2008), GroEL (4.2 A ˚ ; Ludtke et al., 2008), cytoplasmic (3.88 A ˚ ˚; polyhedrosis virus (4.0 A; Yu et al., 2008), and bacteriophage e15 (4.5 A Jiang et al., 2008), the pitch of a-helices and the separation of b-strands were visualized. Though these structures did not have the resolution to utilize standard X-ray crystallographic tools for model construction (typically ˚ resolution), de novo Ca backbone models were built starting at 3.5 A from the cryo-EM density maps of cytoplasmic polyhedrosis virus, GroEL, and bacteriophage e15 using a combination of computational and geometric tools (Baker et al., 2007, Ju et al., 2007). The de novo models built directly from these density maps relied almost entirely on visual interpretation of the density and manual structure assignment. Several recent state-of-the-art reconstructions have now resolved sidechain densities and allowed for the construction of complete atomic models directly from the density map (Cong et al., 2010; Yu et al., 2008; Zhang et al., 2010).

2. Features in a Subnanometer Resolution Density Map Regardless of the resolution, computational tools are critical in analyzing, interpreting, and annotating structural information in cryo-EM density maps. As such, a number of specialized computational tools have been

4

Matthew L. Baker et al.

developed. However, before describing these tools and the general protocol for analyzing a cryo-EM density map, it is important to establish an understanding of the features visible in a density map as a function of resolution (Fig. 1.1). Apparent in many of the density maps deposited in the EMDB (http://emdatabank.org), subnanometer resolution structures have distinct boundaries that allow for subunits and individual domains to be identified and segmented from the entire map (Fig. 1.1A). These well-defined densities help to accurately fit known structural models to a density map (Fig. 1.1B). a-Helices appear as long rod-like densities, and b-sheets appear ˚ resolution (Fig. 1.1C). At slightly higher as thin, continuous planes at 8 A resolutions, connectivity between the secondary structure elements (SSEs) becomes evident (Fig. 1.1D). a-Helices begin to develop features and the ˚ resolution. Beyond 4.5 A ˚ resolution, the pitch becomes visible at 5 A thin, flat planes of b-sheets become broken, as individual strands are ˚ resolution, sidechain densities become recogresolved (Fig. 1.1E). By 4 A nizable (Fig. 1.1F) and a relatively unambiguous trace of a protein backbone can be seen (Fig. 1.1G).

3. Tools: Analyzing a Subnanometer Resolution Density Map While subnanometer resolutions span a wide range of detectable features, tools for analyzing their structure can be grouped into two classes: feature recognition and data integration. In principle, many of these tools are not restricted to subnanometer resolutions, although higher resolution features do provide significant advantages in validation. Feature recognition tools, however, are generally resolution-specific. The following briefly describes - some “standard” tools and techniques for analyzing subnanometer resolution cryo-EM density maps (Table 1.1).

3.1. Fitting atomic models Perhaps the most common method for analyzing a subnanometer resolution density map is fitting a known atomic model into a density map. There are a number of different approaches to fit atomic models to the density map, though an exhaustive rotational and translational search of the model within the map is generally used (reviewed in Rossmann et al., 2005). “Rigidbody” fitting attempts to identify the maximum overlap of the model with the density and minimize the amount of model unaccounted for within the density. Various fitting programs report scores differently, so it is important to consult the program documentation, as well as visually inspect the fit of a model to the density map. Aside from program-specific fitting scores, tools

B A

C

F

D

E

G

Figure 1.1 Features at subnanometer resolutions. A gallery of structural features from ˚ cryo-EM reconstructions is shown. (A) Domains in the clamp region of the 9.5 A resolution reconstruction of RyR1 can be observed (Serysheva et al., 2008). (B) The atomic models of VP5* (lower left) and VP8* (upper right) are fit to the density map ˚ resolution structure of rotavirus (Li et al., corresponding to the VP4 spikes in the 9.5 A 2009). (C) At slightly higher resolutions, secondary structures (a-helices are depicted as cylinders and b-sheets are depicted as planes) can be clearly seen in the capsid proteins ˚ resolution (Zhou et al., 2001). (D) Around this resolution, of rice dwarf virus at 6.8 A possible connections between secondary structure elements can be identified computa˚ resolution rice tionally using density skeletonization (red), again as seen in the 6.8 A dwarf virus capsid protein. (E) Increasing resolution reveals the separation of b-strands ˚ resolution (Ludtke et al., 2008). (F) Large, bulky sidechains begin to in GroEL at 4.2 A ˚ resolution (Cong et al., 2010). (G) An unambigappear in TriC reconstruction at 4.0 A ˚ resolution (Zhang et al., 2008). uous backbone is apparent in VP6 of rotavirus at 3.8 A

6

Matthew L. Baker et al.

Table 1.1 Programs for analyzing subnanometer resolution cryo-EM density maps

Segmentation

Fitting atomic models

Secondary structure identification

Modeling

Visualization

Manual Amira (Visage Imaging, GmbH) Avizo (VSG, France) Chimera (Pettersen et al., 2004) Automatic CoDiv (Volkmann, 2002) EMAN (Ludtke et al., 1999; Tang et al., 2007) Segger (Pintilie et al., 2010) VolRover (Baker et al., 2006b) Rigid body Chimera (Pettersen et al., 2004) CoFi (Volkmann and Hanein, 1999) Coot (Emsley and Cowtan, 2004) DockEM (Roseman, 2000) EMFit (Rossmann, 2000) Foldhunter ( Jiang et al., 2001) Mod-EM (Topf et al., 2005) O ( Jones et al., 1991) Situs (Wriggers et al., 1999) UROX (Siebert and Navaza, 2009) Flexible DireX (Schro¨der et al., 2007) Flex-EM (Topf et al., 2008) MDFF (Trabuco et al., 2009) NMFF (Tama et al., 2004) NORMA (Suhre et al., 2006) Yup.scx (Tan et al., 2008) Situs (Rusu et al., 2008) Validation FH-stat (Serysheva et al., 2005) Helixhunter ( Jiang et al., 2001) Sheetminer/Sheetracer (Kong and Ma, 2003; Kong et al., 2004) SSEHunter (Baker et al., 2007) EM-IMO (Zhu et al., 2010) Gorgon (http://gorgon.wustl.edu) Modeller (Topf et al., 2005, 2006) Rosetta (Baker et al., 2006a; DiMaio et al., 2009) Amira (Visage Imaging, GmbH) Avizo (VSG, France) Chimera (Pettersen et al., 2004) PyMol (DeLano Scientific LLC, USA) VMD (Humphrey et al., 1996)

Common computational tools used in the analysis of subnanometer resolution cryo-EM density maps are listed. Numerous other algorithms have also been published, though only currently downloadable tools are listed.

Cryo-M Map Analysis

7

are available for independently reporting the quality of fit for a model within a density map (Serysheva et al., 2005; Volkmann, 2009). In addition to fitting an atomic model as a “rigid-body” within the density map, atomic models can be morphed, or flexibly fit, into the density map (Schro¨der et al., 2007; Suhre et al., 2006; Tama et al., 2004; Tan et al., 2008; Topf et al., 2008; Trabuco et al., 2009). In these types of programs an atomic model is allowed to relax or bend at certain points to better fit the density map. This is particularly useful when fitting structures of different conformations or homologous structures.

3.2. Constrained modeling with cryo-EM density If a structure for one or more of the protein subunits in the assembly is not known, computational modeling approaches can be used to generate homology models or ab initio models for proteins. In these cases, the cryoEM density map can be used directly to facilitate the construction and evaluation of a protein/domain structural model (Baker et al., 2006a; DiMaio et al., 2009; Topf et al., 2005, 2006; Zhu et al., 2010). In a constrained comparative modeling approach, an initial sequence–structure alignment is allowed to evolve, simultaneously improving the homology model and its fit to the density map (Topf et al., 2005, 2006). This type of approach can also be used to improve local regions of a model within the density map. When a template structure is not known, constrained ab initio modeling may be used to build domains or small proteins (Baker et al., 2006a). In this approach, a gallery of models are built computationally and then later evaluated by their fit to the density map. The resolution of the density map is key in determining the accuracy of the models (Topf et al., 2005). While not restricted to subnanometer resolutions, models built within a subnanometer resolution density map will likely have the correct fold, though atom placement may only be approximate.

3.3. Extracting protein subunits from a density map Like fitting atomic models to a density map, density segmentation, the process of identifying and isolating a single protein or domain from the cryo-EM density map, is not exclusive to subnanometer resolutions. However, at subnanometer resolutions, cryo-EM density maps move from being “blob-like” to having distinct features, including sharp drop-offs in the density that characterizes the edges and boundaries between subunits (Chiu et al., 2005). The actual process of segmenting a subunit can be done using interactive tools commonly found in visualization programs such as Chimera (Pettersen et al., 2004), Amira (Visage Imaging, GmbH), and Avizo (VSG, France) or computational techniques such as those based on the watershed transform (Ludtke et al., 1999; Pintilie et al., 2010; Volkmann, 2002).

8

Matthew L. Baker et al.

3.4. Secondary structure identification In cryo-EM, feature recognition tools have been typically used to identify SSEs within a subnanometer resolution density map. At subnanometer resolutions, helices are clearly resolved as long cylinders with relatively high density. Helixhunter ( Jiang et al., 2001) was the first tool designed to computationally detect a-helices in a density map using a simple crosscorrelation search with a prototypical helix template over three translational and two rotational degrees of freedom. b-Sheets, which resemble thin planes, are more diverse in their structure and not amenable to correlation-based approaches. Rather, morphological analysis of the density was used to first localize b-sheets and strands in subnanometer resolution density maps (Kong and Ma, 2003; Kong et al., 2004). Later developments lead to SSEHunter, a single tool to detect both a-helices and b-sheets (Baker et al., 2007). With SSEHunter, the helix correlation routine is paired with a local geometry analysis and a density skeleton to detect SSEs. With any of these feature recognition tools, it is advisable to operate on segmented density maps rather than the entire map when detecting SSEs.

3.5. De novo modeling In conjunction with development of SSEHunter, a new density skeletonization routine ( Ju et al., 2007) was developed that preserves both the features and the topology in a density map. This density skeleton provided clear connections between observable SSEs. Coupling density skeletonization with the aforementioned feature recognition tools has given rise to a new protocol for de novo structural modeling ( Jiang et al., 2008; Ludtke et al., 2008). The de novo method attempts to construct a model directly from a density map without the aid of an existing structural template. Borrowing methods for constructing atomic models from X-ray crystallographic density maps, SSEs in the sequence and the density map are correlated, providing initiation points for model construction (Abeysinghe et al., 2008). This sequence-to-structure correspondence assigns residue positions to SSEs in the map. Once SSE anchor points are established, models can be constructed using various X-ray crystallographic model building toolkits, such as O ( Jones et al., 1991) and Coot (Emsley and Cowtan, 2004), both of which offer tools for placing atoms within a cryo-EM density map. Gorgon (http://gorgon.wustl.edu), a molecular modeling toolkit tailored to subnanometer resolution cryo-EM density maps, provides a comprehensive suite of utilities to analyze subnanometer resolution density maps, including the ability to generate secondary structure assignments, a sequence-to-structure correspondence routine and Ca model construction utilities.

9

Cryo-M Map Analysis

4. Protocol: From Density Map to Atomic Model With an extensive software library (Table 1.1), analyzing a subnanometer resolution density map can be an overwhelming task. Simply identifying the proper tools and understanding the expected results can be difficult. Figure 1.2 provides a simplified flowchart describing the assorted

Subnanometer resolution density map

Segment density map

Segmented subunit

Identify SSE

Homologous structure

Yes

Known or related atomic model?

No

No model

Yes Known atomic model

Sequence/structure alignment

Secondary structure analysis

SSE model

No Construct homology model

Fit atomic model

Constrained ab initio model

Yes

Small, single domain protein?

Structural model?

Yes

Fit atomic model

No Homologous model

No

Suitable SSE correspondence?

Agrees with density and SSE?

Dejavu SSE search

Fold model

Yes

Yes Flexible fitting

No

Preliminary model

Manual refinement of model

Optimize model

Construct atomic model

Near-atomic resolution?

No

Calculate SSE connectivities

Yes

Yes Visible sidechain density

De novo modeling

Topology model

No Final atomic model

Backbone model

Figure 1.2 Analyzing a subnanometer resolution cryo-EM density map. A general scheme for analyzing subnanometer resolution cryo-EM density maps is depicted. Different projects may take advantage of additional information during the analysis process and thus deviate from the overall scheme.

10

Matthew L. Baker et al.

paths and decisions one may encounter in analyzing a subnanometer resolution density map. In the following sections, we describe a general protocol (based on Fig. 1.2) for analyzing a subnanometer resolution density map. In a subsequent section, we also detail the analysis of the 4.3-A˚ resolution structure of Methanococcus maripaludis chaperonin (Mm-cpn), a group II chaperonin. It should be noted that this protocol is not restricted to single-particle cryo-EM and can be used to analyze any type of subnanometer resolution density map. The steps described in this protocol illustrate a complete pathway for constructing an atomic model from a cryo-EM density map; however, most of the steps can be used independently depending on numerous constraints, such as resolution and available structural information. As such, consulting the various user-guides and documentation for the individual software packages will help the user to optimally select parameters for each particular project. In this protocol, several image processing, visualization and modeling software packages are utilized, including Coot (Emsley and Cowtan, 2004), EMAN/EMAN2 (Ludtke et al., 2005; Tang et al., 2007), Gorgon (http:// gorgon.wustl.edu), Rosetta (Bradley et al., 2005), and UCSF Chimera (Pettersen et al., 2004). Other programs may be substituted, though their exact usage, including inputs and outputs, will vary. Additionally, the amount of time and computational resources required by each of these tools will vary considerably based on the project and experience of the user.

4.1. Segmentation To begin modeling an individual protein subunit, one subunit/domain must be extracted from the macromolecular assembly. Generally, individual protein densities within subnanometer resolution reconstructions have relatively sharp fall-offs at their boundaries. Identifying the boundaries between subunits can be accomplished using a variety of different approaches. Manual segmentation requires the user to visually identify and demarcate the subunits. By adjusting the isosurface display threshold in visualization programs, such as Chimera and Gorgon, these boundaries can usually be seen clearly. The user can then create a mask as either a set of 2D slices or a single 3D volume, or “erase” spurious density to produce an initial segmentation of a protein subunit. The initial segmentation is usually relatively crude to avoid removing too much density but still adequate for analysis. Once all subunits have been segmented, the user can then check to see that all density has been assigned a segment and no overlapping segments are present. This may require several iterations to improve the initial segmentation. Alternatively, automated approaches like EMAN’s segment3d (Ludtke et al., 1999) and Segger (Pintilie et al., 2010)

Cryo-M Map Analysis

11

will cluster every voxel in a density map based on a set of user-provided parameters (e.g., number of subunits, symmetry, etc.). Visual inspection and manual adjustment of the subunits may be required.

4.2. Identifying secondary structure elements Once the map has been segmented, SSEs are then identified using SSEHunter (Baker et al., 2007). This step requires a cubic density map with an even number of voxels. Additionally, to maximize compatibility with other modeling programs, the origin of the segmented subunit should be reset to zero. To accomplish this, proc3d from EMAN or e2proc3d.py from EMAN2 can be used with the “clip” and “origin” options. SSEHunter can be executed in three different ways: (1) directly from the command line (ssehunter3.py in EMAN and e2ssehunter.py in EMAN2), (2) as a plug-in to UCSF’s Chimera, or (3) within Gorgon. The first two options require the installation of EMAN or EMAN2, while Gorgon’s version of SSEHunter is built-in. In each of these cases, the user is required ˚ /pixel) and a to provide the resolution in A˚, the sampling of the map (A threshold corresponding to the highest isosurface value at which all density of the segmented subunit appears to be connected (obtained visually, 2–4s above the mean density). Map sizes between 483 and 1603 generally require less than 15 min to run on a modern desktop and return a set of pseudoatoms, typically corresponding to 50% of the total number of amino acids represented in the segmented volume. Encoded in the B-factor column of the SSEHunter PDB file are the per-pseudoatom SSEHunter scores, which range between 3 and 3. These values represent the likelihood of a density region to be either a-helix (0 3) or b-sheet (3 0). In Gorgon, the pseudoatoms are automatically colored from blue (3, b-sheet) to white (0) to red, (3, a-helix) where the intensity of the color reflects the score and confidence of the prediction. In addition to these pseudoatoms, a density skeleton is calculated with SSEHunter. As mentioned previously, this skeleton is a simplified geometrical representation of the density map that preserves both features and topology. Gorgon offers an improved threshold-free, grayscale density skeleton that can be substituted place of the SSEHunter skeleton in later steps for identifying the connectivity between SSEs.

4.3. Secondary structure annotation Next, pseudoatoms of similar values are grouped into their respective SSEs using SSEBuilder (Baker et al., 2007; Fig. 1.3). Like SSEHunter, SSEBuilder can be accessed from within Chimera or Gorgon. In SSEBuilder,

12

Matthew L. Baker et al.

Figure 1.3 Secondary structure identification. SSEHunter and SSEBuilder, both EMAN programs, can be run as plug-ins to UCSF’s Chimera. The results for the apical domain of Mm-cpn are shown: red spheres represent helix like regions and the blue spheres represent sheet like regions. Regions of similar scoring pseudoatoms from SSEHunter are grouped and built using SSEBuilder. Helices are depicted as cylinders and sheets are depicted as planes.

pseudoatoms with similar scores are manually grouped into individual SSEs and then automatically constructed as VRML objects. The process continues until all visible SSEs are assigned. When selecting SSEs, generally, the minimum size is three pseudoatoms for an a-helix and five pseudoatoms for a b-sheet. False positives may occur; therefore, only groups of pseudoatoms that resemble SSEs in the density (a-helices appear as long cylinders, b-sheets resemble thin, curved surfaces) should be annotated.

Cryo-M Map Analysis

13

4.4. Structural homologues from sequence Until this point, analysis of the density map has focused on observing features within the density map itself. However, it is possible that an atomic model for one or more of the protein components in the density map is known. This may be a single protein, a domain from the macromolecular assembly or a related structure. While it is relatively straightforward to search and retrieve a protein or domain with a known structure from the Protein Data Bank (http://pdb. org), identifying and building structural models from related structures is more complicated. Homologous structures can be identified using a number of sequence-based search methods including BLAST (Altschul et al., 1990) and FASTA (Lipman and Pearson, 1985). These types of searches generally return a sequence alignment between the sequence of interest and a sequence with a known structure from which a homologous model can be constructed. A number of tools are available for this procedure, ranging from the fairly automated Swiss-Model web service (Arnold et al., 2006) to more flexible modeling suites like Modeller (Sali et al., 1995) and Rosetta (Bradley et al., 2005). In addition to these tools, there are several web-based tools that integrate the search and model building steps including Phyre (Kelley and Sternberg, 2009), Fugue (Shi et al., 2001), and 3D Jigsaw (Bates et al., 2001). Simplifying the process even further, meta-prediction servers, like BioInfoBank (Ginalski et al., 2003), provide a convenient way to submit a sequence to multiple prediction servers and view the results. Regardless of the method or service used, fairly reliable atomic models can be produced when a suitable template structure is identified. As each method reports potential models differently, it is important to consult the documentation for each of the tools in determining the validity of the model.

4.5. Identifying structural homologues from SSEs If structural homologues cannot be identified from the sequence, it is still possible to detect homologues based on the locations and orientations of the SSEs (Baker et al., 2003, 2005). The identified SSEs can be used as inputs into two structural similarity search programs, DejaVu (Kleywegt and Jones, 1997) and COSEC (Mizuguchi and Go, 1995). Results obtained from SSEHunter and SSEBuilder are compatible with either program; an additional format conversion utility is provided with EMAN (dejavu2sse.py). While a sequence-based search will return a sequence alignment, a structure-based search does not. Rather, a structure-based search returns a list of possible folds that match a set of SSEs regardless of sequence.

14

Matthew L. Baker et al.

4.6. Fitting atomic models in cryo-EM density maps Once a suitable atomic model has been found or constructed, the model can be fit into the density map with any of the aforementioned tools. Fitting can be done either as a rigid body or flexibly fit, where the model is allowed to morph to better fit the density. To avoid mis- or overinterpretation, it is important to note the features in the density map. The shape and overall fold are likely to be recognizable in the segmented density map at subnanometer resolutions. More specifically, the atomic model should correspond well to the observed SSEs. Independent validation tools ( fh-stat.py in EMAN; Serysheva et al., 2005) and confidence intervals (Volkmann, 2009) may also be used to assess the statistical likelihood of the model positions in the map.

4.7. Predicting SSEs from sequence When no known or homologous structure is available, it is still possible to construct structural models for individual proteins or subunits de novo. In this case, a combination of secondary structure prediction and feature recognition can assign sequence elements, such as helices and strands, to SSEs identified within the density map, a process termed SSE correspondence (Abeysinghe et al., 2008; Ludtke et al., 2008). Having identified the positions of the SSEs within a segmented density map, the next step is to define the SSEs within the sequence of interest. A number of web-based programs can be used to predict secondary structure including SSPro (Pollastri et al., 2002), JPred (Cole et al., 2008), and PsiPred (McGuffin et al., 2000). These programs generally provide a secondary structure assignment and confidence score to each amino acid. Due to errors in prediction, a consensus alignment built from multiple predictions may be better than the results from a single secondary structure prediction server. For convenience, Gorgon contains a tool that will remotely run the sequence prediction, retrieve the predictions, and format the results for the subsequent SSE correspondence routine.

4.8. SSE correspondence Once SSEs have been identified in both the sequence and density map, a correspondence search can be preformed within Gorgon (Fig. 1.4A). This step provides the initial anchor points to place Ca atoms and construct a protein backbone. Determining the SSE correspondence requires four inputs: helix and sheet locations produced by SSEBuilder, the cryo-EM skeleton from SSEHunter or Gorgon, and the sequence/prediction from the previous step. Once entered, a sequence–structure correspondence is calculated and displayed graphically. In Gorgon, the SSE correspondence results are shown as a list ranked in order from best to worst correspondence.

A

B

Figure 1.4 Model construction with Gorgon. The results from the SSE correspon˚ resolution structure of GroEL in Gorgon dence search on the apical domain of the 4.2-A are shown in (A). Helices are shown as cylinders, while sheets are shown as planes. Potential connectivity is depicted as solid lines. A corresponding color scheme for these elements is shown in the SSE correspondence window on the right. (B) Gorgon contains several methods for assigning atoms to the density in the semi-automated atom placement tool. The atomic editor function illustrates the addition of Ca atoms along a density skeleton (not shown). The user can cycle through the possible locations, select the desired position and proceed to the next residue.

16

Matthew L. Baker et al.

To assess individual correspondences, the lengths of the SSEHunter/SSEBuilder helices are compared to the sequence-predicted helices. Ideally, the lengths of correctly matched SSEs should not differ by more than three amino acids. In a correspondence, one or more SSEs may have been incorrectly assigned. If this occurs, the user may constrain the correct individual SSE correspondences by selecting only correctly paired SSEs and re-running the correspondence search. This process is relatively quick (< 5 s for most cases) and can be done repeatedly until the user is satisfied with the correspondence.

4.9. Ca placement From the previous step, an initial topology for the protein structure is established. Starting from this topology assignment, placement of the Ca backbone atoms begins with helices, followed by sheets and loops. The following steps are described using Gorgon.

4.10. Assigning Ca positions in helices Based on the chosen SSE correspondence, Ca positions of the helices are registered with the corresponding sequence. This process and the following modeling steps can be done with the “Semi-Automatic Atom Placement” tool found in Gorgon. Possible errors in the SSE correspondence may require the user to adjust helix length and directionality when assigning the helix residues. Helices are initially represented as cylinders; manual adjustment of the helix position may be necessary to best fit the density. Bulky sidechains in the helices provide visual cues and help anchor the position and pitch of the helix within the density ( Jiang et al., 2008; Ludtke et al., 2008; Yu et al., 2008). Again, Gorgon contains a variety of mechanisms for optimizing helix position, including an option to show relative sidechain size at Ca atom positions. Once the assignment of helices is complete, strands and loops can be assigned.

4.11. Assigning Ca positions in sheets and loops For the purposes of modeling a Ca backbone, b-strands are treated as loops. Extending from the assigned residues in the a-helices, the remaining residues can be assigned with two unique options found in Gorgon’s “SemiAutomatic Atom Placement” tool. With the “Atomic Editor” the last assigned residue before an unassigned section of sequence is selected. The possible positions for the next unassigned amino acid are shown in the density map along the skeleton for which the Ca–Ca distance is satisfied (Fig. 1.4B). The user then interactively selects a position for the

Cryo-M Map Analysis

17

unassigned amino acid. The model is updated and possible positions for the next amino acid are shown. This process continues until a previously assigned Ca is joined. Alternatively, loops may be assigned using the “Loop Editor.” With the endpoints of the loops selected, the “Loop Editor” allows the user to sketch out the approximate path through the density. Unlike the “Atomic Editor,” no Ca–Ca distance constraints are enforced, though Gorgon provides a convenient way to visualize Ca–Ca distances. Red bond distances are too long, blue bonds are too short, and white bonds are approximately the correct length (3.8 0.5 A˚). To aid in the placement of Ca atoms, the density skeleton, calculated by SSEHunter or in Gorgon, can provide possible paths between the SSEs. Using these two techniques, Ca atoms are placed for any unassigned amino acids.

4.12. Fixing an atomic model In models utilizing a known or homologous structure, adjustment of the Ca positions begins with the SSEs. Entire SSEs are first moved to register with the corresponding features in the density map using modeling programs such as Coot, Gorgon, or Chimera. Once the SSEs of the model have been fit to the density map, the remaining Ca atoms are moved individually to best fit the density. It is possible that some portions of the atomic model structure may be absent. If this is the case, the previous steps can be used to build any missing residues before proceeding.

4.13. Ca optimization After all Ca atoms have been assigned, the next step is to adjust the atom positions to optimally fit the density while maintaining reasonable Ca–Ca ˚ ) and angles (60–120 between three consecubond distances ( 3.8 A tive Cas), proper secondary structure features and no atom/bond clashes. Optimization of Ca positions begins with helices; the pitch of the model helices should register well with the pitch of the helix observed in the density. In the event that b-strands are resolved, the distance between ˚ . Once Ca neighboring strand Ca atoms should be between 4.5 and 5 A positions in SSEs have been optimized Ca positions in the loops can be adjusted.

4.14. Building a macromolecular model Following the initial construction of a Ca model, the models are then placed back into the context of the full cryo-EM density map with all other models. As previously mentioned, a number of fitting routines may be used for this step.

18

Matthew L. Baker et al.

At this point, the Ca models are assessed by visually examining how well they fit into the density map. A “good model” will occupy the entire subunit density map and have no clashes with neighboring subunits. Residues in neighboring subunit models should also be readjusted so that the minimum ˚ . The last two steps will likely be distance between any Ca atoms is 4.5 A iterated multiple times to improve the model until no inter- and intrasubunit clashes are evident and the models account for all the density in the reconstruction. The final refined backbone model is then saved as a PDB file.

4.15. Map rescaling To enhance the high-resolution information in the density maps, such as sidechains, for subsequent map building, the density map can be rescaled in Fourier space using the Ca backbone model (Ferna´ndez et al., 2008; Zhang et al., 2010). The model is first blurred to approximately the same resolution as the original density map. Structure factors can be calculated from this blurred map and applied to the original density map in EMAN (proc3d or e2proc3d.py in EMAN2). A low-pass filter is then applied to the rescaled map at the calculated resolution and normalized so that the positive density values are between 0 and 1.

4.16. Ca to atomic model An approximate atomic model can be reconstructed from the Ca model using a Ca-to-backbone builder such as the web-based SABBAC (Maupetit et al., 2006). It is likely that the atomic model produced will have some missing or unassigned residues. If residues are missing, they can be manually added and fit to the density using X-ray crystallographic modeling packages like Coot. It is important to note that building atomic models may not be possible even at near-atomic resolutions. Features in the map dictate the possibility and level of accuracy in model construction. Further model optimization and refinement requires a significant number of sidechain densities to be recognizable.

4.17. Model optimization At near-atomic resolutions, all sidechains larger than Valine should be evident in the density map though this is generally not the case in practice. Therefore, precedence is typically given to mainchain atom positions within the density. However, sidechain density associated with positively charged and aromatic amino acids are discernable more frequently than other amino acids (Cong et al., 2010; Kimura et al., 1997; Zhang et al., 2010) and thus provide landmarks for optimizing the model in the density map.

Cryo-M Map Analysis

19

Conversely, Proline and Glycine are almost always marked by weak or broken density, though this can also be used to register the model to the density map as well. In model optimization, small stretches of residues in the atomic model are adjusted to better fit the density map while maintaining good stereochemistry and enforcing sidechain and mainchain restraints. This process is generally interactive and begins in well-resolved regions such as a-helices, where the pitch of the helix and large, positively charged sidechains provide sufficient points to anchor the placement of atoms. Real-space refinement options found in computational modeling software, like Coot, allow the user to manually adjust atom positions while maintaining realistic biochemical properties. Additionally, sidechain positions can be optimized using a rotamer search such that the corresponding atoms occupy any visible sidechain density. To achieve optimal rotamer assignment, mainchain atom position may need to be altered; optimization at this point should be done only over a small (3–5) number of amino acids. The entire model optimization procedure is relatively subjective and will likely be iterated until all the atoms are placed in the density map and registered with any visible features.

4.18. Monitoring model quality Results of the previous optimization step can be monitored using a Ramachandran plot (Fig. 1.5). When completed, all amino acids should fall in favorable or acceptable positions on the Ramachandran plot. Several iterations may be required to optimally assign all atoms. Once the final structure has been achieved, the model may be fit back into the entire assembly along with any other components.

4.19. Ca to atomic model optimization with Rosetta The process of model optimization described above is relatively interactive. Alternatively, computational modeling is capable of semiautomatically optimizing an initial model, bypassing much of the user-intensive optimization steps. Rosetta (Bradley et al., 2005) can refine structures constrained by experimental electron density maps by optimizing an all-atom energy function that includes both statistical and chemical potential energy terms. When refining a model, a scoring term that assesses the fit of a model to the density map is simultaneously optimized with Rosetta’s standard energy function (DiMaio et al., 2009). From a Ca trace, Rosetta’s rebuilding-and-refinement protocol is used to refine the structure. The Rosetta protocol ca_to_allatom infers an all-atom model and performs structure refinement. The protocol generates models by sampling different conformations of individual SSEs. The Ca positions in

20

Matthew L. Baker et al.

Figure 1.5 Model optimization and validation. Coot can be used to adjust mainchain and sidechain atom positions, optimizing the fit of the atomic model in the density map. A Ramachandran plot is shown overlaid with the map and model of Mm-cpn after optimization.

the starting model guide placement of the initial structure with a usercontrollable parameter specifying how far Cas are allowed to deviate from the starting model. In the next stage, each atom is explicitly modeled and then evaluated using the complete all-atom energy function. Loops are rebuilt, sidechains are placed on the structure and the entire structure is relaxed with Rosetta’s high-resolution energy function. Throughout the entire process, harmonic constraints keep Ca positions from deviating too far from their initial positions. This protocol generally requires significant sampling, on the order of thousands to tens of thousands of models. In addition, even the best models produced may still have loops and other features outside of the density contours. Thus, it is often necessary to follow this protocol with iterative rebuilding using Rosetta’s loopmodel. A final all-atom optimization with a high-resolution energy function is performed to sample less-common sidechain rotamers and sidechain torsions.

Cryo-M Map Analysis

21

4.20. Model optimization with Rosetta Not limited to near-atomic resolutions, the aforementioned Rosetta protocols can also be adapted for use with lower resolution structures assuming a-helices can be identified within the density and a clear correspondence can be established with the predicted secondary structure. Similarly, Rosetta’s relax and rebuilding-and-refinement protocols may also be applied to homology models to improve model accuracy and fit into a subnanometer resolution density map.

5. Case Study: Mm-cpn To illustrate the above process, we have chosen to detail the process of constructing an atomic model directly from a cryo-EM density map, as ˚ resolution structure of Mm-cpn (Fig. 1.6A; Zhang et al., done for the 4.3 A 2010). Where appropriate, specific reference is made to the programs, parameters and results obtained for Mm-cpn using this protocol. It should be noted that the level of detail found in the Mm-cpn density map is not required for analyzing and annotating subnanometer resolution protein structure. Rather, Mm-cpn simply provides a convenient and accessible vehicle to describe the variety of tools available for analyzing subnanometer resolution protein structure. 1. Sixteen subunits were isolated using segment3d from EMAN (segment3d mmcpn.mrc segmented-mmcpn.mrc nseg ¼ 16 split apix ¼ 1.33 sym ¼ d8), which uses a K-means approach to identify subunits (Fig. 1.6A). 2. For Mm-cpn, a single subunit was padded to a 1283 density map and the origin was reset to zero with EMAN using the following command: proc3d mmcpn1.mrc mmcpn-monomer128.mrc clip ¼ 128,128,128 origin ¼ 0,0,0. Note: mmcpn1.mrc is one the 16 segmented monomers from the prior step. 3. SSEHunter was used to identify SSEs using the following command in EMAN: ssehunter3.py mmcpn-monomer128.mrc 1.33 4.5 0.4. A single Mm-cpn subunit (543 amino acids per subunit) returned 201 pseudoatoms. 4. The results were loaded into Chimera along with the density map. Pseudoatoms were represented as spheres and bonds were hidden. The pseudoatoms were colored using Chimera’s “Render by Attribute” option such that the most negative value in the B-factor column (3) was set to blue, the most positive value (3) was set to red, and zero was set to white. 5. For a single Mm-cpn subunit, five b-sheets and 17 a-helices were identified and built using SSEBuilder (Fig. 1.6B).

22

A

Matthew L. Baker et al.

B

C

E

General case

180

517 ALA 481 GLU

E.

D Psi

0 494 VAL

167 LYS

-180 -180

15 TYR

Phi

204 ASP 452 ALA 217 GLU

180

˚ resolution structure of Mm-cpn is Figure 1.6 Structure of Mm-cpn. (A) The 4.3-A shown (Zhang et al., 2010). (B) Using SSEHunter, the secondary structure elements in the Mm-cpn subunit were identified: a-helices are shown as cylinders and b-strands are shown as planes. (C) Using the de novo modeling approach, an atomic model (residues 1– 532) was constructed for one subunit of Mm-cpn. (D) Large, bulky sidechains in the model could be seen in the density. (E) The Ramachandran plot of an Mm-cpn monomer shows greater than 98% of all residues with allowable phi–psi angles.

6. The structure of a related chaperone, the thermosome KS-1 (PDB ID: 1Q3Q; Shomura et al., 2004), was used to construct a homology model for a single Mm-cpn subunit as done in a previous study (Booth et al., 2008). 7. The homology model for the Mm-cpn subunit was fit to the segmented density map using Foldhunter ( Jiang et al., 2001) from EMAN with the following command: foldhunter.py mmcpn-monomer128.mrc mmcpn-homology-model.pdb res ¼ 4.3 apix ¼ 1.33. The resulting transformed PDB model was then loaded into Chimera to verify the fit to the density map and the agreement with the SSEs identified in the previous steps.

Cryo-M Map Analysis

23

8. The fitted homology model was adjusted in Coot and Gorgon such that the model agreed with the a-helices identified by SSEHunter. The helices required only slight rotations and translations. 9. Once the helix positions were optimized strands and loops were adjusted to optimally fit the density while maintaining 3.8 A˚ Ca– Ca distances. The density skeleton was used to identify potential paths through the density. 10. In the Mm-cpn homology model, 30 amino acids were unresolved at the termini. In Coot, Ca atoms were added consecutively to the model starting at the ends of the model until the N-terminus was reached or until the density was not visible (C-terminus). The final model for the Mm-cpn monomer contained residues 1–532 (Fig. 1.6C). 11. After the addition of missing residues, optimization of the complete Ca model was performed to maximize the occupancy of the model in the subunit density map, while maintaining appropriate distance constraints. 12. Sixteen copies of the Mm-cpn Ca model were loaded into Chimera, manually moved into a subunit and fit to the density using the “Fit in map” option (Pettersen et al., 2004). 13. After eight iterations of refinement and fitting (steps 11 and 12) of the entire macromolecular assembly, all clashes were eliminated and model occupancy was optimized. A single PDB file was saved containing all 16 copies of the optimized Mm-cpn Ca model as separate chains. 14. From the full Mm-cpn Ca model, the structure factors for the model were calculated and applied to the original density map using EMAN (detailed description of the parameters can be found in the EMAN documentation). pdb2mrc mmcpn.pdb mmcpn-simulated.mrc res ¼ 4.3 apix ¼ 1.33 box ¼ 192 proc3d mmcpn-simulated.mrc junk.mrc calcsf ¼ mmcpn-sf.txt apix ¼ 1.33 proc3d mmcpn.mrc mmcpn-rescaled.mrc setsf ¼ mmcpn-sf.txt apix ¼ 1.33 proc3d mmcpn-rescaled.mrc mmcpn-lp.mrc lp ¼ 4.3 apix ¼ 1.33 proc3d mmcpn-lp.mrc mmcpn-final-map.mrc mult ¼ 0.3 15. Using the SABBAC web server, the Ca model for one subunit was transformed into an atomic model. The resulting atomic model was loaded into Coot along with the rescaled density map. 16. Portions of the Mm-cpn atomic model created by SABBAC contained breaks in the polypeptide chain. The “Model/Fit/Refine” tools in Coot were used to move/add residues such that a complete polypeptide chain was constructed. 17. Once a complete all-atom model for an Mm-cpn subunit was produced, the “Model/Fit/Refine” tools in Coot were used to move the mainchain atoms in the density map. Small stretches of residues, from three amino acids to entire helices, were fit to the density such that (1) the mainchain atoms were encompassed by density, (2) secondary

24

18.

19.

20. 21.

Matthew L. Baker et al.

structure constraints were maintained and registered with the density and (3) potential sidechain density for large, bulky sidechains were proximal to residues in the model containing corresponding sidechains. In this refinement round, torsion angles, planar peptide constraints, and Ramachandran constraints were enforced for the appropriate type of secondary structure. For every amino acid, a rotamer search was performed allowing sidechains to be placed within corresponding sidechain density. In some instances, the “Model/Fit/Refine” tools in Coot were used to adjust the mainchain and sidechain positions to best fit the density (Fig. 1.6D). The Ramachandran plot in Coot plots the phi–psi angles. Residues falling outside of the acceptable range are plotted in red. Clicking on these outliers, the map and model are recentered in the main Coot display. The steps 17 and 18 were performed over small stretches of residues (3–5 amino acids) containing the outliers until the residues were in acceptable or favorable conformations. After eight iterations of steps 17–19, the Mm-cpn model had greater than 98% of all residues with acceptable Ramachandran angles (Fig. 1.6E). Steps 11 and 12 were repeated except with a rescaled density map and the refined atomic model. A final model containing all 16 subunits was saved and deposited in the PDB.

6. Discussion While the above protocol describes a complete approach to build and refine an atomic model from a cryo-EM density map at subnanometer resolution (Fig. 1.2), individual tools can be, and most often are, used independently. Thus, it is important to know when and what tools are most appropriate for a specific problem. Limitations in the analysis of cryo-EM density maps, due in large part to resolution, are obvious, though not prohibitive, in describing salient structural features and functions in macromolecular assemblies. The ability to analyze a cryo-EM density map hinges on the map itself; size, complexity, and quality of the density map all play critical roles in annotating structure at any resolution. Even the most experienced scientists may not be able to reliably describe features in poorly resolved regions of density maps. One ˚ resolution; conversely, the would not expect to see sidechain density at 9 A ˚ absence of b-strand separation at 3.5 A resolution may indicate potential problems. As such, the analysis process requires a significant investment in time and understanding the quality of the original 2D data and the reconstructed volume.

Cryo-M Map Analysis

25

Illustrating the dependance on the resolvability of density features, de novo model building is based on establishing a sequence-to-structure correspondence using SSEs, features unique to subnanometer resolutions. This necessitates the presence of clearly identifiable SSEs in the density map, though connecting loops may be ambiguous. Density maps vary in composition, quality, and resolution making it difficult to assign a clear resolution cut-off. Model building may be easier and more reliable at near-atomic ˚ ) but still possible at lower resolutions depending on the resolutions (3.5–5 A features that are resolved in the map, as in the case of Hepatitis B (Bo¨ttcher et al., 1997; Conway et al., 1997). At higher resolutions, sidechain density can aid in the placement of Ca atoms, thereby increasing the accuracy and reliability of models. Thus, in all cases of interpreting a subnanometer resolution density map, precedence must be given to the observable features in the density map and not the stated resolution.

ACKNOWLEDGMENTS This work is supported by grants from NIH (P41RR02250, R01GM079429, R01AI0175208) and NSF (IIS-0705644, IIS-0705474). M. R. Baker is supported by a postdoctoral training fellowship from the National Library of Medicine Training Program in Computational Biology and Biomedical Informatics provided by the Keck Center and Gulf Coast Consortia (T15LM007093).

REFERENCES Abeysinghe, S., Ju, T., Baker, M. L., and Chiu, W. (2008). Shape modeling and matching in identifying 3D protein structures. Comput. Aided Des. 40, 708–720. Altschul, S. F., Gish, W., Miller, W., Myers, E. W., and Lipman, D. J. (1990). Basic local alignment search tool. J. Mol. Biol. 215, 403–410. Arnold, K., Bordoli, L., Kopp, J., and Schwede, T. (2006). The SWISS-MODEL workspace: A web-based environment for protein structure homology modelling. Bioinformatics 22, 195–201. Baker, M. L., Jiang, W., Bowman, B. R., Zhou, Z. H., Quiocho, F. A., Rixon, F. J., and Chiu, W. (2003). Architecture of the herpes simplex virus major capsid protein derived from structural bioinformatics. J. Mol. Biol. 331, 447–456. Baker, M. L., Jiang, W., Rixon, F. J., and Chiu, W. (2005). Common ancestry of herpesviruses and tailed DNA bacteriophages. J. Virol. 79, 14967–14970. Baker, M. L., Jiang, W., Wedemeyer, W. J., Rixon, F. J., Baker, D., and Chiu, W. (2006a). Ab initio modeling of the herpesvirus VP26 core domain assessed by CryoEM density. PLoS Comput. Biol. 2, e146. Baker, M. L., Yu, Z., Chiu, W., and Bajaj, C. (2006b). Automated segmentation of molecular subunits in electron cryomicroscopy density maps. J. Struct. Biol. 156, 432–441. Baker, M. L., Ju, T., and Chiu, W. (2007). Identification of secondary structure elements in intermediate-resolution density maps. Structure 15, 7–19.

26

Matthew L. Baker et al.

Bates, P. A., Kelley, L. A., MacCallum, R. M., and Sternberg, M. J. (2001). Enhancement of protein modeling by human intervention in applying the automatic programs 3D-JIGSAW and 3D-PSSM. Proteins (Suppl. 5), 39–46. Baumeister, W., and Steven, A. C. (2000). Macromolecular electron microscopy in the era of structural genomics. Trends Biochem. Sci. 25, 624–631. Booth, C. R., Meyer, A. S., Cong, Y., Topf, M., Sali, A., Ludtke, S. J., Chiu, W., and Frydman, J. (2008). Mechanism of lid closure in the eukaryotic chaperonin TRiC/CCT. Nat. Struct. Mol. Biol. 15, 746–753. Bo¨ttcher, B., Wynne, S. A., and Crowther, R. A. (1997). Determination of the fold of the core protein of hepatitis B virus by electron cryomicroscopy. Nature 386, 88–91. Bradley, P., Malmstro¨m, L., Qian, B., Schonbrun, J., Chivian, D., Kim, D. E., Meiler, J., Misura, K. M., and Baker, D. (2005). Free modeling with Rosetta in CASP6. Proteins 61 (Suppl 7), 128–134. Chiu, W., Baker, M. L., Jiang, W., Dougherty, M., and Schmid, M. F. (2005). Electron cryomicroscopy of biological machines at subnanometer resolution. Structure 13, 363–372. Cole, C., Barber, J. D., and Barton, G. J. (2008). The Jpred 3 secondary structure prediction server. Nucleic Acids Res. 36, W197–W201. Cong, Y., and Ludtke, S. J. (2010). Single particle analysis at high resolution. Methods Enzymol. 482, 211–236. Cong, Y., Baker, M. L., Jakana, J., Woolford, D., Miller, E. J., Reissmann, S., Kumar, R. N., Redding-Johanson, A. M., Batth, T. S., Mukhopadhyay, A., et al. (2010). 4.0-A˚ resolution cryo-EM structure of the mammalian chaperonin TRiC/ CCT reveals its unique subunit arrangement. Proc. Natl. Acad. Sci. USA 107, 4967–4972. Conway, J. F., Cheng, N., Zlotnick, A., Wingfield, P. T., Stahl, S. J., and Steven, A. C. (1997). Visualization of a 4-helix bundle in the hepatitis B virus capsid by cryo-electron microscopy. Nature 386, 91–94. DiMaio, F., Tyka, M. D., Baker, M. L., Chiu, W., and Baker, D. (2009). Refinement of protein structures into low-resolution density maps using rosetta. J. Mol. Biol. 392, 181–190. Emsley, P., and Cowtan, K. (2004). Coot: Model-building tools for molecular graphics. Acta Crystallogr. D Biol. Crystallogr. 60, 2126–2132. Ferna´ndez, J. J., Luque, D., Casto´n, J. R., and Carrascosa, J. L. (2008). Sharpening high resolution information in single particle electron cryomicroscopy. J. Struct. Biol. 164, 170–175. Frank, J. (2002). Single-particle imaging of macromolecules by cryo-electron microscopy. Annu. Rev. Biophys. Biomol. Struct. 31, 303–319. Ginalski, K., Elofsson, A., Fischer, D., and Rychlewski, L. (2003). 3D-Jury: A simple approach to improve protein structure predictions. Bioinformatics 19, 1015–1018. Gonen, T., Cheng, Y., Sliz, P., Hiroaki, Y., Fujiyoshi, Y., Harrison, S. C., and Walz, T. (2005). Lipid-protein interactions in double-layered two-dimensional AQP0 crystals. Nature 438, 633–638. Henderson, R., and Unwin, N. R. (1975). Three-dimensional model of purple membrane obtained by electron microscopy. Nature 257, 28–32. Henderson, R., Baldwin, J. M., Ceska, T. A., Zemlin, F., Beckmann, E., and Downing, K. H. (1990). Model for the structure of bacteriorhodopsin based on highresolution electron cryo-microscopy. J. Mol. Biol. 213, 899–929. Humphrey, W., Dalke, A., and Schulten, K. (1996). VMD: Visual molecular dynamics. J. Mol. Graph. 14(33–8), 27–28. Jiang, W., Baker, M. L., Ludtke, S. J., and Chiu, W. (2001). Bridging the information gap: Computational tools for intermediate resolution structure interpretation. J. Mol. Biol. 308, 1033–1044.

Cryo-M Map Analysis

27

Jiang, W., Baker, M. L., Jakana, J., Weigele, P. R., King, J., and Chiu, W. (2008). Backbone structure of the infectious epsilon15 virus capsid revealed by electron cryomicroscopy. Nature 451, 1130–1134. Jones, T. A., Zou, J. Y., Cowan, S. W., and Kjeldgaard, M. (1991). Improved methods for building protein models in electron density maps and the location of errors in these models. Acta Crystallogr. A 47(Pt 2), 110–119. Ju, T., Baker, M. L., and Chiu, W. (2007). Computing a family of skeletons of volumetric models for shape description. Comput. Aided Des. 39, 352–360. Kelley, L. A., and Sternberg, M. J. (2009). Protein structure prediction on the Web: A case study using the Phyre server. Nat. Protoc. 4, 363–371. Kimura, Y., Vassylyev, D. G., Miyazawa, A., Kidera, A., Matsushima, M., Mitsuoka, K., Murata, K., Hirai, T., and Fujiyoshi, Y. (1997). Surface of bacteriorhodopsin revealed by high-resolution electron crystallography. Nature 389, 206–211. Kleywegt, G. J., and Jones, T. A. (1997). Detecting folding motifs and similarities in protein structures. Methods Enzymol. 277, 525–545. Kong, Y., and Ma, J. (2003). A structural-informatics approach for mining beta-sheets: Locating sheets in intermediate-resolution density maps. J. Mol. Biol. 332, 399–413. Kong, Y., Zhang, X., Baker, T. S., and Ma, J. (2004). A structural-informatics approach for tracing beta-sheets: Building pseudo-C(alpha) traces for beta-strands in intermediateresolution density maps. J. Mol. Biol. 339, 117–130. Li, Z., Baker, M. L., Jiang, W., Estes, M. K., and Prasad, B. V. (2009). Rotavirus architecture at subnanometer resolution. J. Virol. 83, 1754–1766. Lipman, D. J., and Pearson, W. R. (1985). Rapid and sensitive protein similarity searches. Science 227, 1435–1441. Ludtke, S. J., Baldwin, P. R., and Chiu, W. (1999). EMAN: Semiautomated software for high-resolution single-particle reconstructions. J. Struct. Biol. 128, 82–97. Ludtke, S. J., Serysheva, I. I., Hamilton, S. L., and Chiu, W. (2005). The pore structure of the closed RyR1 channel. Structure 13, 1203–1211. Ludtke, S. J., Baker, M. L., Chen, D. H., Song, J. L., Chuang, D. T., and Chiu, W. (2008). De novo backbone trace of GroEL from single particle electron cryomicroscopy. Structure 16, 441–448. Maupetit, J., Gautier, R., and Tuffe´ry, P. (2006). SABBAC: Online Structural Alphabetbased protein BackBone reconstruction from Alpha-Carbon trace. Nucleic Acids Res. 34, W147–W151. McGuffin, L. J., Bryson, K., and Jones, D. T. (2000). The PSIPRED protein structure prediction server. Bioinformatics 16, 404–405. Mizuguchi, K., and Go, N. (1995). Comparison of spatial arrangements of secondary structural elements in proteins. Protein Eng. 8, 353–362. Nakagawa, A., Miyazaki, N., Taka, J., Naitow, H., Ogawa, A., Fujimoto, Z., Mizuno, H., Higashi, T., Watanabe, Y., Omura, T., et al. (2003). The atomic structure of rice dwarf virus reveals the self-assembly mechanism of component proteins. Structure 11, 1227–1238. Pettersen, E. F., Goddard, T. D., Huang, C. C., Couch, G. S., Greenblatt, D. M., Meng, E. C., and Ferrin, T. E. (2004). UCSF Chimera–a visualization system for exploratory research and analysis. J. Comput. Chem. 25, 1605–1612. Pintilie, G. D., Zhang, J., Goddard, T. D., Chiu, W., and Gossard, D. C. (2010). Quantitative analysis of cryo-EM density map segmentation by watershed and scale-space filtering, and fitting of structures by alignment to regions. J. Struct. Biol. 170, 427–438. Pollastri, G., Przybylski, D., Rost, B., and Baldi, P. (2002). Improving the prediction of protein secondary structure in three and eight classes using recurrent neural networks and profiles. Proteins 47, 228–235. Roseman, A. M. (2000). Docking structures of domains into maps from cryo-electron microscopy using local correlation. Acta Crystallogr. D Biol. Crystallogr. 56, 1332–1340.

28

Matthew L. Baker et al.

Rossmann, M. G. (2000). Fitting atomic models into electron-microscopy maps. Acta Crystallogr. D Biol. Crystallogr. 56, 1341–1349. Rossmann, M. G., Morais, M. C., Leiman, P. G., and Zhang, W. (2005). Combining X-ray crystallography and electron microscopy. Structure 13, 355–362. Rusu, M., Birmanns, S., and Wriggers, W. (2008). Biomolecular pleiomorphism probed by spatial interpolation of coarse models. Bioinformatics 24, 2460–2466. Sali, A., Potterton, L., Yuan, F., van Vlijmen, H., and Karplus, M. (1995). Evaluation of comparative protein modeling by MODELLER. Proteins 23, 318–326. Schro¨der, G. F., Brunger, A. T., and Levitt, M. (2007). Combining efficient conformational sampling with a deformable elastic network model facilitates structure refinement at low resolution. Structure 15, 1630–1641. Serysheva, I. I., Hamilton, S. L., Chiu, W., and Ludtke, S. J. (2005). Structure of Ca2þ release channel at 14 A˚ resolution. J. Mol. Biol. 345, 427–431. Serysheva, I. I., Ludtke, S. J., Baker, M. L., Cong, Y., Topf, M., Eramian, D., Sali, A., Hamilton, S. L., and Chiu, W. (2008). Subnanometer-resolution electron cryomicroscopy-based domain models for the cytoplasmic region of skeletal muscle RyR channel. Proc. Natl. Acad. Sci. USA 105, 9610–9615. Shi, J., Blundell, T. L., and Mizuguchi, K. (2001). FUGUE: Sequence-structure homology recognition using environment-specific substitution tables and structure-dependent gap penalties. J. Mol. Biol. 310, 243–257. Shomura, Y., Yoshida, T., Iizuka, R., Maruyama, T., Yohda, M., and Miki, K. (2004). Crystal structures of the group II chaperonin from Thermococcus strain KS-1: Steric hindrance by the substituted amino acid, and inter-subunit rearrangement between two crystal forms. J. Mol. Biol. 335, 1265–1278. Siebert, X., and Navaza, J. (2009). UROX 2.0: An interactive tool for fitting atomic models into electron-microscopy reconstructions. Acta Crystallogr. D Biol. Crystallogr. 65, 651–658. Suhre, K., Navaza, J., and Sanejouand, Y. H. (2006). NORMA: A tool for flexible fitting of high-resolution protein structures into low-resolution electron-microscopy-derived density maps. Acta Crystallogr. D Biol. Crystallogr. 62, 1098–1100. Tama, F., Miyashita, O., and Brooks, C. L. (2004). Normal mode based flexible fitting of high-resolution structure into low-resolution experimental data from cryo-EM. J. Struct. Biol. 147, 315–326. Tan, R. K., Devkota, B., and Harvey, S. C. (2008). YUP.SCX: Coaxing atomic models into medium resolution electron density maps. J. Struct. Biol. 163, 163–174. Tang, G., Peng, L., Baldwin, P. R., Mann, D. S., Jiang, W., Rees, I., and Ludtke, S. J. (2007). EMAN2: An extensible image processing suite for electron microscopy. J. Struct. Biol. 157, 38–46. Topf, M., Baker, M. L., John, B., Chiu, W., and Sali, A. (2005). Structural characterization of components of protein assemblies by comparative modeling and electron cryomicroscopy. J. Struct. Biol. 149, 191–203. Topf, M., Baker, M. L., Marti-Renom, M. A., Chiu, W., and Sali, A. (2006). Refinement of protein structures by iterative comparative modeling and CryoEM density fitting. J. Mol. Biol. 357, 1655–1668. Topf, M., Lasker, K., Webb, B., Wolfson, H., Chiu, W., and Sali, A. (2008). Protein structure fitting and refinement guided by cryo-EM density. Structure 16, 295–307. Trabuco, L. G., Villa, E., Schreiner, E., Harrison, C. B., and Schulten, K. (2009). Molecular dynamics flexible fitting: A practical guide to combine cryo-electron microscopy and X-ray crystallography. Methods 49, 174–180. Volkmann, N. (2002). A novel three-dimensional variant of the watershed transform for segmentation of electron density maps. J. Struct. Biol. 138, 123–129.

Cryo-M Map Analysis

29

Volkmann, N. (2009). Confidence intervals for fitting of atomic models into low-resolution densities. Acta Crystallogr. D Biol. Crystallogr. 65, 679–689. Volkmann, N., and Hanein, D. (1999). Quantitative fitting of atomic models into observed densities derived by electron microscopy. J. Struct. Biol. 125, 176–184. 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. Wynne, S. A., Crowther, R. A., and Leslie, A. G. (1999). The crystal structure of the human hepatitis B virus capsid. Mol. Cell 3, 771–780. Yu, X., Jin, L., and Zhou, Z. H. (2008). 3.88 A˚ structure of cytoplasmic polyhedrosis virus by cryo-electron microscopy. Nature 453, 415–419. Zhang, X., Settembre, E., Xu, C., Dormitzer, P. R., Bellamy, R., Harrison, S. C., and Grigorieff, N. (2008). Near-atomic resolution using electron cryomicroscopy and singleparticle reconstruction. Proc. Natl. Acad. Sci. USA 105, 1867–1872. Zhang, J., Baker, M. L., Schro¨der, G. F., Douglas, N. R., Reissmann, S., Jakana, J., Dougherty, M., Fu, C. J., Levitt, M., Ludtke, S. J., et al. (2010). Mechanism of folding chamber closure in a group II chaperonin. Nature 463, 379–383. Zhou, Z. H., Baker, M. L., Jiang, W., Dougherty, M., Jakana, J., Dong, G., Lu, G., and Chiu, W. (2001). Electron cryomicroscopy and bioinformatics suggest protein fold models for rice dwarf virus. Nat. Struct. Biol. 8, 868–873. Zhu, J., Cheng, L., Fang, Q., Zhou, Z. H., and Honig, B. (2010). Building and refining protein models within cryo-electron microscopy density maps based on homology modeling and multiscale structure refinement. J. Mol. Biol. 397, 835–851.

C H A P T E R

T W O

Methods for Segmentation and Interpretation of Electron Tomographic Reconstructions Niels Volkmann Contents 32 35 36 38 39 40 41 42

1. Introduction 2. Noise Reduction 3. Segmentation 4. Detection and Mapping of Macromolecular Assemblies 5. Classification and Averaging 6. Validation Acknowledgments References

Abstract Electron tomography has become a powerful tool for revealing the molecular architecture of biological cells and tissues. In principle, electron tomography can provide high-resolution mapping of entire proteomes. The achievable resolution (3–8 nm) is capable of bridging the gap between live-cell imaging and atomic resolution structures. However, the relevant information is not readily accessible from the data and needs to be identified, extracted, and processed before it can be used. Because electron tomography imaging and image acquisition technologies have enjoyed major advances in the last few years and continue to increase data throughput, the need for approaches that allow automatic and objective interpretation of electron tomograms becomes more and more urgent. This chapter provides an overview of the state of the art in this field and attempts to identify the major bottlenecks that prevent approaches for interpreting electron tomography data to develop their full potential.

Sanford-Burnham Medical Research Institute, La Jolla, California, USA Methods in Enzymology, Volume 483 ISSN 0076-6879, DOI: 10.1016/S0076-6879(10)83002-2

#

2010 Elsevier Inc. All rights reserved.

31

32

Niels Volkmann

1. Introduction Electron tomography is the most widely applicable method for obtaining 3D information by electron microscopy. In fact, it is the only method suitable for investigating polymorphic structures such as organelles, cells, and tissue at high resolution (current estimates 3–8 nm). Its principle is based on illuminating the sample from many different directions, usually tilt series around one or two axes, and to reconstruct it from those projection images. While the principles of electron tomography have been known for decades, its use has gathered momentum only in recent years. It has been realized that electron tomography, especially its cryovariant, is capable of providing a complete, molecular resolution 3D mapping of entire cellular proteomes including their detailed interactions (Leis et al., 2009; Nickell et al., 2006; Robinson et al., 2007; Tocheva et al., 2010; Volkmann and Hanein, 2009). Electron tomography can depict unique structures and scenes but, due to the fact that all electron tomography preparations can only sustain a limited electron dose, the resulting reconstructions will inevitably suffer from low signal-to-noise ratios and relatively low resolution as compared to other electron microscopy techniques that can take advantage of some form of averaging. Consequently, maps obtained by electron tomography are difficult to interpret. This difficulty in interpretation is further aggravated in highly complex systems (Gru¨newald et al., 2003). Despite a recent surge in dedicated method development toward automatic interpretation of electron tomograms (reviewed in Best et al., 2007; Frangakis and Fo¨rster, 2004; Sandberg, 2007), only relatively few algorithms for reliable detection and extraction of structural features from electron tomograms are available. Instead, the tasks of extracting and interpreting information from the highly complex, 3D scenes that make up cellular tomograms are, for the most part, painstakingly carried out manually. Apart from the subjectivity of the process, the time-consuming (and tiring) nature of this manual task all but precludes the prospects of the high throughput necessary to take full advantage of the method’s potential. For example, it took over 9 months to manually segment and interpret roughly 1% of the volume of a pancreatic beta cell (Marsh, 2005; Marsh et al., 2001). By combining sectioning with automatic data collection schemes (Chapter 12, Vol. 481) and large-scale montaging, it is now technically possible to reconstruct entire mammalian cells with high fidelity within a reasonable timeframe (Noske et al., 2008). Extrapolating from the time required for segmenting 1% of a cell results in 75 man years for manually segmenting a single cell at similar detail. Conducting a meaningful study comparing a number of cells under disease conditions with a control set would literally take hundreds of man years. The need for computational tools to efficiently aid the process and automate the structure recognition, extraction, and interpretation process as much as possible is clearly vital for making these types of studies viable.

Methods for Segmentation and Interpretation of Electron Tomographic Reconstructions

33

The quality of cryotomographic reconstructions can be correlated with ˚ 2 appears to give reasonable the electron dose. A total dose of 50–300 e/A ˚2 results with a ‘‘sweet spot’’ around 120 e /A (Iancu et al., 2006). Because this dose needs to be spread over the whole data set, the dose for each image needs to be kept low enough as to not exceed a total dose of 120 e/A˚2. For a 70 double tilt series with a 2 increment, the dose available for a single image is below 1 e/A˚2, which gives rise to extremely high noise levels in the individual images. The signal in the resulting 3D reconstructions is improved by the dose fractionation effect (McEwen et al., 1995) but the signal-to-noise ratio for these tomograms is still well below 1. Signal-tonoise ratios in cryo-images collected for single-particle reconstructions of ribosomes have been experimentally determined to be in the 0.05 range (Baxter et al., 2009). Given the fact that most contributions to the noise tend to be considerably worse in cryo-tomograms than in these relatively well behaved samples, signal-to-noise ratios in the neighborhood of 0.01 or less should be expected for cryo-tomograms. Together with complications from missing data and the electron microscope’s contrast transfer, this makes noise from many other image-possessing disciplines look mundane. Multiple scattering events in samples thicker than the mean free path length of the illuminating electrons dictate an upper limit of 1 mm in thickness for high-resolution electron tomography, even if relatively high acceleration voltage and energy filtering are used (Grimm et al., 1996). Owing to the technically demanding nature of cryosample sectioning (AlAmoudi et al., 2004), ‘‘conventional’’ electron tomography, which involves staining and plastic embedding, is often preferred in practice for samples that require sectioning (McEwen and Marko, 2001). In fact, cryosectioning is currently an art form practiced in only a handful of laboratories worldwide (see, e.g., Al-Amoudi et al., 2007; Gruska et al., 2008; Hsieh et al., 2006; Pierson et al., 2010; Salje et al., 2009). In embedded material, the main adverse effect of electron irradiation is not direct damage to the particles (which have been mostly exchanged for heavy metal stains anyway) it is damage to the embedding material. For example, the beam induces serious thinning (25–50%) perpendicular to the beam (Luther et al., 1988). The usual strategy around this is to preexpose the specimen to allow completion of a rapid initial thinning phase. The tilt series are then collected with low electron dose (up to about 100 e/A˚2 per image is deemed tolerable) to avoid beam-induced buckling of the sample or further thinning during data collection. While the signal-to-noise ratio is improved in these samples as compared to cryosamples, the resulting images and reconstructions still tend to be rather noisy. In addition, these samples can suffer additional problems such as uneven staining and other sample preparation artifacts that may complicate subsequent analysis. While there have been substantial efforts during the last few years which specifically address data from electron tomography, progress has been much

34

Niels Volkmann

slower than in related imaging fields. As a consequence, the relative lack of adequate tools for automatic and/or objective interpretation in electron tomography has been recognized as a critical barrier to progress in the field (Ben-Harush et al., 2010; Frey et al., 2006; Koning and Koster, 2009; Leis et al., 2006; Marsh, 2005). Most development efforts in biological image processing are geared toward clinical medicine: Clinical data sets typically contain organs of welldefined shape with smooth and distinct boundaries and include only a few continuous objects. In contrast, electron tomograms contain amorphous structures, ragged contours, and numerous small objects which would make it extremely challenging to interpret electron tomograms even if noise would be absent. Unlike medical data sets, for which considerable knowledge is available to validate the results and tune parameters, the interpretation of electron tomograms often leave the user uncertain about the accuracy of the analysis: it is difficult to distinguish artifacts from real structures. Other reasons why image and signal processing methods developed for other domains are not straight forward to apply to electron tomography data include: (i) As a consequence of the damaging effect of the electron beam which limits the amount of electrons available for image formation, electron tomograms tend to exhibit extremely high noise levels. (ii) The geometry of the electron tomography sample holders and the shape of the electron microscopy chamber do not allow tilting more than 70 so a good third of data space is not accessible. In accordance with the projection theorem (Radon, 1917), the nonsampled region generates a wedge-shaped segment in Fourier space that contains no information (the ‘‘missing wedge’’). This problem can be partially alleviated experimentally by taking a second data set after rotating the sample by 90 around the optical axis (Mastronarde, 1997; Nickell et al., 2003; Penczek et al., 1995), yet some of the data space is still not accessible so that some missing data artifacts will always remain. (iii) Aberrations of the electron microscope optics give rise to a pointspread function that can be best described in Fourier space by its Fourier transform, the contrast transfer function. This function mainly depends on the amount of defocus used for imaging and changes the amplitudes and phases of the signal. In an electron tomography setting, the contrast transfer function is not well defined within the sample, especially for thick specimens where there can be a significant variance in focus. In addition, the tilting introduces a focus gradient, further obstructing the underlying signal. (iv) The noise in the reconstruction results from a complex combination of different noise sources including signal-dependent shot noise due to the quantum nature of the electrons, digitization noise, and ‘‘structural noise’’ due to the support/embedding medium (Baxter et al., 2009). In

Methods for Segmentation and Interpretation of Electron Tomographic Reconstructions

35

addition, the noise is highly correlated in space, and is possibly corrupted by the contrast transfer function (Scheres et al., 2007) and the missing wedge. Thus, it is extremely hard to devise adequate noise models to take advantage of algorithms that explicitly take noise characteristics into account to improve performance. The combination of these issues makes devising automatic and objective interpretation tools extremely challenging. While the best course of action will heavily depend on the actual data and the biological question to be addressed, the general workflow for interpreting electron tomograms can be divided into a number of successive stages. Each of these stages will be described in more detail below. Most of the approaches described have been implemented in publicly available computer programs (see Chapter 15, Vol. 482 for a comprehensive list).

2. Noise Reduction Noise reduction has proven to be an indispensable tool for visualization and preprocessing of multidimensional images in the bioimaging field in general and in electron tomography in particular. Noise reduction is especially important for cryotomograms that suffer from the lowest contrast and highest noise levels. Noise-reduction schemes that were adapted for electron tomography relatively early on include wavelet transform techniques (Stoschek and Hegerl, 1997), and nonlinear anisotropic diffusion (Frangakis and Hegerl, 2001). A comparison between the two showed that nonlinear anisotropic diffusion appears to be preferable as a result of faster performance and the presence of better filtering properties (Frangakis et al., 2001). The nonlinear anisotropic diffusion approach was later expanded by using more sophisticated local kernels that result in enhancement of curvilinear and planar structures (Fernandez and Li, 2003). In addition, the bilateral filter (Tomasi and Manduchi, 1998) which is mathematically essentially equivalent to nonlinear anisotropic diffusion (Barash and Comaniciu, 2004) was adapted for electron tomography ( Jiang et al., 2003), achieving—not entirely surprisingly—similar results. Some improvements in performance over the standard bilateral filter can be achieved by explicitly accounting for impulse noise during the filtering step (Pantelic et al., 2006). The price to pay for the improvement is the need to tune one additional adjustable parameter. In general, the noise-reduction performance of all approaches outlined above relies heavily on the choice of a set of user-specified parameters which are hard to predict for any given tomogram. In addition, these approaches can be very demanding in terms of computation time and memory requirements. These drawbacks make these algorithms cumbersome to use for nonexpert users. Approaches that are easier to use, less demanding on the computational

36

Niels Volkmann

side and that maintain fairly reasonable performance for typical electron tomograms include iterative median filtering (van der Heide et al., 2007) and filtering based on a geometric diffusion flow called Beltrami flow (Kimmel et al., 2000) that was recently adapted for electron tomography (Fernandez, 2009). For many practical purposes such as reducing oversegmentation in automatic segmentation procedures a simple low-pass filter, despite its tendency to blur edges and features, may also be adequate (Pintilie et al., 2009).

3. Segmentation Interpreting an electron tomogram, even at the ultrastructural level, requires its decomposition into structural components such as membrane compartments, filamentous structures, or clusters of loosely associated macromolecules like polysomes. Various techniques have been proposed for automated or semiautomated segmentation in the field of image processing. Commonly used approaches include segmentation based on region growing, edge detection, active contours, and model-based segmentation. However, none of these proved to be straight-forward in use with electron tomography data and, until recently, most segmentations of electron tomographic volumes were carried out manually, using programs that allow tracing of volumes within slices to create an isocontour model of the structures of interest (Kremer et al., 1996; Li et al., 1997). This hand tracing tends to be tedious, time-consuming, and subjective (Frey et al., 2006). Many of the computational segmentation approaches developed for electron tomography attempt to improve upon manual segmentation using various types of surface fitting approaches. These range from simple spatial gradient optimization in two dimensions (Ress et al., 2004) through the use of 3D geodesic active contours (Bartesaghi et al., 2005), to implementations of static (fast marching method; Bajaj et al., 2003), as well as full-fledged dynamic levelset based approaches (Osher and Sethian, 1988; Whitaker and Elangovan, 2002). Drawbacks for these energy-minimization based algorithms are their tendency to be subject to local optima and scalability issues resulting in the requirement of reasonably good initial surface models and careful fine-tuning and preconditioning of parameters to ensure correct convergence. An alternative to these energy-based boundary detection algorithms is the use of region-based approaches where distinct regions are detected by some characteristic (intensity, texture) and their boundaries naturally become the output of the segmentation. In particular, the immersionbased watershed algorithm (Beucher and Meyer, 1993; Vincent and Soille, 1991) has been adapted specifically for use with electron tomography (Volkmann, 2002) and has been used successfully for various segmentation tasks in a wide variety of tomography projects (see, e.g., Auer et al., 2008;

Methods for Segmentation and Interpretation of Electron Tomographic Reconstructions

37

Janssen et al., 2006; Marsh et al., 2004; Rouiller et al., 2008; Salvi et al., 2008; Schietroma et al., 2009). The algorithm is based on an analogy with a step-wise flooding of a topological relief by a fluid, with dams being built where independent flows meet. The method is in principle capable of fully automatic segmentation but, in practice, is more appropriately used in a semiautomatic fashion in order to optimize the few operating parameters. The speed of the method allows interactive refinement of these parameters (if the volume is not too large) and, consequently, has now been implemented in the popular graphics packages Amira (Pruggnaller et al., 2008; Stalling et al., 2004) and Chimera (Goddard et al., 2007; Pintilie et al., 2009). Other segmentation approaches that have shown promising potential with electron tomography data are based on normalized graph cut methods and eigenvector analysis (Frangakis and Hegerl, 2002), on orientation fields and line segment detection (Sandberg and Brega, 2007), and on fuzzy sets theory (Gardun˜o et al., 2008). In principle, there are three possible sources of information that can guide segmentation algorithms in their task: (i) features that define an actual boundary point; (ii) features that define the inside or outside of a region; and (iii) shape information of the object to be segmented. All segmentation methods described above use only one of these information sources. A general strength of energy-based approaches is their ability to incorporate shape information. This is true for weak constraints such as boundary smoothness as well as strong constraints like adherence to an absolute shape. For many of the structures in electron tomograms, while the actual shape is not generally predictable, some distinctive geometric properties are known that can be exploited. For example, the tubular shape and distinct size of microtubules can be used in conjunction with active contours to extract them from tomograms of kinetochores ( Jiang et al., 2006). Similarly, template-based iterative boundary detection together with elliptic shape models can be used to generate high fidelity segmentations of Caulobacter crescentus cell membranes (Moussavi et al., 2010). The downside of these approaches is that the models used are highly case specific and likely need non-trivial modifications and/or adjustments for each new application. In contrast to energy-based algorithms, shape information is not easily incorporated into region-based approaches such as the watershed. While it is possible to impose weak constraints such as boundary smoothness a posterioi, the direct inclusion of more sophisticated shape models can not readily be done. A method called watersnakes (Nguyen et al., 2003) combines the watershed transform and active contours (snakes) into a region growing technique with an energy function, allowing it to include shape information. This approach was used to generate high fidelity segmentation of membranes (Nguyen and Ji, 2008). However, the current implementation requires a rough manual segmentation in the form of a number of manually traced slices to define the shape model. Model-only approaches have also been used with encouraging results using patch templates for different types of membranes

38

Niels Volkmann

(coated vs. uncoated Lebbink et al., 2007, 2010). However, these studies were limited to stained, plastic embedded sections. In summary, there is now a fair number of segmentation approaches available that were tested with electron tomographic data. However, from a practical point of view, it is hard to predict which approach has the most promising prospects to yield the desired results. In addition, segmentation performance will likely also depend on which noise-reduction approach was used, making predictions even less certain.

4. Detection and Mapping of Macromolecular Assemblies While features like membranes, vesicles and, to some extent, filaments, can be detected, identified, and extracted with above-mentioned segmentation algorithms, macromolecular assemblies need to be addressed in a more direct fashion. One way is through explicit labeling of molecules of interest but this will only give access to a subset of assemblies. The more general and more attractive way of detection is through computational methods sometimes dubbed ‘‘visual proteomics’’ (Nickell et al., 2006). The first feasibility test for detecting macromolecules in tomographic reconstructions using an algorithmic approach was done using correlationbased template matching on tomograms systems of purified thermosomes, 20S proteasomes, and GroEL, respectively (Bo¨hm et al., 2000). The results were quite encouraging with very high detection fidelity but the conditions of the specimen were far removed from the situation in cellular tomograms: cells are highly crowded with many different constituents and interacting partners. Follow-up tests were done with ‘‘phantom cells’’ (liposomes) filled with 20S proteasomes, thermosomes, or both (Frangakis et al., 2002). In this environment, results are less convincing, but still encouraging. However, this constitutes still a best case scenario where crowding is absent and the templates are not only essentially perfect but also fairly dissimilar. This type of template matching consists of using a ‘‘matched filter’’ which can be shown to be a Bayesian classifier (minimizing the probability of identification errors), as long as the template and the target are nearly identical and the noise is independent and identically distributed, Gaussian, and additive (Sigworth, 2004; van Trees, 1968). These conditions are not very well met for electron tomographic reconstructions: the noise is spatially correlated by the reconstruction process and the point-spread function; the tails of the noise distribution are often quite heavy, especially in stained samples (van der Heide et al., 2007), making the noise distribution distinctly non-Gaussian; and the uncertainty in the magnification, the potential mix of conformations, and/or the presence of stain make it difficult to obtain

Methods for Segmentation and Interpretation of Electron Tomographic Reconstructions

39

sufficiently accurate templates. As a consequence, false hits tend to be generated by this method in areas of high density such as membranes or dense vesicles when used with cellular tomograms (Ortiz et al., 2006; Rath et al., 2003). Sensitivity of the detection performance to the template definition has also been observed (Rath et al., 2003). In summary, feasibility tests with simplified systems indicate that macromolecular assemblies in the size range of 0.5–1 MDa can be identified with satisfactory fidelity using correlation-based template matching in electron cryotomograms. However, these tests were done in the absence of molecular crowding and with essentially perfect templates. Applications to actual cellular tomograms indicate that there is still room for improvement, possibly through combination with other, alternative algorithms that are more robust to the noise features and to inaccuracies in the templates.

5. Classification and Averaging It has been realized that the quality and resolution of the raw densities of macromolecular assemblies extracted from electron tomograms are generally not good enough for direct structural interpretation or meaningful docking of atomic models. In order to boost the signal to make this feasible, the motifs must be aligned, classified, and averaged. The quality of the average depends critically on the accuracy of the 3D alignment. This alignment is not only hampered by the low signal-to-noise ratio in the tomograms but also by the missing data caused by the experimental setup. In addition, variations between the motifs (e.g., different conformations) may exist and must be sorted out before averaging to avoid blurring of details. These obstacles make the task of averaging motifs extracted from electron tomograms difficult. Several approaches have been developed recently that attempt to address these issues (Bartesaghi et al., 2008; Fo¨rster et al., 2007; Schmid and Booth, 2008; Winkler et al., 2009). An alternative and fairly general way of dealing with the missing data issue is by using weighting functions and Fourier space representations of the correlation coefficient which is then used as a scoring function for classification and alignment. The Pearson correlation coefficient can be expressed without loss of generality in Fourier space (Volkmann et al., 1995). The missing data, which is expressed as an empty wedge or pyramid can easily accounted for during score calculation. We use the following mechanism for achieving this. The score is calculated according to the following formula: P w1 F1 w2 F2 cosðf1 f2 Þ ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi C ¼ qhkl ; ð2:1Þ P 2 P 2 ð w F Þ ð w F Þ 1 1 2 2 hkl hkl

40

Niels Volkmann

where F1 and F2 are the amplitudes of the Fourier coefficients of the two densities, fi are the phases and wi are weights. The sum is over all Fourier coefficients up to the resolution of the study. If all wi ¼ 1, this expression is exactly equivalent to the Pearson correlation coefficient in real space (Lunin and Woolfson, 1993). The weights can be exploited to account for the missing data. They can be set to 0 if a Fourier coefficient falls into the missing wedge and to 1 if it does not. Proper normalization is automatically taken care of. An additional advantage of this formulation is that arbitrary weights can be introduced without loss of generality. This will allow, for example, the use of fuzzy borders rather than sharp wedges, accounting for the uncertainty of the tilt angle determination or to weigh the Fourier terms in the average according to the number of contributing volumes. Application of a prototype version of this wedge-weighted correlation scheme has already proven useful in the determination of the Arp 2/3 branch junction structure, where it led to a resolution improvement from 32 to 26 A˚ (Rouiller et al., 2008). Another advantage of this scheme is that it can be equally well applied to account for missing data during correlation-based fitting of high-resolution structures into the averaged subvolumes (Volkmann, 2009; Volkmann and Hanein, 1999, 2003). Classification or sorting of different conformations is an important step for improving the quality of the averages. This can be accomplished by pairwise scoring and hierarchical ascendant cluster analysis (Fo¨rster et al., 2007; Schmid and Booth, 2008) or alignment-through-classification techniques (Bartesaghi et al., 2008; Winkler et al., 2009), both of which have their strength and weaknesses. An alternative approach for sorting conformations is based on a locally focused classification procedure that uses the density distribution within selected variance peaks of the current average as a sorting criterion. This strategy enabled the separation of two distinctly different populations in the Bovine and Yeast Arp 2/3 mediated branch junction samples that were correlated with the orientation on the sample holder (Rouiller et al., 2008). Focused variance analysis was also used successfully in two dimensions for sorting two alternative positions of Arp2-attached GFP in actin branches (Egile et al., 2005). Through approximating the variance via bootstrapping (Penczek et al., 2006b), the same idea can be applied to single-particle reconstructions (Penczek et al., 2006a).

6. Validation Some of the major questions and reoccurring themes in the interpretation of electron tomograms are: Is this the best that can be done? Does the procedure used destroy information? Does it emphasize artifacts and leads to overinterpretation? Currently, there is no obvious answer to these questions. Other than the calculations for the template matching in the clean

Methods for Segmentation and Interpretation of Electron Tomographic Reconstructions

41

macromolecular and liposome systems, all evaluations are ultimately dependent on subjective human judgment. The associated difficulties are exemplified by the large discrepancies encountered by different human operators even for the relatively simple task of delineating membranes in relatively high contrast tomograms (Gardun˜o et al., 2008; Nguyen and Ji, 2008). Furthermore, 3D scenes at the resolution of these tomograms are generally completely unfamiliar to the human eye and even in the absence of noise would likely appear rather chaotic due to the high degree of crowding in cells (Gru¨newald et al., 2003). As a consequence, the value of ‘‘ground truth’’ determination by human evaluation in this context is of questionable value. Humans are subject to all sorts of biases when evaluating visual clues (see, e.g., Harley et al., 2004; Maloney et al., 2005; Zhaoping and Jingling, 2008). While we could hope to average out individual subjectivity in scene evaluation tasks by averaging over many individuals, these species-dependent biases are not likely to be remedied by averaging. The lack of adequate ‘‘ground truths’’ data dictates that the well-developed theories of signal detection (Green and Swets, 1989) and receiver operator characteristics (Hanley and McNeil, 1982) used in the medical imaging field are not readily applicable to the field of molecular resolution electron tomography. In medical imaging, the correctness of scene evaluation (diagnosis) can be readily evaluated (patient did/did not have cancer). In summary, the perhaps most severe bottleneck for successful interpretation of molecular resolution electron tomograms is not necessarily the lack of tools; it is the lack of ability to evaluate the performance of available tools. Owing to general human biases and the alien nature of the molecular resolution scenes encountered in cellular environments, provision of handannotated data sets to establish ‘‘ground truth’’ for evaluation of algorithms is not likely to provide a satisfactory solution. Thus, the most viable path to remedy this situation may be the generation of realistic simulated data sets. However, this would not only require assembling realistic molecular scenes from atomic level structures, it would also require a push in developing appropriate imaging and noise models. Both are clearly difficult and challenging tasks. On the bright side, an added benefit of accurate noise models would be their potential usefulness in various image processing approaches that make explicit use of noise modeling such as maximum likelihood methods (Scheres et al., 2009, see also Chapter 10, Vol. 482).

ACKNOWLEDGMENTS This work was supported by NIH grants GM076503, GM066311, and the NIGMS Cell Migration Consortium. I thank Dorit Hanein for critically reading the manuscript.

42

Niels Volkmann

REFERENCES Al-Amoudi, A., Chang, J. J., Leforestier, A., McDowall, A., Salamin, L. M., Norlen, L. P., Richter, K., Blanc, N. S., Studer, D., and Dubochet, J. (2004). Cryo-electron microscopy of vitreous sections. EMBO J. 23, 3583–3588. Al-Amoudi, A., Diez, D. C., Betts, M. J., and Frangakis, A. S. (2007). The molecular architecture of cadherins in native epidermal desmosomes. Nature 450, 832–837. Auer, M., Koster, A. J., Ziese, U., Bajaj, C., Volkmann, N., Wang, D. N., and Hudspeth, A. J. (2008). Three-dimensional architecture of hair-bundle linkages revealed by electron-microscopic tomography. J. Assoc. Res. Otolaryngol. (Epub ahead of print). Bajaj, C., Yu, Z. Y., and Auer, M. (2003). Volumetric feature extraction and visualization of tomographic molecular imaging. J. Struct. Biol. 144, 132–143. Barash, D., and Comaniciu, D. (2004). A common framework for nonlinear diffusion, adaptive smoothing, bilateral filtering and mean shift. Image Vis. Comput. 22, 73–81. Bartesaghi, A., Sapiro, G., and Subramaniam, S. (2005). An energy-based three-dimensional segmentation approach for the quantitative interpretation of electron tomograms. IEEE Trans. Image Process. 14, 1314–1323. Bartesaghi, A., Sprechmann, P., Liu, J., Randall, G., Sapiro, G., and Subramaniam, S. (2008). Classification and 3D averaging with missing wedge correction in biological electron tomography. J. Struct. Biol. 162, 436–450. Baxter, W. T., Grassucci, R. A., Gao, H., and Frank, J. (2009). Determination of signal-tonoise ratios and spectral SNRs in cryo-EM low-dose imaging of molecules. J. Struct. Biol. 166, 126–132. Ben-Harush, K., Maimon, T., Patla, I., Villa, E., and Medalia, O. (2010). Visualizing cellular processes at the molecular level by cryo-electron tomography. J. Cell Sci. 123, 7–12. Best, C., Nickell, S., and Baumeister, W. (2007). Localization of protein complexes by pattern recognition. Methods Cell Biol. 79, 615–638. Beucher, S., and Meyer, F. (1993). The morphological approach to segmentation: The watershed transformation. In ‘‘Mathematical morphology in image processing,’’ (E. Dougherty, ed.), pp. 433–481. Marcel Dekker, New York. Bo¨hm, J., Frangakis, A. S., Hegerl, R., Nickell, S., Typke, D., and Baumeister, W. (2000). From the cover: Toward detecting and identifying macromolecules in a cellular context: Template matching applied to electron tomograms. Proc. Natl. Acad. Sci. USA 97, 14245–14250. Egile, C., Rouiller, I., Xu, X. P., Volkmann, N., Li, R., and Hanein, D. (2005). Mechanism of filament nucleation and branch stability revealed by the structure of the Arp2/3 complex at actin branch junctions. PLoS Biol. 3, e383. Fernandez, J. J. (2009). TOMOBFLOW: Feature-preserving noise filtering for electron tomography. BMC Bioinform. 10, 178. Fernandez, J. J., and Li, S. (2003). An improved algorithm for anisotropic nonlinear diffusion for denoising cryo-tomograms. J. Struct. Biol. 144, 152–161. Fo¨rster, F., Pruggnaller, S., Seybert, A., and Frangakis, A. S. (2007). Classification of cryoelectron sub-tomograms using constrained correlation. J. Struct. Biol. 161, 276–286. Frangakis, A. S., and Fo¨rster, F. (2004). Computational exploration of structural information from cryo-electron tomograms. Curr. Opin. Struct. Biol. 14, 325–331. Frangakis, A. S., and Hegerl, R. (2001). Noise reduction in electron tomographic reconstructions using nonlinear anisotropic diffusion. J. Struct. Biol. 135, 239–250. Frangakis, A. S., and Hegerl, R. (2002). Segmentation of two- and three-dimensional data from electron microscopy using eigenvector analysis. J. Struct. Biol. 138, 105–113. Frangakis, A. S., Stoschek, A., and Hegerl, R. (2001). Wavelet transform filtering and nonlinear anisotropic diffusion assessed for signal reconstruction performance on multidimensional biomedical data. IEEE Trans. Biomed. Eng. 48, 213–222.

Methods for Segmentation and Interpretation of Electron Tomographic Reconstructions

43

Frangakis, A. S., Bohm, J., Forster, F., Nickell, S., Nicastro, D., Typke, D., Hegerl, R., and Baumeister, W. (2002). Identification of macromolecular complexes in cryoelectron tomograms of phantom cells. Proc. Natl. Acad. Sci. USA 99, 14153–14158. Frey, T. G., Perkins, G. A., and Ellisman, M. H. (2006). Electron tomography of membrane-bound cellular organelles. Annu. Rev. Biophys. Biomol. Struct. 35, 199–224. Gardun˜o, E., Wong-Barnum, M., Volkmann, N., and Ellisman, M. H. (2008). Segmentation of electron tomographic data sets using fuzzy set theory principles. J. Struct. Biol. 162, 368–379. Goddard, T. D., Huang, C. C., and Ferrin, T. E. (2007). Visualizing density maps with UCSF Chimera. J. Struct. Biol. 157, 281–287. Green, D. M., and Swets, J. A. (1989). Signal Detection Theory and Psychophysics. Peninsula Publishing, Los Altos, CA. Grimm, R., Typke, D., Barmann, M., and Baumeister, W. (1996). Determination of the inelastic mean free path in ice by examination of tilted vesicles and automated most probable loss imaging. Ultramicroscopy 63, 169–179. Gru¨newald, K., Medalia, O., Gross, A., Steven, A. C., and Baumeister, W. (2003). Prospects of electron cryotomography to visualize macromolecular complexes inside cellular compartments: Implications of crowding. Biophys. Chem. 100, 577–591. Gruska, M., Medalia, O., Baumeister, W., and Leis, A. (2008). Electron tomography of vitreous sections from cultured mammalian cells. J. Struct. Biol. 161, 384–392. Hanley, J. A., and McNeil, B. J. (1982). The meaning and use of the area under a receiver operating characteristic (ROC) curve. Radiology 143, 29–36. Harley, E. M., Carlsen, K. A., and Loftus, G. R. (2004). The ‘‘saw-it-all-along’’ effect: Demonstrations of visual hindsight bias. J. Exp. Psychol. Learn. Mem. Cogn. 30, 960–968. Hsieh, C. E., Leith, A., Mannella, C. A., Frank, J., and Marko, M. (2006). Towards highresolution three-dimensional imaging of native mammalian tissue: Electron tomography of frozen-hydrated rat liver sections. J. Struct. Biol. 153, 1–13. Iancu, C. V., Wright, E. R., Heymann, J. B., and Jensen, G. J. (2006). A comparison of liquid nitrogen and liquid helium as cryogens for electron cryotomography. J. Struct. Biol. 153, 231–240. Janssen, M. E., Kim, E., Liu, H., Fujimoto, L. M., Bobkov, A., Volkmann, N., and Hanein, D. (2006). Three-dimensional structure of vinculin bound to actin filaments. Mol. Cell 21, 271–281. Jiang, W., Baker, M. L., Wu, Q., Bajaj, C., and Chiu, W. (2003). Applications of a bilateral denoising filter in biological electron microscopy. J. Struct. Biol. 144, 114–122. Jiang, M., Ji, Q., and McEwen, B. F. (2006). Model-based automated extraction of microtubules from electron tomography volume. IEEE Trans. Inf. Technol. Biomed. 10, 608–617. Kimmel, R., Malladi, R., and Sochen, N. A. (2000). Images as embedded maps and minimal surfaces: Movies, color, texture, and volumetric medical images. Int. J. Comput. Vis. 39, 111–129. Koning, R. I., and Koster, A. J. (2009). Cryo-electron tomography in biology and medicine. Ann. Anat. 191, 427–445. Kremer, J. R., Mastronarde, D. N., and McIntosh, J. R. (1996). Computer visualization of three-dimensional image data using IMOD. J. Struct. Biol. 116, 71–76. Lebbink, M. N., Geerts, W. J., van der Krift, T. P., Bouwhuis, M., Hertzberger, L. O., Verkleij, A. J., and Koster, A. J. (2007). Template matching as a tool for annotation of tomograms of stained biological structures. J. Struct. Biol. 158, 327–335. Lebbink, M. N., Jimenez, N., Vocking, K., Hekking, L. H., Verkleij, A. J., and Post, J. A. (2010). Spiral coating of the endothelial caveolar membranes as revealed by electron tomography and template matching. Traffic 11, 138–150.

44

Niels Volkmann

Leis, A. P., Beck, M., Gruska, M., Best, C., Hegerl, R., Baumeister, W., and Leis, J. W. (2006). Cryo-electron tomography of biological specimens. IEEE Signal Process. Mag. 23, 95–103. Leis, A., Rockel, B., Andrees, L., and Baumeister, W. (2009). Visualizing cells at the nanoscale. Trends Biochem. Sci. 34, 60–70. Li, Y., Leith, A., and Frank, J. (1997). Tinkerbell—A tool for interactive segmentation of 3D data. J. Struct. Biol. 120, 266–275. Lunin, V. Y., and Woolfson, M. M. (1993). Mean phase error and the map-correlation coefficient. Acta Cryst. D49, 530–535. Luther, P. K., Lawrence, M. C., and Crowther, R. A. (1988). A method for monitoring the collapse of plastic sections as a function of electron dose. Ultramicroscopy 24, 7–18. Maloney, L. T., Dal Martello, M. F., Sahm, C., and Spillmann, L. (2005). Past trials influence perception of ambiguous motion quartets through pattern completion. Proc. Natl. Acad. Sci. USA 102, 3164–3169. Marsh, B. J. (2005). Lessons from tomographic studies of the mammalian Golgi. Biochim. Biophys. Acta. Mol. Cell Res. 1744, 273–292. Marsh, B. J., Mastronarde, D. N., Buttle, K. F., Howell, K. E., and McIntosh, J. R. (2001). Organellar relationships in the Golgi region of the pancreatic beta cell line, HIT-T15, visualized by high resolution electron tomography. Proc. Natl. Acad. Sci. USA 98, 2399–2406. Marsh, B. J., Volkmann, N., McIntosh, J. R., and Howell, K. E. (2004). Direct continuities between cisternae at different levels of the Golgi complex in glucose-stimulated mouse islet beta cells. Proc. Natl. Acad. Sci. USA 101, 5565–5570. Mastronarde, D. N. (1997). Dual-axis tomography: An approach with alignment methods that preserve resolution. J. Struct. Biol. 120, 343–352. McEwen, B. F., and Marko, M. (2001). The emergence of electron tomography as an important tool for investigating cellular ultrastructure. J. Histochem. Cytochem. 49, 553–564. McEwen, B. F., Downing, K. H., and Glaeser, R. M. (1995). The relevance of dosefractionation in tomography of radiation-sensitive specimens. Ultramicroscopy 60, 357–373. Moussavi, F., Heitz, G., Amat, F., Comolli, L. R., Koller, D., and Horowitz, M. (2010). 3D segmentation of cell boundaries from whole cell cryogenic electron tomography volumes. J. Struct. Biol. 170, 134–145. Nguyen, H., and Ji, Q. (2008). Shape-driven three-dimensional watersnake segmentation of biological membranes in electron tomography. IEEE Trans. Med. Imaging 27, 616–628. Nguyen, H. T., Worring, M., and van den Boomgaard, R. (2003). Watersnakes: Energydriven watershed segmentation. IEEE Trans. Pattern Anal. Mach. Intell. 25, 330–342. Nickell, S., Hegerl, R., Baumeister, W., and Rachel, R. (2003). Pyrodictium cannulae enter the periplasmic space but do not enter the cytoplasm, as revealed by cryo-electron tomography. J. Struct. Biol. 141, 34–42. Nickell, S., Kofler, C., Leis, A. P., and Baumeister, W. (2006). A visual approach to proteomics. Nat. Rev. Mol. Cell Biol. 7, 225–230. Noske, A. B., Costin, A. J., Morgan, G. P., and Marsh, B. J. (2008). Expedited approaches to whole cell electron tomography and organelle mark-up in situ in high-pressure frozen pancreatic islets. J. Struct. Biol. 161, 298–313. Ortiz, J. O., Forster, F., Kurner, J., Linaroudis, A. A., and Baumeister, W. (2006). Mapping 70S ribosomes in intact cells by cryoelectron tomography and pattern recognition. J. Struct. Biol. 156, 334–341. Osher, S., and Sethian, J. A. (1988). Fronts propagating with curvature-dependent speed— Algorithms based on Hamilton-Jacobi formulations. J. Comput. Phys. 79, 12–29.

Methods for Segmentation and Interpretation of Electron Tomographic Reconstructions

45

Pantelic, R. S., Rothnagel, R., Huang, C. Y., Muller, D., Woolford, D., Landsberg, M. J., McDowall, A., Pailthorpe, B., Young, P. R., Banks, J., Hankamer, B., and Ericksson, G. (2006). The discriminative bilateral filter: An enhanced denoising filter for electron microscopy data. J. Struct. Biol. 155, 395–408. Penczek, P., Marko, M., Buttle, K., and Frank, J. (1995). Double-tilt electron tomography. Ultramicroscopy 60, 393–410. Penczek, P. A., Frank, J., and Spahn, C. M. (2006a). A method of focused classification, based on the bootstrap 3D variance analysis, and its application to EF-G-dependent translocation. J. Struct. Biol. 154, 184–194. Penczek, P. A., Yang, C., Frank, J., and Spahn, C. M. (2006b). Estimation of variance in single-particle reconstruction using the bootstrap technique. J. Struct. Biol. 154, 168–183. Pierson, J., Fernandez, J. J., Bos, E., Amini, S., Gnaegi, H., Vos, M., Bel, B., Adolfsen, F., Carrascosa, J. L., and Peters, P. J. (2010). Improving the technique of vitreous cryosectioning for cryo-electron tomography: Electrostatic charging for section attachment and implementation of an anti-contamination glove box. J. Struct. Biol. 169, 219–225. Pintilie, G., Zhang, J., Chiu, W., and Gossard, D. (2009). Identifying components in 3D density maps of protein nanomachines by multi-scale segmentation. IEEE NIH Life Sci. Syst. Appl. Workshop 2009, 44–47. Pruggnaller, S., Mayr, M., and Frangakis, A. S. (2008). A visualization and segmentation toolbox for electron microscopy. J. Struct. Biol. 164, 161–165. ¨ ber die Bestimmung von Funktionen durch ihre Integralwerte la¨ngs Radon, J. (1917). U gewisser Manningfaltigkeiten. Math. Phys. Klasse 69, 262–277. Rath, B. K., Hegerl, R., Leith, A., Shaikh, T. R., Wagenknecht, T., and Frank, J. (2003). Fast 3D motif search of EM density maps using a locally normalized cross-correlation function. J. Struct. Biol. 144, 95–103. Ress, D. B., Harlow, M. L., Marshall, R. M., and McMahan, U. J. (2004). Methods for generating high-resolution structural models from electron microscope tomography data. Structure 12, 1763–1774. Robinson, C. V., Sali, A., and Baumeister, W. (2007). The molecular sociology of the cell. Nature 450, 973–982. Rouiller, I., Xu, X. P., Amann, K. J., Egile, C., Nickell, S., Nicastro, D., Li, R., Pollard, T. D., Volkmann, N., and Hanein, D. (2008). The structural basis of actin filament branching by Arp2/3 complex. J. Cell Biol. 180, 887–895. Salje, J., Zuber, B., and Lowe, J. (2009). Electron cryomicroscopy of E. coli reveals filament bundles involved in plasmid DNA segregation. Science 323, 509–512. Salvi, E., Cantele, F., Zampighi, L., Fain, N., Pigino, G., Zampighi, G., and Lanzavecchia, S. (2008). JUST (Java User Segmentation Tool) for semi-automatic segmentation of tomographic maps. J. Struct. Biol. 161, 287–297. Sandberg, K. (2007). Methods for image segmentation in cellular tomography. Methods Cell Biol. 79, 769–798. Sandberg, K., and Brega, M. (2007). Segmentation of thin structures in electron micrographs using orientation fields. J. Struct. Biol. 157, 403–415. Scheres, S. H., Nunez-Ramirez, R., Gomez-Llorente, Y., San Martin, C., Eggermont, P. P., and Carazo, J. M. (2007). Modeling experimental image formation for likelihood-based classification of electron microscopy data. Structure 15, 1167–1177. Scheres, S. H., Melero, R., Valle, M., and Carazo, J. M. (2009). Averaging of electron subtomograms and random conical tilt reconstructions through likelihood optimization. Structure 17, 1563–1572. Schietroma, C., Fain, N., Zampighi, L. M., Lanzavecchia, S., and Zampighi, G. A. (2009). The structure of the cytoplasm of lens fibers as determined by conical tomography. Exp. Eye Res. 88, 566–574.

46

Niels Volkmann

Schmid, M. F., and Booth, C. R. (2008). Methods for aligning and for averaging 3D volumes with missing data. J. Struct. Biol. 161, 243–248. Sigworth, F. J. (2004). Classical detection theory and the cryo-EM particle selection problem. J. Struct. Biol. 145, 111–122. Stalling, D., Hege, H.-C., and Westerhoff, M. (2004). Amira—A highly interactive system for visual data analysis. In ‘‘Visualization Handbook,’’ (C. R. Johnson and C. D. Hansen, eds.), Academic Press, Orlando, FL, USA. Stoschek, A., and Hegerl, R. (1997). Denoising of electron tomographic reconstructions using multiscale transformations. J. Struct. Biol. 120, 257–265. Tocheva, E. I., Li, Z., and Jensen, G. J. (2010). Electron cryotomography. Cold Spring Harb. Perspect. Biol. 2, a003442. Tomasi, C., and Manduchi, R. (1998). Bilateral filtering for gray and color images. Proc. ICCV ‘98 839–846. van der Heide, P., Xu, X. P., Marsh, B. J., Hanein, D., and Volkmann, N. (2007). Efficient automatic noise reduction of electron tomographic reconstructions based on iterative median filtering. J. Struct. Biol. 158, 196–204. van Trees, H. L. (1968). Detection, Estimation, and Modulation Theory. Wiley, New York. Vincent, L., and Soille, P. (1991). Watersheds in digital space: An efficient algorithm based on immersion simulations. IEEE Trans. Pattern Anal. Mach. Intell. 13, 583–598. Volkmann, N. (2002). A novel three-dimensional variant of the watershed transform for segmentation of electron density maps. J. Struct. Biol. 138, 123. Volkmann, N. (2009). Confidence intervals for fitting of atomic models into low-resolution densities. Acta Crystallogr. D. Biol. Crystallogr. 65, 679–689. Volkmann, N., and Hanein, D. (2009). Electron microscopy in the context of systems biology. In ‘‘Structural Bioinformatics, 2nd edition,’’ (J. Gu and P. Bourne, eds.), pp. 143–170. Wiley-Blackwell, New York. Volkmann, N., and Hanein, D. (1999). Quantitative fitting of atomic models into observed densities derived by electron microscopy. J. Struct. Biol. 125, 176–184. Volkmann, N., and Hanein, D. (2003). Docking of atomic models into reconstructions from electron microscopy. Methods Enzymol. 374, 204–225. Volkmann, N., Schlu¨nzen, F., Vernoslava, E. A., Urzhumstev, A. G., Podjarny, A. D., Roth, M., Pebay-Peyroula, E., Berkovitch-Yellin, Z., Zaytzev-Bashan, A., and Yonath, A. (1995). On ab initio phasing of ribosomal particles at very low resolution. Joint CCP4 ESF-EACBM Newslett. 31. Whitaker, R. T., and Elangovan, V. (2002). A direct approach to estimating surfaces in tomographic data. Med. Image Anal. 6, 235–249. Winkler, H., Zhu, P., Liu, J., Ye, F., Roux, K. H., and Taylor, K. A. (2009). Tomographic subvolume alignment and subvolume classification applied to myosin V and SIV envelope spikes. J. Struct. Biol. 165, 64–77. Zhaoping, L., and Jingling, L. (2008). Filling-in and suppression of visual perception from context: A Bayesian account of perceptual biases by contextual influences. PLoS Comput. Biol. 4, e14.

C H A P T E R

T H R E E

Integration of Cryo-EM with Atomic and Protein–Protein Interaction Data ¨rster and Elizabeth Villa Friedrich Fo Contents 1. Introduction 2. The Problem of Placing Assembly Subunits into Cryo-EM Maps 3. Structure Prediction of Subunits 3.1. Sequence alignment of target and template and finding appropriate structural templates 3.2. Model building 3.3. Building a structural model for the proteasomal AAA-ATPase hexamer 4. Protein–Protein Interaction Data 4.1. State-of-the art protein–protein experimental methods 4.2. Emerging experimental methods 4.3. Computational interaction and interface prediction 4.4. Gathering interproteasomal interactions 5. Model Building of a Complex Using Cryo-EM and Additional Data 5.1. Assembly representation 5.2. Scoring of assemblies 5.3. Building and assessment of AAA-ATPase models 6. Refinement of Atomic Models Using High-Resolution Maps 6.1. Molecular dynamics flexible fitting 6.2. Refinement of a comparative model of the 20S core particle 7. Conclusion and Outlook References

48 49 50 51 53 53 54 56 57 57 58 58 59 60 63 64 65 66 68 69

Abstract Cryoelectron microscopy (cryo-EM) is an increasingly popular method to elucidate the structures of macromolecular complexes. However, in many applications the resolution of cryo-EM densities is limited to the low or intermediate resolution regime, that is, (10 A˚) 1 or worse. Therefore, unambiguous molecular interpretation of cryo-EM densities requires efficient use of additional information, such as atomic structures of related subunits and protein–protein Max-Planck Institute of Biochemistry, Department of Structural Biology, Martinsried, Germany Methods in Enzymology, Volume 483 ISSN 0076-6879, DOI: 10.1016/S0076-6879(10)83003-4

#

2010 Elsevier Inc. All rights reserved.

47

48

¨rster and Elizabeth Villa Friedrich Fo

interaction data. Here, we describe how information from different sources can be combined to determine the approximate molecular architecture of complexes. Molecular dynamics based flexible fitting protocols allow subsequent refinement of the atomistic models.

1. Introduction Cryoelectron microscopy (cryo-EM) is an increasingly popular method to elucidate the structures of macromolecular complexes. While X-ray crystallography and NMR spectroscopy are typically superior in resolution (both methods provide atomic models of the complex under scrutiny), the advantage of cryo-EM is its versatility. Compared to X-ray crystallography and NMR, cryo-EM is substantially less demanding in terms of sample amount, concentration, purity, and homogeneity. Thus, cryo-EM is particularly useful for obtaining structural insights into those complexes that are difficult to purify in high amounts, such as transient assemblies, membrane-associated protein complexes, and structurally heterogeneous macromolecules. Single-particle analysis (SPA) and cryoelectron tomography (CET) are the cryo-EM methods of choice to study biochemically delicate macromolecular complexes. Both methods are restricted to relatively large assemblies; typically 250 kDa and larger, but continuous developments in hard- and software gradually reduce this limit (see also Vol. 482, Chapter 8). In many cases, structural heterogeneity of the assembly under scrutiny is nonnegligible. Then, the single-particle data need to be sorted into different bins according to the different conformers (see also Vol. 482, Chapters 11 and 12). Whereas in some cases sorting of single particles according to specific features yields high-resolution reconstructions, most notably for different states of the ribosome (Becker et al., 2009; Connell et al., 2007), structural heterogeneity limits the resolution to the medium or even low regime in most studies, that is, (10– ˚ ) 1, respectively. 20 A˚) 1 or worse than (20 A CET is probably the most straightforward method to explore the structures of macromolecules associated to their native membranes. The method is capable of imaging pleiomorphic objects, such as virions or organelles, in 3D. When applying criteria from SPA, the resolution of cryoelectron ˚ ) 1 (Gru¨newald et al., 2003). However, tomograms does not exceed (50 A the resolution can be increased by averaging subvolumes containing the same type of macromolecule, analogous to SPA (Fo¨rster et al., 2005). Using this approach, membrane-associated complexes can be resolved to ˚ ) 1 (Fo¨rster and Hegerl, 2007). (20–30 A Molecular interpretation of cryo-EM maps at low or medium resolutions is a challenge for a number of reasons: (i) Atomic models typically cannot be fitted into low or medium resolution maps precisely. As a rule of

Integration of Cryo-EM with Atomic and Protein–Protein Interaction Data

49

˚ ) 1 fits will be ambiguous. Nevertheless, thumb, below a resolution of (20 A this criterion depends ultimately on the size of the model: large models, for example whole subcomplexes, may often be fitted unambiguously at lower resolutions, while fitting of small fragments such as helices will require at least subnanometer-resolutions (Section 6). (ii) Accurate atomic models of many components may not be accessible. Many complexes studied by cryo-EM fail to crystallize due to structural flexibility of the subunits. For example, solenoid folds, which are present in many large eukaryotic complexes, have a substantial degree of structural heterogeneity to accommodate binding to different surfaces (Brohawn et al., 2009). (iii) Many complexes consist of evolutionary related subunits. Thus, the corresponding structures of these subunits are similar to each other, making it extremely hard to position these subunits based on geometric data only. Here, we describe how to build models of assemblies using varied sources of data. The structures of many proteins or some of their domains can be predicted by comparative modeling based on known atomic structures. Using subunit models, EM maps, and protein–protein interaction data, the approximate quaternary structure of the subunits can be modeled. When the EM maps are of high resolution, the atomic models can be refined using flexible fitting methods.

2. The Problem of Placing Assembly Subunits into Cryo-EM Maps As a prototypical application for interpretation of a cryo-EM map, we study the 26S proteasome. The 26S proteasome consists of the cylindershaped 20S core particle (CP), which is solved to atomic detail, and the regulatory particles (RPs), which associate to both cylinder ends (Fo¨rster et al., 2010). The RP possesses a high degree of inherent structural variability, which makes it hard to obtain high-resolution insights into the fully assem˚ ) 1 of the bled 26S holocomplex. To date, the best-resolved structure (20 A 26S proteasome has been obtained using cryo-EM SPA (Nickell et al., 2009). The RP contains six AAA-ATPases, Rpt1–6, which share a high degree of sequence similarity (>45%; Finley et al., 1998). AAA-ATPases typically assemble into hexameric rings (Vale, 2000). Thus, we expect that the six ATPase subunits assemble to a ring, but we do not know the topology of the six subunits within the ring. From the 26S proteasome EM map a density can be segmented consisting of two rings: the upper ring exhibits approximate threefold rotational symmetry, whereas the lower one is approximately sixfold symmetrical (Fig. 3.1A). The segmented density is placed at the cylinder ends, where numerous protein–protein interactions suggest that the AAA-ATPase is expected. Moreover, the estimated mass from

50

¨rster and Elizabeth Villa Friedrich Fo

A

B Residue index 1

200

100 3h43_A

300

400

2ce7_A 2dhr_A 1iy2_A 1lv7_A 3h4m_A 1ixz_A

Figure 3.1 (A) Threefold rotationally symmetrized EM map of the proteasomal AAAATPases Rpt1–6. The upper ring is approximately threefold symmetrical, whereas the lower ring is approximately sixfold symmetrical. (B) Template search for Drosophila melanogaster Rpt1 using HHpred. The search results in one template covering residues 90–145 (3h43A) and several templates covering its AAA-ATPase domain (2ce7A, 2dhrA, 1iy2A, 1lv7A, 3h4mA, 1ixzA) ranging approximately from Rpt1 residue 150–420.

the extracted density (300 kDa) matches the mass of the Rpt1–6 (Nickell et al., 2009). Thus, the segmented density almost certainly corresponds to the AAA-ATPase hexamer. While the resolution of 20 A˚ is sufficient to suggest quaternary structures, it is by no means sufficient to discern the different subunits, which are presumably highly similar to each other, as suggested by the high sequence identity. Thus, inferring the subunit topology can only be accomplished by a hybrid approach. We describe how atomic models of the respective subunits are obtained and then fitted into the EM map in different topologies. These different candidate models are then assessed based on protein–protein interaction data. When high-resolution EM maps are available, the atomic coordinates can be refined further using molecular dynamics (MD) simulations.

3. Structure Prediction of Subunits Structure prediction of proteins is the task of predicting a protein’s 3D structure from its amino acid sequence. The most widely applicable approach to protein structure prediction is “comparative modeling”: the structure of a protein is predicted based on features that are observed in other protein structures (templates; Marti-Renom et al., 2000). This method is sometimes also referred to as “homology modeling” because the templates are usually derived from homologs. Alternative methods

Integration of Cryo-EM with Atomic and Protein–Protein Interaction Data

51

predict the protein structures without templates (de novo). While de novo structure prediction methods are increasingly successful, in particular the ROSETTA program (Schueler-Furman et al., 2005), the methodology is still confined to relatively short proteins (150 amino acids and less), which is prohibitive for many applications in cryo-EM modeling. In comparative modeling, one structural template or multiple templates are chosen to build a model for the “target” sequence. Such features borrowed from the template are, for example, distances between corresponding atom pairs in the backbones of templates. Specifically, the distances between individual atoms in the template are determined and then imposed as “restraints” on the model. For a detailed description of comparative modeling, we refer to the more detailed book chapters, for example, Eswar and Sali (2007), Eswar et al. (2007).

3.1. Sequence alignment of target and template and finding appropriate structural templates The accuracy of any comparative model will be determined by the correctness of the imposed restraints, and hence the similarity of the templates to the (unknown) structure of the target. Thus, the foremost requirement for building an accurate atomic model of a protein is the selection of appropriate templates. Moreover, the amino acid correspondences (sequence alignment) of target and template must be as precise as possible. Both problems are coupled and are therefore discussed in one section. Table 3.1 lists some of the respective web servers, which we find useful for our research. 3.1.1. Sequence based methods The most straightforward way to compare the sequence of the target to another protein is a pairwise alignment of the target and template sequences. For example, the well-known Basic Local Alignment Search Tool (BLAST) is based on pairwise sequence–sequence comparison. Pairwise alignment methods produce largely correct sequence alignments when the sequence identity of the respective sequences exceeds 30–40%. Accordingly, these methods are successful in identifying (structural) homologs for these cases. However, below 30% sequence identity, these algorithms perform poorly in detecting homologs (Brenner et al., 1998). More sensitive methods use multiple sequences for sequence alignment. In sequence-profile alignment, the target sequence is compared to multiple prealigned sequences, that is, the “profile.” Probably, the most popular variant of this class is Position-Specific Iterated (PSI)-BLAST. Another way to make use of multiple templates in the alignment is building a Hidden Markov Model (HMM) of the templates. The additional computational effort of sequence-profile and sequence-HMM alignment compared to

52

¨rster and Elizabeth Villa Friedrich Fo

Table 3.1 Servers for sequence alignment and detection of structural templates Server

Web address

Method

BLAST

http://blast.ncbi.nlm.nih. Pairwise sequence alignment gov/Blast.cgi of target and proteins from PDB or other databases CLUSTALW2 http://www.ebi.ac.uk/ Multiple-sequence alignment Tools/clustalw2/ MUSCLE http://www.drive5.com/ Multiple-sequence alignment muscle/ T-COFFEE http://tcoffee.vital-it.ch/ Multiple-sequence alignment; optionally builds a consensus alignment using different methods PSI-BLAST http://blast.ncbi.nlm.nih. Sequence-profile alignment of target and proteins from gov/Blast.cgi PDB or other databases http://www.ebi.ac.uk/ Tools/psiblast/ HMM–HMM alignment of HHpred http://toolkit.lmb.unitarget and proteins from muenchen.de/hhpred PDB http://toolkit.tuebingen. mpg.de/hhpred FUGUE http://tardis.nibio.go.jp/ Threading-based homolog fugue/prfsearch.html recognition GenTHREADER http://bioinf4.cs.ucl.ac. Threading-based homolog uk:3000/psipred/ recognition

The field is developing rapidly and the list is by no means comprehensive. For more extensive coverage of web servers we refer to Eswar and Sali (2007).

sequence–sequence alignment pays off in retrieving correct homologs, in particular, at low sequence identities. The most sensitive methods employ multiple sequences instead of a single target sequence. Again, the methods can be categorized into profile–profilebased alignment and HMM–HMM-based alignment. For example, HHpred is a server for detecting appropriate structural templates from the PDB using HMM–HMM comparison (Soding, 2005; Soding et al., 2005). 3.1.2. Threading Another method for detecting remote homologs makes use of template structures. A coarse model of the target is built based on the target structure, typically by “threading” the target residues on the template structure, which is then assessed by a statistical potential. Popular softwares are FUGUE (Shi et al., 2001) and GenTHREADER ( Jones, 1999; Table 3.1).

Integration of Cryo-EM with Atomic and Protein–Protein Interaction Data

53

3.2. Model building In essence, the comparative target model is built using “features” observed in the template, for example, the distances between the Ca atoms of two residues, and generic features observed in protein structures, for example, bond angles, as restraints. Input for comparative modeling software, such as MODELLER (Sali and Blundell, 1993) are the structures of templates and the alignment of the target sequence and the templates. Optimization algorithms produce models, where the target features are as similar as possible to the corresponding template features. Thus, by definition, comparative modeling requires that the template(s) cover all parts of the modeled target. The generic restraints based on molecular force fields only allow modeling very short segments ( R1 þ R2. The half width of the Gaussian reflects the accuracy of the experiment and the coordinates. (D) As a score S, we use the negative logarithm of P.

Integration of Cryo-EM with Atomic and Protein–Protein Interaction Data

63

and only the two resulting distances are used for subsequent scoring. In essence, a model is penalized when the two distances exceed the sum of the radii of the interacting particles (Fig. 3.4C and D).

5.3. Building and assessment of AAA-ATPase models For the example of the proteasomal AAA-ATPases, we use the EM data and the interaction restraints sequentially because they provide complementary information. From the EM data, we can deduce the possible positions of the subunits (Fig. 3.1A). Since the map and the template are threefold rotationally symmetrical, the EM data do not provide any information for inferring the subunit topology. In contrast, the interaction data contain information on subunit topology, but due to the low resolution it does provide information for subunit positioning. First, we fit the PAN templates into the EM map using UCSF Chimera (Goddard et al., 2007): the rings are first manually positioned and then refined using the Fitting tool (Fig. 3.5A). The threefold symmetrical N-ring fits neatly into the threefold symmetrical upper ring of the EM map. The sixfold symmetrical AAA-ring can be positioned in two different ways into the approximately sixfold symmetrical lower ring: the Ca–Ca distances between the N-PAN C-terminus (aa 149) and the AAA-PAN N-terminus (aa 158) measure 25 and 26 A˚, respectively, and they can be bridged by the eight residues in both arrangements, whereas the distances to the other four subunits are too large. We then shuffle the order of the subunits by varying the alignment file for model building (see above). The subunits Rpt2, Rpt3, and Rpt5 are confined to the cis positions (Fig. 3.5B), according to the aforementioned structural argument. Thus, we have 60 different subunit topologies for the Rpt hexamer: 30 different subunit topologies for the N-ring, and for each

A

B

C

Figure 3.5 (A) N-PAN (green, cyan) and AAA-PAN (blue). (B) N-PAN hexamer. cis- (cyan) and trans-positions (green) of the N-PAN monomers in the N-ring. (C) Bestscoring configuration of comparative models (Rpt1/Rpt2/Rpt6/Rpt3/Rpt4/Rpt5) fitted into the EM map.

64

¨rster and Elizabeth Villa Friedrich Fo

N-ring topology two different options to position the AAA-ring (rotated by 60 with respect to each other). We score the 60 different subunits based on the number of violated contacts using the subcomplex restraint. Specifically, we used subcomplex data of the Rpt subunits obtained by chemical cross-linking (8 restraints), two-hybrid assays (6), in vitro binding assays (4), and coexpression (1) (Table 1 in Fo¨rster et al., 2009). We consider a restraint violated if the closest distance between two atoms from interacting proteins exceeds the standard ˚ , which we estimate to be the accuracy of the comparative deviation 2.5 A models given the 40% sequence identity between targets and template (Eswar et al., 2007). The scoring is performed in IMP, which we interface by our python library sphIMP (Fo¨rster and Lasker; www.biochem.mpg.de/ foerster). A single N-ring topology (Rpt1/Rpt2/Rpt6/Rpt3/Rpt4/Rpt5) violates a minimum of two restraints. The second-best scoring solution (Rpt1/Rpt2/ Rpt4/Rpt5/Rpt6/Rpt3) violates three restraints. The protein–protein interaction data cannot discriminate between the two corresponding AAA-ring placements. Recently, the ring order of the proteasomal AAA-ATPases has been confirmed experimentally to be Rpt1/Rpt2/Rpt6/Rpt3/Rpt4/Rpt5 (Tomko et al., 2010). We can find the most probable position of the AAA-ring in the context of the 26S proteasome using the EM map of the entire 26S proteasome (Fo¨rster et al., 2009): the 20S CP can be placed unambiguously into the map and the six different orientations of the AAA-ring (each rotated by 60 ) can be scored using the protein–protein interactions of CP and AAA-ATPase, as described for the AAA-ATPase topology. Using the AAA-ring position, which results in minimum violations, we obtain a model for the quaternary structure of the whole AAA-ATPase (Fig. 3.5C). However, the quaternary structure of AAA-ring and N-ring is based on only three CP-AAA restraints and it will need to be solidified by further data.

6. Refinement of Atomic Models Using High-Resolution Maps When the macromolecule under scrutiny is structurally homogeneous, cryo-EM maps may reach subnanometer-resolution (see also Vol. 482, Chapter 8). At subnanometer-resolution, secondary structure elements of proteins are often discernible from the density. Atoms may be positioned into an EM map with an accuracy greatly exceeding the resolution of the EM map, similar to X-ray crystallography model building. In the latter, the amino acids are not necessarily discernible from the density, but their chemical structure and preferred isomers are inferred from a priori knowledge

Integration of Cryo-EM with Atomic and Protein–Protein Interaction Data

65

(Brunger et al., 1987). In flexible fitting, the atom positions of an input model (e.g., a crystal structure of the protein in a different context or a comparative model) are refined using the EM data. Flexible fitting is not only useful to study structural changes of proteins, but also permits the accurate segmentation of an EM map (Seidelt et al., 2009). Here, we describe the flexible fitting of models in EM maps using molecular dynamics flexible fitting (MDFF). Since the whole 26S proteasome is not available at high resolution yet, we illustrate the protocol for refining an atomic model of the 20S CP.

6.1. Molecular dynamics flexible fitting MD simulations have been enormously successful in X-ray crystallographic refinement (Brunger et al., 1987) and can likewise be employed for EM fitting. MD simulations are based on force fields that define the interactions between all atoms in a protein complex (Karplus and McCammon, 2002). Thus, MD preserves the correct stereochemistry of the protein complex, while morphing the protein according to its intrinsic flexibility. MD can be extended to incorporate other sources of data; notably, 3D EM density maps can be incorporated as an additional term in the force field that acts on the atomic model by driving atoms into high-density areas following the steepest descent of the 3D density (Fig. 3.6A; Trabuco et al., 2008). It is of utmost importance to note that flexibly fitting atomic models into EM maps bear the risk of overfitting; when the resolution of the EM data is too low, the ratio of unknown coordinates versus data points is poor. Moreover, EM maps, in particular, at moderate resolutions, may contain errors due to flexibility of the protein complex, but also artifacts introduced by imaging and processing. In general, the risk of overfitting is extremely high when the resolution is worse than (10 A˚) 1. It should also be kept in mind that resolution is a global measure (see also Vol. 482, Chapter 3) and that the quality of the map may be locally substantially worse due to structural heterogeneity. To avoid overfitting, further restraints can be added to the force field in the form of springs to preserve the relative positions between pairs of atoms (distances or “bonds”), triads of atoms (angles), or between four atoms (dihedrals). MDFF imposes restraints to dihedral angles of secondary structure elements in proteins, that is, alpha helices and beta strands. The MDFF simulations yield an ensemble of structures that conform to the EM data, which are representative of the uncertainty of the models. The quality of the fit is monitored by CCC along the simulation. When the value of this metric has equilibrated, all the conformations visited are likely to be present in the original vitrified sample. The deviation of the different conformations is a measure for the precision of the fitted model. The interpretation of flexibly fitted models must be done taking into account

66

¨rster and Elizabeth Villa Friedrich Fo

A

3.0 2.5 2.0 1.5 1.0 0 kcal/mol

B

3.5 3.0 2.5 2.0 1.5 1.0 0.5 0 nm

Figure 3.6 Molecular dynamics flexible fitting. (A) Cross-section of a 2D slice of the potential derived from the EM density of elongation factor Tu (EMD-5036) represented as a contour plot. Arrows represent forces driving the atomic structure toward high-density regions. The circular areas correspond to cross-sections of alpha helices. (B) We use carbon monoxide dehydrogenase to illustrate MDFF: carbon monoxide dehydrogenase adopts two different conformations in the same crystal (PDB code: ˚ ) 1 1OAO). The EM density of the first structure (gray ribbon) is simulated to (10 A resolution (gray mesh). We use the second crystal structure (left; color scheme corresponds to root-mean-squared deviation to the target structure) as a starting model for flexible fitting an atomic model to the EM density. The fitted model (right) exhibits dramatically reduced RMSD deviation.

the quality of the initial atomic model and the EM data. Figure 3.6B and C shows an example of MDFF to model the conformational change of AcetylCoA synthase. A detailed tutorial of MDFF can be found on http://www. ks.uiuc.edu/Research/mdff/.

6.2. Refinement of a comparative model of the 20S core particle Spanning the central region of the 26S proteasome, the 20S CP displays little structural variability compared to the RP. Therefore, the 20S CP can be resolved to subnanometer resolution comparably easily. Homologs of the eukaryotic 20S CP can be found in bacteria and archea, where they are activated by complexes different from the eukaryotic RP (Striebel et al., 2009). 20S CPs from all three domains of life have a common overall

Integration of Cryo-EM with Atomic and Protein–Protein Interaction Data

67

architecture (Voges et al., 1999): they are built of 28 subunits arranged in a stack of four seven-membered rings. All CPs possess a twofold symmetry axis; the two polar rings and the two inner rings are identical, respectively. In its simplest form, for example, in the archeon Thermoplasma acidophilum, the seven CP subunits in the polar ring (a-subunits) and in the inner ring (b-subunits) are identical. In eukaryotes, both rings consist of seven different subunits each, which have typically 30% identical sequences. The crystallographic structure of T. acidophilum CP revealed that the proteolytic sites are located in the inner cavity formed by the b-rings (Lowe et al., 1995). Thus, the a-ring and the b-ring form two identical cavities above and below the central cavity, a phenomenon observed in proteases coined selfcompartmentalization (Lupas et al., 1997). In addition to T. acidophilum, CP crystal structures have been obtained for Rhodococcus erythropolis, Saccharomyces cerevisiae, and Bos taurus. Here, we illustrate the use of flexible fitting for the following hypothetical scientific scenario: we assume the T. acidophilum crystal structure has not been solved, but the one of S. cerevisiae has been. We aim to obtain an atomic model of the T. acidophilum CP using the S. cerevisiae crystal ˚ 1 resolution (Yu et al., structure and a T. acidophilum EM map at 7.5 A 2010). To build a starting model, we generated a sequence alignment of the CP subunits from T. acidophilum and S. cerevisiae using T-COFFEE. Using the sequence alignment and the crystal structure (PDB code: 1RYP) as input, we built a comparative model using MODELLER. Subsequently, this model was rigid-body fitted to the EM map (downloaded from the Electron Microscopy Database; EMDB-5130) using SITUS (Wriggers et al., 1999). An MDFF simulation was set up with default parameters (see www.ks.uiuc. edu/Research/MDFF) and ran for 50,000 integration steps, taking a few hours on a typical desktop computer. For the MDFF simulation, the EM map was B-factor sharpened (Rosenthal and Henderson, 2003), which reverts the typical damping of high-frequencies in EM maps resulting from SPA (Saad et al., 2001). In our experience, B-factor sharpening is generally beneficial for MDFF; however, it is only recommended if the ˚ ) 1 as it might be otherwise resolution of the EM map approaches (10–15 A unreliable. The initial model has a maximum CCC to the map of 0.74, which increases to 0.81 after the fit (Fig. 3.7A). The refinement process particularly changes the positions of the distal a subunits whereas the central b subunits change less because the initial model fits the EM map better (Fig. 3.7B). A comparison between the fitted model and the crystal structure (Yu et al., 2010) can serve as a ground truth for the quality of the model. For example, ˚ during the RMSD of an a-subunit model improves from 5.06 to 2.59 A the fit (Fig. 3.7D). The necessary files to reproduce this exercise can be downloaded from www.biochem.mpg.de/foerster.

68

A

¨rster and Elizabeth Villa Friedrich Fo

B

C

D

20) capture variations in noise and are ignored as a basis for classifying images. There are several approaches to clustering images including nonlinear mapping (Radermacher and Frank, 1985) and self-organizing maps (Pascual et al., 2000). The method used for analyzing FAS, and perhaps the most conceptually simple, is hierarchical ascendant classification (Frank et al., 1988). This algorithm begins by clustering images that are closest in factor space and then successively combines neighboring clusters (Ward, 1963).

188

Edward J. Brignole and Francisco Asturias

An advantage of this approach is that through SPIDER’s Web interface, a dendrogram that relates the classes can be interactively examined to assess the homogeneity of each class (Fig. 9.3C). Therefore, the number of classes needed to adequately describe the data does not need to be decided beforehand as with K-means clustering, another popular and much faster method. Another advantage of hierarchical clustering is reproducibility. Given the same images with the same number of factors, the resulting dendrogram does not change, as opposed to methods that initiate from random seeds and therefore result in variable clustering at each attempt. For practical purposes, because both statistical analysis and clustering are computationally intensive, and because images were sampled well above the achievable resolution, particle images were binned two- to fourfold to reduce the time required for computations, possibly providing a further benefit by improving the signal-to-noise ratio. After clustering, class averages that reflect differences in the particle images were generated from full-sized images to reveal conformational changes of FAS.

3.7. Realignment of particle images within classes Statistical analysis and clustering of particle images separate similar images from images that represent distinct projections of a structure. Once segregated into different views, alignment parameters can be refined because reference-free alignment performs best when applied to a homogeneous set of images. Therefore, after sorting FAS particles into more homogeneous groups, particle images within each class were realigned (reference-free) to yield averages with improved resolution. As implemented in SPIDER, reference-free alignment of particle images is biased by the image randomly selected to start the alignment process, which exerts a disproportionate influence on the alignment of later images and on the final outcome (Penczek et al., 1992). Because the initial alignment was performed on the entire heterogeneous set of FAS particle images, the entire alignmentclassification-realignment scheme was repeated in 10 independent attempts to evaluate multiple possible outcomes, reflecting the stability of the alignment and clustering results (Fig. 9.2C).

3.8. Merging oversampled classes The 10 independent rounds of alignment–classification–realignment typically produce many similar averages and oversample many outcomes. By aligning and classifying averages, redundant classes were objectively merged while retaining dissimilar classes. After merging, each particle is represented in 10 different classes or, for consistently aligned and classified particles, 10 times in a single class. Particles in each merged class were again realigned (Fig. 9.2D) such that redundant membership weights a particle’s influence

EM Analysis of FAS Conformational Flexibility

189

on the final alignment outcome according to the reproducibility of its match to a particular class. This “weighted” classification and alignment is less sophisticated than the computationally intensive likelihood approaches (see Chapters 10 and 11 in Volume 482) that formally assign weights to each particle over all classes and alignment parameters (Scheres et al., 2005; Sigworth, 1998). However, we have found it to generate reliable results because of its emphasis on testing the stability of alignment and clustering.

3.9. Refinement of the 2D class averages When dealing with a heterogeneous set of particle images representing a mixture of conformations, compositions, and/or orientations, initial alignment and clustering do not provide an optimal solution. Therefore, the resulting classes were refined by iterating classification and reference-free alignment to separate particles into progressively more homogeneous and distinct groups (Fig. 9.2E). The classification steps were accomplished by multireference alignment that, unlike the unsupervised clustering described above, simultaneously rotates, translates, and mirrors images. After sorting each particle according to its best correlated class average, reference-free alignment was used to realign particles within each class. Twenty iterations of alternating classification and alignment were performed, and after the final iteration, the particle images within each class were realigned to their average. This iterative classification–alignment procedure has been implemented in a way that facilitates visual comparison of class averages by orienting them at the end of each iteration against a reference image. Additionally, for compositionally heterogeneous samples, classes containing images of smaller subcomplexes may be present that require class-specific radii for alignment. Individual radii are calculated for each class by thresholding the average and 1D projection onto the x- and y-axes. This approach does not require the average to be centered or globular as would a radius calculated from a 1D rotational average. As a useful byproduct, the radius calculations provide the x- and y-shifts to center each average, which is critical during supervised classification because only features of the references within the alignment radius are considered. With these intermediate steps, the approach that we used to address conformational variability in metazoan FAS has been applicable to other macromolecules that exhibit conformational or compositional heterogeneity, such as Mediator (Cai et al., 2010). This procedure for iterative classification and alignment has been incorporated into the Appion pipeline (Lander et al., 2009). The class averages of FAS that result from the iterative classification– alignment procedure reveal structures that could be related by orientational changes and/or conformational variations. To distinguish between these possibilities, class averages were identified in which detailed features of the

190

Edward J. Brignole and Francisco Asturias

upper portion of FAS remain essentially unchanged but those in the lower portion have dramatically moved, clearly indicative of conformation rather than orientation changes. On the other hand, averages of particles in slightly rocked orientations can be easily recognized due to overlap of domains in the upper and lower portions of the structure; however, a wide distribution of views was not observed. Preferential adsorption orientation is an advantage when evaluating conformational heterogeneity by reducing the total number of images that must be collected and analyzed. As a result of adopting a preferred orientation, 2D analysis can separate this modest set of FAS particle images into a manageable number of classes that upon visual inspection predominantly reveal conformational changes.

3.10. Focused alignment and classification Averaging inadequately aligned images of macromolecules displaying variability in composition, conformation, and/or orientation, results in regions of reduced or absent density. In the case of structural conformers, successful analysis requires separation of particles into a discrete number of stable states. For a flexible macromolecule that exhibits a continuum of conformations (dubbed “fleximers” to distinguish from discrete “conformers”), the goal of alignment and classification is to describe the range and distribution of domain mobility (Burgess et al., 2004b). Examination of the 2D class averages of FAS indicates that the relative angle between the upper and lower portions of the structure flexes over a broad continuum (Fig. 9.4A and B). Such dramatic motions present a difficult problem for alignment of particle images—which portion of the structure should be aligned? In any given class, conformational flexibility requires a compromise with both portions of the structure being slightly out of register. Domain motions can be defined by fixing in place one portion of the structure and then examining the mobile portion relative to this fixed frame of reference. Once FAS particles are classified into relatively homogeneous groups, the comparatively less variable upper portion of the structure can be isolated and aligned while excluding the more mobile lower portion of the structure. This strategy and rationale was adapted from methods described for studying fleximers of dynein and myosin (Burgess et al., 2003, 2004a,b; Walker et al., 2000). Focusing alignment on a particular portion of the structure requires a mask to exclude other regions of the images (Fig. 9.4C and D). Masking multiple classes of FAS particles was facilitated by the fact that averages were oriented to a common reference during the preceding iterative classification-alignment (see above). The masked particles (Fig. 9.4E and F) were aligned (reference-free), and the resulting alignment parameters were applied to the original unmasked particles. As observed for images of FAS, focused alignment improves definition of features within the upper portion

191

EM Analysis of FAS Conformational Flexibility

A

C

B

D Mask and center E

F

Align, uncenter, and unmask G

Figure 9.4 Focused alignment. (A) Class averages. (B) Average and variance of particle images from the classes shown in (A). (C) A shape-fitting mask for FAS was constructed by thresholding the average into a binary image (SPIDER’s TH M command). The thresholded average was then segmented to isolate the upper portion. (D) Before masking particles the edge of the mask was blurred by box-convolution (SPIDER’s BC command) and then dilated by the half-width of the filter by repeating the thresholding and filtration. (E) Each class captures a different conformation and consequently the region to be masked was slightly out of alignment. To remedy this situation, before masking the particles in each class, the mask was applied to class averages, masked averages were aligned, and the alignment parameters for each class average were summed with parameters for particles in the class. The particle images were transformed by the combined parameters, the mask was applied, and finally the masked images were centered. (F) Average of all masked particles. (G) Masked particle images were aligned and, in the absence of a command that restricts rotational search, those that exceed a given rotation (20 ) or translation (5% of image dimension) were reset to their initial masked position. After reference-free alignment of the masked particles, the alignment parameters were summed with those prior to masking, accounting for and reversing the centering operation. The cumulative transformation was applied to the original unmasked particles. Average and variance images indicate improved alignment of features in the upper portion at the expense of blurring in the lower portion.

192

Edward J. Brignole and Francisco Asturias

C

B Classify top

E A

D

Classify bottom

Figure 9.5 Focused classification. (A) Average and variance after alignment of the upper portion. Two regions of high variance in the center of the upper portion are indicated by arrows. (B) Mask that focuses statistical analysis on the upper portion of the structure. (C) Averages from classification of the upper portion reveal a continuum of conformational rearrangement. (D) Mask that focuses statistical analysis on the lower portion of the structure. (E) Compared to classification of the upper portion, almost twice as many factors and four times more classes were necessary to describe the motion of domains in the lower portion of the structure.

of the structure, while more variable features excluded by the mask become blurred (Fig. 9.4G). Following alignment of the upper portion, two prominent variations were observed within this region (Fig. 9.5A). To better understand this source of conformational variation, statistical analysis and classification was focused on the upper portion (Fig. 9.5B). Eigenimages indicate that the two variations are anticorrelated, and class averages confirm that an asymmetric opening forms in one half of the upper portion of the structure while closing in the other half (Fig. 9.5C). The class averages can easily be arranged according to their coordinate along the first, dominant factor into a sequence that captures motion of domains in the upper portion of the structure. This domain rearrangement reflects not just in-plane motion but suggests out-of-plane domain rotations that cause changes in overlapping densities that are difficult to interpret in averages of 2D projection images. With the upper portion of the structure aligned, motions of the excluded portion can be analyzed. The average and variance images indicate extensive mobility in the lower portion, and a mask was designed to classify over this entire region (Fig. 9.5D). The resulting class averages (Fig. 9.5E) reveal a dramatic swinging motion as well as a conformational change that obscures the “legs.” This latter movement, like the motion delineated by classification of the upper portion, suggests out-of-plane domain rotation. Because rearrangements that result in overlapping densities are difficult to

EM Analysis of FAS Conformational Flexibility

193

describe from 2D averages, we sought a 3D approach to better comprehend the movement of domains in the upper and lower portions of FAS.

3.11. Random conical tilt: A 3D interpretation of FAS domain movements Although class averages provide significant insight, a more complete structural understanding of macromolecular conformational variability is afforded by 3D structures of the fleximers identified by 2D analysis. Three-dimensional structures that correspond to conformations apparent in 2D averages were calculated by collecting tilted-pair images and then using the random conical tilt reconstruction method (Radermacher et al., 1987). First, Euler angles were calculated for back-projection of tilted particles by combining the rotational alignment of untilted particle images with the angles of tilt and tilt axes that relate each pair of micrographs as defined during particle selection (see above). Despite its simplicity, caution is advised at this step to avoid producing structures with incorrect or mixed handedness due to the orientation in which the micrographs are digitized. The error in this instance results from the fact that tilt axes are measured between 90 and not over a full 360 . In our computational routines, ambiguity in the axis angles is resolved by determining the edge of the tilted micrograph that is farthest from focus. Once projection angles for the tilted particles have been determined, an initial 3D reconstruction for each class was obtained (SPIDER’s BP CG command). Since the tilted particle images were not yet centered, six cycles of translational alignment to corresponding projections of the 3D reconstruction were performed. After each cycle of shift refinement, a new 3D structure is calculated (SPIDER’s BP 3F command). This procedure is repeated for each class of particles, resulting in a 3D structure that corresponds to each 2D average (Fig. 9.6A). The EM density maps have resolutions of ˚ (FSC0.5) reflecting the limited number of particles (500–1000) that 35 A generate the 3D reconstruction for each class, the high defocus of tilted images that lack CTF correction, and compression artifacts from negatively stained specimen preparation. A more useful measure of reliability is comparison with a known high-resolution structure. The FAS crystal structure (Maier et al., 2008) was fitted into each density map as a single unit. Although the comparatively symmetrical conformation captured by the crystal structure can differ dramatically from the conformation apparent in the EM maps, the relative locations of each domain can be clearly recognized by shape and size, indicating that the 3D reconstructions have sufficient detail to give an accurate impression of domain position. To interpret the conformation captured by each class, after fitting the entire FAS crystal structure into the 3D reconstruction for each class, multidomain structural units (i.e., KR-SD, DH2, ER2, [KS-MAT]2) were extracted from

194

Edward J. Brignole and Francisco Asturias

A

B

Figure 9.6 (A) Random conical tilt reconstructions of FAS. (B) For a more complete understanding of conformational changes, domains from the crystal structure were fitted into the density maps, and filtered to the approximate resolution of the EM reconstructions. The domains in the lower portion of the structure can be clearly seen to swing over a 50 range and twist by 90 (also shown as bottom views in lower right panels). Domains in the upper portion of the FAS structure can roll by about 30 in either direction (only clockwise rotation is shown as a side view in lower left panels).

the crystal structure and their positions adjusted locally and stored in a single multimodel PDB-format file (Fig. 9.6B). While excellent software exists for fitting high-resolution structures into EM densities (Volkmann and Hanein, 1999; Wriggers et al., 1999), these do not readily consider constraints imposed by flexible linker sequences that tether loosely connected domains. Because conformational changes between the structures can be easily modeled by simple domain movements, structural units were manually adjusted aided by the local “fit-in-map” tool implemented in Chimera (Goddard et al., 2007) so that appropriate distance constraints were maintained given interdomain linkages. To prevent overlap of adjacent domains, occupied densities were removed using the “volume eraser” tool. Chimera also has the capability to use fitted atomic structures to segment density maps that may be used for this purpose. Once the multidomain structural units were fitted into the EM maps, the angles and axes of domain motions were determined. Previously, this calculation was done manually; however, recent versions of Chimera include the “measure rotation” command to simplify quantitation of domain movements. In the upper portion of the structure, it becomes clear that the DH and ER dimers each roll 30 to generate an asymmetric opening. Domains in the lower portion of the structure are capable of dramatic swinging motions over a 50 range. The most striking conformation captured

EM Analysis of FAS Conformational Flexibility

195

by the EM analysis turns domains in the lower portion perpendicular to the upper portion and indicate clearly that the two subunits of the FAS dimer are capable of twisting around one another, likely by a full 180 .

3.12. Quantitative analysis of conformational distribution of various active-site mutants Given the extensive flexibility of FAS, we sought to understand the catalytic relevance of the observed domain motions. Earlier EM analysis showed that different FAS mutants in the presence of substrates yielded 2D averages that reflect a change in conformation (Asturias et al., 2005). Based on this observation, we reasoned that FAS mutants defective in a particular catalytic activity would, upon addition of substrates, show an altered distribution of conformations related to catalytic arrest. Initially, we analyzed the conformation distributions independently for several FAS mutants prepared in the presence and absence of substrates. However, comparing the averages from each dataset required subjective decisions due to differences in how particle images clustered. To make a more direct comparison, particle images from all of the FAS preparations were merged into a single large dataset and segregated into the same conformational bins through the alignment and classification scheme, described above (Fig. 9.7A). Then, the number of particles from each FAS preparation assigned to a particular conformation were determined and directly compared (Fig. 9.7B). In the presence of substrates, conformations of FAS with the upper domains rolled into an asymmetric arrangement and lower domains in-plane with the upper portion become dominant, indicating that these conformations are directly related to catalysis. In the absence of substrates, the upper domains are more likely to be symmetrically arranged or the lower domains turned perpendicular to the upper portion, indicating that these conformations represent noncatalytic intermediates.

4. EM and FAS: A Versatile Tool for a Flexible Macromolecule The strategy that we employed to study the conformational flexibility of FAS by single-particle EM was based on unsupervised (reference-free) alignment and clustering of particle images and required no a priori knowledge of the FAS structure or any assumptions to sort particles according to their conformation. Gradual subdivision of particle images into progressively more classes revealed intermediate domain positions that made possible a description of the continuous flexibility of FAS. The 2D analysis of conformational heterogeneity was extended to 3D structures using tilted

196

Edward J. Brignole and Francisco Asturias

A

B -

Release +

-

Processing +

Elongation +

Figure 9.7 Comparison of conformational distributions between FAS preparations. (A) After particles were aligned and classified, class averages were sorted into categories, according to symmetry (red) or asymmetry (blue) in the upper portion of the structure, and according to lower domains in-plane (bright) or perpendicular (faded) with the upper portion. (B) The conformational redistribution of particles for each mutant FAS in the presence (þ) or absence () of substrates suggests that catalysis is facilitated by conformations with the lower portion in-plane with an asymmetric upper portion (bright blue). The mutations examined resulted in defective acyl-chain release, processing, or elongation.

images and the random conical tilt method (Radermacher et al., 1987). This approach sidesteps reference-based projection matching and common-lines methods and provides a bias-free 3D description of macromolecular conformations. Finally, a molecular interpretation of 2D and 3D EM structures was generated by considering the published X-ray crystallography structure of a static FAS conformation.

4.1. Single-particle EM as a method for studying macromolecular conformation and flexibility More than a single structure is required to understand how a macromolecular machine functions. Knowledge about structural rearrangements related to association, dissociation, or modification of components is essential to arrive at a complete mechanistic understanding. To facilitate structural analysis, some macromolecular complexes can be induced to populate a single predominant conformation in response to regulators or substrates, as in pH or protease dependent virus maturation (Yu et al., 2008), nucleotide bound states of a group II chaperonin (Zhang et al., 2010), Mediator upon polymerase binding (Cai et al., 2009), and ribosome ratcheting with eEF2/ EF-G (Frank and Agrawal, 2000; Taylor et al., 2007). By contrast, attempts to induce FAS to adopt particular conformations by addition of substrates to

197

EM Analysis of FAS Conformational Flexibility

active-site mutants did not result in population of a single state but only succeeded in perturbing the distribution of particles amongst a conformational continuum (Fig. 9.7B). Discrete conformations or biochemical states of a macromolecule that adopts only a small number of well-defined conformations become evident when separation of images beyond a certain number of groups results in redundant class averages. In contrast, macromolecular flexibility gives rise to continuous motions, and the goal of EM analysis becomes describing the distribution of particles amongst different fleximers and the range of motions sampled by the complex. In our study of FAS, we based our conclusions on relative changes in conformational distribution that resulted from altering catalytic activities by mutation and addition of substrates. In this way, we identified conformations that become more or less abundant to facilitate particular catalytic events and minimized bias related to perturbations arising from the specimen preparation method (e.g., particle adsorption and stain preservation). Fitting domains from the FAS crystal structure into the lowresolution 3D EM structures and examination of catalytic contacts facilitated by particular structures allowed us to describe how large-scale twisting, swinging, and rolling motions were related to specific catalytic events that occur during fatty acid synthesis (Brignole et al., 2009; Fig. 9.8). Our approach to analysis of conformational distribution builds upon earlier investigations of dynamic complexes that pioneered the use of EM as a technique for quantifying macromolecular heterogeneity. Notably, Burgess and colleagues used EM to determine the conformational

Rolling and swinging

Twisting

Rolling and swinging

Figure 9.8 The catalytic cycle of FAS is mediated by conformational flexibility: rolling motions of domains in the upper portion, swinging motions of domains in the lower portion, and twisting about the linkers that connect the upper and lower portions. Sequential interactions between each catalytic domain and the carrier domain (spheres) with its covalently bound cargo (black stick) are illustrated. Both reaction chambers are depicted synthesizing fatty acids asynchronously such that the acyl chain is processed in one chamber but elongated in the other.

198

Edward J. Brignole and Francisco Asturias

distribution between structural elements of dynein and myosin, and perturbed this distribution by addition of nucleotide analogs that mimic particular catalytic states (Burgess et al., 2003, 2004a,b; Walker et al., 2000). Heymann et al. (2003) have also examined a continuous population of maturing herpes virus capsids. More recently, Southworth and Agard (2008) have identified an equilibrium between open and closed conformers of Hsp90 that can be altered by nucleotide and Cai et al. (2010) have used EM to quantify motions in the Mediator head module that can be perturbed by binding of TBP. In a unique example of correlating biochemical kinetics with quantitation of particle conformations, Mulder et al. (2010) examined complex mixtures of 30S ribosome assembly intermediates by EM as a time resolved approach. The power of single-particle EM to provide information about different macromolecular conformations is also demonstrated by new image analysis approaches that make possible the assessment of macromolecular heterogeneity in 3D, ranging from methods to identify structural variations such as local resolution calculation (blocres command in Bsoft; Heymann and Belnap, 2007) or 3D variance analysis (Zhang et al., 2008) to methods that generate several 3D structures from a single dataset by classification of particles within projection groups (Elmlund et al., 2009; Fu et al., 2007) or multimodel maximum likelihood 3D refinement (Scheres et al., 2007). In the past few years, these advances in EM image processing have been used to identify multiple conformational or biochemical states (i.e., subunit occupancy) present in a single specimen preparation, including, for example: SV40 large tumor antigen (Cuesta et al., 2010), ribosome with partial occupancy of EF-G (Gao et al., 2004), RNA polymerase II clamp opening (De Carlo et al., 2003; Kostek et al., 2006), TFIID with partial TBP occupancy (Elmlund et al., 2009; Grob et al., 2006), and RSC chromatin remodeler with and without nucleosome (Chaban et al., 2008). Finally, continuing improvements in EM image formation and acquisition, in particular, phase plates (Danev et al., 2009; Gamm et al., 2010; see Chapter 14 in Volume 481) and direct detectors (Milazzo et al., 2010), promise to enhance contrast and signal, respectively, so that the approach we employed for visualizing extensive conformational heterogeneity of the relatively small FAS particle in stain might soon become feasible in cryo-EM where macromolecular structure would be preserved in a more physiological state.

ACKNOWLEDGMENTS The work described here was supported by the US National Institutes of Health through grants R01 DK16073 to S. Smith and F. J. A., and F32 DK080622 to E. J. B.

EM Analysis of FAS Conformational Flexibility

199

REFERENCES Adrian, M., et al. (1984). Cryo-electron microscopy of viruses. Nature 308, 32–36. Adrian, M., et al. (1998). Cryo-negative staining. Micron 29, 145–160. Aebi, U., and Pollard, T. D. (1987). A glow discharge unit to render electron microscope grids and other surfaces hydrophilic. J. Electron Microsc. Tech. 7, 29–33. Asturias, F. J., et al. (2004). Electron microscopic analysis of the RSC chromatin remodeling complex. Methods Enzymol. 376, 48–62. Asturias, F. J., et al. (2005). Structure and molecular organization of mammalian fatty acid synthase. Nat. Struct. Mol. Biol. 12, 225–232. Boisset, N., et al. (1990). Three-dimensional reconstruction of native Androctonus australis hemocyanin. J. Mol. Biol. 216, 743–760. Boisset, N., et al. (1993). Three-dimensional immunoelectron microscopy of scorpion hemocyanin labeled with a monoclonal Fab fragment. J. Struct. Biol. 111, 234–244. Bretaudiere, J. P., and Frank, J. (1986). Reconstitution of molecule images analysed by correspondence analysis: A tool for structural interpretation. J. Microsc. 144, 1–14. Brignole, E. J., et al. (2009). Conformational flexibility of metazoan fatty acid synthase enables catalysis. Nat. Struct. Mol. Biol. 16, 190–197. Brink, J., et al. (2002). Quaternary structure of human fatty acid synthase by electron cryomicroscopy. Proc. Natl. Acad. Sci. USA 99, 138–143. Brink, J., et al. (2004). Experimental verification of conformational variation of human fatty acid synthase as predicted by normal mode analysis. Structure 12, 185–191. Burgess, S. A., et al. (2003). Dynein structure and power stroke. Nature 421, 715–718. Burgess, S. A., et al. (2004a). The structure of dynein-c by negative stain electron microscopy. J. Struct. Biol. 146, 205–216. Burgess, S. A., et al. (2004b). Use of negative stain and single-particle image processing to explore dynamic properties of flexible macromolecules. J. Struct. Biol. 147, 247–258. Cai, G., et al. (2009). Mediator structural conservation and implications for the regulation mechanism. Structure 17, 559–567. Cai, G., et al. (2010). Mediator head module structure and functional interactions. Nat. Struct. Mol. Biol. 17, 273–279. Chaban, Y., et al. (2008). Structure of a RSC–nucleosome complex and insights into chromatin remodeling. Nat. Struct. Mol. Biol. 15, 1272–1277. Chen, J. Z., et al. (2009). Molecular interactions in rotavirus assembly and uncoating seen by high-resolution cryo-EM. Proc. Natl. Acad. Sci. USA 106, 10644–10648. Cheng, Y., and Walz, T. (2009). The advent of near-atomic resolution in single-particle electron microscopy. Annu. Rev. Biochem. 78, 723–742. Cheng, Y., et al. (2006). Single particle reconstructions of the transferrin–transferrin receptor complex obtained with different specimen preparation techniques. J. Mol. Biol. 355, 1048–1065. Coleman, R. A., and Lee, D. P. (2004). Enzymes of triacylglycerol synthesis and their regulation. Prog. Lipid Res. 43, 134–176. Cong, Y., et al. (2010). 4.0-A resolution cryo-EM structure of the mammalian chaperonin TRiC/CCT reveals its unique subunit arrangement. Proc. Natl. Acad. Sci. USA 107, 4967–4972. Cuesta, I., et al. (2010). Conformational rearrangements of SV40 large T antigen during early replication events. J. Mol. Biol. 397, 1276–1286. Danev, R., et al. (2009). Practical factors affecting the performance of a thin-film phase plate for transmission electron microscopy. Ultramicroscopy 109, 312–325. De Carlo, S., et al. (2003). Cryo-negative staining reveals conformational flexibility within yeast RNA polymerase I. J. Mol. Biol. 329, 891–902.

200

Edward J. Brignole and Francisco Asturias

Elad, N., et al. (2008). Detection and separation of heterogeneity in molecular complexes by statistical analysis of their two-dimensional projections. J. Struct. Biol. 162, 108–120. Elmlund, H., et al. (2009). Cryo-EM reveals promoter DNA binding and conformational flexibility of the general transcription factor TFIID. Structure 17, 1442–1452. Frank, J., and Agrawal, R. K. (2000). A ratchet-like inter-subunit reorganization of the ribosome during translocation. Nature 406, 318–322. Frank, J., and van Heel, M. (1982). Correspondence analysis of aligned images of biological particles. J. Mol. Biol. 161, 134–137. Frank, J., et al. (1978). Reconstruction of glutamine synthetase using computer averaging. Ultramicroscopy 3, 283–290. Frank, J., et al. (1988). Classification of images of biomolecular assemblies: A study of ribosomes and ribosomal subunits of Escherichia coli. J. Microsc. 150, 99–115. Frank, J., et al. (1996). SPIDER and WEB: Processing and visualization of images in 3D electron microscopy and related fields. J. Struct. Biol. 116, 190–199. Fu, J., et al. (2007). Unsupervised classification of single particles by cluster tracking in multi-dimensional space. J. Struct. Biol. 157, 226–239. Gamm, B., et al. (2010). Object wave reconstruction by phase-plate transmission electron microscopy. Ultramicroscopy 110, 807–814. Gao, H., et al. (2004). Dynamics of EF-G interaction with the ribosome explored by classification of a heterogeneous cryo-EM dataset. J. Struct. Biol. 147, 283–290. Goddard, T. D., et al. (2007). Visualizing density maps with UCSF Chimera. J. Struct. Biol. 157, 281–287. Golas, M. M., et al. (2003). Molecular architecture of the multiprotein splicing factor SF3b. Science 300, 980–984. Grob, P., et al. (2006). Cryo-electron microscopy studies of human TFIID: Conformational breathing in the integration of gene regulatory cues. Structure 14, 511–520. Harris, J. R. (2007). Negative staining of thinly spread biological samples. Methods Mol. Biol. 369, 107–142. Henderson, R., et al. (1990). Model for the structure of bacteriorhodopsin based on high-resolution electron cryo-microscopy. J. Mol. Biol. 213, 899–929. Heymann, J. B., and Belnap, D. M. (2007). Bsoft: Image processing and molecular modeling for electron microscopy. J. Struct. Biol. 157, 3–18. Heymann, J. B., et al. (2003). Dynamics of herpes simplex virus capsid maturation visualized by time-lapse cryo-electron microscopy. Nat. Struct. Biol. 10, 334–341. Joshi, A. K., et al. (2005). Effect of modification of the length and flexibility of the acyl carrier protein-thioesterase interdomain linker on functionality of the animal fatty acid synthase. Biochemistry 44, 4100–4107. Kellenberger, E., et al. (1982). The wrapping phenomenon in air-dried and negatively stained preparations. Ultramicroscopy 9, 139–150. Kitamoto, T., et al. (1988). Structure of fatty acid synthetase from the Harderian gland of guinea pig. Proteolytic dissection and electron microscopic studies. J. Mol. Biol. 203, 183–195. Kostek, S. A., et al. (2006). Molecular architecture and conformational flexibility of human RNA polymerase II. Structure 14, 1691–1700. Kuhajda, F. P., et al. (1994). Fatty acid synthesis: A potential selective target for antineoplastic therapy. Proc. Natl. Acad. Sci. USA 91, 6379–6383. Kuhlbrandt, W., et al. (1994). Atomic model of plant light-harvesting complex by electron crystallography. Nature 367, 614–621. Lambert, O., et al. (1994). Quaternary structure of Octopus vulgaris hemocyanin. Threedimensional reconstruction from frozen-hydrated specimens and intramolecular location of functional units Ove and Ovb. J. Mol. Biol. 238, 75–87.

EM Analysis of FAS Conformational Flexibility

201

Lander, G. C., et al. (2009). Appion: An integrated, database-driven pipeline to facilitate EM image processing. J. Struct. Biol. 166, 95–102. Loftus, T. M., et al. (2000). Reduced food intake and body weight in mice treated with fatty acid synthase inhibitors. Science 288, 2379–2381. Maier, T., et al. (2006). Architecture of mammalian fatty acid synthase at 4.5 A resolution. Science 311, 1258–1262. Maier, T., et al. (2008). The crystal structure of a mammalian fatty acid synthase. Science 321, 1315–1322. Menendez, J. A., and Lupu, R. (2007). Fatty acid synthase and the lipogenic phenotype in cancer pathogenesis. Nat. Rev. Cancer 7, 763–777. Milazzo, A. C., et al. (2010). Characterization of a direct detection device imaging camera for transmission electron microscopy. Ultramicroscopy 110, 744–747. Ming, D., et al. (2002). Domain movements in human fatty acid synthase by quantized elastic deformational model. Proc. Natl. Acad. Sci. USA 99, 7895–7899. Miyazawa, A., et al. (2003). Structure and gating mechanism of the acetylcholine receptor pore. Nature 423, 949–955. Mulder, A. M., et al. (2010). Visualizing Ribosome Biogenesis: Time-Resolved Single Particle EM Reveals Parallel Assembly Pathways for 30S Subunit. (in review). Ohi, M., et al. (2004). Negative staining and image classification—Powerful tools in modern electron microscopy. Biol. Proced. Online 6, 23–34. Pascual, A., et al. (2000). Mapping and fuzzy classification of macromolecular images using self-organizing neural networks. Ultramicroscopy 84, 85–99. Penczek, P., et al. (1992). Three-dimensional reconstruction of single particles embedded in ice. Ultramicroscopy 40, 33–53. Radermacher, M., and Frank, J. (1985). Use of nonlinear mapping in multivariate image analysis of molecule projections. Ultramicroscopy 17, 117–126. Radermacher, M., et al. (1987). Three-dimensional reconstruction from a single-exposure, random conical tilt series applied to the 50S ribosomal subunit of Escherichia coli. J. Microsc. 146, 113–136. Radermacher, M., et al. (1994). Cryo-electron microscopy and three-dimensional reconstruction of the calcium release channel/ryanodine receptor from skeletal muscle. J. Cell Biol. 127, 411–423. Rath, B. K., and Frank, J. (2004). Fast automatic particle picking from cryo-electron micrographs using a locally normalized cross-correlation function: A case study. J. Struct. Biol. 145, 84–90. Resh, M. D. (2006). Palmitoylation of ligands, receptors, and intracellular signaling molecules. Sci. STKE, re14. Ruiz, T., et al. (2001). The first three-dimensional structure of phosphofructokinase from Saccharomyces cerevisiae determined by electron microscopy of single particles. J. Struct. Biol. 136, 167–180. Scheres, S. H., et al. (2005). Maximum-likelihood multi-reference refinement for electron microscopy images. J. Mol. Biol. 348, 139–149. Scheres, S. H., et al. (2007). Disentangling conformational states of macromolecules in 3DEM through likelihood optimization. Nat. Methods 4, 27–29. Sherman, M. B., et al. (1981). On the negative straining of the protein crystal structure. Ultramicroscopy 7, 131–138. Sigworth, F. J. (1998). A maximum-likelihood approach to single-particle image refinement. J. Struct. Biol. 122, 328–339. Smith, S., et al. (2003). Structural and functional organization of the animal fatty acid synthase. Prog. Lipid Res. 42, 289–317. Southworth, D. R., and Agard, D. A. (2008). Species-dependent ensembles of conserved conformational states define the Hsp90 chaperone ATPase cycle. Mol. Cell 32, 631–640.

202

Edward J. Brignole and Francisco Asturias

Steven, A. C., and Navia, M. A. (1980). Fidelity of structure representation in electron micrographs of negatively stained protein molecules. Proc. Natl. Acad. Sci. USA 77, 4721–4725. Steven, A. C., and Navia, M. A. (1982). Specificity of stain distribution in electron micrographs of protein molecules contrasted with uranyl acetate. J. Microsc. 128, 145–155. Suloway, C., et al. (2005). Automated molecular microscopy: The new Leginon system. J. Struct. Biol. 151, 41–60. Swinnen, J. V., et al. (2003). Fatty acid synthase drives the synthesis of phospholipids partitioning into detergent-resistant membrane microdomains. Biochem. Biophys. Res. Commun. 302, 898–903. Taylor, K. A., and Glaeser, R. M. (1974). Electron diffraction of frozen, hydrated protein crystals. Science 186, 1036–1037. Taylor, D. J., et al. (2007). Structures of modified eEF2 80S ribosome complexes reveal the role of GTP hydrolysis in translocation. EMBO J. 26, 2421–2431. Tischendorf, G. W., et al. (1974). Determination of the location of proteins L14, L17, L18, L19, L22, L23 on the surface of the 5oS ribosomal subunit of Escherichia coli by immune electron microscopy. Mol. Gen. Genet. 134, 187–208. Unwin, P. N. (1975). Beef liver catalase structure: Interpretation of electron micrographs. J. Mol. Biol. 98, 235–242. van Heel, M., and Frank, J. (1981). Use of multivariate statistics in analysing the images of biological macromolecules. Ultramicroscopy 6, 187–194. Vance, J. E., and Vance, D. E. (2004). Phospholipid biosynthesis in mammalian cells. Biochem. Cell Biol. 82, 113–128. Volkmann, N., and Hanein, D. (1999). Quantitative fitting of atomic models into observed densities derived by electron microscopy. J. Struct. Biol. 125, 176–184. Voss, N. R., et al. (2009). DoG Picker and TiltPicker: Software tools to facilitate particle selection in single particle electron microscopy. J. Struct. Biol. 166, 205–213. Walker, M. L., et al. (2000). Two-headed binding of a processive myosin to F-actin. Nature 405, 804–807. Ward, J. H. (1963). Hierarchical grouping to optimize an objective function. J. Am. Stat. Assoc. 58, 236–244. Wriggers, W., et al. (1999). Situs: A package for docking crystal structures into lowresolution maps from electron microscopy. J. Struct. Biol. 125, 185–195. Yonekura, K., et al. (2003). Complete atomic model of the bacterial flagellar filament by electron cryomicroscopy. Nature 424, 643–650. Yu, I. M., et al. (2008). Structure of the immature dengue virus at low pH primes proteolytic maturation. Science 319, 1834–1837. Zhang, Y. M., et al. (2006). Inhibiting bacterial fatty acid synthesis. J. Biol. Chem. 281, 17541–17544. Zhang, W., et al. (2008). Heterogeneity of large macromolecular complexes revealed by 3D cryo-EM variance analysis. Structure 16, 1770–1776. Zhang, J., et al. (2010). Mechanism of folding chamber closure in a group II chaperonin. Nature 463, 379–383.

C H A P T E R

T E N

Tomography of Actin Cytoskeletal Networks Dorit Hanein Contents 204 205 206 206 208 209 210 212 212

1. Introduction 2. Testing the Lamella Hypothesis 3. Challenges 3.1. First stumbling block, sample preparation 3.2. Second stumbling block, image analysis 3.3. Third stumbling block, transience 3.4. Lastly, fourth stumbling block, heterogeneity Acknowledgments References

Abstract Structural biology research is increasingly focusing on unraveling structural variations at the micro-, meso-, and macroscale aiming at interpreting dynamic biological processes and pathways. Toward this goal, high-resolution transmission cryoelectron microscopy (cryo-EM) and cryoelectron tomography (cryo-ET) are indispensable, as these provide the ability to determine 3D structures of large, dynamic macromolecular assemblies in their native, fully hydrated state in situ. Underlying such analyses is the implicit assumption that specific structural states yield specific cellular outputs. The dependence on this structure– function paradigm is not unique to studies pertaining a particular pathway or biological process but it sets the foundation for all cell biological analyses of macromolecular assemblies. Yet, the paradigm still awaits formal proof. The field of high-resolution electron microscopy (HREM) is in dire need of establishing approaches and technologies to systematic and quantitative determining structure–function correlates in physiologically relevant environment. Here, using the actin cytoskeletal networks as an example, we will provide snapshots of current advances in defining the structures of these highly dynamic networks in situ. We will further detail some of the major stumbling blocks on the way to quantitatively correlate the dynamic state to network morphology in the same window of time and space. Sanford-Burnham Medical Research Institute, La Jolla, California, USA Methods in Enzymology, Volume 483 ISSN 0076-6879, DOI: 10.1016/S0076-6879(10)83010-1

#

2010 Elsevier Inc. All rights reserved.

203

204

Dorit Hanein

1. Introduction Actin is one of the most abundant proteins in eukaryotic cells, with a concentration of 100–500 mM in nonmuscle cells. Actin is remarkably conserved across species, with all 375 residues being identical for humans and chickens. Monomeric or G-actin polymerizes to form actin filaments (F-actin) which upon assembly into higher order networks provide cells with mechanical support and driving force for movement (for recent reviews see Dominguez, 2009; Lee and Dominguez, 2010; Pollard and Cooper, 2009; Reisler and Egelman, 2007). Approximately 40 X-ray crystallographic structures of actin monomers are now available; however, the atomic structure of its polymeric, helical F-actin form(s) remains elusive (Kabsch et al., 1990; Oda et al., 2009). Structurally, F-actin consists of actin monomers arranged in two protofilaments that twist around each other, while actin subunits face the same direction rending a polar filament. A large body of data suggests that individual actin filaments exhibit structural polymorphism, with a fairly constant axial rise and a variable rotation between adjacent subunits within the filament (Egelman et al., 1982; Galkin et al., 2008; Schmid et al., 2004; Stokes and DeRosier, 1987). Eukaryotic actins belong to an ATPase superfamily, defined by shared nucleotide-binding motif consisting of two large domains connected by a hinge to which nucleotide binds (Kabsch and Holmes, 1995). Actin monomers within a filament hydrolyze, through an unidentified structural pathway, a single molecule of ATP to ADP. This process is irreversible in the lifetime of the polymer in vitro (Carlier et al., 1988). It was suggested that hydrolysis is the F-actin “timekeeper” in vivo, demarcating newly polymerized regions from older portions of the filament (Blanchoin and Pollard, 2002; Blanchoin et al., 2000). New ATP-actin monomers preferentially add to the “barbed end” (also known as the fast growing plus end) of the filament and depart the filament primarily from the “pointed end” (minus end) in the form of ADP-actin. The “barbed end” “pointed end” terminology is taken from the appearance of myosin decorated actin filaments imaged in the electron microscope that appear as arrowheads with barbed and pointed ends. The temporal asymmetry of actin monomer behavior in respect to the filaments gives rise to a process known as treadmilling. In vivo, actin filaments perform their function primarily by assembling into higher order cellular structures. In metazoan cells, at least a dozen distinct supramolecular structures F-actin have been identified. Each of these plays unique F-actin and essential functional roles ranging from cellular motility, organelle transport, and pathogen invasion to tumor progression. To mention but a few: immunological synapse, adherent junction, cytokinetic ring, cortactical spectrin-actin, endocytic pits, phagocytic cups, podosomes, invadapodia, filopodia, microvilli, ruffles, and stress fibers (Chhabra and Higgs, 2007).

Seeing Actin Cytoskeletal Networks

205

Many of the highly diverse biological roles critically rely on dynamic properties of these networks, that is, on their treadmilling. Indeed, in vivo transition between G- to F-actin is tightly regulated in a spatiotemporal coordinated manner by a large number of actin-binding and regulatory proteins. Over 100 actin-binding proteins (ABP) have been identified so far. These ABPs are used to initiate polymerization, promote, or restrict elongation of the filaments (severing, capping), maintain a pool of actin monomers (sequestering), promote the assembly of actin networks (crosslinking, bundling, membrane attachment), and regulate the assembly turnover (Lee and Dominguez, 2010; Pollard and Cooper, 2009). Cancer, various neurological disorders, and cardiomyopathies are few of the diseases associated to abnormalities in the assembly or regulation of the actin cytoskeleton. While recent advancements in high-resolution light microscopy techniques (HRLM), molecular biology and biophysical assays provided wealth of significant new knowledge on actin cytoskeleton dynamics in these networks, comprehensive molecular detailed structural correlates are critically lacking. Here, we will use the sheet-like lamellipodia/lamella networks as an example to exemplify the major existing barriers toward progress in providing such structure correlates.

2. Testing the Lamella Hypothesis Much of current biomedical research and development is guided by the structure–function paradigm, which implies that functional outputs can be predicted from structural information and, in reverse direction, that specific functional outputs are mediated by a specific structural configuration. While the paradigm is established as a means to understanding the function of individual molecules, it has remained a mostly at the level of an assumption for larger macromolecular assemblies that drive more complex cellular outputs. For example, the dendritic network model of directed cell migration suggests that the leading edge is pushed forward by an array of actin filaments with a characteristic-branched morphology defined by an actin nucleation complex, the Arp2/3 complex (Pollard and Borisy, 2003). Much of this model has been derived from bulk biochemical analyses of actin polymerization in vitro, from live imaging of single filament assembly outside the cellular context, and from 2D electron microscopy (EM) images of actin networks in detergent extracted, chemically fixed, dehydrated cells. By extrapolation of this data, it is generally assumed that regions with increased branching activity and filament density would be associated with faster protrusion. Although such structure–function relationship has never been shown directly, the interpretation of experiments in the cell migration literature rests on this inference.

206

Dorit Hanein

Among the body of work that challenges this view are studies of actin dynamics during epithelial cell protrusion. Using quantitative fluorescent speckle microscopy (qFSM), a live-cell imaging modality delivering maps of the rates of filament turnover and motion with submicron and second scale resolution (Danuser and Waterman-Storer, 2006), it was inferred that cell edge propulsion may be driven by two partially overlapping, yet differentially regulated actin networks (Delorme et al., 2007; Gupton et al., 2005; Ponti et al., 2004). Molecular and functional analyses of the relationship between edge movement, assembly, and contraction forces suggested that forward motion of the cell edge at the onset of a protrusion cycle may be initiated by elongation of the actin assembly, the lamella that is independent of Arp2/3 complex and then a second, Arp2/3-mediated network, the lamellopodia (branched network) reinforces cytoskeleton expansion against pressures from plasma membrane and extracellular environment ( Ji et al., 2008). This lamella hypothesis is highly controversial (Danuser, 2009; Vallotton and Small, 2009), similar to the dendritic model that still awaits structural validation.

3. Challenges In the following sections, we used the lamella hypothesis as an example to explain the challenges of dissecting the ultrastructure of large and transient assemblies and of putting these data into a functional context. This is only one example and similar challenges are encountered in many other structure–function studies.

3.1. First stumbling block, sample preparation The challenge of structural biology today is to establish approaches that allow capturing the state of a macromolecular assembly at high structural detail during a defined cellular output. It is clear that EM should serve as the gold standard for ultrastructural analyses, nevertheless sample preparation protocols that faithfully preserve these delicate, fragile cytoskeletal networks in situ are still being defined. 3.1.1. Cryo-EM Recent advances of sample preparation and imaging techniques allow us to more reliable correlate live cells morphology to ultrastructure. One of the key advances was that the development of cryo-EM methods. These methods allow biological samples to be imaged frozen in a near-native, physiological environment, thus bypassing the harsh preparative procedures of detergent extraction, chemical fixation, dehydration, metal shadowing, or critical point drying required by traditional EM and opening the way to HR

Seeing Actin Cytoskeletal Networks

207

determination of eukaryotic cytoskeleton ultrastructure (Dubochet et al., 1988; see Volume 483, Chapter 3). 3.1.2. Correlative light and EM More recently, major advances in relating the structure of cytoskeletal networks to cell function in situ were made by correlative light and electron microscopy (LM/EM) approaches. The goal of this approach is to capture the state of a macromolecular assembly at high structural detail preferable at defined cellular output (Lucic et al., 2007, 2008; Muller-Reichert et al., 2007; Sartori et al., 2007; see Volume 481, Chapter 13). This approach aims at combining dynamic information derived from live-cell imaging with high-definition structural characterization of the same region using HR cryo-EM imaging. This approach, in principle, allows us to image a region in the cell with a known dynamic history. 3.1.3. Cryo-ET of eukaryotic cells in toto Electron tomography (cryo-ET) of vitrified whole cells has proven to have great potential for imaging cellular architecture in 3D at a resolution of 4–6 nm of intact cells or cell regions with overall thickness not extending 1 mm. Recently, cryo-ET allows direct proof for the existence of cytoskeletal actin-like networks within bacteria, resolving a long standing controversy for the existence of such assemblies in prokaryotic cells (for recent review see (Li and Jensen, 2009)). The ability to generate 3D volumes of assemblies in situ allows filaments or networks to be followed, so branching, cross-linking, or overlapping arrangements that might seem similar in 2D projection of a single planar section, can be faithfully recognized and followed within the volume (Ben-Harush et al., 2010; Medalia et al., 2002b; Urban et al., 2010; see Volume 481, Chapter 12). Lastly, to overcome thickness of the preparations, for example, Salje et al. (2009) used cryosectioned frozen E. coli cells. Imaging these sections resulted in the first direct in vivo proof of an E. coli cytoskeletal filament assembly. Frozenhydrated sections were also used to extract 3D information using tomographic imaging (Al-Amoudi et al., 2007; Gruska et al., 2008; Leis et al., 2009; Pierson et al., 2010; see Volume 481 Chapter 8). An alternative approach for thinning of the frozen-hydrated specimen is just being introduced via focused ion beam (FIB) technology (Marko et al., 2006, 2007; Rigort et al., 2010). In this new approach, correlative cryofluorescence microscopy allows to navigate the large cellular volumes and to localize specific cellular targets, followed by FIB thinning, and cryo-ET. Despite these advances, sample preparation protocols optimized to promote normal cell growth, that are compatible with HR fluorescence imaging, sustain cryogenic plunging, and are amenable for cellular cryo-ET data acquisition, still need to be identified. Furthermore, to ascertain that these cell culture preparations reproducibly preserve the protein content, 3D

208

Dorit Hanein

architecture, and antigenicity of the macromolecular machineries involved in motility, still remain a challenge. The fidelity of the structural studies is critically dependent on the reproducibility in preservation of the characteristic “composition signature” of these sites. At least 20–40 ABPs have been “colocalized” within the lamellopidia or lamella during their formation and disassembly. Their presence or absence within the preparations contribute to the controversy in proving (disproving) the lamella hypothesis. The elegant work of Medalia et al. (2002a) first demonstrated that cryoET of intact Dictyostelium discoideum cells could reveal the connections of the F-actin network with the plasma membrane as well as possible F-actin branching. Similar views of branched networks were obtained when cells of higher eukaryotes were studied (Ben-Harush et al., 2010; Delorme et al., 2007; Gupton et al., 2005; Medalia et al., 2007). In contrast, cryo-EM and cryo-ET imaging of lamellipodia suggested that F-actins are almost exclusively unbranched (Urban et al., 2010). These studies bring to light the second stumbling block Small et al. (2008).

3.2. Second stumbling block, image analysis The putative spatial overlap of distinct actin networks and the dimensions of the filaments (10 nm) in the volume of the leading edge of a cell will probably be recognizable only by the application of HR image processing tools to the 3D tomographic volumes, supported by sophisticated image segmentation and topology classification algorithms. Visual inspection of projection images is insufficient to test the lamella hypothesis by cryo-ET (see, e.g., Fig. 10.1). Due to the inability to tilt above 70 in all available electron microscopes, reconstruction artifacts such as the missing wedge produced by standard tomographic data collection schemes can significantly hamper the interpretability of the resulting reconstruction. To minimize these artifacts, collecting dual-axis tilt data, thus filling in some of the missing data are essential (Kremer et al., 1996; Penczek et al., 1995). Although challenging on the sample, instrument performance, and volume reconstruction, dual-axis tomography significantly increases the fidelity of filament network determination. We and others have shown the feasibility of this approach to vitrified samples with reconstituted actin networks (Rouiller et al., 2008), and in Pyrodictium (Nickell et al., 2003). High-end TEM instruments that provide functionality to rotate the grid by 90 around the beam axis facilitating dualaxis data acquisition schemes are available. Owing to the extremely low signal-to-noise ratio (0.01), the low contrast, relatively low resolution (>4 nm), and the crowded nature of the specimen, the electron densities are difficult to interpret. Indeed, identification of key macromolecular motifs is the most time-consuming step in the tomography pipeline. Segmentation of the density into more manageable subvolumes, in conjunction with automatic template matching approaches and

209

Seeing Actin Cytoskeletal Networks

C

A

D

E

B

Figure 10.1 Testing the lamella hypothesis. (A) Two-dimensional projection of the leading edge of an eukaryotic cell. The thickness of the cell periphery is 300 nm. Grid bar width: 25 mm. (B) Twelve consecutive dual-axis tilt series will be needed to capture the entire cell edge. A far from trivial task both from imaging and image analysis perspectives. Bar 10 mm. (C) A slice through the volume of one of the movies provided as supplementary information in Urban et al., 2010; ncb2044-s5. This movie was used by the authors for demonstrating the absence of Arp2/3 mediated branch junctions in the captured region of the cell. The authors ascertain the absence of branches in the tomogram by visual inspection of the slices. Bar = 10 nm. (D) We show the segmented density of the central feature in (C) overlay with the fitted 3D branch model determined in (Rouiller et al., 2008). The molecular model was fitted into the extracted density with no modifications, reflecting the good fit between the in vitro derived molecular model and the in vivo imaged density. (E) The extracted branch junction seen in (D) viewed along the mother filament axis. It is unclear why this branch junction was not detected by visual inspection (Urban et al., 2010), however this discrepancy is used to exemplify the need for employing more robust, objective, quantitative image analysis tools rather than by eye inspection.

subvolume averaging should be used to extract structural motifs from the 3D volumes. A major effort in the HR cryo-ET community is to enhance tomographic reconstruction and to devise such fully automated segmentation, and template matching software suites (Volume 482, Chapter 13, Volume 483, Chapters 2, 3). Here, we argue that much of the reported differences in the detection and distinction of filament assemblies with different topology, for example, branched versus straight, and their macromolecular interaction partners is related to the lack of implementation of enabling protocols for unbiased image analysis that allows employment (and scrutiny) of statistical and mathematical tools to ascertain meaningful structure–function correlation (Fig. 10.1C-E).

3.3. Third stumbling block, transience Interactions between the lamellipodium and lamella are transient in space and time. The short-term transience of these systems, the state of motion of the leading edge changes between protrusion and retraction in cycles as short as

210

Dorit Hanein

60–100 s on the length scale of 2–3 mm (Machacek and Danuser, 2006). It can be assumed that the underlying cytoskeleton morphology undergoes a complete restructuring at the same time scale. Thus, the life time of a functional state is probably in the order of 15–20 s, assuming up to five distinct states per cycle. The challenge for EM analyses is the capture of well-defined functional states of cytoskeleton structures at the relevant time-scale and resolution. However, very limited precautions are being taken to guarantee that the structural analyses occurred at the time and length scales of seconds and microns over which cell morphogenic processes are regulated. Therefore, to establish a quantitative structure–function relationship between actin network dynamics and cell motility the states of cell edge and cytoskeleton must be frozen within seconds (20%, v/v) prevent ice formation to certain extent, however, causes osmotic dehydration of samples. High molecular weight sugar polymers (like dextran) were found to minimize this effect, however, at the same time increase crowding due to their protein-like nature. Furthermore, sugars exponentially decrease sample resistance to beam-induced damage, thus impinging of HR structural definition (see Volume 482, Chapter 15). Plunge-freezing is the only technique to achieve vitrification of pure water or physiological buffers in a reproducible way (Cavalier et al., 2009), however, no plunge-freezing vitrification techniques available today can address the need of correlation of live-cell imaging and cryo-EM of vitrified eukaryotic cells with the time-shifts less than 10 s. Vitrification robots are available to warrant controlled, reproducible vitrification. However, these systems do not have the capability of interfacing with a HRLM using oil immersion optics. Technology and protocols need to be developed to allow accommodating vitrification in the relevant time window, while overcoming the obstacles of the geometry and configuration required for HRLM oil immersion optics.

3.4. Lastly, fourth stumbling block, heterogeneity The transience of cell and cytoskeleton states is expected to generate structural heterogeneity. Meaningful statistical evaluation of such structural distributions requires the combination of higher throughput EM imaging

Seeing Actin Cytoskeletal Networks

211

with HR live-cell light microscopy. Transient processes produce a vast spatial heterogeneity of states. Seeing a few snapshots of a structure at different length scales and associating them with a coarse definition of cellular outputs produce incomplete and arbitrary models of structure– function relationships—we suspect a major source of the current disputes in the cell motility field. Different labs may merely see different instances of the same process. To tackle the state heterogeneity, approaches for HR cryo-ET with high-throughput need to be developed (see Volume 481, Chapter 12 and Volume 483, Chapter 11). This will transform HR cryoET data acquisition from a laborious, weakly reproducible imaging method to a quantitative technique that affords employment of state-of-the art statistical and machine learning tools to ascertain significant correlations between cell motility behavior and actin cytoskeleton structure. Driven by the advances of computer-controlled microscopes, digital cameras, and aberration-reducing electron-optics, cellular cryo-ET have become a fast developing imaging technique (Volume 481, Chapter 24). However, to date it is still limited to imaging anecdotal regions of a single eukaryotic cell. Tomographic reconstruction of the full structural variation of the leading edge of a single cell (and subsequent analysis of structurally homogeneous volumes) is a 2-year project for a highly trained and persevering postdoctoral fellow. Establishing structure–function relationships between cytoskeleton and cell morphodynamics by unbiased machine learning of the structural and functional heterogeneity rather requires cryo-ET of tens (30) of cells; and understanding the structural bases of the migration deficiencies requires structure–function relationships from 10 experimental conditions. Clearly, at present, a meaningful structural analysis of a disease condition is outside reach of cryo-ET. Furthermore, the available imaging devices for cryo-ET image acquisition limit the spatial span of image acquisition. Between 8 and 16 consecutive tomograms are required for a full data set of one leading edge of eukaryotic cell (Fig. 10.1A-B). Neither the sample nor the commonly available TEMs would be able to perform for this extent of time (vacuum deterioration, buildup of ice contamination on the sample, etc.). Hence, the development of new instrumentation is necessary to bring cryo-ET to the level of analyzing the structural underpinnings of transient processes in macromolecular assemblies. Structural biology research is increasingly focusing on unraveling biological and biochemical processes and pathways at the micro-, meso-, and macroscale aiming to correlate space and time. Toward this goal, HR transmission electron cryomicroscopy (cryo-TEM) is indispensable, as it provides ability to determine 3D structures of large, dynamic macromolecular assemblies in the native, fully hydrated state in situ. Nevertheless, this technology has not yet been harnessed to quantitatively correlate dynamic states to morphology of fully hydrated sample, in the same window of time,

212

Dorit Hanein

and using rigorous statistical sampling and mathematical tools to ascertain meaningful structure–function correlation in those situations. To harness the power of quantitation of structure–cell function relationships, we need to combine advances in technology development with adequate specimen preparation and correlative mathematical methods that allow quantitative link between HR live-cell imaging and cryo-ET.

ACKNOWLEDGMENTS I thank Drs Danuser and Volkmann for advice and assistance with the manuscript, and Drs Ochoa and Volkmann for providing the data included in Fig 10.1. The funding source for Dr Dorit Hanein for this study is the National Institutes of Health Cell Migration Consortium; Grant Number: U54 GM064346 and NIGMS Grant Number P01 GM066311.

REFERENCES Al-Amoudi, A., Diez, D. C., Betts, M. J., and Frangakis, A. S. (2007). The molecular architecture of cadherins in native epidermal desmosomes. Nature 450, 832–837. Ben-Harush, K., Maimon, T., Patla, I., Villa, E., and Medalia, O. (2010). Visualizing cellular processes at the molecular level by cryo-electron tomography. J. Cell Sci. 123, 7–12. Blanchoin, L., and Pollard, T. D. (2002). Hydrolysis of ATP by polymerized actin depends on the bound divalent cation but not profilin. Biochemistry 41, 597–602. Blanchoin, L., Pollard, T. D., and Mullins, R. D. (2000). Interactions of ADF/cofilin, Arp2/3 complex, capping protein and profilin in remodeling of branched actin filament networks. Curr. Biol. 10, 1273–1282. Carlier, M. F., Pantaloni, D., Evans, J. A., Lambooy, P. K., Korn, E. D., and Webb, M. R. (1988). The hydrolysis of ATP that accompanies actin polymerization is essentially irreversible. FEBS Lett. 235, 211–214. Cavalier, A., Spehner, D., and Humbel, B. M. (2009). Handbook of cryo-preparation methods for electron microscopy. In “Methods in Visualization Series,” (G. Morel, ed.). CRC Press, Boca Raton. Chhabra, E. S., and Higgs, H. N. (2007). The many faces of actin: Matching assembly factors with cellular structures. Nat. Cell Biol. 9, 1110–1121. Danuser, G. (2009). Testing the lamella hypothesis: The next steps on the agenda. J. Cell Sci. 122, 1959–1962. Danuser, G., and Waterman-Storer, C. M. (2006). Quantitative fluorescent speckle microscopy of cytoskeleton dynamics. Annu. Rev. Biophys. Biomol. Struct. 35, 361–387. Delorme, V., Machacek, M., DerMardirossian, C., Anderson, K. L., Wittmann, T., Hanein, D., Waterman-Storer, C., Danuser, G., and Bokoch, G. M. (2007). Cofilin activity downstream of Pak1 regulates cell protrusion efficiency by organizing lamellipodium and lamella actin networks. Dev. Cell 13, 646–662. Dominguez, R. (2009). Actin filament nucleation and elongation factors—structure-function relationships. Crit. Rev. Biochem. Mol. Biol. 44, 351–366. Dubochet, J., Adrian, M., Chang, J.-J., Homo, J.-C., Lepault, J., McDowall, A. W., and Schultz, P. (1988). Cryo-electron microscopy of vitrified specimens. Q. Rev. Biophys. 21, 129–228. Egelman, E. H., Francis, N., and DeRosier, D. J. (1982). F-actin is a helix with a random variable twist. Nature 298, 131–135.

Seeing Actin Cytoskeletal Networks

213

Galkin, V. E., Orlova, A., Cherepanova, O., Lebart, M. C., and Egelman, E. H. (2008). High-resolution cryo-EM structure of the F-actin-fimbrin/plastin ABD2 complex. Proc. Natl. Acad. Sci. USA 105, 1494–1498. Gruska, M., Medalia, O., Baumeister, W., and Leis, A. (2008). Electron tomography of vitreous sections from cultured mammalian cells. J. Struct. Biol. 161, 384–392. Gupton, S. L., Anderson, K. L., Kole, T. P., Fischer, R. S., Ponti, A., HitchcockDeGregori, S. E., Danuser, G., Fowler, V. M., Wirtz, D., Hanein, D., and WatermanStorer, C. M. (2005). Cell migration without a lamellipodium: Translation of actin dynamics into cell movement mediated by tropomyosin. J. Cell Biol. 168, 619–631. Ji, L., Lim, J., and Danuser, G. (2008). Fluctuations of intracellular forces during cell protrusion. Nat. Cell Biol. 10, 1393–1400. Kabsch, W., and Holmes, K. C. (1995). The actin fold. FASEB J. 9, 167–174. Kabsch, W., Mannherz, H. G., Suck, D., Pai, E. F., and Holmes, K. C. (1990). Atomic structure of the actin:DNase I complex. Nature 347, 37–44. Kremer, J. R., Mastronarde, D. N., and McIntosh, J. R. (1996). Computer visualization of three-dimensional image data using IMOD. J. Struct. Biol. 116, 71–76. Lee, S. H., and Dominguez, R. (2010). Regulation of actin cytoskeleton dynamics in cells. Mol. Cells 29, 311–325. Leis, A., Rockel, B., Andrees, L., and Baumeister, W. (2009). Visualizing cells at the nanoscale. Trends Biochem. Sci. 34, 60–70. Li, Z., and Jensen, G. J. (2009). Electron cryotomography: A new view into microbial ultrastructure. Curr. Opin. Microbiol. 12, 333–340. Lucic, V., Kossel, A. H., Yang, T., Bonhoeffer, T., Baumeister, W., and Sartori, A. (2007). Multiscale imaging of neurons grown in culture: From light microscopy to cryo-electron tomography. J. Struct. Biol. 160, 146–156. Lucic, V., Leis, A., and Baumeister, W. (2008). Cryo-electron tomography of cells: Connecting structure and function. Histochem. Cell Biol. 130, 185–196. Machacek, M., and Danuser, G. (2006). Morphodynamic profiling of protrusion phenotypes. Biophys. J. 90, 1439–1452. Marko, M., Hsieh, C., Moberlychan, W., Mannella, C. A., and Frank, J. (2006). Focused ion beam milling of vitreous water: Prospects for an alternative to cryo-ultramicrotomy of frozen-hydrated biological samples. J. Microsc. 222, 42–47. Marko, M., Hsieh, C., Schalek, R., Frank, J., and Mannella, C. (2007). Focused-ion-beam thinning of frozen-hydrated biological specimens for cryo-electron microscopy. Nat. Methods 4, 215–217. Medalia, O., Typke, D., Hegerl, R., Angenitzki, M., Sperling, J., and Sperling, R. (2002a). Cryoelectron microscopy and cryoelectron tomography of the nuclear pre-mRNA processing machine. J. Struct. Biol. 138, 74–84. Medalia, O., Weber, I., Frangakis, A. S., Nicastro, D., Gerisch, G., and Baumeister, W. (2002b). Macromolecular architecture in eukaryotic cells visualized by cryoelectron tomography. Science 298, 1209–1213. Medalia, O., Beck, M., Ecke, M., Weber, I., Neujahr, R., Baumeister, W., and Gerisch, G. (2007). Organization of actin networks in intact filopodia. Curr. Biol. 17, 79–84. Muller-Reichert, T., Srayko, M., Hyman, A., O’Toole, E. T., and McDonald, K. (2007). Correlative light and electron microscopy of early Caenorhabditis elegans embryos in mitosis. Methods Cell Biol. 79, 101–119. Nickell, S., Hegerl, R., Baumeister, W., and Rachel, R. (2003). Pyrodictium cannulae enter the periplasmic space but do not enter the cytoplasm, as revealed by cryo-electron tomography. J. Struct. Biol. 141, 34–42. Oda, T., Iwasa, M., Aihara, T., Maeda, Y., and Narita, A. (2009). The nature of the globular- to fibrous-actin transition. Nature 457, 441–445.

214

Dorit Hanein

Penczek, P., Marko, M., Buttle, K., and Frank, J. (1995). Double-tilt electron tomography. Ultramicroscopy 60, 393–410. Pierson, J., Fernandez, J. J., Bos, E., Amini, S., Gnaegi, H., Vos, M., Bel, B., Adolfsen, F., Carrascosa, J. L., and Peters, P. J. (2010). Improving the technique of vitreous cryosectioning for cryo-electron tomography: Electrostatic charging for section attachment and implementation of an anti-contamination glove box. J. Struct. Biol. 169, 219–225. Pollard, T. D., and Borisy, G. G. (2003). Cellular motility driven by assembly and disassembly of actin filaments. Cell 112, 453–465. Pollard, T. D., and Cooper, J. A. (2009). Actin, a central player in cell shape and movement. Science 326, 1208–1212. Ponti, A., Machacek, M., Gupton, S. L., Waterman-Storer, C. M., and Danuser, G. (2004). Two distinct actin networks drive the protrusion of migrating cells. Science 305, 1782–1786. Reisler, E., and Egelman, E. H. (2007). Actin structure and function: What we still do not understand. J. Biol. Chem. 282, 36133–36137. Rigort, A., Bauerlein, F. J., Leis, A., Gruska, M., Hoffmann, C., Laugks, T., Bohm, U., Eibauer, M., Gnaegi, H., Baumeister, W., and Plitzko, J. M. (2010). Micromachining tools and correlative approaches for cellular cryo-electron tomography. J. Struct. Biol. (in press) Feb 21 Epub ahead of publication. Rouiller, I., Xu, X. P., Amann, K. J., Egile, C., Nickell, S., Nicastro, D., Li, R., Pollard, T. D., Volkmann, N., and Hanein, D. (2008). The structural basis of actin filament branching by the Arp2/3 complex. J. Cell Biol. 180, 887–895. Salje, J., Zuber, B., and Lowe, J. (2009). Electron cryomicroscopy of E. coli reveals filament bundles involved in plasmid DNA segregation. Science 323, 509–512. Sartori, A., Gatz, R., Beck, F., Rigort, A., Baumeister, W., and Plitzko, J. M. (2007). Correlative microscopy: Bridging the gap between fluorescence light microscopy and cryo-electron tomography. J. Struct. Biol. 160, 135–145. Schmid, M. F., Sherman, M. B., Matsudaira, P., and Chiu, W. (2004). Structure of the acrosomal bundle. Nature 431, 104–107. Small, J. V., Auinger, S., Nemethova, M., Koestler, S., Goldie, K. N., Hoenger, A., and Resch, G. P. (2008). Unravelling the structure of the lamellipodium. J. Microsc. 231, 479–485. Stokes, D. L., and DeRosier, D. J. (1987). The variable twist of actin and its modulation by actin-binding proteins. J. Cell Biol. 104, 1005–1017. Urban, E., Jacob, S., Nemethova, M., Resch, G. P., and Small, J. V. (2010). Electron tomography reveals unbranched networks of actin filaments in lamellipodia. Nat. Cell Biol. 12, 429–435. Vallotton, P., and Small, J. V. (2009). Shifting views on the leading role of the lamellipodium in cell migration: Speckle tracking revisited. J. Cell Sci. 122, 1955–1958.

C H A P T E R

E L E V E N

Visual Proteomics ¨rster,* Bong-Gyoon Han,† and Martin Beck‡ Friedrich Fo Contents 1. Introduction 2. Data Acquisition 2.1. Required experimental setup 2.2. Choosing acquisition parameters 2.3. Acquisition and reconstruction of tomograms 3. Templates 3.1. Selecting template structures 3.2. Quaternary structure conservation of protein complexes in Desulfovibrio vulgaris 3.3. Generation of templates from atomic maps 4. Template Matching 4.1. Handling MolMatch 4.2. Creating motif lists 4.3. Scoring 4.4. Visualization of molecular atlases 5. Assessment of Performance 5.1. Estimating true-positive discovery rates from artificial tomograms 5.2. Performance assessment in real data sets 6. The Spatial Proteome of L. interrogans 7. Outlook Acknowledgments References

216 217 217 218 220 221 221 222 225 227 227 229 230 231 232 232 235 237 241 241 242

Abstract Visual proteomics attempts to generate molecular atlases by providing the position and angular orientation of protein complexes inside of cells. This is accomplished by template matching (pattern recognition), a cross-correlationbased process that matches the structure of a specific protein complex to the * Max Planck Institute of Biochemistry, Martinsried, Germany Life Sciences Division, Lawrence Berkeley National Laboratory, University of California, Berkeley, California, USA { European Molecular Biology Laboratory, Heidelberg, Germany {

Methods in Enzymology, Volume 483 ISSN 0076-6879, DOI: 10.1016/S0076-6879(10)83011-3

#

2010 Elsevier Inc. All rights reserved.

215

216

¨rster et al. Friedrich Fo

densities of the whole volume or subvolume of a cell, that is typically acquired by cryoelectron tomography. Thereby, a search is performed that scans the entire volume for structural templates contained in a database. In this chapter, we primarily describe the practical experiences gained with visual proteomics during the Leptospira interrogans proteome project [Beck et al. (2009). Visual proteomics of the human pathogen Leptospira interrogans. Nat. Methods 6, 817.]. We give a practical guide how to implement the method and review critical experimental and computational aspects in detail. Based on a survey that has been undertaken for protein complexes from Desulfovibrio vulgaris, we review the difficulty of generating reference structures in detail. Finally, we discuss the high yield targets for technical improvements.

1. Introduction Proteomics approaches often rely on mass spectrometric (MS) measurements and are carried out on the combined lysates of multiple cells. Thereby, the proteins contained in a sample are identified, but any spatial information about them is lost and cell specific properties are averaged out over the population of lysed cells. The visual proteomics concept promises to overcome these limitations and to identify individual protein complexes in intact cells (Nickell et al., 2006). This method consists of three processing steps that attempt to localize individual structures within cryoelectron tomograms. At first, a library is assembled that contains the reference structures of the targeted protein complexes resampled to the relevant electron optical conditions. Subsequently, the local cross-correlation coefficient between each reference structure and tomogram is calculated for all possible positions and orientations and stored in a cross-correlation volume. Finally, the distribution of cross-correlation values within such volumes is translated into a position list by peak extraction and statistical methods. The combined structural signatures of multiple protein complexes, detected in frozen-hydrated specimens, have the potential to describe the spatial proteome of a specific cell as a molecular atlas (Fig. 11.1). Such information is invaluable to biologists in the age of systems biology, where awareness has grown that individual cells have specific properties that contribute to the behavior of the entire population and that protein interactions are more dynamic than anticipated a decade ago. In 2000, Bohm et al. (2000) provided a proof-of-concept of the approach using simulations as well as a first application to cryoelectron tomograms. However, the unambiguous identification of protein complexes in cryoelectron tomograms of intact cells remains a considerable challenge today. The major obstacles that currently prevent a straightforward implementation are: (i) the strong interaction of electrons with matter limits the application of cryoelectron tomography (CET) to

Generation of Molecular Atlases of Prokaryotes

217

Figure 11.1 Visual proteomics of the human pathogen Leptospira interrogans. Reference structures are shown in the top panel. The ribosome, RNA polymerase II, GroEL, GroEL-ES, Hsp, and ATP synthase were template matched in the tomogram shown as centered slice through the reconstruction on the left. The templates assigned into this volume are shown surface rendered on the right (membrane in blue, cell wall in brown).

relatively thin specimens. Therefore, visual proteomics of whole cells is currently effectively limited to prokaryotes; (ii) the availability of template structures for these species limits the number of targets; (iii) at the currently achievable resolution only large protein complexes have a chance of being identified; and (iv) the low signal-to-noise ratio (SNR) peculiar to cryoelectron tomograms hampers a robust and reliable detection. Technical improvements in specimen thinning, better detectors for electron microscopy, the introduction of phase plates, development of CS correctors, as well as growing structural libraries hold great promise to overcome at least some of these limitations in the future.

2. Data Acquisition 2.1. Required experimental setup The acquisition of cryoelectron tomograms of highest quality in terms of SNR is desirable for the method outlined here. To achieve this goal, the thickness of the frozen-hydrated specimen should not exceed 300 nm in order to avoid extensive multiple electron scattering. Sample preparation is

218

¨rster et al. Friedrich Fo

described in greater detail in Chapter 3, Vol. 481. The use of a transmission electron microscope (TEM) equipped with liquid nitrogen cooled stage, field emission gun, and image filter is essential. The acceleration voltage (av) may range from 200 to 300 kV. While working at 300 kV enables the use of slightly thicker specimen, the use of 200 kV pays off for thinner specimens with an increase in contrast. Any standard tomography acquisition software may be used, however, a good control over the dosage spent on the specimen during the acquisition is advantageous (see Section 2.3). For most of the computational postprocessing steps a standard work station running under any operating system can be used. The template matching itself is computationally expensive and access to a Linux cluster is highly desirable. The protocol outlined in this chapter requires the following software: Matlab version 6.5 or higher, the TOM toolbox for tomography (Nickell et al., 2005), MolMatch for the actual template matching (Fo¨rster, 2005), and a number of scripts and example files that are available online (see also Chapter 15, Vol. 482).

2.2. Choosing acquisition parameters In CET, the experimental setup for the tomographic data collection should be chosen according to the task. One of the factors limiting the resolution in a tomogram is the angular increment Da of the tilt series. The Crowther criterion provides an estimate for the maximal resolution rmax of an object with a diameter (d; Crowther et al., 1970): rmax ¼

1 : dDa

ð11:1Þ

In Eq. (11.1) the angular increment is measured in radians. This equation is based on the fact that the Fourier transformation of a projection corresponds to a central slice of the Fourier transformation of the 3D reconstruction (Chapter 1, Vol. 482). The thickness of each slice is reciprocally proportional to d. The Crowther criterion defines to what resolution the different slices overlap in Fourier space; interpolation can be used to retrieve the information to that resolution. For example, when an object with a diameter of 25 nm is sampled with an angular increment of 9 rmax will not exceed (4 nm) 1. However, it is good practice in image processing to oversample data to minimize the loss of information due to interpolation artifacts, which are in the core of every 3D reconstruction algorithm. Typically, data is oversampled at least by a factor of 2; that is, if the targeted object with a diameter of 25 nm is to be imaged with rmax of (4 nm) 1, it should be sampled with an angular increment of no more than 4.5 . In CET, the resolution is not isotropic: due to the limited tilt-angle of typically 60 to þ 60 , a wedge shaped area in Fourier space remains

219

Generation of Molecular Atlases of Prokaryotes

unsampled. Accordingly, the information is unresolved in the respective sections, which leads to a typical elongation of objects along the z-axis in real space (“missing wedge effect”). In addition to the sampling, we must consider the imaging conditions for data acquisition. The imaging of the TEM can be described as a linear system: in this approximation, an electron micrograph is the projection of the electrostatic potential of the object convoluted with the “contrast transfer function” (CTF). The CTF is an oscillating function of the applied defocus value. Since the contrast largely vanishes in focus, micrographs are taken at a specific defocus value. Due to the oscillation of the CTF, the contrast is reversed in certain resolution ranges. Since it is difficult to deconvolute electron micrographs (“CTF correction”) in CET, we typically low-pass filter tomograms at the first zero-crossing of the estimated CTF. Thus, the chosen defocus value effectively sets an upper resolution limit to the tomographic data. For example, the 1st CTF zero of a TEM operated at 300 kV and a defocus of 8 mm is at (4 nm) 1. Nevertheless, CTF correction of tomograms is a very active field of research (Chapter 1, Vol. 482) and CTF correction methods for tomograms will certainly be applied more commonly in the future, which would make it possible to use data beyond the 1st CTF zero (Fernandez et al., 2006; Winkler, 2007; Zanetti et al., 2009). Furthermore, the chosen magnification is also vitally important for the targeted resolution. The resolution cannot exceed the Nyquist frequency: Ny ¼

1 : 2dpix

ð11:2Þ

In Eq. (11.2) dpix is the pixel size (ps) at the specimen level. When choosing the ps, it must furthermore be considered that the signal transfer of CCD cameras decays substantially as a function of frequency at that image data should generally be oversampled to avoid extensive information loss during image processing. As a consequence, we recommend choosing Ny at least two to three times the targeted resolution. We emphasize that the above considerations only determine the maximal resolution. The precision of the alignment of projections to a common 3D coordinate system can also limit the maximum resolution (Lawrence, 1992; Chapter 13, Vol. 482). However, in most cases the errors from projection alignment are negligible compared to the limitation due to the applicable electron dose. The signal is typically not significant to rmax, that is, the signal is not distinguishable from the noise. When significance criteria from signal-particle analysis (typically Fourier ring correlation) are applied, the signal of cryoelectron tomograms is typically not significant beyond (5–10 nm) 1. The theoretically attainable resolution, that is, if signal loss due to detector imperfections and interpolation could be excluded, is

220

¨rster et al. Friedrich Fo

approximated to be (2 nm) 1 (Henderson, 2004). Nevertheless, considerable information is contained in cryoelectron tomograms beyond the resolution limit set by the electron dose, but it is buried in the noise. Template matching strategies, also coined “matched filters,” can help to retrieve this information. The idea is to use prior knowledge to filter the data accordingly (see below). After summarizing the different factors on signal content in CET, we would recommend choosing the imaging parameters such that the maximum resolution is in the range of (3–4 nm) 1 when no CTF correction is applied. For a typical experimental setup, we recommend a cumulative ˚ at the electron dose of 100 e/A˚2, a defocus of 6–10 mm, a ps of 5–8 A specimen level and an angular increment Da of 1.5–4 .

2.3. Acquisition and reconstruction of tomograms The individual cells of a population will contain different amounts of the target protein complexes in varying subcellular locations. The strength of the visual proteomics approach is that it can reveal such properties. However, in order to compare tomograms taken from different cells or different subvolumes of the same cell, a comparable SNR throughout all data sets is required, because the SNR directly contributes to the true-positive discovery rate (see below). This is a difficult task: The specimen thickness will vary across the EM grid and objects close to grid bars will not be accessible at high tilt-angles. Common acquisition software does not necessarily support a satisfactory dosage control. It is therefore essential to use identical condenser beam settings and exposure times throughout data acquisition. Data sets of similar specimen thickness should be subselected for further analysis following the general principle that the thinner, the higher the SNR. The specimen thickness can be approximated from the electron count of a filtered versus a nonfiltered image acquired at 0 tilt-angle. Three-dimensional reconstructions should be carried out by weighted back projection or using iterative algorithms, for example, the simultaneous iterative reconstruction technique (Leis et al., 2006). A low-pass filter at the first zero of the CTF should be applied to the weighted projections when no CTF correction is attempted. A further comparative assessment between data sets, particularly of the resolution, can be obtained by Fourier ring correlation comparisons between an original projection and the corresponding reprojection of the tomogram calculated from all the other projections (Cardone et al., 2005). When a 2k 2k camera is used, the tomograms are typically reconstructed with a binning factor of 2 and then visually inspected in order to select the desired region that is afterward reconstructed with a binning factor of 1. Finally, a region of the final reconstruction might be selected by segmentation in order to reduce the

Generation of Molecular Atlases of Prokaryotes

221

computational expense for some of the subsequent steps, in particular the peak extraction (see Section 4.2).

3. Templates 3.1. Selecting template structures There are a number of different aspects to be taken into account when reference structures are selected to build a library of templates. Is the reference structure available in high resolution in the protein data bank (PDB file) or as an electron optical density map? PDBs are converted into templates by summing up and resampling the electron density of all the atoms (see Section 3.3). This approach is advantageous because the resulting template maps represent the true electron density and can be directly related to each other. In contrast, EM maps are not on an absolute scale; the gray values are usually normalized to a mean value of zero. However, to simulate particle identification performance for different templates it is crucial to ensure that the signals are on the same scale. Scaling the density of EM-derived maps to other templates derived from EM maps or PDBs is not a trivial task because average densities differ among proteins and the densities of other components, such as nucleic acids vary from proteins (see Section 5.1). As an approximation, the mean value and standard deviation of the density distribution of EM maps can be set to the average mean value and standard deviation observed in all X-ray structures investigated within the same study. Does the targeted protein complex form different oligomeric states or transient structures in the imaged system? To date, visual proteomics studies have targeted highly stable protein complexes that exist primarily in the fully assembled form. If the rigidity assumption is not supported by biochemical data—or even worse: data indicate a high degree of structural variation—an investigation by visual proteomics might be biased. What is the abundance of the targeted protein complex in the cell and to what extent does the template library account for all large protein complexes that exist in the cell? A study of the quantitative proteome of Leptospira interrogans (Malmstrom et al., 2009) has revealed a dynamic range of protein concentrations covering more than 3 orders of magnitude. If, for example, the desired protein complex exists on average in 10 copies per cell, and a single tomogram covers approximately 10% of the cell volume, it will in average contain one target particle. In such a scenario, a statistically significant discrimination of true- from false-positive template matches, as described below (Section 5), is simply impossible. Moreover, other protein complexes might exhibit similar structural signatures that are difficult to distinguish at the given resolution. The protein family of

222

¨rster et al. Friedrich Fo

AAA-ATPases, for example, has numerous members fulfilling diverse functions in the cell (Hanson and Whiteheart, 2005). AAA-ATPases adopt the same fold and they typically assemble to hexameric rings (Vale, 2000). At the typical resolution of cryoelectron tomograms, it will be impossible to distinguish the different members of this diverse protein family. Nevertheless, the Leptospira study has also shown that one specific paralogue is usually by far the most abundant complex of one family. As a consequence, the chances of “overlooking” a large protein complex of high abundance are limited. Can reference structures of homologs account for the desired target protein complex? Thus far, visual proteomics studies have either targeted highly conserved protein complexes using reference structures from other species, or, as in case of Mycoplasma pneumonia, templates were obtained from the same organism using single-particle cryoelectron microscopy (Kuhner et al., 2009). Recently, it was shown that in some cases even subtle changes in the primary structures of homologs may result in substantial changes on the quaternary structure level (Section 3.2).

3.2. Quaternary structure conservation of protein complexes in Desulfovibrio vulgaris A structural-proteomic survey has been performed for the largest and most abundant protein complexes in D. vulgaris Hildenborough (DvH; Han et al., 2009). One of the goals of this survey was to establish the extent to which the previously known structures of homologous protein complexes could be used as templates in the visual proteomics characterization of a new microorganism of interest. Except for the limiting conditions of stability, abundance, and high MW that ultimately determine if a protein complex can be purified, the choice of protein complexes studied was free from other assumptions that would bias the survey. This survey included 15 protein complexes (Table 11.1) purified by a tagless technique (Dong et al., 2008), together with 70S ribosomes purified by a separate protocol. Out of these 16 macromolecular complexes, the quaternary structures of eight were successfully analyzed by single-particle EM structural study at resolutions ˚ . In addition, the subunit compositions and stoichiometries better than 30 A of all the protein complexes were analyzed by biochemical methods and mass spectroscopy. The biochemical identity of 13 complexes could be established with confidence based on sequence homology to other proteins with known functions. For three complexes, however, the sequence homology to other proteins was too low to assign any function. Out of the 13 complexes with reliable functional assignment that were included in this study, only three complexes (the 70S ribosome, GroEL, and the phosphoenolpyruvate synthase complex) showed full conservation of subunit stoichiometry and quaternary structures, while the remaining 10 complexes showed variation between different bacteria, sometimes even

Table 11.1 Biochemical identity, abundance, and stoichiometry of large macromolecular complexes from DvH Database annotation (gene)a

Pyruvate:ferredoxin oxidoreductaseb (DVU3025) Lumazine synthase (riboflavin synthase b-subunit; DVU1198) Riboflavin synthase a-subunit (DVU1200) RNA polymerase b-subunit (DVU1329) RNA polymerase b0 -subunit (DVU2928) RNA polymerase a-subunit (DVU2929) RNA polymerase o-subunit (DVU3242) NusA (DVU0510) 60 kDa chaperonin (GroEL; DVU1976) Putative protein (DVU0671) Inosine-50 -monophosphate dehydrogenase (DVU1044) Phosphoenolpyruvate synthasee (DVU1833) Hemolysin-type calcium-binding repeat protein (DVU1012) Carbohydrate phosphorylase (DVU2349) Putative protein (DVU0631) Predicted phospho-2-dehydro-3-deoxyheptonate aldolase (DVU0460) Alcohol dehydrogenase (DVU2405) Ketol-acid reductoisomerase (DVU1378) Pyruvate carboxylase (DVU1834) Proline dehydrogenase/delta-1-pyrroline-5-carboxylate dehydrogenase (DVU3319)

Particle weight estimated by SEC Number of particles (from EM when known) (kDa) per cell; stoichiometry

1000 (1052) 600 (996)c

4000; [abdg]8 300; a?b60

1100 (885)

500; [bb0 a2oNusA]2

530 (409 and 818) 440 (473) 440 (418) 370 (265) 800 670 ( 584) 600 530 370 370 340 300

700d; a7 (C7) and [a7]2 700; a8 800; a8 1200; a2 1400; a2–3 700; a6–7 100; a10–14 200; a16–20 12,000; a9–10 600; a8–12 800; [ab]2 or [ab]4f 1100; a3

This table is adapted from Table 1 in Han et al. (2009). a Entries in italic font indicate protein complexes for which 3D reconstructions were obtained by single-particle electron microscopy (EM). b Homologs of pyruvate ferredoxin oxidoreductase are sometimes fused and sometimes split into multiple chains. c Contribution of the riboflavin synthase a-subunit to the particle weight is not included. d Particle copy number estimated on the assumption that the protein is present in the cell as a D7 14-mer rather than as the C7 heptamer. e EM result indicates either a dimer or tetramer. Size-exclusion chromatography cannot distinguish between these possibilities. f Pyruvate carboxylase is present in some bacteria as a single polypeptide chain and in other bacteria as two chains. We use ab to represent the single-chain form.

224

¨rster et al. Friedrich Fo

within the same bacterial genus. The type of variation that was observed can be illustrated by four examples taken from the study of Han et al. (2009). First, RNA polymerase II was purified mainly as an unusual dimeric complex of the core enzyme and transcription elongation factor NusA. This dimeric stoichiometry was not observed previously in other bacteria. Second, DvH pyruvate:ferredoxin oxidoreductase exists as octomeric complex while the same protein exists as a dimer even in another species of the same genus, Desulfovibrio africanus (Chabrie`re et al., 2001). Third, lumazine synthase (also known as riboflavin synthase b-subunit) shows various species-dependent multimeric states that include pentamers (Morgunova et al., 2005), decamers (Klinke et al., 2005), and icosahedra. Even though DvH lumazine synthase has an icosahedral form, as it does in B. subtilis (Ritsert et al., 1995) and Aquifex aeolicus (Zhang et al., 2001), the pentameric subunits are rotated about 30 relative to the fivefold symmetry axis (Fig. 11.2), resulting in a cage structure with a bigger diameter than previous X-ray model structures reported. Fourth, a DvH homolog of carbohydrate phosphorylase shows a ring like structure in the EM images and eluted as a complex large enough to be at least a hexamer in size-exclusion A

B

Figure 11.2 Comparison of two types of icosahedral structures of lumazine synthase (riboflavin synthase beta subunit) formed by the proteins from Aquifex aeolicus (A) and D. vulgaris Hildenborough (B). The positions and directions of some of the fivefold axes are indicated with red lines to help the comparison of two structures. A single pentameric ring subunit is shown as a yellow ribbon representation with a single monomer shown in red. Note that the vertices of the pentamers are rotated by different amounts in the two complexes, resulting in two structures with different diameters. (A) Semitransparent isosurface representation of the complex from A. aeolicus, computed at the same resolution as that estimated for the structure obtained by electron microscopy for the complex from DvH. A ribbon diagram of the atomic model of the complex (PDB: 1HQK) is embedded in the low-resolution isosurface. (B) Semitransparent isosurface representation of the complex, obtained by electron microscopy, is shown together with the ribbon diagram of a DvH homology model. The pentameric ring is rotated by 30 around the icosahedral fivefold axis to produce a good fit within the EM density map. The homology model was obtained by using the MODBASE (Pieper et al., 2006) server located at http://modbase.compbio.ucsf.edu/modbase-cgi/index.cgi. This figure is adapted from Fig. 2 in Han et al. (2009).

Generation of Molecular Atlases of Prokaryotes

225

chromatography (SEC). The previously known structures of enzymes in this family of proteins were either monomers or dimers. In addition to these unexpected differences in structure, the DvH GroEL particles were initially eluted as a heptameric, single-ring structure with C7 symmetry. However, after the addition of ATP and Mgþ 2, most of the GroEL particles assembled into the more conventional double-ring form with D7 symmetry, suggesting that the double-ring GroEL is probably the major form under the physiological conditions within DvH. At the same time, it is interesting to note that a single-ring form was also purified from mitochondria (Dubaquie et al., 1998) and a few other bacteria (Ferrer et al., 2004; Ishii et al., 1995; Pannekoek et al., 1992). The high structural diversity observed in the study of Han et al. (2009) suggests that variation is much more likely to occur than conservation as regards the size, shape, and subunit composition of large prokaryotic complexes. This observation warns us that the use of previously determined quaternary structures from homologous proteins as templates for interpreting tomograms should be practiced with caution. In the specific case of DvH, it was essential to set up DvH-specific templates for 10 complexes out of 13 protein complexes with identified functions.

3.3. Generation of templates from atomic maps In order to generate template structures that best account for the signal observed in CET, the imaging process has to be simulated as realistically as possible. First, the electron density needs to be calculated from the coordinates and identities of the atoms specified in the PDB. Subsequently the density is convoluted with the CTF, which describes the imaging in the TEM in linear approximation (Fig. 11.3A). For biological materials, the electron optical density is proportional to the electrostatic potential of the macromolecule. To create a template from an existing atomic model, the coordinates of the individual atoms contained in a PDB file are translated into the 3D electron density. The following code executed in Matlab together with the TOM toolbox will at first define the dimensions of the array and ps; then load the information from the PDB file; transform it into an electron density map at the defined ps; center the density according to the center of mass, and finally display the result: cube ¼ 64; ps ¼ 12.6/2; GroELS ¼ tom_pdbread(’1AON_groE.pdb’); GroELSem ¼ tom_pdb2em(GroELS, ps, cube); shift ¼ tom_cm(GroELSem); GroELSems ¼ tom_shift(GroELSem,[(cube/2)þ1 (cube/2)þ1 (cube/2)þ1]-shift); tom_dspcub(GroELSems,2);

226

¨rster et al. Friedrich Fo

B

A

0.5

CTF

Atomic model

-0.5

Electrostatic potential Convolution with CTF Low pass

1

-1

C

Binning

1/11.5 1/5.76 1/3.84 1/2.88 1/2.3 1/1.92 1/1.65 1/1.44 q (nm-1)

1/11.5 1/5.76 1/3.84 1/2.88 1/2.3 1/1.92 1/1.65 1/1.44 q (nm-1)

0.2 0.1 0

CTF

-0.1 -0.2 -0.3

Template

-0.4 -0.5 -0.6

Figure 11.3 Generation of templates from atomic maps. (A) The electrostatic potential of the target structure is calculated from the atomic model. The potential is then convolved with the CTF, low-pass filtered at the first zero of the CTF and binned. (B) The CTF consists of two additive contributions, phase contrast (blue) and amplitude contrast (red), which are damped by the MTF (green). (C) The resulting CTF is lowpass filtered at the first zero.

The procedure places the density into the center of the specified cube, which has to be chosen large enough to accommodate the entire template. All operations are carried out at half the ps prior to binning in order to avoid sampling artifacts. Next, the template needs to be convoluted with the relevant CTF (Fig. 11.3B). The following code defines ps, defocus, av, and Nyquist frequency; calculates the CTF; convolutes the template, and displays the result: ps ¼ 1.26 / 2; defocus ¼ 6.5; av ¼ 200; Ny ¼ 1/(2*ps); ctf ¼ tom_create_ctf(defocus, ones(64,64,64), ps, av); GroELSconv ¼ real(tom_ifourier(ifftshift( fftshift(tom_ fourier(GroELSems)). *ctf))); tom_dspcub(GroELSconv, 2); ˚ for tom_pdb2em The ps needs to be redefined because it is to be given in A and in nm for tom_ctf. In the example above, any negative contrast values

Generation of Molecular Atlases of Prokaryotes

227

beyond the 1st zero of the CTF are eliminated. If the modulation transfer function (MTF) of the detector is known, the template may be convoluted with the MTF as shown above. Finally, the template is filtered to the relevant rmax of the tomograms (Fig. 11.3C; Section 2.3); the volume is binned to the relevant ps; the mean value and standard deviation are normalized to 0 and 1, respectively; the results are displayed and the volume is stored: Resolution ¼ 5; Shell ¼ (cube/2) * (1/(Resolution * Ny)); GroELSconv ¼ tom_bandpass(GroELSems,0,shell,shell/10); GroELSconv ¼ tom_bin(GroELSconv); [mean max min std] ¼ tom_dev(GroELSconv,’noinfo’); GroELS ¼ GroELSconv-mean; GroELS ¼ GroELS./std; tom_dspcub(GroELS, 2); tom_emwrite(’GroELS_12.6A_conv.em’,GroELS); Finally, a spherical mask needs to be generated that defines the area around the template, to which the template matching is constrained. To achieve optimal performance a mask tightly enclosing the reference structure is preferable because densities, which are adjacent to the putative particles in the tomograms, but not considered in the template will otherwise decrease correlation values. The following code generates a spherical mask with a radius of 10 pixels and 2 pixels Gaussian smoothing at the edges, displays the masked template for visual inspection, and stores the mask: mask ¼ tom_spheremask(ones(32,32,32),10,2); tom_dspcub(GroELS.*mask,2); tom_emwrite(’GroELS_mask10.em’,mask); A template and mask generated and stored in this way can be directly used with MolMatch for template matching (Section 4). All of the described steps will have to be repeated for all the desired templates and mask radii should be adjusted based on visual inspection.

4. Template Matching 4.1. Handling MolMatch MolMatch is software to calculate a measure for the local similarity of a tomogram and a template (Fo¨rster, 2005) and is freely available online (www. biochem.mpg.de/foerster/Content_Software/MOLMATCH). Probably the most common measure for similarity of two sets of N-dimensional data is crosscorrelation. In essence, the cross-correlation corresponds to the normalized scalar product of the two data sets (interpreted as an N-dimensional vector).

228

¨rster et al. Friedrich Fo

The normalization comprises subtraction of the respective mean-values and subsequent division by their standard division. The advantage is that no absolute scale for the data sets is required. For detection of patterns in tomograms, the concept has to be extended in two respects: (i) only small local features within the tomograms are searched, that is, the dimensions of tomogram and template differ vastly. (ii) Electron tomograms are incompletely sampled leading to typical distortions of the volume in the beam direction (“missing wedge effect”). To account for both, we implemented a local, constrained correlation function (CCF) to correlate a volume (V), and a template (T; Fo¨rster, 2005). The CCF is constrained to the experimentally sampled fraction of the Fourier space by convoluting T with a point-spread function (PSF) that causes the same “smearing” along the beam direction as V: CCF ðr; j; #; cÞ ¼ PðNx ;Ny ;Nz Þ

VrþD TD;j;#;c PSFD YD rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2 PðNx ;Ny ;Nz Þ PðNx ;Ny ;Nz Þ TD;j;#;c PSFD YD YD ðVrþD V r Þ2 D¼ð1;1;1Þ

D¼ð1;1;1Þ

D¼ð1;1;1Þ

ð11:3Þ In Eq. (11.3) the CCF measures the local similarity of V and the mean-free T at every tomogram voxel r, given a mask Y specifying the local environment and a PSF. The CCF is further a function of the template orientation, specified by three angles (j, #, c) (subsequent rotations around the z-, x-, and z-axis). The three volumes T, Y, and PSF are of much smaller dimensions than the tomogram V; they are pasted into the tomogram such that r ¼ (1,1,1) corresponds to the center of all three volumes of the same dimensions. All expressions of the CCF can be calculated very efficiently using Fourier transformations (Fo¨rster, 2005; Ortiz et al., 2006; Roseman, 2004). MolMatch is a C-based program to calculate the above expression. More precisely, the maximum value of the CCF with respect to the orientation (j, #, c) and the corresponding orientation are stored because the amount of data would be immense otherwise. All input volumes need to be prepared externally, for example, in MATLAB. The mask has been prepared above and the PSF for a volume of a specified dimension (dim), minimum- and maximum-tilt-angles (mintilt and maxtilt) can be generated with the following script: dim ¼ 32;mintilt ¼60;maxtilt ¼ 60; psf ¼ zeros(dim,dim,dim); psf(1,1,1) ¼ 1; wedge ¼ av3_wedge(psf, mintilt, maxtilt); psf ¼ real(tom_ifourier(tom_ fourier (psf ).*fftshift(wedge))); tom_emwrite(’psf_60þ60.em’,psf );

Generation of Molecular Atlases of Prokaryotes

229

After preparation of the required input volumes MolMatch can be invoked in a Unix shell. For the given example of GroEL, MolMatch could be run the following way: mpirun -np 36 molmatch.exe Lepto_phantom_cell.em_SNR_0.5_BPWeiProj_0.5.em GroELS_12.6A_conv.em Out 0 50 10 0 50 10 0 90 10 psf_-60þ60.em GroELS_mask10.em 256 Here, MolMatch is preceded by the specification of the number of used processors for the MPI library (36 processors in this case). The first three arguments following MolMatch are the tomogram file (Lepto_phantom_cell. em_SNR_0.5_BPWeiProj_0.5.em), the template (GroELS_12.6A_conv.em), and the output filename (Out). The next nine numbers specify the angular sampling: for j, c, and # the minimum, maximum, and the increment are specified. Since GroEL exhibits a D7 point symmetry, j and c can be restricted to 0–50 and # to 0–90 at the chosen increment of 10 for all angles. Asymmetric templates will generally require the full range of 0–360 for j and c and 0–180 for #. The angular increment should be chosen according to the targeted resolution. For example, if the template volume spans 20 nm in one dimension and the targeted resolution is 4 nm 1, the increment should be at least a sin(4.2/20) 20 , but better only 10 as oversampling is generally beneficial. The angular range is followed by the PSF (psf_60þ60.em) and the mask (GroELS_mask10.em). The last parameter is of entirely technical nature: it specifies how the volumes are subdivided into smaller cubes to facilitate computation of large volumes that exceed the available memory in the used computers. For typical applications, it should be chosen equal to the smallest dimension of the tomogram, typically the third dimension (z).

4.2. Creating motif lists The MolMatch output is used to generate “motif lists,” that is, lists containing coordinates and orientations of putative targets. MolMatch produces two output volumes, from which the motif lists are generated: Out.ccf.norm (maximum of the local constrained correlation with respect to the orientation), and Out.ang (indices of orientations corresponding to maximum correlation). The correlation function should ideally exhibit distinct maxima at putative particles, whereas the orientation indices are only interpretable when the angular range used as input is known. The motif list is generated sequentially: First, Out.ccf.norm is screened for its maximum value. The value and the corresponding coordinates and orientation (three Euler angles calculated from the index in Out.ang and the rotational parameters for MOLMATCH) are stored in the motif list. Out.ccf.norm is then set to a small value (1) in a sphere with defined radius centered at the peak to prevent recurrent detection. The procedure is repeated nparts times to detect nparts putative particles.

230

¨rster et al. Friedrich Fo

Optionally, particles that are too close to the border of the tomogram (specified by tomoedge) are rejected from the motif list. corr ¼ tom_emread(‘Out.ccf.norm’); angfile ¼ ’Out.ang’; nparts ¼ 1000; radius ¼ 10; tomoedge ¼ 15; maxphi ¼ 50; maxpsi ¼ 50; maxthe ¼ 90; minphi ¼ 0; minpsi ¼ 0; minthe ¼ 0; incr ¼ 10; motlfile ¼ ’motl_GroEL.em’; motiflist ¼ av3_createmotl(corr, angfile, nparts, radius, tomoedge, maxphi, maxpsi, maxthe, incr, minphi, minpsi, minthe); tom_emwrite(motlfile, motiflist); The motif lists then need to be assessed to discard false-positive detections (Section 5). In addition, motif lists can be used to refine and classify 3D averages of the detected particles using the AV3 package.

4.3. Scoring A scoring function provides a measure for the probability of an event. For example, the CCF described in Eq. (11.3) is a score for the probability that a feature in a tomogram corresponds a specific macromolecular complex. To assess the performance of a scoring function, it has to be calibrated using a defined test data set (also called training data). Thereby, the likelihood of falsepositive discoveries is calculated as a function of the score threshold that discriminates true from false hits. Such concepts for assessment of performance are widely used in other areas of biology, for example, for assigning peptides to tandem MS spectra (Keller et al., 2002) or for assessing the statistical significance of sequence alignments (Henikoff, 1996; Rost, 2002). How the performance of visual proteomics can be assessed is described in detail in Section 5. Scoring functions often rely on more than one single readout but combine several readouts as linear combinations of subscores. Thereby, the weight of such subscores is optimized in training data sets using a linear discriminate analysis of true and false positives. Once optimized, the combined discrimination power of all subscores is likely superior to single readouts. A scoring function SF for visual proteomics (Beck et al., 2009) that relies on three different knowledge-based, empirical readouts has been used during the Leptospira proteome project: SF ¼ A∗CCF Par þ B∗

CCF Par CCF Par þ C∗ CCC TopComp CCF TopDecoy

ð11:4Þ

In Eq. (11.4), A, B, and C are weighting factors for the subscores, CCFPar is the CCF of the targeted template as described in Eq. (11.3), CCFTopComp is the highest CCF value of any other (competing) template within the same

Generation of Molecular Atlases of Prokaryotes

231

position of the tomogram, and CCFTopDecoy is the highest of any nontargeted, decoy template, which can be random structures or shapes. The rational of this SF is to reward a penalty to hits in positions where competing or decoy templates have a high CCF. In principle, the list of subscores could be extended by quantities that are not derived from cross-correlations (e.g., occupied volume of templates or the local SNR of the tomogram). To apply the scoring function to the CCFs calculated by MolMatch for all targets and decoy templates used in a study, the individual motif lists have to be merged as layers into a single 3D motif list file. This can be done automatically and for the entire template library at once using motlgen, a wrapper script for av3_createmotl: load Templates load Decoys help motlgen fileroot ¼ ’PhantomCell_SNR_0.5_BPWeiProj_0.5’; motlgen ( fileroot, Templates, Decoys); Thereby, Templates is an array that contains the library annotation, a table containing names, sizes, and other definitions of the template library. The command “help motlgen” will display the specifics of the definitions contained in the annotation table in addition to the help text on the motlgen script. Decoys is the same as Templates but for additional nontargeted templates that are used for scoring. The motlgen wrapper executes the peak extraction for multiple templates and stores the retrieved information in motif lists that are structured as usual but contain competing assignments from other templates as layers along the third dimension. Afterward, the score can be calculated: score_motl ¼ score_matches(’motl_’,Templates,Decoys); Thereby, the number of matches to be assigned positive (class 1) based on the score threshold defined by Templates. This number strongly influences the performance and must be carefully chosen (see Section 5 for detail). Finally, the resulting hits can be assessed for double assignments (coordinates that are assigned to more than target) in space: fetch_double_hits(score_motl, Templates); This script identifies double hits, displays statistics, and reassigns the competing hits that have a lower relative score within their score distribution as false prior to visualization (see below).

4.4. Visualization of molecular atlases For visualization of the detection results the template can be rotated and positioned at the determined coordinates and the corresponding angles. These particles can be displayed in the context of the typically manually

232

¨rster et al. Friedrich Fo

segmented tomogram. In most cases, an isosurface representation is chosen because it is much easier to capture the content of a tomogram in this representation compared to common volume-rendered approaches. A convenient, albeit costly, program for 3D visualization is Amira. It is also possible to develop plug-ins for Amira. The Frangakis laboratory has developed such a plug-in to display templates in a tomogram at positions and orientations specified in a motif list (Pruggnaller et al., 2008). The plug-in is freely available (www.biophys.uni-frankfurt.de/frangakis/Amiratools.htm). An alternative, professional visualization software is 3D Studio Max. The software uses realistic ray-tracing technology and scripting is possible as well. This software has been used to visualize ribosomal morphologies in different studies (Brandt et al., 2009; Ortiz et al., 2006).

5. Assessment of Performance A major challenge in the visual proteomics approach is to discriminate true-positive from false-positive detections. Due to the moderate SNR of the original data, the observed distribution of cross-correlation values comprises an overlay of the distributions of true- and false-positive hits. In order to assess the performance, the amount of overlap of both distributions has to be estimated. Such an approach is commonly used in MS-based proteomics to determine false-positive discovery rates for peptides assigned to tandem MS spectra (Keller et al., 2002). It does not attempt to associate individual hits with the true- or false-positive distribution, but builds a statistical model that allows calculating the likelihood for all matches in the observed distribution to be false-positive hits. The performance of visual proteomics depends on a variety of tomogram-specific parameters that affect the SNR and consequently all the targeted protein complexes in a similar way. These are primarily acquisition settings, specimen thickness, and molecular crowding. However, also target-specific parameters, such as molecular weights, cellular abundance of the target, as well as cellular abundance of protein complexes competing for assignments play a critical role. Therefore, the performance cannot be determined in general but needs to be estimated separately for each individual template contained in the library.

5.1. Estimating true-positive discovery rates from artificial tomograms One way of estimating the performance of visual proteomics is to simulate the entire process in silico. Thereby, artificial tomograms are generated as realistically as possible and subjected to template matching. Since the

Generation of Molecular Atlases of Prokaryotes

233

position and orientation of all templates in the artificial tomogram is known, the distribution of true- and false-positive hits can be calculated separately and false as well as true-positive discovery rates can be determined. This approach does not only take the image formation mechanism into account, but also a priori information about template abundances, and, if procurable, the abundances of protein complexes competing for assignments as well as the total protein concentration inside the cell. This information can be approximated using quantitative proteomics approaches (Malmstrom et al., 2009). As an additional benefit, the performance for the desired template can be assessed in conjunction with acquisition parameters before the actual data collection. The first step to in silico assessment is to build an artificial tomogram (phantom cell) that contains the correct copy number of the template structures to account for their concentration within the cell. The following code makes the necessary definitions, generates an empty phantom cell and a “book keeping” volume that is required to avoid multiple assignments: nTemp ¼ 10; radius ¼ 8; offset ¼ 10; volFile ¼ ’phantom_cell.em ’; bkFile ¼ ’book_keeping_volume.em ’; bkVol ¼ ones (128,128,128); tom_emwrite(bkFile, bkVol); vol ¼ zeros (128,128,128); tom_emwrite(volFile, vol); Thereby, nTemp defines the number of molecules to paste into the phantom cell, radius defines the local proximity around the center of mass occupied by the template, offset defines the size of an empty edge of the phantom cell, volFile the file name of the phantom cell, and bkFile the name of the “book keeping” volume. Next, the template structure is loaded and pasted 10 times into the phantom cell at random positions and orientations. These positions are bookmarked in the “book keeping” volume by setting the density in their proximity to zero. The pasting procedure will not place templates into previously bookmarked positions. Finally, the modified phantom cell and book keeping volume as well as the respective motif list (motl) are stored. GroELS ¼ tom_emread(’GroELS_12.6A.em ’); [motl bkVol vol] ¼ paste_template(nTemp, GroELS.Value, radius, bkFile, volFile, offset); tom_emwrite(’GroELS_motl.em’, motl); tom_emwrite(bkFile, bkVol); tom_emwrite(volFile, vol); This procedure can be repeated multiple times to account for multiple templates. In addition, decoy density can be added analogously to simulate molecular crowding. Next, adequate noise has to be added to simulate the image formation process. This can be done by adding a contribution where the CTF- and

234

¨rster et al. Friedrich Fo

MTF-convoluted signal is multiplied with noise (“quantum noise”) and a pure MTF-convoluted noise contribution (“background noise”; Forster et al., 2008). To simulate the tomogram, noisy projections are calculated and then reconstructed in 3D. The following code makes definitions as described above and simulates the imaging of a given phantom cell at a given ps, defocus, av, angular increment scheme (angles), and SNR: ps ¼ 1.26; defocus ¼ 6.5; av ¼ 200; angles ¼ [63:1.5:40 38:2:38 40:1.5:63]; SNR ¼ 0.5; tomogen(’Lepto_phantom_cell.em’, SNR, angles, defocus, ps, av); The resulting volume can be directly used for template matching as described in Section 4. The noise level should be estimated from real data sets. This can, for example, be done in the following way: Artificial tomograms are generated at different noise levels and subjected to template matching as well as real data sets. Thereby, an easily detectable template of sufficient size and abundance, for example the ribosome, should be targeted. To select the adequate noise level, the distributions of cross-correlation values of candidate hits from real data sets and phantom cells are compared. Once the noise level is calibrated, all reference structures contained in the library template matched to the phantom data using MolMatch as described in Section 4. Afterward, motif lists are generated for all templates using the following code (see Section 4.4 for detail): load Templates fileroot ¼ ’PhantomCell_SNR_0.5_BPWeiProj_0.5’; motlgen ( fileroot, Templates); Finally, the performance is assessed by comparing two sets of motif lists: the one containing positions of templates originally pasted into the phantom cell and the other containing matched positions from the same cell. The following code loads the template library annotation, defines the file locations of both motif lists and launches the performance calculation: load Templates MatchedMotlRoot ¼ ’MatchMotl_’; PastedMotlRoot ¼ ’PasteMotl_’; assess_ performance(MatchedMotlRoot, PastedMotlRoot,Templates); This procedure produces a number of plots. The observed distribution of cross-correlation values is overlaid with the corresponding distributions of the true- and false-positive hits. The specificity Sp and the sensitivity Se are calculated and plotted against the cross-correlation threshold, whereby specificity is defined in Eq. (11.5): Sp ¼

nTruePos nTruePos þ nFalsePos

ð11:5Þ

235

Generation of Molecular Atlases of Prokaryotes

with nTruePos being the number of true-positive hits and nFalsePos being the number of false-positive hits at a given cross-correlation threshold; and sensitivity is defined in Eq. (11.6), Se ¼

nTruePos nTemplates

ð11:6Þ

with nTemplates being the number of templates accounting for the given target protein complex that are contained in the tomogram. The output allows an assessment if the desired protein complex can be detected under the given cellular and imaging conditions and it reveals the dependency of the performance on the CCC threshold (Fig. 11.4).

5.2. Performance assessment in real data sets Alternatively, the performance of visual proteomics can be assessed in real data sets based on a priori knowledge about the template structures or their spatial distribution. This approach, although preferable over simulations, is not generally applicable to all template structures. Matching with a mirrored template of “nonnative” handedness can provide complementary information about the distribution of cross-correlation values. When the 1

120

0.75 80 n

0.5 40

0.25

0.45

0.5 CCC

0.55

0.45 0.5 0.55 CCC threshold

Figure 11.4 Performance analysis of template matching in artificial tomograms. RNA polymerase, a high abundant protein complex of only moderate size was chosen as an example. The observed distribution of cross-correlation coefficients (blue, left) is overlaid with the corresponding distributions of the true- (green) and false-positive hits (red). The latter two do overlap to a large extend, therefore the performance is imperfect. Specificity (red) and sensitivity (green) are shown on the right. At an arbitrarily chosen CCC threshold of 0.48 about 75% of the discovered hits are true positive, however, less than 50% of all targets are discovered.

236

¨rster et al. Friedrich Fo

90 Native handedness Mirrored

n

60

30

0.15

0.35 CCC

Figure 11.5 Distribution of the cross-correlation coefficients (CCCs) extracted obtained from a tomogram of a Leptospira cell after template matching with the ribosome as template (black) and the mirrored ribosome as decoy template (grey).

distribution of cross-correlation values observed using the 70S ribosome as a template is compared to the corresponding distribution observed for the mirrored ribosome, a shift is apparent (Fig. 11.5). The number of single particles to the right of the cross-over of both curves roughly matches the cellular abundance of the ribosome indicating that at this point both distributions decline into noise correlation. Another promising approach is to match template structures of different subunits of a protein complex separately, for example, the small and the large ribosomal subunit, to check if they reasonably colocalize in translating ribosomes. The implementation of this control for tightly associated subunits is, however, quite challenging: The local cross-correlation has to be tightly constrained in order to mask out the adjacent subunits, barely including the surrounding solvent. However, the contrast arises primarily from the difference in density between the target and the surrounding solvent and hence template matching performance decreases. Only if resolution and SNR are sufficient to recognize intrinsic features within the different subunits, this approach might be employed, which is not realistic with the current experimental setup, even for very large protein complexes. Alternatively, loosely associated structures provide a good target for similar controls. The colocalization of ribosomes engaged in polysomal arrays (Beck et al., 2009; Brandt et al., 2009) or even RNA polymerases associated with translating ribosomes can account as examples for such cases. Furthermore, a priori knowledge of the cellular localization or orientation of protein complexes can provide clues about the performance. For example, membrane associated protein complexes exhibit a specific positioning and orientation relative to the membrane (Ortiz et al., 2006) and RNA polymerases

Generation of Molecular Atlases of Prokaryotes

237

cluster to the genomic DNA. Finally, cellular treatments such as drug or stress stimuli might induce a considerable up- or downregulation of individual target protein complexes (Beck et al., 2009; Malmstrom et al., 2009). Such perturbations should change the observed distribution of crosscorrelation values; however, impose the challenge of comparing data obtained from different tomograms and cells (Section 2.3). Subtle changes in the SNR and stochastic variations of the expression level within the cell population complicate the data evaluation in this case.

6. The Spatial Proteome of L. interrogans L. interrogans is a Spirochete with a strongly elongated and helically coiled cell shape and a pathogen that causes Leptosirosis in a wide range of species. Since the diameter of a cross-section of a typical cell is no more than 100–180 nm, L. interrogans is an ideal specimen for cryo-ET allowing for excellent electron beam penetration and the acquisition of tomograms with relatively good SNR (Beck et al., 2009). At first, the proteome was measured by tandem MS in shotgun mode. Gene products accounting for about 62% of the 3700 open reading frames were detected. A query for the expressed protein complexes in structural databases retrieved 26 candidate protein complexes of a certain minimal size (>250 kDa) and a certain sequence conservation that might be suitable for template matching (Table 11.2). Targeted proteomics was used to determine the cellular protein concentration on a proteome-wide scale (Malmstrom et al., 2009). Protein abundance was found to range over 3.5 orders of magnitude. The most abundant protein detected in this study was LipL32, an integral part of the peptido glycan layer, with 40,000 copies per cell. The most abundant candidate protein complex was the ribosome with about 4500 copies per cell, occupying 20% of the cytoplasmic volume. Another 10 protein complexes had abundances of at least 100 copies per cell, all others were of low abundance. Next, a subset of protein complexes was selected to estimate the performance of visual proteomics in silico (Fig. 11.6A). For this purpose, phantom cells were generated that contained the templates at their cytoplasmic concentration in a membrane enclosed space and the image formation process was simulated to generate artificial tomograms and subjected to template matching (Fig. 11.6B). Ideally, the distribution of cross-correlation values obtained by template matching for a specific protein complex can be described by an overlay of two Gaussian functions, one accounting for the false-positive and the other for the true-positive distribution. The falsepositive distribution is, however, likely composed of signals from several different species. Therefore, the data interpretation based curve shape per se

Table 11.2 Structures of protein complex candidates for template matching

Template

Reference

Oligomeric state

Molecular weight (kDa)

Complexes per cell

Ribosome RNA polymerase ATP synthase GroEL/ES GroEL Glutamine synthase Citrate synthase Transaldolase clpP Carbamoyl phosphatase Enolase Aspartyl-tRNA-synthase Cytosolic amino-peptidase GTP-cyclohydrolase clpB Ornithine carbamoyltransferase Phoshoribosyl-pyrophosphatase Acetyl-CoA-carboxylase Hsp15 Dihydrolipoamide acetyltransferase HslU-V LIC11615_UbiD Aspartate carbamoyltransferase Acetolactate synthase Bacterioferretin Lumazine synthase

pdb_ 2AW7_2AWB pdb_2GHO pdb_1QO1 pdb_1aon pdb_1KP8 pdb_2gls pdb_2h12 pdb_1vpx pdb_1YG6 pdb_1bxr pdb_1w6t pdb_1eqr pdb_1gyt pdb_1GTP emd_1243 pdb_1a1s pdb_1dkr pdb_1vrg pdb_2BYU pdb_1dpb

Multicomponent protomer Multicomponent protomer 6-mer 21-mer 14-mer 12-mer 6-mer 20-mer 14-mer 8-mer 8-mer 3-mer 12-mer 20-mer 6-mer 12-mer 6-mer 6-mer 12-mer 24-mer

2200 340 450 870 810 620 300 520 304 650 390 200 660 500 600 420 210 350 220 640

4500 3000 1500 1100

pdb_1G3I pdb_2idb pdb_1f1b pdb_1OZF pdb_1BFR pdb_1NQU

24-mer 6-mer 12-mer 4-mer 24-mer 60-mer

830 350 300 250 450 1000

20 17 10 10 4 4

Accession numbers, oligomeric states, molecular weights, and cellular abundances are listed.

320 220 150 140 120 100 100 90 75 70 65 55 40 40 30

239

Generation of Molecular Atlases of Prokaryotes

A

1

2

3

B

4

5

6 D

C

9

1

4 5

7

200

8

9

160 120

n

n

2

80

100 20 0

0.44

0.54 CCC

1 Normalized score

Figure 11.6 In silico estimation of the performance of visual proteomics. (A) Subset of protein complexes selected for performance analysis is shown surface rendered with chain traces in black (if applicable): 1. 70S ribosome, 2. RNA polymerase II, 3. GroEL, 4. GroEL-ES, 5. Hsp, 6. Clp B, 7. Clp P, 8. HslU-V, 9. ATP-synthase. (B) Centered slices through a Leptospira phantom cell shown perpendicular (left) and parallel (right) to the electron optical axis, as well as before (top) and after (bottom) simulating the image formation process. (C) Distribution of cross-correlation coefficients obtained from real data sets by template matching. The area marked under the curve corresponds to the expected abundance of the template; the arrowhead marks the intuitive threshold for assignments as indicated by the curve shape. (D) Score distributions obtained from phantom cells of the observed (blue), true- (green) and false-positive hits (red). The arrowhead marks the threshold that accounts for a specificity of 40%.

will not allow the deduction of false-positive discovery rates or specificities (Fig. 11.6C). Since the position and orientation of all protein complexes in the artificial tomograms are known, the distribution of true- and falsepositive hits can be calculated in order to determine specificities. To improve the discrimination of true- from false-positive hits, a scoring function was developed that not only incorporates the cross-correlation value of the target but also the cross-correlation values of competing templates and a number of decoy templates within the same position of the tomogram (Fig. 11.6D, Section 4.3). This scoring scheme outperformed the classical workflow in some cases, particularly when protein complexes of similar structural signature were competing for assignments. A set of 18 different phantom cells was evaluated multiple times to investigate the performance of visual proteomics under the influence of noise, missing wedge direction, protein abundance, and molecular

240

¨rster et al. Friedrich Fo

MW (MDa)

crowding. In the course of these simulations, the following factors were found to have a profound impact: (i) background noise may reduce the performance depending on the MTF of the particular CCD camera used in a study; (ii) the missing wedge can introduce angular bias to the discovery rate of templates with anisotropic signal content, such as the elongated ATP-synthase; (iii) the specificity increases with the template abundance; and (iv) decreases with the degree of molecular crowding. The last two effects are generally more pronounced for smaller than for larger templates (Fig. 11.7). While the specificity achieved for high abundant megadalton complexes is satisfactory, true-positive discovery rates higher than 50% are difficult to achieve when protein complexes of smaller molecular weights are targeted. For protein complexes of very low cellular abundance, it is quite a challenge to obtain robust statistical models. Tomograms covering subvolumes of 37 different L. interrogans cells were acquired. A subset of 12 tomograms of similar SNR was selected and subjected to template matching and scoring (Fig. 11.1). The local concentration of the targeted protein complexes varied within and across data sets: the cells displayed an average ribosome concentration of 20 mM ( 40 mg/ml) in the cytoplasm, but the local concentration ranged from 5 to 30 mM (10–65 mg/ml). The local fluctuations in case of total GroEL together with (GroEL-ES) were larger and ranged from 8 to 100 mM ( 0.5–6.5 mg/ml). In most, but not all tomograms, the ratio of ribosomes to RNA polymerases was locally maintained. Stress-induced changes in the GroEL to GroEL-ES ratio and Hsp abundance were apparent, but could not

2.0 1.5 1.0 0.5 0.25 0.4

1 Specificity

Figure 11.7 The performance of visual proteomics depends on the size of the targeted protein complexes. The specificity for nine different protein complexes with molecular weights ranging from 0.2 to 2.5 MDa was determined in silico at a pixel size of 1.26 nm, SNR of 0.5 with an underfocus of 6.5 mm (first zero of CTF at 4 nm). The specificity refers to the fraction of true-positive hits at a discovery rate of 0.5.

Generation of Molecular Atlases of Prokaryotes

241

be substantiated by statistical significance testing. Some templates provided auxiliary information for the orthogonal validation of the data obtained to assess the performance in real data sets as described above (Section 5.2).

7. Outlook A number of limitations to visual proteomics might be overcome by further technical developments. Currently, the most critical limitation is the moderate SNR of CET. The development of alternative detection concepts is probably most important to capture the entire signal emanated from the sample. In addition, the introduction of phase plates in conjunction with CS correctors may optimize the contrast transfer. Specimen thinning techniques will be pivotal to image cell types, which exceed 500 nm or more; thus, application of visual proteomics to eukaryotic cells depends on progress in this field. In-depth characterization of the vast majority of large protein complexes in systems under study, by biochemical as well as structural techniques, has become feasible and will largely contribute to template library reliability and completeness. The further development of classifiers and scoring functions for pattern recognition holds great potential to increase true-positive discovery rates. To detect low abundant protein complexes with confidence, the throughput has to be increased on the data acquisition and postprocessing side. The implementation of a better dosage control during the data acquisition will be necessary to compare large number of data sets. The signal in visual proteomics primarily arises from the contrast given by the difference in density between the targeted protein complex and the surrounding solvent. Improved contrast, for example through phase plates, is therefore likely to increase the performance in the future. One major limitation is given by nature and will remain: The dosage of electrons that can be applied to a specimen ultimately limits the attainable resolution of CET. As a consequence, visual proteomics in conjunction with CET is not likely to become broadly applicable. The biological questions that can be answered will be limited to protein complexes of a certain minimal molecular weight that can be targeted by template matching. Such relatively large protein complexes, however, function in central cell biological processes, such as transcription, translation, protein folding, and degradation.

ACKNOWLEDGMENTS B. G. H. was supported by the U.S. Department of Energy Contract DE-AC0205CH11231. F. F. is grateful to a Career Development Award from the Human Frontier Science Program.

242

¨rster et al. Friedrich Fo

REFERENCES Beck, M., et al. (2009). Visual proteomics of the human pathogen Leptospira interrogans. Nat. Methods 6, 817. Bohm, J., et al. (2000). Toward detecting and identifying macromolecules in a cellular context: Template matching applied to electron tomograms. Proc. Natl. Acad. Sci. USA 97, 142–145. Brandt, F., et al. (2009). The native 3D organization of bacterial polysomes. Cell 136, 261. Cardone, G., et al. (2005). A resolution criterion for electron tomography based on crossvalidation. J. Struct. Biol. 151, 117. Chabrie`re, E., et al. (2001). Crystal structure of the free radical intermediate of pyruvate: Ferredoxin oxidoreductase. Science 294, 2559. Crowther, R. A., et al. (1970). Three dimensional reconstructions of spherical viruses by Fourier synthesis from electron micrographs. Nature 226, 421. Dong, M., et al. (2008). A “tagless” strategy for identification of stable protein complexes genome-wide by multidimensional orthogonal chromatographic separation and iTRAQ reagent tracking. J. Proteome Res. 7, 1836. Dubaquie, Y., et al. (1998). Purification of yeast mitochondrial chaperonin 60 and co-chaperonin 10. Methods Enzymol. 290, 193. Fernandez, J. J., et al. (2006). CTF determination and correction in electron cryotomography. Ultramicroscopy 106, 587. Ferrer, M., et al. (2004). Functional consequences of single:double ring transitions in chaperonins: Life in the cold. Mol. Microbiol. 53, 167. Fo¨rster, F. (2005). Quantitative Analysis of macromolecules in cryoelectron tomograms using correlation methods. Dissertation. Technical University of Munich, Munich. Fo¨rster, F., et al. (2008). Classification of cryo-electron sub-tomograms using constrained correlation. J. Struct. Biol. 161, 276. Han, B. G., et al. (2009). Survey of large protein complexes in D. vulgaris reveals great structural diversity. Proc. Natl. Acad. Sci. USA 106, 16580. Hanson, P. I., and Whiteheart, S. W. (2005). AAAþ proteins: Have engine, will work. Nat. Rev. 6, 519. Henderson, R. (2004). Realizing the potential of electron cryo-microscopy. Q. Rev. Biophys. 37, 3. Henikoff, S. (1996). Scores for sequence searches and alignments. Curr. Opin. Struct. Biol. 6, 353. Ishii, H., et al. (1995). Equatorial split of holo-chaperonin from Thermus thermophilus by ATP and Kþ. FEBS Lett. 362, 121. Keller, A., et al. (2002). Empirical statistical model to estimate the accuracy of peptide identifications made by MS/MS and database search. Anal. Chem. 74, 5383. Klinke, S., et al. (2005). Crystallographic studies on decameric Brucella spp. Lumazine synthase: A novel quaternary arrangement evolved for a new function? J. Mol. Biol. 353, 124. Kuhner, S., et al. (2009). Proteome organization in a genome-reduced bacterium. Science 326, 1235. Lawrence, M. C. (1992). Least-squares method of alignment using markers. In “Electron Tomography,” ( J. Frank, ed.), p. 197. Plenum Press, New York. Leis, A. P., et al. (2006). Cryo-electron tomography of biological specimens. IEEE Signal. Process. Mag. 23, 95. Malmstrom, J., et al. (2009). Proteome-wide cellular protein concentrations of the human pathogen Leptospira interrogans. Nature 460, 762.

Generation of Molecular Atlases of Prokaryotes

243

Morgunova, E., et al. (2005). Crystal structure of lumazine synthase from Mycobacterium tuberculosis as a target for rational drug design: Binding mode of a new class of purinetrione inhibitors. Biochemistry 44, 2746. Nickell, S., et al. (2005). TOM software toolbox: Acquisition and analysis for electron tomography. J. Struct. Biol. 149, 227. Nickell, S., et al. (2006). A visual approach to proteomics. Nat. Rev. 7, 225. Ortiz, J. O., et al. (2006). Mapping 70S ribosomes in intact cells by cryoelectron tomography and pattern recognition. J. Struct. Biol. 156, 334. Pannekoek, Y., et al. (1992). Identification and molecular analysis of a 63-kilodalton stress protein from Neisseria gonorrhoeae. J. Bacteriol. 174, 6928. Pieper, U., et al. (2006). MODBASE: A database of annotated comparative protein structure models and associated resources. Nucleic Acids Res. 34, D291. Pruggnaller, S., et al. (2008). A visualization and segmentation toolbox for electron microscopy. J. Struct. Biol. 164, 161. Ritsert, K., et al. (1995). Studies on the lumazine synthase/riboflavin synthase complex of Bacillus subtilis: Crystal structure analysis of reconstituted, icosahedral beta-subunit capsids with bound substrate analogue inhibitor at 2.4 A˚ resolution. J. Mol. Biol. 253, 151. Roseman, A. M. (2004). FindEM—A fast, efficient program for automatic selection of particles from electron micrographs. J. Struct. Biol. 145, 91. Rost, B. (2002). Enzyme function less conserved than anticipated. J. Mol. Biol. 318, 595. Vale, R. D. (2000). AAA proteins. Lords of the ring. J. Cell Biol. 150, F13. Winkler, H. (2007). 3D reconstruction and processing of volumetric data in cryo-electron tomography. J. Struct. Biol. 157, 126. Zanetti, G., et al. (2009). Contrast transfer function correction applied to cryo-electron tomography and sub-tomogram averaging. J. Struct. Biol. 168, 305. Zhang, X., et al. (2001). X-ray structure analysis and crystallographic refinement of lumazine synthase from the hyperthermophile Aquifex aeolicus at 1.6 A resolution: Determinants of thermostability revealed from structural comparisons. J. Mol. Biol. 306, 1099.

C H A P T E R

T W E LV E

Cryoelectron Tomography of Eukaryotic Cells Asaf Mader,*,† Nadav Elad,*,† and Ohad Medalia*,† Contents 1. Introduction 2. Specimen Preparation 3. Relying on Correlative Light and Electron Microscope for Cellular Structural Study of Eukaryotic Cells 4. Cryoelectron Tomography of Cytoskeleton-Driven Processes 5. Cryotomography of Midbodies 6. Structural Analysis of the Nuclear Pore Complex by Cryo-ET 7. Concluding Remarks Acknowledgments References

246 248 252 254 256 257 260 260 260

Abstract Biological processes involve a high degree of protein dynamics resulting in a constant remodeling of the cellular landscape at the molecular level. Orchestrated changes lead to significant rearrangement of the eukaryotic cytoskeleton and nuclear structures. Visualization of the cellular landscape in the unperturbed state is essential for understanding these processes. The development of cryoelectron tomography (cryo-ET) and its application to eukaryotic cells has provided a major step forward toward better realizing these processes. In conjunction with rapid freezing techniques, that is, vitrification by plunge-freezing and high-pressure freezing, cryo-ET is most suitable for investigating cellular ultrastructures in a close-to-life state. Here, we review the application of cryo-ET to the study of eukaryotic cells, with special emphasis on sample preparation, cytoskeleton organization, and macromolecular structures observed at a resolution of 4–6 nm.

* Department of Life Sciences, Ben Gurion University of the Negev, Beer-Sheva, Israel The National Institute for Biotechnology in the Negev, Ben Gurion University of the Negev, Beer-Sheva, Israel

{

Methods in Enzymology, Volume 483 ISSN 0076-6879, DOI: 10.1016/S0076-6879(10)83012-5

#

2010 Elsevier Inc. All rights reserved.

245

246

Asaf Mader et al.

1. Introduction Nearly every major process that occurs in a cell is conducted by multiprotein complexes, often referred to as molecular machines. These complexes, composed of proteins and nucleic acids, form functional units capable of performing specific cellular tasks. Detailed structural studies could shed light on understanding individual cellular processes. Such information is needed to comprehend the complexity of the cell and to understand how multiple protein assemblies participate in tightly regulated processes within the crowded cellular environment (Campbell, 2008). Since its inception, electron microscopy (EM) has been central for the visualization of cellular structures with high spatial resolution. Indeed, our basic knowledge of cells, their various organelles, membranous compartments, cytoskeleton filaments, and even small cytoplasmatic particles, such as ribosomes, have been provided by EM imaging. Conventional sample preparation techniques, however, require physical and chemical perturbations, such as chemical fixation, dehydration, plastic-embedding, sectioning, and heavy metal staining. Technical advents in microscopic imaging techniques, such as cryoelectron microscopy (cryo-EM), have led to unprecedented capabilities for examining unperturbed cellular structures. Cryo-EM, in conjunction with automated electron tomography (ET; Dierksen et al., 1992, 1993), now allows three-dimensional (3D) visualization of polymorphic cellular structures through the recording of projection series obtained at varying angles around one or two axes, followed by reconstruction of the 3D volume (Koning and Koster, 2009; Lucic et al., 2005). Thus, cryoelectron tomography (cryo-ET) enables 3D investigation of prokaryotes (Gan et al., 2008; Iancu et al., 2010; Konorty et al., 2008; Li and Jensen, 2009; Milne and Subramaniam, 2009), eukaryotes (Henderson et al., 2007; Kurner et al., 2004; Medalia et al., 2002), and purified molecular complexes (Saibil, 2000; Sali et al., 2003) embedded in amorphous, vitrified buffers. As such, cryoET bridges structural studies with molecular and cell biology approaches, providing complimentary information toward unraveling the functional networks found within cells (Ben-Harush et al., 2010). Vitrification by rapid freezing (10 5 s) arrests cellular processes instantly (McDowall et al., 2010), circumventing molecular changes owing to chemical fixation, heavy metal staining, or dehydration and, therefore, ensures close-to-life representation of the proteome. Still, many eukaryotic cells are too thick to be imaged directly by cryo-ET, due to the inelastic scattering of electrons that occurs while penetrating a sample thicker than a micron. When sample thickness exceeds the mean free path of the electrons, multiple inelastic scattering occurs, degrading the quality of the image, even when medium voltages (300–400 keV) and energy filters are used (Grimm

247

Cryoelectron Tomography of Eukaryotic Cells

et al., 1996). Dealing with thicker cellular regions and tissues requires physically processing the samples using high-pressure freezing (Moor, 1987) followed by cryosectioning (Al-Amoudi et al., 2004; Ladinski et al., 2010), cryosectioning with a focused-ion-beam (FIB; Marko et al., 2007) or thinning freeze-plunged cells (Rigort et al., 2010). Hence, cryo-ET has been more successfully applied to prokaryotes and viruses, to provide novel structural information. Nevertheless, cryo-ET of eukaryotic cells is possible but is mostly limited to flat thin cells or cell peripheries. Exceptionally, thin eukaryotic cells can be studied as a whole, such as Ostreococcus tauri (Henderson et al., 2007), where all parts of the cells can be reconstructed within a single tomogram (Fig. 12.1). Traditionally, cellular components and events have been tracked and captured using fluorescent-light microscopy. In contrast, cryo-ET only allows study of an area of 2 mm2, representing only 1.5% of the peripheral areas of a typical eukaryotic cell. Therefore, it is difficult to localize specific biological processes or complexes of interest using low-dose cryo-EM conditions. Accordingly, correlating cryo-EM with fluorescent-light microscopy imaging (i.e., correlative light-electron microscopy—CLEM) was recently introduced to unambiguously identify specific cellular components and processes (Chen and Briegel, 2010; Lucic et al., 2008; Plitzko et al., 2009; Robinson et al., 2001; Vicidomini et al., 2010; Weston et al., 2010). Thus, fluorescent microscopy assists by locating and mapping a desired cellular feature, which can later be found under the electron beam

m

A ne

B

n

c g r

gr

250 nm

Figure 12.1 Cryo-ET of entire eukaryotic cell. (A) Single x-y slice through the middle of a single O. tauri cell. In these experiments, the cells were found to be mainly non-dividing, 21.6-nm. Shown are nuclei (n), nuclear envelope (ne), chloroplasts (c), mitochondria (m), Golgi bodies (g), granules (gr), and ribosome-like particles (r). (B) 3-D rendered view of O. tauri cell, 36-nm. Adapted from Henderson et al. (2007).

248

Asaf Mader et al.

and studied in 3D at higher magnification by cryo-ET. Hence, combining these two methodologies allows for a resolving of the molecular architecture of cells in terms of specific functional processes and states. In this review, we address the application of cryo-ET to the analysis of the cytoplasm and nuclear periphery of eukaryotic cells. In particular, we will focus on cytoskeleton-based structures, macromolecular complexes such as the nuclear pore complexes (NPCs), and different methods allowing for localization of cellular processes.

2. Specimen Preparation Cryo-ET provides a unique opportunity to study macromolecules in situ. However, this technique is strictly limited by specimen thickness, as described above. Therefore, when growing cells on EM grids, conditions should be optimized to allow cells to spread maximally. Unprecedented information has been deduced by studying whole thin prokaryotic cells by cryo-ET, without the need for sectioning (Komeili et al., 2006; Ku¨rner and Baumeister, 2006; Seybert et al., 2006; Zhang et al., 2007). On the other hand, sample preparation of eukaryotic cells for cryoET is more challenging, since most eukaryotic cells exceed the sample thickness limitation (1000 nm thickness for 300 keV). Hence, to directly visualize unaltered eukaryotic cells, we are restricted to sufficiently flat cells or to cell peripheries. Dealing with thicker specimens, such as concentrated cell suspensions or tissue samples, requires high-pressure freezing (Moor, 1987), a technique that enables the vitrification of an about 200 mm thick biological sample through the use of a 2000 bar jet of liquid nitrogen (Studer et al., 2008). The frozen-hydrated specimen is then cryosectioned into 50–200 nm thick sections, by cryoultramicrotomy (Al-Amoudi et al., 2004; Hsieh et al., 2006; McDowall et al., 1983), or thinned by an FIB (Marko et al., 2007). FIB thinning usually utilizes a gallium ion beam for gentle removal of a few nanometers of material, while keeping the specimen in the vitrified state, thereby minimizing artifacts, such as roughness, knife marks, or deformation attributed to cryosectioning. Cryoplaning can also facilitate thinning of large areas of vitreous ice prior to cryofluorescence, FIB thinning, and cryo-ET. Through the use of a customized 35 diamond knife with a clearance angle of 6 , planing the ice surface can be carried out using a nominal “feed” (increment) of 6 nm and a “block” speed of 5 mm/s at 150 C. An antistatic device was reported to assist in clearing debris from the knife edge. Thus, cryoplaning was introduced as a new concept for mechanical prethinning samples for cryo-ET (Rigort et al., 2010). The first application of cryo-ET for the study of an intact eukaryotic cell was documented by Medalia et al. (2002). In this study, the authors

Cryoelectron Tomography of Eukaryotic Cells

249

addressed the soil-living amoeba, Dictyostelium discoideum, resolving macromolecular complexes, namely ribosomes and 26S proteasomes, as well as actin filaments. These cells remained viable and were able to adhere and spread on the electron microscope grid, a prerequisite for studying intact eukaryotic cells. However, it is noteworthy that D. discoideum are fast motile cells that spread and adhere to surfaces within minutes. The unique eukaryotic cell, O. tauri, can also be studied as a whole. This unicellular green alga does not need to be attached to the EM grid for enabling the application of cryo-ET to resolve its tightly packed organelles and ultrastructure (Henderson et al., 2007). Figure 12.1 shows a cross-section and 3D segmentation of a cell harvested at the dark-to-light transition. Since its organelles were not visibly dividing, this example was classified as “nondividing” cell. So, while most eukaryotic cells are too thick to be captured in the intact state, their interactions with the extracellular matrix (ECM) and the external surfaces of their peripheral regions, such as adhesion sites, lamelleapodia, and other membrane protrusions, can be studied. For example, the actin network arrangement of intact D. discoideum filopodia was described in detail (Medalia et al., 2007), as were luminal particles within mammalian cellular microtubules (MT; Garvalov et al., 2006). A beautiful example of the use of cryo-ET to structurally characterize a eukaryotic structure came from the work of Nicastro et al. (2005, 2006). Their reconstruction of a specific molecular complex, namely the MT-based scaffold structure of the flagella and axonemal dynein of Chlamydomonas and sea urchin sperm, exemplify the possibility for studying native eukaryotic structure at medium resolution. The eukaryotic flagellum is thin enough to allow for structural analysis addressing the molecular mechanism underlying flagellar beating. Since this organelle is highly complex and its motor protein, dynein, is large enough for tomographic reconstruction, this study showed the power of studying a given process at a close-to-life state. Figure 12.2A shows a transmission electron micrograph of an intact quiescent, frozenhydrated sea urchin sperm flagellum. Due to the superposition of elements along the direction of the beam, fine details are not attainable However, after reconstruction, a detailed 3D view was achieved. The structure of the dynein motor, that is, the motor that moves each doublet MT relative to its neighbor, was identified and shown to include links connecting dynein arms with one another along the axoneme, as well as structures that connect inner and outer dynein arms (Fig. 12.2B). Comparing axoneme from wild-type Chlamydomonas and from the pf9-3 mutant (Myster et al., 1997) which fails to assemble the I1 inner arm complex, a two-headed dynein isoform composed of two dynein heavy chains (1 alpha and 1 beta) and three intermediate chains, revealed differences in structure (Fig. 12.2C). Adherent eukaryotic cells can be cultured on EM grids. Therefore, growth conditions for each cell line should be adjusted to allow optimal

250

Asaf Mader et al.

A

100 nm

B

40 nm

C

b a

IL

DRC

Figure 12.2 Cryo-ET of sperm flagellum from sea urchin and Chlamydomonas. (A) Transmission electron micrograph of an intact quiescent, frozen-hydrated sea urchin sperm flagellum. (B) A cross-section through a surface-rendered view of an axoneme. The structures were translationally averaged using 96-nm axial periodicity. Shown are the plasma membrane (brown), microtubules (gray), outer dynein arms (ODAs) and inner dynein arms (IDAs; red), radial spokes (orange), central pair protrusions (yellow), and bipartite bridges (pink). Scale bar: 40 nm. (C) Volume rendering presents the organization of the wild-type Chlamydomonas dynein arm viewed from the B tubule of the adjacent doublet, with the proximal end on the left. The 1-alpha and 1beta motor domains of the I1complex, the IL which is the DIC/DLC-tail complex (dynein intermediate chains /dynein light chains) and the dynein regulatory complex (DRC) that regulates phosphatase and kinase activities on the doublet MTs are indicated. Adapted from Nicastro et al. (2005, 2006).

cell growth and maximal attachment. Parameters which are often adjusted include grid geometry, continuous carbon versus holey grids, and protein coating of the carbon-coated grids. Carbon-coated gold and platinum EM grids are typically favored over the widely used copper grids, due to the toxicity of copper ions to cells. To circumvent the toxic effect of copper ions, D. discoideum cells were

Cryoelectron Tomography of Eukaryotic Cells

251

cultured on copper grids coated with carbon on both sides (Medalia et al., 2002), such that the surface of interaction between the buffer and the metal was minimized. Carbon-coated EM grids are the most commonly used support for EM studies. Traditionally, cryo-EM is conducted using perforated carbon films, as the microscopy is performed on macromolecular complexes which are not absorbed to the support. However, cells can only spread when adhering to a support. Although a perforated support allows for removal of excessive solution without direct contact with cells, not all eukaryotes adhere to these EM grids. An alternative support which may be of some advantage is silicon membrane-coated grids. These commercially available EM grids (QUANTIFOIL, Jena) present different surface properties that allow certain cells to spread better, while allowing cryo-ET to be performed. The grid surface can be coated with proteins or other polymers by modifying the general protocol used for EM grid coating (Grassucci et al., 2007). Since most eukaryotic cells interact with ECM proteins, coating the EM grids with such proteins prior to cell seeding can often assist the cells to better spread on the coated grid. Such coating of grids can be achieved by incubating glow-discharged grids on a drop of an ECM protein solution before placing the grids into a culture dish. For example, overlying the grids on 50 mg/ml fibronectin (Calbiochem, Darmstadt, Germany) for 45 min with the carbon side facing toward the solution. Cells cultured on grids should be carefully examined to permit normal adherence and spreading to achieve an appearance indistinguishable from cells grown in a conventional tissue culture dish. The density of cells should be kept at the subconfluence level to achieve optimal results, since high cell density will eventually lead to thicker cellular structure that cannot be studied by cryo-ET. Consequently, we suggest the following steps for successful investigation: 1. Place the grids inside a culture dish containing growth medium. The grid carbon film should face up. 2. Seed the culture gently, by pipetting the adherent cells into the culture dish. 3. Leave the plate for 10–15 min to allow for cell attachment. Monitor cell density before incubating at 37 C. 4. Incubate the grid in an incubator (37 C, 5% CO2) until the cells spread. Finally, for cryo-ET, fiducial markers are often used to align projection images. Therefore, colloidal gold (10–15 nm in diameter) is routinely added before cell vitrification (Dubochet et al., 1982). However, fiducial markers should be resuspended in physiological buffer when studying eukaryotic cells. For this purpose, the colloidal gold should be synthesized and protected by physisorption of a protein, such as BSA, as previously described (Slot and Geuze, 1985).

252

Asaf Mader et al.

3. Relying on Correlative Light and Electron Microscope for Cellular Structural Study of Eukaryotic Cells The immense efforts devoted to uncovering protein–protein interactions and characterizing the interactome (Rual et al., 2005) has also aimed at positioning macromolecular complexes within their native cellular environment. Approaches such as visual proteomics enable characterizing macromolecular complexes within their native cellular environment in situ, allowing for a full view of functional protein interactions (Beck et al., 2010; Nickell et al., 2006). Cryo-ET provides the means to visualize single macromolecular assemblies in their native cellular context. Still, identifying specific macromolecules in the cellular context or in a cellular process remains challenging. The limited field of view and lack of genetically expressed markers associated with EM hampers the unambiguous identification of cellular process under cryoconditions. Correlating fluorescent light with cryo-EM thus offers a route to circumvent this obstacle, in some cases. Fluorescence microscopy has revolutionized cell biology (Yuste, 2005). With the discovery of the green fluorescent protein (GFP) and its derivates, specific organelles, cellular events, and macromolecular complexes can be tracked by live cell microscopy (Chalfie et al., 1994; Heim et al., 1995; Shimomura et al., 1962). Integrating cryo-ET with fluorescent-light microscopy can thus provide invaluable information relating the functional state of cellular processes and the location of specific organelles. Fluorescence microscopy was previously correlated with EM studies using detergent-extracted, chemically fixed cells (Nemethova et al., 2008). The design of a cryochamber for light and fluorescent microscopy (Sartori et al., 2007; Schwartz et al., 2007; van Driel et al., 2009) now allows one to scan vitrified samples under the fluorescent microscope and use the same sample for cryo-ET. Cells, grown on grids, can be vitally labeled with one or more fluorescent dyes, such as fluorescent protein, vitrified, and directly examined. When considering a long process, the nonfixed biological specimen can undergo live fluorescent inspection and only then be vitrified for cryo-ET. The correlative approach combining light (i.e., fluorescence) and EM (i.e., cryo-ET) is illustrated in Fig. 12.3. Cells cultured on an EM grid were mapped under the fluorescent microscope, with regions of interest, for example, focal adhesion sites, identified, and imaged prior to vitrification of the EM grid, without the use of fixative (Fig. 12.3A). Next, the exact region of interest was located by cryo-EM and a tilt series was collected and reconstructed (Fig. 12.3B). A technical difficulty in CLEM is to pinpoint the exact fluorescent spot observed in the fluorescent microscope in the electron microscope under

253

Cryoelectron Tomography of Eukaryotic Cells

A

B

Figure 12.3 Correlative fluorescent and cryoelectron microscopy. (A) Cells expressing GFP-paxilin were examined under a fluorescent microscope. One cell is shown with the fluorescent region under scrutiny. (B) The cells were vitrified and cryo-ET was conducted on the protrusion shown in (A). The mitochondrion is clearly seen (black arrow). The actin cytoskeleton, as well as macromolecular complexes, can also be observed in the tomographic section, 10 nm in thickness. Scale bars: (A) 5 mm; (B) 200 nm.

cryoconditions. Grids containing markers, that is finder grids, are most suitable for this purpose. A major development came with the introduction of holders that can convert the coordinates from fluorescent microscope to cryo-EM, allowing easy localization of the region of interest in a fully automatic, computerized fashion (Lucic et al., 2007; Sartori et al., 2007), thereby minimizing the time needed to localize a given cellular structure under the electron beam. A different strategy utilizes noninvasive genetic manipulations to generate covalent fusions with a protein, such as metallothionein (i.e., a protein that can bind gold atoms) designed to serve as an electron-dense GFP analogue for EM (Mercogliano and DeRosier, 2006). In this report, aurothiomalate, an antiarthritic gold compound, reacted with metallothionein in vitro to form metallothionein–gold complexes, containing 17 gold atoms. This is a promising approach but needs to be further developed so as to attach larger quantities of gold atom to allow for an unambiguous visualization of heavy metal clusters in situ. Another promising approach uses quantum dots (QD), namely inorganic nanocrystals that fluoresce at distinct wavelengths, depending on their size (2–12 nm) and shape (Liu et al., 2005). The core, typically a CdSe or CdTe crystal, is electron-dense, enabling discrimination of the distinct QDs at the EM level (Giepmans et al., 2005). Since QDs possess high fluorescence, they allow single molecule detection. Moreover, the electron-dense crystal is directly visible by EM, and because of the different sizes and shapes, three different QDs can potentially be simultaneously used to label distinct proteins at the EM level (Deerinck et al., 2007; Nisman et al., 2004). QDs can be conjugated to streptavidin-tagged antibodies target to specific biotinylated-receptor proteins (Howarth et al., 2008). This approach

254

Asaf Mader et al.

can potentially be applied for cryo-CLEM, allowing for the visualization of specific cellular processes. A variety of techniques have been used to incorporate QDs into cells and label cellular proteins, including mechanical delivery, receptor-mediated internalization, chemical transfection, and passive uptake. Conjugation of cell-penetrating peptides to QDs is in progress (Akerman et al., 2002; Xue et al., 2007), although some successful studies have already been reported (Hoshino et al., 2004). Still, QDs exhibit relative low contrast in the electron microscope and, therefore, are not easily identified. Although the optical resolution of an ordinary fluorescent microscope is about 200 nm, new light imaging modalities recently introduced new super spatial resolution capabilities. These include photoactivated localization microscopy (PALM; Betzig et al., 2006), stimulated emission depletion microscopy (STED; Hell, 2003), stochastic optical reconstruction microscopy (STORM; Rust et al., 2006), and structured illumination (Schermelleh et al., 2008), with >40 nm resolution (as achieved with 3DSI, STED, PALM, STORM) providing the most suitable platform for CLEM. However, applying these methods without the use of chemical fixation is challenging and still needs additional development. Since most of the techniques that can break Abbe’s law of resolution rely on multiple image collection of the sample and switching fluorophores between two states, the sample is typically fixed to achieve maximal resolution.

4. Cryoelectron Tomography of CytoskeletonDriven Processes Eukaryotic cells move by an intricate mechanism of extension and retraction of their filamentous actin (F-actin) cytoskeleton network. During locomotion, cells extend thin dynamic protrusions at their leading edge, such as lamellipodia and filopodia. The actin cytoskeleton in eukaryotic cell is also the infrastructure for many fundamental processes and a major constituent of the cell shape regulation, adhesion, division, and motility machineries. EM of actin cytoskeletons was traditionally performed on sectioned chemically fixed, detergent-extracted (Brown et al., 1976; Svitkina et al., 1997) or ventral membranes of cells (Heuser and Kirschner, 1980). While analysis of such preparations provided insight into the architecture of actin networks (Small et al., 1994), the spatial resolution of the structures revealed was limited. For instance, interactions between filamentous cytoplasmic structures were difficult to resolve after metal decoration or replica formation. Moreover, artificial alteration of the cytoskeleton by detergent treatment prohibited the visualization of F-actin–plasma membrane interactions. In any case, bundling and cross-linking proteins, elements crucial for

255

Cryoelectron Tomography of Eukaryotic Cells

maintaining actin network architecture, may dissociate from the filaments and redistribute within a specimen during the course of traditional EM preparation procedures. EM of vitrified yet unaltered cells has proven to be a key technique for the 3D reconstruction of actin architecture (Medalia et al., 2002, 2007). As shown in Fig. 12.4, cryo-ET of intact D. discoideum cells reveals the intricate actin network, membranes, and cytoplasmatic macromolecular complexes (Fig. 12.4A) without the distortion or artifacts associated with detergent treatment. Using cryo-ET also permits viewing and study of the unaltered A

B

C

D

E

Figure 12.4 Visualizing the cytoplasm of a Dictyostelium discoideum cell. (A) Surfacerendered view of a volume of 815 nm 870 nm 97 nm. Colors were subjectively attributed to linear elements to mark the actin filaments (reddish); other macromolecular complexes, mostly ribosomes (green), and membranes (blue) are shown. (B, D) Actin anchorage to the membrane is illustrated in a surface-rendering of a cortical region demonstrating the plasma membrane associated with actin filaments ((B) is a rendered volume of 500 nm 300 nm 20 nm). The high magnification in (D) shows a kink-like structure close to the filament membrane-associating site, creating a filament-membrane bridge at a nonperpendicular angle. (C, E) Actin filament bundles in the cell cortex, and rendered volumes showing actin lateral connections to the membrane (C) or lateral bridges between two filaments (E), (extracted from (C)). Adapted from Medalia et al. (2002).

256

Asaf Mader et al.

anchorage of the actin filament network with the plasma membrane (Fig. 12.4B and 4D), as well as massive actin filament bundling at the cell cortex (Fig. 12.4C and E). Thus, cryo-ET is becoming a major structural tool to study cytoskeleton network architecture (Urban et al., 2010), comprising actin and MTs, using the identical experimental setups routinely used in the field of cell biology, namely intact cells.

5. Cryotomography of Midbodies Another organelle we have recently characterized by cryo-ET is the midbody, a tightly packed MT-based bridge which transiently connects the two daughter cells at the end of cytokinesis. Assembled from condensed, central spindle-MTs and numerous associated proteins, the midbody gradually narrows until daughter cell partitioning occurs at this site (Barr and Gruneberg, 2007; Eggert et al., 2006; Glotzer, 2005). Unlike the flagellum, the midbody features several hundreds of MTs arranged in parallel when assembled, together with several tens of MTs at it narrowest point, where abscission occurs. Furthermore, the midbody is extremely populated and highly electron-dense, properties which make it a challenging sample for cryo-ET, in particular, when one attempts to track individual MTs and detect macromolecular complexes. Fluorescence microscopy and proteomic studies localize an ever-growing number of proteins to the midbody, including MT-associated and actinassociated proteins, as well as membrane trafficking and regulatory proteins (Echard et al., 2004; Otegui et al., 2005; Skop et al., 2004; Steigemann and Gerlich, 2009). Some of these show distinctive sublocalizations when tagged fluorescently, for example PRC1, the MT-bundling protein that localizes to the overlap region (stem body; Mollinari et al., 2002), the MTsevering protein, spastin, that is localized to two narrow segments on either side of the overlap region (Connell et al., 2009), and centriolin, which is involved in membrane vesicle targeting and fusion and localized to a ring structure around the midbody (Gromley et al., 2005). Nevertheless, the detailed macromolecular composition and MT organization that occur during the abscission process cannot be determined by current light microscopy techniques due to the compactness of the structure. Such highresolution details require EM. Midbody morphology was viewed early on by EM using sectioning and heavy metal staining (Byers and Abramson, 1968; McDonald et al., 1979; McIntosh et al., 1975; Mullins and Biesele, 1977; Paweletz, 1967). These studies and others (e.g., Euteneuer and McIntosh, 1980; McDonald et al., 1979; Tippit et al., 1980) determined that interdigitating antiparallel MTs, ending at the center of the telophase spindle (termed “polar MTs”), create the

Cryoelectron Tomography of Eukaryotic Cells

257

thicker overlap region. Tracking midbody MTs through serial sections confirmed this observation, and revealed that the midbody featured an additional population of MTs which cross the overlap region (termed “continuous” or “free” MTs) (Mastronarde et al., 1993; McIntosh et al., 1975). Cryo-ET of midbodies enabled in-depth 3D analysis, providing a novel view of MT organization within this organelle and insight into its components. Since the investigated cells are typically round during mitosis, their midbody was obscured by thick ice or by the flanking daughter cells. Cryo-ET was, therefore, not unfeasible for intact midbodies, and midbody isolation was carried out according to established protocols (Mullins and McIntosh, 1982; Sellitto and Kuriyama, 1988). Isolated midbodies were easily identified under the electron beam and were suitable for cryo-ET, although their dense overlap region and its thickness prevented the gaining of detailed information in most tomograms. Nevertheless, tracking of MTs was possible in several tomograms and provided interesting results. At late telophase, the midbody was shown to be dominated by a core-bundle of MTs that transverse the electron-dense region, with their plus ends found outside the overlap region. The polar MTs that terminate in the overlap region surround this continuous bundle in an outer shell. A marked change in architecture was, however, observed in the outer shell of late-stage midbodies, where the polar MTs lost their interdigitation and retracted from the overlap region. These observations suggest that the midbody, having acquired a distinct MT architecture, as compared to the preceding central spindle, actively facilitates the final stage of cytokinesis. The spindle network, initially designed for separating the duplicated chromosome and later, for pushing the poles of the mitotic cell apart, is apparently redesigned upon formation of the midbody and adapts to its new role in daughter cell partitioning. This example indicates that cytoskeleton elements can be detected even in an extremely dense environment and when sample is as thick as a micron. In some cases, as shown above, the resolution of the 3D map is sufficient to permit biological meaningful experiments and conclusions. Such detailed MT architecture could only be investigated when the whole mount midbodies are considered in their hydrated state, emphasizing the ability of cryo-EM to unravel detailed structural organization.

6. Structural Analysis of the Nuclear Pore Complex by Cryo-ET The nuclear envelope (NE) is perforated by NPCs, which fuse the outer nuclear membrane (ONM) and the inner nuclear membrane (INM) to form aqueous translocation channels. These multiprotein assemblies allow passive diffusion of small molecules, as well as receptor-dependent,

258

Asaf Mader et al.

mediated translocation of large proteins and ribonucleoproteins. Macromolecular cargo usually harbor specific nuclear localization signals (NLSs) or nuclear export signals (NESs) that are recognized by transport receptors, mediating cargo passage through the NPC. Such receptors, referred to as karyopherins, chaperone cargo during transport across the NPC by means of hydrophobic interactions with phenylalanine-glycine-rich nucleoporin repeat domains (FG-repeats; Stewart, 2007; Suntharalingam and Wente, 2003). NPCs are composed of 30 different proteins termed nucleoporins (Nups) that are arranged as multimers containing multiple copies (Cronshaw et al., 2002; Rout et al., 2000; Terry et al., 2007). As such, the NPC is one of the largest molecular machines in the cell, exhibiting a molecular weight of over 100 MDa in vertebrates (Beck and Medalia, 2008; Lim et al., 2008). The structural organization of the NPC is largely conserved from yeast to man, comprising a pseudoeightfold symmetric central framework termed the spoke complex, a central pore of about 50 nm diameter, and filamentous structures on the cytoplasmic and nuclear sides of the complex (Elad et al., 2009; Fig. 12.5). At its nuclear face, the NPC is found in close interaction with the

A

240 nm B

Figure 12.5 Structural analysis of the nuclear pore complex by Cryo-ET. (A) A 50-nm thick tomographic slice through an intact nucleus of Dictyostelium discoideum. The D. discoideum NPC was observed in cryotomograms of intact nuclei. (B) A reconstruction of the NPC from native spread NE of Xenopus oocytes. Luminal and cytoplasmic faces of the NPCs are shown in the upper and lower panels, respectively. Adapted from Beck et al. (2007), Elad et al. (2009), and Frenkiel-Krispin et al. (2010).

Cryoelectron Tomography of Eukaryotic Cells

259

nuclear lamina, a meshwork of lamin filamentous structures (Ben-Harush et al., 2009), and other associated proteins. Structural examination of NPCs began several decades ago (Callan and Tomlin, 1950). However, the development of cryo-EM (Adrian et al., 1984) has permitted detailed and artifact-free analysis of NPCs embedded in physiological buffers (Akey and Radermacher, 1993; Yang et al., 1998). A landmark achievement in the field came with the application of cryo-ET to the intact NE and nuclei, enabling enhanced structural preservation with minimal purification steps (Beck et al., 2004; Stoffler et al., 2003). The nucleus had been considered as a dense and a thick specimen, beyond the limits of cryo-ET analysis. However, as will be discussed, cryo-ET of intact nuclei revealed the structure of the NPC, one of the largest macromolecular complexes in cells, in unprecedented detail. While structural analysis of active NPCs has obvious advantages, the integrity of the NE is crucial for retaining pore activity (Gorlich and Kutay, 1999). Additionally, biochemical procedures which are traditionally used for the purification of NPCs can lead to a loss of components of the transport machinery and the cargo. Nuclei can be readily prepared from D. discoideum by filtering the cells through a polycarbonate filter. These nuclei retain their nuclear transport activity and can be instantly vitrified on EM grids (Beck et al., 2004). Nuclei from this lower eukaryote are small ( 2 mm) and are connected to the bulk of the ER by thin tubular connections (Muller-Taubenberger et al., 2001). Figure 12.5A shows a slice through the reconstructed volume of such a nucleus, with discontinuity of the NE being reflected in a side view of NPC. Using a 3D alignment algorithm (Forster et al., 2005), the structure of the D. discoideum NPC was described without imposing eightfold symmetry. Instead, the structure was analyzed as eight protomers in silico (Beck et al., 2007). Such analysis resulted in an improvement in resolution of the structure of the central framework to about 6 nm. Due to its inherent flexibility, such analysis did not, however, consider the nuclear basket. The estimated orientation of the NPC was estimated manually, since it is embedded on an ellipsoid surface, the nuclear membrane, reducing the need for large angular scan during the averaging procedure. Luminal and cytoplasmic aspects of the D. discoideum NPC are presented in Fig. 12.5B. The central framework is 60 nm in height and has an outer diameter of 120 nm. The diameter of the main channel is about 50 nm. The structure of the metazoan NPC, as reconstructed from spread NEs of Xenopus oocytes, resembles that of D. discoideum. In both cases, the outer diameter of the structure and the opening of the main channel are very similar in dimension, namely 120 and 50 nm, respectively, although the Xenopus NPC seems to be taller (100 nm, Fig. 12.5C; Stoffler et al., 2003). Moreover, the distribution of densities appears to vary between the two species.

260

Asaf Mader et al.

7. Concluding Remarks Cryo-ET of eukaryotic cells has the potential to become a major approach in answering outstanding questions in cellular structural biology. The ability to structurally characterize systems which were routinely studied by means of cell biology techniques, for example, fluorescently based microscopy of wild-type and mutant cell lines, now using structural tools is intriguing. However, a major challenge still facing this technique concerns identifying macromolecular complexes. The variety of orientations macromolecules can assume and the current resolution of cellular tomograms, that is 4–6 nm, prohibit unbiased identification of many complexes, although some successful template-matching approaches have been introduced (Frangakis et al., 2002; Seybert et al., 2006). An alternative approach, based on electron-dense labeling, should be developed, focused on the design of a GFP analogue for cryo-EM. Such a strategy would provide a general solution for identifying complexes whose structure is not yet determined or which transiently interact to form large assemblies in cells. In conclusion, to make cryo-ET applicable not only to cellular protrusions and thin regions of the eukaryotic cell but also to allow a structural view of organisms, development of a reliable freeze-hydrated, artifact-free sectioning technique that can be applied to tissues and cells is called for. Alternative microdissection techniques based on ion beam milling are currently being developed (Marko et al., 2007; Rigort et al., 2010). Realization of this technology would open a window for the entry of cryo-ET into other branches of biology and provide, for instance, a bridge between structural and developmental biology. Thus, we are currently at the first stages of applying cryo-ET to eukaryotic cells, entering biological territories which are not yet charted.

ACKNOWLEDGMENTS This study was supported grants from the Israeli Science Foundation and the German–Israeli Cooperation Project (DIP) (H.2.2) and by an ERC Starting Grant to O. M.

REFERENCES Adrian, M., et al. (1984). Cryo-electron microscopy of viruses. Nature 308, 32–36. Akerman, M. E., et al. (2002). Nanocrystal targeting in vivo. Proc. Natl. Acad. Sci. USA 99, 12617–12621. Akey, C. W., and Radermacher, M. (1993). Architecture of the Xenopus nuclear pore complex revealed by three-dimensional cryo-electron microscopy. J. Cell Biol. 122, 1–19.

Cryoelectron Tomography of Eukaryotic Cells

261

Al-Amoudi, A., et al. (2004). Cryo-electron microscopy of vitreous sections. EMBO J. 23, 3583–3588. Barr, F. A., and Gruneberg, U. (2007). Cytokinesis: Placing and making the final cut. Cell 131, 847–860. Beck, M., and Medalia, O. (2008). Structural and functional insights into nucleocytoplasmic transport. Histol. Histopathol. 23, 1025–1033. Beck, M., et al. (2004). Nuclear pore complex structure and dynamics revealed by cryoelectron tomography. Science 306, 1387–1390. Beck, M., et al. (2007). Snapshots of nuclear pore complexes in action captured by cryoelectron tomography. Nature 449, 611–615. Beck, M., et al. (2010). Visual proteomics in Leptospira interrogans. Methods Enzymol. 1. CryoEM. Part 2 Lessons learned from specific examples chapter. Ben-Harush, K., et al. (2009). The supramolecular organization of the C. elegans nuclear lamin filament. J. Mol. Biol. 386, 1392–1402. Ben-Harush, K., et al. (2010). Visualizing cellular processes at the molecular level by cryoelectron tomography. J. Cell Sci. 123, 7–12. Betzig, E., et al. (2006). Imaging intracellular fluorescent proteins at nanometer resolution. Science 313, 1642–1645. Brown, S., et al. (1976). Cytoskeletal elements of chick embryo fibroblasts revealed by detergent extraction. J. Supramol. Struct. 5, 119–130. Byers, B., and Abramson, D. H. (1968). Cytokinesis in HeLa: Post-telophase delay and microtubule-associated motility. Protoplasma 66, 413–435. Callan, H. G., and Tomlin, S. G. (1950). Experimental studies on amphibian oocyte nuclei. 1. Investigation of the structure of the nuclear membrane by means of the electron microscope. Proc. R. Soc. Lond. B Biol. Sci. 137, 367–378. Campbell, I. D. (2008). The Croonian lecture 2006. Structure of the living cell. Philos. Trans. R. Soc. Lond. B Biol. Sci. 363, 2379–2391. Chalfie, M., et al. (1994). Green fluorescent protein as a marker for gene expression. Science 263, 802–805. Chen, S., and Briegel, A. (2010). Correlative cryo-light and -electron microscopy. Methods Enzymol. 1. Cryo-EM. Part 1 Data collection chapter. Connell, J. W., et al. (2009). Spastin couples microtubule severing to membrane traffic in completion of cytokinesis and secretion. Traffic 10, 42–56. Cronshaw, J. M., et al. (2002). Proteomic analysis of the mammalian nuclear pore complex. J. Cell Biol. 158, 915–927. Deerinck, T. J., et al. (2007). Light and electron microscopic localization of multiple proteins using quantum dots. Methods Mol. Biol. 374, 43–53. Dierksen, K., et al. (1992). Towards automatic electron tomography. Ultramicroscopy 40, 71–87. Dierksen, K., et al. (1993). Towards automatic electron tomography. 2. Implementation of autofocus and low-dose procedures. Ultramicroscopy 49, 109–120. Dubochet, J., et al. (1982). Electron-microscopy of frozen water and aqueous-solutions. J. Microsc. Oxf. 128, 219–237. Echard, A., et al. (2004). Terminal cytokinesis events uncovered after an RNAi screen. Curr. Biol. 14, 1685–1693. Eggert, U. S., et al. (2006). Animal cytokinesis: From parts list to mechanisms. Annu. Rev. Biochem. 75, 543–566. Elad, N., et al. (2009). Structural analysis of the nuclear pore complex by integrated approaches. Curr. Opin. Struct. Biol. 19, 226–232. Euteneuer, U., and McIntosh, J. R. (1980). Polarity of midbody and phragmoplast microtubules. J. Cell Biol. 87, 509–515.

262

Asaf Mader et al.

Forster, F., et al. (2005). Retrovirus envelope protein complex structure in situ studied by cryo-electron tomography. Proc. Natl. Acad. Sci. USA 102, 4729–4734. Frangakis, A. S., et al. (2002). Identification of macromolecular complexes in cryoelectron tomograms of phantom cells. Proc. Natl. Acad. Sci. USA 99, 14153–14158. Frenkiel-Krispin, D., et al. (2010). Structural analysis of a metazoan nuclear pore complex reveals a fused concentric ring architecture. J. Mol. Biol. 395, 578–586. Gan, L., et al. (2008). Molecular organization of Gram-negative peptidoglycan. Proc. Natl. Acad. Sci. USA 105, 18953–18957. Garvalov, B. K., et al. (2006). Luminal particles within cellular microtubules. J. Cell Biol. 174, 759–765. Giepmans, B. N., et al. (2005). Correlated light and electron microscopic imaging of multiple endogenous proteins using Quantum dots. Nat. Methods 2, 743–749. Glotzer, M. (2005). The molecular requirements for cytokinesis. Science 307, 1735–1739. Gorlich, D., and Kutay, U. (1999). Transport between the cell nucleus and the cytoplasm. Annu. Rev. Cell Dev. Biol. 15, 607–660. Grassucci, R. A., et al. (2007). Preparation of macromolecular complexes for cryo-electron microscopy. Nat. Protoc. 2, 3239–3246. Grimm, R., et al. (1996). Determination of the inelastic mean free path in ice by examination of tilted vesicles and automated most probable loss imaging. Ultramicroscopy 63, 169–179. Gromley, A., et al. (2005). Centriolin anchoring of exocyst and SNARE complexes at the midbody is required for secretory-vesicle-mediated abscission. Cell 123, 75–87. Heim, R., et al. (1995). Improved green fluorescence. Nature 373, 663–664. Hell, S. W. (2003). Toward fluorescence nanoscopy. Nat. Biotechnol. 21, 1347–1355. Henderson, G. P., et al. (2007). 3-D ultrastructure of O. tauri: Electron cryotomography of an entire eukaryotic cell. PLoS ONE 2, e749. Heuser, J. E., and Kirschner, M. W. (1980). Filament organization revealed in platinum replicas of freeze-dried cytoskeletons. J. Cell Biol. 86, 212–234. Hoshino, A., et al. (2004). Quantum dots targeted to the assigned organelle in living cells. Microbiol. Immunol. 48, 985–994. Howarth, M., et al. (2008). Monovalent, reduced-size quantum dots for imaging receptors on living cells. Nat. Methods 5, 397–399. Hsieh, C. E., et al. (2006). Towards high-resolution three-dimensional imaging of native mammalian tissue: Electron tomography of frozen-hydrated rat liver sections. J. Struct. Biol. 153, 1–13. Iancu, C. V., et al. (2010). Organization, structure, and assembly of alpha-carboxysomes determined by electron cryotomography of intact cells. J. Mol. Biol. 396, 105–117. Komeili, A., et al. (2006). Magnetosomes are cell membrane invaginations organized by the actin-like protein MamK. Science 311, 242–245. Koning, R. I., and Koster, A. J. (2009). Cryo-electron tomography in biology and medicine. Ann. Anat. 191, 427–445. Konorty, M., et al. (2008). Structural analysis of photosynthetic membranes by cryo-electron tomography of intact Rhodopseudomonas viridis cells. J. Struct. Biol. 161, 393–400. Ku¨rner, J., and Baumeister, W. (2006). Cryo-Electron Tomography Reveals the Architecture of a Bacterial Cytoskeleton, in Complex Intracellular Structures in Prokaryotes, Springer-Verlag, Berlin-Heidelberg, pp. 313–318. Kurner, J., et al. (2004). New insights into the structural organization of eukaryotic and prokaryotic cytoskeletons using cryo-electron tomography. Exp. Cell Res. 301, 38–42. Ladinski, M., et al. (2010). High-pressure freezing and cryosectioning thick samples. Methods Enzymol. 1. Cryo-EM. Part 1 Sample preparation chapter. Li, Z., and Jensen, G. J. (2009). Electron cryotomography: A new view into microbial ultrastructure. Curr. Opin. Microbiol. 12, 333–340.

Cryoelectron Tomography of Eukaryotic Cells

263

Lim, R. Y., et al. (2008). Towards reconciling structure and function in the nuclear pore complex. Histochem. Cell Biol. 129, 105–116. Liu, T., et al. (2005). The fluorescence bioassay platforms on quantum dots nanoparticles. J. Fluoresc. 15, 729–733. Lucic, V., et al. (2005). Structural studies by electron tomography: From cells to molecules. Annu. Rev. Biochem. 74, 833–865. Lucic, V., et al. (2007). Multiscale imaging of neurons grown in culture: From light microscopy to cryo-electron tomography. J. Struct. Biol. 160, 146–156. Lucic, V., et al. (2008). Cryo-electron tomography of cells: Connecting structure and function. Histochem. Cell Biol. 130, 185–196. Marko, M., et al. (2007). Focused-ion-beam thinning of frozen-hydrated biological specimens for cryo-electron microscopy. Nat. Methods 4, 215–217. Mastronarde, D. N., et al. (1993). Interpolar spindle microtubules in PTK cells. J. Cell Biol. 123, 1475–1489. McDonald, K. L., et al. (1979). Cross-sectional structure of the central mitotic spindle of Diatoma vulgare. Evidence for specific interactions between antiparallel microtubules. J. Cell Biol. 83, 443–461. McDowall, A. W., et al. (1983). Electron microscopy of frozen hydrated sections of vitreous ice and vitrified biological samples. J. Microsc. 131, 1–9. McDowall, A. W., et al. (2010). Plunge-freezing isolated macomolecular complexes, viruses, and small cells. Methods Enzymol. 1. Cryo-EM. Part 1 Sample preparation chapter. McIntosh, J. R., et al. (1975). Structure and physiology of the mammalian mitotic spindle. Soc. Gen. Physiol. Ser. 30, 31–76. Medalia, O., et al. (2002). Macromolecular architecture in eukaryotic cells visualized by cryoelectron tomography. Science 298, 1209–1213. Medalia, O., et al. (2007). Organization of actin networks in intact filopodia. Curr. Biol. 17, 79–84. Mercogliano, C. P., and DeRosier, D. J. (2006). Gold nanocluster formation using metallothionein: Mass spectrometry and electron microscopy. J. Mol. Biol. 355, 211–223. Milne, J. L., and Subramaniam, S. (2009). Cryo-electron tomography of bacteria: Progress, challenges and future prospects. Nat. Rev. Microbiol. 7, 666–675. Mollinari, C., et al. (2002). PRC1 is a microtubule binding and bundling protein essential to maintain the mitotic spindle midzone. J. Cell Biol. 157, 1175–1186. Moor, H. (1987). Theory and practice of high-pressure freezing. In “Cryotechniques in Biological Electron Microscopy,” (R. A. Steinbrecht and K. Zierold, eds.), pp. 175–191. Springer-Verlag, Berlin. Muller-Taubenberger, A., et al. (2001). Calreticulin and calnexin in the endoplasmic reticulum are important for phagocytosis. EMBO J. 20, 6772–6782. Mullins, J. M., and Biesele, J. J. (1977). Terminal phase of cytokinesis in D-98s cells. J. Cell Biol. 73, 672–684. Mullins, J. M., and McIntosh, J. R. (1982). Isolation and initial characterization of the mammalian midbody. J. Cell Biol. 94, 654–661. Myster, S. H., et al. (1997). The Chlamydomonas Dhc1 gene encodes a dynein heavy chain subunit required for assembly of the I1 inner arm complex. Mol. Biol. Cell 8, 607–620. Nemethova, M., et al. (2008). Building the actin cytoskeleton: Filopodia contribute to the construction of contractile bundles in the lamella. J. Cell Biol. 180, 1233–1244. Nicastro, D., et al. (2005). 3D structure of eukaryotic flagella in a quiescent state revealed by cryo-electron tomography. Proc. Natl. Acad. Sci. USA 102, 15889–15894. Nicastro, D., et al. (2006). The molecular architecture of axonemes revealed by cryoelectron tomography. Science 313, 944–948. Nickell, S., et al. (2006). A visual approach to proteomics. Nat. Rev. Mol. Cell Biol. 7, 225–230.

264

Asaf Mader et al.

Nisman, R., et al. (2004). Application of quantum dots as probes for correlative fluorescence, conventional, and energy-filtered transmission electron microscopy. J. Histochem. Cytochem. 52, 13–18. Otegui, M. S., et al. (2005). Midbodies and phragmoplasts: Analogous structures involved in cytokinesis. Trends Cell Biol. 15, 404–413. Paweletz, N. (1967). On the function of the “Flemming body” during division of animal cells. Naturwissenschaften 54, 533–535. Plitzko, J. M., et al. (2009). Correlative cryo-light microscopy and cryo-electron tomography: From cellular territories to molecular landscapes. Curr. Opin. Biotechnol. 20, 83–89. Rigort, A., et al. (2010). Micromachining tools and correlative approaches for cellular cryoelectron tomography. J. Struct. Biol. (In press) Robinson, J. M., et al. (2001). Correlative fluorescence and electron microscopy on ultrathin cryosections: Bridging the resolution gap. J. Histochem. Cytochem. 49, 803–808. Rout, M. P., et al. (2000). The yeast nuclear pore complex: Composition, architecture, and transport mechanism. J. Cell Biol. 148, 635–651. Rual, J. F., et al. (2005). Towards a proteome-scale map of the human protein–protein interaction network. Nature 437, 1173–1178. Rust, M. J., et al. (2006). Sub-diffraction-limit imaging by stochastic optical reconstruction microscopy (STORM). Nat. Methods 3, 793–795. Saibil, H. R. (2000). Conformational changes studied by cryo-electron microscopy. Nat. Struct. Biol. 7, 711–714. Sali, A., et al. (2003). From words to literature in structural proteomics. Nature 422, 216–225. Sartori, A., et al. (2007). Correlative microscopy: Bridging the gap between fluorescence light microscopy and cryo-electron tomography. J. Struct. Biol. 160, 135–145. Schermelleh, L., et al. (2008). Subdiffraction multicolor imaging of the nuclear periphery with 3D structured illumination microscopy. Science 320, 1332–1336. Schwartz, C. L., et al. (2007). Cryo-fluorescence microscopy facilitates correlations between light and cryo-electron microscopy and reduces the rate of photobleaching. J. Microsc. 227, 98–109. Sellitto, C., and Kuriyama, R. (1988). Distribution of a matrix component of the midbody during the cell cycle in Chinese hamster ovary cells. J. Cell Biol. 106, 431–439. Seybert, A., et al. (2006). Structural analysis of Mycoplasma pneumoniae by cryo-electron tomography. J. Struct. Biol. 156, 342–354. Shimomura, O., et al. (1962). Extraction, purification and properties of aequorin, a bioluminescent protein from the luminous hydromedusan, Aequorea. J. Cell. Comp. Physiol. 59, 223–239. Skop, A. R., et al. (2004). Dissection of the mammalian midbody proteome reveals conserved cytokinesis mechanisms. Science 305, 61–66. Slot, J. W., and Geuze, H. J. (1985). A new method of preparing gold probes for multiplelabeling cytochemistry. Eur. J. Cell Biol. 38, 87–93. Small, J. V., et al. (1994). Visualization of actin filaments in keratocyte lamellipodia: Negative staining compared with freeze-drying. J. Struct. Biol. 113, 135–141. Steigemann, P., and Gerlich, D. W. (2009). Cytokinetic abscission: Cellular dynamics at the midbody. Trends Cell Biol. 19, 606–616. Stewart, M. (2007). Molecular mechanism of the nuclear protein import cycle. Nat. Rev. Mol. Cell Biol. 8, 195–208. Stoffler, D., et al. (2003). Cryo-electron tomography provides novel insights into nuclear pore architecture: Implications for nucleocytoplasmic transport. J. Mol. Biol. 328, 119–130. Studer, D., et al. (2008). Electron microscopy of high pressure frozen samples: Bridging the gap between cellular ultrastructure and atomic resolution. Histochem. Cell Biol. 130, 877–889.

Cryoelectron Tomography of Eukaryotic Cells

265

Suntharalingam, M., and Wente, S. R. (2003). Peering through the pore: Nuclear pore complex structure, assembly, and function. Dev. Cell 4, 775–789. Svitkina, T. M., et al. (1997). Analysis of the actin-myosin II system in fish epidermal keratocytes: Mechanism of cell body translocation. J. Cell Biol. 139, 397–415. Terry, L. J., et al. (2007). Crossing the nuclear envelope: Hierarchical regulation of nucleocytoplasmic transport. Science 318, 1412–1416. Tippit, D. H., et al. (1980). Organization of spindle microtubules in Ochromonas danica. J. Cell Biol. 87, 531–545. Urban, E., et al. (2010). Electron tomography reveals unbranched networks of actin filaments in lamellipodia. Nat. Cell Biol. 12, 429–435. van Driel, L. F., et al. (2009). Tools for correlative cryo-fluorescence microscopy and cryoelectron tomography applied to whole mitochondria in human endothelial cells. Eur. J. Cell Biol. 88, 669–684. Vicidomini, G., et al. (2010). A novel approach for correlative light electron microscopy analysis. Microsc. Res. Tech. 73, 215–224. Weston, A. E., et al. (2010). Towards native-state imaging in biological context in the electron microscope. J. Chem. Biol. 3, 101–112. Xue, F. L., et al. (2007). Enhancement of intracellular delivery of CdTe quantum dots (QDs) to living cells by Tat conjugation. J. Fluoresc. 17, 149–154. Yang, Q., et al. (1998). Three-dimensional architecture of the isolated yeast nuclear pore complex: Functional and evolutionary implications. Mol. Cell 1, 223–234. Yuste, R. (2005). Fluorescence microscopy today. Nat. Methods 2, 902–904. Zhang, P., et al. (2007). Direct visualization of Escherichia coli chemotaxis receptor arrays using cryo-electron microscopy. Proc. Natl. Acad. Sci. USA 104, 3777–3781.

C H A P T E R

T H I R T E E N

3D Visualization of HIV Virions by Cryoelectron Tomography Jun Liu,* Elizabeth R. Wright,† and Hanspeter Winkler‡ Contents 1. Introduction 2. Cryoelectron Tomography 2.1. Sample preparation: Frozen-hydrated specimen 2.2. Dose series: Determining the optimal electron dose 2.3. Data acquisition: Low-dose tilt series 2.4. 3D tomographic reconstruction 2.5. Subvolume analysis 2.6. Alignment strategies 3. 3D Visualization of Intact HIV Virion 3.1. Cryoelectron tomography of HIV virions 3.2. Molecular structure of the gag shells 3.3. Molecular architecture of HIV Env 3.4. Molecular details of neutralizing antibody b12-Env complex 3.5. CD4 induces major conformational change of HIV Env 4. Conclusions and Perspective Acknowledgments References

268 270 270 271 272 273 274 276 278 278 278 279 282 285 286 286 286

Abstract The structure of the human immunodeficiency virus (HIV) and some of its components have been difficult to study in three-dimensions (3D) primarily because of their intrinsic structural variability. Recent advances in cryoelectron tomography (cryo-ET) have provided a new approach for determining the 3D structures of the intact virus, the HIV capsid, and the envelope glycoproteins located on the viral surface. A number of cryo-ET procedures related to specimen preservation, data collection, and image processing are presented in this

* Department of Pathology and Laboratory Medicine, University of Texas Medical School at Houston, Houston, Texas, USA Department of Pediatrics, Emory University School of Medicine, Atlanta, Georgia, USA { Institute of Molecular Biophysics, Florida State University, Tallahassee, Florida, USA {

Methods in Enzymology, Volume 483 ISSN 0076-6879, DOI: 10.1016/S0076-6879(10)83014-9

#

2010 Elsevier Inc. All rights reserved.

267

268

Jun Liu et al.

chapter. The techniques described herein are well suited for determining the ultrastructure of bacterial and viral pathogens and their associated molecular machines in situ at nanometer resolution.

1. Introduction Human immunodeficiency virus (HIV) is a spherical retrovirus with a diameter of 100–150 nm. The lipid-enveloped virus first buds from the host cell in an immature form (Freed, 1998). The major structural proteins are unprocessed at this point and are referred to as the Gag polyprotein. The proteins are arranged radially in the following order from the lipid bilayer: matrix (MA), capsid (CA), SP1, nucleocapsid (NC), SP2, and p6 (Fuller et al., 1997; Wilk et al., 2001). During the budding process, the viral protease (PR) becomes active and initiates cleavage of the major protein domains of the Gag polyprotein, which results in a major structural rearrangement known as “maturation.” It is during this process that the CA domains reorganize to form the bullet-shaped core or “capsid” that encompasses the viral genome of the mature virus. Throughout the entire process of budding and maturation, the virus is enclosed by a viral envelope consisting of a lipid bilayer in which the envelope glycoprotein (Env) is embedded. HIV Env binds CD4 receptors and coreceptors on target cells, thereby initiating viral entry and infection. Native HIV Env is a trimeric complex organized as three heterodimers. The heterodimers consist of a noncovalently associated extracellular subunit gp120 and a transmembrane subunit gp41. Although several crystal structures of the gp41 and the gp120 monomer are known (Chan et al., 1997; Chen et al., 2005, 2009; Huang et al., 2005; Kwong et al., 1998; Zhou et al., 2007), the crystal structure of the HIV Env trimer has not been determined. Electron microscopy (EM) of sectioned materials and negatively stained virions has contributed significantly to our understanding of the morphology and fine structure of HIV (Gelderblom et al., 1987). However, conventional EM methods have been unable to provide information about the 3D architecture of the virus. The advent of cryoelectron microscopy (cryoEM) made it possible for the structure of both the immature and mature forms of the virus to be readdressed. In several landmark papers, cryo-EM studies of the immature virus revealed the radial arrangement of the Gag polyprotein domains; determined that the CA domain forms a hexagonal lattice with a distinctive spacing of 8.0 nm; and defined the number of individual CA proteins required to compose the core of the mature virus (Briggs et al., 2004, 2006b; Fuller et al., 1997; Nermut et al., 1998). Concurrently, the mature virus was examined by cryo-EM and it was determined that virion size remains constant through the maturation process;

Cryoelectron Tomography of HIV

269

virions can have conical or altered shaped cores; virions can have multiple cores; subunits of the CA protein of the cores form a hexagonal lattice with a 9.6-nm spacing; and the assembly of the core is most-likely template driven (Briggs et al., 2003). The 2D cryo-EM structures of intact immature and mature HIV virions dramatically changed how researchers thought about HIV, its assembly, and maturation. Simultaneous to studies of the intact virus, research groups have sought to address questions related to the individual proteins as they rearrange during viral maturation. A significant effort has been placed on studies of the CA protein and the dramatic conformation rearrangement it undergoes during maturation. Studies of the individual proteins have been driven forward through the combined use of computational modeling, 2D cryo-EM, electron crystallography, NMR, and X-ray crystallography methods (Byeon et al., 2009; Douglas et al., 2004; Ganser et al., 1999, 2003; Ganser-Pornillos et al., 2007; Lanman et al., 2002, 2003; Li et al., 2000; Massiah et al., 1994, 1996; Pornillos et al., 2009). Researchers still endeavored to study the 3D structure of intact HIV virions. It was with this in mind that electron tomography was used to visualize the trilobed structure of Env spikes on negatively stained HIV virions (Zhu et al., 2003) and the architecture of the virus–cell contact region from chemically fixed and stained specimens (Sougrat et al., 2007). The leap of using electron tomography to examine an external structure of the intact HIV virions was remarkable; however, studies of the internal architecture of the virus was not possible because of conventional sample preservation artifacts. Soon after the electron tomography results of HIV Env spikes were published, microscope manufactures developed streamlined, computer controlled, automated microscopes. It was due to these significant technological advances that cryoelectron tomography (cryo-ET) became a powerful technology for studying the structure of intact virions and whole cells (Grunewald et al., 2003; Medalia et al., 2002). Through a number of recent investigations, cryo-ET has now been established as the ideal approach for studying the distinct architecture of HIV. Research groups have recently determined the 3D structure of both immature and mature virions (Benjamin et al., 2005; Briggs et al., 2006a, 2009; Wright et al., 2007). Cryo-ET examinations of mature HIV-1 virions revealed that the conical cores were unique in structure and position, but they also demonstrated certain similarities with respect to size and shape, the distance of the cone’s base from the envelope/MA layer, the range of the cone angle. It was also observed that the conical CA shape was preferred in vivo, which provided additional evidence to support the templatedirected model of CA formation (Benjamin et al., 2005; Briggs et al., 2006a). Structural studies of the immature virion soon followed and some of the methods and results are described in detail in the section below (Briggs et al., 2009; Wright et al., 2007). Concurrent to the cryo-ET studies

270

Jun Liu et al.

of HIV maturation, cryo-ET, combined with advanced computational methods, was used to determine the molecular architecture of native Env spikes in situ at nanometer resolution (Liu et al., 2008; Zanetti et al., 2006; Zhu et al., 2006). The studies of HIV Env spike structure, arrangement, and Env in-complex with neutralizing antibodies has provided a new driving force for the investigation and development of Env-specific therapeutics. In this chapter, we present a number of cryo-ET procedures related to specimen preservation, data collection, and image processing that have proved to be useful for establishing a better understanding of HIV viral assembly, viral entry, and antibody neutralization at the nanometer level.

2. Cryoelectron Tomography Cryo-ET is established as an important 3D imaging technique that serves to bridge the information gap between atomic structures and EM reconstructions that reveal the cellular or viral architectures (Baumeister et al., 1999; Lucic et al., 2005; Subramaniam, 2005). In combination with advanced computational methods, cryo-ET is the most promising approach to determine the molecular architecture of nanomachines in situ. The potential of cryo-ET lies in its ability to investigate cellular and viral components in their native state without fixation, dehydration, embedding, or sectioning artifacts. However, the poor signal-to-noise ratio (SNR) of the collected image data and the lack of effective protein labeling techniques are major impediments in cryo-ET, which confound researchers’ efforts to identify specific molecules reliably within a cell or pleomorphic virus. Our interest has gravitated around the improvement of cryo-ET methods for determining 3D structure of large macromolecular complexes in situ. The goal is to render molecular detail at higher resolution and to automate and expedite the reconstruction process and the subsequent data analysis. The high-throughput cryo-ET processing pipeline includes data acquisition, fast tilt series alignment, 3D reconstruction, subvolume extraction, and analysis. The combination of these techniques has enabled us to determine 3D structures of macromolecular assemblies at resolutions in the range of 2–4 nm (Liu et al., 2008, 2009), which permits an interpretation of structure/ function relationships at the nanometer level.

2.1. Sample preparation: Frozen-hydrated specimen The preparation of frozen-hydrated viral specimens is the key step for the direct visualization of virus particles by cryo-ET. The general procedure for preparing frozen-hydrated biological specimens is well described in detail (Dubochet et al., 1988; Chapter 3, Volume 481). For cryo-EM and

Cryoelectron Tomography of HIV

271

cryo-ET, the basic steps for making frozen-hydrated specimens include: EM grid preparation, sample preparation, and plunge freezing. 2.1.1. EM grid preparation The type of EM holey carbon grid used is an important component for the consistent, high-quality frozen-hydrated viral sample preparation. There are several grid options: lacey or holey grids, Quantifoil grids, and C-flat grids. Quantifoil and C-flat 200 mesh grids are preferred, because of the wellordered pattern of the holes in the carbon film and the 200 mesh size allows for maximal tilting of the grid in the microscope. It is essential for the formation of a thin layer of ice-suspended virions that the grids are cleaned and processed by glow discharging prior to plunge freezing in order to remove manufacturing by-products and increase the hydrophilic nature of the grid surface. 2.1.2. Sample preparation If the tilt series are to be aligned by the use of gold markers, there are two basic options for applying the 5 or 10 nm gold particles to the EM grid. (1) A solution of the colloidal gold particles is applied to the EM grid and allowed to air-dry. (2) A concentrated stock of the gold particles is mixed with the viral sample and the mixture is applied to the EM grid during the freezing process. The second procedure is commonly used for tomography applications. 2.1.3. Plunge freezing A virus suspension of 4 ml is placed on a freshly glow discharged holey carbon grid. The excess solution is removed manually with a piece of filter paper and the grid is then plunged into liquid ethane. Alternatively, the sample can also be frozen using a semiautomatic or automatic plunge freezing apparatus. There are a number of apparatuses available, including the FEI Vitrobot (FEI, Hillsboro, OR) and the Gatan CP3 (Gatan, Pleasanton, CA), and the newly introduced Leica EM GP (Leica Microsystems GmbH, Wetzlar, Germany). Both FEI and Gatan systems as well as a manual plunge freezing apparatus have been used to prepare grids of frozen-hydrated viral samples.

2.2. Dose series: Determining the optimal electron dose It has been shown that a statistically well-defined 3D reconstruction can be obtained from low-dose and noisy projection images, as long as the total dose is sufficient (McEwen et al., 1995). However, radiation damage is a fundamental problem for frozen-hydrated biological specimens (Chapter 15, Volume 481). Higher total dose and higher defocus increases contrast, but with the introduction of significant radiation damage and a reduction in

272

Jun Liu et al.

resolution. Therefore, it is important for each application to find the optimal tradeoff between dose, defocus, contrast, and resolution before ˚ 2) images collecting tilt series. For this purpose, twenty high-dose (10 e/A at 0 tilt were recorded (Fig. 13.1). Comparative analysis or visual inspection of a movie of aligned images provides a good estimation of beam damage and image distortion that occurred during dose series. The tilt series ˚ 2 for selected for processing were usually exposed to a total dose of 80 e/A 2 ˚ HIV Env studies and 120 e/A for HIV maturation analysis.

2.3. Data acquisition: Low-dose tilt series Because the time required for acquiring a tilt series and computing a tomographic reconstruction is significant, it is important to locate as many well-preserved virions as possible within a sample. The regions of interest that exhibit a sufficiently high virion concentration and appropriate vitreous ice thickness are typically selected at low magnification (4700) and high defocus (300 mm) under low-dose conditions. The low-dose “search” images have reasonable contrast to readily locate the area of interest. The use of the latest advances in automated image acquisition (Koster et al., 1992, 1997) is essential for data acquisition of frozen-hydrated biological specimens. Currently, there are several software packages available for automatic data acquisition: SerialEM (Mastronarde, 2005), UCSF Tomo (Zheng et al., 2004), TOM (Nickell et al., 2005), Xplore3D (FEI), EMMenu (Tietz), and Leginon (Suloway et al., 2009; chapter 14, Volume 483). In particular, Leginon and batch tomography from FEI Xplore3D were used successfully to

A

B

C

Figure 13.1 Dose series of virus: determining the optimal electron dosage. A series of ˚ 2) images of viruses at 0 tilt are recorded. Three of them (10e, 80e, high-dose (10 e/A and 160e) are shown in (A)–(C), respectively. The difference among these images is subtle. The contrast of image (C) is very good, but the fine detail of viruses (e.g., membrane bilayer) disappears. In order to conserve the fine details of virus, a total electron dose of between 80e and 120e was selected to collected tilt series. The scale bar is 100 nm.

Cryoelectron Tomography of HIV

273

collect large amounts of data (Liu et al., 2008, 2009; Suloway et al., 2009). The major benefit of batch tomography is the automatic collection of data without human intervention. In the FEI Polara electron microscope, the specimen can be maintained at liquid nitrogen temperature for 24 h, and as many as 3000 2K 2K CCD images can be taken during this period. For single-axis tilt series acquisition, the angular range is usually from 70 to þ70 with fixed angular increments of 1 or 2 . Batch tomography in FEI Xplore3D and Leginon starts at 0 , then tilts to 70 with the chosen increment. After reaching the highest tilt angle, the stage will return to 0 and then tilt to þ70 with the same increment. Two images at 0 tilt are collected, one in the beginning and the other one on the return of the stage to 0 . A comparison of these two images provides information about radiation-induced changes to the sample or to the added gold markers, as well as any changes of other microscope parameters during the image acquisition. Frozen-hydrated virus specimens were imaged at liquid nitrogen temperatures using a Polara field emission gun electron microscope (FEI) equipped with a 2K 2K CCD placed at the end of either a GIF 2002 or Tridium energy filter (Gatan). The electron microscopes were operated in two slightly different conditions, but for similar purposes of enhancing image contrast (Liu et al., 2008; Wright et al., 2007). In order to determine 2.0 nm resolution structures without taking CTF correction into consideration, about 2 mm defocus was chosen to collect low-dose single-axis tilt series for HIV Env studies. At this defocus, energy filtering with a narrow slit (20 eV) becomes critical for enhancing the image contrast.

2.4. 3D tomographic reconstruction The generation of a 3D reconstruction from a tilt series can be subdivided into three tasks: (1) preprocessing of the raw micrographs, (2) alignment of the tilt series, and (3) the computation of the tomogram. The preprocessing removes image imperfections, such as density gradients or density outliers (“hot” or “cold” pixels of the CCD camera). The images in the tilt series must be aligned prior to the computation of a 3D map, because a common reference frame and accurate geometric parameters must be determined before a 3D map can be computed. The collected images are not in register for various reasons: noneucentricity or mechanical instability of the stage, specimen drift, or poor tracking and recentering. The commonly used alignment methods fall into two categories, marker-based and marker-free methods (Chapter 13, Volume 482). In earlier studies (Zhu et al., 2006, 2008), a purely marker-free approach (Winkler and Taylor, 2006) was used. More recently, a hybrid approach which is based on an initial fiducial alignment with IMOD (Kremer et al., 1996) or Inspect3D (FEI) followed by a marker-free projection-matching

274

Jun Liu et al.

refinement was used to analyze large numbers of tilt series (Liu et al., 2008, 2009). The marker-free method as implemented in the protomo software package (Winkler and Taylor, 2006) usually works well if specimens are less than 200 nm in thickness (Dai et al., 2008; Liu et al., 2006; Ye et al., 2008; Zhu et al., 2006, 2008). The advantage of this method is that by using crosscorrelation techniques, the alignment is based directly on the signal produced by structural features of the specimen and not on artificially introduced gold markers. The alignment principle is similar to the projection-matching approach in single particle reconstruction and thus permits iterative refinement. A preliminary map computed from already aligned images in the tilt series is reprojected and used to refine the alignment. An additional step is the reevaluation of more accurate geometric parameters. Alignment and geometry refinement are alternated and it generally takes multiple cycles of alignment and geometry refinement to process a tilt series, so that this method is computationally more expensive than the markerbased method. For fast and reliable marker-based alignment, IMOD (Kremer et al., 1996) and Inspect3D (FEI) are useful packages presently available. The disadvantage of this approach is the relatively large size of gold markers, which are about 5 nm or larger in diameter so that their positions can only be determined with limited accuracy. Furthermore, the markers may not be immobile during the course of data acquisition, especially in frozenhydrated specimens. Thus, the alignment based on gold markers may not guarantee the optimal alignment of specimen features. Taking into account these considerations, a hybrid method combining the marker-based and marker-free approaches was developed for the Env project. As a first step, the tilt series were initially aligned with gold markers, and as a second step, projection matching was used to refine the alignment. The cross-correlation in the second step is carried out with a band-pass filter of which the high-pass filter component suppresses very low spatial frequencies, and the low-pass filter component excludes any signal beyond the first zero of the contrast transfer function. Computation of the final tomograms is carried out with protomo (Winkler and Taylor, 2006) which uses a weighted backprojection algorithm implementing general weighting functions (Harauz and van Heel, 1986).

2.5. Subvolume analysis The goal of subvolume analysis is the extraction of smaller volumes from tomograms that contain structural motifs of interest and the rendering of the motifs with improved SNR and resolution for further study. This process may include a classification step if the motifs are structurally heterogeneous, in order to average only similar motifs. The subvolume

Cryoelectron Tomography of HIV

275

analysis also needs to take into account the problem of missing information in reciprocal space, which is the cause of image artifacts in the tomograms. The missing information in a single-axis tilt series is usually referred to as the “missing wedge,” which arises due to the fact that the data is collected from a limited angular tilt range in the microscope (e.g., from 70 to þ70 ). The effects of the missing wedge can be eliminated in the final averaged image only if the motifs occur in various orientations in the tomograms. In other words, the orientation of the missing wedge relative to the motif structure varies correspondingly, and the gaps in reciprocal space of one motif overlap with sampled regions in other motifs, so that in the final merged image missing regions are effectively eliminated. The processing pipeline of subvolume analysis starts with locating structural motifs in the tomograms. In general, these motifs are oriented variably within the tomogram; so that the initial search must take into account not only the position but also the orientation. The orientation search considerably increases the computational effort of automatically identifying the motifs. Algorithms are available for 3D template matching which are primarily based on cross-correlation techniques (Frangakis, 2006). Threedimensional template matching is much more demanding than its 2D counterpart mainly because of effects of the missing wedge, the low contrast and the poor SNR of typical cryotomograms. In the subvolume analysis of HIV Env, the motifs (the Env spikes) were manually picked using the graphical display program “tomopick” in protomo that lets the user scroll through the volume and identify spike locations at various depths in the tomogram (Winkler, 2007). Alternatively, IMOD (Kremer et al., 1996) and Chimera (Pettersen et al., 2004) can be used for this purpose. The orientation in space was derived from the measured spike positions under the assumption that the virions can be approximated by a spherical or ellipsoidal surface. The unknown coefficients of the equation for an ellipsoidal surface are calculated with the measured positions as input by a least squares fit. At each measured point, a surface normal is computed, based on the equation of the fitted ellipsoid. An approximate direction of the spike axis can then be derived from the surface normal. Only the rotation about the spike axis remains unknown, and must be determined later at the alignment stage. In order to form an average of the raw motifs, subvolumes are extracted from the tomograms and aligned with respect to each other. The extraction or windowing is carried out either explicitly, by copying the subvolume to a new image, or implicitly, by simply recording the motif position and orientation. In the former case, the size of the data set can be reduced if the motifs are widely separated in the original tomograms. The latter case is more flexible, though, since window sizes can be changed easily at any point during processing. Subvolumes are aligned by cross-correlation and effects of the missing wedge are compensated by the use of constrained correlation

276

Jun Liu et al.

(Fo¨rster et al., 2005; chapter 11, Volume 483). The rotational alignment is performed with an orientation search by maximizing the cross-correlation coefficient; the translational alignment is simultaneously derived from the position of the correlation peak maximum. If the approximate orientation of the motif is known, as is the case for the HIV Env, the grid search can be limited in accordance with the expected deviation of the orientation estimates; otherwise the full range of the three Euler angles must be scanned, which comes at a higher computational cost and is less reliable. Averaging is carried out with a merging procedure in reciprocal space that takes into account the regions with missing information in the raw motifs that are averaged.

2.6. Alignment strategies If the examined motifs are heterogeneous, the alignment procedure outlined above cannot be directly applied. In the studies of HIV Env a contributing factor to the structural heterogeneity are orientational differences. The motif orientation relative to the tomogram determines which regions of reciprocal space fall in the sector covered by the missing wedge. The missing information affects the rendering of structural details. For instance, the membranes of spikes picked from the top or bottom of a virion almost disappear, whereas they are clearly delineated in the side views. Unless spikes are selected according to orientation, say from the top or bottom of virions, classification methods must be employed to ensure that only similarly oriented motifs are averaged. One strategy that is commonly used in single particle analysis is multireference alignment and classification to separate various projections of a macromolecule (Penczek et al., 1994). In the case of heterogeneous 3D motifs the procedure of aligning and classifying the motifs is similar: (1) create multiple references for alignment that are representative of the heterogeneous population, (2) align each raw motif to all references and select the best according to cross-correlation peak height, (3) classify the aligned motifs, and (4) average each class separately. Class averages can then be used for further alignment and classification cycles. The number of classes that are computed are chosen according to the number of expected conformations in the heterogeneous data set. This is impossible if there is a continuum of conformations rather than discrete states, for instance, if flexible parts are present in a molecule. Since some clustering algorithms, such as hierarchical ascendant classification, require a specific number of classes as input, several sets of class averages were computed, usually 4, 10, and 20 classes for the Env spike data, which were subsequently compared visually to identify significant structural differences. The second strategy called “alignment by classification” was developed recently (Winkler et al., 2009) to minimize possible reference bias problem,

Cryoelectron Tomography of HIV

277

that is, the choice of the references may influence the outcome of the alignment: (1) classify all motifs, (2) compute class averages, (3) align all class averages with respect to each other, and (4) apply the alignment transformations of each class to the respective class members. This procedure is carried out repeatedly. No bias is introduced since no arbitrary reference selection takes place. Also, the alignment is carried out with class averages that have a higher SNR so that it is more robust than a multireference alignment of raw motifs. The number of classes computed in this procedure should be chosen as high as possible, because the goal is to capture as many different spatial orientations as possible in addition to the conformational differences. There is a tradeoff to consider, however, between a large number of classes versus the number of members per class. The number of averaged class members directly determines the improvement of the SNR, and the optimal choice depends on the total number of motifs in the entire data set. After completion of the alignment procedure, the same criteria as for multireference alignment apply for the final classification. The initial alignment of the Env spikes warrants a further discussion. As mentioned above, only two of the three Euler angles can be determined with the described fitting method in the beginning, namely those that indicate the direction of the spike axis. In order to determine the third one, the rotation about the spike axis, the directional angles are first refined by an alignment to a global average, that was rotationally averaged about the spike axis, so that only two angles need to be scanned. A classification is then carried out and the class averages are aligned with respect to each other rotationally about the spike axis. This will identify motifs with structural differences and unify similar motifs that differ only in orientation, rotationally. The procedure resembles the second strategy, with the exception that one or two angles are restricted in the alignment. A further problem is the effect of the missing wedge at the classification stage that tends to group subvolumes according to the orientation of the missing wedge rather than structural differences (Walz et al., 1997). One way to overcome this problem is the use of constrained cross-correlation as a similarity measure for the classification (Fo¨rster et al., 2008). A similar approach based on the signal overlap of a pair of subvolumes in reciprocal space is described by Bartesaghi et al. (2008). This method uses a variant of the multireference alignment method described above which includes refinement steps in the formation of class averages. There is evidence that classification without missing wedge compensation does not necessarily lead to an orientation based grouping of the motifs. It has been shown that a judicious choice of the classification mask can essentially eliminate the effects of the missing wedge (Winkler et al., 2009). This was verified by plotting the directions of the tilt axis for each class member, and inspecting the resulting distribution pattern visually. If the missing wedge were the

278

Jun Liu et al.

driving factor in the classification, it would result in a particular direction within a class being preferred.

3. 3D Visualization of Intact HIV Virion 3.1. Cryoelectron tomography of HIV virions A number of approaches have been used for the production and purification of concentrated samples of noninfectious HIV for cryo-ET analysis. For studies of viral maturation, all concentrated HIV preparations were kindly provided by Dr Wesley I. Sundquist’s laboratory located at the University of Utah, Salt Lake City, UT. For studies of HIV Env, concentrated virus samples (HIVBaL; from AIDS Vaccine Program, SAIC-Frederick, National Cancer Institute) were purified by sucrose gradient centrifugation and inactivated by treatment with Aldrithiol-2 (AT-2). The use of AT-2 inactivates viral infectivity through covalent modification of internal viral proteins while retaining the functional and structural integrity of HIV Env (Arthur et al., 1998).

3.2. Molecular structure of the gag shells Tilt series images of immature HIV-1 virions were binned twofold and tomographic reconstructions were generated by weighted backprojection using the IMOD package (Kremer et al., 1996). Individual immature virions were selected and cropped out of the completed tomograms using IMOD tools and were subsequently denoised by nonlinear anisotropic diffusion as implemented in BSOFT (Heymann, 2001). In order to examine and analyze the structure of the Gag shells within the immature HIV virions, the Amira software package (Visage Imaging, Inc.) was used to generate radial density profiles. The profiles were produced by sampling spherical shells of the virions and plotting the density per unit area as a function of radius. Surface projections were created to better represent the surface area covered by each of the Gag proteins over each spherical shell. The surface projections were created by first generating a sphere to represent each Gag shell. Second, the electron densities observed in each shell were sampled along vectors normal to a triangle mesh surface imposed onto each sphere. Then, the summed densities of the vectors were each assigned vertex points on the surface of each shell. To determine the percentage of the surface area within the CA NTD shell covered by ordered and disordered lattice, the surface area of the sphere was calculated with the Amira surface area module. Boundaries around disordered regions (or patches) were drawn manually and the surface area calculated.

Cryoelectron Tomography of HIV

279

To locate the unit cells, the points on the projected surfaces with the least density, that is, regions without protein, within a specified radius were selected. Those points were then labeled with the standard deviation of the distance to their six nearest neighbors, as a measure of the local degree of order in the hexagonal lattice. Amira extension modules created within the Jensen lab executed the preceding functions. For unit cell averaging, the set of extrema on the projected surface layers having the least variation in the distance to their neighbors was selected. For each of the selected points, a local coordinate system was established comprising the normal vector to the surface (the Z axis), the vector from the point to its nearest neighbor (the X axis), and their cross-product (the Y axis). The entire volume containing the virion was then rotated into alignment with the global axes and translated so as to superimpose all of the selected surface extrema. Density renderings of the CA and SP1 domains and atomic model fitting were performed with the UCSF Chimera (Pettersen et al., 2004). While in 2D projection, we observed the radial spoke arrangement of Gag, in 3D only the CA and SP1 domains of Gag contained patches of hexagonal order, which were resolved in the tomographic reconstructions (Fig. 13.2). The most significant finding was the arrangement of the doublelayered hexagonal lattice formed from the CA and SP1 domains. Upon further analysis, we proposed a model in which individual CA hexamers are stabilized by a bundle of six SP1 helices (Fig. 13.3). At the resolution of our data, we could not determine which portions of the associated CA and NC domains were involved in the formation of the helices. However, this structural discovery suggests why the SP1 spacer is essential for assembly of the Gag lattice and how cleavage between SP1 and CA acts as one of the structural switches controlling maturation. The structural interpretation was supported by evidence demonstrating that the Gag SP1 region is critical for viral assembly (Accola et al., 1998; Gay et al., 1998; Li et al., 2003; Liang et al., 2002) and functions as a maturation “switch” (Accola et al., 1998; Gross et al., 2000). The junction between SP1 and CA of the immature Gag lattice has also become a target for the development of retroviral maturation inhibitors. One group of compounds, the betulinic acid derivatives, has been demonstrated to be successful in blocking the cleavage of the CA-SP1 connection (Li et al., 2003; Zhou et al., 2005). One derivative, PA-457 or Bevirimat, has recently entered Phase 2 clinical trials in order to assess its efficacy for suppression of the virus.

3.3. Molecular architecture of HIV Env Typically, for each tilt series, a 1900 1900 500 reconstruction ( 7 GB) was generated by weighted backprojection as implemented in protomo (Winkler and Taylor, 2006). After 4 4 4 binning, contrast inversion and low-pass filtering, the tomograms were examined carefully

280

Jun Liu et al.

A

B

C

D

Figure 13.2 Raw image and slices through the 3D reconstruction of an HIV-1 virion. (A) The central image from a tilt series of an immature HIV-1 virion. (B–D) 5.6 nm slices through the 3D reconstruction, displaying the order of the Gag lattice. In (B), large gaps are present in the Gag lattice below the lipid bilayer (black arrows and bars mark one ordered region of the Gag lattice). In regions above visibly ordered Gag, the membrane-MA layer appears bilaminar (black arrowhead). In regions where there is no visibly ordered Gag, it appears unilaminar (white arrowhead). Scale bar 50 nm (from Wright et al., 2007, with permission).

by IMOD and the virions with visible spikes were identified and extracted from the original tomograms, based on the coordinates of their centers. Each extracted 320 320 320 volume contains one individual virion ˚ . The same virion was also stored in with an original sampling size of 4.1 A an 80 80 80 volume after 4 4 4 binning as described above. The significantly enhanced contrast of the binned maps is the key for visualizing the intact virions and picking Env spikes (Fig. 13.4). Surface spikes on each virion were identified by visual inspection, using “tomopick” in protomo (Winkler, 2007) or UCSF Chimera (Pettersen et al., 2004). In total, 4741 spikes were selected from 382 HIV virions, 4323 spikes from 306 HIV virions complexed with b12-fab, 4849 spikes from 292 HIV virions complexed with full-length b12 and 4900 spikes from 503 HIV virions complexed with CD4 and 17b-Fab (Liu et al., 2008). To speed up the image analysis, 2 2 2 binned subvolumes of the spikes were generated for the initial rounds of alignment and classification. Binning also increases the SNR, and thus results in a more reliable initial alignment and classification. This has proven to be critical for tomographic subvolume analysis, especially if large amounts of low contrast data need to

Cryoelectron Tomography of HIV

281

Figure 13.3 A model of the arrangement of Gag domains within the immature lattice. Top and side views (top and bottom, respectively) of the averaged Gag lattice (gray surface) with atomic models of CA NTD (cyan), CA CTD (yellow), and SP1 (magenta) fit into the density. Note that this is one interpretation of the density in terms of gross molecular architecture. Scale bar 8 nm (from Wright et al., 2007, with permission). (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this chapter.)

be processed. As a first step, a global average of all the extracted volumes ( 4000) is formed and if the positions and orientations were determined correctly, the average should reveal the membrane and the mushroom shaped spike (Fig. 13.5A). Since only two of the three Euler angles were determined, the spikes are not yet in register axially, thus the mushroom shaped appearance. A translational alignment was carried out based on the center of mass of the spike density, and an ellipsoidal mask (left panels in Fig. 13.5A) that included the spike volume was applied. Initial classification within a mask (defined in right panels of Fig. 13.5A) clearly showed the inherent threefold symmetry in the spike structure (Fig. 13.5B–E). The further alignment, multivariate statistical analysis and classification were performed as described by (Winkler, 2007) and later confirmed with an

282

Jun Liu et al.

A

B

Figure 13.4 The 4 4 4 binned map of a tomographic reconstruction reveals significant contrast for 3D visualization of intact virion (B) and identification of Env spikes. The square size in (A) is 120 120 nm.

alternative approach (Bartesaghi et al., 2008). Once the structures were convincingly ascertained, threefold symmetry was imposed for subsequent rounds of refinement. Fourier shell correlation (FSC) was used to estimate the resolution of the final structures. The molecular architecture of the HIV Env trimer is shown in Fig. 13.6D. There are several atomic models of monomeric gp120 available in the PDB, and three of them are displayed within their 2.0 nm resolution envelopes which were calculated by “pdb2mrc” in EMAN (Ludtke et al., 1999). At 2.0 nm resolution, CD4 binding and b12 binding conformations of gp120 are very different from the ligand free conformation of gp120 from SIV. Although visual comparison of these maps suggests that two models (CD4 binding and b12 binding conformations) match the corresponding densities in the HIV Env map better, the resolution is insufficient for a quantitative comparison.

3.4. Molecular details of neutralizing antibody b12-Env complex b12 is a well known broadly cross-reactive, neutralizing antibody (Burton et al., 1994). A recent crystal structure of monomeric gp120 core complexed with Fab fragments from b12 not only reveals the conformationally invariant surface for initial CD4 attachment but also provides atomic-level detail of the b12 epitope (Zhou et al., 2007). Therefore, antibody b12 can also be utilized to map the CD4 binding site on HIV Env. Both b12 and its fab were incubated with intact virions separately. Due to the limited resolution, the resulting tomographic reconstructions show little evidence of antibody binding. However, the class averages of b12 (or b12 fab) and Env complex

283

Cryoelectron Tomography of HIV

HIV Env

HIV Env + b12fab

A

B

C

D

E

Side view

Top view

Side view

Top view

Figure 13.5 Initial classification of HIV Env spikes. Global average shows the membrane and mushroom shaped spike (A). Two different masks were used for 3D alignment (left) and 3D classification (right), respectively. Four class averages from initial classification were generated from HIV Env spikes without (left) and with b12 fab (right). The square size is 30 30 nm.

clearly reveal extra density, compared to the native Env without any ligand (Fig. 13.5). The extra density from a refined structure (Fig. 13.7) corresponds very well with one antibody fab. A pseudo-atomic model of the HIV gp120 trimer (Liu et al., 2008) was produced by fitting the crystal structure of b12 in complex with an HIV gp120 core (Zhou et al., 2007) into EM density maps using Chimera (Pettersen et al., 2004). The unassigned densities (Fig. 13.7C) at the apex of HIV Env likely correspond to the variable loops (e.g., V1/V2), which are supported by recent V1/V2 deletion mutant studies (Hu, Liu, Taylor, and Roux, unpublished work). The unassigned densities at the bottom of gp120 represent N/C-termini, which are confirmed by a recent crystal structure (Pancera et al., 2010). This result indicates that cryo-ET is becoming a powerful and reliable technique for

284

Jun Liu et al.

A

D Env PDB:1GC1 CD4-binding conformation

B PDB:2NY7 B12-binding conformation

Viral membrane

C PDB:2BF1 ligand-free conformation

Figure 13.6 3D density map of HIV Env at 2.0 nm resolution (Liu et al., 2008). Comparison of EM map (D) with simulated maps calculated from three gp120 monomer crystal structures: CD4-binding conformation (A), b12-binding conformation (B), and ligand free conformation (C) indicates that the molecular docking remains challenging at 2.0 nm resolution. The scale bar is 5 nm.

A

B

b12

C

b12fab

Env

gp120 Viral membrane

Viral membrane

Figure 13.7 3D structure of HIV Env and b12 complex clearly reveal extra density (B), compared to that of the native Env without any ligand (A). The extra density corresponds very well with one antibody fab. A pseudo-atomic model (Liu et al., 2008) of HIV gp120 trimer as shown in (C) is derived based on the rigid body fit of crystal structure (Zhou et al., 2007) from the complex of gp120 and b12 fab. The scale bar is 5 nm.

determining the native structures of HIV Env and its interactions with antibodies at molecular resolution. Currently, there are considerable interests in understanding the interaction between HIV Env and other well known broadly neutralizing antibodies (2G12 against gp120, 2F5, and 4E10 against gp41). It is critical to identify functionally conserved regions of the highly variable HIV Env for better understanding of their structural basis for neutralization. It will certainly enhance the development of novel antibodies capable of neutralizing diverse HIV isolates.

285

Cryoelectron Tomography of HIV

3.5. CD4 induces major conformational change of HIV Env It was believed that HIV Env undergoes a series of conformational changes when it interacts with receptor (CD4) and coreceptor on the host cell surface, leading to fusion of viral and cellular membranes. In the light of high resolution crystal structures, especially of the ternary complex (gp120 with both CD4 and CD4 induced antibody 17b; Kwong et al., 1998), the 3D structure of native HIV Env in a ternary complex with CD4 and 17b was determined (Liu et al., 2008) within a month from data acquisition to modeling. This could be achieved mainly because of the application of a high-throughput cryo-ET procedure. Most remarkably, the intact crystal structure of the ternary complex can be fitted into EM map as one rigid body (Fig. 13.8A). This structure suggests that the binding of Env to CD4 results in a major reorganization (Fig. 13.8B) of the Env trimer and a close contact between the virus and target cell coreceptor, thus facilitating viral entry (Fig. 13.8C).

A

B 17b

gp120

CD4 CD4 gp120 Viral membrane CD4

C

Cell membrane Coreceptor

CD4

“Closed”

gp120

“Open”

gp41

Viral membrane

Figure 13.8 3D structure of HIV Env with CD4 and CD4 induced antibody 17b (Liu et al., 2008). The crystal structure from a ternary complex of gp120 with CD4 and 17b (Kwong et al., 1998) was fitted in EM density map as rigid body (A). gp120 was colored in magenta, CD4 in yellow, 17b fab in light green. CD4 induces huge conformational change in the HIV Env spike (B). A model of viral entry was proposed in (C). The scale bar is 5 nm.

286

Jun Liu et al.

4. Conclusions and Perspective Cryo-ET is the method of choice to study intact heterogeneous viruses at molecular resolution (Grunewald et al., 2003; Subramaniam et al., 2007). The recent technological innovations have established cryoET as the most advanced technique for determining 3D structures of macromolecular complexes in situ (Beck et al., 2007; Liu et al., 2008, 2009; Murphy et al., 2006). However, poor SNR and the lack of effective protein labeling techniques remain two significant limitations of cryo-ET methodologies. We believe that further advances in cryo-ET through the development of innovative techniques for increasing throughput and resolution, adapting traditional antibody labeling, and combining genetic approaches with cryo-ET for the specific characterization of macromolecular complexes in situ will provide greater insights into the fascinating cellular processes of living organisms.

ACKNOWLEDGMENTS J. L. thanks Dr James Stoops for comments on and discussion of the manuscript. The authors are thankful to Dr Grant Jensen for helpful suggestions. J. L. thanks Dr Ken Roux for sharing unpublished results. The work reported here was done in the laboratories of Drs Grant Jensen, Sriram Subramanian and Ken Taylor. J. L. is supported by a Welch Foundation Grant AU-1714, NIH grant 1R01AI087946.

REFERENCES Accola, M. A., Hoglund, S., and Gottlinger, H. G. (1998). A putative alpha-helical structure which overlaps the capsid-p2 boundary in the human immunodeficiency virus type 1 Gag precursor is crucial for viral particle assembly. J. Virol. 72, 2072–2078. Arthur, L. O., Bess, J. W., Jr., Chertova, E. N., Rossio, J. L., Esser, M. T., Benveniste, R. E., Henderson, L. E., and Lifson, J. D. (1998). Chemical inactivation of retroviral infectivity by targeting nucleocapsid protein zinc fingers: A candidate SIV vaccine. AIDS Res. Hum. Retroviruses 14(Suppl 3), S311–S319. Bartesaghi, A., Sprechmann, P., Liu, J., Randall, G., Sapiro, G., and Subramaniam, S. (2008). Classification and 3D averaging with missing wedge correction in biological electron tomography. J. Struct. Biol. 162, 436–450. Baumeister, W., Grimm, R., and Walz, J. (1999). Electron tomography of molecules and cells. Trends Cell Biol. 9, 81–85. Beck, M., Lucic, V., Fo¨rster, F., Baumeister, W., and Medalia, O. (2007). Snapshots of nuclear pore complexes in action captured by cryo-electron tomography. Nature 449, 611–615. Benjamin, J., Ganser-Pornillos, B. K., Tivol, W. F., Sundquist, W. I., and Jensen, G. J. (2005). Three-dimensional structure of HIV-1 virus-like particles by electron cryotomography. J. Mol. Biol. 346, 577–588.

Cryoelectron Tomography of HIV

287

Briggs, J. A., Wilk, T., Welker, R., Krausslich, H. G., and Fuller, S. D. (2003). Structural organization of authentic, mature HIV-1 virions and cores. EMBO J. 22, 1707–1715. Briggs, J. A., Simon, M. N., Gross, I., Krausslich, H. G., Fuller, S. D., Vogt, V. M., and Johnson, M. C. (2004). The stoichiometry of Gag protein in HIV-1. Nat. Struct. Mol. Biol. 11, 672–675. Briggs, J. A., Grunewald, K., Glass, B., Fo¨rster, F., Krausslich, H. G., and Fuller, S. D. (2006a). The mechanism of HIV-1 core assembly: Insights from three-dimensional reconstructions of authentic virions. Structure 14, 15–20. Briggs, J. A., Johnson, M. C., Simon, M. N., Fuller, S. D., and Vogt, V. M. (2006b). Cryoelectron microscopy reveals conserved and divergent features of gag packing in immature particles of Rous sarcoma virus and human immunodeficiency virus. J. Mol. Biol. 355, 157–168. Briggs, J. A., Riches, J. D., Glass, B., Bartonova, V., Zanetti, G., and Krausslich, H. G. (2009). Structure and assembly of immature HIV. Proc. Natl. Acad. Sci. USA 106, 11090–11095. Burton, D. R., Pyati, J., Koduri, R., Sharp, S. J., Thornton, G. B., Parren, P. W., Sawyer, L. S., Hendry, R. M., Dunlop, N., Nara, P. L., et al. (1994). Efficient neutralization of primary isolates of HIV-1 by a recombinant human monoclonal antibody. Science 266, 1024–1027. Byeon, I. J., Meng, X., Jung, J., Zhao, G., Yang, R., Ahn, J., Shi, J., Concel, J., Aiken, C., Zhang, P., and Gronenborn, A. M. (2009). Structural convergence between Cryo-EM and NMR reveals intersubunit interactions critical for HIV-1 capsid function. Cell 139, 780–790. Chan, D. C., Fass, D., Berger, J. M., and Kim, P. S. (1997). Core structure of gp41 from the HIV envelope glycoprotein. Cell 89, 263–273. Chen, B., Vogan, E. M., Gong, H., Skehel, J. J., Wiley, D. C., and Harrison, S. C. (2005). Structure of an unliganded simian immunodeficiency virus gp120 core. Nature 433, 834–841. Chen, L., Kwon, Y. D., Zhou, T., Wu, X., O’Dell, S., Cavacini, L., Hessell, A. J., Pancera, M., Tang, M., Xu, L., Yang, Z. Y., Zhang, M. Y., et al. (2009). Structural basis of immune evasion at the site of CD4 attachment on HIV-1 gp120. Science 326, 1123–1127. Dai, W., Jia, Q., Bortz, E., Shah, S., Liu, J., Atanasov, I., Li, X., Taylor, K. A., Sun, R., and Zhou, Z. H. (2008). Unique structures in a tumor herpesvirus revealed by cryo-electron tomography and microscopy. J. Struct. Biol. 161, 428–438. Douglas, C. C., Thomas, D., Lanman, J., and Prevelige, P. E., Jr. (2004). Investigation of N-terminal domain charged residues on the assembly and stability of HIV-1 CA. Biochemistry 43, 10435–10441. Dubochet, J., Adrian, M., Chang, J. J., Homo, J. C., Lepault, J., McDowall, A. W., and Schultz, P. (1988). Cryo-electron microscopy of vitrified specimens. Q. Rev. Biophys. 21, 129–228. Fo¨rster, F., Medalia, O., Zauberman, N., Baumeister, W., and Fass, D. (2005). Retrovirus envelope protein complex structure in situ studied by cryo-electron tomography. Proc. Natl. Acad. Sci. USA 102, 4729–4734. Fo¨rster, F., Pruggnaller, S., Seybert, A., and Frangakis, A. S. (2008). Classification of cryoelectron sub-tomograms using constrained correlation. J. Struct. Biol. 161, 276–286. Frangakis, A. S., and Rath, B. K. (2006). Motif search in electron tomography. In “Electron Tomography: Methods for Three-Dimensional Visualization of Structures in the Cell,” ( J. Frank, ed.). Springer, New York, London. Freed, E. O. (1998). HIV-1 gag proteins: Diverse functions in the virus life cycle. Virology 251, 1–15. Fuller, S. D., Wilk, T., Gowen, B. E., Krausslich, H. G., and Vogt, V. M. (1997). Cryoelectron microscopy reveals ordered domains in the immature HIV-1 particle. Curr. Biol. 7, 729–738.

288

Jun Liu et al.

Ganser, B. K., Li, S., Klishko, V. Y., Finch, J. T., and Sundquist, W. I. (1999). Assembly and analysis of conical models for the HIV-1 core. Science 283, 80–83. Ganser, B. K., Cheng, A., Sundquist, W. I., and Yeager, M. (2003). Three-dimensional structure of the M-MuLV CA protein on a lipid monolayer: A general model for retroviral capsid assembly. EMBO J. 22, 2886–2892. Ganser-Pornillos, B. K., Cheng, A., and Yeager, M. (2007). Structure of full-length HIV-1 CA: A model for the mature capsid lattice. Cell 131, 70–79. Gay, B., Tournier, J., Chazal, N., Carriere, C., and Boulanger, P. (1998). Morphopoietic determinants of HIV-1 Gag particles assembled in baculovirus-infected cells. Virology 247, 160–169. Gelderblom, H. R., Hausmann, E. H., Ozel, M., Pauli, G., and Koch, M. A. (1987). Fine structure of human immunodeficiency virus (HIV) and immunolocalization of structural proteins. Virology 156, 171–176. Gross, I., Hohenberg, H., Wilk, T., Wiegers, K., Grattinger, M., Muller, B., Fuller, S., and Krausslich, H. G. (2000). A conformational switch controlling HIV-1 morphogenesis. EMBO J. 19, 103–113. Grunewald, K., Desai, P., Winkler, D. C., Heymann, J. B., Belnap, D. M., Baumeister, W., and Steven, A. C. (2003). Three-dimensional structure of herpes simplex virus from cryo-electron tomography. Science 302, 1396–1398. Harauz, G., and van Heel, M. (1986). Exact filters for general geometry three dimensional reconstruction. Optik 73, 11. Heymann, J. B. (2001). Bsoft: Image and molecular processing in electron microscopy. J. Struct. Biol. 133, 156–169. Huang, C. C., Tang, M., Zhang, M. Y., Majeed, S., Montabana, E., Stanfield, R. L., Dimitrov, D. S., Korber, B., Sodroski, J., Wilson, I. A., Wyatt, R., and Kwong, P. D. (2005). Structure of a V3-containing HIV-1 gp120 core. Science 310, 1025–1028. Koster, A. J., Chen, H., Sedat, J. W., and Agard, D. A. (1992). Automated microscopy for electron tomography. Ultramicroscopy 46, 207–227. Koster, A. J., Grimm, R., Typke, D., Hegerl, R., Stoschek, A., Walz, J., and Baumeister, W. (1997). Perspectives of molecular and cellular electron tomography. J. Struct. Biol. 120, 276–308. Kremer, J. R., Mastronarde, D. N., and McIntosh, J. R. (1996). Computer visualization of three-dimensional image data using IMOD. J. Struct. Biol. 116, 71–76. 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, 648–659. Lanman, J., Sexton, J., Sakalian, M., and Prevelige, P. E., Jr. (2002). Kinetic analysis of the role of intersubunit interactions in human immunodeficiency virus type 1 capsid protein assembly in vitro. J. Virol. 76, 6900–6908. Lanman, J., Lam, T. T., Barnes, S., Sakalian, M., Emmett, M. R., Marshall, A. G., and Prevelige, P. E., Jr. (2003). Identification of novel interactions in HIV-1 capsid protein assembly by high-resolution mass spectrometry. J. Mol. Biol. 325, 759–772. Li, S., Hill, C. P., Sundquist, W. I., and Finch, J. T. (2000). Image reconstructions of helical assemblies of the HIV-1 CA protein. Nature 407, 409–413. Li, F., Goila-Gaur, R., Salzwedel, K., Kilgore, N. R., Reddick, M., Matallana, C., Castillo, A., Zoumplis, D., Martin, D. E., Orenstein, J. M., Allaway, G. P., Freed, E. O., et al. (2003). PA-457: A potent HIV inhibitor that disrupts core condensation by targeting a late step in Gag processing. Proc. Natl. Acad. Sci. USA 100, 13555–13560. Liang, C., Hu, J., Russell, R. S., Roldan, A., Kleiman, L., and Wainberg, M. A. (2002). Characterization of a putative alpha-helix across the capsid-SP1 boundary that is critical

Cryoelectron Tomography of HIV

289

for the multimerization of human immunodeficiency virus type 1 gag. J. Virol. 76, 11729–11737. Liu, J., Taylor, D. W., Krementsova, E. B., Trybus, K. M., and Taylor, K. A. (2006). Threedimensional structure of the myosin V inhibited state by cryoelectron tomography. Nature 442, 208–211. Liu, J., Bartesaghi, A., Borgnia, M. J., Sapiro, G., and Subramaniam, S. (2008). Molecular architecture of native HIV-1 gp120 trimers. Nature 455, 109–113. Liu, J., Lin, T., Botkin, D. J., McCrum, E., Winkler, H., and Norris, S. J. (2009). Intact flagellar motor of Borrelia burgdorferi revealed by cryo-electron tomography: Evidence for stator ring curvature and rotor/C-ring assembly flexion. J. Bacteriol. 191, 5026–5036. Lucic, V., Fo¨rster, F., and Baumeister, W. (2005). Structural studies by electron tomography: From cells to molecules. Annu. Rev. Biochem. 74, 833–865. Ludtke, S. J., Baldwin, P. R., and Chiu, W. (1999). EMAN: Semiautomated software for high-resolution single-particle reconstructions. J. Struct. Biol. 128, 82–97. Massiah, M. A., Starich, M. R., Paschall, C., Summers, M. F., Christensen, A. M., and Sundquist, W. I. (1994). Three-dimensional structure of the human immunodeficiency virus type 1 matrix protein. J. Mol. Biol. 244, 198–223. Massiah, M. A., Worthylake, D., Christensen, A. M., Sundquist, W. I., Hill, C. P., and Summers, M. F. (1996). Comparison of the NMR and X-ray structures of the HIV-1 matrix protein: Evidence for conformational changes during viral assembly. Protein Sci. 5, 2391–2398. Mastronarde, D. N. (2005). Automated electron microscope tomography using robust prediction of specimen movements. J. Struct. Biol. 152, 36–51. McEwen, B. F., Downing, K. H., and Glaeser, R. M. (1995). The relevance of dosefractionation in tomography of radiation-sensitive specimens. Ultramicroscopy 60, 357–373. Medalia, O., Weber, I., Frangakis, A. S., Nicastro, D., Gerisch, G., and Baumeister, W. (2002). Macromolecular architecture in eukaryotic cells visualized by cryoelectron tomography. Science 298, 1209–1213. Murphy, G. E., Leadbetter, J. R., and Jensen, G. J. (2006). In situ structure of the complete Treponema primitia flagellar motor. Nature 442, 1062–1064. Nermut, M. V., Hockley, D. J., Bron, P., Thomas, D., Zhang, W. H., and Jones, I. M. (1998). Further evidence for hexagonal organization of HIV gag protein in prebudding assemblies and immature virus-like particles. J. Struct. Biol. 123, 143–149. Nickell, S., Fo¨rster, F., Linaroudis, A., Net, W. D., Beck, F., Hegerl, R., Baumeister, W., and Plitzko, J. M. (2005). TOM software toolbox: Acquisition and analysis for electron tomography. J. Struct. Biol. 149, 227–234. Pancera, M., Majeed, S., Ban, Y. E., Chen, L., Huang, C. C., Kong, L., Kwon, Y. D., Stuckey, J., Zhou, T., Robinson, J. E., Schief, W. R., Sodroski, J., et al. (2010). Structure of HIV-1 gp120 with gp41-interactive region reveals layered envelope architecture and basis of conformational mobility. Proc. Natl. Acad. Sci. USA 107, 1166–1171. Penczek, P. A., Grassucci, R. A., and Frank, J. (1994). The ribosome at improved resolution: New techniques for merging and orientation refinement in 3D cryo-electron microscopy of biological particles. Ultramicroscopy 53, 251–270. Pettersen, E. F., Goddard, T. D., Huang, C. C., Couch, G. S., Greenblatt, D. M., Meng, E. C., and Ferrin, T. E. (2004). UCSF Chimera—A visualization system for exploratory research and analysis. J. Comput. Chem. 25, 1605–1612. Pornillos, O., Ganser-Pornillos, B. K., Kelly, B. N., Hua, Y., Whitby, F. G., Stout, C. D., Sundquist, W. I., Hill, C. P., and Yeager, M. (2009). X-ray structures of the hexameric building block of the HIV capsid. Cell 137, 1282–1292. Sougrat, R., Bartesaghi, A., Lifson, J. D., Bennett, A. E., Bess, J. W., Zabransky, D. J., and Subramaniam, S. (2007). Electron tomography of the contact between T cells and SIV/HIV-1: Implications for viral entry. PLoS Pathog. 3, e63.

290

Jun Liu et al.

Subramaniam, S. (2005). Bridging the imaging gap: Visualizing subcellular architecture with electron tomography. Curr. Opin. Microbiol. 8, 316–322. Subramaniam, S., Bartesaghi, A., Liu, J., Bennett, A. E., and Sougrat, R. (2007). Electron tomography of viruses. Curr. Opin. Struct. Biol. 17, 596–602. Suloway, C., Shi, J., Cheng, A., Pulokas, J., Carragher, B., Potter, C. S., Zheng, S. Q., Agard, D. A., and Jensen, G. J. (2009). Fully automated, sequential tilt-series acquisition with Leginon. J. Struct. Biol. 167, 11–18. Walz, J., Typke, D., Nitsch, M., Koster, A. J., Hegerl, R., and Baumeister, W. (1997). Electron tomography of single ice-embedded macromolecules: Three-dimensional alignment and classification. J. Struct. Biol. 120, 387–395. Wilk, T., Gross, I., Gowen, B. E., Rutten, T., de Haas, F., Welker, R., Krausslich, H. G., Boulanger, P., and Fuller, S. D. (2001). Organization of immature human immunodeficiency virus type 1. J. Virol. 75, 759–771. Winkler, H. (2007). 3D reconstruction and processing of volumetric data in cryo-electron tomography. J. Struct. Biol. 157, 126–137. Winkler, H., and Taylor, K. A. (2006). Accurate marker-free alignment with simultaneous geometry determination and reconstruction of tilt series in electron tomography. Ultramicroscopy 106, 240–254. Winkler, H., Zhu, P., Liu, J., Ye, F., Roux, K. H., and Taylor, K. A. (2009). Tomographic subvolume alignment and subvolume classification applied to myosin V and SIV envelope spikes. J. Struct. Biol. 165, 64–77. Wright, E. R., Schooler, J. B., Ding, H. J., Kieffer, C., Fillmore, C., Sundquist, W. I., and Jensen, G. J. (2007). Electron cryotomography of immature HIV-1 virions reveals the structure of the CA and SP1 Gag shells. EMBO J. 26, 2218–2226. Ye, F., Liu, J., Winkler, H., and Taylor, K. A. (2008). Integrin alpha IIb beta 3 in a membrane environment remains the same height after Mn2þ activation when observed by cryoelectron tomography. J. Mol. Biol. 378, 976–986. Zanetti, G., Briggs, J. A., Grunewald, K., Sattentau, Q. J., and Fuller, S. D. (2006). Cryoelectron tomographic structure of an immunodeficiency virus envelope complex in situ. PLoS Pathog. 2, e83. Zheng, Q. S., Braunfeld, M. B., Sedat, J. W., and Agard, D. A. (2004). An improved strategy for automated electron microscopic tomography. J. Struct. Biol. 147, 91–101. Zhou, J., Huang, L., Hachey, D. L., Chen, C. H., and Aiken, C. (2005). Inhibition of HIV-1 maturation via drug association with the viral Gag protein in immature HIV-1 particles. J. Biol. Chem. 280, 42149–42155. Zhou, T., Xu, L., Dey, B., Hessell, A. J., Van Ryk, D., Xiang, S. H., Yang, X., Zhang, M. Y., Zwick, M. B., Arthos, J., Burton, D. R., Dimitrov, D. S., et al. (2007). Structural definition of a conserved neutralization epitope on HIV-1 gp120. Nature 445, 732–737. Zhu, P., Chertova, E., Bess, J., Jr., Lifson, J. D., Arthur, L. O., Liu, J., Taylor, K. A., and Roux, K. H. (2003). Electron tomography analysis of envelope glycoprotein trimers on HIV and simian immunodeficiency virus virions. Proc. Natl. Acad. Sci. USA 100, 15812–15817. Zhu, P., Liu, J., Bess, J., Jr., Chertova, E., Lifson, J. D., Grise, H., Ofek, G. A., Taylor, K. A., and Roux, K. H. (2006). Distribution and three-dimensional structure of AIDS virus envelope spikes. Nature 441, 847–852. Zhu, P., Winkler, H., Chertova, E., Taylor, K. A., and Roux, K. H. (2008). Cryoelectron tomography of HIV-1 envelope spikes: Further evidence for tripod-like legs. PLoS Pathog. 4, e1000203.

C H A P T E R

F O U R T E E N

Automation in Single-Particle Electron Microscopy: Connecting the Pieces Dmitry Lyumkis, Arne Moeller, Anchi Cheng, Amber Herold, Eric Hou, Christopher Irving, Erica L. Jacovetty, Pick-Wei Lau, Anke M. Mulder, James Pulokas, Joel D. Quispe, Neil R. Voss, Clinton S. Potter, and Bridget Carragher Contents 292 295 300 304 311 314 316 317 319 321 322 327 330 330

1. 2. 3. 4. 5.

Introduction Specimen Preparation Autoloading and Robotic Screening Microscopy Image Processing 5.1. CTF estimation and correction 5.2. Particle selection and stack creation 5.3. 2D alignment and classification 5.4. Ab initio reconstructions 5.5. 3D refinement 6. Assessment and Integration 7. The Future of Automation Acknowledgments References

Abstract Throughout the history of single-particle electron microscopy (EM), automated technologies have seen varying degrees of emphasis and development, usually depending upon the contemporary demands of the field. We are currently faced with increasingly sophisticated devices for specimen preparation, vast increases in the size of collected data sets, comprehensive algorithms for image processing, sophisticated tools for quality assessment, and an influx of interested scientists from outside the field who might lack the skills of experienced microscopists. This situation places automated techniques in high National Resource for Automated Molecular Microscopy, Department of Cell Biology, The Scripps Research Institute, La Jolla, California, USA Methods in Enzymology, Volume 483 ISSN 0076-6879, DOI: 10.1016/S0076-6879(10)83015-0

#

2010 Elsevier Inc. All rights reserved.

291

292

Dmitry Lyumkis et al.

demand. In this chapter, we provide a generic definition of and discuss some of the most important advances in automated approaches to specimen preparation, grid handling, robotic screening, microscope calibrations, data acquisition, image processing, and computational infrastructure. Each section describes the general problem and then provides examples of how that problem has been addressed through automation, highlighting available processing packages, and sometimes describing the particular approach at the National Resource for Automated Molecular Microscopy (NRAMM). We contrast the more familiar manual procedures with automated approaches, emphasizing breakthroughs as well as current limitations. Finally, we speculate on future directions and improvements in automated technologies. Our overall goal is to present automation as more than simply a tool to save time. Rather, we aim to illustrate that automation is a comprehensive and versatile strategy that can deliver biological information on an unprecedented scale beyond the scope available with classical manual approaches.

1. Introduction The field of cryo-electron microscopy (cryo-EM) has witnessed an explosion of activity during the recent past with the emergence of increasingly sophisticated, automated technologies for performing the steps required to obtain a refined electron density map. As a result, cryo-EM is rapidly moving from a method practiced by a small group of experts to one that is applicable to a variety of scientific disciplines. Our increased ability to collect, store, and process previously unimaginable amounts of data has not only pushed the limits of questions addressable by the electron microscopy (EM) community, but has also attracted significant interest from experts outside our field, requiring structural insights but lacking the time and resources available to experienced EM practitioners. The success and general scientific credibility of our field will largely depend on how we handle the impending transition from EM as an esoteric methodology to a readily accessible technique in the scientific toolbox. As a research group, our goal has been to develop automated data collection and processing software that aid both expert and novice users by combining automation and streamlined user-friendly interfaces with an underlying transparent framework that is both accessible to neophytes and extendable by seasoned EM practitioners. Our field is in a phase of expansion, and new software packages are being released at a remarkable rate, some contributed by outsiders to the field. It is our belief that the future of cryo-EM lies in bridging the gap between expert users developing sophisticated technologies and novices who benefit from user-friendly automation. We have always acknowledged, however, that “black box” automation of any aspects of cryo-EM is undesirable. The user must maintain control over

Automation in Single-Particle Electron Microscopy

293

automated routines, and a careful balance must be maintained between manual control and automated efficiency so as to pose minimal risks to the versatility, expandability, and quality of existing and future methods. Automation should strive to be more than a mere elimination of repetitive tasks. It should also serve as an aid to expert users, encouraging development of next generation algorithms and technical advances. Our overall focus has been on the construction of a “pipeline” for singleparticle EM to provide a streamlined but transparent route from specimen preparation to the reconstruction of a refined electron density map. In an earlier chapter describing our automated data collection software, we quoted the following definition of automation: “Automation is the use of computers to control machinery and processes, replacing human operators. It is a step beyond mechanization, where human operators are provided with machinery to help them do their jobs. Some advantages are repeatability, tighter quality control, waste reduction, integration with systems, increased productivity, and reduction of labor. Some disadvantages are high initial costs and increased dependence on maintenance” (Suloway et al., 2005). Given our recent focus on automated image processing, we expand on the definition to add that: automation in EM enables the elimination of repetitive, predictable user interactions required to manipulate, process, translate or move data between image-processing modules, using an intuitive mode of user learning. In addressing the problems posed by inter-package incompatibility of data conventions and definitions, we are not simply constructing a de novo pipeline of novel image-processing algorithms. Rather, we are assembling existing modules, interpreting their functionality, and identifying inter-connecting paths, as if putting together pieces of a large and complicated puzzle. As new components are added, the pipeline grows, but not necessarily in a linear fashion. There are many paths to solving structures using EM, and our goal is to supply the user with the means to connect any selected set of pieces together, while also ensuring that interchangeable parts fit together in a seamless way and providing the ability to add new components as needed (Fig. 14.1). Given the breadth of the topic of automation, we have adopted a “problem-centric” rather than “package-centric” approach to our description of the various areas of cryo-EM where automation has a distinct contribution. The manuscript is divided into sections on: (1) sample preparation, (2) robotic loading and screening, (3) electron microscope operation, alignment, and data collection, (4) single-particle image processing, and (5) assessment and integration. We begin each section with a general description of the problem and follow with a summary of the automated, as contrasted to manual, approaches that have been undertaken to address the particular issue, providing references and highlights to specific packages where appropriate. We discuss some of the major advantages and limitations of the existing techniques and hypothesize how future questions or

294

Dmitry Lyumkis et al.

Figure 14.1 Each aspect of cryo-EM can be automated to some degree. Our overall goal is to show that automation can be more than simply a tool to save time. Rather, we aim to present automation as a comprehensive and versatile strategy, one where the overall picture becomes more valuable than the sum of its individual parts. When the pieces are connected, electron microscopy becomes a mature and sophisticated technology that is increasingly capable of addressing complex biological questions.

problems might be addressed via automation. Finally, in the method boxes, we provide specific step-by-step protocols for how these procedures are implemented at the National Resource for Automated Molecular Microscopy (NRAMM). We regret any omissions, errors, or oversights that are almost inevitable in a review of this wide scope.

Automation in Single-Particle Electron Microscopy

295

2. Specimen Preparation Reproducibility is the primary concern of experimentalists, and yet, in the field of single-particle EM, reproducing a good data set is far from trivial. Specimen preparation, still considered an art form perfected by few, presents the major bottleneck to extracting identical structural information from independently prepared biological samples. Poor specimen preparation will adversely complicate data collection, image processing, and interpretation of resulting structures. The grid, of course, is a very large piece of real estate compared to the size of an average macromolecule, and when only a small number of fields needs to be acquired, a well embedded and suitably distributed specimen can likely be found somewhere on the grid, even if at the expense of time. However, the push toward higher resolution and the need to examine conformationally variable macromolecules poses higher demands for the extent of the grid that must be suitable for data collection and encourages the development of more consistent and reproducible specimen preparation methods. Traditionally, heavy metal stains have been frequently used to embed the sample for viewing inside the electron microscope, providing a “footprint” of the specimen on the adsorption substrate. Following optional modification of the substrate’s hydrophobic properties, several microliters of sample are applied to the grid, which is then followed by application of the stain of choice. Specific protocols for negative staining have varied among the EM community (e.g. reviewed in Ohi et al., 2004). Deviations from conventional practices involve differences in the adsorption substrate on the grid (e.g., Quantifoil, lacey carbon, or C-flat holey-carbon grids), the method of substrate preparation (e.g., glow-discharging or plasma-cleaning), and the specifics of the technique for applying the specimen (e.g. single-blot or deep-staining). Much is still a matter of controversy and possibly also depends on the specimen itself. A special protocol for negative staining has been developed that avoids the need for blotting with filter paper by making use of a liquid-handling system to control the application of solutions to the grid (Ku¨hlbrandt et al., 2003) and has enabled the automation of screens for 2D crystallization conditions of membrane proteins (Cheng et al., 2007; Iacovache et al., 2010; Vink et al., 2007). For more

296

Dmitry Lyumkis et al.

general purposes, such systems could be used to investigate the effect of buffer conditions, test different sample dilutions (ideal for a liquid-handling robot with a small well), or collect time points in a time-resolved study. Vitrification has presented more opportunities for automated technology development (Chapter 3, Vol. 481). Since the early days of manual cryo-blotting and freeze-plunging (Dubochet et al., 1982; Taylor and Glaeser, 1976) the entire process has evolved to one that is reasonably controllable and reproducible with automated systems. In contrast to manual methods where frequent practice is important to maintain the skill set, an automated process can be learned quickly and returned to with reasonable facility even after a long hiatus. The first and most widely recognized of the automated grid preparation devices is the FEI Vitrobot, a commercially available robot based on earlier work by several groups (Frederik and Hubert, 2005; White et al., 1998, 2003). More recently, two newer models have come onto the market—the Gatan Cp3 (Melanson, 2009) and the Leica EM GP (Microsystems, 2009). Several homemade automated devices have also been developed, but will not be discussed here due to their lack of public availability. They all perform the same basic operation. The grid is held in place inside an isolated chamber that can be adjusted to a desired temperature and humidity by the user. After sample application, a robotic mechanism applies filter paper to the grid and blots away excess liquid using parameters specified by the user (e.g., blotting time, number of blots, time between blots). Finally, the grid is plunged into a liquid ethane medium at a temperature of about 185 C. This combination of conditions creates vitreous, rather than crystalline, ice and effectively preserves the sample in its natural state by avoiding deformation caused by intruding ice crystals (a comprehensive review on sample vitrification can be found in Dubochet et al., 1988). Each preparation device possesses distinct advantages. For example, Gatan Cp3 allows for one-sided or two-sided blotting. Newer models of the FEI Vitrobot allow for automated transfer of the grid from the vitrification medium into liquid nitrogen. The Leica EM GP is capable of sensor controlled blot timing and promises to provide the most accurate and reproducible sample preparation parameters. When accurately calibrated, these devices all increase the reproducibility of cryo-freezing and can provide grids with large areas of well-embedded particles. One interesting area of recent developments in automated vitrification technologies is time-resolved cryo-EM, wherein reactants essential to a

Automation in Single-Particle Electron Microscopy

297

structural event are sprayed onto the sample within a few milliseconds prior to plunging. This allows for a range of reactants to be delivered as aerosols to the system of interest, whose conformational and/or compositional changes can then be studied. Time-resolved cryo-EM has been successfully used on a number of occasions (Berriman and Unwin, 1994; Walker et al., 1995, 1999; White et al., 1998, 2003), but the intricacy of the technique, and perhaps the lack of an easily controlled automated device, has precluded its widespread use in the EM community. Two major challenges for automation of time-resolved vitrification have arisen. From a mechanical standpoint, there is a need for a system that enables reproducible production of a thin film of primary solution immediately before applying activating solution. From a computational standpoint, additional complexity introduced by compositional heterogeneity in reactant mixing regions places high demands on classification and sorting algorithms in later image-processing steps. To address these needs, a first-generation monolithic device has recently been described that enables mixing of two reactant solutions in a microfluidic channel immediately prior to being sprayed onto a conventional EM grid as it is being plunged into cryogen (Lu et al., 2009). This setup allows for adjustable combinations of reactants and requires no blotting, removing variability in contact between the grid and filter paper. While this device has yet to be commercially manufactured, the preliminary results look promising, making it a likely candidate for future automation. When optimizing specimen preparation, the grid substrate is not only an important factor for embedding the sample but also affects the distribution of the particles and the ease with which we can automate the subsequent stage of data collection. For cryo-EM, samples are typically spread over either continuous or fenestrated carbon substrates. Automated data acquisition requires identifying areas of ice of suitable thickness, either across the continuous carbon or suspended across the holes. Maximizing the efficiency of data acquisition requires that the distribution of particles is neither too crowded nor too sparse, a factor that is influenced by the size of the holes and the thickness of the carbon at the edge of the hole. Methods for producing fenestrated carbon have existed for half a century (Drahos and Delong, 1960; Harris, 1962), but they produce holes in a wide array of sizes and shapes. This poses challenges (although not

298

Dmitry Lyumkis et al.

insurmountable ones) for automatic identification of a suitable area for focusing, drift monitoring, and overcoming charging effects. If holes are arranged in a regular geometry, these issues are more easily addressed. Quantifoil grids (Ermantraut et al., 1998) were the first commercially offered grids with predictably sized and spaced holes. They were rapidly adopted by the community and continue to be widely used. About a decade later, C-flat grids were developed at NRAMM (Quispe et al., 2007) and have been commercialized by Protochips, Inc. The C-flat grids are composed purely of carbon, avoiding the need to remove underlying substrate, and providing a smaller lip around the edge of the hole. These grids are particularly well suited to ensuring an even distribution of small particles embedded in thin ice, which might otherwise be swept to the edge of the hole. More recently, Yoshioka et al. tested a new prototype fenestrated grid from Protochips, Inc., called Cryomesh, which is composed of doped silicon carbide using processes adapted from the semiconductor industry. The improved low temperature conductivity and extreme rigidity of these grids appear to have a significant impact on reducing beam-induced movement in TEM images of tilted cryopreserved samples, and thus has the potential for improving the resolution of three-dimensional (3D) cryo-reconstructions in general (Yoshioka et al., 2010). Similarly, Pantelic et al. (2010) showed that graphene oxide may provide a substrate superior to amorphous carbon. Another interesting class of grids that has emerged, termed affinity grids (Kelly et al., 2008), contains a monolayer of lipids spread over the holes (Chapter 4, Vol. 481). The lipids carry a Ni-NTA head group, thereby providing a rapid and convenient technique for concentrating His-tagged particles out of solution onto the surface of the grid without biochemical intervention. Using a similar strategy, TM Dune Sciences has now started to commercialize SMART Grids with specific types of affinities for various biomolecules. In the near future, we can expect plenty of interesting biological uses arising directly from these developing technologies. We envision that the future of EM sample and grid preparation will follow the path of X-ray crystallography, where robotics and microfluidic devices are increasingly used for the preliminary stages of optimizing conditions for crystal formation. Although a standard protocol can be followed for the majority of analyzed samples (Box 1), improvements can be conceived for various aspects of specimen preparation. For negative stain, robotic machines that precisely monitor (1) the volume placed on the grid, (2) the number of “washes,” (3) the amount of time allotted to its absorbance, and (4) the amount of remaining stain after wicking, can be automated to provide increased reproducibility and efficiency. For vitreous ice, we might optimize the blotting procedure, or develop a new method entirely, to decrease variability in ice thickness across the sample grid. The automated transfer of grids after staining or vitrification would limit the handling (and mangling) of grids and reduce exposure of cryogrids to contamination. Automated or robotic screening procedures (a few of

Automation in Single-Particle Electron Microscopy

299

which will be discussed below) would also facilitate comparisons between different preparation conditions. Box 1 NRAMM Specimen Preparation Protocol

Purpose: Prepare a sample for imaging in the electron microscope. 1. Select the adsorption substrate: While carbon films are the most commonly available substrate, others are gaining popularity (see text). For cryofrozen specimens, the most commonly used substrates are holey-carbon films (typically C-flat or Quantifoils), thin continuous carbon over holey carbon, or continuous carbon. We typically choose to use C-flats, as these often provide the most well distributed particles embedded in large uniform areas of vitreous ice. Thin continuous carbon over holes can be used for small particles that prefer to adhere to carbon, and continuous carbon can be suitable for larger particles and potentially provides for better estimation of the CTF and reduction in beam-induced specimen movement at high tilt. 2. Prepare the adsorption substrate: Prior to sample application the hydrophobicity properties of the adsorption substrate for the specimen must be modified. Plasma cleaning or glow discharging the grid will modify the hydrophilic properties of the substrate, so as to improve buffer dispersion and sample adherence to the grid. We use the Gatan Solarus Plasma cleaner, typically set for 5–10 s, 25% O2, 75% argon. 3. Apply the sample to the grid: The standard protocol is to apply 3–5 mL of sample and allow for 30–60 s adherence time, but these parameters are specimen dependent. Prior to embedding the sample in either heavy metal stain or vitrified ice, filter paper is used to wick away any excess sample. In the case of automated freeze-plunging, this step is performed by the FEI Vitrobot. 4. Embed the sample: The sample is fixed using negative staining or freezeplunging protocols. Automated negative staining protocols are available but are not widely used, and in order to maintain the fidelity of the true structure, cryo-freezing is often preferred. Samples are vitrified using the FEI Vitrobot (typically set to 4C, 95% relative humidity). Our usual protocol is to use a range of blotting times, typically from 3 to 9 seconds, to make approximately four grids. This provides a bracket for ice thickness that frequently produces at least one grid that is suitable for data collection (Fig. 14.2). 5. Store the sample: Negatively stained grids are stored in standard grid boxes in a dessicator. Vitrified grids are stored in smaller four-grid boxes, which are in (continued )

300

Dmitry Lyumkis et al.

Box 1 (continued )

turn placed into storage tubes suspended under liquid nitrogen inside long-term storage dewars.

Figure 14.2 Vitrified sample prepared over C-flat grid, freshly cleaned using Gatan Solarus Plasma cleaner and vitrified using the FEI Vitrobot.

3. Autoloading and Robotic Screening Since the introduction of the electron microscope, physical exchange of specimen grids has been the job of a human operator. For small-scale operations where only a few grids are exchanged in a daily session, automation may not prove more efficient than a human being. However, for large-scale projects requiring routine screening procedures, automation pays off profoundly. This section will focus on technologies that relieve the burden of routine and highly repetitive grid loading and screening procedures. Software-controlled grid-exchange appliances have been developed in both academic and commercial environments. The first, developed at NRAMM (Cheng et al., 2007; Potter et al., 2004), uses a six-axis articulate robot arm mounted separately from the microscope to mimic the human operation of inserting the loaded grid holder through the air lock. In the current design, suction created at the tip of a fine nozzle attached to the same robot arm is used to pick up grids from a preloaded 96

Automation in Single-Particle Electron Microscopy

301

grid tray and load them into the specimen holder aided by an actuator to raise and lower the specimen holder lever arm. A second implementation, developed at the New York Structural Biology Center (NYSBC) shares the same principle but uses a Cartesian robot (which moves back and forth along a single, fixed axis) for the task of inserting the loaded grid holder and a Selective Compliant Assembly Robot Arm (SCARA, limited to motion within a defined xy-plane) for loading grids from the grid tray to the holder (Hu et al., 2010). Some advantages of subdividing the grid loading and holder insertion tasks include a decrease in the requirements for robotic alignments, an increase in fine-tuned motion specific to EM grid loading, and an overall decrease in cost. The grid pickup mechanism on this robot has additionally evolved to consist of three spaced-out vacuum nozzles instead of one. A third implementation (Lefman et al., 2007), developed by Gatan Inc., mounts its grid tray and insertion mechanism directly on the microscope column and under vacuum. The grid “tray” is a transfer cylinder in a Gatling gun design where grids are preloaded in individual cartridge-type grid holders, effectively bypassing the air lock entrance and exit for individual grids. The fourth implementation uses a modified FEI “autoloader” that can accept, from a cylindrical grid reservoir, one of the 8 Titan autoloader cassettes with pre-loaded grids (Coudray, et al., 2010). It can then use the mechanism of the autoloader to put the grid into the imaging path. All four systems are used exclusively at room temperature. For cryo-specimens, JEOL microscopes currently provide side-entry grid holders and cartridge-based cryo-grid loaders that allow for their exchange in and out of the imaging path without operating the air lock. The user loads a cryo-grid cartridge, thereby eliminating possible atmosphere-based grid contamination during transfer. However, the task of physically exchanging grids still requires manual mechanical operations. Two instruments developed by FEI (the Polara and the Titan Krios) allow for the exchange of multiple grids, preloaded into cartridges, using a specially designed cryo-grid loader controlled by a software interface. This is a highly beneficial development for improving the throughput when acquiring large cryo-data sets, where multiple grids might be used during the collection process. The current alternative—individual grids mounted in manually inserted side-entry stages—is slow, inefficient, and leads to contamination of the grid and the microscope vacuum. While the cost for the current generation of automated cryo-grid exchange instruments is extremely high, the obvious advantages should lead to further developments in this area. Grid-exchange robots, when effectively coupled to automated techniques for target selection and image acquisition, can drastically increase the throughput of screening projects. To date, the only processes that have been set up are for performing initial screens of 2D crystallization conditions for

302

Dmitry Lyumkis et al.

membrane proteins (Cheng et al., 2007; Vink et al., 2007). The Leginon package (Suloway et al., 2005) provides an application for communicating with a robotic specimen loader and randomly sampling selected areas of the grid without requiring user intervention (Box 2). Apart from its usefulness in large screen-based studies, the large number of high magnification images output by the application can subsequently focus data collection to a specified region. Simple image-processing calculations can be employed to determine, in a more quantitative manner, optimal imaging conditions. For example, regions exhibiting variations in ice thickness, tendency to specimen drift, or impurities are quantifiable data that have already been extensively analyzed with simple algorithms (Stagg et al., 2006). When coupled to “smart” screening protocols that learn to distinguish different conditions, we could limit image acquisition to optimal areas of the grid, thereby systematizing data collection and reducing the amount of micrograph post-collection processing. Provided that we can improve automated loading and screening procedures, such that these operations could be performed quickly, efficiently, and routinely, we would open up opportunities for conducting experiments that would otherwise remain undone. For example, development of robotic devices and screening protocols will greatly facilitate the examination of more heterogeneous samples in a high-throughput fashion, perhaps with the intent of examining the effects of different conditions on specimen heterogeneity. Such screening capabilities, coupled with finely tuned time-resolved methods could be used for increased resolution and efficiency in large-scale analyses (Mulder et al., 2010). Whole cell and cell section studies might benefit from the decrease in operator time required for finding the optimal area of the grid for imaging structures of interest in electron tomographic experiments. Given the increasing role that EM is playing in addressing various biological questions, we can envision that the number of laboratories requiring automated screening procedures is likely to increase in the future, which should drive the development of commercial systems to facilitate these applications. Box 2 NRAMM Autoloading and Robotic Screening

Purpose: Utilize robotic grid insertion with an automated screening application to systematically survey multiple areas on different grids. At NRAMM, this procedure is largely limited to negatively stained grids but the protocol and software can be easily extended to enable microscopes offering cryo-grid exchange to perform iterative screening protocols (Fig. 14.3).

Automation in Single-Particle Electron Microscopy

303

Figure 14.3 Top: Grid tray loaded with 96 grids. Bottom: The robotic arm is mounted from the ceiling next to the Tecnai F20 electron microscope.

1. Prepare the sample tray: at NRAMM, up to 96 individual grids are placed onto a tray. Without a mode for preserving the controlled environment necessary for cryo-frozen specimens, this procedure is largely limited to negatively stained grids. The MSI-Robot application available in Leginon will enable microscopes offering cryo-grid exchange to perform iterative screening protocols. 2. Calibrate the insertion robot: when the grids are not preloaded into a specially designed cartridge, the robotic mechanism for insertion requires proper calibration. Allowable errors in translational movement vary with the specifics of the robot design. (continued )

304

Dmitry Lyumkis et al.

Box 2 (continued )

3. Test grid insertion: it is crucial to make sure that the first grid insertion by the robot proceeds smoothly, ensuring that subsequent grids do not damage the microscope when the human operator is absent. 4. Establish Leginon/robot communication: Start Leginon and start robot server. 5. Start the MSI-Robot 1st Pass application: this Leginon application allows the user to systematically screen multiple areas on different grids. 6. Select the grids: this is carried out through the “Robot” node within Leginon. 7. The robot inserts the first selected grid before returning the control back to Leginon. 8. Obtain an atlas: a comprehensive illustration of the specimen grid can be obtained within the “Grid_Survey_Targeting” node. Modifying the settings will determine the number of acquired fields for constructing an atlas, and thus the area to be screened for the given grid. 9. Obtain higher magnification images: the procedure for obtaining highmagnification images is analogous to standard Leginon protocols (see Box 3). The “Square_Targeting” node samples grid squares, while the “Mid_Mag_Survey_Targeting” node specifies the areas for surveying the grid through higher magnification images. 10. Robot returns the grid to the tray. 11. Repeat steps 7–10 for remaining grids.

4. Microscopy Traditional methods for collecting micrographs from a TEM require careful hand-eye coordination—the operator views the image on a phosphor screen or video display and makes adjustments to parameters by turning knobs and pushing buttons. In this way, areas of interest on the grid are selected and the imaging conditions calibrated using adjustments to stage position, magnification, defocus, and other microscope parameters. Once the conditions are acceptable, a final micrograph is recorded to film or a digital camera. These manual procedures have several drawbacks. For one, visual surveys of a specimen grid can be tedious, and impractical when acquiring thousands of micrographs. Additionally, as a general rule for radiation-sensitive specimens, searching for suitable target areas using the

Automation in Single-Particle Electron Microscopy

305

same microscope parameters as used for the final exposure will subject the sample to excessive radiation. This was realized early on with tomographic data collection, where automation is critical to reducing the overall specimen radiation. Computational approaches have since greatly increase the efficiency of the process (Chapter 12, Vol. 481). Single-particle data collection, though much simpler than tomography, can benefit from similar automation. Prior to the development of automated image acquisition packages, the first protocols for automation focused on electron beam calibrations, with particular emphases placed on the ability to set the defocus, correct astigmatism, and achieve coma-free alignment of beam tilt. Several practical procedures have been implemented for determining these three image aberrations (Kimoto et al., 2003; Koster et al., 1987; Saxton et al., 1983; Typke and Dierksen, 1995; Typke and Ko¨stler, 1977; Zemlin, 1979; Zemlin et al., 1978). The crucial findings impacting automation protocols involved the realization that all three aberrations can be efficiently determined by examining either diffractograms of the images (Typke and Ko¨stler, 1977; Zemlin, 1979) or the images themselves (Koster et al., 1987) using induced beam-tilt image pairs. The latter idea was further clarified to describe the calibration matrices for measuring defocus, astigmatism, and beam-tilt misalignment using beam-tilt-induced image displacement (Koster and de Ruijter, 1992) and quickly adopted in a number of early automation packages for defocus measurements (Dierksen et al., 1992, 1993; Fung et al., 1996; Kisseberth et al., 1997; Rath et al., 1997). Today, as the intended resolution during data collection increases, these established methods for microscope alignments remain central to defocus measurements and astigmatism correction, but the push for near-atomic resolution structures has demanded further improvements. Coma-free beam-tilt alignments, which primarily impact high-resolution information captured on the micrograph, have become the next area of development. To our knowledge, only the current release of the Leginon package (Suloway et al., 2005) has incorporated a fully automated procedure for correcting axial coma, based on the diffractogram tableau analysis outlined by Zemlin (1979). Automation packages have been established for acquiring tilt-series images in both academic (Marsh et al., 2007; Mastronarde, 2005; Nickell et al., 2005; Suloway et al., 2009; Zheng et al., 2007a,b, 2009) and commercial releases (software from FEI, JEOL, Gatan, TVIPS, and others), for acquiring random-conical tilt (RCT) (Zheng et al., 2007b) or orthogonal tilt (OT) image pairs (Yoshioka et al., 2007), for screening 2D crystallization conditions (Cheng et al., 2007; Coudray et al., 2008; Hu et al., 2010),

306

Dmitry Lyumkis et al.

for acquiring montages of serial sections of biological tissues (Cardona et al., 2010; Mastronarde, 2005), and for routine single-particle, helical, and 2D-crystal analyses employing either single or multiple exposures at a fixed tilt angle (Carragher et al., 2000; Lei and Frank, 2005; Marsh et al., 2007; Potter et al., 1999; Shi et al., 2008; Suloway et al., 2005; Zhang et al., 2009, 2001, 2003). For tomography and RCT/OT applications, the need for acquiring multiple images of the same final targets under defined, but different microscope parameters requires, by itself, a high level of automation for tracking features and defocus. Untilted data collection packages have expanded upon the ideas initially developed for tomography with improvements to methods for selecting and locating specified targets. By examining the approaches enabling low-dose targeting, we can divide the software developed in the last ten years into three distinct levels. At the lowest level, the software provides multiple user-defined microscope states for the different tasks that need to be performed, such as search, focus, and exposure. Target selection and centering are performed by direct user feedback while observing the image on the microscope viewing screen. Microscope manufacturers routinely provide this level of automation in the form of a low-dose kit. With CCD cameras gaining popularity, several packages also provide the option of saving the acquired digital images (e.g., EMmenu from TVIPS, Digital Micrograph from Gatan, or TEM Imaging and Analysis (TIA) from FEI). The JAMES package (Marsh et al., 2007), developed specifically to unify software from JEOL microscopes and Gatan CCD cameras into a single Python integrator, provides a few additional capabilities for directly communicating with the microscope. At the next level of automation, targets are selected on a CCD image acquired at a low magnification and the stage is driven to the target location using a transformation calibration, as in SerialEM (Mastronarde, 2005), TOM (Nickell et al., 2005), and SAM (Shi et al., 2008). However specimen drift, common with cryosamples, is not monitored, thus presenting the possibility that selected targets will have moved by the time the final high magnification image is acquired. Algorithms for drift monitoring or compensation improve the likelihood that targets selected at low magnification will be centered in the field of view during the final high magnification image acquisition. AutoEM (Lei and Frank, 2005; Zhang et al., 2001, 2003) monitors drift, while JADAS (Zhang et al., 2009) compensates for it using a piezo drive that ensures greater precision in stage movement. Even with these additional features, targeting algorithms based purely on a transformation matrix cannot predict long-term changes and, thus, are only reliable for immediate selections on a lower magnification parent image. At the highest level of automation,

Automation in Single-Particle Electron Microscopy

307

individualized adjustment for each target allows the user to queue up a large set of targets and extend data collection over many hours or even days. For example, the user might queue up several low magnification squares on the grid that contain multiple high magnification acquisition targets. If, during the processing of the first square, the software detects specimen drift, it measures the exact distance by which the specimen has shifted, and adjusts all low and high magnification targets by that value. This advantage enables batch queuing and processing, as is done in Leginon (Suloway et al., 2005), UCSFTomo (Zheng et al., 2009), and FEI tomography. Multiple-scale adjustment also allows UCSFTomo to return to the same area after an inplane rotation of the stage and for Leginon to return to the same target after reinsertion of a grid by a grid-loading robot (Cheng et al., 2007). At the current state, automation in image acquisition has relieved users from many tedious tasks such as estimating the defocus, tracking micrograph features, and monitoring specimen drift. In many cases, it has also improved the accuracy of targeting and reduced the accumulated dose on the specimen. However, one cannot argue that the power of automation has offset the importance of specimen and grid preparation, nor that automation can replace a good operator in controlling the quality of the final images. For one, most current packages lack controlled feedback loops for target selection, which would aim to emulate the sophistication and reliability of a human brain in adjusting the targets selected at low magnification based on an analysis of the quality of the images acquired at high magnification. Second, the high level of automation in packages such as Leginon, comes with the price of demanding both accurate and precise microscope alignments in order to maintain consistent calibrations. Effectively, the sensitivity of the performance of automated procedures to microscope alignments and calibrations often requires the user to possess significant knowledge about these issues. With the development of algorithms that automatically align and calibrate the microscope using digital analysis, one can envision surpassing these obstacles. Thus far, only beam shift, rotation center, and coma-free beam tilt can be adjusted by the click of a button in Leginon. Other alignment routines are limited by the availability of adjustment functions supplied by the microscope manufacturers, and, crucially, quality control feedback loops. For example, the task of moving the stage to a target of interest currently requires the operator to interrupt the process for recalibration if the target is not properly centered in the field of view. With feedback loops, errors in the acquisition process could be detected automatically and invoke the required calibrations. In its optimal form, automation software should perform a thorough alignment and calibration inspection before, and periodically during, data acquisition, and then optimize alignments as needed. Such automated monitoring would also provide an objective log of the exact parameter settings at the time, which can be used to reproduce the behavior of the microscope in a future session.

308

Dmitry Lyumkis et al.

We implemented much of our vision for automation when we began redesigning our original Leginon package in 2002 (Suloway et al., 2005) (Box 3). We believe that a generalized image acquisition automation package should be: (1) extendable to new functionality and applications, (2) flexible for different imaging and specimen conditions, (3) portable for use with different microscopes, cameras, grid-exchange robots, and novel devices, and (4) tightly integrated with a database to allow for data management and tracking. For extendibility and flexibility, Leginon is written using object-oriented modules called nodes that are combined into applications. Individual users can customize their specific sequence of operations (e.g., different methods of focusing) through a graphical user interface, and are always provided with the option of stepping down from full automation, when desired. A similar concept is implemented in JADAS (Zhang et al. 2009), where a recipe gathers modular operations and conditions to complete a series of image acquisitions and target selections, and TOM (Nickell et al., 2005), which is created under the toolbox concept in Matlab. For portability, Leginon employs the pyScope python extension—a thin wrapper around an application-based programming interface distributed by the device manufacturers. Leginon can adapt to new microscopes and cameras by modification of a small set of functions contained within the pyScope library. The software is in use on FEI Tecnai, Polara, Titan Krios, and CM series platforms as well as selected JEOL instruments (Hu et al., 2010), and supports a variety of digital cameras (manufactured by TVIPS, Gatan, and FEI). The extensive use of a relational database, together with a web-based data viewer has made it possible for us to promptly locate any of the more than 1 million images acquired by Leginon since we began data collection in 2003. Each is associated with metadata that provides a comprehensive summary of the imaging conditions and related parameters at the time. The database and web-based infrastructure developed for Leginon has subsequently been extended to the construction of our image-processing pipeline, Appion (Lander et al., 2009), which provides full processing capabilities to any session at the microscope. Box 3 NRAMM Automated Data Acquisition

Purpose: Collect an untilted data set using Leginon (Fig. 14.4). 1. Align the microscope: Leginon employs beam-tilt-induced image shift for several operations, and thus relies on a well-aligned microscope. Prior to initializing Leginon, it is necessary to (a) calibrate the electron gun, (b) align and stigmate condenser and objective lenses, (c) calibrate the electron beam, and (d) identify the focal plane. 2. Initialize Leginon: Make sure that Leginon Client is connected to the microscope.

Automation in Single-Particle Electron Microscopy

309

Figure 14.4 Leginon Layout. (top) Graphical User Interface for the Leginon software package. Each number refers to the description in Box 3. (bottom) Magnified view of the “Exposure_Targeting” node, displaying the log file, image statistics, the hole-finder routine, and the hole image with chosen targets.

3. Import or manually define magnification presets: For standard data collection, we use the “gr” preset for 200 magnification, the “sq” preset for 500 magnification, the “hl” preset for 5000 magnification, the “fa” preset for autofocusing at 50,000 magnification, and the “en” preset for acquiring final images at a specimen-dependent magnification. (continued )

310

Dmitry Lyumkis et al.

Box 3 (continued )

4. Align image shift between magnifications: Adjustments to beam-tilt-induced image shift between different magnification settings can be done through Leginon with the “align presets to each other” function within the “Presets_Manager” node. This ensures that, regardless of the magnification, stage-movement coordinates correspond to user-defined targets from a parent image. 5. Acquire dark and bright image references: Baseline standards for electron counts with “en” images can be set within the “Correction” node. 6. Determine the electron dose: This is done within the “Presets_Manager” node using an empty area of the grid. 7. Acquire an atlas: The atlas represents a composite view of the entire grid and is acquired using the “Grid_Targeting” node. It is critical to make sure that the objective aperture is removed for this step, so that it does not obstruct the view for lower magnification images. 8. Select a square: Within the “Square_Targeting” node, the user can specify square targets. While an automatic square detection algorithm is available, it is usually faster to perform this manually, as only a limited set of squares will likely be used for data acquisition. 9. Select a hole: The image of the selected square will appear within the “Hole_Targeting” node. In an analogous procedure to square targeting, the user can queue targets at this magnification and select a target for automatic high magnification focusing, along with a focus point for setting the Z-height of the stage. Alternatively, the hole-finder algorithm guides the user through setting up an automated hole-detection procedure. 10. Select high magnification images: The image of the selected hole will appear within the “Exposure_Targeting” node, where final acquisition images are selected. A single autofocus image is usually defined for all acquisition images within this node. As in the “Hole_Targeting” procedure, an automated hole-detection protocol can be used here. 11. Monitor initial data collection for accuracy: In the initial stages of data collection, it is important to make sure that every procedure executes in the intended manner. For example, monitoring the “Z_Focus” node, the “Focus” node, and the “Drift_Monitor” node is critical to make sure that all automated algorithms are proceeding smoothly. It might also be necessary to monitor the accuracy of the hole-finder, and adjust if necessary. When this has been ascertained, Leginon can be left unaltered until all targets have been acquired. 12. Refill liquid nitrogen: For cryo-specimens, liquid nitrogen must be added to the stage every 3 h. The cold trap can be refilled at the same time or on a 6 hourly schedule. For negative stain specimens a specially adapted cold trap is available that can last for 12 h or an autofill system can be setup.

Automation in Single-Particle Electron Microscopy

311

5. Image Processing A question often faced by single-particle electron microscopists is: what is the best manner in which to process large amount of data? In our lab, rather than relying on a single package, we have developed Appion, a wrapper software for high-throughput image processing in single-particle EM (Lander et al., 2009). Appion incorporates individual routines from a number of packages and integrates them into a standardized Python-scripted wrapper. It is intimately connected to a graphical web-based user interface, as well as a centralized SQL database, which stores all input and output parameters and enables complex queries of the metadata. Our goal in developing Appion is to provide a streamlined, intuitive, integrated, and transparent process, to thoroughly monitor and track relevant parameters and results associated with processing runs, and to enable users with various levels of expertise to run procedures and analyze the outcome without the need to decipher myriads of cryptic data files (see Box 4 for a conventional protocol for obtaining a refined electron density map using Appion). This transparency is fundamentally important for the advancement of automated EM techniques, but it remains subordinate to the actual content underneath, which consist of image-processing and computational algorithms that were developed by hundreds of scientists in the field (Baxter et al., 2007; Crowther et al., 1996; Frank et al., 1996; Heymann, 2001; Heymann and Belnap, 2007; Heymann et al., 2008; Hohn et al., 2007; Lander et al., 2009; Ludtke et al., 1999; Marabini et al., 1996; Sorzano et al., 2004; Tang et al., 2007; van Heel and Keegstra, 1981; van Heel et al., 1996; also see Chapter 15, Vol. 482). In this section, we describe the automated advancements dealing with the some of the basic areas of singleparticle image processing. Additionally, we highlight recent advances in making each particular task more transparent, integrated, and streamlined, thereby contributing to the overall usability of the separate operations. Box 4 NRAMM Image Processing

Purpose: Reconstruct a 3D electron density map in Appion using a streamlined protocol (Fig. 14.5). Appion presents the user with a menu of options for image processing that is dynamically updated as each step is completed. When the user clicks on one of the menu options, Appion generates a new web page specific to the selected operation that requests inputs and allows the user to launch jobs on one of several processing machines or clusters. The job progress is monitored by updates to the menu. Once a completed job shows up in the menu, the user (continued )

312

Dmitry Lyumkis et al.

Box 4 (continued )

may click on its entry to generate a web page that reports on the results. Most input options are provided with defaults. Help options for each input are provided as pop-ups on the Appion web pages. Detailed step-by-step instructions for most of the procedures are available within the Appion documentation.

Figure 14.5 Appion Webpage Layout (A) Immediately above images in the imageviewer are tools for viewing particle picks, FFT and ACE results, downloading images in jpeg, mrc, or tiff format, making on-image measurements, and sorting images using the “hide” and “exemplar” buttons. All processing pages contain (B) a header with project name, session name, and file path information and (C) the Appion sidebar with available processing functionalities. Clicking on a processing functionality opens a web-based form for launching jobs. (D) General information and parameters that must be changed by the user are displayed on the left hand side, (E) whereas more specialized input parameters are displayed on the right hand side with default values specified. (F) Buttons at the bottom of the form will show the command or launch the job from the web page. (G) Citations are provided for the software packages and/or algorithms contained within a particular functionality. (H) Running, complete, and queued jobs can be tracked within in the Appion sidebar, and job log-files are accessible by clicking on the “# running ” link.

Automation in Single-Particle Electron Microscopy

313

1. View the raw micrographs: Images may be acquired using Leginon or uploaded to the database using the “Upload images” functionality in Appion. Once images are uploaded they are automatically tracked by the database and can be viewed using a web-based Imageviewer. Clicking the “processing” button in the Imageviewer takes the user to the main processing page of Appion. 2. Select particles: Appion provides several methods: “DoG Picker” is a reference-free approach, “Template Correlator” uses the reference-based approach in FindEM, and “Manual Picker” provides for interactive particle picking by the user. An example workflow for particle selection includes: (a) Selecting a subset of particles via DoG Picker or Manual Picker, (b) Aligning and classifying the subset to produce class averages, (c) Using selected class averages for Template Correlator. Users always have the option of cleaning up picks from any of the automated picking runs using Manual Picker. Overall results are provided on Appion summary pages or can be viewed overlaid on the individual images in the Image viewer by selecting the “P” button. As with CTF estimation, particle picking can be started concurrently with image acquisition. 3. Estimate the CTF: ACE, and ACE 2 provide fairly robust algorithms forCTF estimation on untilted images that generally require no adjustments to the provided default settings. CTFTilt can be used for CTF estimates on tilted micrographs. A summary of results can be viewed by clicking on the “complete” CTF items in the Appion menu. Results for individual images can also be viewed on the Imageviewer pages, by selecting the “ACE” button. Individual results include graphical overlays, estimated parameters and associated fitness values. We generally find that fitness values of > 0.8 are acceptable. CTF correction can be started concurrently with image acquisition; the estimation program, once started, will keep querying for new images as they come in. 4. Create a particle stack: The “Stack Creation” page is used to extract particles from the images based on the picks from a particle-selection run, or the picks associated with a previously created and modified stack. Inputs include options for filtering, binning, CTF correction, etc. Particles can be rejected based on CTF fitness parameters, particle correlation values, or defocus range. Results pages provide a summary of the stack, a link to view the stack as individual particles, and further options to clean up the stack using a variety of filters. 5. Align the raw particles: Reference-free procedures include Xmipp maximum-likelihood and SPIDER reference-free alignment, which can be used to create references for subsequent reference-based alignment. Reference-based procedures include Xmipp reference-based maximum likelihood, SPIDER multi-reference, IMAGIC multi(continued )

314

Dmitry Lyumkis et al.

Box 4 (continued )

reference, and EMAN multi-reference alignments. Most procedures can be run initially using default input parameters. The aligned particles can be examined and manipulated further from the summary pages. 6. Classify the aligned particles: As with the alignment routines, Appion provides several different options of classification that can be applied to any alignment run. These include: SPIDER Correspondence Analysis, IMAGIC Multivariate Statistics Analysis, or the Xmipp Kerden Self-Organizing Map routine. The specified feature analysis routine locks the user into a clustering procedure, which generates summed class averages. Class averages can be viewed and manipulated further from the summary pages after requesting “View montage as a stack for further processing.” Options include viewing the raw particles associated with each class, creating templates or substacks from selected classes or running common-lines procedures to create an initial model form selected classes. 7. Generate an initial model: Many options are available. Models can be uploaded from the PDB or EMDB, read in from a file, or imported from previously reconstructed datasets. If tilted data has been acquired, it is possible to perform “one-click” RCT or OTR, and tomographic reconstructions from selected 2D class averages or Z-projected subtomogram averages. These options are presented when viewing the stacks or class averages of appropriate datasets. Other options include common-lines approaches either utilizing EMAN’s cross-common-lines protocol or an automated version of IMAGIC’s angular reconstitution. These options are available when viewing class averages or from the ab initio reconstruction menu option. 8. Refine: Options include procedures using EMAN1, Frealign, or Spider application packages. Results can be viewed on summary pages, which includes more detailed information for each iteration, such as data and graphical output for Resolution curves, Euler angle distributions, snapshots of 3D maps, class averages, and particles contributing to the map.

5.1. CTF estimation and correction Image formation in EM is distorted by the modulation of a contrast transfer function (CTF), which affects the amplitude of the Fourier coefficients at all spatial frequencies of the electrons contributing to the projection of the object. The distortion depends upon the physical parameters of the electron microscope, such as accelerating voltage, lens aberrations, and, crucially, the displacement of the focal plane at the time of imaging. Correcting for this aberration is critical for high-resolution reconstructions. It has long been

Automation in Single-Particle Electron Microscopy

315

known that power spectra representing the spatial frequencies present in an electron micrograph can be used as robust estimators for the CTF (Welch, 1967), allowing the user to manually generate a theoretical CTF with the intent of finding the highest correlation to the experimentally observed power spectral density (PSD). This task, however, becomes burdensome, errorprone, and simply impractical when very large numbers of micrographs are collected on a daily basis, when Thon rings in the PSD are not clearly delineated, or when the presence of astigmatism or stage tilt introduces additional unknowns into the equation. The requirement of accurate CTF correction for high-resolution reconstructions has encouraged a number of automated approaches to CTF estimation. In its most fundamental form, algorithms for defining a theoretical CTF model require optimizing some correlation function between a predicted and an experimentally observed PSD for each electron micrograph (Fernandez et al., 1997; Huang et al., 2003; Mallick et al., 2005; Mindell and Grigorieff, 2003; Sander et al., 2003; Sorzano et al., 2007; Velasquez-Muriel et al., 2003; Welch, 1967; Zhou et al., 1996). Any optimization of an automated CTF estimation algorithm must focus on minimizing estimation errors inherent in the algorithm itself or the input data. Such errors can result from the low signal-to-noise ratio of cryo-micrographs (Baxter et al., 2009), inconveniently located local minima in CTF estimation functions, and cross-talk between different CTF parameters that lead to identical results. Recently, Sorzano et al. published a study in which a comprehensive mathematical description was devised to analyze the effects of different CTF parameters and corresponding estimation errors (Sorzano et al., 2009b). In doing so, the authors provide a quantitative model which describes the way in which errors in the input parameter values translate into errors in the output values (i.e., parameter sensitivity). Studies like this pave the road for future improvements in automated CTF estimation algorithms, demonstrating the need for reliable quality control criteria that take into account parameter sensitivities, lens aberrations and astigmatism, and low SNR inherent to cryo-micrographs. Comprehensive confidence scores, rather than simple correlation values between an experimental and a theoretical PSD, provide a more robust metric for assessing the accuracy of CTF estimation. They can be used to not only systematize a challenging manual task, but to accelerate the pace at which beginning stages of data processing can be completed.

316

Dmitry Lyumkis et al.

5.2. Particle selection and stack creation Ideally, the specimen of interest resides on the grid in a uniform and homogeneous array. However, we can think of many examples that deviate from this scenario. These might include structural specimen heterogeneity arising from compositional or conformational variations, foreign substances resulting from inadequate specimen purification, specimen aggregation, or artifacts and contaminants arising from the preparation procedures. The outcome is a conglomeration of “stuff,” which must be sorted to identify the particles of interest while keeping the false positive and false negative detection rates as low as possible. Once factors such as variation in ice thickness (and thus particle contrast), particle orientation, specimen heterogeneity, and the sheer number of potential picks are considered, the task of distinguishing the “good” from the “bad” becomes far from trivial. Depending on the stage of image processing, different approaches can be taken to automate particle selection. Particle selection algorithms can be loosely classified into reference-free or reference-based categories, with an occasional hybrid of the two (for a review see Zhu et al., 2004). In working with a novel sample, it may be wise to choose a reference-free algorithm wherein particles are segmented based on a measure of “saliency,” such as variance, Markovian fields, or segmented edges. Such approaches (Ogura and Sato, 2004a; Plaisier et al., 2004; Singh et al., 2004; Umesh Adiga et al., 2004; Voss et al., 2009; Zhu et al., 2003) need not necessarily be templatefree, and several of the algorithms effectively apply a template-matching strategy, in which the template is generated without a priori knowledge of the sample. An alternative strategy is to supply a reference based on some prior structure (Huang and Penczek, 2004; Rath and Frank, 2004; Roseman, 2004; Wong et al., 2004). In this case, rather than detecting salient features, the primary task is to compute an optimum similarity measurement between the supplied reference and all possible particle orientations, a procedure that works well for later stages of processing where the primary structures and directional views have been established and the goal is to optimize the raw particle stack for refinement. A third approach combines reference-free and reference-based methods, whereby the algorithm “learns” which particles to select in a supervised manner (Hall and Patwardhan, 2004; Mallick et al., 2004; Ogura and Sato, 2004b; Plaisier et al., 2004; Short, 2004; Sorzano et al., 2009c; Volkmann, 2004). Naturally, combinations of these algorithms are often employed. For example, one can

Automation in Single-Particle Electron Microscopy

317

begin with a reference-free approach, then align and classify the particles to obtain templates for subsequent rounds of reference-based picking. The crucial step lies in careful visual examination of the actual picks in order to assess the algorithm’s applicability to the particularities of the data. Classification schemes (Chen and Grigorieff, 2007; Shaikh et al., 2008), filtering tactics employing “trap templates” for false-negative detection (Chen and Grigorieff, 2007), multistage picking classifiers (Sorzano et al., 2009c), and other strategies often achieve equivalent results to rigorous manual optimization of the selected particles. On the other hand, one might want to completely eliminate bias by utilizing a reference-free approach to pick any specimen-resembling species, as was successfully performed by Mulder et al. to collect over a million ribosomal particles at various stages of assembly (Mulder et al., 2010)—an inconceivable task without automated particle selection. Sometimes, therefore, the best strategy might be to simply pick everything.

5.3. 2D alignment and classification In order to extract quantitative information out of the inherently low SNR data obtained by cryo-EM, 2D averaging must be applied to homogenous subsets of single particles. This requires the particles to be brought into alignment with one another, so that the signal of common motifs is amplified. Alignment protocols typically operate by shifting, rotating, and mirroring each particle in the data set in order to find the orientation of particle A that maximizes some similarity function (of which many variations exist; Chen and Grigorieff, 2007; Roseman, 2004; Saxton and Frank, 1977; Stewart and Grigorieff, 2004; van Heel et al., 1992) with particle B (reviewed in Frank, 2006b; Joyeux and Penczek, 2002). As with particle-selection algorithms, alignment procedures are separated into reference-free and reference-based approaches, depending upon the existence of a template obtained from a priori information about the specimen. Within automated alignment procedures the “do loop” has long freed the user from manual calculation of similarity functions and tracking of parameters for thousands of images, thereby setting the scene for iterative alignments, wherein vectorial addition of parameters is tracked and applied to each particle after all iterations. Despite their utility for fine comparisons, the iterative strategy is sometimes hindered by over-fitting of noise, which presents a critical drawback and

318

Dmitry Lyumkis et al.

requires particular scrutiny when dealing with small or conformationally heterogeneous specimens (Stewart and Grigorieff, 2004). Maximum-likelihood algorithms (Scheres et al., 2005b; Sigworth, 1998; Sjors and Sigworth et al., this volume) have gained popularity with numerous incorporations into the Xmipp processing package (Marabini et al., 1996; Scheres et al., 2008; Sorzano et al., 2004). These techniques integrate the two disparate branches of alignment by iteratively extracting references directly from the data without any a priori knowledge. They create a formalism for the task of aligning images to this set of references that includes estimates for the noise and angular distribution of particles and have shown the ability to overcome false minima and reference bias (Sigworth, 1998). These methods also require significantly more computing power that can be partly addressed by reducing the search space over the alignment parameters (Scheres et al., 2005a), but in the end relies upon advances in computing and processing power. Following alignment, the particles require grouping into classes in accordance with compositional, conformational, and orientational similarity. Classification separates a heterogeneous data set into homogenous subsets of classes with minimal intraclass variation based on a pixel-bypixel comparison of its members. If a priori knowledge about the specimen includes a reference, supervised classification allows grouping of data in accordance with similarity to the references (van Heel and StofflerMeilicke, 1985). However, in cases where a reference is not available, unsupervised classification is required. Given the massive amount of information contained in large data sets, classification programs typically make use of data reduction techniques before proceeding with a mathematical description of inter-image variance. There are many variations on the initial approach proposed by Marin van Heel and Joachim Frank (Frank and van Heel, 1982; van Heel, 1984; van Heel and Frank, 1980, 1981, reviewed in Bonnet, 1998; Bonnet, 2000; Frank, 2006b), but all are conceptually similar to principal components analysis (PCA) (Pearson, 1901), a technique that has become the ideal basis for algorithms dealing with complicated image analysis such as facial recognition. Available classification modules have, for the most part, automated data reduction and the description of inter-image variance, requiring little to no manual input from the user and are available in most processing packages. After reducing the dimensionality, particles are ordered and summed according to their relative proximity within the reduced multidimensional space, traditionally using k-means or hierarchical clustering methods (Frank et al., 1988). Other classification techniques have included “fuzzy” definitions of class memberships (Carazo et al., 1990; Scheres et al., 2005b) and neural networks, which find patterns in a data set through a learning process and heuristically sort images onto a 2D self-organizing map (SOM) (Marabini and Carazo, 1994; Pascual-Montano et al., 2001). All of these methods still require manual user

Automation in Single-Particle Electron Microscopy

319

input to specify the number of desired classes. One promising recent development uses an algorithm entitled affinity propagation (Frey and Dueck, 2007) that no longer requires this specification and has been implemented for single-particle analysis within the latest version of Appion (Lander et al., 2009). Existing alignment and classification modules have done much to mathematically describe variance within the data, but have not always succeeded in placing identically oriented single particles into homogeneous classes using automation. In theory, this means eliminating rotational and translational errors during alignment, and minimizing intraclass variance while maximizing interclass variance during classification. Currently, this combinatorial procedure remains the manual task of the user, who might iteratively reclassify the aligned stack using modified parameters, carefully inspect and subclassify distinct classes to account for additional variations, or remove and realign raw particles based on classification results. Efforts toward aiding the user in the decision-making process have been made in the form of computational methods for estimating alignment errors (Baldwin and Penczek, 2005) and graphical user interfaces that display information relevant to quality assessment (Lander et al., 2009), such as spectral signal-to-noise ratio or Fourier ring correlation values, correlation distributions, variance analyses, and comprehensive particle tracking, which permit “one-click” particle manipulations based on class affiliation. Ideally, these decisions can be encoded into algorithms that would intelligently seek the most likely set of outcomes.

5.4. Ab initio reconstructions With an absence of prior knowledge about specimen structure, the precise orientation of the object with respect to a stationary camera remains unknown. This is one of the central problems in EM and poses the question— how do we determine the correct orientational Euler angles to each recorded 2D projection and combine these into a 3D reconstructed map? It is interesting to note that the human mind is fully capable of extrapolating 3D structures when only 2D information is present. By examining different views (2D projections) of an unknown object, one might superimpose the projection images in various orientations with one another, as if putting together a 3D puzzle, a task that was performed with remarkable accuracy for early structures of the ribosome (Lake, 1976). This type of manual interpretation, even when accurate, depends on the predilection of the

320

Dmitry Lyumkis et al.

user. Automated methods for performing ab initio reconstructions largely reduce the bias associated with manual analyses and fall into one of two general categories—those that do and those that do not require physical specimen tilting inside the microscope. Specific advantages and disadvantages to each will be mentioned below, for which comprehensive discussions can be found in (Cheng et al., 2006; Voss et al., 2010). Tomographic reconstructions (Frank, 2006a), random-conical tilt (RCT) reconstructions (Radermacher et al., 1986), and orthogonal tilt (OT) reconstructions (Leschziner and Nogales, 2006; Leschziner et al., this volume) rely upon physical specimen tilt. With the tomographic technique, the specimen stage is sequentially tilted from about 60 to þ60 using a small angular increment, providing nearly all directional views with the exception of the missing wedge of information. A limitation of this straightforward technique is that the biological specimen acquires radiation damage with each electron dose causing damage to the specimen. The combination of serial electron dosing and the resulting missing wedge of ˚, information currently limits the resolution of tomography to 30 A although algorithmic improvements for tilted CTF estimation (Fernandez et al., 2006; Philippsen et al., 2007; Sorzano et al., 2009a), tilt-series alignment (Sorzano et al., 2009a; Winkler and Taylor, 2006), and subtomogram averaging (Winkler et al., 2009) are pushing this limit. RCT and OT methods also utilize specimen tilt, but require only the collection of image pairs. In RCT, images are taken at 0 and 45–60 . This allows the former to be used for 2D alignment and classification, and the latter for 3D reconstruction. OT differs from RCT in that images are collected at þ45 and 45 , providing the equivalent of a 90 rotation and eliminating any missing Fourier information. In practice, all of these tilted acquisition methods are challenging to perform manually, primarily due to the significant amount of bookkeeping required to track the precise alignment shifts and rotations, classification dependencies, and tilt angles for each particle. In Appion, a complete record of all particle parameters is retained within the database, and therefore RCT, OT, and tomographic reconstructions can be achieved in a single step following automated alignment and classification (Voss et al., 2010). Thus, with appropriate checkpoints to provide for user intervention as needed, all of the ab initio methods that rely on titled image acquisition are amenable to full automation. When physical specimen tilt is not an option, it is possible to obtain an initial model exclusively from the transmitted information composing each image. Common-lines based methods (Crowther et al., 1970; Farrow and Ottensmeyer, 1992; Goncharov and Vainshtein, 1986; Penczek et al., 1996; Singer et al., 2010; van Heel, 1987) rely on the central section theorem, which permits the identification of identical intersecting 1D lines for all combinatorial pairs of projections and the resulting assignment of Euler angles, although this strategy is only viable

Automation in Single-Particle Electron Microscopy

321

when the particles do not exhibit preferred orientation. The problem with any common-lines based approach is that, given a specific set of input values, a single globally or locally “optimal” structure will inevitably emerge and there is no reliable way to distinguish between correct and incorrect reconstructions. Some groups have relied on random methods of generating initial models, initializing either from random noise or a symmetric Gaussian sphere (Ludtke et al., 1999; Yan et al., 2007), on the principle that iterative convergence to a common motif indicates an appropriate starting model and avoids the introduction of bias into the reconstruction procedure. This impartial approach is sometimes successful when applied to large and highly symmetric particles (Liu et al., 2007; Ochoa et al., 2008; Pan et al., 2009; Parent et al., 2010) but is limited by the unfavorable, low signal-to-noise nature of cryo-EM images and can lead to incorrect structures within local minima for smaller, less symmetric objects. Automation and increased computing power has the potential to address the issue of structural reliability by applying statistical validation and the ability to repeat reconstructions using different starting models. Nevertheless, the benefit of simply gathering auxiliary information to validate 3D reconstructions as in (Singer et al., 2010), is crucial.

5.5. 3D refinement Most initial models establish nothing more than a preliminary sense of the overall shape of the biological specimen. In order to reveal structural information that can answer specific biological questions, the model requires refining. In single-particle analysis, a refinement is an iterative procedure, which sequentially aligns the raw particles, assign to them appropriate spatial orientations (Euler angles) by comparing them against the model, and then backprojects them into 3D space to form a new model. Effectively, a full refinement takes as input a raw particle stack and an initial model and is usually carried out until no further improvement of the structure can be observed, often measured by convergence to some resolution criterion (Frank et al., 1981; Harauz and van Heel, 1986; Penczek et al., 1994; Sousa and Grigorieff, 2007; Unser et al., 1987, 1989). Iterative refinement of a 3D model relies on the incremental adjustment of various parameters for full exploitation of the information content of the data. For example, decreasing the angular projection increment of the input model allows for a finer

322

Dmitry Lyumkis et al.

sampling of the data during classification and back-projection. Resolutiondependent band-pass filtering procedures are routinely applied in order to eliminate irrelevant information that exceeds the resolution limit of the map. Density sharpening using b-factor weighting or low-pass filtering (Fernandez et al., 2008; Rosenthal and Henderson, 2003) boosts the amplitudes of the high-frequency information within the map, lost during image acquisition. Many packages also offer automasking procedures in which a structure-specific mask is applied based on densities in the digital image and the particle size (Grigorieff, 2007; Ludtke et al., 1999; Tang et al., 2007; van Heel et al., 1996). Analysis of the Euler angles and correlation values assigned to raw particles can ensure that only the most reliable particles contribute to the final reconstruction (Stagg et al., 2006). The ability to identify and appropriately apply these parameters is what distinguishes a “manual” from an “automated” iterative refinement. In the manual approach, user intervention partitions the large combination of steps pertaining to each iteration into distinct protocols involving frequent decision-making (e.g., how many particles to exclude or how to filter the stack for the next iteration). Automated iterative refinements take such decisions into account, and have already been variously implemented into the major EM packages, as for example the “semiautomated” projection-matching reconstruction scheme within EMAN (Ludtke et al., 1999). Similar intelligent algorithms that take into consideration information limits or inconsistencies in the data before proceeding with the next step of a refinement can be expected in the future. The key is to impose enough transparency on the process such that tweaking parameters at any step remains feasible without sacrificing streamlined efficiency.

6. Assessment and Integration Perhaps the most important aspect of any automated, or for that matter, manual, operation is the means to assess its outcome (a list of automated quality assessment steps for both Leginon and Appion is described in Box 5). Automation, certainly at this stage of its development, should not be thought of as a replacement for intelligent and critical evaluation by an operator. Thus, a major goal of any automated procedure should be to provide feedback to the end-user in a transparent and easily accessible form. In areas such as specimen preparation,

Automation in Single-Particle Electron Microscopy

323

this kind of feedback is not readily available until the grid is in the electron microscope. This has led to proposals for devices that examine vitrified grids using light microscopy, but so far their use is limited to fluorescently labeled specimens (Sartori et al., 2007). However, if specimen transfer itself becomes more automated, the need for pre-screening devices will be reduced by an improvement in the transfer operation and resulting reduction in grid and vacuum contamination. Once the grid is in the microscope and automated image acquisition begins, quality assessment becomes a critical component at multiple points in the process. For example, without a user constantly monitoring the instrument, as during manual acquisition, the automated software is responsible for monitoring changes in parameters such as beam intensity, image and beam shifts, and microscope alignments. If these parameters depart from expected ranges, they must either be automatically adjusted or user intervention must be requested. The user should also be able to readily examine the images being acquired, preferably from any remote computer. For example, in Leginon, this form of monitoring is available with web-based viewing tools that provide immediate access to images as they are acquired. Using any standard web browser, the user can parse through images, examine the relationship between the quality of highmagnification images and targets selected on lower magnification images (a hierarchical viewing strategy), compute power spectra to evaluate the defocus and astigmatism settings, view particle picking and CTF estimation results overlaid on the raw micrographs, and perform simple measurements on the data. Box 5 NRAMM Quality Assessment

Purpose: Describe tools used for routine quality assessment in Leginon and Appion. 1. Assess the quality of automated data collection: In Leginon, the results of any operation can be tested within the interactive GUI prior to target queuing, as, for example, when assessing the hole-finder algorithm. This ensures that optimal criteria are set prior to data collection. Leginon displays output to the user when it encounters errors during data collection, as in the case when an autofocus procedure fails or when the specimen drifts for an extended period of time. At all times, the functional state of the software and all relevant data collection parameters are monitored and tracked using the database to ensure reproducibility in future sessions. 2. Assess the quality of the collected data: Each Leginon session keeps track of statistics during data collection, which are used to assess the quality of the images. For each experiment and for each collected image, Leginon (continued )

324

Dmitry Lyumkis et al.

Box 5 (continued )

monitors: (a) all instrument parameters associated with the microscope, (b) the duration of collection, (c) camera readout statistics and electron dose, (d) the magnitude of specimen drift over time, (e) the ice thickness for each hole, (f) image and beam shift values, (g) the accuracy of the autofocus procedure, and other additional parameters. This type of record keeping allows us to track the state of the instruments and analyze the quality of the collected data. 3. Eliminate undesired micrographs or regions: Images can be manually selected to be “hidden” from view using buttons on the Imageviewer. Note that images can also be marked as “Exemplars.” In Appion, “Junk Finding” and “Manual Masking” options exclude undesired areas of the micrograph for particle selection. The “Multi-Image Assessment” tool allows the user to reject entire micrographs, usually after a particle-selection run, while the “Image Rejector” function automatically rejects images that either do not have CTF parameters, particle picks, or associated tilt pairs, or are outside a specified fitness factor of defocus range after CTF estimation. Images that have been “rejected” or “hidden” can be optionally excluded from processing runs. Processing runs can also be set to use only “exemplar” images, often useful for testing. Rejected or hidden images, or exemplars, can still be viewed in the Imageviewer by selecting them form the pulldown menu options. Hidden images can be restored while viewing them by selecting the hidden button again. Images are never permanently removed from the database. 4. Assess the quality of processing algorithms: Appion’s imageviewer allows the user to assess some of the initial steps of image processing. The “ACE” button provides graphical displays of CTF estimation as well as fitting parameters and fitness values. The user can visualize images of the edges detected by the algorithm and compare them to the Thon rings present in the PSD of the image. The “P” button displays all particle picks from the current micrograph for the particle-selection run specified by the user. Histograms of confidence scores and correlation values are displayed on the Appion summary pages for CTF estimation and particle picking. 5. Assess the quality of particle stacks: Stacks can be directly examined as a montage of particles. Summary pages provide graphical summaries of the intensity and standard deviation of the particle images and these outputs can be useful for cleaning up the stack by rejecting particles on the edge of a hole or over the carbon. The “Xmipp_sort_by_statistics” algorithm can also be used to identify junk particles. During subsequent image-processing steps, particles may be rejected based on poor scores within 2D alignment and classification steps, high Euler jumper statistics, or poor correlations for class assignment during 3D refinement.

Automation in Single-Particle Electron Microscopy

325

6. Assess the accuracy of 2D alignment and classification: Appion report pages display Eigenimages, variance averages, and/or correlation histograms for each alignment and classification run; some outputs are algorithm dependent. All translations, rotations, and mirror reflections applied to the images, as well as the coordinates within the original micrograph are stored in the database. 7. Assess veracity of an electron density map: Appion displays a variety of output related to the reconstruction refinement as an aid in determining the consistency and veracity of the reconstructed map. Each method for generating initial models provides combinatorial assessment criteria consistent with the routine (exemplified by the random-conical reconstruction method in Fig. 14.6). Later refinement runs automatically generate summary pages providing output at each iteration that includes resolution derived from the Fourier Shell and Rmeasure criteria, distribution of Euler angles, Euler angle differences between the current and previous iteration as well as the average median Euler difference for all iterations, side-by-side comparison of input projections with associated reprojections of the map, and snapshots of the reconstructed model.

Figure 14.6 Assessment Pages for a random-conical tilt reconstruction (A) A table summarizes the number of particles, FSC, R-measure resolution, and description for all available 3D reconstructions. (B) Clicking on a reconstruction name opens a detailed summary page that includes iterative snapshots and links to (C) 2D alignment plots and (D) class averages used for 3D reconstruction. Within the viewer, individual 2D class averages can be selected and their raw particles viewed. (E) Individual raw particles can be marked for exclusion during sub-stack creation.

326

Dmitry Lyumkis et al.

Similarly, during processing and analysis of the images the requirement for assessment and critical evaluation along all steps of the procedure is obvious. Sometimes assessment might be fairly straightforward. For example, when picking particles on a micrograph or estimating the position of Thon rings in the micrograph PSD, the user has ready access to graphical overlays on the images and can visually inspect the result of the automated procedures. However, many of the most exciting structural challenges of the future will likely require the processing of hundreds of thousands, or millions, of particles, and in this case visual assessment of every image or every particle becomes unreasonable. Instead, the automated algorithms must also provide figures of merit that provide the means to accept or reject the results, ideally in conjunction with visual inspection of a limited number of images. We note here that automated algorithms are unlikely to ever be in 100% agreement with user decisions and thus must be reevaluated at several points along the pipeline. Ultimately, the goal of the automated algorithms is not to precisely emulate a human operator, but rather to provide the highest quality results using the given data. At the later processing stages of alignment, classification, and 3D reconstruction a significant level of automation exists, and the current goal is to provide the user with as much information as possible to facilitate decisions about the validity of the outcome. For example, it would be dangerous to believe a 3D map based only on an inspection of the map itself without a close, critical examination of all associated data. This last should include Euler angle statistics, resolution curves, comparisons between classes and reprojections, summaries of selected vs. rejected particles, variance maps, and other validations. All of these factors should also be assessed over the entire course of the iterative procedures. One of our goals in developing Appion (Lander et al., 2009) was to make sure that the user has ready access to this information no matter what set of procedures is used for processing, analysis, and refinement. To this end, Appion uses a web browser to present summaries of relevant data in a single format regardless of the individual programs used for processing or refinement. This frees the user from the need for detailed knowledge of how and where each of the different software suites stores and presents data. While the final evaluation of a calculated map is currently left almost entirely to the judgment of an experienced user, we can envisage future figures of merit based on an integrated evaluation of the entire set of outcomes, or comparisons to previous results for similar structures. The idea of an integrated system that has access to all data and meta data for all processing steps has been a significant driving force in the development of

Automation in Single-Particle Electron Microscopy

327

both Leginon and Appion. Both systems are intimately connected to a relational database that tracks every item of data (Fellmann et al., 2002). For Leginon this includes the calibration and setup parameters, and a complete record of the state of the instrument for every image acquired. In Appion, the database tracks every procedure used and all input and output data and parameters, creating links and calculating statistics where necessary. This provides a complete and permanent record of the processing steps applied to any data set that can be accessed at any time in the future, independent of an individual’s record keeping skills. For example, in Appion and Leginon, the exact time, stage position inside the microscope, and ambient air and column temperature are known for each particle in a given 3D reconstruction. We should also note that the term “database,” as used in Appion and Leginon, refers to more than just a set of linearly searchable, well-organized files. Rather, it is a collection of logically structured, related data, together with a system to query and update that data in an efficient, non-linear, manner. This implementation combines the database management system and the database itself, providing the tools to create schemas and perform complex information queries using the Structured Query Language (SQL). Mining the database allows us to efficiently identify relevant parameters in a processing run, readily repeat precise processing conditions, output valuable quantitative statistics, and inter-connect paths from different processing stages. An important goal for Appion is to use this infrastructure and the resulting accessibility of meta-data for individual experiments to provide intelligent feedback during processing and analysis.

7. The Future of Automation The unfortunate limitation of our field is that the theoretical Nyquist frequency of an electron micrograph does not predict the achievable resolution for the specimen of interest. Early in the development of the field, impressive illustrations of the potential of EM to attain atomic detailed reconstructions were provided using 2D electron crystallography (e.g., Henderson et al., 1990; Nogales et al., 1998). While these methods obviously benefited from the inherent averaging imposed by specimen crystallization, they provided hope for attaining high-resolution

328

Dmitry Lyumkis et al.

reconstructions for single particles. “Chasing the train,” a pun that indicated the potential to trace the polypeptide chain into an EM density map, became somewhat popular at the turn of the century (van Heel et al., 2000). Less than a decade later, backbone traces and atomic resolution maps for single particles have begun to appear, although still primarily limited to specimens with high symmetry and large molecular weight (Chen et al., 2009; Cong et al., 2010; Jiang et al., 2008; Ludtke et al., 2008; Yu et al., 2008; Zhang et al., 2008). Significant progress has also been made in improving the resolution of particles with low symmetry and molecular weight; the last several years have seen 3D reconstructions of the 290 kDa Transferrin–Transferrin Receptor complex at 7.5 A˚ resolution (Cheng et al., 2004), the 180 kDa ˚ resolution (Okorokov et al., tetrameric P53 tumor suppressor at 13.7 A 2006), and a DNA nanostructure at just 78 kDa solved to 12 A˚ resolution (Kato et al., 2009). But while the theoretical potential of the technology seems increasingly within reach, the ability to achieve high resolution remains the exception rather than the norm. In practice, most specimens are unlikely to achieve a backbone trace, especially given the conformational and compositional heterogeneity of many biological samples. Recently, in a seminal paper covering the last 30 years of single-particle reconstructions, Joachim Frank proposed that finding recipes to achieve resolutions beyond what has already been achieved, particularly for asymmetric particles, remains a daunting task, which will require pushing the quantity of data being both collected and processed (Frank, 2009). We believe that automation has a crucial role to play in this arena. We return once again to our original definition of automation and investigate why this term is so appropriate to EM. In the most fundamental case, automation allows for the acquisition of thousands of micrographs and hundreds of thousands of single particles. Beyond this, we can ensure reproducibility when preparing multiple vitrified samples in an identical fashion, tighter quality control by collecting images exclusively within a specific range of ice thickness, waste reduction by focusing solely on the particles that have been grouped into homogeneous subsets through classification, and integration with systems by enabling inter-package processing compatibility. With the incorporation of machine-learning algorithms, comprehensive graphical user interfaces, and integrated centralized databases for image processing, automated technologies begin to resemble human inductive methods of learning. They begin to adapt to the particularities of each novel dataset, allow for user-based intervention, and contain, at their core, the connective glue to interlink the assortment of operations from initial data collection through construction of a final map. The result of this is increased productivity and reduction of labor. Automation, when done right, opens up possibilities for performing experiments that might otherwise remain out of reach.

Automation in Single-Particle Electron Microscopy

329

In an effort to push the limits of EM technologies, many laboratories around the world, both in academia and industry, are addressing current technical challenges with sophisticated microscope and computational accessories. Spherical and chromatic aberration correctors will allow finer sampling by reducing contrast delocalization in high-resolution areas. Energy filters eliminate inelastically scattered electrons contributing to background noise. Phase plates positioned in the back focal plane of the objective lens modify the CTF and amplify contrast of vitrified specimens (Danev and Nagayama, 2001; Nagayama and Danev, 2008; Danev et al., this volume), potentially allowing for reconstruction of very small structures. Cameras equipped with direct electron detectors could soon surplant current fiber optic-coupled CCD cameras and improve the modulation transfer function (MTF) of acquired images (Faruqi and Henderson, 2007). General-purpose computing on graphics processing units (GPUs) (reviewed in Schmeisser et al., 2009) have demonstrated the capability to provide more than 50-fold speedups of matrix multiplication, fast Fourier transforms, and multireference particle comparisons (Castano Diez et al., 2007, 2008; Govindaraju et al., 2008) and may replace CPU processing for computationally expensive tasks like maximum-likelihood or bootstrap calculations. Collectively, these technical improvements promise to expand the scope of questions addressable by single-particle EM, but will likely increase the complexity of our instruments. Automation will undoubtedly play a role in providing maintenance, alignment, and calibration of these and other increasingly sophisticated devices. While technical advancements in many aspects of EM may signify that atomic resolution will be achievable for some suitable macromolecules, there are many interesting biological questions that can be answered at considerably lower resolutions. As a resource facility, we aim to provide technology to a wide array of researchers, including biologists, chemists, material scientists, and other practitioners from outside the cryo-EM field and a high level of automation and streamlining is critical to this mission. We are increasingly seeing EM used as one of a wide variety of hybrid methods. For example, cryo-EM has been combined with multiangle laser light scattering (MALLS), small-angle X-ray scattering (SAXS), mass-spectrometry, and biochemical mutational analyses to identify the dimerization domain of ALIX during its involvement in membrane remodeling (Pires et al., 2009). It has been used in combination with nuclear magnetic resonance (NMR) spectroscopy, X-ray crystallography, SAXS, and computational docking, to determine the structure of the actin-binding domain of Talin and provide direct evidence for its interaction with cytoskeletal actin (Gingras et al., 2008). It has been combined with data from ultracentrifugation, quantitative immunoblotting, affinity purification, overlay assays, immuno-labeling, and bioinformatics in a tour-deforce structural analysis of the entire nuclear pore complex (Alber et al.,

330

Dmitry Lyumkis et al.

2007a,b). EM, as a field, has long since ceased to belong purely to the physicists and is rapidly transforming into a mature technology embraced by many scientific disciplines. Groups around the world are assembling the puzzle pieces for all of the distinct operations, collaboratively putting together a streamlined EM pipeline, polishing the building materials, repairing the leaks, and broadening the bottlenecks in the pipeline. The successful future of EM does not simply depend upon its ability to rival contenders in attaining atomic resolution, but rather the broader integration of the technology as a hybrid approach to addressing increasingly complex questions.

ACKNOWLEDGMENTS We acknowledge primary support from the National Institutes of Health (NIH) through the National Center for Research Resources (NCRR) P41 program (Grants RR17573 and RR023093). We are also grateful for the support provided to Arne Moeller by the Joint Center for Innovation in Membrane Protein Production for Structure Determination (Grant RFA-RM-08-019) and to Pick-Wei Lau by the American Heart Association. We ask for forgiveness for any omissions, errors, or oversights that are almost inevitable in a review of this wide scope.

REFERENCES Alber, F., et al. (2007a). Determining the architectures of macromolecular assemblies. Nature 450, 683–694. Alber, F., et al. (2007b). The molecular architecture of the nuclear pore complex. Nature 450, 695–701. Baldwin, P. R., and Penczek, P. A. (2005). Estimating alignment errors in sets of 2-D images. J. Struct. Biol. 150, 211–225. Baxter, W. T., et al. (2007). SPIRE: The SPIDER reconstruction engine. J. Struct. Biol. 157, 56–63. Baxter, W. T., et al. (2009). Determination of signal-to-noise ratios and spectral SNRs in cryo-EM low-dose imaging of molecules. J. Struct. Biol. 166, 126–132. Berriman, J., and Unwin, N. (1994). Analysis of transient structures by cryo-microscopy combined with rapid mixing of spray droplets. Ultramicroscopy 56, 241–252. Bonnet, N. (1998). Multivariate statistical methods for the analysis of microscope image series: Applications in materials science. J. Microsc. 190, 2–18. Bonnet, N. (2000). Artificial intelligence and pattern recognition techniques in microscope image processing and analysis. Adv. Imaging Electron Phys. 114, 1–77. Carazo, J. M., et al. (1990). Fuzzy sets-based classification of electron microscopy images of biological macromolecules with an application to ribosomal particles. J. Microsc. 157, 187–203. Cardona, A., et al. (2010). An integrated micro- and macroarchitectural analysis of the Drosophila brain by computer-assisted serial section electron microscopy (submitted). Carragher, B., et al. (2000). Leginon: An automated system for acquisition of images from vitreous ice specimens. J. Struct. Biol. 132, 33–45. Castano Diez, D., et al. (2007). Implementation and performance evaluation of reconstruction algorithms on graphics processors. J. Struct. Biol. 157, 288–295.

Automation in Single-Particle Electron Microscopy

331

Castano-Diez, D., et al. (2008). Performance evaluation of image processing algorithms on the GPU. J. Struct. Biol. 164, 153–160. Chen, J. Z., and Grigorieff, N. (2007). SIGNATURE: A single-particle selection system for molecular electron microscopy. J. Struct. Biol. 157, 168–173. Chen, J. Z., et al. (2009). Molecular interactions in rotavirus assembly and uncoating seen by high-resolution cryo-EM. Proc. Natl. Acad. Sci. USA 106, 10644–10648. Cheng, Y., et al. (2004). Structure of the human transferrin receptor-transferrin complex. Cell 116, 565–576. Cheng, Y., et al. (2006). Single particle reconstructions of the transferrin-transferrin receptor complex obtained with different specimen preparation techniques. J. Mol. Biol. 355, 1048–1065. Cheng, A., et al. (2007). Towards automated screening of two-dimensional crystals. J. Struct. Biol. 160, 324–331. Cong, Y., et al. (2010). 4.0-A resolution cryo-EM structure of the mammalian chaperonin TRiC/CCT reveals its unique subunit arrangement. Proc. Natl. Acad. Sci. USA 107, 4967–4972. Coudray, N., et al. (2008). Automatic acquisition and image analysis of 2D crystals. Micros. Today 16, 48–49. Coudray, N., et al. (2010). Automated Screening of 2D crystallization trials using transmission electron microscopy: A high-throughput tool-chain for sample preparation and microscopic analysis. Submitted. Crowther, R. A., et al. (1996). Three dimensional reconstructions of spherical viruses by Fourier synthesis from electron micrographs. Nature 226, 421–425. Crowther, R. A., et al. (1970). MRC image processing programs. J. Struct. Biol. 116, 9–16. Danev, R., and Nagayama, K. (2001). Transmission electron microscopy with Zernike phase plate. Ultramicroscopy 88, 243–252. Dierksen, K., et al. (1992). Towards automatic electron tomography. Ultramicroscopy 40, 71–87. Dierksen, K., et al. (1993). Towards automatic electron tomography 2. Implementation of autofocus and low-dose procedures. Ultramicroscopy 49, 109–120. Drahos, V., and Delong, A. (1960). A simple method for obtaining perforated supporting membranes for electron microscopy. Nature 186, 104. Dubochet, J., et al. (1982). Electron microscopy of frozen water and aqueous solutions. J. Microsc. 128, 219–237. Dubochet, J., et al. (1988). Cryo-electron microscopy of vitrified specimens. Q. Rev. Biophys. 21, 129–228. Ermantraut, E., et al. (1998). Perforated support foils with re-defined hole size, shape and arrangement. Ultramicroscopy 74, 75–81. Farrow, N. A., and Ottensmeyer, F. P. (1992). A posteriori determination of relative projection directions of arbitrarily oriented macromolecules. J. Opt. Soc. Am. A 9, 1749–1760. Faruqi, A. R., and Henderson, R. (2007). Electronic detectors for electron microscopy. Curr. Opin. Struct. Biol. 17, 549–555. Fellmann, D., et al. (2002). A relational database for cryoEM: Experience at one year and 50000 images. J. Struct. Biol. 137, 273–282. Fernandez, J. J., et al. (1997). A spectral estimation approach to contrast transfer function detection in electron microscopy. Ultramicroscopy 68, 267–295. Fernandez, J. J., et al. (2006). CTF determination and correction in electron cryotomography. Ultramicroscopy 106, 587–596. Fernandez, J. J., et al. (2008). Sharpening high resolution information in single particle electron cryomicroscopy. J. Struct. Biol. 164, 170–175.

332

Dmitry Lyumkis et al.

Frank, J. (ed.) (2006). Electron Tomography—Methods for Three-Dimensional Visualization of Structures in the Cell, Springer, New York. Frank, J. (2006b). Three-dimensional electron microscopy of macromolecular assemblies: Visualization of biological molecules in their native state Oxford University Press, New York. Frank, J. (2009). Single-particle reconstruction of biological macromolecules in electron microscopy—30 years. Q. Rev. Biophys. 42, 139–158. Frank, J., and van Heel, M. (1982). Correspondence analysis of aligned images of biological particles. J. Mol. Biol. 161, 134–137. Frank, J., et al. (1981). Computer averaging of electron micrographs of 40S ribosomal subunits. Science 214, 1353–1355. Frank, J., et al. (1988). Classification of images of biomolecular assemblies: A study of ribosomes and ribosomal subunits of Escherichia coli. J. Microsc. 150, 99–115. Frank, J., et al. (1996). SPIDER and WEB: Processing and visualization of images in 3D electron microscopy and related fields. J. Struct. Biol. 116, 190–199. Frederik, P. M., and Hubert, D. H. W. (2005). Cryoelectron microscopy of liposomes. Methods in Enzymol. 391, 431–448. Frey, B. J., and Dueck, D. (2007). Clustering by passing messages between data points. Science 315, 972–976. Fung, J. C., et al. (1996). Toward fully automated high-resolution electron tomography. J. Struct. Biol. 116, 181–189. Gingras, A. R., et al. (2008). The structure of the C-terminal actin-binding domain of talin. EMBO J. 27, 458–469. Goncharov, A., and Vainshtein, B. K. (1986). Determining the spatial orientation of arbitrarily arranged particles given their projections. Dokl. Acad. Sci. USSR 287, 1131–1134. Govindaraju, N. K., et al. (2008). High performance discrete fourier transforms on graphics processors. Proceedings of the 2008 ACM/IEEE Conference on Supercomputing, pp. 1–12. IEEE Press, Austin, TX. Grigorieff, N. (2007). FREALIGN: High-resolution refinement of single particle structures. J. Struct. Biol. 157, 117–125. Hall, R. J., and Patwardhan, A. (2004). A two step approach for semi-automated particle selection from low contrast cryo-electron micrographs. J. Struct. Biol. 145, 19–28. Harauz, G., and van Heel, M. (1986). Exact filters for general geometry 3-dimensional reconstruction. Optik 73, 146–156. Harris, W. J. (1962). Holey films for electron microscopy. Nature 196, 499–500. Henderson, R., et al. (1990). Model for the structure of bacteriorhodopsin based on highresolution electron cryo-microscopy. J. Mol. Biol. 213, 899–929. Heymann, J. B. (2001). Bsoft: Image and molecular processing in electron microscopy. J. Struct. Biol. 133, 156–169. Heymann, J. B., and Belnap, D. M. (2007). Bsoft: Image processing and molecular modeling for electron microscopy. J. Struct. Biol. 157, 3–18. Heymann, J. B., et al. (2008). Computational resources for cryo-electron tomography in Bsoft. J. Struct. Biol. 161, 232–242. Hohn, M., et al. (2007). SPARX, a new environment for Cryo-EM image processing. J. Struct. Biol. 157, 47–55. Hu, M., et al. (2010). Automated electron microscopy for evaluating two-dimensional crystallization of membrane proteins. J. Struct. Biol. (submitted). Huang, Z., and Penczek, P. A. (2004). Application of template matching technique to particle detection in electron micrographs. J. Struct. Biol. 145, 29–40. Huang, Z., et al. (2003). Automated determination of parameters describing power spectra of micrograph images in electron microscopy. J. Struct. Biol. 144, 79–94.

Automation in Single-Particle Electron Microscopy

333

Iacovache, I., et al. (2010). The 2DX robot: A membrane protein 2D crystallization Swiss Army knife. J. Struct. Biol. 169, 370–378. Jiang, W., et al. (2008). Backbone structure of the infectious epsilon15 virus capsid revealed by electron cryomicroscopy. Nature 451, 1130–1134. Joyeux, L., and Penczek, P. A. (2002). Efficiency of 2D alignment methods. Ultramicroscopy 92, 33–46. Kato, T., et al. (2009). High-resolution structural analysis of a DNA nanostructure by cryoEM. Nano Lett. 9, 2747–2750. Kelly, D. F., et al. (2008). The Affinity Grid: A pre-fabricated EM grid for monolayer purification. J. Mol. Biol. 382, 423–433. Kimoto, K., et al. (2003). Practical procedure for coma-free alignment using caustic figure. Ultramicroscopy 96, 219–227. Kisseberth, N., et al. (1997). emScope: A tool kit for control and automation of a remote electron microscope. J. Struct. Biol. 120, 309–319. Koster, A. J., and de Ruijter, W. J. (1992). Practical autoalignment of transmission electron microscopes. Ultramicroscopy 40, 89–107. Koster, A. J., et al. (1987). An autofocus method for a TEM. Ultramicroscopy 21, 209–222. Ku¨hlbrandt, W. (2003). Two-dimensional crystallization of membrane proteins: A practical guide. In “Membrane Protein Purification and Crystallization: A Practical Guide,” (C. Hunte, et al., eds.), pp. 253–284. Academic Press, San Diego. Lake, J. A. (1976). Ribosome structure determined by electron microscopy of Escherichia coli small subunits, large subunits and monomeric ribosomes. J. Mol. Biol. 105, 131–139. Lander, G. C., et al. (2009). Appion: An integrated, database-driven pipeline to facilitate EM image processing. J. Struct. Biol. 166, 95–102. Lefman, J., et al. (2007). Automated 100-position specimen loader and image acquisition system for transmission electron microscopy. J. Struct. Biol. 158, 318–326. Lei, J., and Frank, J. (2005). Automated acquisition of cryo-electron micrographs for single particle reconstruction on an FEI Tecnai electron microscope. J. Struct. Biol. 150, 69–80. Leschziner, A. E., and Nogales, E. (2006). The orthogonal tilt reconstruction method: An approach to generating single-class volumes with no missing cone for ab initio reconstruction of asymmetric particles. J. Struct. Biol. 153, 284–299. Liu, X., et al. (2007). Averaging tens to hundreds of icosahedral particle images to resolve protein secondary structure elements using a Multi-Path Simulated Annealing optimization algorithm. J. Struct. Biol. 160, 11–27. Lu, Z., et al. (2009). Monolithic microfluidic mixing-spraying devices for time-resolved cryo-electron microscopy. J. Struct. Biol. 168, 388–395. Ludtke, S. J., et al. (1999). EMAN: Semiautomated Software for High-Resolution SingleParticle Reconstructions. J. Struct. Biol. 128, 82–97. Ludtke, S. J., et al. (2008). De novo backbone trace of GroEL from single particle electron cryomicroscopy. Structure 16, 441–448. Mallick, S. P., et al. (2004). Detecting particles in cryo-EM micrographs using learned features. J. Struct. Biol. 145, 52–62. Mallick, S. P., et al. (2005). ACE: Automated CTF estimation. Ultramicroscopy 104, 8–29. Marabini, R., and Carazo, J. M. (1994). Pattern recognition and classification of images of biological macromolecules using artificial neural networks. Biophys. J. 66, 1804–1814. Marabini, R., et al. (1996). Xmipp an image processing package for electron microscopy. J. Struct. Biol. 116, 237–240. Marsh, M. P., et al. (2007). Modular software platform for low-dose electron microscopy and tomography. J. Microsc. 228, 384–389. Mastronarde, D. N. (2005). Automated electron microscope tomography using robust prediction of specimen movements. J. Struct. Biol. 152, 36–51.

334

Dmitry Lyumkis et al.

Melanson, L. A. (2009). A versatile and affordable plunge freezing instrument for preparing specimens for cryo transmission electron microscopy (cryoEM). Micros. Today 2, 14–17. Microsystems (2009). L., Leica EM GP Product Brochure. Mindell, J. A., and Grigorieff, N. (2003). Accurate determination of local defocus and specimen tilt in electron microscopy. J. Struct. Biol. 142, 334–347. Mulder, A. M., et al. (2010). Time resolved single particle EM reveals structure-based mechanism for assembly of the 30S ribosomal subunit (submitted).. Nagayama, K., and Danev, R. (2008). Phase contrast electron microscopy: Development of thin-film phase plates and biological applications. Philos. Trans. R. Soc. Lond. B Biol. Sci. 363, 2153–2162. Nickell, S., et al. (2005). TOM software toolbox: Acquisition and analysis for electron tomography. J. Struct. Biol. 149, 227–234. Nogales, E., et al. (1998). Structure of the alpha beta tubulin dimer by electron crystallography. Nature 391, 199–203. Ochoa, W. F., et al. (2008). Partitivirus structure reveals a 120-subunit, helix-rich capsid with distinctive surface arches formed by quasisymmetric coat-protein dimers. Structure 16, 776–786. Ogura, T., and Sato, C. (2004a). Auto-accumulation method using simulated annealing enables fully automatic particle pickup completely free from a matching template or learning data. J. Struct. Biol. 146, 344–358. Ogura, T., and Sato, C. (2004b). Automatic particle pickup method using a neural network has high accuracy by applying an initial weight derived from eigenimages: A new reference free method for single-particle analysis. J. Struct. Biol. 145, 63–75. Ohi, M., et al. (2004). Negative staining and image classification—Powerful tools in modern electron microscopy. Biol. Proced. Online 6, 23–34. Okorokov, A. L., et al. (2006). The structure of p53 tumour suppressor protein reveals the basis for its functional plasticity. EMBO J. 25, 5191–5200. Pan, J., et al. (2009). Atomic structure reveals the unique capsid organization of a dsRNA virus. Proc. Natl. Acad. Sci. USA 106, 4225–4230. Pantelic, R. S., et al. (2010). Graphene oxide: A substrate for optimizing preparations of frozen-hydrated samples. J. Struct. Biol. 170, 152–156. Parent, K. N., et al. (2010). P22 coat protein structures reveal a novel mechanism for capsid maturation: Stability without auxiliary proteins or chemical crosslinks. Structure 18, 390–401. Pascual-Montano, A., et al. (2001). A novel neural network technique for analysis and classification of EM single-particle images. J. Struct. Biol. 133, 233–245. Pearson, K. (1901). On lines and planes of closest fit to systems of points in space. Philos. Mag. 2, 559–572. Penczek, P. A., et al. (1994). The ribosome at improved resolution: New techniques for merging and orientation refinement in 3D cryo-electron microscopy of biological particles. Ultramicroscopy 53, 251–270. Penczek, P. A., Zhu, J., and Frank, J. (1996). A common-lines based method for determining orientations for N > 3 particle projections simultaneously. Ultramicroscopy 63, 205–218. Philippsen, A., et al. (2007). The contrast-imaging function for tilted specimens. Ultramicroscopy 107, 202–212. Pires, R., et al. (2009). A crescent-shaped ALIX dimer targets ESCRT-III CHMP4 filaments. Structure 17, 843–856. Plaisier, J. R., et al. (2004). TYSON: Robust searching, sorting, and selecting of single particles in electron micrographs. J. Struct. Biol. 145, 76–83. Potter, C. S., et al. (1999). Leginon: A system for fully automated acquisition of 1000 electron micrographs a day. Ultramicroscopy 77, 153–161.

Automation in Single-Particle Electron Microscopy

335

Potter, C. S., et al. (2004). Robotic grid loading system for a transmission electron microscope. J. Struct. Biol. 146, 431–440. Quispe, J., et al. (2007). An improved holey carbon film for cryo-electron microscopy. Microsc. Microanal. 13, 365–371. Radermacher, M., et al. (1986). A new 3-D reconstruction scheme applied to the 50 S ribosomal subunit of E.coli. J. Microsc. 141, RP1–RP2. Rath, B. K., and Frank, J. (2004). Fast automatic particle picking from cryo-electron micrographs using a locally normalized cross-correlation function: A case study. J. Struct. Biol. 145, 84–90. Rath, B. K., et al. (1997). Low-dose automated electron tomography: A recent implementation. J. Struct. Biol. 120, 210–218. Roseman, A. M. (2004). FindEM—A fast, efficient program for automatic selection of particles from electron micrographs. J. Struct. Biol. 145, 91–99. Rosenthal, P. B., and Henderson, R. (2003). Optimal determination of particle orientation, absolute hand, and contrast loss in single-particle electron cryomicroscopy. J. Mol. Biol. 333, 721–745. Sander, B., et al. (2003). Automatic CTF correction for single particles based upon multivariate statistical analysis of individual power spectra. J. Struct. Biol. 142, 392–401. Sartori, A., et al. (2007). Correlative microscopy: Bridging the gap between fluorescence light microscopy and cryo-electron tomography. J. Struct. Biol. 160, 135–145. Saxton, W. O., and Frank, J. (1977). Motif detection in quantum noise-limited electron micrographs by cross-correlation. Ultramicroscopy 2, 219–227. Saxton, W. O., et al. (1983). Procedures for focusing, stigmating, and alignment in highresolution electron microscopy. J. Microsc. 130, 187–201. Scheres, S. H., et al. (2005a). Fast maximum-likelihood refinement of electron microscopy images. Bioinformatics 21(Suppl. 2), ii243–ii244. Scheres, S. H. W., et al. (2005b). Maximum-likelihood multi-reference refinement for electron microscopy images. J. Mol. Biol. 348, 139–149. Scheres, S. H., et al. (2008). Image processing for electron microscopy single-particle analysis using XMIPP. Nat. Protoc. 3, 977–990. Schmeisser, M., et al. (2009). Parallel, distributed and GPU computing technologies in single-particle electron microscopy. Acta Crystallogr. D Biol. Crystallogr. 65, 659–671. Shaikh, T. R., et al. (2008). Particle-verification for single-particle, reference-based reconstruction using multivariate data analysis and classification. J. Struct. Biol. 164, 41–48. Shi, J., et al. (2008). A Script-Assisted Microscopy (SAM) package to improve data acquisition rates on FEI Tecnai electron microscopes equipped with Gatan CCD cameras. J. Struct. Biol. 164, 166–169. Short, J. M. (2004). SLEUTH—A fast computer program for automatically detecting particles in electron microscope images. J. Struct. Biol. 145, 100–110. Sigworth, F. J. (1998). A maximum-likelihood approach to single-particle image refinement. J. Struct. Biol. 122, 328–339. Singer, A., et al. (2010). Detecting consistent common lines in cryo-EM by voting. J. Struct. Biol. (in press; corrected proof). Singh, V., et al. (2004). Image segmentation for automatic particle identification in electron micrographs based on hidden Markov random field models and expectation maximization. J. Struct. Biol. 145, 123–141. Sorzano, C. O., et al. (2004). Xmipp a new generation of an open-source image processing package for electron microscopy. J. Struct. Biol. 148, 194–204. Sorzano, C. O. S., et al. (2007). Fast, robust, and accurate determination of transmission electron microscopy contrast transfer function. J. Struct. Biol. 160, 249–262.

336

Dmitry Lyumkis et al.

Sorzano, C. O., et al. (2009a). Marker-free image registration of electron tomography tiltseries. BMC Bioinform. 10, 124. Sorzano, C. O., et al. (2009b). Error analysis in the determination of the electron microscopical contrast transfer function parameters from experimental power Spectra. BMC Struct. Biol. 9, 18. Sorzano, C. O. S., et al. (2009c). Automatic particle selection from electron micrographs using machine learning techniques. J. Struct. Biol. 167, 252–260. Sousa, D., and Grigorieff, N. (2007). Ab initio resolution measurement for single particle structures. J. Struct. Biol. 157, 201–210. Stagg, S. M., et al. (2006). Automated cryoEM data acquisition and analysis of 284, 742 particles of GroEL. J. Struct. Biol. 155, 470–481. Stewart, A., and Grigorieff, N. (2004). Noise bias in the refinement of structures derived from single particles. Ultramicroscopy 102, 67–84. Suloway, C., et al. (2005). Automated molecular microscopy: The new Leginon system. J. Struct. Biol. 151, 41–60. Suloway, C., et al. (2009). Fully automated, sequential tilt-series acquisition with Leginon. J. Struct. Biol. 167, 11–18. Tang, G., et al. (2007). EMAN2: An extensible image processing suite for electron microscopy. J. Struct. Biol. 157, 38–46. Taylor, K., and Glaeser, R. M. (1976). Electron microscopy of frozen-hydrated biological specimens. J. Ultrastruct. Res. 55, 448–456. Typke, D., and Dierksen, K. (1995). Determination of image aberrations in high-resolution electron microscopy using diffractogram and cross-correlation methods. Optik 99, 155–166. Typke, D., and Ko¨stler, D. (1977). Determination of the wave aberration of electron lenses from superposition diffractograms of images with differently tilted illumination. Ultramicroscopy 2, 285–295. Umesh Adiga, P. S., et al. (2004). A binary segmentation approach for boxing ribosome particles in cryo EM micrographs. J. Struct. Biol. 145, 142–151. Unser, M., et al. (1987). A new resolution criterion based on spectral signal-to-noise ratios. Ultramicroscopy 23, 39–51. Unser, M., et al. (1989). The spectral signal-to-noise ratio resolution criterion: Computational efficiency and statistical precision. Ultramicroscopy 30, 429–433. van Heel, M. (1984). Multivariate statistical classification of noisy images (randomly oriented biological macromolecules). Ultramicroscopy 13, 165–183. van Heel, M. (1987). Angular reconstitution: A posteriori assignment of projection directions for 3D reconstruction. Ultramicroscopy 21, 111–124. van Heel, M., and Frank, J. (1980). Classification of particles in noisy electron micrographs using correspondence analysis. (E. S. Gelsema and L. N. Kanal, eds.), vol. 1, pp. 235–243. Elsevier, Amsterdam. van Heel, M., and Frank, J. (1981). Use of multivariate statistics in analysing the images of biological macromolecules. Ultramicroscopy 6, 187–194. van Heel, M., and Keegstra, W. (1981). IMAGIC: A fast flexible and friendly image analysis software system. Ultramicroscopy 7, 113–130. van Heel, M., and Stoffler-Meilicke, M. (1985). Characteristic views of E. coli and B. stearothermophilus 30S ribosomal subunits in the electron microscope. EMBO J. 4, 2389–2395. van Heel, M., et al. (1992). Correlation functions revisited. Ultramicroscopy 46, 307–316. van Heel, M., et al. (1996). A new generation of the IMAGIC image processing system. J. Struct. Biol. 116, 17–24. van Heel, M., et al. (2000). Single-particle electron cryo-microscopy: Towards atomic resolution. Q. Rev. Biophys. 33, 307–369.

Automation in Single-Particle Electron Microscopy

337

Velasquez-Muriel, J. A., et al. (2003). A method for estimating the CTF in electron microscopy based on ARMA models and parameter adjustment. Ultramicroscopy 96, 17–35. Vink, M., et al. (2007). A high-throughput strategy to screen 2D crystallization trials of membrane proteins. J. Struct. Biol. 160, 295–304. Volkmann, N. (2004). An approach to automated particle picking from electron micrographs based on reduced representation templates. J. Struct. Biol. 145, 152–156. Voss, N. R., et al. (2009). DoG Picker and TiltPicker: Software tools to facilitate particle selection in single particle electron microscopy. J. Struct. Biol. 166, 205–213. Voss, N. R., et al. (2010). A toolbox for ab initio 3-D reconstructions in single-particle electron microscopy. J. Struct. Biol. 169, 389–398. Walker, M., et al. (1995). Millisecond time resolution electron cryo-microscopy of the M-ATP transient kinetic state of the acto-myosin ATPase. Biophys. J. 68, 87S–91S. Walker, M., et al. (1999). Observation of transient disorder during myosin subfragment-1 binding to actin by stopped-flow fluorescence and millisecond time resolution electron cryomicroscopy: Evidence that the start of the crossbridge power stroke in muscle has variable geometry. Proc. Natl. Acad. Sci. USA 96, 465–470. Welch, P. D. (1967). Use of Fast Fourier Transform for Estimation of Power Spectra - a Method Based on Time Averaging over Short Modified Periodograms. IEEE Trans. Audio Electroacoustics 70. White, H. D., et al. (1998). A computer-controlled spraying-freezing apparatus for millisecond time-resolution electron cryomicroscopy. J. Struct. Biol. 121, 306–313. White, H. D., et al. (2003). A second generation apparatus for time-resolved electron cryomicroscopy using stepper motors and electrospray. J. Struct. Biol. 144, 246–252. Winkler, H., and Taylor, K. A. (2006). Accurate marker-free alignment with simultaneous geometry determination and reconstruction of tilt series in electron tomography. Ultramicroscopy 106, 240–254. Winkler, H., et al. (2009). Tomographic subvolume alignment and subvolume classification applied to myosin V and SIV envelope spikes. J. Struct. Biol. 165, 64–77. Wong, H. C., et al. (2004). Model-based particle picking for cryo-electron microscopy. J. Struct. Biol. 145, 157–167. Yan, X., et al. (2007). Ab initio random model method facilitates 3D reconstruction of icosahedral particles. J. Struct. Biol. 157, 211–225. Yoshioka, C., et al. (2007). Automation of random conical tilt and orthogonal tilt data collection using feature-based correlation. J. Struct. Biol. 159, 335–346. Yoshioka, C., et al. (2010). Cryomesh: A new substrate for cryo-electron microscopy. Microsc. Microanal. 16, 43–53. Yu, X., et al. (2008). 3.88 A structure of cytoplasmic polyhedrosis virus by cryo-electron microscopy. Nature 453, 415–419. Zemlin, F. (1979). A practical procedure for alignment of a high resolution electron microscope. Ultramicroscopy 4, 241–245. Zemlin, F., et al. (1978). Coma-free alignment of high resolution electron microscopes with the aid of optical diffractograms. Ultramicroscopy 3, 49–60. Zhang, P., et al. (2001). Automated data collection with a Tecnai 12 electron microscope: Applications for molecular imaging by cryomicroscopy. J. Struct. Biol. 135, 251–261. Zhang, P., et al. (2003). Automated image acquisition and processing using a new generation of 4K 4K CCD cameras for cryo electron microscopic studies of macromolecular assemblies. J. Struct. Biol. 143, 135–144. Zhang, X., et al. (2008). Near-atomic resolution using electron cryomicroscopy and singleparticle reconstruction. Proc. Natl. Acad. Sci. 105, 1867–1872. Zhang, J., et al. (2009). JADAS: A customizable automated data acquisition system and its application to ice-embedded single particles. J. Struct. Biol. 165, 1–9.

338

Dmitry Lyumkis et al.

Zheng, S. Q., et al. (2007a). UCSF tomography: An integrated software suite for real-time electron microscopic tomographic data collection, alignment, and reconstruction. J. Struct. Biol. 157, 138–147. Zheng, S. Q., et al. (2007b). Automated acquisition of electron microscopic random conical tilt sets. J. Struct. Biol. 157, 148–155. Zheng, S. Q., et al. (2009). Dual-axis target mapping and automated sequential acquisition of dual-axis EM tomographic data. J. Struct. Biol. 168, 323–331. Zhou, Z. H., et al. (1996). CTF determination of images of ice-embedded single particles using a graphics interface. J. Struct. Biol. 116, 216–222. Zhu, Y., et al. (2003). Automatic particle detection through efficient Hough transforms. IEEE Trans. Med. Imaging 22, 1053–1062. Zhu, Y., et al. (2004). Automatic particle selection: Results of a comparative study. J. Struct. Biol. 145, 3–14.

Author Index

A Abeysinghe, S., 8, 14 Abramson, D. H., 256 Accola, M. A., 279 Adolfsen, F., 33, 207 Adrian, M., 122, 182–183, 207, 259, 270 Aebi, U., 102, 183 Agard, D. A., 198, 272–273 Agasse, A., 94 Agrawal, R. K., 163, 166, 196 Agre, P., 93–97, 99, 101–103, 110 Ahn, J., 269 Aihara, T., 204 Aiken, C., 269, 279 Akerman, M. E., 254 Akey, C. W., 259 Akhavan, D., 96 Aksyuk, A. A., 83 Al-Amoudi, A., 33, 207, 247–248 Alber, F., 62, 69, 329 Alberts, B., 174 Allaway, G. P., 279 Allen, G. S., 163 Aloy, P., 56–57 Altschul, S. F., 13 Amann, K. J., 37, 40, 208–209 Amat, F., 37 Ambudkar, S. V., 94 Amini, S., 33, 207 Amiry-Moghaddam, M., 96, 99 Amos, L. A., 99, 122–123, 128, 134 Anderson, K. L., 206, 208 Andrees, L., 32, 207 Andrews, S., 94 Angenitzki, M., 208 Arkin, I. T., 96 Armache, J.-P., 162 Arnal, I., 134 Arnold, K., 13 Arthos, J., 268, 282–284 Arthur, L. O., 269, 278 Asturias, F. J., 179, 182–183, 195 Atanasov, I., 274 Auer, M., 36 Auinger, S., 208

B Bajaj, C., 6–7, 35–36 Baker, D., 6–7, 10, 13, 19, 68 Baker, M. L., 1, 35, 68, 84–85 Baker, M. R., 1 Baker, T. S., 6, 8, 122–123 Baldi, P., 14 Baldwin, J. M., 2, 92, 99, 122–123 Baldwin, P. R., 6–7, 10, 282, 319 Banks, J., 35 Ban, N., 84 Ban, Y. E., 283 Baraban, J. M., 96 Barash, D., 35 Barber, J. D., 14 Barmann, M., 33 Barnes, S., 269 Barr, F. A., 256 Bartesaghi, A., 36, 39–40, 269–270, 273–274, 277, 280, 282–286 Barton, G. J., 14 Bartonova, V., 269 Bates, P. A., 13 Batth, T. S., 3, 5, 18 Bauerlein, F. J., 207 Baumeister, W., 2, 32–34, 38–39, 41, 74, 102, 124, 207–208, 248, 269–270, 272, 276–277, 286 Baxter, W. T., 33–34, 162, 311, 315 Becker, T., 48, 162 Beck, F., 207, 272 Beck, M., 34, 208, 215, 252, 258–259, 276, 286 Beckman, E., 123 Beckmann, E., 2, 92, 99 Beckmann, R., 162 Behera, R. N., 112 Bel, B., 33, 207 Bellamy, R., 3, 5 Belnap, D. M., 198, 269, 286, 311 Benedetti, E. L., 95 Ben-Harush, K., 34, 207–208, 246, 259 Benjamin, J., 269 Bennett, A. E., 269, 286 Benveniste, R. E., 278 Berger, J. M., 268 Berger, U. V., 99

339

340 Berkovitch-Yellin, Z., 39 Berman, H. M., 76 Berneche, S., 107–108 Bernstein, F. C., 76 Berriman, J., 297 Bess, J. W. Jr., 269–270, 278 Best, C., 32, 34 Betts, M. J., 33, 207 Betzig, E., 254 Beucher, S., 36 Beulshausen, S., 99 Bhakat, P., 100 Bhat, R. V., 96 Biasini, M., 103–104 Biegert, A., 54 Biesele, J. J., 256 Birmanns, S., 6 Bishai, W. R., 97 Blanchard, S. C., 162–163, 174 Blanchoin, L., 204 Blanc, N. S., 33 Blommers, M. J., 137 Blundell, T. L., 13, 53 Bobkov, A., 37 Bo¨hm, J., 38, 216 Bohm, U., 207 Boisset, N., 183 Bok, D., 95 Bokoch, G. M., 206, 208 Bonhoeffer, T., 207 Bonnet, N., 318 Booth, C. R., 22, 39–40 Booy, F. P., 100 Bordoli, L., 13 Borgnia, M. J., 97, 270, 273–274, 280, 283–286 Borhani, D. W., 96 Borisy, G. G., 205 Bortz, E., 274 Bosch, J., 84 Bos, E., 33, 207 Botkin, D. J., 270, 273–274, 286 Bo¨ttcher, B., 3, 25 Boulanger, P., 268, 279 Bouley, R., 96 Bourque, C., 96 Bouwhuis, M., 38 Bowman, B. R., 13 Bradley, P., 10, 13, 19 Brandt, F., 232, 236 Braunfeld, M. B., 272 Braun, T., 97–98, 101, 105–106 Brega, M., 37 Brenner, S. E., 51 Bretaudiere, J. P., 187 Briegel, A., 247 Briggs, J. A., 268–270 Brignole, E. J., 179 Brimacombe, R., 171

Author Index

Brink, J., 182 Brohawn, S. G., 49 Bron, P., 268 Brooks, C. L., 6–7 Brown, D., 96 Brown, S., 254 Brunger, A. T., 6–7, 65 Bru¨nner, M., 168–170 Bryson, K., 14 Bui, K. H., 123, 137–138 Burgess, S. A., 183, 190, 197–198 Burkhardt, N., 163 Burton, D. R., 268, 282–284 Butcher, S. J., 78 Buttle, K. F., 32, 34, 208 Byeon, I. J., 269 Byers, B., 256 C Cai, G., 189, 196, 198 Calamita, G., 97 Callan, H. G., 259 Campbell, I. D., 246 Cantele, F., 37 Caplow, M., 129 Carazo, J.-M., 35, 41, 164, 318 Carbrey, J. M., 99 Cardona, A., 306 Cardone, G., 220 Carlier, M. F., 204 Carlomagno, T., 137 Carlsen, K. A., 41 Carragher, B., 272–273, 291 Carrascosa, J. L., 18, 33, 207 Carriere, C., 279 Carroll, T. P., 93–94 Casagrande, F., 108–109 Castano-Diez, D., 329 Castillo, A., 279 Casto´n, J. R., 18 Cate, J. H., 84 Caujolle-Bert, D., 103 Cavacini, L., 268 Cavalier, A., 210 Ceska, T. A., 2, 92, 99, 123 Chaban, Y., 198 Chabrie`re, E., 224 Cha´con, P., 60 Chalfie, M., 252 Chami, M., 103–104, 108–109 Chan, D. C., 268 Chang, J.-J., 33, 80, 207, 270 Chapman, M. S., 60 Chaumont, F., 94 Chazal, N., 279 Chen, B., 268 Chen, C. H., 279

341

Author Index

Chen, D. H., 3, 5, 8, 14, 16 Cheng, A., 94–95, 101, 110, 269, 272–273, 291 Cheng, L., 6–7 Cheng, N., 3, 25 Cheng, Y., 2, 92, 94–96, 101–102, 105–109, 111, 183, 320, 328 Chen, H., 272 Chen, J. Z., 166, 180, 317–318, 328 Chen, L., 268, 283 Chen, S., 247 Cherepanova, O., 204 Chernomordik, L. V., 152 Chertova, E. N., 270, 273–274, 278 Chettova, E., 269 Chhabra, E. S., 204 Chiu, W., 3, 5–8, 10–11, 13–14, 16, 19, 22, 35–37, 68, 78, 162, 204, 282 Chivian, D., 10, 13, 19 Chorny, I., 97 Chre´tien, D., 129, 132–133 Chrispeels, M. J., 93, 98 Christensen, A. M., 269 Chuang, D. T., 3, 5, 8, 14, 16 Cline, J., 95 Cole, C., 14 Coleman, R. A., 180 Comaniciu, D., 35 Comolli, L. R., 37 Concel, J., 269 Cong, Y., 3, 5, 18, 22, 82, 180, 328 Connell, J. W., 256 Connell, S. R., 48, 162, 165, 168–171 Conway, J. F., 3, 25 Cooper, J. A., 204–205 Cornett, B., 123, 137 Costantino, D., 172 Costin, A. J., 32 Couch, G. S., 6–7, 10, 23, 172, 275, 279–280, 283 Coudray, N., 301, 306 Cowan, S. W., 6, 8 Cowtan, K., 6, 8, 10, 81 Crepeau, R. H., 122 Cronshaw, J. M., 258 Cross, R. A., 134 Crowther, R. A., 3, 25, 33, 100, 108, 125, 218, 311, 320 Cuesta, I., 198 Cullis, P. R., 152 Curmi, P. A., 133 Cyrklaff, M., 139 D Dabrowski, M., 162, 168, 170–171 Dai, W., 274 Dalke, A., 6 Dal Martello, M. F., 41

Danev, R., 198, 329 Daniel, H., 108 Daniels, B. V., 97 Daniels, M. J., 98 Danuser, G., 206, 208, 210 Davidson, K. G., 96 DeBonis, S., 134 De Carlo, S., 198 Deen, P. M., 95, 98 Deerinck, T. J., 253 de Groot, B. L., 95, 97–99, 101, 105–107, 110 de Haas, F., 268 De Kruijff, B., 152 Delano, W. L., 81 Delong, A., 298 Delorme, V., 206, 208 Delrot, S., 94 DerMardirossian, C., 206, 208 DeRosier, D. J., 130, 204, 253 de Ruijter, W. J., 305 Desai, P., 269, 286 Devidas, S., 99 Devkota, B., 6–7 Dey, B., 268, 282–284 Diaconu, M., 83 Dierksen, K., 246, 305 Diez, D. C., 33, 207 DiMaio, F., 1, 68, 85 Dimitrov, D. S., 268, 282–284 Ding, H. J., 269, 273, 280–281 Djuranovic, S., 54 Dockstader, J., 95 Doenhoefer, A., 168–170 Dokland, T., 84 Dominguez, R., 204–205 Dong, G., 3, 5 Dong, M., 222 Dormitzer, P. R., 3, 5 Dougherty, M., 3, 5, 7, 18, 21–22 Douglas, C. C., 269 Douglas, N. R., 3, 18, 21–22 Downing, K. H., 2, 33, 92, 99–100, 121, 130, 271 Drahos, V., 298 Dror, R. O., 96 Dubaquie, Y., 225 Dubochet, J., 33, 122, 207, 251, 270, 296 Dueck, D., 319 Dunham, C., 171 Dunia, I., 95 Dunlop, N., 282 Dutta, S., 75 Dux, L., 147 E Eastwood, M. P., 96 Echard, A., 256 Ecke, M., 208

342

Author Index

Edelstein, S. J., 122 Edstrom, A., 122–123 Egea, P. F., 97 Egelman, E. H., 76, 204 Eggermont, P. P. B., 35, 164 Eggert, U. S., 256 Egile, C., 37, 40, 208–209 Ehrenberg, M., 162–163, 171 Eibauer, M., 207 Ekvall, M., 98 Elad, N., 187, 245 Elangovan, V., 36 Eletr, S., 148 Elkjaer, M., 99 Ellisman, M. H., 34, 36–37, 41 Elmlund, H., 198 Elofsson, A., 13 Emmett, M. R., 269 Emsley, P., 6, 8, 10, 81 Engel, A., 91 Eramian, D., 5 Ericksson, G., 35 Ermantraut, E., 298 Esser, M. T., 278 Estes, M. K., 5 Eswar, N., 51, 53–54, 64 Euteneuer, U., 256 Evans, J. A., 204 F Fabiola, F., 60 Fain, N., 37 Fang, P., 96 Fang, Q., 6–7 Farrow, N. A., 320 Faruqi, A. R., 329 Fass, D., 268, 276 Fellmann, D., 327 Ferna´ndez, J. J., 18, 33, 35–36, 85, 207, 219, 315, 320, 322 Ferrer, M., 225 Ferrin, T. E., 6–7, 10, 23, 37, 172, 275, 279–280, 283 Fields, S., 56 Fillmore, C., 269, 273, 280–281 Finch, J. T., 269 Finley, D., 49 Fischer, D., 13 Fischer, R. S., 206, 208 Fisher, L., 143 Fletterick, R. J., 134 Fo¨rster, F., 32, 38–40, 47, 215, 259, 269–270, 272, 276–277, 286 Fotiadis, D., 96, 98, 101, 105–106, 108–109 Fotin, A., 80 Fowler, V. M., 206, 208 Fram, E. K., 122

Francis, N., 204 Frangakis, A. S. R., 32–33, 35, 37–40, 207, 260, 269, 275, 277 Frank, J., 2, 33–34, 36, 39–40, 79, 162–166, 170–172, 184–187, 196, 207–208, 276, 306, 311, 316–318, 320–321, 328 Fraysse, L. C., 98 Frederik, P. M., 296 Freed, E. O., 268, 279 Frenkiel-Krispin, D., 258 Frey, B. J., 319 Frey, T. G., 34, 36 Frick, A., 98 Frickey, T., 53 Frigon, R. P., 132 Frkiaer, J., 99 Frydman, J., 22 Fucini, P., 165, 168–170 Fu, C. J., 3, 18, 21–22 Fu, D., 95, 97, 110 Fu, J., 198 Fujimoto, L. M., 37 Fujimoto, Z., 3 Fujiyoshi, Y., 2, 18, 91, 123, 125 Fuller, S. D., 75, 132–133, 268–270, 279 Fung, J. C., 305 Furman, C. S., 96 Fushimi, K., 98 G Gabashvili, I. S., 74, 166 Galkin, V. E., 204 Gamm, B., 198 Ganesh, T., 136 Gan, L., 246 Ganser, B. K., 269 Ganser-Pornillos, B. K., 269 Gao, H., 33–34, 163–164, 198 Gardun˜o, E., 37, 41 Gartmann, M., 162 Garvalov, B. K., 249 Gaskin, F., 122 Gatz, R., 207 Gaub, H. E., 124 Gaudette, R., 123 Gautier, R., 18 Gay, B., 279 Geerts, W. J., 38 Gelderblom, H. R., 268 Gensen, G. J., 286 Gerisch, G., 207–208, 269 Gerlich, D. W., 256 Gero´s, H., 94 Gerstein, M., 136 Geuze, H. J., 251 Gheorghe, K. R., 100 Giepmans, B. N., 253

343

Author Index

Giesebrecht, J., 161 Gigant, B., 133 Gilula, N. B., 105 Ginalski, K., 13 Gingras, A. R., 329 Gipson, B., 108 Gish, W., 13 Glaeser, R. M., 33, 100, 124, 182, 271, 296 Glass, B., 269 Glaves, J. P., 143 Glotzer, M., 256 Gnaegi, H., 33, 207 Gobin, R., 96 Goddard, T. D., 6–7, 10, 23, 37, 61, 63, 85, 172, 194, 275, 279–280, 283 Goila-Gaur, R., 279 Golas, M. M., 183 Goldie, K. N., 208 Gomes, D., 94 Gomez-Llorente, Y., 35 Go, N., 13 Goncharov, A., 317 Gonen, T., 2, 74, 94–99, 101–102, 105–109, 111–112 Gong, H., 268 Gorelick-Feldman, D. A., 96, 99 Gorin, M. B., 95 Gorlich, D., 259 Gosavi, S., 172 Gossard, D. C., 6–7, 10, 36–37 Gottlinger, H. G., 279 Govindaraju, N. K., 329 Gowen, B. E., 268 Grassucci, R. A., 33–34, 162–163, 166, 251, 276 Grattinger, M., 279 Greenblatt, D. M., 6–7, 10, 23, 172, 275, 279–280, 283 Green, D. M., 41 Green, N. M., 148 Griesinger, C., 137 Grigorieff, N., 3, 5, 85, 108–109, 166, 315, 317, 321–322 Grimes, J. M., 84 Grimm, R., 33, 247, 270, 272 Grise, H., 270 Grob, P., 198 Gromley, A., 256 Gronenborn, A. M., 269 Gross, A., 32, 41 Gross, H., 96 Gross, I., 268, 279 Grubmu¨ller, H., 95, 110 Gruneberg, U., 256 Gru¨newald, K., 32, 41, 48, 269–270, 286 Gruska, M., 33–34, 207 Guggino, W. B., 93–97, 99 Guliaev, A. B., 96 Gupton, S. L., 206, 208

Gursky, R., 163 Gyobu, N., 97, 100, 102–106, 108, 125 H Haase, W., 101 Habeck, M., 69 Hachey, D. L., 279 Halic, M., 162 Hallgren, K., 98 Hall, R. J., 316 Hamilton, S. L., 5–7, 10, 14 Han, B.-G., 95–96, 100, 110, 215 Hanein, D., 6, 32, 36–38, 40, 60–61, 194, 203–212 Hankamer, B., 35 Hanley, J. A., 41 Hanna, S. L., 83 Hanson, P. I., 222 Harauz, G., 274, 321 Hara, Y., 98 Harder, D., 108 Harley, E. M., 41 Harlow, M. L., 36 Harries, W. E., 96–97 Harris, J. R., 183 Harrison, C. B., 6–7 Harrison, S. C., 2–3, 5, 94, 96, 105, 109, 111–112, 268 Harris, W. J., 298 Hartmann, T., 124 Hartshorn, M. J., 81 Harvey, S. C., 6–7 Hasler, L., 96, 102 Hausmann, E. H., 268 Hayward, S. B., 100 Heagle, A. B., 166 Hebert, H., 100 Hedfalk, K., 98 Hediger, M. A., 99 Heel, M., 186 Hefti, A., 102 Hege, H.-C., 37 Hegerl, R., 34–35, 37–39, 48, 208, 272, 277 Heim, R., 252 Heitz, G., 37 Hekking, L. H., 38 Helgstrand, C., 84 Hell, S. W., 254 Henderson, G. P., 246–247, 249 Henderson, L. E., 278 Henderson, R., 2, 67, 74, 76, 92, 99–100, 108, 122–123, 125, 136, 180, 220, 322, 327, 329 Hendrickson, W. A., 268, 285 Hendry, R. M., 282 Henikoff, S., 230 Henrick, K., 75 Herman, G. T., 164

344 Herold, A., 291–330 Hertzberger, L. O., 38 Hessell, A. J., 268, 282–284 Heuser, J. E., 254 Heuser, T., 137, 250 Heymann, J. B., 33, 77, 94–95, 101–103, 110, 198, 269, 278, 286, 311 Higashi, T., 3 Higgs, H. N., 204 Hildebrand, P. W., 168–170 Hill, C. P., 269 Hirai, T., 18, 92, 94–95, 100–103, 110 Hirata, Y., 98 Hiroaki, Y., 2, 94, 96–97, 102–106, 108–109, 111, 125 Hirokawa, N., 134 Hirose, K., 134 Hitchcock-DeGregori, S. E., 206, 208 Hite, R. K., 91 Hockley, D. J., 268 Hodgkinson, J. L., 80 Hoenger, A., 134, 208 Hoffmann, C., 207 Ho¨fte, H., 93 Hoglund, S., 279 Hohenberg, H., 279 Hohn, M., 311 Ho, J. D., 97 Holmes, K. C., 204 Homo, J. C., 207, 270 Honig, B., 6–7 Horowitz, M., 37 Horwitz, J., 95 Hoshino, A., 254 Hou, E., 291–330 Howarth, M., 253 Howell, K. E., 32, 37 Hryc, C. F., 1 Hsieh, C. E., 33, 207, 248 Hsin, J., 81 Huang, C. C., 6–7, 10, 23, 37, 172, 268, 275, 279–280, 283 Huang, C. Y., 35 Huang, L., 279 Huang, Y., 96 Huang, Z., 315–316 Hua, Y., 269 Hubert, D. H. W., 296 Hub, J. S., 95 Hudson, C. S., 96 Hudspeth, A. J., 36 Hu, G. B., 153 Huiskonen, J. T., 78 Hu, J., 279 Hu, M., 301, 306, 308 Humbel, B. M., 210 Humphrey, W., 6 Hyman, A., 207

Author Index I Iacovache, I., 103–104, 296 Iancu, C. V., 33, 246 Iizuka, R., 22 Inesi, G., 148, 150 Innis, C. A., 162 Irikura, D., 101 Irving, C., 291 Ishida, H., 84 Ishii, H., 225 Ishikawa, T., 123, 134, 137–138 Iwasa, M., 204 J Jacob, S., 207–209 Jacovetty, E. L., 291 Jakana, J., 3, 5, 8, 16, 18, 21–22, 162 Jakobsson, P.-J., 100 Janssen, M. E., 37 Jap, B. K., 94–96, 101–102, 110 Jaynes, E. T., 60 Jegerscho¨ld, C., 100 Jensen, G. J., 32–33, 74, 207, 246, 269, 272–273, 280–281 Jensen, M. ., 96, 99 Jensen, U. B., 99 Jiang, J., 97 Jiang, M., 37 Jiang, W., 3, 5–8, 10, 13, 16, 22, 35, 61, 85, 328 Jia, Q., 274 Ji, L., 206 Jimenez, N., 38 Jingling, L., 41 Jin, L., 3, 16 Ji, Q., 37, 41 Job, D., 129 Johanson, U., 98, 101, 105–106 John, B., 6–7 Johnson, J. E., 78 Johnson, K. D., 93 Johnson, M. C., 101, 268 Jones, D. T., 14, 52 Jones, I. M., 268 Jones, T. A., 6, 8, 13 Jontes, J., 123 Joshi, A. K., 182 Jourdain, I., 133 Joyeux, L., 317 Jung, J. S., 96, 269 Ju, T., 3, 6, 8, 11, 14 K Kabsch, W., 204 Kaelin, W. G. Jr., 56 Kamegawa, A., 97, 102–106, 108, 125 Kanaoka, Y., 101

345

Author Index

Karlsson, M., 98 Karplus, M., 13, 65 Karsenti, E., 132–133 Kato, T., 328 Kaufmann, B., 80 Kaufmann, T. C., 103 Keegstra, W., 311 Kellenberger, E., 183 Keller, A., 230, 232 Kelley, A. C., 162, 168–171 Kelley, L. A., 13 Kelly, B. N., 269 Kelly, D. F., 298 Khademi, S., 96 Khandelia, H., 98 Kidera, A., 18, 92, 100, 110 Kieffer, C., 269, 273, 280–281 Kieft, J. S., 172 Kikkawa, M., 134 Kilgore, N. R., 279 Kim, D. E., 10, 13, 19 Kim, E., 37 Kimmel, R., 36 Kimoto, K., 305 Kim, P. S., 268 Kimura, K., 97, 102, 104–106, 108 Kimura, Y., 18, 92, 100, 110 King, J., 3, 8, 16 Kirchberger, M., 150 Kirschner, M. W., 254 Kisseberth, N., 305 Kistler, J., 95–96, 101–102, 105–108 Kitamoto, T., 182 Kjeldgaard, M., 6, 8 Kjellbom, P., 98, 101, 105–106 Kleiman, L., 279 Kleywegt, G. J., 13 Klinke, S., 224 Klishko, V. Y., 269 Klug, A., 122 Klussmann, E., 99 Knoers, N. V., 98 Knossow, M., 133 Kobilka, B. K., 112 Koch, M. A., 268 Koduri, R., 282 Koestler, S., 208 Kofler, C., 32, 38 Kole, T. P., 206, 208 Koller, D., 37 Komeili, A., 248 Kong, L., 283 Kong, Y., 6, 18 Koning, R. I., 34, 246 Konorty, M., 246 Kopp, J., 13 Korber, B., 268 Korn, E. D., 204

Kossel, A. H., 207 Kostek, S. A., 198 Koster, A. J., 34, 36, 38, 246, 272, 277, 305 Ko¨stler, D., 305 Kostyuchenko, V. A., 80 Kowal, J., 103–104 Kozono, D., 99 Krahn, J. M., 123, 137 Krausslich, H. G., 268–269, 279 Krell, J., 137 Krementsova, E. B., 274 Kremer, J. R., 36, 208, 273–275, 278 Kress, Y., 122 Krucinski, J., 95, 97, 110 Kubicek, K., 137 Kuhajda, F. P., 180 Ku¨hlbrandt, W., 97, 100, 123, 145, 180, 295 Kuhner, S., 222 Kuhn, R. J., 83 Kukulski, W., 98, 101, 103–106 Kumar, N. M., 105 Kumar, R. N., 3, 5, 18 Kuriyama, R., 257 Ku¨rner, J., 39, 246, 248 Kutay, U., 259 Kwong, P. D., 268, 285 Kwon, T., 99 Kwon, Y. D., 268, 283 L Lacapere, J. J., 103, 147 Lachkar, S., 133 Ladinski, M., 247 Ladjadj, M., 166 Lake, J. A., 319 Lambert, O., 185 Lam, B. K., 101 Lambooy, P. K., 204 Lam, T. T., 269 Lander, G. C., 189, 308, 311, 319, 326 Landis, D. M., 96 Landsberg, M. J., 35 Lanman, J., 269 Lanzavecchia, S., 37 Larsson, H., 122–123 Lasker, K., 6–7, 59, 69 Laugks, T., 207 Lau, P.-W., 291 Lawrence, M. C., 33, 219 Lawson, C. L., 73 Leadbetter, J. R., 286 LeBarron, J., 162 Lebart, M. C., 204 Lebbink, M. N., 38 Ledbetter, M. C., 122 Ledizet, M., 83 Lee, D. P., 180

346 Lee, J. K., 95, 110 Lee, K. K., 78 Lee, S. H., 204–205 Lefman, J., 301 Leforestier, A., 33 Lei, J., 306 Leiman, P. G., 4 Leis, A. P., 32–34, 38, 207, 220 Leis, J. W., 34 Leith, A., 33, 36, 39, 166 Leitner, A., 57 Lepault, J., 122, 207, 270 Leschziner, A. E., 320 Lescoute, A., 162, 168, 170–171 Leslie, A. G., 3 Levitt, M., 3, 6–7, 18, 21–22 Levy, D., 103, 147, 149 Liang, C., 279 Libson, A., 95, 97, 110 Li, F., 279 Lifson, J. D., 269–270, 278 Li, H. L., 94, 101, 123, 125–126, 130, 137 Likas, A., 164 Lim, J., 206 Lim, R. Y., 258 Linaroudis, A. A., 39, 272 Lin, T., 270, 273–274, 286 Lipman, D. J., 13 Li, R., 37, 40, 208–209 Li, S., 35, 269 Liu, H., 37 Liu, J., 39–40, 267–286 Liu, Q., 57 Liu, T., 253 Liu, X., 318 Li, X., 274 Li, Y., 36, 166 Li, Z., 5, 32, 94, 96, 108, 207, 246 Lockhart, A., 134 Loerke, J., 161 Loftus, G. R., 41 Loftus, T. M., 180 Lopez, J. M., 108–109 Lorentzen, E., 57 Lo¨we, J., 33, 67, 123, 125–126, 128, 207 Lucic, V., 207, 246–247, 253, 270, 276, 286 Ludtke, S. J., 3, 5–8, 10, 14, 16, 18, 21–22, 162, 282, 311, 321–322, 328 Lu, G., 3, 5 Lunin, V. Y., 40 Lupas, A. N., 53, 67 Lupu, R., 180 Luque, D., 18 Luther, P. K., 33 Luu, D. T., 98 Lu, Z., 297 Lyumkis, D., 291

Author Index M MacCallum, R. M., 13 Machacek, M., 206, 208, 210 Mackenzie, B., 99 Mader, A., 245 Maeda, Y., 204 Maier, T., 182, 193 Maimon, T., 34, 207–208 Maiolica, A., 57 Ma, J., 6, 8 Majeed, S., 268, 283 Malladi, R., 36 Mallick, S. P., 315–316 Malmstrom, J., 221, 233, 237 Malmstro¨m, L., 10, 13, 19 Maloney, L. T., 41 Maloney, P. C., 97 Mandelkow, E.-M., 123, 129, 131 Manduchi, R., 35 Manenti, S., 95 Mann, D. S., 6, 10 Mannella, C. A., 33, 207 Mannherz, H. G., 204 Marabini, R., 311, 318 Mariani, V., 107–108 Marko, M., 33–34, 207–208, 247–248, 260 Marshall, A. G., 269 Marshall, R. M., 36 Marsh, B. J., 32, 34, 36–38 Marsh, M. P., 305–306 Martin, D. E., 279 Marti-Renom, M. A., 6–7, 50 Martonosi, A., 147 Marumo, F., 98 Maruyama, T., 22 Massiah, M. A., 269 Mastronarde, D. N., 32, 34, 36, 208, 257, 272–275, 278, 305–306 Ma, T., 96 Matallana, C., 279 Matsudaira, P., 204 Matsumoto, A., 84 Matsushima, M., 18, 100 Maupetit, J., 18 Maurel, C., 98 Mayr, M., 37 McCammon, J. A., 6, 65, 172 McCann, M. C., 98 McCrum, E., 270, 273–274, 286 McDonald, K. L., 207, 256 McDowall, A. W., 33, 35, 122, 207, 246, 248, 270 McEwen, B. F., 33, 37, 122, 271 McGuffin, L. J., 14 McIntosh, J. R., 32, 36–37, 123, 208, 256–257, 273–275, 278 McMahan, U. J., 36

347

Author Index

McNeil, B. J., 41 Medalia, O., 32–34, 41, 207–208, 245, 269, 276, 286 Meiler, J., 10, 13, 19, 137 Melanson, L. A., 296 Melero, R., 41 Melnyk, P., 95, 101, 110 Menendez, J. A., 180 Meng, E. C., 6–7, 10, 23, 172, 275, 279–280, 283 Meng, X., 269 Mercogliano, C. P., 253 Metoz, F., 134 Meury, M., 108 Meyer, A. S., 22 Meyer, F., 36 Mielke, T., 162, 168–171 Miercke, L. J., 95–97, 110 Miki, K., 22 Mikolajka, A., 168–170 Milazzo, A. C., 198 Miller, E. J., 3, 5, 18 Miller, W., 13 Milligan, R. A., 6, 123, 129, 131, 134, 172 Milne, J. L., 246 Mindell, J. A., 166, 315 Ming, D., 182 Misra, M., 147–148 Misura, K. M., 10, 13, 19 Mitra, A. K., 94–95, 101, 110 Mitra, K., 79, 172 Mitsuma, T., 97 Mitsuoka, K., 18, 92, 94–95, 97, 100–106, 108, 110, 125 Miyashita, O., 6–7 Miyazaki, N., 3 Miyazawa, A., 18, 92, 100, 110, 180 Mizoguchi, A., 97, 102, 104–106, 108 Mizuguchi, K., 13 Mizuno, H., 3 Mizutani, T., 148 Moberlychan, W., 207 Modis, Y., 83 Moeller, A., 291 Mollinari, C., 256 Moncoq, K., 156–157 Monnens, L. A., 98 Montabana, E., 268 Moor, H., 247–248 Morais, M. C., 4 Morgan, G. P., 32 Morgenstern, R., 100 Morgunova, E., 224 Morth, J. P., 148 Mosser, G., 103 Mouritsen, O. G., 98 Moussavi, F., 37 Movassagh, T., 123, 137–138 Muench, S. P., 80

Mukhopadhyay, A., 3, 5, 18 Mukhopadhyay, S., 83 Mulder, A. M., 198, 291 Mulders, S. M., 95 Muller, B., 279 Mu¨ller, D. J., 35, 96 Muller-Reichert, T., 207 Mu¨ller, S. A., 99, 105, 107 Muller-Taubenberger, A., 259 Mullins, J. M., 256–257 Mullins, R. D., 204 Munro, J. B., 162–163, 174 Murata, K., 18, 92, 94–95, 100–103, 110 Murayama, K., 162 Murphy, F. V., 162, 168–171 Murphy, G. E., 74, 286 Myers, E. W., 13 Myster, S. H., 249 N Nagayama, K., 329 Nagelhus, E. A., 96 Naitow, H., 3 Nakagawa, A., 3 Nara, P. L., 282 Narita, A., 204 Natarajan, P., 81 Navaza, J., 6–7 Navia, M. A., 184 Nedvetsky, P. I., 99 Neely, J. D., 96 Nejsum, L. N., 99 Nemethova, M., 207–209, 252 Nermut, M. V., 268 Nettles, J. H., 123, 137 Net, W. D., 272 Neujahr, R., 208 Neutze, R., 98 Nguyen, H. T., 37, 41 Nicastro, D., 37–38, 40, 123, 137, 207–209, 249–250, 269 Nicholson, W., 130 Nickell, S., 32, 34, 37–38, 40, 49–50, 208–209, 216, 218, 252, 272, 305–306, 308 Nielsen, S., 96, 99 Nierhaus, K. N., 163, 165, 168 Nishikawa, K., 97, 102, 104–106, 108 Nisman, R., 253 Nissen, P., 162 Nitsch, M., 277 Noel, J. K., 172 Nogales, E., 80, 92, 100, 121, 320, 327 Nollert, P., 95, 97, 110 Nomura, H., 148 Norlen, L. P., 33 Norris, S. J., 270, 273–274, 286 Noske, A. B., 32

348

Author Index

Nunez-Ramirez, R., 35 Nybakken, G. E., 83 Nyblom, M., 98 O Ochoa, W. F., 321 O’Connell, J. D. R., 97 Oda, T., 204 O’Dell, S., 268 Oesterhelt, D., 136 Ofek, G. A., 270 Ogawa, A., 3 Ogura, T., 316 Ohi, M., 183, 295 Oiwa, K., 123, 137–138 Okorokov, A. L., 328 Olesen, C., 148 Olofsson, A., 103 Olson, A. J., 93 Olson, N. H., 74 Omura, T., 3 Onuchic, J. N., 168–170, 172 Orenstein, J. M., 279 Orlova, A., 204 Ortiz, J. O., 39, 228, 232, 236 Orzechowski, M., 172 Osher, S., 36 Oshima, A., 105 Otegui, M. S., 256 O’Toole, E. T., 207 Ottensmeyer, F. P., 320 Ottersen, O. P., 96 Ozel, M., 268 P Pai, E. F., 204 Pailthorpe, B., 35 Palacin, M., 108–109 Pancera, M., 268, 283 Pan, J., 321 Pannekoek, Y., 225 Pantaloni, D., 204 Pantelic, R. S., 35, 298 Pao, G. M., 93 Parent, K. N., 321 Parren, P. W., 282 Paschall, C., 269 Pascual, A., 187 Pascual-Montano, A., 318 Pastor-Soler, N., 96 Patla, I., 34, 207–208 Patwardhan, A., 316 Pauli, G., 268 Paweletz, N., 256 Pawelzik, S., 100 Pawley, J. B., 100 Pearson, K., 318

Pearson, W. R., 13 Pebay-Peyroula, E., 39 Pedersen, B. P., 148 Penczek, P. A., 34, 40, 162–163, 165–166, 168–171, 185, 188, 208, 276, 316–317, 319–321 Peng, L., 6, 10 Perkins, G. A., 34, 36 Peters, P. J., 33, 207 Petry, S., 171 Pettersen, E. F., 6–7, 10, 23, 54, 81, 85, 172, 275, 279–280, 283 Pfingsten, J. S., 172 Philippsen, A., 97, 99, 105, 107–108, 320 Pieper, U., 224 Pierson, J., 33, 123, 207 Pigino, G., 37 Pintilie, G. D., 6–7, 10, 36–37, 85 Pires, R., 329 Plaisier, J. R., 316 Plitzko, J. M., 207, 247, 272 Podjarny, A. D., 39 Pollard, T. D., 37, 40, 183, 204–205, 208–209 Pollastri, G., 14 Ponti, A., 206, 208 Pool, M. R., 162 Pornillos, O., 269 Porter, K. R., 122 Porter, M. E., 123, 137 Post, J. A., 38 Potter, C. S., 272–273, 291 Potterton, L., 13 Praetorius, J., 99 Prasad, B. V., 5, 84 Preston, G. M., 93–97 Prevelige, P. E. Jr., 269 Pruggnaller, S., 37, 39–40, 232, 277 Przybylski, D., 14 Pulokas, J., 272–273, 291 Purhonen, P., 100 Pyati, J., 282 Q Qian, B., 10, 13, 19 Qin, S., 58 Qin, Y., 165, 168 Quiocho, F. A., 13 Quispe, J. D., 291, 298 R Rachel, R., 34, 208 Radermacher, M., 124, 166, 183–184, 187, 193, 196, 259, 317 Radon, J., 34 Raeburn, C., 100 Raijmakers, R., 57 Ramakrishnan, V., 162, 168–171

349

Author Index

Ranck, J. L., 103 Randall, G., 39–40, 277, 282 Rash, J. E., 96 Rasmussen, S. G., 112 Ratera, M., 108–109 Rath, B. K., 39, 184, 305, 316 Ratje, A. H., 168–170 Raunser, S., 93 Ravelli, R. B. G., 133 Raytchev, M., 137 Reddick, M., 279 Redding-Johanson, A. M., 3, 5, 18 Reddy, V. S., 95, 101, 110 Reese, M., 137 Reese, T. S., 96 Rees, I., 6, 10 Reichow, S. L., 94 Reinisch, K. M., 84 Reisler, E., 204 Reissmann, S., 3, 5, 18, 21–22 Re´migy, H.-W., 103–104 Ren, G., 95, 101, 110 Resch, G. P., 207–209 Resh, M. D., 180 Ress, D. B., 36 Revel, J. P., 95 Rice, S., 134 Rice, W. J., 153 Riches, J. D., 269 Richter, K., 33 Rigaud, J. L., 103 Rigort, A., 207, 247–248, 260 Ringler, P., 97 Rinke-Appel, J., 171 Ritsert, K., 224 Rixon, F. J., 6–7, 13, 78, 162 Robbins, R. A., 97 Robinson, C. V., 32, 57 Robinson, J. E., 268, 283, 285 Robinson, J. M., 247 Robles-Colmenares, Y., 97 Rockel, B., 32, 207 Rodnina, M. V., 171 Roldan, A., 279 Roseman, A. M., 6, 228, 316–317 Rosenbaum, D. M., 112 Rosenthal, P. B., 67, 322 Rosenthal, W., 99 Rossio, J. L., 278 Rossmann, M. G., 4, 6 Rost, B., 14, 230 Rost, M., 162 Roth, M., 39 Rothnagel, R., 35 Rouiller, I., 37, 40, 208–209 Rousselet, A., 95 Rout, M. P., 258

Roux, K. H., 39–40, 269–270, 273–274, 276–277 Rual, J. F., 252 Ruhlen, R. L., 129 Ruiz, T., 183 Russell, R. B., 56–57 Russell, R. S., 279 Rust, M. J., 254 Rusu, M., 6, 84 Rutten, T., 268 Rychlewski, L., 13 S Saad, A., 67, 162 Sablin, E. P., 134 Sagara, Y., 150 Sahm, C., 41 Saibil, H. R., 83, 246 Saier, M. H. J., 93 Sakakibara, H., 123, 137–138 Sakalian, M., 269 Salamin, L. M., 33 Sali, A., 5–7, 13, 22, 32, 51, 53, 246 Salje, J., 33, 207 Salvi, E., 37 Salzwedel, K., 279 Sanbonmatsu, K. Y., 162–163, 172, 174 Sanchez-Pedregal, V. M., 137 Sandal, N. N., 93 Sandberg, K., 32, 37 Sander, B., 315 Sandstrom, A., 97 Sanejouand, Y. H., 6–7 San Martin, C., 35 Sanner, M. F., 93 Santoni, V., 98 Sapiro, G., 39–40, 270, 273–274, 277, 280, 282–286 Sartori, A., 207, 252–253, 323 Sasaki, S., 93, 97–98, 102, 104–106, 108 Sato, C., 316 Sattentau, Q. J., 270 Savage, D. F., 97 Sawyer, L. S., 282 Saxton, W. O., 305, 317 Schalek, R., 207 Schenk, A. D., 91, 103, 108 Scheres, S. H. W., 35, 41, 164, 189, 198, 318 Schermelleh, L., 254 Scheuring, S., 99, 105, 107 Scheybani, T., 102 Schief, W. R., 283 Schietroma, C., 37 Schlu¨nzen, F., 39 Schmeisser, M., 329 Schmid, M. F., 3, 7, 39–40, 74, 204 Schmidt-Krey, I., 101

350 Schnell, J. R., 101 Schonbrun, J., 10, 13, 19 Schooler, J. B., 269, 273, 280–281 Schreiner, E., 6–7 Schro¨der, G. F., 3, 6–7, 18, 21–22, 84 Schroeer, B., 162, 168, 170–171 Schueler-Furman, O., 51 Schuette, J.-C., 162, 168–171 Schug, A., 172 Schu¨ler, M., 162, 168, 170–171 Schulten, K., 6–7, 162, 172 Schultz, P., 207, 270 Schwartz, C. L., 123, 252 Schwede, T., 13 Sedat, J. W., 272 Seidelt, B., 65, 162 Sellitto, C., 257 Selmer, M., 171 Sengupta, D., 112 Sengupta, J., 162–163 Serysheva, I. I., 5–7, 10, 14 Sethian, J. A., 36 Settembre, E., 3, 5 Sexton, J., 269 Seybert, A., 39–40, 248, 260, 277 Shah, S., 274 Shaikh, T. R., 39, 162, 317 Shanks, J., 129 Sharon, M., 57 Sharp, S. J., 282 Shaw, D. E., 96 Shayakul, C., 99 Sherman, M. B., 184, 204 Shi, J., 13, 52, 269, 272–273, 306 Shimomura, O., 252 Shirouzu, M., 162 Shomura, Y., 22 Short, J. M., 316 Siebert, X., 6 Signorell, G. A., 103, 107–108 Sigworth, F. J., 38, 189, 318 Silberstein, C., 96 Simon, M. N., 268 Sindelar, C. V., 123, 134–135 Singer, A., 320–321 Singh, V., 316 Sinz, A., 56 Skehel, J. J., 268 Skop, A. R., 256 Sliz, P., 2, 94, 96, 102, 105–106, 108–109, 111 Slot, J. W., 251 Small, J. V., 206–209, 254 Smith, B. L., 94 Smith, J. C., 112 Smith, J. M., 100, 108, 125 Smith, S., 182 Snegupta, J., 171 Snyder, J. P., 123, 137

Author Index

Sobel, A., 133 Sochen, N. A., 36 Soding, J., 52 Sodroski, J., 268, 283, 285 Soille, P., 36 Song, J. L., 3, 5, 8, 14, 16 Song, Y. H., 129 Sorensen, T., 148 Sorzano, C. O. S., 311, 315–317, 320 Sosinsky, G. E., 105 Sougrat, R., 269, 286 Sousa, D., 85, 318 Southworth, D. R., 198 Spahn, C. M. T., 40, 161 Spehner, D., 210 Spehner, J. C., 93 Sperling, J., 208 Sperling, R., 208 Spillmann, L., 41 Sprechmann, P., 39–40, 277, 282 Srayko, M., 207 Srivastava, S., 162 Stagg, S. M., 302, 319 Stahlberg, H., 96–97, 99, 105, 107–109 Stahl, S. J., 3, 25 Stalling, D., 37 Stanfield, R. L., 268 Starich, M. R., 269 Stark, H., 171 Starosta, A., 168–170 Steigemann, P., 256 Steitz, T. A., 162 Sternberg, M. J., 13 Steven, A. C., 2–3, 25, 32, 41, 184, 269, 286 Stewart, A., 317 Stewart, M., 258 Stoffler, D., 259 Stoffler-Meilicke, M., 318 Stokes, D. L., 148, 153, 155, 157, 204 Stoschek, A., 35, 272 Stout, C. D., 269 Striebel, F., 66 Stroud, R. M., 95–97, 110 Stuckey, J., 283 Studer, D., 33, 248 Subramaniam, S., 36, 39–40, 136, 246, 269–270, 273–274, 277, 280, 282–286 Suck, D., 204 Suda, K., 103 Suhre, K., 6–7, 61 Sui, H., 95, 110, 123, 130, 138–139 Suloway, C., 184, 272–273, 293, 302, 305–308 Summers, M. F., 269 Sundquist, W. I., 269, 273, 280–281 Sun, R., 274 Suntharalingam, M., 258 Suzuki, H., 97, 102, 104–106, 108 Svergun, D. I., 166

351

Author Index

Svitkina, T. M., 254 Swami, N. K., 163 Sweet, G., 93 Sweet, R. W., 268, 285 Swets, J. A., 41 Swinnen, J. V., 180 T Tagari, M., 75 Tajkhorshid, E., 98 Taka, J., 3 Takemoto, C., 162 Tama, F., 6–7, 84, 172 Tamma, G., 99 Tang, G., 6, 10, 311, 322 Tang, M., 268 Tani, K., 97, 102–106, 108, 125 Tanimura, Y., 97 Tan, R. K., 6–7, 84 Taylor, D. J., 196 Taylor, D. W., 274 Taylor, K. A., 39–40, 151, 182, 269–270, 273–274, 276–277, 279, 296, 320 Terada, T., 162 Terry, L. J., 258 Thie´baud, P., 94 Thomas, D., 268–269 Thomton, G. B., 282 Thygesen, J., 84 Timasheff, S. N., 132 Tippit, D. H., 256 Tischendorf, G. W., 183 Tittmann, P., 96 Tivol, W. F., 269 Tocheva, E. I., 32 Tomasi, C., 35 Tomko, R. J. Jr., 64 Tomlin, S. G., 259 Topf, M., 5–7, 22, 61, 84, 165, 168 To¨rnroth-Horsefield, S., 98 Torrents, D., 108–109 Tournier, J., 279 Towbin, H., 56 Toyoshima, C., 148, 153 Trabuco, L. G., 6–7, 65, 69, 84, 162, 172 Trapani, S., 84 Trieber, C., 157 Trybus, K. M., 274 Tsukaguchi, H., 99 Tsuruta, H., 162 Tuffe´ry, P., 18 Tyka, M. D., 6–7, 19, 68 Typke, D., 33, 38, 94, 208, 272, 277, 305 U Uchida, S., 98 Ucurum, Z., 108

Ullmann, G. M., 112 Umesh Adiga, P. S., 316 Unger, V. M., 105 Unser, M., 318 Unwin, N. R., 2, 74, 93, 153, 297 Unwin, P. N., 99–100, 184 Urban, E., 207–209, 256 Urzhumstev, A. G., 39 V Vainshtein, B. K., 320 Vajda, Z., 99 Valencia, E., 108–109 Valenti, G., 99 Vale, R. D., 49, 134, 222 Valle, M., 41, 163–164, 171 Vallotton, P., 206 Vance, D. E., 180 Vance, J. E., 180 van den Boomgaard, R., 37 van der Goot, F. G., 103–104 van der Heide, P., 36, 38 van der Krift, T. P., 38 van Driel, L. F., 252 van Heel, M., 171, 186, 274, 311, 317–318, 320–322, 328 van Hoek, A. N., 94, 96, 99 van Oost, B. A., 98 van Os, C. H., 95, 98 Van Ryk, D., 268, 282–284 van Trees, H. L., 38 van Vlijmen, H., 13 Vassylyev, D. G., 18, 100 Velasquez-Muriel, J. A., 315 Verbavatz, J. M., 96 Verbeek, J. J., 164 Verdijk, M. A., 98 Verdoucq, L., 98 Verkleij, A. J., 38 Verkman, A. S., 94, 96 Vernoslava, E. A., 39 Viadiu, H., 99 Vicidomini, G., 247 Vigers, G., 100 Villa, E., 6–7, 34, 47, 80, 162, 172, 207–208 Vincent, L., 36 Vink, M., 296, 302 Vlassis, N., 164 Vocking, K., 38 Vogan, E. M., 268 Voges, D., 67 Vogt, V. M., 268 Volkmann, N., 6–7, 14, 31, 60–61, 194, 208–209, 316 Vonck, J., 100, 136 Vos, M., 33, 207 Voss, N. R., 184, 291

352

Author Index W

Wade, R. H., 129, 134, 162 Wagenknecht, T., 39 Wainberg, M. A., 279 Wakabayashi, T., 134 Walian, P. J., 95–96, 110 Walker, M. L., 190, 198, 297 Wallin, M., 122–123 Walz, J., 270, 272, 277 Walz, T., 2, 91, 183 Wang, D. N., 36, 100, 123 Wang, H. W., 80, 128, 132–133, 162 Wang, J., 84 Wang, Y. A., 74, 84, 98 Ward, A., 143–157 Ward, J. H., 187 Warren, G. B., 149 Watanabe, Y., 3 Waterman-Storer, C. M., 206, 208 Webb, B., 6–7 Webb, M. R., 204 Weber, I., 207–208, 269 Wedemeyer, W. J., 6–7 Weigele, P. R., 3, 8, 16 Weir, J. R., 162, 168–171 Weitz, D., 108 Weitzman, C., 95, 97, 110 Weixlbaumer, A., 171 Welch, P. D., 315 Welker, R., 268–269 Wells, B., 98 Wendler, P., 83 Wente, S. R., 258 Werten, P. J., 99, 105, 107 Westerhoff, M., 37 Westhof, E., 162, 168, 170–171 Weston, A. E., 247 Whitaker, R. T., 36 Whitby, F. G., 269 White, H. D., 296–297 Whiteheart, S. W., 222 Whitford, P. C., 168–170, 172 Whittaker, M., 129 Wiegers, K., 279 Wiener, M. C., 94 Wieringa, B., 98 Wiley, D. C., 268 Wilk, T., 268–269, 279 Williams, R. C., 100 Wilson, D. N., 162, 165, 168 Wilson, I. A., 268 Wingfield, P. T., 3, 25 Winkler, D. C., 269, 286 Winkler, H., 39–40, 219, 267–286, 320 Wintermeyer, W., 171 Wirtz, D., 206, 208

Wirtz, S., 97 Wittmann, T., 206, 208 Wolf, S. G., 92, 100, 123, 125–126, 128 Wolfson, H., 6–7 Wong-Barnum, M., 37, 41 Wong, H. C., 316 Woolford, D., 3, 5, 18, 35 Woolfson, M. M., 40 Worring, M., 37 Worthylake, D., 269 Wriggers, W., 6, 60–61, 67, 84, 172, 194 Wright, E. R., 33, 267 Wu, L. F., 93 Wu, Q., 35 Wu, X., 268 Wyatt, R., 268, 285 Wynne, S. A., 3, 25, 84 X Xiang, S. H., 268, 282–284 Xiang, Y., 80 Xiong, Y., 84 Xu, C., 3, 5, 153 Xue, F. L., 254 Xu, H., 96 Xu, L., 268, 282–284 Xu, X. P., 36–38, 40, 208–209 Y Yancey, S. B., 95 Yang, C., 40 Yang, Q., 259 Yang, R., 269 Yang, T., 207 Yang, X., 268, 282–284 Yang, Z. Y., 268 Yan, X., 318 Yasumura, T., 96 Yeager, M., 94, 98, 105, 269 Ye, F., 39–40, 274, 276–277 Yeh, R., 97 Yohda, M., 22 Yonath, A., 39 Yonekura, K., 180 Yoshida, T., 22 Yoshioka, C., 298, 305 Young, H. S., 143, 153, 157 Young, P. R., 35 Yuan, F., 13 Yu, I. M., 196 Yuste, R., 252 Yu, X., 3, 16, 76, 328 Yu, Y., 67 Yu, Z. Y., 6–7, 36

353

Author Index Z Zabransky, D. J., 269 Zampighi, G. A., 37 Zampighi, L. M., 37 Zanetti, G., 219, 269–270 Zauberman, N., 276 Zavialov, A., 163, 171 Zaytzev-Bashan, A., 39 Zeidel, M. L., 94 Zemlin, F., 2, 92, 99, 122–123, 305 Zeng, X., 108–109 Zhang, F., 54 Zhang, J., 3, 6–7, 10, 18, 21–22, 36–37, 180, 196, 306, 308 Zhang, M. Y., 268, 282–284 Zhang, P., 153, 248, 269, 306 Zhang, S. X., 123 Zhang, W., 4, 162, 168–171, 198 Zhang, W. H., 268 Zhang, X., 3, 5–6, 8, 224, 328 Zhang, Y. M., 180

Zhang, Z. Y., 108 Zhao, G., 101, 269 Zhaoping, L., 41 Zheng, Q. S., 272 Zheng, S. Q., 272–273, 305, 307 Zhou, H. X., 58 Zhou, J., 279 Zhou, T., 268, 282–284 Zhou, Z. H., 3, 5–7, 13, 16, 74, 92, 274, 315 Zhu, J., 6–7, 166, 320 Zhu, P., 39–40, 269–270, 273, 276–277 Zhu, Y., 316 Ziese, U., 36 Zilker, A., 124 Zimmerberg, J., 152 Zlotnick, A., 3, 25 Zou, J. Y., 6, 8 Zoumplis, D., 279 Zuber, B., 33, 207 Zulauf, M., 102 Zwick, M. B., 268, 282–284

Subject Index

A AAA-ATPase models, building and assessment, 63–64 proteasomal hexamer, 53–55 Actin cytoskeletal, tomography barbed end, 204 binding protein, 205 eukaryotic, 204 heterogeneity, 210–212 high-resolution light microscopy techniques (HRLM), 205 image analysis dual-axis tilt, 208 lamella hypothesis, test, 208–209 two-dimensional projection, 209 lamella hypothesis, testing, 205–206 pointed end, 204 rapid transfer apparatus (RTS), 210 sample preparation correlative light, EM, 207 cryo-EM, 206–207 cryo-ET, 207–208 transience, 209–210 transmission electron cryomicroscopy (cryo-TEM), 211 Animal fatty acid synthase analytical method active-site mutants, 195 catalytic domains, 181 classification, 192–193 class merging, 188–189 conical tilt, 193–195 2D class average, 189–190 focused alignment, 190–191 image alignment, preliminary, 185 image preprocessing, particle, 185 microscopy and data collection, 184 particle images, realignment of, 188 particle selection, 184 preliminary image classification, 185–188 specimen preparation, 182–183 catalytic cycle, 180–182 electron microscopy (EM), 180 FAS homodimer, 181 flexible macromolecule catalytic cycle, 197 single-particle EM, 196–198

Aquaporins (AQPs). See also Electron crystallography and aquaporins adhesive properties, AQP0, 95 AQP4M1 and AQP4M23, 96 AQP1 structure, 95 bR structure, 100 data collection, benefits, 99 dioleyl phosphatidylcholine (DOPC), 94 electron crystallographic studies, 94 AQP2, 98–99 AQP9, 99 AQPZ, 97 GlpF, 97 SoPIP2, 98 a-TIP, 98 image processing library and toolbox (IPLT), 106 mean square deviations (RMSD), 111 methyl-b-cyclodextrin (MBCD), 103 MIP family, 93 signal-to-noise ratio (SNR), 106 Atomic and protein–protein interaction data. See Cryoelectron microscopy (cryo-EM) integration Atomic model protocol annotation, secondary structure, 11–12 atomic map template generation, 225–227 TOM toolbox, 225 Ca, 18 atomic model optimization, Rosetta, 19–20 optimization, 17 placement, 16 positions, helices, 16 positions, sheets and loops, 16–17 fitting atomic models, 14 fixing of, 17 flowchart, 9 macromolecular model, 17–18 map rescaling, 18 model optimization, 18–19, 21 model quality monitoring, 19 secondary structure elements identification, 11 segmentation, 10–11 SSEs correspondence, 14–16 prediction, sequence, 14 structural homologues identification, 13

355

356

Subject Index B

Back injection technique, 124 Bacteriorhodopsin, atomic models, 93 C 2þ

Ca -ATPase ATPase activity, 153 density gradient centrifugation, 149 detergent effect, 152 freeze–thaw cycles, 151 lipid composition, 152 lipid types, 149 proteoliposomes, 149 reactive red affinity chromatography, 148 sarcoplasmic reticulum (SR), 147 thapsigargin/phospholamban, 150 Contrast transfer function (CTF), 314–315 Cryoelectron microscopy (cryo-EM) integration atomic models refinement, high-resolution maps molecular dynamics flexible fitting, 65–66 20S core particle, 66–68 cryoelectron tomography (CET), 48 model building, complex AAA-ATPase models, building and assessment, 63–64 assembly representation, 59 scoring of, 60–63 molecular interpretation, 48 placing assembly subunits, problem molecular dynamics (MD) simulations, 50 regulatory particles (RPs), 49 26S proteasome, 49 protein–protein interaction data, 54 chemical cross-linking, 56 coexpression, 57 computational interaction and interface prediction, 57–58 experimental methods, 57 interproteasomal interactions, 58–59 in vitro binding assays, 56 in vivo pulldown, 56 two-hybrid assay, 56 single-particle analysis (SPA), 48 structure prediction, subunits comparative modeling, 50 de novo model, 51 model building, 53 proteasomal AAA-ATPase hexamer, 53–55 sequence based methods, 51–52 threading, 52 Cryoelectron tomography, eukaryotes cryosectioning, 247 cytoskeleton-driven processes Dictyostelium discoideum, 255 filamentous actin (F-actin), 254

vitrification, EM, 255 fluorescent-light microscopy, 247 imaging technique, 246 midbodies cryo-ET, 257 localization, 256 morphology, 256–257 polar MT, 257 structre, 256 nuclear pore complex active, 259 3D alignment algorithm, 259 Dictyostelium discoideum, 259 nuclear envelope, 257 nuclear membrane, 257–258 nucleoporin, 258 structural analysis, 258 specimen preparation carbon-coated EM grids, 251 cryoplaning, 248 cryoultramicrotomy, 248 Dictyostelium discoideum, 249 fiducial marker, 251 flagellum, 249–250 grid, 250 nondividing cell, 249 visualization, 248 structural study fluorescencemicroscopy, 252 green fluorescent protein (GFP), 252 grids, 253 macromolecules visualization, 252 optical resolution, 254 quantum dots (QD), 253 whole cell, 247 Cryoelectron tomography, HIV virions alignment strategies by classification, 276–277 3D motifs, 276 env spikes, 277 data acquisition angular range, 273 virus, images, 272 3D tomographic reconstruction alignment methods, 273 protomo software package, 274 optimal electron dose, 271–272 sample preparation EM grid preparation, 271 frozen-hydrated specimen, 270 plunge freezing, 271 subvolume analysis HIV env, 275 missing wedge, 275 rotational alignment, 276 Cryoelectron tomography, prokaryotes and viruses, 247

357

Subject Index D Dictyostelium discoideum, 208 Dimyristoyl phosphatidylcholine (DMPC), 94, 101 E Electron crystallography and aquaporins bacteriorhodopsin, atomic models, 93 structural biology aquaporin-0, 95–96 aquaporin-1, 94–95 aquaporin-4, 96–97 studies of, 97–99 of structural studies data processing, 108–109 2D crystallization, 101–103 electron crystallography vs. X-ray crystallography, 109–112 single-layered vs. double-layered 2D crystals, 105–107 specimen preparation, 103–105 Electron tomographic reconstructions. See Segmentation and interpretation methods EM Data Bank (EMDB), 75 F F-actin, 204 Fatty acid synthase (FAS). See also Animal fatty acid synthase analytical method active-site mutants, 195 catalytic domains, 181 classification, 192–193 class merging, 188–189 conical tilt, 193–195 crystal structure, 193 2D class average, 189–190 focused alignment, 190–191 image alignment, preliminary, 185 image preprocessing, particle, 185 microscopy and data collection, 184 particle images, realignment of, 188 particle selection, 184 preliminary image classification, 185–188 specimen preparation, 182–183 conformational changes, 182 crystal structure, 193 vs. EM catalytic cycle, 197 single-particle EM, 196–198 fleximers, 190 hierarchical clustering, 187–188 homodimer, 181 Ferredoxin oxidoreductase, visual proteomics, 224

G G-actin, 204 Grid affinity, 298 C-flat, 298 FEI vitrobot, 296 H Helical crystallization, membrane proteins Ca2þ-ATPase ATPase activity, 153 density gradient centrifugation, 149 detergent effect, 152 freeze–thaw cycles, 151 lipid composition, 152 lipid types, 149 proteoliposomes, 149 reactive red affinity chromatography, 148 sarcoplasmic reticulum (SR), 147 thapsigargin/phospholamban, 150 MsbA adsorption, 145 ATP-binding cassette (ABC), 144 density maps, 148 electron cryomicroscopy, 147 helical crystal, 146 n-dodecyl b-D-maltoside (bDDM), 145 n-undecyl b-D-maltoside (UDM), 145 Vibrio cholerae, 145 and two-dimensional crystals, lipid dependent lipid mixture effect, 154 magnesium, 155–156 HIV virions, 3D visualization. See also Cryoelectron tomography, HIV virions b12 env complex CD4 attachment, 282 classification, 283f 3D map, 284f CD4 receptor, 285 conical cores, 269 cryoelectron tomography alignment strategies, 276–278 data acquisition, 272–273 3D tomographic reconstruction, 273–274 optimal electron dose, 271–272 sample preparation, 270–271 subvolume analysis, 274–276 2D cryo-EM, 269 electron microsopy, 268 lipid-enveloped, 268 molecular architecture env protein, 279–282 gag shell, 278–279 structural proteins, 268

358

Subject Index L

Leptospira interrogans, 237 M Methanococcus maripaludis chaperonin (Mm-cpn), 10, 21–24 Microtubules (MTs) assembly/disassembly intermediates GTP and GDP interfaces, 132 protofilaments, 132 structure, 131 tubulin bound, 133 kinesin movement mechanism crystal structure, 133 higher resolution cryo-EM, 136 kinesin–microtubule interaction., 134–135 neck-linker docking, 134 nucleotide state, 134 structure depolymerization, 129 monomers arrangement, 128 three-dimensional reconstruction, 130 Molmatch handling beam direction, 228 constrained correlation function (CCF), 228 N-dimensional data, 227 MsbA adsorption, 145 ATP-binding cassette (ABC), 144 density maps, 148 electron cryomicroscopy, 147 helical crystal, 146 n-dodecyl b-D-maltoside (bDDM), 145 n-undecyl b-D-maltoside (UDM), 145 Vibrio cholerae, 145 Multiparticle cryo-EM, ribosomes field emission electron guns (FEG), 162 heterogeneity, 3D sorting classification methods, 163 protein synthesis, 163 refinement strategy, 164–165 interpretation atomic models, 171 CrPV IRES, 171–172 kirromycin complex, 173 reconstruction, 171 subnanometer multireference refinement, 166–171 sample preparation, 165–166 N National Resource for Automated Molecular Microscopy (NRAMM) autoloading and robotic screening, 300–304 automated data acquisition, 308–310 grids, 298

image processing, 311–314 quality assessment, 323–325 specimen preparation protocol, 299 P Pearson correlation coefficient, 39 Protein–protein interaction data, 54 chemical cross-linking, 56 coexpression, 57 computational interaction and interface prediction, 57–58 experimental methods, 57 interproteasomal interactions, 58–59 in vitro binding assays, 56 in vivo pulldown, 56 two-hybrid assay, 56 Pyrodictium, 208 R Rapid transfer apparatus (RTS), 210 S Segmentation and interpretation methods classification and averaging normalization, 40 Pearson correlation coefficient, 39 clinical data sets, 34 cryotomographic reconstructions, 33 electron tomogram interpretation, 36 energy-based algorithms, 37 macromolecular assemblies, detection and mapping feasibility tests, 39 matched filter, 38–39 visual proteomics, 38 noise reduction, 35–36 validation, 40–41 watersnakes, 37 Single-particle electron microscopy, automation appion, 322 assessment, 322–327 black box, 292 contrast transfer function (CTF), 314–315 cryo-EM automation, 297 combined, 329 time-resolved, 296 definition, 293 grid loading, 300–301 image processing ab initio reconstructions, 319–321 appion, 311 CTF estimation and correction, 314–315 2D alignment, 320 2D classification, 317 3D refinement, 321–322

359

Subject Index

NRAMM image processing, 311–314 particle selection and stack creation, 316–317 integrated system, 326 leginon, 308 microscopy image aberrations, 305 image acquisition, 306 pyScope python extension, 308 specimen radiation, 305 target selection, 308 modulation transfer function, 329 NRAMM autoloading and robotic screening, 300–304 automated data acquisition, 308–310 quality assessment, 323–325 specimen preparation protocol, 299 robotic screening, 300–304 specimen preparation adsorption substrate, 295 Cryomesh, 298 Gatan Cp3, 296 grid, 295 heavy metal stains, 295 negative staining, 295 NRAMM, protocol, 299–300 optimization, 298 reproducibility, 295 robotic mechanism, 296 sample, vitrified, 300 vitrification, 296–297 x-ray crystallography, 298 transferrin receptor complex, 328 Single-particle EM. See Animal fatty acid synthase Spastin, 256 Spatial proteome in silico, 239 Leptospira interrogans, 237 template matching, 238 Subnanometer multiparticle cryo-EM multireference refinement alignment and classification phase, 168 high-frequency enhancement, 170 loops, 167 pixel size, 170 procedure, 166 size, 166 strong low-pass filtering, 169 sample preparation, 165–166 Subnanometer resolution cryo-EM density maps features, 3–5 limitations, 24 Mm-cpn, 21–24 protocol, atomic model Ca, 18

Ca atomic model optimization, Rosetta, 19–20 Ca optimization, 17 Ca placement, 16 Ca positions, helices, 16 Ca positions, sheets and loops, 16–17 fitting atomic models, 14 fixing of, 17 flowchart, 9 macromolecular model, 17–18 map rescaling, 18 model optimization, 18–19, 21 model quality monitoring, 19 secondary structure annotation, 11–12 secondary structure elements identification, 11 segmentation, 10–11 SSE correspondence, 14–16 SSEs prediction, sequence, 14 SSEs, structural homologues identification, 13 structural homologues, 13 sequence-to-structure correspondence, 25 tools, analysis constrained modeling, 7 de novo modeling, 8 fitting atomic models, 4, 7 for programs, 6 protein subunits extraction, 7 secondary structure identification, 8 T Tubulin structure and interactions drug binding study, 136–137 electron crystallography alpha and beta-tubulin difference, 128 atomic model, 126 back injection technique, 124 crystal size, 123 3D density map, 126 diffraction pattern, 124 double carbon film, 125 factors, 125 P-loop, 127 ribbon diagram, 127 tannin embedment, 123 microtubules (MTs) assembly/disassembly intermediates, 131–133 kinesin movement mechanism, 133–136 structure, 128–130 tomography axoneme, 137 Chlamydomonas, 138 cryosections, 139

360

Subject Index

Tubulin structure and interactions (cont.) sea urchin, 139 signal-to-noise ratio, 137 U Unified data resource, cryo-EM access 3D visualization tools, 82 EMDB atlas page, 81 EMSEARCH, 80 accession ID, 87 deposition and content, 77–80 EM Data Bank (EMDB), 75 experiment info, 86 external database IDs, 86 fitted model deposition, 87 map deposition, EMDEP, 86 map file, 85–86 model file, 86 structural data archives dictionary development, EM, 76–77 maps, 75 models, 76 unified resource, 77 uses, examples comparative studies, 84 crystal structure phasing, 83–84 data interpretation, 83 molecular pictures, 83 software development, 84–85

V Visual proteomics cryoelectron tomography (CET), 216 data acquisition experimental setup, 217–218 parameters, 218–220 tomogram reconstruction, 220–221 TOM toolbox, 225 human pathogen, 217 lumazine synthase, 224 mass spectrometer (MS), 216 performance assessment real data sets, 235–237 true-positive discovery, 232–235 prokaryotes, 217 segmentation and interpretation methods, 38 spatial proteome in silico, 239 Leptospira interrogans, 237 template matching, 238 template matching molecular atlases, visualization, 231–232 Molmatch, handling, 227–229 motif list, creation, 229–230 scoring, 230–231 templates Desulfovibrio vulgaris, 222–225 quaternary structure conservation, 222–225 structure selection, 221–222 template generation, atomic map, 225–227 Vitrification, 296–297

Methods in enzymology 483 - Cryo-EM, Part C: Analyses, Interpretation, and Case Studies - PDF Free Download (2024)

References

Top Articles
Latest Posts
Recommended Articles
Article information

Author: Mr. See Jast

Last Updated:

Views: 5513

Rating: 4.4 / 5 (55 voted)

Reviews: 94% of readers found this page helpful

Author information

Name: Mr. See Jast

Birthday: 1999-07-30

Address: 8409 Megan Mountain, New Mathew, MT 44997-8193

Phone: +5023589614038

Job: Chief Executive

Hobby: Leather crafting, Flag Football, Candle making, Flying, Poi, Gunsmithing, Swimming

Introduction: My name is Mr. See Jast, I am a open, jolly, gorgeous, courageous, inexpensive, friendly, homely person who loves writing and wants to share my knowledge and understanding with you.