Lipid-Protein Interactions: Methods and Protocols [2nd ed.] 978-1-4939-9511-0;978-1-4939-9512-7

Table of contents :
Front Matter ....Pages i-xiv
Multiscale Modeling and Simulation Approaches to Lipid–Protein Interactions (Roland G. Huber, Timothy S. Carpenter, Namita Dube, Daniel A. Holdbrook, Helgi I. Ingólfsson, William A. Irvine et al.)....Pages 1-30
Quartz Crystal Microbalances as Tools for Probing Protein–Membrane Interactions (Søren B. Nielsen, Daniel E. Otzen)....Pages 31-52
Surface Plasmon Resonance for Measuring Interactions of Proteins with Lipids and Lipid Membranes (Aleksandra Šakanovič, Vesna Hodnik, Gregor Anderluh)....Pages 53-70
Thermodynamic Analysis of Protein–Lipid Interactions by Isothermal Titration Calorimetry (Musti J. Swamy, Rajeshwer S. Sankhala, Bhanu Pratap Singh)....Pages 71-89
Differential Scanning Calorimetry of Protein–Lipid Interactions (Olga Cañadas, Cristina Casals)....Pages 91-106
Imaging and Force Spectroscopy of Single Transmembrane Proteins with the Atomic Force Microscope (K. Tanuj Sapra)....Pages 107-144
Kinetics of Insertion and Folding of Outer Membrane Proteins by Gel Electrophoresis (Andre Schüßler, Sascha Herwig, Jörg H. Kleinschmidt)....Pages 145-162
Optimized Negative-Staining Protocol for Lipid–Protein Interactions Investigated by Electron Microscopy (Jianfang Liu, Hao Wu, Changyu Huang, Dongsheng Lei, Meng Zhang, Wei Xie et al.)....Pages 163-173
Probing Heterogeneous Lipid Interactions with Membrane Proteins Using Mass Spectrometry (John W. Patrick, Arthur Laganowsky)....Pages 175-190
Protein Microarrays and Liposome: A Method for Studying Lipid–Protein Interactions (Samuel Herianto, Chien-Sheng Chen, Heng Zhu)....Pages 191-199
Structural Investigations of Protein–Lipid Complexes Using Neutron Scattering (Luke A. Clifton, Stephen C. L. Hall, Najet Mahmoudi, Timothy J. Knowles, Frank Heinrich, Jeremy H. Lakey)....Pages 201-251
Circular-Dichroism and Synchrotron-Radiation Circular-Dichroism Spectroscopy as Tools to Monitor Protein Structure in a Lipid Environment (Koichi Matsuo, Kunihiko Gekko)....Pages 253-279
FTIR Analysis of Proteins and Protein–Membrane Interactions (Suren A. Tatulian)....Pages 281-325
UV Resonance Raman Spectroscopy as a Tool to Probe Membrane Protein Structure and Dynamics (DeeAnn K. Asamoto, Judy E. Kim)....Pages 327-349
Analyzing Transmembrane Protein and Hydrophobic Helix Topography by Dual Fluorescence Quenching (Gregory A. Caputo, Erwin London)....Pages 351-368
Förster Resonance Energy Transfer as a Tool for Quantification of Protein–Lipid Selectivity (Luís M. S. Loura, Manuel Prieto, Fábio Fernandes)....Pages 369-382
A Guide to Tracking Single Membrane Proteins and Their Interactions in Supported Lipid Bilayers (Evan L. Taylor, Kumud Raj Poudel, James A. Brozik)....Pages 383-414
Fluorescence Correlation Spectroscopy to Examine Protein–Lipid Interactions in Membranes (Viktoria Betaneli, Jonas Mücksch, Petra Schwille)....Pages 415-447
Membrane Pore Formation by Peptides Studied by Fluorescence Techniques (Suren A. Tatulian, Nabin Kandel)....Pages 449-464
Folding of β-Barrel Membrane Proteins into Lipid Membranes by Site-Directed Fluorescence Spectroscopy (Lisa Gerlach, Omkolsum Gholami, Nicole Schürmann, Jörg H. Kleinschmidt)....Pages 465-492
EPR Techniques to Probe Insertion and Conformation of Spin-Labeled Proteins in Lipid Bilayers (Enrica Bordignon, Svetlana Kucher, Yevhen Polyhach)....Pages 493-528
Studying Lipid–Protein Interactions with Electron Paramagnetic Resonance Spectroscopy of Spin-Labeled Lipids (Tibor Páli, Zoltán Kóta)....Pages 529-561
Solid-State NMR Approaches to Study Protein Structure and Protein–Lipid Interactions (Christopher Aisenbrey, Evgeniy S. Salnikov, Jesus Raya, Matthias Michalek, Burkhard Bechinger)....Pages 563-598
Solution NMR Spectroscopy for the Determination of Structures of Membrane Proteins in a Lipid Environment (Ashish Arora)....Pages 599-643
Nanodiscs as a New Tool to Examine Lipid–Protein Interactions (Ilia G. Denisov, Mary A. Schuler, Stephen G. Sligar)....Pages 645-671
Back Matter ....Pages 673-682

Citation preview

Methods in Molecular Biology 2003

Jörg H. Kleinschmidt Editor

Lipid-Protein Interactions Methods and Protocols Second Edition

METHODS

IN

MOLECULAR BIOLOGY

Series Editor John M. Walker School of Life and Medical Sciences University of Hertfordshire Hatfield, Hertfordshire, UK

For further volumes: http://www.springer.com/series/7651

For over 35 years, biological scientists have come to rely on the research protocols and methodologies in the critically acclaimed Methods in Molecular Biology series. The series was the first to introduce the step-by-step protocols approach that has become the standard in all biomedical protocol publishing. Each protocol is provided in readily-reproducible step-bystep fashion, opening with an introductory overview, a list of the materials and reagents needed to complete the experiment, and followed by a detailed procedure that is supported with a helpful notes section offering tips and tricks of the trade as well as troubleshooting advice. These hallmark features were introduced by series editor Dr. John Walker and constitute the key ingredient in each and every volume of the Methods in Molecular Biology series. Tested and trusted, comprehensive and reliable, all protocols from the series are indexed in Pub Med.

Lipid-Protein Interactions Methods and Protocols Second Edition

Edited by

Jörg H. Kleinschmidt Department of Biophysics, Institute of Biology, FB10 and Center for Interdisciplinary Nanostructure Science and Technology (CINSaT), University of Kassel, Kassel, Germany

Editor Jo¨rg H. Kleinschmidt Department of Biophysics, Institute of Biology FB10 and Center for Interdisciplinary Nanostructure Science and Technology (CINSaT), University of Kassel Kassel, Germany

ISSN 1064-3745 ISSN 1940-6029 (electronic) Methods in Molecular Biology ISBN 978-1-4939-9511-0 ISBN 978-1-4939-9512-7 (eBook) https://doi.org/10.1007/978-1-4939-9512-7 © Springer Science+Business Media, LLC, part of Springer Nature 2019 This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors, and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, express or implied, with respect to the material contained herein or for any errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. This Humana imprint is published by the registered company Springer Science+Business Media, LLC, part of Springer Nature The registered company address is: 233 Spring Street, New York, NY 10013, U.S.A.

Preface Proteins and lipids are the main components of biological membranes that are essential structuring elements of all living cells. Natural phospholipids spontaneously form lipid bilayers. In biomembranes, the phospholipid bilayer constitutes a hydrophobic barrier that prevents the arbitrary exchange of solutes, while membrane-spanning proteins allow the regulated exchange of solutes or the transduction of signals from one side of the membrane to the other. Many enzymatic reactions take place at the membrane-water interface. Specific lipid-protein interactions are important for the stable integration and activity of integral and peripheral membrane proteins. The unique structure of the lipid bilayer requires specific surface properties of integral and peripheral proteins for their function. Protein surfaces exposed to the fatty core of the membrane are typically hydrophobic, while protein surfaces exposed to the aqueous space usually are composed of polar amino acid residues. The polar-apolar interface of the lipid bilayer is formed by the glycerol backbone and by the polar head group of the various phospholipid species and therefore an important region for lipid-protein interactions. The fatty acyl chains of the lipids can vary a lot in length and degree of unsaturation and the membranes may contain cholesterol, sphingolipids, etc. This has consequences, e.g., for membrane thickness and fluidity. Furthermore, the complex membrane composition often leads to the formation of microdomains with distinct physicochemical properties. To gain detailed insight into membrane properties, it is therefore of great importance to understand the complex nature of the interactions of membrane proteins with lipids. This volume provides a selection of protocols to examine protein-lipid interactions, membrane and membrane protein structure, how membrane proteins affect lipids, and how they are in turn affected by the lipid bilayer and lipid properties. Numerous methodologies have been developed in the past, each with its own advantages and limitations. The methods described here are all actively used, complementary, and necessary to obtain comprehensive information about membrane structure and function. The method of choice is determined by the information that is sought, but is dependent on the properties of the sample, the available quantity, and the required sensitivity. Label-free approaches described in this book include methodologies like quartz-crystal microbalances with dissipation, surface plasmon resonance, isothermal titration calorimetry, and differential scanning calorimetry. These are useful methods, e.g., to monitor binding events and to obtain the free energies, enthalpies, and entropies of protein-lipid interactions. Imaging techniques like electron microscopy and atomic force microscopy are used to examine the structure and organization of protein-lipid complexes in membranes. Atomic force spectroscopy allows the probing of mechanical properties of macromolecules, e.g., the force necessary to unfold a single protein in a lipid environment. Neutron scattering is an emerging technique to study the structure of proteinlipid complexes, which in combination with deuteration of either lipids or proteins, allows resolving the inner structure of big and dynamic lipid-protein complexes. The secondary structure of native and nonnative proteins in lipid membranes can conveniently be monitored by circular dichroism spectroscopy and synchrotron radiation circular dichroism spectroscopy. The development of the latter greatly extended the recordable wavelength range, strengthening structural investigations. Secondary structure and in addition the orientation and order parameters of membrane proteins in lipid bilayers can be obtained

v

vi

Preface

from infrared-spectroscopic methods. In combination with isotope editing methods, these are also used to determine changes in local protein conformation. The specificity and selectivity of protein interactions with lipid species is efficiently investigated by combining the labeling of lipids and proteins with either fluorescence or electron paramagnetic resonance spectroscopy, which are both very sensitive techniques to examine the protein-lipid interface. These methods are applied with great success to probe the topology of peptides and proteins in membranes. Fluorescence quenching is a fast and reliable technique to determine the location of fluorescent amino acids especially tryptophan residues in lipid bilayers. Fo¨rster energy transfer is a highly sensitive fluorescence method that is useful to measure distances up to 10 nm and able to detect direct binding as well as deviations from homogeneity of the lipid distribution around a protein. Fluorescence methodologies have also been successfully applied, e.g., to track single molecules like transmembrane proteins in planar supported membranes. Single particle tracking allows imaging and tracking single fluorescent molecules with good spatial and temporal resolution. Particle association and dissociation events can be monitored. Fluorescence correlation spectroscopy uses the time correlation of temporal fluctuations of fluorescence, which are detected in a focal volume, to explore dynamic events with high temporal resolution and statistical accuracy. Lipid spin labeling and electron paramagnetic resonance (EPR also called electron spin resonance, ESR) spectroscopy have been highly successful in the determination of the protein-solvating lipid shell, that is, the stoichiometry of the lipid interactions with integral membrane proteins. Since the technique allows the estimation of mobile and protein-immobilized lipids and their exchange rates, it has also been very successful in probing the lipid selectivity of transmembrane proteins. In combination with site-directed mutagenesis, EPR of spinlabeled proteins has become a powerful tool to examine protein structure and dynamics, even in complex systems that are not accessible with other approaches. Nuclear magnetic resonance (NMR) spectroscopy, combined with isotopic labeling, is widely used to probe the structure and dynamics of proteins either in solution or in a lipid environment. Solutionphase NMR is performed with detergent solubilized membrane proteins and usually yields high-resolution structures. Solid-state NMR is performed either as magic angle spinning (MAS) NMR to obtain highly resolved protein structures with spectra resembling fast isotropically tumbling proteins in solution or with oriented samples. When applied to oriented samples, NMR gives valuable information on the dynamics and orientation of lipids and proteins in membranes and allows determining, e.g., peptide orientation, lipidorder parameters, and lipid phases. Lipid-protein interactions have been investigated in small, large, and giant vesicles, in supported lipid bilayers, in lipid monolayers, with shortchain lipid micelles, or even with single lipids. More recently lipid nanodiscs have been developed, which are disc-like fragments of lipid bilayers that are stabilized by amphipathic helical proteins. Nanodiscs have been shown to be a robust means for stabilizing and investigating protein-lipid interactions, and nanodisc applications are reviewed here. A range of molecular dynamics simulation approaches are applied to examine lipid-protein interactions in detail. For the present second edition, all chapters have been updated to include recent developments. Six new chapters were added. The new chapters cover examinations of lipid-protein interactions by mass spectrometry, the use of protein microarrays to identify and examine lipid interactions of large sets of different proteins, investigations on membrane protein structure and dynamics by UV resonance Raman spectroscopy, examinations of peptide-induced pore formation in membranes, and investigations on lipid-protein interactions in folding and insertion of β-barrel membrane proteins. Lipid-protein interactions are

Preface

vii

heavily investigated by many teams worldwide and while it is clear that not all methods can be described in a single book, the present volume covers a wide range of the methods used in this area of research. I thank all contributing authors for providing interesting and highly valuable methods, insights, and reviews for this volume of Methods in Molecular Biology. Kassel, Germany

Jo¨rg H. Kleinschmidt

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

1 Multiscale Modeling and Simulation Approaches to Lipid–Protein Interactions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Roland G. Huber, Timothy S. Carpenter, Namita Dube, Daniel A. Holdbrook, Helgi I. Ing
o lfsson, William A. Irvine, Jan K. Marzinek, Firdaus Samsudin, Jane R. Allison, Syma Khalid, and Peter J. Bond 2 Quartz Crystal Microbalances as Tools for Probing Protein–Membrane Interactions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Søren B. Nielsen and Daniel E. Otzen 3 Surface Plasmon Resonance for Measuring Interactions of Proteins with Lipids and Lipid Membranes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Aleksandra Sˇakanovicˇ, Vesna Hodnik, and Gregor Anderluh 4 Thermodynamic Analysis of Protein–Lipid Interactions by Isothermal Titration Calorimetry. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Musti J. Swamy, Rajeshwer S. Sankhala, and Bhanu Pratap Singh 5 Differential Scanning Calorimetry of Protein–Lipid Interactions . . . . . . . . . . . . . . ˜ adas and Cristina Casals Olga Can 6 Imaging and Force Spectroscopy of Single Transmembrane Proteins with the Atomic Force Microscope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . K. Tanuj Sapra 7 Kinetics of Insertion and Folding of Outer Membrane Proteins by Gel Electrophoresis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ¨ ßler, Sascha Herwig, and Jo¨rg H. Kleinschmidt Andre Schu 8 Optimized Negative-Staining Protocol for Lipid–Protein Interactions Investigated by Electron Microscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Jianfang Liu, Hao Wu, Changyu Huang, Dongsheng Lei, Meng Zhang, Wei Xie, Jinping Li, and Gang Ren 9 Probing Heterogeneous Lipid Interactions with Membrane Proteins Using Mass Spectrometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . John W. Patrick and Arthur Laganowsky 10 Protein Microarrays and Liposome: A Method for Studying Lipid–Protein Interactions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Samuel Herianto, Chien-Sheng Chen, and Heng Zhu 11 Structural Investigations of Protein–Lipid Complexes Using Neutron Scattering. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Luke A. Clifton, Stephen C. L. Hall, Najet Mahmoudi, Timothy J. Knowles, Frank Heinrich, and Jeremy H. Lakey

ix

v xi

1

31

53

71 91

107

145

163

175

191

201

x

12

13 14

15

16

17

18

19

20

21

22

23

24

25

Contents

Circular-Dichroism and Synchrotron-Radiation Circular-Dichroism Spectroscopy as Tools to Monitor Protein Structure in a Lipid Environment . . . Koichi Matsuo and Kunihiko Gekko FTIR Analysis of Proteins and Protein–Membrane Interactions. . . . . . . . . . . . . . . Suren A. Tatulian UV Resonance Raman Spectroscopy as a Tool to Probe Membrane Protein Structure and Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . DeeAnn K. Asamoto and Judy E. Kim Analyzing Transmembrane Protein and Hydrophobic Helix Topography by Dual Fluorescence Quenching . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Gregory A. Caputo and Erwin London Fo¨rster Resonance Energy Transfer as a Tool for Quantification of Protein–Lipid Selectivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Luı´s M. S. Loura, Manuel Prieto, and Fa´bio Fernandes A Guide to Tracking Single Membrane Proteins and Their Interactions in Supported Lipid Bilayers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Evan L. Taylor, Kumud Raj Poudel, and James A. Brozik Fluorescence Correlation Spectroscopy to Examine Protein–Lipid Interactions in Membranes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ¨ cksch, and Petra Schwille Viktoria Betaneli, Jonas Mu Membrane Pore Formation by Peptides Studied by Fluorescence Techniques. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Suren A. Tatulian and Nabin Kandel Folding of β-Barrel Membrane Proteins into Lipid Membranes by Site-Directed Fluorescence Spectroscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ¨ rmann, Lisa Gerlach, Omkolsum Gholami, Nicole Schu ¨ and Jorg H. Kleinschmidt EPR Techniques to Probe Insertion and Conformation of Spin-Labeled Proteins in Lipid Bilayers. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Enrica Bordignon, Svetlana Kucher, and Yevhen Polyhach Studying Lipid–Protein Interactions with Electron Paramagnetic Resonance Spectroscopy of Spin-Labeled Lipids . . . . . . . . . . . . . . . . . . . . . . . . . . . . Tibor Pa´li and Zolta´n K
ota Solid-State NMR Approaches to Study Protein Structure and Protein–Lipid Interactions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Christopher Aisenbrey, Evgeniy S. Salnikov, Jesus Raya, Matthias Michalek, and Burkhard Bechinger Solution NMR Spectroscopy for the Determination of Structures of Membrane Proteins in a Lipid Environment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Ashish Arora Nanodiscs as a New Tool to Examine Lipid–Protein Interactions . . . . . . . . . . . . . Ilia G. Denisov, Mary A. Schuler, and Stephen G. Sligar

Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

253 281

327

351

369

383

415

449

465

493

529

563

599 645 673

Contributors CHRISTOPHER AISENBREY  Institut de Chimie, CNRS, Universite´ de Strasbourg, UMR 7177, Strasbourg, France JANE R. ALLISON  School of Biological Sciences and Maurice Wilkins Centre for Molecular Biodiscovery, The University of Auckland, Auckland, New Zealand; Biomolecular Interaction Centre, University of Canterbury, Christchurch, New Zealand GREGOR ANDERLUH  Department of Molecular Biology and Nanobiotechnology, National Institute of Chemistry, Ljubljana, Slovenia ASHISH ARORA  Molecular and Structural Biology Division, CSIR-Central Drug Research Institute, Lucknow, India DEEANN K. ASAMOTO  Department of Chemistry and Biochemistry, University of California, San Diego, La Jolla, CA, USA BURKHARD BECHINGER  Institut de Chimie, CNRS, Universite´ de Strasbourg, UMR 7177, Strasbourg, France VIKTORIA BETANELI  Medical Faculty “Carl Gustav Carus”, Institute of Physiological Chemistry, Technische Universit€ at Dresden, Dresden, Germany PETER J. BOND  Bioinformatics Institute (BII), Agency for Science, Technology and Research (A∗STAR), Singapore, Singapore; Department of Biological Sciences, National University of Singapore, Singapore, Singapore ENRICA BORDIGNON  Faculty of Chemistry and Biochemistry, Ruhr University Bochum, Bochum, Germany JAMES A. BROZIK  Department of Chemistry, Washington State University, Pullman, WA, USA OLGA CAN˜ADAS  Department of Biochemistry and Molecular Biology, Faculty of Chemistry, Complutense University of Madrid, Madrid, Spain; Centro de Investigaci
on Biome´dica en Red de Enfermedades Respiratorias (CIBERES), Instituto de Salud Carlos III, Madrid, Spain GREGORY A. CAPUTO  Department of Chemistry and Biochemistry, Rowan University, Glassboro, NJ, USA TIMOTHY S. CARPENTER  Biosciences and Biotechnology Division, Physical and Life Sciences Directorate, Lawrence Livermore National Laboratory, Livermore, CA, USA CRISTINA CASALS  Department of Biochemistry and Molecular Biology, Faculty of Chemistry, Complutense University of Madrid, Madrid, Spain; Centro de Investigaci
on Biome´dica en Red de Enfermedades Respiratorias (CIBERES), Instituto de Salud Carlos III, Madrid, Spain CHIEN-SHENG CHEN  High Throughput Biosensing Laboratory, Department of Food Safety/Hygiene and Risk Management, College of Medicine, National Cheng Kung University, Tainan City, Taiwan LUKE A. CLIFTON  Rutherford Appleton Laboratory, Science and Technology Facilities Council, Didcot, Oxfordshire, UK ILIA G. DENISOV  Department of Biochemistry, University of Illinois, Urbana, IL, USA NAMITA DUBE  Department of Chemistry, University of Cambridge, Cambridge, UK

xi

xii

Contributors

FA´BIO FERNANDES  Centro de Quı´mica Fı´sica Molecular, Institute of Nanosciences and Nanotechnologies, Instituto Superior Te´cnico, Lisbon, Portugal; IBB-Institute for Bioengineering and Biosciences, Instituto Superior Te´cnico, Universidade de Lisboa, Lisbon, Portugal KUNIHIKO GEKKO  Hiroshima Synchrotron Radiation Center, Hiroshima University, Higashi-Hiroshima, Japan LISA GERLACH  Department of Biophysics, Institute of Biology, FB10 and Center for Interdisciplinary Nanostructure Science and Technology (CINSaT), University of Kassel, Kassel, Germany OMKOLSUM GHOLAMI  Department of Biophysics, Institute of Biology, FB10 and Center for Interdisciplinary Nanostructure Science and Technology (CINSaT), University of Kassel, Kassel, Germany STEPHEN C. L. HALL  School of Biosciences, University of Birmingham, Edgbaston, Birmingham, UK FRANK HEINRICH  Department of Physics, Carnegie Mellon University, Pittsburgh, PA, USA; National Institute of Standards and Technology Centre for Neutron Research, Gaithersburg, MD, USA SAMUEL HERIANTO  High Throughput Biosensing Laboratory, Department of Food Safety/ Hygiene and Risk Management, College of Medicine, National Cheng Kung University, Tainan City, Taiwan SASCHA HERWIG  Department of Biophysics, Institute of Biology, FB10 and Center for Interdisciplinary Nanostructure Science and Technology (CINSaT), University of Kassel, Kassel, Germany VESNA HODNIK  Department of Molecular Biology and Nanobiotechnology, National Institute of Chemistry, Ljubljana, Slovenia; Department of Biology, Biotechnical Faculty, University of Ljubljana, Ljubljana, Slovenia DANIEL A. HOLDBROOK  Bioinformatics Institute (BII), Agency for Science, Technology and Research (A∗STAR), Singapore, Singapore CHANGYU HUANG  Molecular Foundry, Lawrence Berkeley National Laboratory, Berkeley, CA, USA ROLAND G. HUBER  Bioinformatics Institute (BII), Agency for Science, Technology and Research (A∗STAR), Singapore, Singapore HELGI I. INGO´LFSSON  Biosciences and Biotechnology Division, Physical and Life Sciences Directorate, Lawrence Livermore National Laboratory, Livermore, CA, USA WILLIAM A. IRVINE  Centre for Theoretical Chemistry and Physics, Institute of Natural and Mathematical Sciences, Massey University, Auckland, New Zealand NABIN KANDEL  Department of Physics, University of Central Florida, Orlando, FL, USA SYMA KHALID  School of Chemistry, University of Southampton, Southampton, UK JUDY E. KIM  Department of Chemistry and Biochemistry, University of California, San Diego, La Jolla, CA, USA JO¨RG H. KLEINSCHMIDT  Department of Biophysics, Institute of Biology, FB10 and Center for Interdisciplinary Nanostructure Science and Technology (CINSaT), University of Kassel, Kassel, Germany TIMOTHY J. KNOWLES  School of Biosciences, University of Birmingham, Edgbaston, Birmingham, UK ZOLTA´N KO´TA  Biological Research Centre, Institute of Biophysics, Szeged, Hungary SVETLANA KUCHER  Faculty of Chemistry and Biochemistry, Ruhr University Bochum, Bochum, Germany

Contributors

xiii

ARTHUR LAGANOWSKY  Department of Chemistry, Texas A&M University, College Station, TX, USA JEREMY H. LAKEY  Institute for Cell and Molecular Biosciences, Newcastle University, Newcastle Upon Tyne, UK DONGSHENG LEI  Molecular Foundry, Lawrence Berkeley National Laboratory, Berkeley, CA, USA JINPING LI  Department of Biomedical Sciences, Mercer University School of Medicine, Savannah, GA, USA JIANFANG LIU  Molecular Foundry, Lawrence Berkeley National Laboratory, Berkeley, CA, USA ERWIN LONDON  Department of Biochemistry and Cell Biology, Stony Brook University, Stony Brook, NY, USA LUI´S M. S. LOURA  Faculdade de Farma´cia, Universidade de Coimbra, Coimbra, Portugal; Centro de Quı´mica de Coimbra, Universidade de Coimbra, Coimbra, Portugal NAJET MAHMOUDI  Rutherford Appleton Laboratory, Science and Technology Facilities Council, Didcot, Oxfordshire, UK JAN K. MARZINEK  Bioinformatics Institute (BII), Agency for Science, Technology and Research (A∗STAR), Singapore, Singapore KOICHI MATSUO  Hiroshima Synchrotron Radiation Center, Hiroshima University, Higashi-Hiroshima, Japan MATTHIAS MICHALEK  Institut de Chimie, CNRS, Universite´ de Strasbourg, UMR 7177, Strasbourg, France JONAS MU¨CKSCH  Max Planck Institute of Biochemistry, Martinsried, Germany SØREN B. NIELSEN  Technology and Functionality, R&D, Protein Chemistry & Functionality, Arla Foods Ingredients Group P/S, Videbæk, Denmark; Department of Molecular Biology and Genetics, Interdisciplinary Nanoscience Center (iNANO), Aarhus University, Aarhus C, Denmark DANIEL E. OTZEN  Department of Molecular Biology and Genetics, Interdisciplinary Nanoscience Center (iNANO), Aarhus University, Aarhus C, Denmark TIBOR PA´LI  Biological Research Centre, Institute of Biophysics, Szeged, Hungary JOHN W. PATRICK  Department of Chemistry, Texas A&M University, College Station, TX, USA YEVHEN POLYHACH  Laboratory of Physical Chemistry, ETH Zurich, Zurich, Switzerland KUMUD RAJ POUDEL  Analytical Development, Juno Therapeutics, Seattle, WA, USA MANUEL PRIETO  Centro de Quı´mica Fı´sica Molecular, Institute of Nanosciences and Nanotechnologies, Instituto Superior Te´cnico, Lisbon, Portugal; IBB-Institute for Bioengineering and Biosciences, Instituto Superior Te´cnico, Universidade de Lisboa, Lisbon, Portugal JESUS RAYA  Institut de Chimie, CNRS, Universite´ de Strasbourg, UMR 7177, Strasbourg, France GANG REN  Molecular Foundry, Lawrence Berkeley National Laboratory, Berkeley, CA, USA ALEKSANDRA SˇAKANOVICˇ  Department of Molecular Biology and Nanobiotechnology, National Institute of Chemistry, Ljubljana, Slovenia EVGENIY S. SALNIKOV  Institut de Chimie, CNRS, Universite´ de Strasbourg, UMR 7177, Strasbourg, France FIRDAUS SAMSUDIN  School of Chemistry, University of Southampton, Southampton, UK RAJESHWER S. SANKHALA  School of Chemistry, University of Hyderabad, Hyderabad, India

xiv

Contributors

K. TANUJ SAPRA  Department of Biosystems Science and Engineering (D-BSSE), Eidgeno¨ssische Technische Hochschule (ETH) Zurich, Basel, Switzerland MARY A. SCHULER  Department of Biochemistry, University of Illinois, Urbana, IL, USA; Department of Cell and Developmental Biology, University of Illinois, Urbana, IL, USA NICOLE SCHU¨RMANN  Department of Biophysics, Institute of Biology, FB10 and Center for Interdisciplinary Nanostructure Science and Technology (CINSaT), University of Kassel, Kassel, Germany ANDRE SCHU¨ßLER  Department of Biophysics, Institute of Biology, FB10 and Center for Interdisciplinary Nanostructure Science and Technology (CINSaT), University of Kassel, Kassel, Germany PETRA SCHWILLE  Max Planck Institute of Biochemistry, Martinsried, Germany BHANU PRATAP SINGH  School of Chemistry, University of Hyderabad, Hyderabad, India STEPHEN G. SLIGAR  Department of Biochemistry, University of Illinois, Urbana, IL, USA; Department of Chemistry, University of Illinois, Urbana, IL, USA MUSTI J. SWAMY  School of Chemistry, University of Hyderabad, Hyderabad, India SUREN A. TATULIAN  Department of Physics, University of Central Florida, Orlando, FL, USA EVAN L. TAYLOR  Department of Chemistry, Washington State University, Pullman, WA, USA HAO WU  Molecular Foundry, Lawrence Berkeley National Laboratory, Berkeley, CA, USA; Department of Computer Science, College of Information Science and Technology, Beijing Normal University, Beijing, China WEI XIE  State Key Laboratory for Biocontrol, School of Life Sciences, The Sun Yat-Sen University, Guangzhou, Guangdong, China; Center for Cellular and Structural Biology, The Sun Yat-Sen University, Guangzhou, Guangdong, China MENG ZHANG  Molecular Foundry, Lawrence Berkeley National Laboratory, Berkeley, CA, USA HENG ZHU  Department of Pharmacology and Molecular Sciences/High-Throughput Biology Center, School of Medicine, Johns Hopkins University, Baltimore, MD, USA

Chapter 1 Multiscale Modeling and Simulation Approaches to Lipid–Protein Interactions Roland G. Huber, Timothy S. Carpenter, Namita Dube, Daniel A. Holdbrook, Helgi I. Ingo´lfsson, William A. Irvine, Jan K. Marzinek, Firdaus Samsudin, Jane R. Allison, Syma Khalid, and Peter J. Bond Abstract Lipid membranes play a crucial role in living systems by compartmentalizing biological processes and forming a barrier between these processes and the environment. Naturally, a large apparatus of biomolecules is responsible for construction, maintenance, transport, and degradation of these lipid barriers. Additional classes of biomolecules are tasked with transport of specific substances or transduction of signals from the environment across lipid membranes. In this article, we intend to describe a set of techniques that enable one to build accurate models of lipid systems and their associated proteins, and to simulate their dynamics over a variety of time and length scales. We discuss the methods and challenges that allow us to derive structural, mechanistic, and thermodynamic information from these models. We also show how these models have recently been applied in research to study some of the most complex lipid–protein systems to date, including bacterial and viral envelopes, neuronal membranes, and mammalian signaling systems. Key words Molecular dynamics (MD) simulation, Molecular modeling, Protein–lipid interactions, Lipid-binding protein, Membrane proteins, Membrane peptides, Multiscale, Coarse-grained (CG) models

1

Introduction Lipid membranes are ubiquitously used by nearly all life forms to separate themselves from their environment, as well as compartmentalize different functional areas internally. This compartmentalization necessitates on the one hand a mechanism for construction and maintenance of the lipid barriers, and on the other hand a means to transport substances and transmit signals across these membranes to allow for the uptake of nutrients, reaction to changes in environmental conditions, cell–cell communication, etc. [1, 2]. These functions are usually provided by various proteins. In the context of lipid homeostasis, the low solubility of

Jo¨rg H. Kleinschmidt (ed.), Lipid-Protein Interactions: Methods and Protocols, Methods in Molecular Biology, vol. 2003, https://doi.org/10.1007/978-1-4939-9512-7_1, © Springer Science+Business Media, LLC, part of Springer Nature 2019

1

2

Roland G. Huber et al.

lipid species in water necessitates an array of transport proteins to shuttle the lipids from their sites of synthesis to their terminal locations. In the context of signaling and cross-membrane transport, membrane-associated and membrane-embedded proteins play a crucial role. These can be roughly divided into peripheral membrane-associated and transmembrane peptides and proteins. Major classes of transmembrane proteins are cellular receptors and a variety of pores and channels. It is estimated that 20–30% of genes encode membrane proteins in most species [3] and their crucial role in cellular signaling makes them highly attractive drug targets [4]. Unfortunately, the study of this interesting class of proteins is made more challenging by a variety of biophysical phenomena. Membrane proteins are notoriously hard to crystallize, and hence it is difficult to obtain reliable structural information in many cases [5]. Moreover, membrane proteins may not easily adopt their native fold in the absence of a lipid bilayer environment [6]. Their exposed hydrophobic surfaces can also interfere with expression and purification [5]. Accurate modeling of transmembrane and membraneassociated peptides and proteins can therefore provide an attractive pathway to the investigation of these important biomolecules. Computational modeling of proteins in aqueous solution is a well-established research tool and is used ubiquitously in biophysical research and drug discovery [7–13]. The special attraction for computational modeling for protein–lipid interactions lies in the fact that a suite of highly optimized modeling techniques can be applied to research problems that are difficult to tackle with conventional experimental techniques. Molecular dynamics (MD) simulations represent the biological systems in atomistic or nearatomistic resolution, and apply physics-based interaction potentials among the interaction sites [7]. In conjunction with application of appropriate thermostats/barostats for maintenance of the system ensemble (see Note 1), the equations of motions can be propagated forward in time to obtain a realistic trajectory of the particles in the system. A large amount of effort has gone into the derivation of appropriate interaction potentials over recent decades, yielding a variety of different force fields that can be applied to a system of interest (see Notes 2–4). Most force fields consist of a similar set of additive functions, which may be separated into bonded and nonbonded terms. The former broadly consist of harmonic bond and angle terms, and typically sinusoidal functions for torsion angles. The latter consist of pairwise potentials for charge–charge interactions, described using Coulomb’s law, and van der Waals interactions, described via the Lennard–Jones potential. The MD simulation technique is in principle relatively independent of scale; interaction sites may consist either of all atoms (fully atomistic), heavy atoms grouped together with nonpolar hydrogens (united atom description, in which only the polar hydrogens

Multiscale Modeling of Lipid–Protein Interactions

3

are treated explicitly), or of even larger groups of atoms (coarsegrained description). Extremely coarse-grained (CG) models can group hundreds of atoms into single interaction sites (e.g., to investigate the dynamics of large viruses) [14–16]. The key challenge in generating accurate models is the derivation of model parameters. Whereas there exists a well-established set of fully atomistic and united-atom force fields for both proteins (mainly CHARMM [17, 18], Amber [19, 20], GROMOS [21]) and lipids (Slipids [22–25], Lipid14 and 17 [26] CHARMM [83], GROMOS [27, 28]), only a limited number of coarser, largerscale force fields readily exist, the most prominent of which is MARTINI [29–32] which offers parameters for proteins, lipids, and other biomolecules. However, a multiscale approach can be used to validate a customized, coarser model to ensure that it faithfully reproduces the desired observables of a more detailed approach. Although a CG model will omit details that could be discerned in atomistic simulations, these models offer access to time and length scales inaccessible at higher detail [29, 33–35]. Moreover, mapping between different representations is possible, and mapping procedures are an active area of research [29, 35–38]. Although GPU acceleration has allowed for a considerable speedup of molecular simulations [20, 39], atomistic resolution simulations are currently limited to the hundreds of nanoseconds to microsecond time scale for a typical membraneembedded protein system, unless specialized hardware is used [40, 41]. CG simulation approaches at the scale of the MARTINI force field, which maps approximately four heavy atoms to one interaction site, typically allow for simulations of the same systems over tens to hundreds of microseconds, or more commonly, for studies of much larger systems, for example, to study aggregation or oligomerization phenomena [42, 43]. We will now outline the basic prerequisites, key tools and techniques, and common analysis approaches to the computational modeling of protein–lipid systems. We will also present an expose´ of the successful application of the presented techniques to biological research problems.

2

Materials

2.1 Survey of Structural Information

In order to construct a computational model of a protein–lipid system, structural information is crucial. Most known structures of biomolecules are deposited in the Protein Data Bank [44, 45] (http://www.rcsb.org/) and are freely accessible for download. The contents of the PDB predominantly comprise of protein structures determined by X-ray crystallography, but in recent years, the quantity of solution NMR structures [46], and cryo-electron microscopy (EM) structures [47] has been increasing steadily.

4

Roland G. Huber et al.

Available structures range from small peptides and single domain proteins to large-scale multiprotein or heterogeneous complexes such as the human ribosome or complete virus particles. A good starting structure for a molecular simulation may contain proteins, nucleic acids, or ligands of all sorts, but it is necessary that the structure has a sufficient resolution and that the key areas of interest are resolved [48, 49]. Resolution is mainly a limiting factor in the case of cryo-EM structures, but recent technological advances [47, 50] now allow for structure determination at near-atomistic resolution. NMR and X-ray structures may have missing dynamic or disordered regions, as these properties preclude the collection of high-quality experimental data in these regions. This is of special concern for membrane proteins [51]. Such missing regions must be carefully considered during the modeling process (see Note 5). At this stage, it is also crucial to familiarize oneself with the particular systems for which the structural measurements were undertaken, as sometimes the coordinates may have been determined for truncated subdomains, mutant or hybrid variants, or obtained under specific pH/solvent conditions that may or may not correspond to the conditions that are to be modeled. 2.2 Choice of Representation

Following a survey of available structural information, one needs to determine how the system shall be represented in the model. A selection of different force fields and resolution scales is readily available at the atomistic and near-atomistic scales (see Notes 2 and 3). Whereas some force fields try to cover as broad a chemical space as possible (a property highly valued for modeling specific ligands), other force fields specialize in a particular classes of system (e.g., lipids, proteins, or nucleic acids) and are not necessarily easily combined with others. Moreover, the choice of representation is also influenced by the research objectives: modeling the specific interactions of small ligands or individual lipids with a protein may call for a fully atomistic representation, whereas in studying, for example, large-scale aggregation behavior, such a choice may be detrimental, as the resulting loss in accessible time and length scales outweighs the more accurate description of interactions and hence a coarser representation would be appropriate.

2.3 Structure Preparation

Before a specific protein–lipid system can be modeled, several preparatory steps are necessary to ensure that the modeled system approximates the state of the biological system as close as possible. Common tasks in this pipeline include: The modeling of missing loops [52–54] and the reversal or introduction of specific mutations [52, 53] (see Note 5); deciding on the protonation state of acidic and basic side chains appropriate to the pH conditions that are to be modeled; inclusion of physiologically appropriate disulfide bonds; capping of protein chains at the N- and/or C-termini if a truncated protein is to be modeled; and addition of solvent,

Multiscale Modeling of Lipid–Protein Interactions

5

counterions, and/or structural ions to achieve a biologically realistic system (see Note 6). In addition, modeling a lipid environment of appropriate composition and phase should be a key aim for studying any biologically relevant membrane phenomena. The ultimate goal is to obtain a model of the protein–lipid system that faithfully reproduces as many properties of the biological system as possible. It is therefore highly recommended that this step is undertaken with the utmost diligence, and that all desired inputs and properties are double-checked and verified, as at this point no significant computational resources have been expended, and later discovery of mistakes at this stage may invalidate weeks or months of simulation calculations. 2.4 Membrane Embedding

One of the key tasks for the simulation of membrane-associated and membrane-embedded proteins is their placement inside a representative membrane, for which existing coordinates and topologies are available (see Note 2). Several tools and methods are available that assist in the construction of lipid bilayers as well as the embedding [55–58] of proteins within them (see Note 4 and Subheading 3 for further details). A key facet of membrane protein simulations is that there is a fundamental anisotropy within the system, as the surface tension of the membrane needs to be modeled independently from the global pressure scaling of the full system (see Note 1).

2.5 Multiscale Models

It is often desirable to conduct simulations at a CG level, as the spatial and temporal scales accessible are usually 1–2 orders of magnitude increased. However, it is necessary to ensure that the CG representations are able to accurately reflect the behavior observed at a higher level of accuracy in the key properties of interest. Hence, a multiscale approach [36, 38, 59] is often pursued, where high-resolution atomic-scale models are used to derive and validate coarser models, which can then in turn be used to sample slow processes in large systems, while having confidence that the individual model components retain appropriate behavior. Conversely, parts of CG simulations can be extracted and can be back-mapped to higher-resolution representations, where after some refinement these structures are either used as the starting point for further simulations or are analyzed with respect to specific interactions.

2.6 The Simulation Loop

The basic idea behind MD simulations is to start at a specific structure close to the system equilibrium, assign velocities corresponding to a specific temperature, and then propagate the motions of all particles forward for a small time step. At each time step, the forces that act on each particle are calculated from the gradients of the pair potentials that constitute the force field. These forces are then applied to each individual particle, changing their velocity, which is then again propagated forward in time yielding

6

Roland G. Huber et al.

new particle positions. At these new positions the forces are again evaluated, thus concluding the simulation loop. This simulation loop is performed for as many steps as desired to obtain a trajectory for each individual particle at each point in time. The key limiting factor in this procedure is the magnitude of the time step, which has to be small enough to prevent even high-velocity particles from deviating too far outside equilibrium bond lengths, angles, etc. Such deviation is unphysical and leads to very large reaction forces, which introduces numerical instability that may cause the simulation to crash. Advanced integration methods, such as a Verlet-type integrator, offer added numerical stability and energy conservation. Typical time step durations for atomistic simulations are 1–2 fs, subject to the use of constraints to stabilize high-frequency bonds or angles such as those associated with hydrogen atoms, whereas even moderately CG simulations typically allow time steps upward of 15 fs. As a consequence, obtaining a 1 μs trajectory of an atomistic simulation requires the simulation loop to be evaluated 0.5–1 billion times, which requires considerable computational resources. A variety of software packages is available to perform these calculations, the most broadly used of which are GROMACS [60–62], NAMD [63, 64], CHARMM [17], and Amber [19, 20, 65]. 2.7 Simulation Protocol

Following careful system preparation, a simulation run is usually preceded by an energy minimization protocol. The main purpose of this minimization is to reduce or eliminate close contacts, overlaps, or otherwise energetically highly unfavorable situations that would introduce a steep energy gradient and hence produce large forces in the initial steps of a simulation. A thorough minimization may be conducted in several steps, keeping different parts of the system restrained (e.g., one may initially only allow the solvent to move). Following minimization, an equilibration procedure is generally performed. This may take the form of a “thermalization” protocol to ensure that the system is equilibrated at the temperature of interest, ideally along with one or more simulations in which initially strong harmonic restraints are placed on protein and/or lipid components, allowing for temperature and pressure to stabilize and the solvent to adjust to the solute. These restraints are gradually relaxed during equilibration until a stable system is obtained without restraints. A proper equilibration protocol should retain all key structural characteristics of the investigated proteins. Production simulations of membrane–protein systems are usually conducted as unrestrained NPT calculations using semiisotropic pressure scaling to account for membrane surface tension, at a temperature above that of the gel-to-liquid phase transition. A low root-mean-square deviation (RMSD) of the protein backbone with respect to its initial/experimental coordinates is usually a good metric to judge stability of a protein during equilibration, while it is advisable to check for stability of key bilayer properties over time, such as mean

Multiscale Modeling of Lipid–Protein Interactions

7

area per lipid (APL), and density of individual lipid components with respect to the membrane normal. It is difficult to propose a specific equilibration protocol for all systems, as the amount and style of equilibration required will differ from system to system, depending amongst other things upon the quality of the initial structure, the number and character of changes introduced during structure preparation, removal or addition of ligands and ions, and the insertion protocol within the lipid membrane. Basic Analysis

All simulations will yield a trajectory, but not all simulation trajectories are meaningful. Hence, a simulation result needs to be carefully inspected to ensure that the data produced is robust. Key facets of successful simulations are that the proteins largely retain their equilibrium conformation, lipid membranes are stable, cofactors and ligands remain bound, and temperature, pressure, and energy of the system remain stable. Most of these parameters can be assessed by plotting simple graphs (RMSD, energy vs. time plots, etc.), and by visual inspection of the trajectory using a molecular graphics viewer such as VMD [66]. It is not unusual to observe a certain amount of drift in some or all of these parameters in the early stages of the simulation, as the components find their equilibrium state under the influence of the chosen force field. However, during further analysis, such periods of drift should be excluded and the processing should focus on the converged parts of the trajectory, where system equilibrium states are sampled. If the system does not reach an equilibrium state, it might be helpful to extend the duration of the simulation, or to reconsider earlier choices in how the system model was constructed.

2.9 Computational Considerations

As previously suggested, the computational demand of molecular simulations is considerable whenever large systems and/or long time scales are of interest for a particular research problem. The computational cost of the MD technique is in large part due to the calculation of the nonbonded interactions, in particular due to electrostatics (see Note 7). Whereas simple calculations can be carried out on a desktop computer, it is usually desirable to have some sort of high-performance computing (HPC) facility available. Modern-day simulation packages are designed to make efficient use of parallel processing, and can utilize a variety of different computer architectures and coprocessors [61]. In recent years, GPU computing has contributed to a significant speedup of simulation runs, as the massively parallel nature of GPUs is well suited to the force computations characteristic of MD [20, 65]. Most simulation packages are distributed in source code form and can hence be compiled and/or modified to take advantage of specific architectures. In addition to general-purpose HPC hardware, some work has gone into the design of application-specific integrated circuits (ASIC) dedicated to MD simulations. The Anton machines created

2.8

8

Roland G. Huber et al.

by DE Shaw Research [40, 41] make use of such ASICs and have enabled access to millisecond time scales for biologically relevant systems in full atomic resolution.

3

Methods

3.1 Setting Up Membrane Protein and Membrane Peptide Systems

Once the coordinates for the protein model have been prepared, it is necessary to embed them in an appropriate lipid environment prior to solvation (see Note 4). The slow dynamics associated with the exchange of lipids means that care must be taken to correctly position a protein within a model membrane or mimetic environment, to ensure that biologically meaningful results are obtained from the subsequent production simulation. Several approaches are outlined below for building biologically or experimentally relevant systems. The starting configurations for transmembrane protein simulations are typically created by using one, or a combination of, two strategies: (a) the lipid membrane is constructed around the protein; or (b) the protein is inserted directly into a preexisting, equilibrated lipid membrane (see Note 2). Both require some relaxation of the system before production data can be collected. The first of these strategies is utilized by the popular CHARMM‑GUI online server [56] and “insane” script for CG simulations [58]. In certain circumstances, however, the second of these strategies is preferred, particularly if the insertion process minimally perturbs the membrane environment and thereby reduces the time required for the system to relax.

3.1.1 Inserting a Protein Within a Lipid Bilayer Environment

1. The coordinates of some preequilibrated lipid bilayer mixtures are available freely online (see Note 2), but in cases where a complex mixture is required the bilayer will need be constructed. This may be carried out via a number of available online tools (see Note 4), or may be performed “in house” using techniques similar to those that construct the lipid bilayer with the protein in situ. The speed of construction of an isolated lipid membrane can be increased by first forming a small, equilibrated bilayer “patch” containing ~16 or so lipids in each leaflet. The initial configuration of the lipids in the patch is biased so that they closely resemble a lipid bilayer: by placing the individual lipids on grid points or by algorithmically packing them into a bilayer configuration (see also Subheading 3.3). The assembled bilayer patch is then relaxed for approximately 100–1000 ns, or until an observable property such as the APL has plateaued. After the lipid bilayer has stabilized, the final frame is taken and the patch is replicated in the x and y dimensions (assuming z to be the normal to the bilayer) until the membrane is large enough in size to accommodate the transmembrane protein of interest.

Multiscale Modeling of Lipid–Protein Interactions

9

2. The correct initial position of the protein relative to the plane of the bilayer can be estimated by eye, using the position of the amphipathic aromatic “girdles” that anchor themselves in the lipid–head group interface [67]. Alternatively, empirical approaches can be used to facilitate positioning, such as the Orientations of Proteins in Membranes (OPM) database [68] and associated server (http://opm.phar.umich.edu/server. php) which predicts the protein location based on the transfer energy of accessible amino acids between water and the membrane core. 3. A straightforward approach for membrane embedding termed g_membed developed by Wolf et al. [55], related to another method described by Yesylevskyy [69], involves growing an initially “contracted” protein back to its correct size in the membrane plane while pushing lipids away during a short MD simulation, minimally disrupting the lipid bilayer. Similarly, the method of Faraldo-Gomez et al. [70], uses an implicit protein grid-based force field for specificity at the protein–lipid interface and applies weak repulsive forces to nearby lipid molecules perpendicular to the solvent-accessible protein surface during multiple short MD simulations, resulting in a volume adapted to the protein surface with only minimal perturbation of the existing bilayer structure. 4. Another approach, based on the LAMBADA and inflateGRO2 [71] tools, works sequentially to perform the positioning and insertion of the protein. LAMBADA decides the appropriate orientation for insertion by calculating a hydrophilicity score along the protein’s axis as it is tilted at angles relative to the bilayer plane. Once positioned, inflateGRO2 differentiates lipids based on their overlap with the protein, removing those with high overlap and relaxing the lipids that exist in the inner most annular layers surrounding the protein. 5. Alternatively, one of the simplest strategies to insert a transmembrane peptide or protein in an equilibrated lipid bilayer is to carefully superimpose the coordinates of both and then to remove any protein-overlapping lipid molecules. This insertion method can result in excessively large holes in the bilayer with the lipids suboptimally packed around the protein, especially for nonuniform proteins with irregular surface shapes. However, with increasing computational power and faster algorithms, it is now often possible to simply use an extended “equilibration period” to fix these issues. In this approach, positional restraints are applied to the protein coordinates to prevent structural drift during an MD simulation, during which local lipid and solvent molecules can relax freely around the protein surface. Subsequently, protein restraints may be gradually released, prior to production MD.

10

Roland G. Huber et al.

6. The above methodologies may fail in the case of proteins whose surfaces are particularly irregular or asymmetric, potentially leading to large gaps within the lipid bilayer that are not easily equilibrated. A possible solution to this problem is to use alchemical transformation to “grow” the peptide within the membrane phase. Alchemical methods have traditionally been used to calculate the free energy (ΔG) between two thermodynamic states [72]; in practice, this takes the form of a series of transformations between nonphysical intermediates defined as a function of the coupling parameter λ. Such an approach has proven useful for peptides that exist parallel to the membrane surface, but buried within the bilayer. For example, immunoreceptor tyrosine-based activation motifs (ITAMs) containing conserved YxxL/I sequences have been proposed to be embedded in this manner. Phosphorylation of ITAM tyrosines present in the cytoplasmic tails of T-cell receptor (TCR) associated chains serve to propagate antigen-induced activation, and it has been proposed that this may be regulated by the extent of peptide sequestration within the lipid bilayer [73]. To investigate this further, the CD3ε ITAM peptide [73] was alchemically “grown” at different positions within a membrane model. This was achieved via a series of short MD simulations, with the protein coordinates weakly restrained, starting from λ ¼ 0 (peptide absent) and increasing in successive λ ¼ 0.05 increments until λ ¼ 1 (peptide fully inserted). To avoid sudden atomic overlaps due to nascent peptide-lipid interactions, which may cause simulation instabilities, best results were achieved by: (a) independently “switching on” the Lennard–Jones interactions first, followed next by introducing the point charges of the peptide atoms; and (b) using a soft-core potential to avoid endpoint errors [74]. Jefferys et al. coined the term “alchembed” for a variant of this strategy, and have made available a tutorial for the GROMACS simulation package [75]. The ITAM simulations suggested that burial of the peptide within the bilayer hydrophobic core would likely lead to significant lipid deformation and membrane instability (Fig. 1). Instead, in the resting state, the ITAM peptide likely sits at the membrane interface, where the tyrosine residues can still interact with lipids. 7. Finally, a “brute force” simulation strategy entails placing a protein in a random mixture of lipid and solvent, prior to an extended simulation, leading to spontaneous self-assembly into a membrane-inserted state, thereby avoiding the need for user intervention [76]. With appropriate lipid–solvent ratios, micelles or other nonlamellar phases may also be achieved. Unfortunately, the compute times for performing such simulations in full-atomic resolution are costly, and this method can

Multiscale Modeling of Lipid–Protein Interactions

11

Fig. 1 Alchemical insertion of ITAM peptide into a lipid bilayer within the hydrophobic core (top row) or at the membrane interface (bottom row). Representative snapshots are shown at three different values of coupling parameter λ, indicated inset. Lipid head groups are indicated in CPK wireframe format; the ITAM peptide is shown in green ribbons format, with basic side chains (blue) and tyrosines (red) represented in licorice format. Lipid tails and water molecules are omitted for clarity

be somewhat unpredictable in terms of the final system configuration that is obtained. Fortunately, CG approaches can help to solve this problem [77], yielding significant speed-up and enabling rapid testing of multiple starting conditions, such as system component concentrations and lipid composition. Such a strategy can be useful for complex, multicomponent systems; for example, it has been applied to understand why some thrombin-derived C-terminal peptides in wound fluids form antibacterial amyloid-like particles [43], while others exert antiseptic activity by binding to endotoxin aggregates or immune receptors [78]. Furthermore, once a desired CG protein–lipid assembly has been generated, it is possible to “backmap” or “reverse transform” the coordinates to all-atom representation for detailed refinement. This may be achieved via alignment against libraries of molecular fragments combined with homology modeling approaches [79], or using a tool based on geometric projection and subsequent cycles of relaxation based on energy minimization and position-restrained MD [80], which enables straightforward conversion of MARTINI systems to their atomic counterparts for a variety of common force fields. Such a strategy can, for example, prove useful in predicting both the global conformation and detailed atomic interactions of membrane protein oligomers in a lipid bilayer environment [81]. 3.1.2 Setting Up Peripheral Membrane Protein Systems

Peripheral membrane-bound proteins adhere temporarily to the surface of the lipid bilayer, and are important in a wide variety of cellular functions, including, for example, regulatory roles in channels and receptors, enzyme targeting, and signaling in

12

Roland G. Huber et al.

protein–protein complexes. Stable binding of such proteins often involves electrostatic interactions, which act over long distances between charged amino acids and lipids. While approximate methods are available to predict association of peripheral proteins with membranes, these tend to lack details of specific interactions which may be biologically important. Thus, a new method named Rotational Interaction Energy Profiling (RIEP) has recently been developed to rapidly evaluate the electrostatically optimal orientation of a protein with a lipid bilayer of specific composition, yielding configurations for subsequent seeding of MD simulations [82]: 1. The aim is to rapidly evaluate and identify optimal protein orientation(s) with respect to the membrane on the basis of electrostatics. The procedure allows the characterization of membrane–protein association, the identification of important residues, and initiation of MD simulation of the binding process from predetermined ideal orientation(s). 2. The requirements for this method are: (a) Python wrapper script for calling GROMACS [61] tools, available from https://github.com/allison-group/riep. This can run across multiple nodes of a compute cluster. (b) GROMACS software, version 3 onward. (c) Separate coordinate, topology, and index files for the equilibrated protein and membrane components. 3. The procedure, as outlined in Fig. 2, is as follows: (a) Rotate protein coordinates around pitch, roll, and yaw in user-determined degree increments (rotation increments of 30 are recommended). (b) Place the rotated protein at a user-determined minimum distance from the membrane. A minimum protein–mem˚ is recommended. Note that this brane distance of 5 A distance may change during energy minimization, typically becoming closer for favorable orientations. (c) Combine the rotated protein coordinates with the nonrotated membrane coordinates, and subsequently solvate and energy minimize the system. (d) Initiate a short MD simulation of the protein–membrane system for a user-determined number of steps. Fewer than 25 integration steps are recommended; the goal of this and the energy minimization is to relax the system at minimal computational expense. (e) Calculate protein–membrane Coulombic and Lennard–Jones potential energies for each set of protein–membrane coordinates. Optimal orientations of the protein for membrane association are those with the lowest energy. Many extrinsic membrane proteins associate with the membrane via attractive electrostatic interactions; thus, the

Multiscale Modeling of Lipid–Protein Interactions

13

Fig. 2 An efficient protocol for setting up optimal peripheral membrane protein systems. Note that each of the labeled steps correspond to those in the text

Coulombic potential energy is typically of higher magnitude and most informative. (f) Visualize the optimal coordinates, for example, with VMD [66]. (g) Finally, to calculate per-residue decomposition of protein–membrane energy for a given orientation, residues are selected within the Coulombic cutoff distance of the membrane, and the short MD simulation rerun, using the identified residues of interest as energy groups. 3.2 Preparation of Complex Membrane Protein Systems 3.2.1 Lamellar and Nonlamellar Starting Structures with CHARMM-GUI

It is often desirable to investigate the properties of a specific combination of lipids organized in a nonplanar configuration. When a CG self-assembly-based strategy (see Subheading 3.1.1) is insufficient, the CHARMM-GUI server (http://www.charmm-gui.org) may be used to create such systems [56]. Recently, the membrane builder has expanded its repertoire of lipids to include over 180 different variants [83], and it additionally supports the use of CG lipids from the MARTINI force field. The chosen lipids can be assembled into the lamellar and nonlamellar lipid structures of vesicles, micelles, and hexagonal phase membranes [57, 84].

14

Roland G. Huber et al.

3.2.2 A Protocol for Simulating Enveloped Virus Particles

Beyond the challenge of predicting the orientation of membrane peptides and proteins within a lipid membrane, MD simulations can also be used in refining large multicomponent protein–lipid complexes, such as enveloped viruses, in which a lipid vesicle derived from the cell is “coated” by embedded viral proteins [14]. This was recently demonstrated for the entire dengue virus envelope particle, with a diameter of ~50 nm, based on a combination of data from atomic-resolution and CG MARTINI-based simulations along with cryo-EM [42]. The steps required to achieve this are outlined below, and illustrated schematically in Fig. 3. 1. Initially, atomic-resolution simulations should be performed for isolated viral envelope protein subunits and/or small assemblies thereof, either for the ectodomains in solution, or for the full-length constructs containing transmembrane regions embedded within a small lipid bilayer patch. This provides a measure of the dynamics and stable structural properties of the proteins, enabling subsequent calibration of the CG model. A straightforward way to map such dynamics between resolutions is to utilize an elastic network within the CG MARTINI model [29], and to iteratively tune the associated parameters (i.e., cutoff distances and harmonic potential force constants) until comparable protein flexibility is achieved at both resolutions. 2. At the same time, regions such as flexible loops or transmembrane helices missing from the cryo-EM structure should be reconstructed (see Subheading 2). Subsequently, mapping of the entire viral atomic protein coordinates into MARTINI representation should be conducted, along with an energy minimization protocol to ensure the absence of steric clashes (Fig. 3b). 3. A CG viral vesicle may now be built, for example, using the CHARMM-GUI Martini Maker [57], in accordance with the diameter estimated from cryo-EM maps or other experimental measurements, and with a lipid composition guided by available lipidomics data. Due to the tightly packed mesh of proteins typical of (pseudo)icosahedral viral envelopes, a simple process of overlay of protein coordinates and deletion of overlapping lipids is typically insufficient, since most lipid molecules in the vesicle will at least partially coincide with the protein. To overcome this, a procedure involving shrinking of lipids along their principal axes, deletion of remaining overlapping lipids, followed by multiple iterative rounds of energy minimization and protein position-restrained equilibration (Fig. 3a) should result in a reasonable model of the viral protein embedded vesicle [42].

Multiscale Modeling of Lipid–Protein Interactions

15

Fig. 3 Protocol for multiscale modeling and refinement of viral envelope against cryo-EM data. (a) Schematic for stepwise method to embed viral proteins within a lipid vesicle. (b) Multiscale illustration of dengue virus envelope. The simulation construct is shown for the entire viral envelope complex in surface representation on the right. A “zoomed” snapshot on the left shows the dimeric envelope proteins in both atomistic and CG representations (left). Protein is colored according to radial distance from the center of the virus, and lipids are colored green. (c) Alignment of density maps determined by cryo-EM (red) and refined during simulation (blue)

4. The CG enveloped viral protein–lipid complex should next be solvated within a simulation box (typically a dodecahedron or truncated octahedron is useful in minimizing excessive solvation of the spherical virus particle) and progressively equilibrated. Initially, sets of ~10,000–100,000 step protein backbone-restrained equilibration simulations in the NVT ensemble should be run with a short integration time step, incrementally increasing this by ~2–4 fs for each successive simulation. Subsequently, once the maximum integration

16

Roland G. Huber et al.

time step has been reached, longer equilibration runs (e.g., ~10–100 ns) should be run in the NPT ensemble, during which the position restraints on protein coordinates are gradually reduced. This may finally be followed by an unrestrained NPT production run. 5. MDanalysis [85] is an object-oriented Python library which enables analysis of trajectories derived from many MD packages (https://www.mdanalysis.org), and includes tools to generate theoretical density maps from the underlying simulation frames. This may be used to generate a simulation-averaged density map, using a grid spacing in accordance with the resolution of the experimental cryo-EM map (Fig. 3c). The Chimera [86] software package (https://www.cgl.ucsf.edu/ chimera/) enables alignment of the experimental and theoretical maps, by minimizing the mean cosine angle between vectors obtained via trilinear interpolation; this also yields a correlation, providing a measure of agreement between simulation and experiment. 6. Finally, refined viral envelope coordinates may be back-mapped and simulated at atomic resolution (see Subheading 3.1.1), depending on availability of computational resources, enabling fine-grained analysis of lipid–protein interactions (Fig. 3b). 3.3 Setting Up Biologically Realistic Membrane Systems

Building good starting structures for lipid bilayer simulations can be quite involved due to the interwoven nature and long relaxation timescales of lipids. The complexity increases further with bilayer size and number of lipids types due to slow equilibrium of lipid mixing, lipid flip-flop, and bilayer undulation dynamics. Improved bilayer building tools (see Note 4), use of coarser, more forgiving force fields, and faster computation have made bilayer construction easier, but interest in larger membranes (up to ~500  500 nm), complex bilayer geometries (various capsids and organelles), and more complex lipid mixtures (>60 different lipid types) have complicated matters—see for example [87–90]. As simulation complexity begins to approach that of a physiologically relevant membrane composition, one must begin to consider membrane asymmetry and the associated technical difficulties related to its setup and simulation. The primary concern when generating starting configurations for asymmetric membranes is that the average APL varies based on the lipid type and the particular lipid mixture the lipid is in; therefore, for a periotic system, the number of lipids in each membrane leaflet should likely differ. If the aggregated area occupancy in the two leaflets is not the same, then an artificial membrane “frustration” will arise that affects the lateral pressure profile, membrane curvature, and leaflet surface tension. Adjusting the number of lipids in each leaflet based on their projected APL reduces some of the artifacts but does not fully

Multiscale Modeling of Lipid–Protein Interactions

17

Fig. 4 Setup of an asymmetric neuronal plasma membrane. (a) Two simulations of the symmetric outer and inner leaflet mixtures are run to estimate the APL. After setup with the insane tool the asymmetrical mixture is simulated and lipid flip-flop is monitored. If flip-flop is “fast” and asymmetric the setup needs to be iterated with an adjusted lipid mixture. (b) Snapshot of the asymmetric neuronal plasma membrane simulation described in [89] is shown from the top, showing the outer leaflet, and the bottom, showing the inner leaflet. Pie charts with the overall distribution of lipids head group types in the outer/inner leaflet are also shown

resolve the asymmetry. APLs cannot simply be reverse-engineered from measured APLs, as these (a) are usually identified at a specific temperature that may not be appropriate for the simulation in question, (b) are dependent on the local lipid environment, and/or (c) may not even exist for a specific lipid in isolation. Another approach is to artificially induce lipid flip-flop, allowing lipids to equilibrate between the leaflet; this will relieve any area asymmetry, but will also alter the lipid mixtures of the two leaflets. Here, a third method is described based on separate APL measurements for the two leaflets (Fig. 4), which has been used in, for example, [87, 89]: 1. The outer/inner leaflet lipid mixtures are used to instigate two individual simulations; one with the outer leaflet composition as a symmetrical membrane, and another with the inner leaflet composition as a symmetrical membrane. These simulations are run concurrently, and the global lateral areas (in terms of box size) of these membranes are monitored until they reach an

18

Roland G. Huber et al.

equilibrium value. Note that this can take several microseconds for complex membranes. These simulations produce the equilibrated average area for both the inner and outer leaflets. The ratio between these values provides a scaling function to calculate the number of lipids required for each leaflet. This data can be utilized by tools such as insane [58], which provides an “offset” flag to indicate the asymmetry of lipid numbers between the leaflets. 2. Bilayers undulate and their tendency to undulate is lipid mixture dependent. Therefore, to get a better estimate of the average APL of the two different mixtures, either the bilayer surface should be fitted, or the bilayers should be restrained to be similarly flat. This can be done by applying a weak position restraint potential on the head group particles of a major constituent of one of the leaflets. This restraint potential is applied only in the direction of the normal of the bilayer, and only on a single leaflet so as not to affect the bilayer thickness. 3. Having determined the appropriate number of lipids present in each leaflet, one must also then consider the effects of membrane components that have the ability to flip-flop (such as cholesterol) within the timeframe of the simulation. This is especially important for CG simulations that easily reach timescales in which cholesterol can equilibrate between the two leaflets. Deviation of this cholesterol distribution from its original leaflet fractions causes buildup of cholesterol in one of the leaflets and will again lead to artificial membrane “frustration.” 4. In order to adjust for cholesterol flip-flop (and other fast flipflopping lipids), the original, asymmetric lipid system (with corrected leaflet-dependent densities) is simulated until the cholesterol distribution has equilibrated. Again, this may take many microseconds of simulation. If cholesterol displays significant deviation from the starting configuration then these new cholesterol inner/outer leaflet fractions are incorporated back into the original leaflet compositions (and any minor adjustments made accordingly). These updated leaflet compositions with adjusted cholesterol content again need to have their APL/densities calculated via simulation of a pair of symmetric bilayers (repeating step 1). The revised asymmetric bilayer (adjusted for cholesterol equilibration and leaflet offset) is simulated again (repeating steps 2 and 3) to monitor for any further drift in the between leaflet lipid distributions. This process is iteratively repeated until the cholesterol leaflet distributions do not drift from their starting values during simulation. It should be noted that this method is not without its faults. By definition, any bilayer that has an asymmetric leaflet lipid

Multiscale Modeling of Lipid–Protein Interactions

19

composition is not at equilibrium, therefore, what constitutes as fast lipid flip-flop will depend on the application in question. Furthermore, there is the assumption that the APL of the lipids in a restrained, flat membrane is representative of how the lipid would behave in a freely undulating bilayer. 3.4 Beyond Membranes: Bacterial Envelopes and Cell Walls

The cell wall is an often-forgotten component of the bacterial cell envelope. Sandwiched between the inner and outer membranes in an aqueous compartment called periplasm, the cell wall is made of a network of peptide and sugar molecules commonly known as peptidoglycan [91, 92]. The peptidoglycan mesh is linked covalently to the outer membrane via Braun’s lipoproteins [93] and noncovalently to both membranes via integral membrane proteins like OmpA and TolR [94–96]. Earlier simulation studies of the cell wall by Gumbart et al. focused on elucidating its physical properties such as elasticity, pore size, and thickness [97]. However, molecular details of how the cell wall is positioned and interact with other numerous components of the cell envelope remained sparse. We therefore developed atomistic parameters for simulation of peptidoglycan network (Fig. 5) in the presence of Braun’s lipoprotein and OmpA [98, 99]. The Braun’s lipoprotein is anchored to the outer membrane via a lipidated N-terminus and binds the peptide chain of the cell wall on its C-terminus. The length of the Braun’s lipoprotein, therefore, has a direct influence on the distance between the cell wall layer and the outer membrane [100]. Our simulations, however, showed that this is not quite as simple as often suggested. The Braun’s lipoprotein was able to tilt and bend significantly with respect to the outer membrane, effectively shifting the cell wall closer to the latter, during simulations. This smaller gap in turn facilitated the initial binding of OmpA periplasmic C-terminal domain to the cell wall.

Fig. 5 Model of the gram-negative bacterial E. coli inner membrane and peptidoglycan layer. In the inner membrane model below, phospholipid head group phosphorus atoms are shown in cyan. Above, the molecular surface of the peptidoglycan model is shown, with the inset depicting that the strands are 50% cross-linked

20

Roland G. Huber et al.

In fact, in the absence of Braun’s lipoprotein the C-terminal domain of OmpA monomer showed a high propensity to bind to the inner leaflet of the outer membrane instead of the cell wall, further corroborating the instrumental role of Braun’s lipoprotein. Dimerization of OmpA, interestingly, eliminated this dependency as the C-terminal domain bound to the cell wall even without Braun’s lipoprotein, perhaps due to the stronger electrostatic interaction formed in the OmpA dimer. While the functional relevance of OmpA dimerization is still unknown, in vivo cross-linking studies and mass spectrometry showed that the dimeric interface is largely localized in the C-terminal domain [101, 102], suggesting a likely role in improving cell wall binding. Our simulations also indicated that, once bound, the linker connecting the membrane embedded N-terminal beta barrel domain and the peptidoglycan bound C-terminal domain of OmpA to be highly adaptable, which is potentially important to provide a flexible mechanical support for the underlying cell wall network. In essence, these simulations have uncovered important molecular details of the dynamic interplay between the cell wall and some of the components of the gramnegative bacterial cell envelope. 3.5 Thermodynamics of Lipid–Protein Complex Interactions

Free energy calculations allow the assessment of thermodynamics of a given biological process. A convenient approach for binding and transfer processes is the calculation of a potential of mean force (PMF) [103]. The PMF may be obtained by defining a reaction coordinate (e.g., the distance between two binding partners), and then sampling the forces acting on the respective solutes when restrained at a specific distance. Using the weighted histogram analysis method (WHAM) allows to reconstruct the PMF from these forces [104]. In the case of no interaction, the average force should be zero, whereas attractive interactions are indicated by forces pulling the particles to one another. Sampling the PMF along the entire reaction coordinate allows one to obtain the free energy difference between the bound minimum energy and the unbound energy at larger solute distances, which is equivalent to the difference in free energy between these states. The technique of using discrete restrained sampling points to obtain the PMF is called umbrella sampling [105], which is implemented in most major simulation codes. Using this approach, we recently demonstrated how an extensive series of free energy calculations in combination with multiscale models allow us to trace the thermodynamics of transfer of a ligand between multiple protein partners and membranes along a “biological relay” [106] (Fig. 6). The relay in question is the Toll-like receptor 4 (TLR4) system [107, 108]; TLR4 is part of the mammalian innate immune pathway which is associated with the plasma membrane and signals the presence of pathogen invasion by detecting lipopolysaccharide (LPS), the dominant

Multiscale Modeling of Lipid–Protein Interactions

21

Fig. 6 Thermodynamics of LPS transfer from the bacterial membrane to the terminal TLR4/MD-2 receptor complex. In gram-negative bacteria, the outer leaflet of the outer membrane of is largely constituted of LPS, while the inner leaflet contains a mixture of phospholipid species, and the membrane is also interspersed with a variety of outer membrane proteins such as the modeled OmpF porin. Extracting LPS from this membrane requires considerable energetic effort. LPS is subsequently handed from protein to protein in a cascade, following a path of increasing affinity until ultimately bound in the TLR4/MD2 complex. The energetic cost estimated from PMF calculations of each stepwise transfer process in this cascade is indicated inset. Dimerization of the terminal receptor activates downstream signaling, thus enabling an innate immune response

component of the outer leaflet of gram-negative bacterial outer membranes [109]. Due to its high hydrophobicity, the energy required to extract LPS from bacterial membranes is likely to be considerable. Thus, a cascade transport system exists, including the essential membrane-associated GPI-anchored CD14 receptor [110–113]. We applied PMF calculations to derive energetic information for the stepwise processes, from the extraction of LPS from biologically realistic bacterial membrane models, LPS binding to CD14 and to the lipid-binding coreceptor of TLR4, MD-2 [114, 115], and finally LPS/MD-2 transfer to the terminal TLR4 receptor. This technique enabled us to demonstrate that LPS follows a thermodynamic funnel, leading to a favorable net change in

22

Roland G. Huber et al.

free energy upon traversing the complete pathway. Moreover, subsequent CG simulations enabled us to observe the formation of a hydrophobic bridge as part of a transient complex between CD14, MD-2, and TLR4 during LPS transfer [106], thus circumventing the energetic penalty of exposing the highly hydrophobic acyl tails to aqueous solvent.

4

Notes 1. Usually, biological systems should be modeled in the isothermal–isobaric ensemble, with maintenance of a constant number of particles, pressure, and temperature (NPT), as this reflects experimental and in vivo conditions. Whereas globular proteins can be modeled with isotropic pressure scaling, this is generally not appropriate for membrane systems. A semiisotropic pressure coupling that uses distinct barostats for the membrane plane and the normal direction may be employed to allow for independent changes in lipid area. Alternatively, simulations can be conducted in the canonical ensemble (NVT) at preset box dimensions or with a constant area in the membrane plane to constrain the system to a preset area per lipid, or alternatively, with an additional surface tension term. 2. Topologies with example coordinates for specific lipids and/or lipid bilayers can now be obtained from many websites, as exemplified by the following nonexhaustive list: http://wcm. ucalgary.ca/tieleman/downloads (Tieleman group); https:// lipidbook.bioch.ox.ac.uk (Sansom group); http://ter pconnect.umd.edu/~jbklauda/research/download.html (Klauda group); http://www.softsimu.net/downloads.shtml (Karttunen group); http://www.dsimb.inserm.fr/~luca/ downloads/ (Monticelli group); http://www.charmm-gui. org/?doc¼archive&lib¼lipid_pure (CHARMM-GUI site); http://mackerell.umaryland.edu/charmm_ff.shtml (MacKerell group); https://biophys.uni-saarland.de/downloads.html (Hub group); http://cgmartini.nl/index.php/force-fieldparameters/ (Marrink group); and https://atb.uq.edu.au/ index.py?tab¼existing_molecules (ATB and Repository). 3. Parameters for nonstandard (i.e., nonprotein, nonlipid, non-nucleic acid) molecules, such as for drug-like molecules, need separate topology files describing their connectivity, bond, angle, and dihedral parameters and charges. These can be constructed by combining preexisting topologies of smaller fragments, generated manually, or generated in an automated fashion using one of several web servers or local tools such as “antechamber” [116]. Popular servers include: ParamChem (www.paramchem.org); SwissParam (http://www.swissparam.

Multiscale Modeling of Lipid–Protein Interactions

23

ch), the GlycoBioChem PRODRG2 Server (http://davapc1. bioch.dundee.ac.uk/cgi-bin/prodrg); the Automated Topology Builder (ATB) and Repository (https://atb.uq.edu.au); and the CHARMM General Force Field (CGenFF) program (https://cgenff.paramchem.org). Parameters obtained from an automated tool always need to be carefully examined to ensure that connectivity, atom types, and bond orders have been properly applied, and where possible, tested for reproduction of accurate experimentally validated properties during simulations. 4. Common tools for membrane building include: cellPACK—A specialized version of the autoPACK tool for large-scale packing of biological macromolecules into arbitrary geometries, also supports packing of lipids [117] (autopack.org); CmME—CELLmicrocosmos Membrane Editor is a Javabased tool to generate heterogeneous membranes based on lipid shape [118] (www.cellmicrocosmos.org); CHARMMGUI—A versatile web-based Graphical User Interface that supports building both atomistic (CHARMM) and CG (Martini) bilayers from a large list of supported lipid types [57, 83, 119] (www.charmm-gui.org); insane—“INSert membrANE” is a flexible python-based command-line tool for building Martini bilayers and inserting proteins in bilayers; it supports numerous lipids and uses extendable lipid templates that allow for easy additions of new lipid types [58] (cgmartini.nl/ index.php/insane); LipidBuilder—A webserver that can build bilayers of CHARMM lipids created from a library of head group and hydrocarbon tails building blocks [120] (lipidbuilder.epfl.ch); LipidWrapper—A tool to curve bilayers into an arbitrary shape [121] (nbcr.ucsd.edu/lipidwrapper); MemBuilder—A webserver to build heterogeneous atomistic lipid bilayers [122] (www.membuilder.org); MemGen—A force field in dependent webserver for setting up heterogenous bilayers based on uploaded lipid structures [123] (memgen. uni-goettingen.de); PACKMOL—A program to generate initial coordinates for molecular dynamics simulations, including bilayers [124] (m3g.iqm.unicamp.br/packmol); VMD—The Visual Molecular Dynamics, molecular modeling and visualization, program has a membrane plugin that can replicate and trim preequilibrated bilayer patches to the desired size [66] (www.ks.uiuc.edu/Research/vmd). 5. It may be necessary to “repair” parts of an experimentally solved structure, for example, in the case of dynamic and hence unresolved loops. For small modifications such as missing amino acid side chain atoms, the simulation package itself will likely be sufficient. For more complicated problems, homology modeling approaches may be necessary. Homology

24

Roland G. Huber et al.

modeling enables the derivation of unknown “target” protein structures from homologous “template” proteins of known structure. This is based on the observation that structures tend to be more conserved than sequences. Useful structural homologues typically have sequence identities of at least ~30% to generate models with confidence. Known structural elements (domains, TM helices, etc.) can be incorporated as constraints, which improves reliability. Conserved features, such as inverted topological repeats, may be used to model alternative conformational states of membrane proteins [125]. Homology modeling is usually performed by advanced stand-alone programs such as Modeller [53]. Additionally, online servers may be used for homology modeling (e.g., SWISS-PROT (http:// swissmodel.expasy.org)) and/or for ab initio protein structure prediction (e.g., Rosetta (https://www.rosettacommons.org/ software/servers), I-TASSER (https://zhanglab.ccmb.med. umich.edu/I-TASSER/)). 6. Solvation of simulation systems is usually performed by superimposing preequilibrated solvent boxes with other system components and deleting overlapping water molecules. Additionally, water may be manually removed from the inner hydrophobic core of lipid bilayer models, or a restrained equilibration could be performed to ensure any water molecules spontaneously exit the membrane. Usually, a physiological salt concentration of 0.1–0.15 M NaCl is established by replacing water molecules with the respective ions, either via random replacement or based on the electrostatic potential of the system. These steps have generally been at least semi-automated in most modern simulation software packages. 7. In contrast to bonded and van der Waals interactions, electrostatic interactions do not become negligible with distance. While the Coulomb potential decays with 1/r, the number of interaction partners increases with r3, and hence applying a cutoff for electrostatic interactions is not appropriate. Two approaches are generally used to deal with this effect: at distances larger than a specific cutoff (usually ~1.5 nm), electrostatics are modeled by a response from a uniform background dielectric, a so-called reaction field. Alternatively, in periodic boundary conditions, Ewald summation may be used to calculate exact electrostatic interactions, usually in the form of FFT-based Particle Mesh Ewald (PME) [126]. This can however introduce periodicity artifacts into the system [127]. In the absence of additional information, the choice of treatment or electrostatics should be guided by the current standards of the force field.

Multiscale Modeling of Lipid–Protein Interactions

25

Acknowledgments T.S.C. and H.I.I. acknowledge that this work has been supported in part by the Joint Design of Advanced Computing Solutions for Cancer (JDACS4C) program established by the U.S. Department of Energy (DOE) and the National Cancer Institute (NCI) of the National Institutes of Health. T.S.C. and H.I.I. note that this work was performed under the auspices of the U.S. Department of Energy by Lawrence Livermore National Laboratory under Contract DE-AC52-07NA27344. LLNL-JRNL-752805. J.R.A. acknowledges the following sources of funding: Rutherford Discovery Fellowship (15-MAU-001); Marsden grant (15-UOA-105); New Zealand Ministry of Business, Innovation and Employment (MBIE) Endeavour Smart Ideas grant (UOCX1706); Maurice Wilkins Centre for Molecular Biodiscovery Flagship Project grant (MWC 3716850). W.A.I. was supported by the following sources of funding: Massey University Doctoral Scholarship; Massey University Doctoral Dissemination grant. N.D. thanks the Nehru trust for Cambridge University and Rajiv Gandhi (UK) foundation for funding. P.J.B. and J.K.M. acknowledge funding from the Ministry of Education in Singapore (MOE AcRF Tier 3 Grant Number MOE2012-T3-1008). References 1. Bernlohr DA, Simpson MA, Hertzel AV, Banaszak LJ (1997) Intracellular lipidbinding proteins and their genes. Annu Rev Nutr 17:277–303 2. De Libero G, Mori L (2005) Recognition of lipid antigens by T cells. Nat Rev Immunol 5:485–496 3. Russ AP, Lampel S (2005) The druggable genome: an update. Drug Discov Today 10:1607–1610 4. Overington JP, Al-Lazikani B, Hopkins AL (2006) How many drug targets are there? Nat Rev Drug Discov 5:993–996 5. Arora A, Tamm LK (2001) Biophysical approaches to membrane protein structure determination. Curr Opin Struct Biol 11:540–547 6. Phillips R, Ursell T, Wiggins P, Sens P (2009) Emerging roles for lipids in shaping membrane-protein function. Nature 459:379–385 7. Karplus M, McCammon JA (2002) Molecular dynamics simulations of biomolecules. Nat Struct Biol 9:646–652

8. Karplus M, Kuriyan J (2005) Molecular dynamics and protein function. Proc Natl Acad Sci U S A 102:6679–6685 9. Durrant JD, McCammon JA (2011) Molecular dynamics simulations and drug discovery. BMC Biol 9:71 10. Ash WL, Zlomislic MR, Oloo EO, Tieleman DP (2004) Computer simulations of membrane proteins. Biochim Biophys Acta 1666:158–189 11. Domene C, Bond PJ, Sansom MSP (2003) Membrane protein simulations: ion channels and bacterial outer membrane proteins. In: Protein simulations. Elsevier, Amsterdam, pp 159–193 12. Bond PJ, Sansom MSP (2006) Insertion and assembly of membrane proteins via simulation. J Am Chem Soc 128:2697–2704 13. Sansom MSP, Bond P, Beckstein O et al (2008) Water in ion channels and poressimulation studies. In: Ion channels: from atomic resolution physiology to functional genomics. John Wiley & Sons, Ltd, New York, NY, pp 66–83

26

Roland G. Huber et al.

14. Huber RG, Marzinek JK, Holdbrook DA, Bond PJ (2017) Multiscale molecular dynamics simulation approaches to the structure and dynamics of viruses. Prog Biophys Mol Biol 128:121–132 15. Angelescu DG, Linse P (2008) Viruses as supramolecular self-assemblies: modelling of capsid formation and genome packaging. Soft Matter 4:1981 16. Reddy T, Sansom MSP (2016) Computational virology: from the inside out. Biochim Biophys Acta 1858:1610–1618 17. Brooks BR, Brooks CL, Mackerell AD et al (2009) CHARMM: the biomolecular simulation program. J Comput Chem 30:1545–1614 18. Huang J, Mackerell AD (2013) CHARMM36 all-atom additive protein force field: validation based on comparison to NMR data. J Comput Chem 34:2135–2145 19. Case DA, Cheatham TE, Darden T et al (2005) The Amber biomolecular simulation programs. J Comput Chem 26:1668–1688 20. Salomon-Ferrer R, Go¨tz AW, Poole D et al (2013) Routine microsecond molecular dynamics simulations with AMBER on GPUs. 2. Explicit solvent particle mesh ewald. J Chem Theory Comput 9:3878–3888 21. Schmid N, Eichenberger AP, Choutko A et al (2011) Definition and testing of the GROMOS force-field versions 54A7 and 54B7. Eur Biophys J 40:843 22. J€ambeck JPM, Lyubartsev AP (2012) Derivation and systematic validation of a refined all-atom force field for phosphatidylcholine lipids. J Phys Chem B 116:3164–3179 23. J€ambeck JPM, Lyubartsev AP (2012) An extension and further validation of an all-atomistic force field for biological membranes. J Chem Theory Comput 8:2938–2948 24. J€ambeck JPM, Lyubartsev AP (2012) Another piece of the membrane puzzle: extending slipids further. J Chem Theory Comput 9:774–784 25. Ermilova I, Lyubartsev AP (2016) Extension of the slipids force field to polyunsaturated lipids. J Phys Chem B 120:12826–12842 26. Dickson CJ, Madej BD, Skjevik A˚A et al (2014) Lipid14: the amber lipid force field. J Chem Theory Comput 10:865–879 27. Schuler LD, Daura X, van Gunsteren WF (2001) An improved GROMOS96 force field for aliphatic hydrocarbons in the condensed phase. J Comput Chem 22:1205–1218

28. Berger O, Edholm O, J€ahnig F (1997) Molecular dynamics simulations of a fluid bilayer of dipalmitoylphosphatidylcholine at full hydration, constant pressure, and constant temperature. Biophys J 72:2002–2013 29. Marrink SJ, Risselada HJ, Yefimov S et al (2007) The MARTINI force field: coarse grained model for biomolecular simulations. J Phys Chem B 111:7812–7824 30. Monticelli L, Kandasamy SK, Periole X et al (2008) The MARTINI coarse-grained force field: extension to proteins. J Chem Theory Comput 4:819–834 31. De Jong DH, Singh G, Bennett WFD et al (2013) Improved parameters for the martini coarse-grained protein force field. J Chem Theory Comput 9:687–697 32. Uusitalo JJ, Ingo´lfsson HI, Akhshi P et al (2015) Martini coarse-grained force field: extension to DNA. J Chem Theory Comput 11:3932–3945 33. Nielsen SO, Lopez CF, Srinivas G, Klein ML (2004) Coarse grain models and the computer simulation of soft materials. J Phys Condens Matter 16:R481–R512 34. Saunders MG, Voth GA (2013) Coarsegraining methods for computational biology. Annu Rev Biophys 42:73–93 35. Zhang Z, Pfaendtner J, Grafmu¨ller A, Voth GA (2009) Defining coarse-grained representations of large biomolecules and biomolecular complexes from elastic network models. Biophys J 97:2327–2337 36. Noid WG, Chu JW, Ayton GS et al (2008) The multiscale coarse-graining method. I. A rigorous bridge between atomistic and coarsegrained models. J Chem Phys 128:244114 37. Noid WG, Liu P, Wang Y et al (2008) The multiscale coarse-graining method. II. Numerical implementation for coarsegrained molecular models. J Chem Phys 128:244115 38. Izvekov S, Voth GA (2005) A multiscale coarse-graining method for biomolecular systems. J Phys Chem B 109:2469–2473 39. Go¨tz AW, Williamson MJ, Xu D et al (2012) Routine microsecond molecular dynamics simulations with AMBER on GPUs. 1. generalized born. J Chem Theory Comput 8:1542–1555 40. Shaw DE, Bowers KJ, Chow E et al (2009) Millisecond-scale molecular dynamics simulations on Anton. In: Proc Conf High Perform Comput Netw Storage Anal SC 09 1. ACM, New York, NY 41. Shaw DE, Grossman JP, Bank JA et al (2014) Anton 2: raising the bar for performance and

Multiscale Modeling of Lipid–Protein Interactions programmability in a special-purpose molecular dynamics supercomputer. In: Proceedings of the International Conference for High Performance Computing, Networking. Storage and Analysis. IEEE Press, Piscataway, NJ, USA, pp 41–53 42. Marzinek JK, Holdbrook DA, Huber RG et al (2016) Pushing the envelope: dengue viral membrane coaxed into shape by molecular simulations. Structure 24:1410–1420 43. Petrlova J, Hansen FC, van der Plas MJA et al (2017) Aggregation of thrombin-derived C-terminal fragments as a previously undisclosed host defense mechanism. Proc Natl Acad Sci U S A 114:E4213–E4222 44. Berman HM, Westbrook J, Feng Z et al (2000) The protein data bank. Nucleic Acids Res 28:235–242 45. Kleywegt GJ, Jones TA (1998) Databases in protein crystallography. Acta Crystallogr Sect D Biol Crystallogr 54:1119–1131 46. Wuthrich K (1986) NMR of proteins and nucleic acids. Wiley, New York, NY 47. Cheng Y (2015) Single-particle Cryo-EM at crystallographic resolution. Cell 161:450–457 48. Wlodawer A, Minor W, Dauter Z, Jaskolski M (2008) Protein crystallography for non‐crystallographers, or how to get the best (but not more) from published macromolecular structures. FEBS J 275:1–21 49. Brown EN, Ramaswamy S (2007) Quality of protein crystal structures. Acta Crystallogr Sect D Biol Crystallogr 63:941–950 50. Hryc CF, Chen DH, Chiu W (2011) Nearatomic resolution cryo-EM for molecular virology. Curr Opin Virol 1:110–117 51. Carpenter EP, Beis K, Cameron AD, Iwata S (2008) Overcoming the challenges of membrane protein crystallography. Curr Opin Struct Biol 18:581–586 52. Fiser A, Sˇali A (2003) Modeller: generation and refinement of homology-based protein structure models. In: Methods in enzymology. Elsevier, Amsterdam, pp 461–491 53. Sali A (2008) MODELLER a program for protein structure modeling release 9v4, r6262. Structure:779–815 54. Shen M, Sali A (2006) Statistical potential for assessment and prediction of protein structures. Protein Sci 15:2507–2524 55. Wolf MG, Hoefling M, Aponte-Santamarı´a C et al (2010) g_membed: efficient insertion of a membrane protein into an equilibrated lipid bilayer with minimal perturbation. J Comput Chem 31:2169–2174

27

56. Jo S, Lim JB, Klauda JB, Im W (2009) CHARMM-GUI membrane builder for mixed bilayers and its application to yeast membranes. Biophys J 97:50–58 57. Qi Y, Ingo´lfsson HI, Cheng X et al (2015) CHARMM-GUI martini maker for coarsegrained simulations with the martini force field. J Chem Theory Comput 11:4486–4494 58. Wassenaar TA, Ingo´lfsson HI, Bo¨ckmann RA et al (2015) Computational lipidomics with insane: a versatile tool for generating custom membranes for molecular simulations. J Chem Theory Comput 11:2144–2155 59. Chang R, Ayton GS, Voth GA (2005) Multiscale coupling of mesoscopic- and atomisticlevel lipid bilayer simulations. J Chem Phys 122:244716 60. Abraham MJ, Murtola T, Schulz R et al (2015) GROMACS: high performance molecular simulations through multi-level parallelism from laptops to supercomputers. SoftwareX 1–2:19–25 61. Berendsen HJC, van der Spoel D, van Drunen R (1995) GROMACS: a message-passing parallel molecular dynamics implementation. Comput Phys Commun 91:43–56 62. Van Der Spoel D, Lindahl E, Hess B et al (2005) GROMACS: fast, flexible, and free. J Comput Chem 26:1701–1718 63. Phillips JC, Braun R, Wang W et al (2005) Scalable molecular dynamics with NAMD. J Comput Chem 26:1781–1802 64. Sanbonmatsu KY, Tung CS (2007) High performance computing in biology: multimillion atom simulations of nanoscale systems. J Struct Biol 157:470–480 65. Go¨tz AW, Williamson MJ, Xu D et al (2012) Routine microsecond molecular dynamics simulations with amber - Part I: Generalized born. J Chem Theory Comput 8:1542–1555 66. Humphrey W, Dalke A, Schulten K (1996) VMD: visual molecular dynamics. J Mol Graph 14(27-28):33–38 67. Ulmschneider MB, Sansom MSP (2001) Amino acid distributions in integral membrane protein structures. Biochim Biophys Acta 1512:1–14 68. Lomize MA, Lomize AL, Pogozheva ID, Mosberg HI (2006) OPM: orientations of proteins in membranes database. Bioinformatics 22:623–625 69. Yesylevskyy SO (2007) ProtSqueeze: simple and effective automated tool for setting up membrane protein simulations. J Chem Inf Model 47:1986–1994 70. Faraldo-Go´mez J, Smith G, Sansom M (2002) Setting up and optimization of

28

Roland G. Huber et al.

membrane protein simulations. Eur Biophys J 31:217–227 71. Schmidt TH, Kandt C (2012) LAMBADA and InflateGRO2: efficient membrane alignment and insertion of membrane proteins for molecular dynamics simulations. J Chem Inf Model 52:2657–2669 72. Shirts MR, Mobley DL, Chodera JD (2007) Chapter 4 Alchemical free energy calculations: ready for prime time? Annu Rep Comput Chem 3:41–59 73. Xu C, Gagnon E, Call ME et al (2008) Regulation of T cell receptor activation by dynamic membrane binding of the CD3ε cytoplasmic tyrosine-based motif. Cell 135:702–713 74. Shirts MR, Pande VS (2005) Solvation free energies of amino acid side chain analogs for common molecular mechanics water models. J Chem Phys 122:134508 75. Jefferys E, Sands ZA, Shi J et al (2015) Alchembed: a computational method for incorporating multiple proteins into complex lipid geometries. J Chem Theory Comput 11:2743–2754 76. Bond PJ, Cuthbertson JM, Deol SS, Sansom MSP (2004) MD simulations of spontaneous membrane protein/detergent micelle formation. J Am Chem Soc 126:15948–15949 77. Scott KA, Bond PJ, Ivetac A et al (2008) Coarse-grained MD simulations of membrane protein-bilayer self-assembly. Structure 16:621–630 78. Saravanan R, Holdbrook DA, Petrlova J et al (2018) Structural basis for endotoxin neutralization and anti-inflammatory activity of thrombin-derived C-terminal peptides. Nat Commun 9:2762 79. Stansfeld PJ, Hopkinson R, Ashcroft FM, Sansom MSP (2009) PIP2-binding site in kir channels: definition by multiscale biomolecular simulations. Biochemistry 48:10926–10933 80. Wassenaar TA, Pluhackova K, Bockmann RA et al (2014) Going backward: a flexible geometric approach to reverse transformation from coarse grained to atomistic models. J Chem Theory Comput 10:676–690 81. Kargas V, Marzinek JK, Holdbrook DA et al (2017) A polar SxxS motif drives assembly of the transmembrane domains of Toll-like receptor 4. Biochim Biophys Acta 1859:2086–2095 82. Irvine WA, Flanagan JU, Allison JR (2018) Computational prediction of amino acids governing protein-membrane interaction for the PIP3 cell signalling system. Structure 27:371

83. Wu EL, Cheng X, Jo S et al (2014) CHARMM-GUI membrane builder toward realistic biological membrane simulations. J Comput Chem 35:1997–2004 84. Cheng X, Jo S, Lee HS et al (2013) CHARMM-GUI micelle builder for pure/ mixed micelle and protein/micelle complex systems. J Chem Inf Model 53:2171–2180 85. Michaud‐Agrawal N, Denning EJ, Woolf TB, Beckstein O (2011) MDAnalysis: a toolkit for the analysis of molecular dynamics simulations. J Comput Chem 32:2319–2327 86. Huang CC, Couch GS, Pettersen EF, Ferrin TE (1996) Chimera: an extensible molecular modeling application constructed using standard components. In: Pac. Symp. Biocomput. World Scientific, Hackensack, NJ, p 724 87. Ingo´lfsson HI, Melo MN, van Eerden FJ et al (2014) Lipid organization of the plasma membrane. J Am Chem Soc 136:14554–14559 88. Reddy T, Sansom MSP (2016) The role of the membrane in the structure and biophysical robustness of the dengue virion envelope. Structure 24(3):375–382 89. Ingo´lfsson HI, Carpenter TS, Bhatia H et al (2017) Computational lipidomics of the neuronal plasma membrane. Biophys J 113:2271–2280 90. Koldsø H, Shorthouse D, He´lie J, Sansom MSP (2014) Lipid clustering correlates with membrane curvature as revealed by molecular simulations of complex lipid bilayers. PLoS Comput Biol 10:e1003911 91. Vollmer W, Blanot D, De Pedro MA (2008) Peptidoglycan structure and architecture. FEMS Microbiol Rev 32:149–167 92. Vollmer W, Seligman SJ (2010) Architecture of peptidoglycan: more data and more models. Trends Microbiol 18:59–66 93. Braun V (1975) Covalent lipoprotein from the outer membrane of Escherichia coli. Biochim Biophys Acta 415:335–377 94. Koebnik R (1995) Proposal for a peptidoglycan‐associating alpha‐helical motif in the C‐ terminal regions of some bacterial cell‐surface proteins. Mol Microbiol 16:1269–1270 95. Parsons LM, Lin F, Orban J (2006) Peptidoglycan recognition by Pal, an outer membrane lipoprotein. Biochemistry 45:2122–2128 96. Roujeinikova A (2008) Crystal structure of the cell wall anchor domain of MotB, a stator component of the bacterial flagellar motor: implications for peptidoglycan recognition. Proc Natl Acad Sci 105:10348–10353 97. Gumbart JC, Beeby M, Jensen GJ, Roux B (2014) Escherichia coli peptidoglycan

Multiscale Modeling of Lipid–Protein Interactions structure and mechanics as predicted by atomic-scale simulations. PLoS Comput Biol 10:e1003475 98. Samsudin F, Ortiz-Suarez ML, Piggot TJ et al (2016) OmpA: a flexible clamp for bacterial cell wall attachment. Structure 24:2227–2235 99. Samsudin F, Boags A, Piggot TJ, Khalid S (2017) Braun’s lipoprotein facilitates OmpA interaction with the escherichia coli cell wall. Biophys J 113:1496–1504 100. Cohen EJ, Ferreira JL, Ladinsky MS et al (2017) Nanoscale-length control of the flagellar driveshaft requires hitting the tethered outer membrane. Science 356:197–200 101. Zheng C, Yang L, Hoopmann MR et al (2011) Cross-linking measurements of in vivo protein complex topologies. Mol Cell Proteomics 10:M110-006841 102. Marcoux J, Politis A, Rinehart D et al (2014) Mass spectrometry defines the C-terminal dimerization domain and enables modeling of the structure of full-length OmpA. Structure 22:781–790 103. Doudou S, Burton NA, Henchman RH (2009) Standard free energy of binding from a one-dimensional potential of mean force. J Chem Theory Comput 5:909–918 104. Kumar S, Rosenberg JM, Bouzida D et al (1992) THE weighted histogram analysis method for free-energy calculations on biomolecules. I. The method. J Comput Chem 13:1011–1021 105. K€astner J (2011) Umbrella sampling. Wiley Interdiscip Rev Comput Mol Sci 1:932–942 106. Huber RG, Berglund NA, Kargas V et al (2018) A thermodynamic funnel drives bacterial lipopolysaccharide transfer in the TLR4 pathway. Structure 26:1151–1161 107. Lu Y-C, Yeh W-C, Ohashi PS (2008) LPS/TLR4 signal transduction pathway. Cytokine 42:145–151 108. Kim HM, Park BS, Kim J-I et al (2007) Crystal structure of the TLR4-MD-2 complex with bound endotoxin antagonist eritoran. Cell 130:906–917 109. Brandenburg K, Seydel U (1984) Physical aspects of structure and function of membranes made from lipopolysaccharides and free lipid A. Biochim Biophys Acta 775:225–238 110. Ryu J-K, Kim SJ, Rah S-H et al (2017) Reconstruction of LPS transfer cascade reveals structural determinants within LBP, CD14, and TLR4-MD2 for efficient LPS recognition and transfer. Immunity 46:38.

29

https://doi.org/10.1016/j.immuni.2016. 11.007 111. Juan TS-C, Kelley MJ, Johnson DA et al (1995) Soluble CD14 truncated at amino acid 152 binds lipopolysaccharide (LPS) and enables cellular response to LPS. J Biol Chem 270:1382–1387 112. Kelley SL, Lukk T, Nair SK, Tapping RI (2013) The crystal structure of human soluble CD14 reveals a bent solenoid with a hydrophobic amino-terminal pocket. J Immunol 190:1304–1311 113. Kim J-I, Lee CJ, Jin MS et al (2005) Crystal structure of CD14 and its implications for lipopolysaccharide signaling. J Biol Chem 280:11347–11351 114. Zimmer SM, Liu J, Clayton JL et al (2008) Paclitaxel binding to human and murine MD-2. J Biol Chem 283:27916–27926 115. Ohto U, Fukase K, Miyake K, Shimizu T (2012) Structural basis of species-specific endotoxin sensing by innate immune receptor TLR4/MD-2. Proc Natl Acad Sci U S A 109:7421–7426 116. Wang J, Wang W, Kollman PA, Case DA (2001) Antechamber, an accessory software package for molecular mechanical calculations. J Am Chem Soc 222:U403 117. Johnson GT, Autin L, Al-Alusi M et al (2014) cellPACK: a virtual mesoscope to model and visualize structural systems biology. Nat Methods 12:85–91 118. Sommer B, Dingersen T, Gamroth C et al (2011) CELLmicrocosmos 2.2 MembraneEditor: a modular interactive shape-based software approach to solve heterogeneous membrane packing problems. J Chem Inf Model 51:1165 119. Jo S, Kim T, Iyer VG, Im W (2008) CHARMM-GUI: a web-based graphical user interface for CHARMM. J Comput Chem 29:1859–1865 120. Bovigny C, Tamo` G, Lemmin T et al (2015) LipidBuilder: a framework to build realistic models for biological membranes. J Chem Inf Model 55:2491–2499 121. Durrant JD, Amaro RE (2014) LipidWrapper: an algorithm for generating large-scale membrane models of arbitrary geometry. PLoS Comput Biol 10:e1003720 122. Ghahremanpour MM, Arab SS, Aghazadeh SB et al (2013) MemBuilder: a web-based graphical interface to build heterogeneously mixed membrane bilayers for the GROMACS biomolecular simulation program. Bioinformatics 30:439–441

30

Roland G. Huber et al.

123. Knight CJ, Hub JS (2015) MemGen: a general web server for the setup of lipid membrane simulation systems: Fig. 1. Bioinformatics 31:2897–2899 124. Martinez L, Andrade R, Birgin EG, Martı´nez JM (2009) PACKMOL: a package for building initial configurations for molecular dynamics simulations. J Comput Chem 30:2157–2164 125. Vergara-Jaque A, Fenollar-Ferrer C, Kaufmann D, Forrest LR (2015) Repeatswap homology modeling of secondary active

transporters: updated protocol and prediction of elevator-type mechanisms. Front Pharmacol 6:183 126. Darden T, York D, Pedersen L (1993) Particle mesh Ewald: an Nlog(N) method for Ewald sums in large systems. J Chem Phys 98:10089–10092 127. Patra M, Karttunen M, Hyvo¨nen MT et al (2003) Molecular dynamics simulations of lipid bilayers: major artifacts due to truncating electrostatic interactions. Biophys J 84:3636–3645

Chapter 2 Quartz Crystal Microbalances as Tools for Probing Protein–Membrane Interactions Søren B. Nielsen and Daniel E. Otzen Abstract Extensive studies on the spontaneous collapse of phospholipid vesicles into supported lipid bilayers (SLBs) have led to procedures which allow SLB formation on a wealth of substrates and lipid compositions. SLBs provide a widely accepted and versatile model system which mimics the natural cell membrane separating the extracellular and intracellular fluids of the living cell. The quartz crystal microbalance with dissipation monitoring (QCM-D) has been central in both the understanding of vesicle collapse into SLBs on various substrates but also in probing the kinetics and mechanisms of biomolecular interactions with SLBs in real time. We describe a robust procedure to form SLBs of zwitterionic and charged lipids on SiO2 sensor crystals which subsequently can be exploited to probe the interaction between proteins and peptides with the SLB. Key words Supported lipid bilayer, SiO2, Quartz crystal microbalance with dissipation (QCM-D), Interaction

1

Introduction A quartz crystal microbalance consists of a small circular piece of quartz (the sensor crystal), whose piezoelectric properties make it oscillate at a characteristic resonance frequency f in response to an oscillating electric field. The QCM-D technique simultaneously measures changes in the resonance frequency (Δf ) and the energy dissipation (ΔD). These values depend not only on the sensor crystal itself but also the molecules (e.g., water, ions, lipid bilayers, proteins, and peptides) trapped in the oscillation through contact with the crystal surface [1, 2]. Provided this material absorbs as a rigid film, Δf is directly proportional to the mass adsorbed to the sensor surface (the Sauerbrey relation). However, this relation breaks down if the film is not sufficiently rigid. Information about film rigidity is provided by ΔD. Consequently, the combination of Δf and ΔD (described in more detail in Subheading 3.4.2) provides detailed information about the general viscoelastic properties of

Jo¨rg H. Kleinschmidt (ed.), Lipid-Protein Interactions: Methods and Protocols, Methods in Molecular Biology, vol. 2003, https://doi.org/10.1007/978-1-4939-9512-7_2, © Springer Science+Business Media, LLC, part of Springer Nature 2019

31

32

Søren B. Nielsen and Daniel E. Otzen

material bound to the sensor surface, such as those associated with various biological processes. The application of QCM-D has been demonstrated in numerous studies of protein adsorption [1, 3–6], antibody reactions with antigens [7, 8], attachment and spreading of cells [9, 10], and highly hydrated protein films [11]. Another extensively studied yet not completely understood phenomenon is the formation of planar supported lipid bilayers (SLBs) on various substrate surfaces. SiO2, Si3N4 and mica supports are popular choices for SLB formation while recent progress has further allowed the formation of SLBs on TiO2 and Au substrates. QCM-D has been key in providing mechanistic insights leading to the identification and understanding of conditions which determine whether adsorbing vesicles (1) remain intact (e.g., oxidized Au), (2) form SLBs through direct collapse upon interaction with the substrate (e.g., cationic lipids binding on SiO2) or (3) form SLBs when adsorbed vesicles rupture and fuse at a critical surface coverage level [12–15]. The latter process depends on factors such as solvent ions (divalent cations bridge anionic lipids to SiO2 surfaces, though to different extents [16], while monovalent ions will have a more modest effect through charge shielding), lipid composition (charge repulsion of anionic lipids binding to the SiO2 surface), surface chemistry (e.g., charge and roughness), pH [12, 17], and flow rates [18]. Fusion of charged lipids can be promoted by charged mercaptopropionic acid monolayers [19]. Once SLBs are formed on the sensor surface (as confirmed by characteristic Δf and ΔD signals), it is possible to monitor interactions with biomolecules which interact with SLBs or interacts with specific phospholipids or functional/reactive groups intentionally added to the SLB studied such as maleimide-activated or biotindoped bilayers [20–22]. By coupling these functionalities to specific biomolecules, such functionalities can be exploited to probe interactions with, for example, collagen [23], hyaluronan [24], DNA or specific peptide sequences [21, 22]. SLBs thus provide an excellent model system for biological cell membranes which separate and regulate the transport of biomolecules between the exterior and interior environments in vivo. In this chapter, we present a robust protocol to explore interactions of different proteins with SLBs on SiO2. Examples of applications include protein interactions with specific lipid head groups (Ca2+-dependent interaction of annexin A5 with anionic phosphatidylserine head groups [25]), enzymatic hydrolysis of SLBs by phospholipase A2 [26, 27], binding and membrane perturbation by antimicrobial peptides on SLBs [28–34], the interaction of protein–lipid complexes with SLBs [35], binding of extracellular matrix components to lipid surfaces [36] and incorporation of neuronal adhesion proteins to promote cell adhesion and growth [37]. Nonspecific protein–SLB interactions are only observed to a

Quartz Crystal Microbalances as Tools for Probing Protein–Membrane Interactions

33

very small extent [38], making SPBs an attractive platform for experiments in which low nonspecific binding is essential and for further functionalization of SPBs such as biotinylation and glycosylation. In addition to the closed QCM-D chamber which is the main focus of this chapter, Q-sense provides open modules which allow direct pipetting of the protein to the sensor surface. Although open modules are more prone to temperature variation and drift due to sample evaporation or variations in bulk volume, they allow precise control of the amount of protein added and may (depending on the bulk volume) require less sample owing to decreased loss of unused protein within the flow system.

2

Materials

2.1 QCM-D System and Accessories

1. Q-sense E4, E1 or D300 QCM-D system from Q-sense AB (V€astra Fro¨lunda, Sweden). 2. 5 MHz silicon dioxide (SiO2) coated QCM-D crystal sensors (Cat# QSX 303) from Q-sense AB (V€astra Fro¨lunda, Sweden). 3. Q-sense Sensor Holder. 4. Ozone lamp (e.g., Bioforce Nanosciences UV/Ozone Procleaner). 5. 2% SDS in MilliQ H2O. 6. 20 mM Tris–HCl, 150 mM NaCl. 7. 1 M CaCl2 or MgCl2.

2.2

Lipid Bilayers

1. 1,2-Diooleoyl-sn-glycero-3-phosphocholine (DOPC), 1,2-diooleoyl-sn-glycero-3-phosphoserine (DOPS) and/or 1,2-diooleoyl-sn-glycero-3-phosphatidylglycerol (DOPG) from Avanti Polar Lipids (Alabaster, AL). 2. Sonicator probe or Avanti Polar Lipids MiniExtruder (incl. suitable supports and polycarbonate membranes in 30–100 nm range for vesicle sizing). 3. Liquid nitrogen or dry ice bath. 4. Water bath.

2.3

Software

1. Qsoft 401 v2.5.10 for Q-sense E1 and E4 QCM-D systems for data acquisition. 2. Qsoft v1.6.18 for Q-sense D300 QCM-D systems for data acquisition. 3. QTools v3.0.13 for data analysis. The software is supplied by the manufacturer and updates can be downloaded from the user area at www.q-sense.com (registration requires instrument serial number and user name).

34

3 3.1

Søren B. Nielsen and Daniel E. Otzen

Methods Preparations

3.1.1 Cleaning of SiO2 Coated Sensors

Maintaining clean QCM-D sensor crystals and QCM-D flow system is essential to obtain reliable results and we here describe a procedure for cleaning of SiO2 sensor crystals. The reader is referred to Notes 1–3 for guidelines and implications of (improper) system cleaning. 1. Place sensor crystals in sensor holder. 2. Immerse sensor crystal in 2% sodium dodecyl sulphate (SDS) for 30 min at room temperature to remove adsorbed proteins and lipids. Do not allow the sensor to dry after exposure to SDS until they have been washed in MilliQ H2O. 3. Rinse thoroughly with MilliQ H2O. 4. Dry under a stream of nitrogen gas. 5. Subject to UV/Ozone treatment for 10 min to oxidize residual organic contaminants. 6. Rinse thoroughly with MilliQ H2O. 7. Dry under a stream of nitrogen gas.

3.1.2 Vesicle Preparation

Vesicles can be prepared by several methods including but not limited to detergent removal, sonication and extrusion. Of these methods, extrusion of multilamellar vesicles through polycarbonate membranes with defined pore sizes as described in the text below is preferred in our laboratory. Refer to Note 4 for an alternative procedure to produce small unilamellar vesicles by sonication. 1. Dissolve individual lipids to 10 mg/mL in 1:1 chloroform: methanol in a clean glass vial. 2. Mix a total amount of 5–10 mg lipid to desired molar or weight ratio in a clean glass vial (1.5 mL). Up to 50% anionic lipids such as PS can be tolerated [14, 39]. A higher content of anionic lipids will increase electrostatic repulsion between vesicles and increase the activation barrier to collapse and vesicle fusion. 3. Remove solvent in a stream of nitrogen gas while slowly turning the glass vial to create a thin lipid film on the bottom and wall of the glass vial. 4. Continue flushing with nitrogen gas for at least 1 h or incubate lipid film overnight in vacuum desiccator to remove residual solvent. 5. Add 1 mL 10 mM Tris–HCl pH 7.4, 150 mM NaCl to a final lipid concentration of 5–10 mg/mL and allow the film to swell for at least 10 min at a temperature 5–10
C above the highest tm of the lipids employed (see Note 5). Gently vortex the

Quartz Crystal Microbalances as Tools for Probing Protein–Membrane Interactions

35

suspension to remove remaining lipid film from the glass vial walls. This procedure will create large multilamellar vesicles (LMVs). 6. Subject lipid suspension to 4–5 freeze-thaw cycles by successive transfer between liquid nitrogen (or dry ice bath) and a water bath (NB! Keep water bath temperature above tm). 7. Prepare large unilamellar vesicles (LUVs) by at least 11 passages through a 30–100 nm polycarbonate membrane mounted between support discs in the Avanti Polar Lipids MiniExtruder. LUVs will generally be slightly larger than the pore size employed. 8. Vesicles can be stored for up to 3–4 days at 4
C. Avoid freezing after extrusion. 3.1.3 Sensor Mounting and Equilibration

1. After thorough cleaning of the sensor crystals, the dry crystal is mounted in the QCM-D sensor chamber. 2. Flow through at least 1 mL buffer at 250 μL/min. To avoid buffer dependent changes in F and D in subsequent steps, this buffer should be the same as that intended for use in protein–SLB interaction study. 3. Start the Qsoft program and set to desired temperature (this temperature should be above ambient). The QCM-D apparatus does not allow cooling or heating above 60
C unless an external heating/cooling source is attached. Allow temperature equilibration for 5–15 min (depending on temperature setpoint); the higher above ambient, the longer equilibration time is necessary to avoid drift in ΔF and ΔD due to temperature drift. 4. Start measurement by sweeping frequencies to locate resonance peaks at multiple harmonics (e.g., first to ninth harmonic). Recording data at multiple harmonics allow more detailed data analysis as described in (see Notes 6–8). 5. Obtain at least 5 min of stable (preferably 0.1  106 Hz1) which require compensation for viscous losses as described in following sections. The ΔD/Δf value of a proper SLB is typically in the order of 0.01 Hz1 or less [35, 40]. Δf and ΔD data for the third to seventh harmonic of a typical SLB formation on SiO2 surface are shown in Fig. 2. After obtaining a stable baseline of at least 5 min, 0.1 mg/mL SUVs made of POPC were injected, leading to a rapid decrease in Δf and an increase in ΔD at all harmonics, indicative of the adsorption of intact vesicles. Around 8 min of incubation time, a critical surface coverage is reached and vesicles begin to rupture onto the SiO2 surface as shown by the maxima in Δf and ΔD. After complete vesicle rupture, the Δf and ΔD values stabilize around 25 Hz and ~0.1  106, respectively. Notice that the Sauerbrey relation is sufficient to describe the final SLB but not valid during vesicle adsorption and rupture. Under these conditions, ΔD is high and individual harmonics spread out as an indication of a viscoelastic film. This requires modeling of the viscoelastic properties for accurate mass estimates.

Quartz Crystal Microbalances as Tools for Probing Protein–Membrane Interactions

41

The difference between harmonics arise due from two effects: (1) decreasing penetration depths (δn) with increasing frequency (harmonic number, n) and (2) films which do not couple fully to the oscillation of the sensor crystal. The penetration depth (δn) at which the amplitude of the shear wave has decreased by a factor e at a given overtone can be estimated by: rffiffiffiffiffiffiffiffiffi η δn ¼ ρf n π where ρ is the density of the layer and η is the shear viscosity. Thus, the penetration depth is inversely proportional to the overtone number and equals ~137 nm for the third overtone and ~90 nm for the seventh overtone in water at 25
C (ρ ¼ 997 kg/m3, η ¼ 8.9  104 kg/ms). The Sauerbrey relation has been shown to provide valid estimate of film thicknesses of up to ~2δ [52] under the limitations previously described. SLBs are around ~4.5–5 nm thick, so a substantial region can be probed beyond the lipid surface. This difference in penetration depths can give rise to some interesting observations. The antimicrobial peptide aurein 1.2 binds to DMPC membranes, giving rise to positive Δf values (mass loss) for the third and fifth overtones but negative Δf values (mass gain) for the seventh and ninth overtones [31]. The higher overtones register binding closer to the membrane surface than the lower overtones. Accordingly, the data may be interpreted to indicate that peptide binding to the surface reduces the coupling between the membrane and the bulk solution [31]. 3.4.1 Protein Adsorption onto SLBs

SLBs are generally highly resistant toward nonspecific adsorption of proteins. No or weak interactions have been observed for HSA, hIgG, fibrinogen, hemoglobin, cytochrome c, and serum [38], and not many have been shown to adsorb as rigid adlayers. Specific binding is however demonstrated by annexin-A1, A5 and A6, which bind in a Ca2+ dependent fashion to phosphatidylserine lipid head groups and prothrombin has been used to probe the interleaflet distribution of DOPS lipids [39, 49, 50, 53]. Buzhynskyy et al. [49] prepared SLBs from 0.1 mg/mL SUVs consisting of 4:1 DOPC–DOPS in the presence of 2 mM Ca2+ to facilitate collapse of anionic lipids on SiO2 and monitored the binding of annexin-A5 monomer (Fig. 3a) and dimer (Fig. 3b) species at 20 μg/mL and 40 μg/mL protein, respectively (~570 nM). The use of SiO2 support has been shown to produce SLBs with an equal distribution of anionic DOPS between bilayer leaflets [39]. With Δf ¼ 16  1 Hz and low ΔD at 0.1  106  0.05  106 (ΔD/Δ f ~ 0.006  106 Hz1), the adsorbed annexin-A5 monomer film is rigid and may be described by the Sauerbrey relation (Eq. 1), yielding an adsorbed mass of approximately 280  18 ng/cm2

42

Søren B. Nielsen and Daniel E. Otzen 4:1 DOPC:DOPS SUVs

(a)

0

EGTA

ΔF (Hz)

- 20

6

SUVs 4

- 40

ΔD X106

AnxA5

2 - 60 0 0

10

20 Time (min)

30

AnxA5 dimer

(b)

0

13

SUVs

- 25

5

- 125

ΔD X106

ΔF (Hz)

9 - 75

1

- 175 0

50 Time (min)

100

Fig. 3 Interaction of AnxA5 and AnxA5 dimers with 4:1 DOPC:DOPS SLBs on SiO2 support. (a) The characteristic overshoot of ΔF (black) and ΔD (orange) during SUV collapse into SLBs is followed by the injection of 20 μg/mL Annexin-A5 (AnxA5) monomer, leading to adsorption of 280  18 ng/cm2 AnxA5. This layer of adsorbed monomeric protein is resistant toward nonspecific interaction with SUVs. However, rinsing with EGTA eliminates the Ca2+-dependent AnxA5:SLB interaction. (b) The profile starts with preformed SLB, after which 40 μg/mL AnxA5 dimer is injected, leading to adsorption of ~670  18 ng/cm2 AnxA5 dimer. The dimer allows subsequent binding of SUVs. The models illustrate the proposed adsorption of AnxA5 monomer and dimers and the specific association of 4:1 DOPC:DOPS SUVs to AnxA5 dimers. Modified with permission from [49]. Copyright 2009 Elsevier Press

[49]. Adsorption of the dimer yielded larger Δf and ΔD values at 38  1 Hz and 1  106  0.1  106 (ΔD/Δf ~ 0.026  106 Hz1), respectively. This corresponds to an adsorbed dimer mass of ~670  18 ng/cm2 which is more than twice that of the monomer and the excess mass may be ascribed to the fact that QCM-D in addition to the protein also measures coupled water trapped in the cavities between protein molecules as also emphasized by the larger ΔD. However, the film remains sufficiently rigid to obtain appropriate mass estimates by the Sauerbrey relation for comparison. After protein adsorption, the authors further demonstrate that SLB-bound Annexin A5 dimers but not monomers are

Quartz Crystal Microbalances as Tools for Probing Protein–Membrane Interactions

43

able to bind to injected SUVs and that Annexin A5 binding to the SLB is Ca2+ dependent and reversible upon injection of EGTA containing buffer [49]. Note that while the SLB-Annexin A5 dimer film is sufficiently rigid to be described by Eq. 1, the relation is not valid when intact SUVs adsorb onto Annexin A5 dimers (Fig. 3b). 3.4.2 Viscoelastic Films

For soft films, which are also formed in the initial stages of SLB formation, the Sauerbrey relation underestimates the mass, since the oscillation of the film is not fully coupled to the sensor surface, that is, the film “slips.” Further data analysis using Voigt-based viscoelastic models is in such cases necessary and requires use of the change in dissipation ΔD accompanying film formation [54]. As a guideline, a rigid nonslipping film shows (1) low dissipation (D ~ 0–0.2  106) and (2) a low increase in dissipation at increasing frequency shifts (low slope in frequency-dissipation plots as discussed later), generally in the order of ΔD/Δf < 1  108 Hz1. If the layer is sufficiently rigid (SLBs are rigid in the QCM-D regime), mass estimates from the Sauerbrey equation are generally in good agreement with viscoelastic models described in the following but do not provide estimates of film density, viscosity and shear modulus. The Voigt model is included in the Qtools software package provided by Q-sense and provides means to quantify not only the adsorbed mass but also film thickness, shear and viscous moduli if the effective film density is known a priori. The value for the effective density of the film lie between the density of proteins (~1.2–1.35 g/cm3) and water (~1.0 g/cm3), taking into account that coupled water is detected by the QCM-D [3, 55]. For SLBs, a density of 1.1 g/cm3 can be assumed whereas the density of a densely packed DNA film is 1.7 g/cm3 [55]. More precise estimates of the effective film density can be obtained using optical techniques such as SPR and OWLS in combination with QCM-D [3, 56]. However, a good estimate of the adsorbed mass can still be obtained even if good estimates of the film density is not known a priori since the mass estimate is independent of the choice of density (Δm ¼ density  thickness) within the biologically relevant range. However, estimates of viscosity and shear modulus are dependent on the choice of film density [55]. In practice QCM-D provides estimates of the adsorbed mass and coupled water (“wet mass”) whereas optical techniques such as SPR and OWLS provide estimates of the adsorbed protein alone (“dry mass”). In Voigt viscoelastic modeling, it is essential to verify that estimates of shear modulus and shear viscosities are within a biologically relevant range. The shear modulus of Mefp-1 protein films was previously reported to be 0.066 MPa, increasing to 0.3 MPa upon cross-linking [1]. Concurrently, estimates of shear viscosity

44

Søren B. Nielsen and Daniel E. Otzen

increased from 0.002 kg/m s to 0.006 kg/m s for the cross-linked Mefp-1 film [1]. Similarly, cross-linking of polylysine films by glutaraldehyde increased shear modulus from 0.25 to 2.5 MPa and shear viscosity from 0.003 to 0.0175 kg/m s [57]. BSA adsorption onto stainless steel and alumina led to shear elastic modulus of 0.5  0.1 MPa and 0.3  0.1 MPa, respectively and shear viscosity estimates of 0.0028  0.001 kg/m s and 0.0022  0.0003 kg/m s, respectively [58]. Lastly, β-lactoglobulin adsorption onto polyethersulfone films led to shear viscosity estimates of ~0.0023–0.01 kg/m s and shear modulus of ~0.1–0.92 MPa [59]. Thus, biologically relevant estimates of shear modulus and shear viscosity (for both protein and lipid films) should in the range of ~0.01–2.5 MPa and ~0.001–0.02 kg/m s, respectively. Very few examples of viscoelastic modeling applied to protein films on SLBs are available in the literature, presumably due to the complexity of data treatment and verification. However, the Δf and ΔD data during adsorption and cross-linking of Mefp-1 onto a hydrophobic QCM-D sensor surface in Fig. 4 serve to illustrate the relevant biological range of shear viscosity and shear modulus estimates, and can be used as a guideline for experiments in which protein adsorbs to SLBs. 3.4.3 Membrane Modulation/Disintegration

It is challenging to make a qualitative description of the interaction between SPBs and proteins/peptides from Δf and ΔD values alone. However, useful insight can often be provided by frequencydissipation plots (Δf–ΔD plots). These provide a “phase diagram” describing how the conformation or structure of proteins (and lipids) changes for a mass unit adsorbing to the film, eliminating time as an explicit parameter. Δf–ΔD plots consist of discrete points each representing a single point in time. Thus, a linear relation in the Δf–ΔD plot suggests a simple adsorption process in which added mass leads to a constant change in dissipation (i.e., film rigidity) per mass unit. In these plots, dissipation may either be constant or increase, depending on whether the adsorbed film is essentially rigid throughout or softens upon adsorption of a given mass, respectively. A larger increase in dissipation per mass unit suggests a more flexible or unfolded conformation of the adsorbed proteins, leading to a more hydrated film. In contrast, a low increase in dissipation per mass unit indicates the adsorption of a rigid film, which may involve a more compact protein state. Such information may shed more light on the mechanism of interaction, and can be exploited to examine how factors such as pH, temperature, lipid composition, salts, or mutations in the polypeptide chain affect protein adsorption [4, 7, 60, 61]. Further, deviations from linear relationships (not easily picked up from raw Δf and ΔD time plots) usually indicate one or more secondary processes occurring at the surface, and provide essential mechanistic information about the adsorption and/or

Quartz Crystal Microbalances as Tools for Probing Protein–Membrane Interactions

0 -10 -20 -30 -40 -50 -60 -70 -80

14

NalO4

Δfn=3/3 Δfn=5/5 Δf /3 (fit)

12 10

n=3

Δfn=5/5 (fit)

8

ΔDn=3 ΔDn=5 ΔDn=3 (fit) ΔD (fit) n=5

6

0

4

ΔD (10-6)

Δf (Hz)

(a)

45

2 20 40 time (min)

(b) 25

0 80

60 NalO4

dvoight(nm)

20 15 10 5 0 0

10

20

30

40 50 time (min)

60

70

80 6

NalO4

6

5

5

4

4

3

3

2

2

1

1 0 0

10

20

30

40 50 time (min)

60

70

80

0

shear modulus (x105 N/m2)

viscosity (x10-3 Ns/m2)

(c) 7

Fig. 4 (a) Raw f–D data for the adsorption and cross-linking (by NaIO4) of a Mefp1 film on an alkane terminated QCM-D sensor surface. NaIO4 is added around 56 min (arrows). Subsequently Voigt viscoelastic modeling is used to obtain estimates of (b) film thickness and (c) shear viscosity and shear modules. Reproduced with permission from [1]. Copyright 2001 American Chemical Society

46

Søren B. Nielsen and Daniel E. Otzen 14 1

12

0.8 3mM + 10mg/ml BSA

10 mM

0.6 0.4

Dissipation (x106)

10

0.2 0

0.3 mM

-25 -26 -27 -28

5 mM

8

6

320 mM OA

4

1 mM

3 mM

2

0 0

-10

-20

-30

-40

-50

-60

Resonance frequency [Hz] Fig. 5 Δf–ΔD plot of equine lysozyme–oleic acid complexes with oleic acid and 4:1 DOPC:DOPG SLBs. Reproduced with permission from [35]. Copyright 2010 Elsevier Press

film rearrangement process(es) [28, 35]. Deviations from linear relations may be observed as the protein concentration and/or lipid composition changes. We have found Δf–ΔD plots particularly helpful in determining the mechanism of interaction between equine lysozyme complexes with oleic acid (ELOA) and SLBs. As shown in Fig. 5, Δf–ΔD plots reveal the typical vesicle adsorption-rupture process resulting in SLBs with Δf around 25 Hz and ΔD < 0.2  106 [35]. On exposure to low concentrations of ELOA (7 M) for folding and insertion into a lipid membrane or membrane mimetic environment. For OmpA, which contains five Ws in the transmembrane βbarrel domain and two C in the periplasmic domain, it has been shown that its native Ws can be replaced by F without affecting its function [29]. After preparation of the site-specifically labeled βTMPs, a range of fluorescence methods including kinetic studies (Fig. 2) can be used to investigate local protein–lipid interactions, membrane adsorption and/or insertion, local protein folding [30] or local protein stability of β-TMPs [31–33], and also interactions of unfolded β-TMPs with molecular chaperones like Skp [34] (Fig. 2c upper panel). 1.4 Site-Directed Fluorescence Quenching

The range of methods using site-directed fluorescence spectroscopy to examine protein–lipid interactions can be extended by inclusion of an additional label (e.g., lipid-bound fluorescence quenchers in the membrane like spin-labeled lipids [35–37] or brominated lipids [38–40]). Labeled lipids can be used to examine specific interactions and protein or peptide topology in membranes (see also Chapter 15 in this book) or sequential events in membrane protein folding [30]. The combination of site-directed fluorescence spectroscopy with quenching methods using either sets of brominated or sets of spin-labeled lipids that carry the quencher at different distance from the bilayer center is a very powerful methodology in membrane protein folding [27], also to examine insertion of individual strands of β-TMPs [30]. The impact of lipid–protein interactions on intramolecular conformation changes in membrane proteins can also be studied by introducing a second probe into the same protein. For folding studies, single W, single C mutants were prepared and the C was labeled spectroscopically (e.g., either by a short-range fluorescence quencher or by a suitable fluorescence energy transfer acceptor) to study folding of membrane proteins [41, 42]. Intramolecular W-fluorescence quenching was determined from the ratio of the

470

Lisa Gerlach et al.

Fig. 2 (a) OmpA (1BXW) contains five W residues. With the exception of W102, these are exposed to the hydrophobic core of the lipid membrane. (b) Membrane adsorption of unfolded OmpA at 2
C, monitored by the increase of the fluorescence of the five single W mutants of OmpA. In each mutant, the four other Ws were replaced by F. W102, which is directed into the lumen of the β-barrel domain, shows the smallest fluorescence increase. Adapted from ref. 30 (c) Wavelength of the maximum of the fluorescence intensity of unfolded forms of single W mutants of OmpA in the absence and in the presence of the molecular chaperone Skp (upper panel) or lipopolysaccharide (lower panel). Whereas most positions in OmpA are affected by interactions with Skp, only few locations in OmpA show interactions with lipopolysaccharide (adapted from ref. 34)

fluorescence intensities in the presence and in the absence of a C-linked nitroxide spin-label. A lower ratio indicates closer proximity between the fluorophore W and the quencher linked to the C (Fig. 3). Intramolecular W fluorescence quenching by a neighboring spin-labeled C demonstrates that the β-TMP OmpA does not fold to any native-like tertiary structure in aqueous solution, but via several distinct folding intermediates in the presence of lipid bilayers [41], characterized by different proximities between the fluorescent Ws and the labeled Cs. The intermediates correlate with intermediates previously identified using lipid-bound fluorescence quenchers [27, 30], for a recent review on the folding of OmpA cf. [1]. Results obtained under various folding conditions indicated that membrane-bound folding intermediates with different proximities between W as a fluorophore and the nitroxide group at the C as a quencher can be trapped when folding is performed at temperatures between 2 and 40
C. Similarly, single W, single C mutants of β-TMPs may be labeled with 5-((2-iodoacetamido) ethyl)-1-aminonapthalene sulfate (IAEDANS). Fo¨rster resonance energy transfer (FRET), see also Chapter 16 in this book, can then be used to determine distances and distance changes upon folding of β-TMPs [42].

Insertion of β-barrel membrane proteins

471

Fig. 3 Insertion and folding of the β-TMP OmpA into lipid membranes by intramolecular site-directed fluorescence quenching (ISFQ). (a) Membrane topology (lipid-oriented amino acid side chains indicated in yellow) and crystal structure (1BXW) of the transmembrane domain of OmpA. (b) Relative proximities of single W, single C mutants as deduced from intramolecular fluorescence quenching [41]. The single-W-singleC-mutants W57C35, W15C35, W15C162, W7C170, and W7C43 of OmpA were labeled with a nitroxide spin-label at the C. The relative fluorescence intensity, that is, the ratio of the fluorescence intensity in the presence to the intensity in the absence of the nitroxide-label was obtained 1 h after initiation of folding and is given in the Table below for folding reactions performed at temperatures between 2 and 40
C. The lower these ratios, the closer the W and the labeled C. Trapped folding intermediates correlate with intermediates identified previously using lipid bound quenchers, demonstrating β-barrel formation and insertion into the lipid bilayer are correlated, see ref. 1 for a detailed review

472

2

Lisa Gerlach et al.

Materials

2.1 Expression Systems

2.2 General Laboratory Equipment and Lab-Supplies

To perform the folding and membrane insertion experiments described here, a suitable expression system for the β-TMP of interest is required. β-TMPs have been overexpressed in various E. coli strains, e.g., E. coli MC4100 or E. coli BL21(DE3) [43]. Typically β-TMPs are isolated either after overexpression of their precursors containing the signal sequence for transport to the outer membrane [44–46] or after overexpression without signal sequence into cytoplasmic inclusion bodies (IBs), see for example refs. 8, 10, 12, 13, 47. Expression into the outer membrane leads to natively folded functional β-TMPs but often complicates purification as the native proteins must be extracted from the membrane. Membranes contain many components leading to difficulties in obtaining highly pure β-TMP or requiring additional purification steps, reducing the yields of β-TMPs after purification. On the other hand, yields are much higher when β-TMPs are overexpressed as IBs, that is, in an unfolded aggregated form. Therefore, this strategy requires solubilization of β-TMPs in unfolded form in concentrated solutions of chaotropic denaturants like urea and subsequent refolding/renaturation. Expression into inclusion bodies also simplifies the purification protocol. To date, most β-TMPs successfully refold in a suitable detergent or lipid environment when the denaturant is strongly diluted. As folding studies are the focus of this chapter, only expression in unfolded form into IBs is described here. E. coli BL21(DE3) is also a suitable expression system for β-TMPs from mitochondria of higher organisms like the voltage-dependent anion-selective channel, human isoform 1 (hVDAC1) [9, 48] or Tom40 [49]. Typically, pET vectors are used as expression vectors. We here describe expression and purification for OmpA and hVDAC1. For site-directed mutagenesis and cloning, readers are referred to well-established protocols described elsewhere (e.g., [50, 51]), see also Methods in Molecular Biology, volume 1498, In Vitro Mutagenesis. 1. Refrigerated tabletop centrifuge for 1.5, 30 and 50 mL tubes (Eppendorf 5810 R with rotors F-34-6-38 and FA-45-30-11 or similar). 2. UV/VIS Absorption spectrophotometer. 3. Quartz cuvettes for UV-Absorbance measurements. 4. Thermomixer (e.g., Eppendorf or similar). 5. Vortex Mixer. 6. Lab balance. 7. Precision balance accurate to 0.1 mg or better.

Insertion of β-barrel membrane proteins

473

8. Precision pipettors, 0.2–10 μL, 2–20 μL, 10–100 μL, 100–1000 μL, 5 mL (see Note 1). 9. Precision Hamilton syringes, 1–10 μL, 10–100 μL, calibrated. 10. Laboratory flasks, beakers, and tubes for buffers and reagents. 11. Dialysis membranes (e.g., Spectra/Por) or desalt spin columns (e.g., PD10, GE Healthcare). 12. Concentrators (Amicon Ultra 0.5 mL centrifugal filters, Merck or similar). 13. Gas tank with nitrogen or argon gas with regulator to allow for a steady gentle stream for both, degassing solutions and for drying 1–2 mL of lipid solutions in organic solvent. 2.3 Protein Expression and Purification 2.3.1 Basic Materials for the Expression of β-TMPs

In these protocols, 2 L (OmpA) or 6 L (hVDAC1) of cell culture are prepared to isolate 40–200 mg of β-TMP.

1. An E. coli strain like BL21(DE3) Omp8 [43] for the overexpression of a β-TMP, either as a stab culture or a glycerol stock, harboring an expression plasmid for the β-TMP and a gene for antibiotic resistance (e.g., against ampicillin). 2. Ampicillin (or other antibiotic, for which resistance is encoded on the expression vector). 3. Media: 2 or 6 L of autoclaved Miller LB medium: 10 g/L tryptone, 5 g/L yeast extract, 10 g/L NaCl, adjusted to pH 7.0 with NaOH), supplemented with 0.1 g/L ampicillin after autoclaving. 4. Isopropyl β-D-1-thiogalactopyranoside (IPTG) for induction of protein expression. 5. Baffled Erlenmeyer flasks, Erlenmeyer flasks or fermenter for cell culture. 6. Incubator with temperature control to shake cell cultures. 7. Laboratory centrifuge to harvest cells from 2 to 6 L of cell culture at 5000–7000  g. 8. Magnetic stirrer. 9. Ultrasonifier (Branson ultrasonifier W-450D or similar) with macro tip for cell lysis. 10. Fast protein liquid chromatography system (FPLC) for gradient-based elutions.

2.3.2 Protein Purification of OmpA

1. Tris resuspension buffer: 20 mM 2-amino-2-(hydroxymethyl) propane-1,3-diol (tris (hydroxy-methyl)aminomethane, Tris), pH 8.0.

474

Lisa Gerlach et al.

2. Tris solubilization buffer: 20 mM Tris, 0.1% β-mercaptoethanol, 8 M urea, pH 8.0 (see Note 2). 3. Tris wash buffer: 15 mM Tris, 0.1% β-mercaptoethanol, 8 M urea, pH 8.5. 4. Tris elution buffer: 15 mM Tris, 100 mM NaCl, 0.05% βmercaptoethanol, 8 M urea, pH 8.5. 5. Protease inhibitor tablet. 6. Lysozyme (chicken egg white lysozyme). 7. Q-Sepharose fast flow anion exchange column. 2.3.3 Protein Purification of hVDAC1

1. Tris buffer A: 25 mM 2-amino-2-(hydroxymethyl)propane1,3-diol (tris (hydroxy-methyl)aminomethane, Tris), pH 8.0), 150 mM NaCl, 0.2 mM PMSF. 2. Tris buffer B: 25 mM Tris, pH 8.0, 150 mM NaCl, 1% Triton X-100. 3. Tris buffer C: 100 mM Tris, 10 mM Dithiothreitol (DTT), 1 mM EDTA, 8 M urea, pH 8.0. 4. Tris buffer D: 50 mM Tris, pH 8.0, 8 M urea. 5. MES buffer A: 50 mM MES, pH 6.0, 8 M urea. 6. MES buffer B: 50 mM MES, pH 6.0, 500 mM NaCl, 8 M urea. 7. Protease inhibitor tablet. 8. Lysozyme (chicken egg white lysozyme). 9. DEAE-Sepharose Fast Flow anion exchange column. 10. CM-Sepharose Fast Flow cation exchange column.

2.4

Protein Labeling

1. Black Eppendorf tubes/reactions tubes or tubes wrapped in aluminum foil to shield samples from light. 2. Labeling buffer: 20 mM Tris, pH 7.2, 1 mM EDTA, 7 M urea (see Note 2). 3. Tris(2-carboxyethyl) phosphine hydrochloride (TCEP), dissolved just before use in distilled water at a fourfold molar excess over the β-TMP to be labeled. 4. Labeling reagent: Either fluorescence label or spin-label. Prepare freshly before use or use from a 100 mM stock solution, stored at 80
C in the dark. Always protect labeling reagents and labeling reactions from light exposure in a black reaction tubes (Eppendorf or similar). (a) As thiol-reactive fluorescence label, 5-((((2-Iodoacetyl) amino) ethyl) amino) naphthalene-1-sulfonic acid (1,5-IAEDANS) or similar reactive labels may be used. (b) As a fluorescence quencher, a spin-label can be used (e.g., (1-Oxyl-2,2,5,5-tetra-methyl pyrroline-3-methyl)

Insertion of β-barrel membrane proteins

475

methanethiosulfonate (MTSL)). Spin-labeling reagents are available from Santa Cruz Biotechnology, Toronto Research Chemicals or others, see also Chapter 21. 5. Fluorescence labeling reagents are dissolved in DMSO or DMF, MTSL spin-labeling reagent is dissolved in ethanol, acetonitrile, or DMSO. Use Hamilton syringes for volume measurements of organic solvents. Concentrations of the labeling reagent are typically chosen to add a small volume of the labeling reagent to the reactions mixture (about 1 mL) to result in a tenfold molar excess over the β-TMP. 2.5 Protein Methylation

1. Borate buffer: 50 mM, pH 9.0 (see Note 2), 1 mM EDTA, 7 M urea. 2. Tris(2-carboxyethyl) phosphine hydrochloride (TCEP), dissolved just before use in distilled water to a concentration of 12 mM. 3. N-Methyl-N-nitrosobenzamide (MNB), dissolved in acetonitrile, 100 mM.

2.6 Determination of Free Thiol Groups in Unlabeled Cysteines

1. Tris buffer: 10 mM Tris, 1 mM EDTA, pH 7.2.

2.7 Preparation of Lipid Bilayer Vesicles (Liposomes)

1. Synthetic phospholipid(s) of choice (for example 1,2-dilauroyl-sn-glycero-3-phosphocholine (diC12:0PC), 1,2-dilauroylsn-glycero-3-phosphoethanolamine (diC12:0PE), 1,2-dilauroyl-sn-glycero-3-phosphoglycerol (diC12:0PG), 1-palmitoyl, 2-oleoyl-sn-glycero-3-phosphocholine (C16:0C18:1PC), 1,2-dimyristoyl-sn-glycero-3-phosphocholine (diC14:0PC), 1,2-dioleoyl-sn-glycero-3-phosphocholine (diC18:1PC), etc. or extracted phospholipids from natural sources, either as a lyophilized powder or in organic solvent). Lipids can be purchased from companies like Avanti Polar Lipids (Alabaster, AL) and should be stored at 20
C or below. Lipids may include labeled lipids, such as 1-palmitoyl-2-(m,m+1-dibromo)stearoyl-sn-glycero-3-phosphocholine (m,m+1-DiBrPC) that are commercially available as 4,5-DiBrPC, 6,7-DiBrPC, 9,10-DiBrPC, or 11,12-DiBrPC (Avanti Polar Lipids, Alabaster, AL).

2. 5,50 -dithiobis-2-nitrobenzoic acid (DTNB, Ellman’s reagent).

2. Chloroform. 3. Methanol. 4. Buffer of choice (e.g., glycine buffer): 10 mM glycine, pH 8.0, 2 mM EDTA, citrate Buffer: 10 mM citrate, pH 3, 2 mM EDTA, HEPES buffer: 10 mM, pH 7.2, 2 mM EDTA, 10 mM NaCl.

476

Lisa Gerlach et al.

5. Liquid nitrogen for freeze-fracturing lipid dispersions in aqueous buffer. 6. Water bath for thawing frozen lipid samples. 7. Oil pump for the generation of high vacuum. 8. Vacuum desiccator. 9. Either for large unilamellar vesicles (LUVs) or small unilamellar vesicles (SUVs): (a) LUVs: Extruder or Mini-Extruder, available from Avestin (Ottawa, Canada), Avanti Polar Lipids (Alabaster, AL), T & T Scientific (Knoxville, TN), etc. and suitable polycarbonate membranes of defined pore size (e.g., 100 nm (Nucleopore, Whatman, Clifton, NJ)) for the preparation of LUVs. (b) SUVs: Ultrasonifier (Branson ultrasonifier W-450D or similar) with microtip for the preparation of SUVs and 5 mL pear-shaped flask. 2.8 Fluorescence Measurements

1. Calibrated spectrofluorometer providing research-grade signal/noise ratio in fluorescence data acquisition and software to average multiple scans (e.g., Horiba Jobin-Yvon Fluoromax/Fluorolog, ISS Chronos/K2/PC1, PTI Quantamaster, JASCO FP-8000 series, PerkinElmer LS55, Hitachi F-7000, Shimadzu RF 6000, OLIS, or others). 2. Temperature bath controlling the temperature of the sample compartment of the spectrofluorometer. 3. Quartz cuvettes for fluorimetry (e.g., 1 mL (10 mm excitation direction  4 mm emission direction) or 0.5 mL (10 mm excitation direction  2 mm emission direction)).

3

Methods

3.1 Protein Expression and Purification 3.1.1 Expression of β-TMPs in Form of Cytoplasmic Inclusion Bodies

1. For a preculture, cells from a single colony on an agar plate are picked and grown at 37
C in 40 mL of LB medium supplemented with ampicillin (or other antibiotic, for which resistance is encoded on the expression vector). 2. 2 L (OmpA) or 6 L (hVDAC1) of autoclaved LB Medium are supplemented with 0.1 g/L ampicillin and inoculated with 40 mL of the preculture. Cells are shaken in an incubator at 37
C until they have grown to an OD600 ~0.6 (OmpA) or to an OD600 ~0.9 (hVDAC1) (see Note 3). 3. A final concentration of 0.2–1 mM IPTG is added to the cell culture to induce expression of the β-TMP. After 4–6 h of incubation, bacteria are harvested by centrifugation at 6000  g for 15 min.

Insertion of β-barrel membrane proteins 3.1.2 Extraction and Purification of OmpA

477

1. The harvested wet cell paste is resuspended in a round flask with 40 mL of Tris suspension buffer and kept cool in an ice/water bath at 0
C. Add a protease inhibitor tablet. 2. A final concentration of 10–50 μg/mL of lysozyme is added and the resuspended paste is stirred on a magnetic stirrer for 30 min while cooling in the ice/water bath. 3. To lyse the cells, the resuspended paste is sonicated under cooling in an ice/water bath for 30 min using a sonifier equipped with a macrotip at 20% power and at a duty cycle of 0.5 s on/0.5 s off. 4. IBs containing OmpA and membrane fractions are separated from soluble proteins by centrifugation for 20 min at 8000  g and at 4
C. 5. IBs are solubilized in 20 mL Tris solubilization buffer. Insoluble membrane fractions are removed by 30 min of centrifugation at 3000  g and at 20
C. The clear supernatant contains unfolded OmpA. 6. OmpA (pI ~5.5) is purified by anion exchange chromatography. When air is present in the FPLC system, purge the system as described in the instructions of the manufacturer. 7. A Q-Sepharose Fast Flow column is connected to the FLPC system and first equilibrated with 3 column volumes (cv) of Tris wash buffer at a flow rate of 2.5–3 mL/min. 8. IBs with OmpA are loaded onto the column at a reduced flow rate 1 mL/min. 9. Nonspecifically bound material is removed by washing the column with 3 cv of Tris wash buffer at a flow rate of 3 mL/ min. The UV-absorbance of the flow-through is monitored at 280 nm and the flow-through is collected in 50 mL falcon tubes. 10. Purified OmpA is then eluted, monitoring the UV-absorbance at 280 nm. Tris elution buffer is used for a linear elution gradient from 0 to 100 mM NaCl over ~10 cv. All fractions including those showing absorption at 280 nm are collected in glass tubes of 5 mL volume. 11. Eluted fractions showing absorbance at 280 nm are analyzed for OmpA content by SDS PAGE [52, 53]. 12. The tubes containing purified OmpA in unfolded solubilized form are pooled and when necessary dialyzed against Tris solubilization buffer to remove NaCl. 13. Pooled fractions containing purified OmpA are concentrated to obtain high concentrations of unfolded OmpA in Tris buffer containing 8 M urea. 14. The concentration of OmpA is carefully determined, for example, using the method by Lowry [54].

478

Lisa Gerlach et al.

3.1.3 Extraction and Purification of hVDAC1

1. The harvested wet cell paste is resuspended in ~40 mL of Tris buffer A and cooled in an ice/water bath. 2. A final concentration of 10–50 μg/mL of lysozyme is added and the resuspended paste is stirred on a magnetic stirrer for 30 min while cooling in the ice/water bath. 3. To lyse the cells, the resuspended paste is sonicated under cooling in an ice/water bath for 15–40 min using a sonifier equipped with a macrotip at 20% power and at a duty cycle of 0.5 s on/0.5 s off. 4. IBs containing hVDAC1 and membrane fractions are separated by 30 min of ultracentrifugation at 51,500  g and at 4
C. 5. IBs are homogenized with Tris buffer A and centrifuged for 30 min at 51,500  g and at 4
C. 6. The pellet is washed with Tris buffer B. 7. The suspension is centrifuged at 33,000  g and at 4
C for 30 min. 8. IBs are resuspended and solubilized in Tris buffer C. 9. hVDAC1 is prepurified by a DEAE-Sepharose Fast Flow anion exchange column. hVDAC1 has a pI of ~8.6. In buffer D, hVDAC1 is only slightly positively charged and does not bind as well to DEAE-Sepharose as the protein impurities from E. coli. Thus, hVDAC1 can be collected in the flow-through. The column is attached to the FPLC system and first equilibrated with 4 cv of Tris buffer D at a flow rate of 2.5 mL/min. 10. Solubilized inclusion bodies of hVDAC1 are loaded onto the DEAE Sepharose column and collected in the flow-through. 11. hVDAC1 is first dialyzed overnight against MES buffer A. In this buffer, hVDAC1 is positively charged. 12. A CM-Sepharose Fast Flow cation exchange column is equilibrated with 4 cv of MES buffer, pH 6.0, at a flow rate of 2.5 mL/min. 13. hVDAC1 is loaded onto the CM-Sepharose at a flow rate of 0.5 mL/min. 14. The column is washed with 4 cv of MES buffer A to remove unspecifically bound material. 15. Monitoring the UV absorbance at 280 nm, hVDAC1 is eluted with MES buffer B by applying a linear gradient of 0–500 mM NaCl at a flow rate of 2.5 mL/min over ~10 cv. Fractions of hVDAC1 are collected in glass tubes of 5 mL volume. 16. The fractions that showed UV-absorption by hVDAC1 during elution are analyzed for hVDAC1 by SDS-PAGE [52, 53].

Insertion of β-barrel membrane proteins

479

17. Contents of glass tubes containing purified hVDAC1 are pooled and then concentrated to ~20–30 mg/mL. 18. The concentration of hVDAC1 is carefully determined, for example, using the method by Lowry [54]. 3.2 Labeling of Cysteine Residues by Fluorophores or by Spin-Labels as Fluorescence Quenchers

For site-directed fluorescence spectroscopy with thiol-bound fluorescence labels and for site-directed fluorescence quenching with thiol-bound spin-labels, single C mutants or single W, single C mutants of the β-TMP must be expressed and purified as described above. Single C-mutants may either be labeled with a fluorescence label for fluorescence spectroscopy or with a fluorophore that is used as an acceptor in FRET experiments. In single W, single C mutants, the C residue may be labeled by a spin-label that acts as an intramolecular fluorescence quencher (see Note 4) or alternatively, by a FRET acceptor for W as a donor. 1. In black reaction tubes, concentrated stock solutions of a single-C β-TMP like single-C-OmpA or single-C-VDAC are mixed with labeling buffer to obtain a final concentration of 30 μM β-TMP in a total volume of 1 mL (i.e., ~0.9 mg hVDAC1 or 1.05 mg OmpA) (see Notes 5–7). 2. To reduce disulfide bonds of C residues, 10 μL of a stock solution of TCEP (12 mM) in distilled water are added for a ~4-fold molar excess of TCEP over β-TMP. 3. The sample is flushed with nitrogen gas for 1 min and incubated for 30 min at room temperature under gentle shaking, for example, in a thermomixer. 4. Mix 3 μL of a 100 mM stock solution of reactive labeling reagent, typically a tenfold molar excess over the β-TMP, to the reduced β-TMP at 4
C. Incubate in the dark under shaking for at least 12 h (overnight). 5. Remove excess label either by dialysis against labeling buffer, by a desalting column or by repeated centrifugation in a centrifuge concentrator and resolubilization in labeling buffer (see also Note 8). 6. Labeled β-TMPs are concentrated in a centrifuge concentrator. 7. The concentration of the β-TMP is estimated, for example, using the method of Lowry [54].

3.3 S-Methylation of C Residues in Proteins

1. For intramolecular fluorescence quenching with spin-labels covalently linked to a thiol group of a C-residue, control experiments in the absence of the quencher are required. As the free thiol group of C already quenches fluorescence and to avoid dimerization of the β-TMP, the thiol group is methylated [55, 56]. For the methylation reaction, the β-TMP is first dissolved from a stock solution in 1 mL of borate buffer to a concentration of ~30 μM.

480

Lisa Gerlach et al.

2. To reduce disulfide bonds of C residues, 10 μL of a stock solution of TCEP (12 mM) in distilled water is added for an ~fourfold molar excess of TCEP over β-TMP. 3. The sample is flushed with nitrogen gas for 1 min and incubated for 30 min at room temperature under gentle shaking, for example, in a thermomixer. 4. For methylation, 6 μL of MNB (a 20-fold molar excess) are added to the reduced β-TMP and the sample is again flushed with nitrogen gas for 1 min then incubated for 2 h at 37
C (see Note 9). 5. After C methylation, excess MNB is removed either by extensive dialysis against 1 L borate buffer with at least five replacements of the dialysis buffer, by a desalting column or by repeated centrifugation in a centrifuge concentrator and resolubilization in labeling buffer. 6. The labeled protein is concentrated using a centrifuge concentrator. 7. The concentration of the methylated β-TMP is carefully estimated, for example, using the method of Lowry [54]. The degree of methylation is determined by estimating the remaining free thiol groups using 5,50 -dithiobis(2-nitrobenzoic Acid) (DTNB, Ellman’s reagent) as described in the protocol by Riddles et al. [57], page 58f. 3.4 Analysis of the Labeling Efficiency Using Ellman’s Method

The labeling efficiency should be estimated either by recording the photometric absorption of the introduced fluorescence label (subtracting the background absorption of the unlabeled protein) or by EPR spectroscopy of the spin-label (used here as a fluorescence quencher). For S-Methylation, spin-labels, or other labels that do not absorb at ADTNB (412 nm), the content of free thiols (nonlabeled SH groups) is estimated using Ellman’s reagent DTNB as described in [57], page 58f and elsewhere: 1. 1.5 mL of Tris buffer are transferred into a Quartz cuvette and the absorbance at 412 nm (A412) is set to zero using the Autozero function of the UV/VIS spectrophotometer. 2. 50 μL of Tris buffer are mixed with 50 μL of labeling buffer, added to the Quartz cuvette and the absorbance of this buffer at 412 nm (ABuffer) is recorded. 3. For the sample measurement, 1.5 mL of Tris buffer are placed into a cuvette and A412 is again set to zero. 4. 50 μL of DTNB solution (for a fivefold molar excess to the protein) is added and ADTNB is recorded at 412 nm. 5. 50 μL of labeled β-TMP is added and AFinal is recorded at 412 nm after the Absorption is stable. 6. Evaluate the concentration of free thiols originally present in the protein from the recorded absorbances at 412 nm:

Insertion of β-barrel membrane proteins

481

ΔA412 ¼ AFinal  ð3:1=3:2Þ ðA DTNB  A Buffer Þ: To determine the free thiol groups of a β-TMP in solutions containing 8 M Urea, calculate the concentration of the formed TNB2 anions corresponding to the original concenand ε412 tration of free thiols from ΔA412 (TNB2) ¼ 13,700 M1 cm1. For estimation of the concentration of TNB2 anions in nondenaturing buffers use ε412 (TNB2) ¼ 14,159 M1 cm1. 3.5 Preparation of Lipid Bilayers

Phospholipids with saturated or unsaturated acyl chains and different head groups may all be suitable for the preparation of lipid bilayers and lipid vesicles. Lipid bilayers can be prepared also from mixtures of different phospholipids. For membrane protein folding studies it should be noted that β-TMPs like OmpA or FomA most often do not fold into lipid bilayers of phospholipids with myristoyl or longer fatty acid chains unless SUVs are prepared and the temperature chosen for the folding experiment of the β-TMP is above the phase transition temperature for the transition from the lamellarordered to the lamellar disordered (fluid) phase [4, 7, 24, 58]. Similar observations have been made also for OmpF [11], FomA [4, 8, 59] and other β-TMPs [12, 13]. For folding experiments with thinner bilayers, formed, for example, by 1,2-dilauroyl-sn-glycero3-phosphocholine (diC12:0PC), LUVs can be used [59, 60]. Here preparation of lipid membranes is described for a lipid bilayer composed of equimolar amounts of diC12:0PC and 1,2-dilauroyl-snglycero-3-phosphoglycerol (diC12:0PG). For investigations on membrane protein topology bilayers may also contain labeled lipids (e.g., brominated lipids) as fluorescence quenchers [27, 30]. 1. Lyophilized diC12:0PC and diC12:0PG are separately dissolved in a mixture of chloroform–methanol (1:1 v/v), for example, at concentrations of 10 mg/mL. Use a Hamilton syringe or organic solvent compatible pipettor for organic solvents. 2. Appropriate volumes of the solutions of diC12:0PC and diC12:0PG are mixed for a selected molar ratio of diC12:0PC/ diC12:0PG using Hamilton syringes or organic solvent compatible pipettors. 3. The mixed solution of diC12:0PC/diC12:0PG is dried in the fume hood under a gentle stream of nitrogen to form thin lipid films. 4. The lipid films are carefully dried in a desiccator connected to an oil pump for 4 h under high vacuum to remove the residual solvent. Lipids films can be stored under Argon gas at 20
C, but preferably they are used after preparation. 5. Dried films are hydrated in a buffer of choice (e.g., glycine buffer) at an appropriate concentration (e.g., 2 mg/mL) and dispersed using a lab vortex mixer.

482

Lisa Gerlach et al.

3.5.1 Preparation of LUVs

1. The lipid dispersion is subjected to seven freeze/thaw cycles in liquid nitrogen and in a water-bath at 30–40
C. 2. The freeze-fractured lipid dispersion is extruded 30 times through polycarbonate membranes of a selected pore size, typically between 50 and 800 nm [60] (see Note 10). The lipid dispersion should become transparent after the first extrusion through the polycarbonate membranes.

3.5.2 Preparation of SUVs

1. Hydrated lipid dispersions (e.g., 2 mg/mL) are transferred to a 5 mL pear-shaped flask and sonicated in for 50 min using the microtip of an ultrasonifier at 50% duty cycle (0.5 s on/0.5 s off) at 5–10% power under concurrent cooling in an ice/water bath. Ensure that the tip is in the lipid solution and does not touch the wall of the pear-shaped flask. After sonication, the lipid dispersion must be transparent and opalescent; otherwise repeat at slightly higher power. 2. Remove the titanium dust from the microtip of the ultrasonifier by centrifugation at 700  g in a tabletop centrifuge. 3. SUVs are incubated at 4
C for at least 10–12 h and used the day after preparation. The diameter of the SUVs prepared by this method is typically 30 nm, as can be checked in dynamic light scattering experiments (see Note 11).

3.6 Setup the Spectrofluorometer for Fluorescence Spectroscopy

1. Ensure the spectrofluorometer is properly calibrated by recording the water Raman signal and if necessary recalibrate both, excitation and emission monochromators as described in the instruction manual for the spectrofluorometer. 2. Turn on the water-bath for temperature control of the sample compartment and set the temperature for the experiment. 3. Set excitation and emission band-width by adjusting the slits. On a Spex-Fluorolog, excitation and emission bandwidths of 2.5–3 nm are good choices. 4. Set the excitation wavelength. For intrinsic W-fluorescence spectroscopy, an excitation wavelength of 295 nm should be chosen for selective excitation of W-residues (see also Fig. 1). 5. Set shutters to auto-open/auto-close. 6. Set other instrument parameters, when necessary; for example, for a Spex Fluorolog 3, high-voltage has to be switched on and the integration time has to be specified. An integration time of 0.05–0.1 s is usually sufficient (see Note 12). 7. Specify the scan range for which fluorescence spectra are recorded, the wavelength increment for the scan and the number of scans that shall be accumulated and averaged to improve the signal/noise ratio.

Insertion of β-barrel membrane proteins

483

8. Set the number of scans that should be averaged to improve the signal/noise ratio of the spectrum. 9. First record a fluorescence spectrum of a solution of the amino acid tryptophan or the labeling reagent in buffer at a defined concentration (e.g., 2 μM tryptophan). A background spectrum of the buffer in the absence of tryptophan is subtracted. The maximum intensity of this spectrum can be used for normalization of spectra recorded for the same settings of the spectrofluorometer. Normalization is useful as the fluorescence intensity depends on instrument parameters, like the light source, quality of mirrors and gratings, and the settings for the band width of the excitation and the emission monochromators. Normalized spectra will be independent of instrumental parameters. 3.7 Kinetics of Folding of β-TMPs by Fluorescence Recordings

3.7.1 Folding of β-TMPs by Recording Fluorescence at Specific Wavelength as a Function of Time

Membrane protein folding of β-TMPs into lipid bilayers is slow. For OmpA [24] and FomA and for folding into SUVs of diC18:1PC [8] it has been reported that insertion and folding into lipid membranes takes up to several hours [4]. It is important to find conditions where insertion and folding of the β-TMP are fast (see Note 13). Rate-limiting for the folding of OmpA are the steps that follow membrane adsorption (at least in the absence of folding machines). For folding events on a longer time scale, fluorescence spectra can be recorded every few minutes to obtain additional information like shifts of the wavelength at which fluorescence intensity has a maximum. Early folding events up to the adsorption at the surface of the lipid bilayer are relatively fast. For faster folding events of β-TMPs, the change of the intensity at a selected emission wavelength should be directly recorded as a function of time [26, 61]. The software for most spectrofluorometers supports time-based recordings. 1. Setup the spectrofluorometer as described in Subheading 3.6, steps 1–6. 2. Select time-based spectrofluorometer.

scan

in

the

software

for

the

3. For acquisition of a time-based scan, set the emission monochromator to an appropriate emission wavelength. For following the time course of W-fluorescence in a folding experiment of a β-TMP, the emission monochromator is typically set to 330 nm. Upon membrane insertion and folding, W-fluorescence increases at this wavelength, because most W-residues of β-TMPs are found in aromatic girdles below the glycerol backbone of the phospholipids within the hydrophobic region of the lipid bilayer. 4. Adjust the concentration of the stock solution of the β-TMP in buffer containing 8 M urea to 100 μM (for OmpA this

484

Lisa Gerlach et al.

corresponds to about 3.5 g/L) (see Notes 13 and 14). Keep samples on ice. 5. The concentration of the lipid or total lipid of a lipid mixture (LUVs or SUVs) should be sufficiently high to obtain the desired lipid/β-TMP ratio after initiation of the folding reaction in 1 mL final volume in the fluorescence cuvette. Prepare a stock dispersion of lipid vesicles in the buffer of choice as described in Subheading 3.5. For a lipid/protein ratio of 1600 mol/mol (see Note 15), stock lipid vesicles are prepared at 3.6 mM (about ~2.0–2.9 g/L, depending on the lipid). 6. Transfer 900 μL of the lipid vesicles into the fluorescence cuvette. Add 80 μL of the corresponding lipid free buffer. 7. As folding of β-TMPs into membranes is slow, folding can be initiated by fast hand mixing of 20 μL of the stock solution of the β-TMP into the cuvette containing the preformed lipid vesicles with a dead-time of about 10 s (see Note 14). The time-based scan is immediately started after mixing. 3.7.2 Kinetics of Folding of β-TMPs by Recording Spectra as a Function of Time

1. Setup the spectrofluorometer as described in Subheading 3.6. 2. Adjust the concentration of the stock solution of the β-TMP in buffer containing 8 M urea to 100 μM (for OmpA this corresponds to about 3.5 g/L) (see Notes 13 and 14). Keep samples on ice. 3. The concentration of the lipid or total lipid of a lipid mixture (LUVs or SUVs) should be sufficiently high to obtain the desired lipid/β-TMP ratio after initiation of the folding reaction in 1 mL final volume in the fluorescence cuvette. Prepare a stock dispersion of lipid vesicles in the buffer of choice as described in Subheading 3.5. For a lipid/protein ratio of 400 mol/mol, stock lipid vesicles are prepared at 1.6 mM (about ~0.9 to 1.3 g/L, depending on the lipid) (see Note 15). 4. Transfer 500 μL of the lipid vesicles into the fluorescence cuvette. Add 480 μL of the corresponding lipid free buffer. 5. Record the background spectrum for subtraction and save it into a file. For a good signal/noise ratio, several scans should be averaged as necessary. 6. On the spectrofluorometer, enter the file name of the spectrum of the lipid background for auto subtraction from the spectra acquired subsequently. 7. Quickly mix 20 μL of the stock solution of the β-TMP in buffer containing 8 M urea into the cuvette, diluting the urea 50-fold in the presence of the lipid vesicles to initiate folding (see Note 14). Start a timer and record the first fluorescence spectrum of the β-TMP. Several scans should be accumulated for a good

Insertion of β-barrel membrane proteins

485

signal/noise ratio. When supported by the software, autosubtract the background spectrum recorded previously. 8. Record additional fluorescence spectra every 2–10 min and take note of the time indicated by the timer. Acquisition of spectra can be stopped when changes between subsequent spectra are small. 9. Plot the intensity at a selected wavelength as a function of time. For W-fluorescence, the intensity at 330 nm is a good choice. 10. Analyze the kinetics. Since at high lipid/β-TMP ratios, folding steps typically follow a rate-law of pseudo first-order, fits of an exponential function or of a sum of two or more exponential functions to the experimental data may be realistic. To validate the kinetic model, it will be necessary to perform additional experiments, for example at different concentrations of β-TMP and/or lipid, or by adjusting other parameters (pH, temperature, etc.). It may also be instructive to complement the fluorescence experiments with other spectroscopic methods (e.g., circular dichroism spectroscopy) [26]. 3.7.3 Determination of Insertion of W-Residues of β-TMPs Using Brominated Lipids

The kinetic methods described above have been successfully expanded to monitor local changes at the surface of a β-TMP. Fluorescence properties of single W mutants of β-TMPs directly reflect changes in the local environment of the region of the protein into which the single W is introduced [30]. This environment changes during membrane insertion and folding of the β-TMP (Fig. 2b). The changes will depend on the region into which the W is introduced. W-fluorescence spectroscopy applied to a set of single W mutants that are separately used in kinetic folding and insertion experiments described in preceding sections can provide important information on the folding mechanism of a β-TMP (see Fig. 2b). This strategy of combining kinetic experiments with sitedirected fluorescence spectroscopy has been further extended by introducing lipid bound fluorescence quenchers that quench fluorescence dependent on the distance of the fluorophore from the center of the lipid bilayer. Again, the same protocols as described in the preceding sections are used, except that lipid membranes are prepared that are composed of the host lipid and a lipid carrying a fluorescence quencher at a suitable concentration, that is, 30 mol-% in the case of lipids carrying bromines on their sn-2 acyl chain. Work on insertion and folding of OmpA took advantage of brominated lipids to examine insertion of W-residues as a function of time at various temperatures [30]. In brominated lipids, two bromine atoms are covalently bound to neighboring carbon atoms of an sn-2 fatty acid chain of a phospholipid. These vicinal bromines act as efficient quenchers for W-fluorescence over short distances [62, 63]. For single W mutants of OmpA, several kinetic

486

Lisa Gerlach et al.

experiments were performed. In each experiment diC18:1PC was mixed with one out of five different brominated lipids (m,m+1 DiBrPC) at a molar ratio of 7/3 to record W-fluorescence changes upon insertion and folding of OmpA according to the protocol given in Subheading 3.7.1. The changes in intensity throughout the time course of insertion and folding of OmpA depend on the distance of the bromines from the center of the lipid bilayer. Fluorescence time courses of folding and insertion of single W-mutants of OmpA were obtained using 4,5-DiBrPC, 6,7-DiBrPC, 9,10DiBrPC or 11,12-DiBrPC [30]. For these brominated lipids, the distances of the bromine atoms from the center of the lipid bilayer, dQ, have been determined to be 12.8, 11.0, 8.3, and 6.5 A˚ [64]. A fifth kinetic experiment has to be performed with lipid bilayers containing diC18:1PC only. This experiment allows normalization of the fluorescence data. Fluorescence at a specific time t after initiation of folding, obtained in the presence of one of the brominated lipids, FQ(t, dQ), is divided by the fluorescence obtained in the absence of quenchers at the same time after initiation of folding, F0(t). The relative fluorescence quenching can be used to estimate the distance of the fluorescent W from the bilayer center at a specific time, for example by parallax method [36, 65], or as in ref. 30 by distribution analysis [38, 66]:   F Q ðt, d Q Þ S 2 ¼ exp  pffiffiffiffiffiffi  expf1=2 ½ðd Q  d W Þ=σ g ð1Þ F 0 ðtÞ σ 2π dQ is the distance of the quencher from the lipid bilayer center, and dW is the distance of the fluorophore (W) from the center. The dispersion, σ, is a function of the sizes of the fluorophore and quencher and thermal fluctuations between the two. S is a function of the quenching efficiency and quencher concentration in the membrane. For the entire set of kinetic experiments, folding trajectories of W-residues in projection to the membrane normal have been obtained using this approach [30]. 3.7.4 Determination of βStrand Association by SiteDirected Fluorescence Quenching in β-TMPs

Site-directed fluorescence quenching has been used to examine the structure formation of β-TMPs in a lipid environment [41]. In a folding experiment, the evolution of fluorescence quenching as a function of time at specific sites within the protein can provide important information on the folding mechanism of the β-TMP. Here, double mutants of β-TMPs are used in the protocols described above to examine the kinetics of folding. Based on a plasmid encoding a mutant of the β-TMP, in which all W are replaced by F and all C are replaced by A, two new mutations are introduced, a fluorescent W and a C, to which a spin-label is covalently attached as a fluorescence quencher as described in Subheading 3.2. The sites for these mutations are selected so that the W is in close proximity to the quencher at the C when

Insertion of β-barrel membrane proteins

487

the β-TMP is folded. To normalize fluorescence time courses, the same kinetic experiment must be performed also in absence of the quencher. This experiment has to be performed with the same mutant, but carrying a methyl group instead of the spin-label at the C. S-Methylation of the C is described in Subheading 3.3 (see Notes 16 and 17). When both experiments are completed, the relative fluorescence quenching, RQ ¼ F(SL-WnCm)/F (Me-WnCm) is calculated, where F(SL-WnCm) is the fluorescence intensity of the spin-labeled form and F(Me-WnCm) the fluorescence intensity of the methylated form of the mutant. The Table shown in Fig. 3b lists the RQ obtained for several single W, single C mutants of OmpA obtained 60 min after initiation of folding at the indicated temperatures (see ref. 41 for details).

4

Notes 1. Calibrated precision pipettors should be used. Pipettors should be carefully checked for accuracy before use and when necessary calibrated. Pipettors should only be used for aqueous solutions unless they are specifically designed for organic solvent. Do not use plastic pipettors or plastic pipettor tips for organic solvents, use Hamilton (glass) syringes or organic solvent compatible pipettors for solvents like chloroform, acetonitrile etc. Use disposable pipettor tips once and do not mix different solutions by using the same tip. 2. All buffers containing high concentrations of urea should be prepared fresh. At ~20
C and at a pH >7.0 a solution of 8 M urea will already contain 4 mM cyanate after 1 week and ~20 mM cyanate when equilibrium is reached. Cyanate reacts with -SH or -NH2 groups of proteins which should be avoided as much as possible. Cyanate can be removed by passing solutions containing urea over a Bio-Rad AG 501-X8 column. Check the conductance of the solution to ensure cyanate is absent. Unfolded proteins in buffers containing urea should be kept frozen. 3. The expression level and optimal expression condition depends on the β-TMP, choice of expression vector, used strain etc. and can be optimized, for example, by adjusting the OD used for induction. 4. Nitroxide spin-labels act as fluorescence quenchers over short distances [35, 36]. For intramolecular fluorescence quenching a reference fluorescence experiment is required, in which the C is methylated, see Subheading 3.3, as free C may also quench fluorescence.

488

Lisa Gerlach et al.

5. Transmembrane segments of β-TMPs like OmpA or hVDAC1 are best labeled in unfolded form as the thiol group of the C residue introduced by site-directed mutagenesis is then well accessible for the labeling reagent. Folded proteins may be labeled at surface exposed C residues that are accessible to the labeling reagent. 6. For efficient labeling a pH range of 7.2–7.5 is used. The thiol groups are then nucleophilic, supporting the reaction with the label. 7. To avoid dimerization of MTSL or comparable labeling reagents, the concentration of the labeling reagent should be kept below 500 μM in the reaction mixture. This also limits the concentration of the β-TMP, as a tenfold molar excess of MTSL is used for labeling. If more labeled β-TMP is needed, it is better to increase the volume of the labeling reaction instead of the concentrations of the reactants. 8. The removal of the excess of label is crucial, further dialysis steps might help. Do not add reducing agents like DTT after labeling. 9. For efficient S-methylation, a basic pH of 9.0 is required. 10. LUVs should always be prepared fresh for direct use in folding experiments. The lipid suspension should be kept above the phase transition temperature of the chosen lipid during hydration and extrusion. 11. SUVs are metastable. They should be equilibrated for at least 12 h (overnight) and used the next day to ensure good reproducibility of folding experiments. SUVs should not be stored. 12. Longer integration times improve the signal/noise ratio but extended light exposure may also cause photobleaching. 13. An accurate estimation of the protein concentration is essential when measuring fluorescence since it can influence the fluorescence intensity dramatically. The concentration of the β-TMP should be sufficiently high to obtain a good signal/noise ratio. However, there is an upper limit of the concentration of the βTMP that can be used in this experiment: For a valid recording of the folding kinetics of a protein, the fluorescence intensity detected by the photomultiplier should increase linearly with the concentration of the β-TMP. However, the response of the photomultiplier (detector) becomes nonlinear when the number of emitted photons is too high. For a SPEX fluorolog 3, the concentration of the β-TMP should be chosen so that the fluorescence intensity does not exceed 2 megacounts per second (Mcps). 14. The dilution of the denaturant is critical [18], the optimal dilution factor depends on the β-TMP.

Insertion of β-barrel membrane proteins

489

15. Aggregation competes with folding into native form. If the concentration of the β-TMP is too high and the lipid concentration too low, the aggregation of the protein may be favored. 16. For comparisons, it is important that the β-TMPs are as pure as possible. Concentrations of both the spin-labeled and the methylated forms must be determined very accurately. 17. It is important that spin-labeling and methylation are both complete. Free thiol groups should be estimated as described in Subheading 3.4. References 1. Kleinschmidt JH (2015) Folding of β-barrel membrane proteins in lipid bilayers - unassisted and assisted folding and insertion. Biochim Biophys Acta 1848(9):1927–1943. https:// doi.org/10.1016/j.bbamem.2015.05.004 2. McMorran LM, Brockwell DJ, Radford SE (2014) Mechanistic studies of the biogenesis and folding of outer membrane proteins in vitro and in vivo: what have we learned to date? Arch Biochem Biophys 564:265–280. https://doi.org/10.1016/j.abb.2014.02.011 3. Otzen DE, Andersen KK (2013) Folding of outer membrane proteins. Arch Biochem Biophys 531(1–2):34–43. https://doi.org/10. 1016/j.abb.2012.10.008 4. Kleinschmidt JH (2006) Folding kinetics of the outer membrane proteins OmpA and FomA into phospholipid bilayers. Chem Phys Lipids 141(1–2):30–47. https://doi.org/10. 1016/j.chemphyslip.2006.02.004 5. Kleinschmidt JH (2003) Membrane protein folding on the example of outer membrane protein A of Escherichia coli. Cell Mol Life Sci 60(8):1547–1558. https://doi.org/10.1007/ s00018-003-3170-0 6. Tamm LK, Arora A, Kleinschmidt JH (2001) Structure and assembly of β-barrel membrane proteins. J Biol Chem 276(35):32399–32402 7. Surrey T, J€ahnig F (1992) Refolding and oriented insertion of a membrane protein into a lipid bilayer. Proc Natl Acad Sci U S A 89 (16):7457–7461 8. Pocanschi CL, Apell H-J, Puntervoll P, Høgh BT, Jensen HB, Welte W, Kleinschmidt JH (2006) The major outer membrane protein of Fusobacterium nucleatum (FomA) folds and inserts into lipid bilayers via parallel folding pathways. J Mol Biol 355:548–561 9. Shanmugavadivu B, Apell HJ, Meins T, Zeth K, Kleinschmidt JH (2007) Correct folding of the β-barrel of the human membrane

protein VDAC requires a lipid bilayer. J Mol Biol 368:66–78 10. Huysmans GH, Radford SE, Brockwell DJ, Baldwin SA (2007) The N-terminal helix is a post-assembly clamp in the bacterial outer membrane protein PagP. J Mol Biol 373 (3):529–540. https://doi.org/10.1016/j. jmb.2007.07.072 11. Surrey T, Schmid A, J€ahnig F (1996) Folding and membrane insertion of the trimeric β-barrel protein OmpF. Biochemistry 35 (7):2283–2288 12. Dewald AH, Hodges JC, Columbus L (2011) Physical determinants of β-barrel membrane protein folding in lipid vesicles. Biophys J 100 (9):2131–2140. https://doi.org/10.1016/j. bpj.2011.03.025 13. Burgess NK, Dao TP, Stanley AM, Fleming KG (2008) β-barrel proteins that reside in the Escherichia coli outer membrane in vivo demonstrate varied folding behavior in vitro. J Biol Chem 283(39):26748–26758. https://doi. org/10.1074/jbc.M802754200 14. Dornmair K, Kiefer H, J€ahnig F (1990) Refolding of an integral membrane protein. OmpA of Escherichia coli. J Biol Chem 265 (31):18907–18911 15. Popot J-L (2014) Folding membrane proteins in vitro: a table and some comments. Arch Biochem Biophys 564:314–326. https://doi. org/10.1016/j.abb.2014.06.029 16. Pocanschi CL, Popot J-L, Kleinschmidt JH (2013) Folding and stability of outer membrane protein A (OmpA) from Escherichia coli in an amphipathic polymer, amphipol A8–35. Eur Biophys J 42(2–3):103–118. https://doi. org/10.1007/s00249-013-0887-z 17. Kleinschmidt JH, Popot JL (2014) Folding and stability of integral membrane proteins in amphipols. Arch Biochem Biophys 564C:327–343. https://doi.org/10.1016/j. abb.2014.10.013

490

Lisa Gerlach et al.

18. Conlan S, Bayley H (2003) Folding of a monomeric porin, OmpG, in detergent solution. Biochemistry 42(31):9453–9465 19. Anbazhagan V, Vijay N, Kleinschmidt JH, Marsh D (2008) Protein-lipid interactions with Fusobacterium nucleatum major outer membrane protein FomA: spin-label EPR and polarized infrared spectroscopy. Biochemistry 47(32):8414–8423. https://doi.org/10. 1021/bi800750s 20. Anbazhagan V, Qu J, Kleinschmidt JH, Marsh D (2008) Incorporation of outer membrane protein OmpG in lipid membranes: proteinlipid interactions and β-barrel orientation. Biochemistry 47(23):6189–6198. https://doi. org/10.1021/bi800203g 21. Ramakrishnan M, Qu J, Pocanschi CL, Kleinschmidt JH, Marsh D (2005) Orientation of β-barrel proteins OmpA and FhuA in lipid membranes. Chain length dependence from infrared dichroism. Biochemistry 44 (9):3515–3523. https://doi.org/10.1021/ bi047603y 22. Ramakrishnan M, Pocanschi CL, Kleinschmidt JH, Marsh D (2004) Association of spinlabeled lipids with β-barrel proteins from the outer membrane of Escherichia coli. Biochemistry 43(37):11630–11636. https://doi.org/ 10.1021/bi048858e 23. Patel GJ, Kleinschmidt JH (2013) The lipid bilayer-inserted membrane protein BamA of Escherichia coli facilitates insertion and folding of outer membrane protein A from its complex with Skp. Biochemistry 52(23):3974–3986. https://doi.org/10.1021/bi400103t 24. Kleinschmidt JH, Tamm LK (1996) Folding intermediates of a β-barrel membrane protein. Kinetic evidence for a multi-step membrane insertion mechanism. Biochemistry 35 (40):12993–13000 25. Surrey T, J€ahnig F (1995) Kinetics of folding and membrane insertion of a β-barrel membrane protein. J Biol Chem 270 (47):28199–28203 26. Kleinschmidt JH, Tamm LK (2002) Secondary and tertiary structure formation of the β-barrel membrane protein OmpA is synchronized and depends on membrane thickness. J Mol Biol 324:319–330 27. Kleinschmidt JH, Tamm LK (1999) Timeresolved distance determination by tryptophan fluorescence quenching: probing intermediates in membrane protein folding. Biochemistry 38 (16):4996–5005 28. Huysmans GH, Radford SE, Baldwin SA, Brockwell DJ (2012) Malleability of the folding mechanism of the outer membrane protein

PagP: parallel pathways and the effect of membrane elasticity. J Mol Biol 416(3):453–464. https://doi.org/10.1016/j.jmb.2011.12.039 29. Arora A, Rinehart D, Szabo G, Tamm LK (2000) Refolded outer membrane protein A of Escherichia coli forms ion channels with two conductance states in planar lipid bilayers. J Biol Chem 275(3):1594–1600 30. Kleinschmidt JH, den Blaauwen T, Driessen A, Tamm LK (1999) Outer membrane protein A of E. coli inserts and folds into lipid bilayers by a concerted mechanism. Biochemistry 38 (16):5006–5016 31. Gupta A, Zadafiya P, Mahalakshmi R (2014) Differential contribution of tryptophans to the folding and stability of the attachment invasion locus transmembrane β-barrel from Yersinia pestis. Sci Rep 4:6508. https://doi.org/10. 1038/srep06508 32. Hong H, Rinehart D, Tamm LK (2013) Membrane depth-dependent energetic contribution of the tryptophan side chain to the stability of integral membrane proteins. Biochemistry 52 (25):4413–4421. https://doi.org/10.1021/ bi400344b 33. Huysmans GH, Baldwin SA, Brockwell DJ, Radford SE (2010) The transition state for folding of an outer membrane protein. Proc Natl Acad Sci U S A 107(9):4099–4104. https://doi.org/10.1073/pnas.0911904107 34. Qu J, Behrens-Kneip S, Holst O, Kleinschmidt JH (2009) Binding regions of outer membrane protein A in complexes with the periplasmic chaperone Skp. A site-directed fluorescence study. Biochemistry 48(22):4926–4936. https://doi.org/10.1021/bi9004039 35. Abrams FS, London E (1993) Extension of the parallax analysis of membrane penetration depth to the polar region of model membranes: use of fluorescence quenching by a spin-label attached to the phospholipid polar headgroup. Biochemistry 32(40):10826–10831 36. Chattopadhyay A, London E (1987) Parallax method for direct measurement of membrane penetration depth utilizing fluorescence quenching by spin-labeled phospholipids. Biochemistry 26(1):39–45 37. London E, Feigenson GW (1981) Fluorescence quenching in model membranes. 1. Characterization of quenching caused by a spin-labeled phospholipid. Biochemistry 20 (7):1932–1938 38. Ladokhin AS (1997) Distribution analysis of depth-dependent fluorescence quenching in membranes: a practical guide. Methods Enzymol 278:462–473

Insertion of β-barrel membrane proteins 39. Ladokhin AS, Wang L, Steggles AW, Holloway PW (1991) Fluorescence study of a mutant cytochrome b5 with a single tryptophan in the membrane-binding domain. Biochemistry 30(42):10200–10206 40. Ladokhin AS (1999) Evaluation of lipid exposure of tryptophan residues in membrane peptides and proteins. Anal Biochem 276 (1):65–71 41. Kleinschmidt JH, Bulieris PV, Qu J, Dogterom M, den Blaauwen T (2011) Association of neighboring β–strands of outer membrane protein A in lipid bilayers revealed by site directed fluorescence quenching. J Mol Biol 407(2):316–332. https://doi.org/10.1016/j. jmb.2011.01.021 42. Kang G, Lopez-Pena I, Oklejas V, Gary CS, Cao W, Kim JE (2012) Fo¨rster resonance energy transfer as a probe of membrane protein folding. Biochim Biophys Acta 1818 (2):154–161. https://doi.org/10.1016/j. bbamem.2011.08.029 43. Prilipov A, Phale PS, Van Gelder P, Rosenbusch JP, Koebnik R (1998) Coupling sitedirected mutagenesis with high-level expression: large scale production of mutant porins from E. coli. FEMS Microbiol Lett 163 (1):65–72 44. Datta DB, Arden B, Henning U (1977) Major proteins of the Escherichia coli outer cell envelope membrane as bacteriophage receptors. J Bacteriol 131(3):821–829 45. Freudl R, MacIntyre S, Degen M, Henning U (1988) Alterations to the signal peptide of an outer membrane protein (OmpA) of Escherichia coli K-12 can promote either the cotranslational or the posttranslational mode of processing. J Biol Chem 263(1):344–349 46. Kleinschmidt JH, Wiener MC, Tamm LK (1999) Outer membrane protein A of E. coli folds into detergent micelles, but not in the presence of monomeric detergent. Protein Sci 8(10):2065–2071 47. Pautsch A, Vogt J, Model K, Siebold C, Schulz GE (1999) Strategy for membrane protein crystallization exemplified with OmpA and OmpX. Proteins 34(2):167–172 48. Maurya SR, Mahalakshmi R (2013) Modulation of human mitochondrial voltagedependent anion channel 2 (hVDAC-2) structural stability by cysteine-assisted barrel-lipid interactions. J Biol Chem 288 (35):25584–25592. https://doi.org/10. 1074/jbc.M113.493692 49. Mager F, Gessmann D, Nussberger S, Zeth K (2011) Functional refolding and characterization of two Tom40 isoforms from human

491

mitochondria. J Membr Biol 242(1):11–21. https://doi.org/10.1007/s00232-011-93728 50. Xia Y, Xun L (2017) Revised mechanism and improved efficiency of the QuikChange sitedirected mutagenesis method. Methods Mol Biol 1498:367–374. https://doi.org/10. 1007/978-1-4939-6472-7_25 51. Taniguchi N, Murakami H (2017) Multiple site-directed and saturation mutagenesis by the patch cloning method. Methods Mol Biol 1498:339–347. https://doi.org/10.1007/ 978-1-4939-6472-7_22 52. Laemmli UK (1970) Cleavage of structural proteins during the assembly of the head of bacteriophage T4. Nature 227:680–685 53. Weber K, Osborne M (1964) The reliability of molecular weight determinations by dodecyl sulfate-polyacrylamide gel electrophoresis. J Biol Chem 244:4406–4412 54. Lowry OH, Rosebrough NJ, Farr AL, Randall RJ (1951) Protein measurement with the Folin phenol reagent. J Biol Chem 193:265–275 55. Heinrikson RL (1971) The selective S-methylation of sulfhydryl groups in proteins and peptides with methyl-p-nitrobenzenesulfonate. J Biol Chem 246(12):4090–4096 56. Hunziker PE (1991) Cysteine modification of metallothionein. Methods Enzymol 205:399–400 57. Riddles PW, Blakeley RL, Zerner B (1983) Reassessment of Ellman’s reagent. Methods Enzymol 91:49–60 58. Rodionova NA, Tatulian SA, Surrey T, J€ahnig F, Tamm LK (1995) Characterization of two membrane-bound forms of OmpA. Biochemistry 34(6):1921–1929 59. Marsh D, Shanmugavadivu B, Kleinschmidt JH (2006) Membrane elastic fluctuations and the insertion and tilt of β-barrel proteins. Biophys J 91:227–232 60. Pocanschi CL, Patel GJ, Marsh D, Kleinschmidt JH (2006) Curvature elasticity and refolding of OmpA in large unilamellar vesicles. Biophys J 91:L75–L78 61. Schiffrin B, Calabrese AN, Higgins AJ, Humes JR, Ashcroft AE, Kalli AC, Brockwell DJ, Radford SE (2017) Effects of periplasmic chaperones and membrane thickness on BamAcatalyzed outer-membrane protein folding. J Mol Biol 429(23):3776–3792. https://doi. org/10.1016/j.jmb.2017.09.008 62. Bolen EJ, Holloway PW (1990) Quenching of tryptophan fluorescence by brominated phospholipid. Biochemistry 29(41):9638–9643 63. Markello T, Zlotnick A, Everett J, Tennyson J, Holloway PW (1985) Determination of the

492

Lisa Gerlach et al.

topography of cytochrome b5 in lipid vesicles by fluorescence quenching. Biochemistry 24 (12):2895–2901 64. McIntosh TJ, Holloway PW (1987) Determination of the depth of bromine atoms in bilayers formed from bromolipid probes. Biochemistry 26(6):1783–1788 65. Kaiser RD, London E (1998) Determination of the depth of BODIPY probes in model

membranes by parallax analysis of fluorescence quenching. Biochim Biophys Acta 1375 (1–2):13–22 66. Ladokhin AS, Holloway PW (1995) Fluorescence of membrane-bound tryptophan octyl ester: a model for studying intrinsic fluorescence of protein-membrane interactions. Biophys J 69(2):506–517

Chapter 21 EPR Techniques to Probe Insertion and Conformation of Spin-Labeled Proteins in Lipid Bilayers Enrica Bordignon, Svetlana Kucher, and Yevhen Polyhach Abstract Electron paramagnetic resonance (EPR) spectroscopy of spin-labeled membrane proteins is a valuable biophysical technique to study structural details and conformational transitions of proteins close to their physiological environment, for example, in liposomes, membrane bilayers, and nanodiscs. Unlike in nuclear magnetic resonance (NMR) spectroscopy, having only one or few specific side chains labeled at a time with paramagnetic probes makes the size of the object under investigation irrelevant in terms of technique sensitivity. As a drawback, extensive site-directed mutagenesis is required in order to analyze the properties of the protein under investigation. EPR can provide detailed information on side chain dynamics of large membrane proteins or protein complexes embedded in membranes with an exquisite sensitivity for flexible regions and on water accessibility profiles across the membrane bilayer. Moreover, distances between the two spin-labeled side chains in membrane proteins can be detected with high precision at cryogenic temperatures. The application of EPR to membrane proteins still presents some challenges in terms of sample preparation, sensitivity and data interpretation, thus it is difficult to give ready-to-go methodological recipes. However, new technological developments (arbitrary waveform generators) and new spin labels spectroscopically orthogonal to nitroxides increased the range of applicability from in vitro toward in-cell EPR experiments. This chapter is an updated version of the one published in the first edition of the book and describes the state of the art in the application of nitroxide-based site-directed spin labeling EPR to membrane proteins, addressing new tools such as arbitrary waveform generators and spectroscopically orthogonal labels, such as Gd(III)-based labels. We will present challenges in sample preparation and data analysis for functional and structural membrane protein studies using site-directed spin labeling techniques and give experimental details on EPR techniques providing information on side chain dynamics and water accessibility using nitroxide probes. An updated optimal Q-band DEER setup for nitroxide probes will be described, and its extension to gadolinium-containing samples will be addressed. Key words Site-directed spin labeling, Nitroxide, Gadolinium, Arbitrary waveform generator, EPR, Membrane proteins, Bilayer, Water accessibility, High field, Mobility, Distances, Double electron–electron resonance (DEER), ODNP

1

Introduction Probing insertion and conformational transitions of proteins in lipid bilayers remains a formidable task for any biophysical technique both in vivo and in vitro. Magnetic resonance techniques as

Jo¨rg H. Kleinschmidt (ed.), Lipid-Protein Interactions: Methods and Protocols, Methods in Molecular Biology, vol. 2003, https://doi.org/10.1007/978-1-4939-9512-7_21, © Springer Science+Business Media, LLC, part of Springer Nature 2019

493

494

Enrica Bordignon et al.

liquid and solid state NMR can be combined to obtain information in the two tumbling regimes (from the aqueous solution to the membrane bilayer), but the bigger the size of the protein under study, the more challenging is to obtain high resolution atomic details. X-ray crystallography can provide the structure of proteins in a membrane-like milieu (most crystals are obtained in detergent micelles), but usually only one or few protein conformations are obtainable. Time-resolved structural information is key to understanding biological mechanism and the recent advances in X-ray free electron lasers (XFEL) will allow to tackle this by exploring various methods to trigger biomolecular transformations. Last but not least, the resolution revolution of cryo-EM paved the way to high throughput of structures of large proteins and protein complexes with atomic details in membrane environments. Sitedirected spin labeling EPR, is a complementary biophysical technique, which in combination with the above mentioned methods (NMR, X-ray, and cryo-EM) can aid the understanding of dynamic processes underlining protein transformations, and can create coarse-grained models of proteins conformational transitions. EPR is particularly useful to study proteins which are partially disordered or that undergo transformation from water-soluble to membrane-embedded states, proteins which are characterized by large rearrangements in the membrane bilayer or that can form heterogeneous complexes with dynamic transient interactions. In fact, one the major advantage is that EPR can be performed with similar sensitivity in both aqueous and lipid environment, thus allowing to follow insertion and conformational changes of membrane proteins in the same sample. Clearly, the drawback of the technique is the need to genetically introduce cysteines in (most of the cases), and subsequently EPR-active paramagnetic centers, for example, nitroxide probes (NO), gadolinium (Gd3+) ions, manganese (Mn2+) ions, and trityl radicals, which give information at the molecular level on the protein. Once a nitroxide spin label is inserted in a protein (Subheading 3.1), it is possible to extract information about the dynamics of the spin-labeled side chain, and of the backbone associated with it (Subheading 3.2). Most other spectroscopically orthogonal labels, for example gadolinium, which will be briefly covered in the DEER section, do not provide such information, and they are usually not detectable by X-band CW EPR at low micromolar concentrations. Additionally, the water accessibility of a site specifically labeled with nitroxide probes can be addressed by several techniques (Subheading 3.3) and the topology of a protein with respect to the membrane-water interface can be investigated. By placing two spin labels in the protein of interest, one can also detect very precisely the mean distance and the distance distribution between the two spins via dipolar spectroscopy techniques (Subheading 3.4), allowing for detection of conformational changes in proteins

Spin labeling EPR on membrane proteins

495

during their functioning. Analysis of interspin distances in different states of a protein is nowadays the major source of information provided by EPR in structural biology, and the development of several new EPR methods with increased signal sensitivity tailored for different spin types is a field of intensive research. Here we will present an optimized Q-band four-pulse DEER sequence, which is the most commonly used technique to reliably extract interspin distances in membrane proteins. Subheading 3 will give detailed guidelines to perform the experiments on membrane proteins and to extract the information described above, focusing on the pitfalls which may be encountered both in data recording and processing. Selected examples from the literature will be used in order to convey the potential of the technique to obtain insights into different aspects of membrane protein structural details.

2

Materials

2.1 Sample Preparation and Instrumentation

In order to perform experiments similar to those outlined in this chapter one needs a molecular biology facility laboratory to express and purify membrane proteins. Additionally, basic lab equipment is needed to spin label and handle membrane protein samples. The chemistry laboratory equipped with pH meter, analytical balances, centrifuges and basic chemicals is needed for preparations of buffers, spin label, and other stock solutions. In the following we will only discuss standard cysteine-based spin-labeling approaches. Minimally required chemicals: 1. 1,4-dithio-DL-threitol (DTT) (Sigma-Aldrich) or tris(2-carboxyethyl)phosphine (TCEP) as reducing agents to prereduce all cysteines prior to spin labeling. 2. Spin labels: (a) (1-Oxyl-2,2,5,5-tetramethylpyrroline-3-methyl)-methanethiosulfonate (spin label, MTSL, Fig. 1) is the most commonly used spin label. The reagent is available from Toronto Research Chemicals, Ontario, Canada. A 100 mM stock solution of the spin label in acetonitrile or DMSO is prepared and stored at
80  C in the dark. (b) 4-Hydroxy-2,2,6,6-tetramethylpiperidine 1-oxyl (TEMPOL) as a standard spin probe can be purchased from Sigma-Aldrich. It is water soluble and can be stored in small aliquots of 100 μM concentration at
80  C for spin counting.

496

Enrica Bordignon et al. O O

-

O

O

S

N

O

N

N

S

N

N O

S

N -

O

Gd(III)-maleimido-DOTA

MTSL

N

N

O

N -

3+

Gd

S

O

O

O

O

Gd

R1

iodacetamido proxyl O N S

N

O

O

maleimido proxyl Fig. 1 Pyrroline-type spin labels and Gd(III)-maleimido-DOTA. Sketch of the structures of three nitroxide spin labels and a Gd(III)-maleimido-DOTA label which are commonly used to label cysteine residues in proteins. The inset shows one representative simulated MTSL side chain rotamer (R1 side chain) and one Gd(III)maleimido-DOTA side chain rotamer attached to an alpha-helix in a protein with the program MMM (http:// www.epr.ethz.ch/software.html)

(c) Other nitroxide probes (PROXYL) with maleimido- or iodoacetamido-functional groups are available from Sigma-Aldrich (Fig. 1). They are soluble in DMSO and stored as 100 mM stock solutions at
80  C in the dark. (d) In some cases, reduction-resistant Gd(III)-based spin labels may me preferred over traditional nitroxides. The most common type is Gd(III)-maleimido-DOTA (Fig. 1). The maleimido-DOTA chelator can be purchased from Macrocyclics (Plano, Texas, US) or CheMatech (Dijon, France) and GdCl3 from Sigma-Aldrich. Additionally, Xylenol Orange (Sigma-Aldrich) dye is needed for determination of free Gd3+ in solution when loading the chelator to prepare the stock solution. 3. Deuterated glycerol and water low-temperature pulse experiments.

(Sigma-Aldrich)

for

4. Liquid nitrogen is at least required for shock-freezing the protein samples and low temperature CW EPR measurements.

Spin labeling EPR on membrane proteins

497

5. EPR spectrometers and accessories: (a) X- and W-band continuous wave EPR spectrometers (Jeol Ltd., Tokyo, Japan; Magnettech GmbH, Berlin, Germany; Bruker BioSpin, Karlsruhe, Germany). (b) X- and Q-band pulse EPR spectrometer (Bruker BioSpin, Karlsruhe, Germany). (c) For X-band ODNP, no commercial setup yet exists, and one needs to combine NMR and EPR consoles (from different distributors), have an EPR resonator with ENDOR coils, or with an independent NMR coil to be inserted, and eventually an additional source of microwave enabling high power (up to 5–10 W). 6. Tubes and capillaries: (a) Glass tubes (0.9 mm outer diameter) for continuous wave room temperature measurements (for this purpose use 50 μl micropipettes closed with a Bunsen from Brand GMBH, Wertheim, Germany). (b) 3 mm outer diameter quartz tubes for low temperature measurements (Wilmad-Labglass, Vineland, NJ, USA or Aachener Quarz-Glas Technologie Heinrich, Aachen, Germany). (c) TPX (polymethylpentene) capillaries for accessibility measurements (Bruker BioSpin, Karlsruhe, Germany; Molecular Specialities, Inc. Milwaukee, WI, USA). (d) W-band capillaries with o.d. of 0.9 mm (WilmadLabglass, Vineland, NJ, USA) are used for low temperature CW EPR measurements; and capillaries with i.d. of 0.5 and 0.6 (VitroCom, NJ, U.S.A.) are used for ODNP measurements with a homemade setup. 7. Dewars for liquid nitrogen. 2.2 Technical Requirements for Continuous Wave and Pulse EPR Measurements

Sufficient conditions for the detection of reliable continuous wave (CW) nitroxide spectra either at room or at low temperature are: 1. Room temperature EPR spectra are detected on 10–20 μl of sample inserted in glass tubes (0.9 mm outer diameter) with a sweep width of 15 mT with 0.2 mT/s recording speed, a 0.15 mT maximal modulation amplitude (corresponding to the natural line-width of a freely tumbling MTSL in water), 0.5–1 mW microwave power. In general, the power saturation curve must be measured for a given spin label at the working temperature at least once for each cavity in use in order to determine the optimal power to avoid signal saturation. Resonators with high quality factor are preferable (e.g., super high Q cavity from Bruker) for all CW EPR measurements. For other spin labels, in general the maximal modulation amplitude

498

Enrica Bordignon et al.

is defined by 1/3 of peak-to-peak linewidth and an optimal sweep width should be adjusted for each spin type. 2. Low temperature CW EPR spectra are detected at 160 K on 20–30 μl of sample inserted in quartz tubes (3–3.8 mm outer diameter) with a sweep width of 25 mT, a 0.25 mT maximal modulation amplitude, 0.2 mW microwave power for nitroxides (optimal power should be obtained by a power saturation curve). For low temperature W-band spectra, 1 μl of sample is inserted in quartz capillaries (0.9 mm outer diameter, 0.5 mm inner diameter) and the spectra are detected with a sweep width of 40 mT, 0.25 mT modulation amplitude (due to the broad spectral linewidth), 5 μW microwave power (to avoid saturation at 160 K, the power saturation curve must be measured to avoid line-shape artifacts prior to the high field experiments). Resonators with high quality factor are preferable as well (e.g., super high Q cavity from Bruker for X band and standard cavity for W band). Pulse measurements require specific characteristics from the hardware. Quite general, broad bandwidth and high microwave power are the most essential requirements. Additionally, the filling factor of the resonator (the amount of sample relative to the active region of the resonator) is important as it directly influences the experimental sensitivity. 1. For distance measurements with the double electron–electron resonance (DEER) experiment (see Subheading 3.4.2), the optimal setup is achieved with a Q-band spectrometer equipped with a microwave TWT amplifier delivering 150 W and an external ELDOR source or an arbitrary waveform generator (AWG). The best performance in DEER experiments is achieved when an AWG with large bandwidth is available (the Bruker SpinJet-AWG commercially available for the E580 spectrometer has currently only 400 MHz bandwidth), which provides an infinite number of independent channels with tunable frequencies, phases and powers. Notably, the new Q-band Bruker bridges consists solely of a digital source at 9.65 GHz with a tripled 8 GHz local oscillator and a Bruker SpinJet-AWG (400 MHz bandwidth, 0.625 ns time resolution, clock 1.6 GS/s) replacing the common SPFU/MPFU channels. Otherwise, at least four independent (in terms of power and phase) MPFU pulse channels should be available: one for the pump pulse (ELDOR channel) and the rest three for the observer pulses. However, DEER is still possible when two phase-locked stripline channels (SPFU) and one ELDOR channel are only available. Among the commercial resonators, the best sensitivity is obtained with the Bruker Q-band resonator accepting 3 or 1.8 mm tubes;

Spin labeling EPR on membrane proteins

499

optimal performance is achieved with homemade resonators accepting 3 mm outer diameter tubes (ETH Zurich, Jeschke group). Finally, a cryostat for the low temperature operation (50 K for nitroxide, 10 K for gadolinium) is required. 2. For Electron Spin Echo Envelope Modulation (ESEEM) measurements (Subheading 3.3) a pulse spectrometer operating at a single microwave frequency suffices. Requirements to the resonators are not that strict as those for DEER. One still achieves optimum performance with the Bruker split-ring MS3, but dielectric MD4 (max. Tube diameter 4 mm) or MD5 (max. Tube diameter 5 mm) resonators are suitable as well. A cryostat for the low temperature operation is required.

3 3.1

Methods Spin Labeling

The most commonly used spin labels for membrane proteins are commercially available pyrroline-type nitroxide radicals (Fig. 1): iodacetamido proxyl label, maleimido proxyl and methanethiosulfonate spin label (known as MTSL or MTSSL) [1]. Pyrroline-type nitroxide radicals are more resistant to reducing stress than the pyrrolidine-type analogous systems [2, 3], which can also be found commercially. Additionally, biocompatible nitroxide probes based on gem-diethyl pyrroline structures became recently available, which open the possibility to use them in cellular context [4–6] together with the reduction-resistant Gd(III)-based spin labels, which were already successfully used for in cell [7, 8]. Since the Gd-based labels cannot be purchased directly as a “ready-touse” labels, the protocol for preparation of the most commonly used Gd(III)-maleimido-DOTA from commercially available components will be given later in this chapter. MTSL binds specifically and reversibly to cysteines (addition of reducing agents will destroy the disulfide bond, removing the label from the protein). Iodoacetamido and maleimido labels are irreversibly bound to cysteines, and if present in limiting quantities and slightly alkaline pH, cysteine modification will be the exclusive reaction. However, iodoacetamido spin labels in nonbuffered solutions can also alkylate amines (lysine, N-termini), thioethers (methionine), imidazoles (histidine), and carboxylates (aspartate, glutamate, acidic C-termini), and maleimido spin labels were found to be able to bind to lysines and arginines. Iodacetamido and MTSL are flexible and do not affect protein structures at most sites. In contrast, the bulkier maleimido group may cause problems at specific sites in proteins and can undergo hydrolysis reactions at high pH (see Note 1). To disentangle the overall motion of the label from the protein backbone motion, more rigid spin labels can be employed. As an

500

Enrica Bordignon et al.

example the bifunctional MTSL reagent which binds to two consecutive cys residues at position i, i + 3 (or i + 4) in a helix has been shown to report on the backbone dynamics of α-helical proteins [9] (see Note 2). The most commonly used spin label for membrane proteins is MTSL, due to its extreme specificity for the free thiol of cysteines, a stoichiometric reaction at most sites, and the relatively small size and flexibility of the modified side chain. The labeling strategies for membrane proteins are similar to those applicable to water-soluble proteins. However, more care must be taken in identifying sites which are accessible for the spin label. In particular, sites located in the middle of the transmembrane region of the protein may not be labeled at all, or once labeled may influence the protein stability (see Note 3). In the following, a recipe for spin labeling membrane proteins with MTSL is given, which can be, however, used as guideline for the labeling with other nitroxide spin labels or with Gd(III)-maleimido-DOTA. 1. Experimentally, the spin labeling procedure is carried out incubating overnight at 4  C detergent-solubilized proteins (10–30 μM) with tenfold molar excess of MTSL under gentle shaking. To avoid biradical formation the MTSL concentration must be 90%), the remaining free Gd3+ ions should be “neutralized” with Ethylenediaminetetraacetic acid (EDTA) to prevent it to bind nonspecifically to the protein under investigation. Since maleimide functional groups are unstable in aqueous environment, small (e.g., 10 μl) aliquots of Gd(III)-maleimido-DOTA should be immediately frozen in liquid nitrogen and stored at
80  C until use. 3.2 Spin Label Dynamics Extracted at Room Temperature

EPR spectra of spin labels attached at protein sites are sensitive to even minor changes in the reorientational motion of the nitroxide spin label with respect to the external magnetic field. Reorientational motions of spin labels that are fast on the EPR timescale effectively average the anisotropic magnetic interactions leading to motionally narrowed spectral features. At X band, this corresponds to correlation times less than 1 ns. Slower motions (5–100 ns) which are characteristic of buried sites in proteins and of global tumbling of proteins in membranes, lead to increased hyperfine splittings and broader spectral linewidths. In intermediate regimes the partial averaging of the anisotropic properties leads to a gradual change in the spectral features (Fig. 2) (see Note 6). The reorientation of the spin label in the protein is strictly anisotropic being a complex function of the spin label molecular structure (e.g., MTSL is characterized by five rotatable bonds, see Fig. 1) and the primary, secondary, tertiary, and eventually quaternary structure of the protein under investigation. Often complex

Enrica Bordignon et al.

a

a

b

b

c

c

d

d

2 mT

B

C

2aiso

D

a b c

gxx gyy

d 2 mT

6 4 2

gzz

Axx Ayy

Hpp

0.2 mT

2Azz

Hpp [mT -1]

X band

A

W band

502

Azz

a

b

c

d

Fig. 2 Dynamics of the spin-labeled side chain encoded in the EPR spectra. (a) X-band derivative EPR spectra simulated with Easyspin (a, routine “garlic,” 0.1 ns rotational correlation time; b, routine “chili,” 1 ns; c, routine “chili,” 5 ns; d, routine “pepper,” rigid limit). The relation between aiso or Azz and the outer peaks separation is presented in spectra a and d. (b) Analogous W-band derivative EPR spectra simulated with Easyspin. The parameters of the g and A tensors are depicted on the spectra. (c) Extraction of the ΔHpp from the central line of the X-band spectra (denoted by the grey bar). (d) Plot of the inverse of the ΔHpp for the four spectra simulated

distributions of motional states can be observed in spectra exhibiting more than one spectral component. For a semiquantitative description of spin label dynamics, a simple empirical “mobility” parameter is extracted from the spectra at room temperature. This parameter is useful to obtain insights in the protein secondary structure, based on the available correlation between mobility and protein structure obtained from a large number of studies, mainly on α-helical proteins (see Note 7). The most commonly used mobility parameter is the inverse of the central EPR line-width (ΔHpp
1), which is an increasing function of the mobility of the spin label [12] (Fig. 2c, d) (see Note 8). One example of the use of the mobility parameter to address the conformational change of the human apoptotic Bcl-2 protein Bax from the inactive water-soluble to the oligomeric active membrane-embedded state is shown in Fig. 3 [13]. Wild type Bax contains two accessible natural cysteines at positions 62 and 126, which were labeled without additional mutagenesis, with a spin labeling efficiency close to 90%. Figure 3a shows the intensitynormalized EPR spectra detected at room temperature in the two protein conformations (see Note 9).

Spin labeling EPR on membrane proteins

A

B

BaxWT-R1

5

Hpp [mT -1] 1 mT

solution reconstituted

4 3 2

sol

C spectral fraction of monomeric Bax

503

rec

100% BaxWT-R1 + liposomes

50%

BaxWT-R1 + cBid + liposomes

2 mT

100% 0% 0%

1

10

100

1000

time [min]

Fig. 3 Bax membrane insertion followed by changes in spin-labeled side chain. (a) X-band intensity-normalized derivative spectra of wild type Bax in solution (black) and reconstituted in liposomes (grey). (b) Mobility parameters extracted from the two spectra presented in panel a. (c) Kinetics of Bax conformational changes at 37  C. Spectra of double-labeled Bax were recorded each 43 s at 37  C with (grey) and without (black) cBid as triggering agent. The intensity of the central EPR line versus incubation time (logarithmic scale) is plotted. The inset in the right shows the spectra at time zero (black dotted) and at the end of the incubation time

The two mobility parameters extracted from the spectra are shown in Fig. 3b. The time course of the slow conformational changes induced by the membrane could be followed at 37  C by detecting the EPR spectra over time (e.g., using the 2D CW experiment in the Bruker software Xepr). The decrease of the spectral amplitude over time denotes the broadening of the spectrum due to the protein conformational change. The intensity of the central line plotted vs. time is presented in Fig. 3c. The spectra detected at time zero (just after insertion of the tube in the spectrometer), characteristic of 100% of the protein in the inactive soluble state, and after 6 h of incubation at 37  C, characteristic of the membrane-embedded state, are presented on the right panel. Analysis of the intensity decay gives the half time of the membrane insertion and conformational change of the protein. Addition of

504

Enrica Bordignon et al.

cleaved Bid, a Bcl-2 protein which triggers Bax insertion, resulted in a faster insertion of spin-labeled Bax, which could clearly be detected by EPR. 3.3 Water Accessibility: Topology of Membrane Proteins

The protein topology with respect to the water and membrane interfaces is an important aspect in membrane protein studies and it is not directly addressable by X-ray crystallography. The three different types of possible side chain environments in a membrane protein, namely the protein neighboring residues, the bulk water or the lipid alkyl chains can be addressed selectively by accessibility measurements in EPR. Several methods are available to obtain such information on nitroxide-labeled membrane proteins both at room and cryogenic temperatures. This section will present the possible methods, focusing on high field, 3p-ESEEM and Overhauser dynamic nuclear polarization (ODNP). The most common EPR technique to obtain information on side chain accessibility at room temperature is power saturation CW EPR [14, 15], which consists in detecting spectra at increasing microwave power to induce saturation, a method which was already detailed in [16]. The amplitude of the central line is plotted versus the square root of the incident microwave power, and the saturation curve is analyzed to obtain the empirical P1/2 value which is inversely proportional to the product of the effective relaxation times T1 and T2. The saturation behavior is characteristic of the local environment of the spin label and can be modified by addition of paramagnetic species soluble in water (chromium oxalate CrOX or NiEDDA) or preferentially partitioning in membrane bilayers (O2) (see Note 10). A second technique consists in the direct measure of the changes induced by paramagnetic agents in the spin-lattice relaxation T1 by saturation recovery, a method detailed in [17]. In this case an intense saturation pulse of microwave is delivered at room temperature at a frequency corresponding to the central EPR line and the return of the spectral intensity is monitored with a weak CW observing microwave field. The advantage of this method is the possibility to selectively measure T1 at room temperature on different spectral components, which are usually present in the spectra of spin-labeled membrane proteins (see Note 11). An example of the application of saturation recovery to proteins to extract accessibility information can be found in [18]. A third approach allows extraction of the polarity (electric fields created by charges or electric dipole moments of the water molecules) and proticity (propensity to form H-bonds) of the nitroxide microenvironment by analysis of the EPR spectra in the rigid limit [19–21]. For membrane proteins in water–glycerol mixture, a temperature of 160 K, although not being exactly the rigid limit for nitroxides, can be safely used. The Azz principal value of the 14N hyperfine tensor (half of the splitting between the positive low field

Spin labeling EPR on membrane proteins

505

peak and the negative high field peak of the X-band CW spectrum, Fig. 2) is proportional to the polarity of the nitroxide microenvironment (for MTSL the Azz value can change from 3.3 to 3.7 mT going from an apolar to a polar environment). The Azz parameter can be already extracted from X band CW EPR spectra. This information can be complemented by detection of the propensity of the nitroxide to form H-bond (proticity) via the gxx principal value of the g tensor (for MTSL the gxx value can change from 2.0089 to 2.0083 mT going from an apolar-aprotic to a polarprotic environment) (Fig. 2). Measurements of gxx require the continuous wave detection of the low temperature nitroxide spectrum at W band (3.4 T/95 GHz). A detailed description of the method to detect W-band spectra and extract the polarity parameters is given in the following. 1. Singly labeled membrane proteins at an optimal concentration of 100 μM (in detergent or liposomes) containing 10% v/v of glycerol as cryoprotectant are prepared and 1 μl is inserted in W-band quartz capillaries (0.9 mm outer, 0.5 mm inner diameter) using for example a thin Pasteur pipette made with a Bunsen burner. The height of the sample in the tube is about 5 mm, enough to fill the W band cavity. 2. Prepare the spectrometer (e.g., Bruker Elexsys E680 spectrometer equipped with a W-band probehead) in “tune mode” and position the microwave dip in the center of the screen. 3. The capillary is shock frozen in liquid nitrogen and inserted in the W-band cavity precooled at 160 K in a helium- or nitrogenflow cryostat. After insertion of the capillary, the resonant frequency moves to lower values. Follow it and position the dip again in the middle of the screen (see Note 12). 4. The temperature of the cavity is allowed to equilibrate for at least 30 min. Reposition eventually the dip in the center of the screen. The equilibrium can also be judged by the invariance of the dip position and shape (see Note 13). 5. Before recording a power saturation curve must be measured at 160 K at least for one mutant of the spin-labeled protein under investigation to obtain the highest microwave power which can be used to still be in the linear regime. Additionally, the phase has to be properly adjusted. Usually 5 μW incident microwave power, 100 kHz field modulation, 0.25 mT modulation amplitude, 40 mT sweep width are good experimental parameters. 6. A first estimation of the gxx and Azz values, diagnostic for polarity and water penetration into the membrane, can be directly obtained from the spectra (Fig. 2). The fit of the spectra can provide more exact values of the parameters (e.g., using the Easyspin routine “pepper” [22] and esfit, see Note 14). A starting g-tensor for MTSL is (2.0085, 2.0061,

506

Enrica Bordignon et al.

2.0021), and starting A-tensor in MHz is (14, 14, 98). Calibration of the B field is mandatory. The gxx region of the spectrum is the most complicated to fit. Due to gxx-strain, there can be significant broadening of the spectral features, which needs to be taken into account in the simulation. Moreover, heterogeneities in the H-bonding may lead to the appearance of shoulders. In this case the fit is performed with two (max. three) gxx components. A detailed study of this heterogeneity can be found in [23]. 7. The values obtained can be inserted in a plot of gxx vs. Azz to obtain information on the relative location of each spin-labeled site with respect to known values. A comprehensive gxx vs. Azz plot containing data from spin labels in organic solvents and from different proteins spin labeled with MTSL and can be found in [20, 23]. An example of a polarity-proticity profile detected in the membrane protein bacteriorhodopsin is presented in Fig. 4a. Spin labels were placed in different position through the proton channel from the periplasmic to the cytoplasmic side of the protein. A clear water penetration barrier in the middle of the protein could be detected by both Azz and gxx (highlighted in red). This method is used to unravel relative polarities of different spin-labeled sites in a membrane protein, or to detect changes in the water accessibility due to protein conformational changes. A fourth approach is to detect the accessibility of the nitroxide toward deuterated water molecules by comparing the amplitude of the 3-pulse ESEEM (electron spin echo envelope modulation) oscillations detected at X-band at cryogenic temperatures [24]. For deuterium ESEEM measurements on spin-labeled membrane protein embedded in liposomes or solubilized in detergent, one needs to deuterate the buffer up to more than 90%. Relative deuterium accessibility parameters for different side chains in a protein can be obtained using the amplitude of a nitroxide in the deuterated buffer solution as a reference. An example of the water accessibility derived by 3-pulse ESEEM for two spin-labeled side chains in the light harvesting complex II compared to a watersoluble reference is presented in Fig. 4b [24]. The advantage of the method is that deuterium ESEEM modulations provide a direct measure of the accessibility of the spin label side chain to the aqueous buffer. In addition, data are intrinsically well reproducible as the depth of the modulations is not very sensitive to the microwave field conditions once resonator is overcoupled. The disadvantages are: the use of cryogenic temperatures, the difficulty to distinguish protein buried from membrane-embedded sites and to disentangle the effects arising from deuterium atoms from water, exchangeable protons, or deuterated glycerol (the latter is used as a cryoprotectant).

Spin labeling EPR on membrane proteins

A

g xx

507

162

Azz

166

100

167

162 166 163

167

162 100 46

46

100

100 171

170

46 93

53

46

89

53

53

53 204

129

129

3.40

3.41

3.42

B field [T]

194

129

129

less accessible 3.5 3.6 3.7 AZZ [mT]

86 85 84 83

(gxx- 2)×104

14

B

V(t) [a.u.]

more accessible

52

water-soluble reference

12 10

52

8 6

196 0

2

4

t [ s]

6

8

12

16

196

I [a.u.]

20

52

10

196

less accessible 0

0

4

8

[MHz]

Fig. 4 Low temperature methods to extract polarity and water accessibility. (a) W-band rigid limit spectra of BR (PDB 1MOL) spin-labeled in the proton channel. The changes in gxx and Azz are highlighted by vertical lines. Right panel, plots of gxx and Azz versus nitroxide position showing the hydrophobic barrier in the channel. (b) Structure of LHCIIb (PDB 2BHW) with two iodoacetamido spin labels at positions 52 and 196 with MMM. Right panel, 3-pulse ESEEM traces detected on a water-soluble standard and on the two spin-labeled positions. Bottom, ESEEM modulation depths correlated with the water accessibility

508

Enrica Bordignon et al.

The last method which we want to highlight is X-band room temperature Overhauser dynamic nuclear polarization [25]. The setup is not available commercially and it requires EPR and NMR detection hardware and an RF coil inserted into an EPR resonator. This method allows monitoring the amount of water molecules approaching the nitroxide spin label at physiological temperatures, providing a direct qualitative measure of the accessibility of the spin-labeled side chain to the aqueous buffer [26, 27]. In addition, based on modeling approaches, the site-specific translation diffusivity of the water molecules around the spin label can be inferred extrapolating the ODNP enhancement at infinite microwave power (see for example [28, 29]). To obtain a qualitative parameter of water accessibility of a spinlabeled site in a membrane protein, one can perform an ODNP experiment using a conventional EPR bridge delivering up to 200 mW power. However, dedicated microwave source must be used for extrapolation of the enhancement at infinite power (Fig. 5a). Briefly, in an ODNP experiment the NMR signal of the water protons is measured as a function of increasing microwave power, which is applied continuously at the central line of the CW EPR spectrum of a nitroxide (Fig. 5a). The proton NMR signal is only enhanced if electron spins are present in solution as it shown in Fig. 5a. The ODNP enhancement depends on the concentration of the electron spin, its rotational dynamics and its water accessibility. Ideally, with similar spin label dynamics, the bigger the enhancement at a given electron spin concentration, the higher the water accessibility at that site. An example of water accessibility changes derived qualitatively from ODNP experiments is presented in Fig. 5b for the vitamin B12 ABC importer. The periplasmic gate is shown to decrease its water accessibility upon binding of BtuF, and concomitantly the cytoplasmic gate becomes more water exposed [26]. 3.4 Methods Based on Interspin Distances Determination

Conformational changes of membrane proteins can be followed by monitoring distances between selected labeled sites on a protein [30]. The existing CW and pulse EPR methods rely on extracting the dipolar coupling between two electron spins to obtain the distance distributions. There are common limitations for the CW and pulse EPR methods of distance determination. Generally, in addition to the dipolar (anisotropic) part, coupling between the spins contains an isotropic contribution known as Heisenberg exchange coupling. Since no general quantitative treatment of the distance dependence of the exchange coupling exists yet, distances at which exchange coupling is significant ( 34 dBm (highlighted by a colored area). (B) ODNP enhancement (parameter ϵ + 1) for detergent-solubilized BtuCD (black triangles) and BtuCD-F (red triangles) with a spin label at position 141 in the cytoplasmic gate or at position 168 in the periplasmic gate [26]. Binding of BtuF decreases water accessibility at the periplasmic gate and increases it at the cytoplasmic gate (see arrows), as seen from the opposite ODNP effects

dipolar tensor. In contrast, in the limiting case of fast tumbling (e.g., at room or higher temperatures), the dipolar interaction is completely averaged on the EPR time scale thus it vanishes from the experimental spectra. In the intermediate situations (spin labels with restricted molecular motions on membrane proteins) only

510

Enrica Bordignon et al.

residual dipolar interaction is visible. Therefore, determination of distances from room temperature spectra ignoring motional effects may be erroneous. 3.4.1 Interspin Distances at Cryogenic Temperatures Detected by CW EPR

This method is used when the magnitude of the dipolar coupling exceeds the natural linewidth of the nitroxide CW EPR spectrum and can therefore be separated from other interactions by lineshape analysis. Usually this is the case at interspin distances