the effects of urea and of ph on protein structure · the effects of urea and of ph on protein...

University of Groningen

The effects of urea and of pH on protein structure studies by molecular dynamics simulationMueller, Daniela Sung-Mi

IMPORTANT NOTE: You are advised to consult the publisher's version (publisher's PDF) if you wish to cite fromit. Please check the document version below.

Document VersionPublisher's PDF, also known as Version of record

Publication date:2010

Link to publication in University of Groningen/UMCG research database

Citation for published version (APA):Mueller, D. S-M. (2010). The effects of urea and of pH on protein structure studies by molecular dynamicssimulation Groningen: s.n.

CopyrightOther than for strictly personal use, it is not permitted to download or to forward/distribute the text or part of it without the consent of theauthor(s) and/or copyright holder(s), unless the work is under an open content license (like Creative Commons).

Take-down policyIf you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediatelyand investigate your claim.

Downloaded from the University of Groningen/UMCG research database (Pure): http://www.rug.nl/research/portal. For technical reasons thenumber of authors shown on this cover page is limited to 10 maximum.

Download date: 07-06-2018

https://www.rug.nl/research/portal/en/publications/the-effects-of-urea-and-of-ph-on-protein-structure-studies-by-molecular-dynamics-simulation(a3f1d0fa-444a-4a4a-a5aa-8b68bf336462).html

The effects of urea and of pHon protein structure

studied by molecular dynamics simulation

Daniela S. Mueller

Author: Daniela S. Mueller

This study was carried out in the Molecular Dynamics group, Groningen BiomolecularSciences and Biotechnology Institute, Faculty of Mathematics and Natural Sciences,University of Groningen, Netherlands.

Copyright c© 2010 Daniela S. Mueller

ISBN 978-90-367-4391-4

Cover design: Daniela S. Mueller

The images in this book and the cover design were prepared with the following softwareprogrammes: GIMP, Grace, OpenOffice, Plot, POV-Ray, VMD [Humphrey et al. 1996].

RIJKSUNIVERSITEIT GRONINGEN

The effects of urea and of pH on protein structurestudied by molecular dynamics simulation

Proefschrift

ter verkrijging van het doctoraat in de

Wiskunde en Natuurwetenschappen

aan de Rijksuniversiteit Groningen

op gezag van de

Rector Magnificus, dr. F. Zwarts,

in het openbaar te verdedigen op

vrijdag 18 juni 2010

om 13.15 uur

door

Daniela Sung-Mi Müllergeboren op 3 december 1975

te Lünen, Duitsland

Promotor: Prof. dr. A.E. Mark

Beoordelingscommissie: Prof. dr. Bauke W. DijkstraPD dr. Jürgen SchlitterProf. dr. Martin Zacharias

to the memory of

my beloved grandmother Maria Müller

and

Axel Pawellek (1972–2007)

ForewordMost of the people who claim that no one ever reads dissertations have written one them-selves. I suspect it’s a form of nonchalance — which can be very trying on those who havenothing yet to be nonchalant about. Or maybe our scholars are trying to stir up a revolution tomake dissertations obsolete. Personally, I would not want to miss the experience of writingand publishing this thesis. And I hope that this book will be observed and received at leasta little, which would make up for the effort put into writing it. I thank the Faculty of Math-ematics and Natural Sciences and the University of Groningen for their generous financialsupport for the printing of this book. I wrote this thesis in the hope that it may be useful. Notto become rich, poor, old, or trying to give the world a text of germanicised English, as somepeople might think.

That said, this thesis is likely to be useful only to a few people. Therefore I would liketo direct the generally interested reader to the summaries at the end of the book. These aretargeted at laypeople and were written in language that should be understandable to most.There is an English, a Dutch and a German version. Next the outlook on future research, justbefore the bibliography, may be of general interest. To keep it understandable I tried to writethe outlook section in general, non-scientific English. Have a look at the figures, in particularthe colourful images of the proteins. If you are interested to learn more, the next step wouldbe to read the general introduction to the thesis and the introductions to the chapters. If nowyou want to learn about the outcomes of the studies, these are summarised in the conclusionat the end of each chapter. Then you might think: This is not making any sense to me at all,I’ll (close the book and) admire the cover for a bit; or: I wish this made more sense to me,I’d better check out the details of this study. Either way, I hope you will enjoy reading — orpeeping into — my book.

Groningen, May 2010

i

Contents

Foreword i

List of Tables vii

List of Figures ix

I Introduction x

1 Introduction 1

1.1 Fold — un-fold . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

1.2 Molecular dynamics simulation . . . . . . . . . . . . . . . . . . . . . . . . 4

II Partial unfolding of bacterial cytochrome c by urea 7

2 Partial unfolding of bacterial cytochrome c by urea 9

2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

2.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

2.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

2.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

2.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

iii

CONTENTS

III Flaviviral envelope glycoprotein E 31

3 Introduction to flaviviral envelope glycoprotein E 31

4 The role of histidine residues in low-pH-mediated viral membrane fusion 37

4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

4.2 Activation of class I fusion proteins . . . . . . . . . . . . . . . . . . . . . 42

4.3 Activation of class II fusion proteins . . . . . . . . . . . . . . . . . . . . . 43

4.4 Environment of histidine residues . . . . . . . . . . . . . . . . . . . . . . 46

4.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

5 The effect of histidine protonation on the dengue viral envelope protein ectodomainsE 49

5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49

5.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54

5.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57

The pre-fusion and the post-fusion conformation . . . . . . . . . . . . . . 57

Subunit interfaces in the dimer and the trimer . . . . . . . . . . . . 58

Domain III interface . . . . . . . . . . . . . . . . . . . . . . . . . 60

pKaaa and solvent accessibility of the histidine residues . . . . . . . . 61

The membrane domains of E and M . . . . . . . . . . . . . . . . . . . . . 67

The prM-E complex . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67

Conservation of the histidine residues and their micro-environments . . . . 72

MD simulation of the sE protein . . . . . . . . . . . . . . . . . . . . . . . 73

Electrostatic interactions . . . . . . . . . . . . . . . . . . . . . . . 73

Number of contacts at the subunit and domain interfaces . . . . . . 78

Structural changes during the MD simulations . . . . . . . . . . . . 79

pKaaa and solvent accessibility of the histidine residues . . . . . . . . 83

Restructuring of the E dimer . . . . . . . . . . . . . . . . . . . . . 84

5.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89

The pre-fusion subunit interface . . . . . . . . . . . . . . . . . . . . . . . 90

iv

The interface of domain III . . . . . . . . . . . . . . . . . . . . . . . . . . 92

The membrane domains . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94

Intra-domain interactions . . . . . . . . . . . . . . . . . . . . . . . . . . . 95

pKa and solvent accessibility of the histidine residues . . . . . . . . . . . . 96

The prM-E complex . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98

Conservation of the histidine residues and their micro-environments . . . . 99

Restructuring and dimer disassembly . . . . . . . . . . . . . . . . . . . . . 99

5.5 Summary and conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . 101

6 Model for the activation of flaviviral fusion proteins 103

6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103

6.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106

6.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108

The low-pH-dependent conformational change of the E protein . . . . . . . 108

Interaction between the ectodomain and the stem-anchor region . . . . . . 109

Structural intermediate in the conformational change of the full-length E pro-tein . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111

6.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115

The low-pH-dependent conformational change of the E protein . . . . . . . 115

Interaction between the ectodomain and the stem-anchor region . . . . . . 116

Model for flaviviral membrane fusion mediated by dimeric E protein . . . . 117

6.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121

Bibliography 139

Outlook 141

Summary 143

Samenvatting 144

v

CONTENTS

Zusammenfassung 145

Acknowledgments 147

CV 151

vi

List of Tables

2.1 The simulation systems of cytochrome c-550 (cytc) in water and urea . . . 14

2.2 Secondary structure of cytc in water and urea . . . . . . . . . . . . . . . . 18

4.1 Sequence alignments of A) influenza hemagglutinin HA, and B) flaviviralenvelope protein sE sequences . . . . . . . . . . . . . . . . . . . . . . . . 44

5.1 Hydrogen bonds and salt bridges across the subunit and the domain III inter-face of the dengue viral sE protein in the pre-fusion form . . . . . . . . . . 59

5.2 Predicted pKa of protonation of the histidine residues in the dengue viral sEprotein . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61

5.3 Solvent accessible surface areas of the histidine residues in the sE protein . 64

5.4 Predicted pKa of protonation of the histidines in the prM-E complex . . . . 70

5.5 Conservation of the histidine residues in flaviviral sE proteins . . . . . . . . 73

5.6 Salt bridges and hydrogen bonds in the crystal structures and in the MDsimulation of the dengue viral sE protein . . . . . . . . . . . . . . . . . . . 76

5.7 Number of contacts and contact atoms at the subunit interface of the sE dimer 78

5.8 Root mean square deviation (RMSD) of the domains of the dengue viral sEprotein in the MD simulations . . . . . . . . . . . . . . . . . . . . . . . . 81

vii

List of Figures

2.1 Crystal structure of bacterial cytochrome c-550 (cytc) . . . . . . . . . . . . 10

2.2 Root mean square deviation (RMSD) of the backbone atoms of cytc duringMD simulation in water and urea . . . . . . . . . . . . . . . . . . . . . . . 17

2.3 Cumulative number (CN) of the radial distribution function (RDF), and RDFof the cytc atoms with respect to the heme-iron . . . . . . . . . . . . . . . 19

2.4 Protein environment and solvation of the heme group in cytc . . . . . . . . 21

2.5 Opening of the cytc heme cavity in 10 M urea . . . . . . . . . . . . . . . . 23

2.6 Conformation of cytc in 10 M urea . . . . . . . . . . . . . . . . . . . . . . 23

2.7 CN of the RDF, and RDF of solvent species with respect to cytc . . . . . . 24

2.8 Hydrogen bonding between Lys99 and the heme group . . . . . . . . . . . 25

3.1 Flaviviral membrane fusion . . . . . . . . . . . . . . . . . . . . . . . . . . 32

3.2 X-ray crystal structures of the dengue viral sE protein . . . . . . . . . . . . 33

3.3 The pre-fusion conformation of the dengue viral sE protein and the positionsof the histidines in the dimer . . . . . . . . . . . . . . . . . . . . . . . . . 34

4.1 Pre-fusion and post-fusion structures of influenza HA and dengue viral Eprotein . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40

4.2 Schematic diagram of the interactions of a histidine residue in a viral fusionprotein in the pre-fusion and the post-fusion state. . . . . . . . . . . . . . . 41

5.1 Titration curves of histidine protonation and of pH-dependent fusion of cellsinfected with dengue virus . . . . . . . . . . . . . . . . . . . . . . . . . . 51

5.2 Model of the flavivirus maturation pathway . . . . . . . . . . . . . . . . . 52

viii

5.3 Ectodomain and membrane domains of the dengue viral envelope proteins Eand M . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53

5.4 Titration curves predicted for the histidine residues in the dengue viral sEprotein . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65

5.5 Model of low-pH-dependent, dengue viral fusion triggered by histidine pro-tonation in the E protein . . . . . . . . . . . . . . . . . . . . . . . . . . . 66

5.6 Histidine interactions in a model of the prM-E heterodimer . . . . . . . . . 69

5.7 Cryo-EM structure of the prM-E heterodimer at pH 6.0 . . . . . . . . . . . 71

5.8 Hydrogen bond His144-Asp42 during MD simulation . . . . . . . . . . . 75

5.9 Histidine protonation-dependent conformational change of the sE dimer af-ter 70 ns of MD simulation . . . . . . . . . . . . . . . . . . . . . . . . . . 79

5.10 RMSD of the sE protein during MD simulation . . . . . . . . . . . . . . . 80

5.11 RMSDx of the Cα-atoms of the sE dimer after ∼ 60 ns of simulation . . . . 82

5.12 Dengue viral sE protein after ∼ 60 ns of MD simulation . . . . . . . . . . . 85

5.13 Conformational change of the subunits of the sE dimer after histidine proto-nation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86

5.14 Contact regions between the subunits in the sE dimer . . . . . . . . . . . . 87

5.15 Histidine residues at the protein surface of the sE dimer after histidine pro-tonation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88

6.1 Models of the mature dimer of the full-length E protein . . . . . . . . . . . 105

6.2 Model of the transition between the pre-fusion and the post-fusion confor-mation of the E protein . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109

6.3 Conformational hybrid model of the E protein dimer . . . . . . . . . . . . 110

6.4 Membrane-facing surface of the full-length E protein . . . . . . . . . . . . 111

6.5 Sequence similarity among 28 flaviviral E proteins . . . . . . . . . . . . . 112

6.6 Surface of the ectodomain and the membrane domain of the pre-fusion Eprotein . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113

6.7 Model of an intermediate from the MD simulation of the sE protein afterhistidine protonation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114

6.8 Model of flaviviral membrane fusion mediated by dimeric E protein . . . . 119

Introduction 11.1 Fold — un-fold

Proteins are compounds of tremendous significance for all life on this planet and are foundin all the living organisms and viruses known to us. Many proteins function as enzymes oras the building blocks of biological matter. Often the function depends on the specific three-dimensional structure of the protein. The structure of a protein is based on a linear sequenceof amino acids in a polymeric chain. The specific sequence and the spatial arrangement —topology, or fold — of a peptide chain determine its three-dimensional structure. Due to themany degrees of freedom in this system, the relationships between the sequence, structure[Levinthal 1968] and function are complex and hard to predict. Although numerous proteinstructures have been experimentally solved, understanding how sequence leads to structureand thus to function remains one of the most fundamental challenges in biology.

The folding of a protein chain is a dynamic process that takes place in all living cellsunder so-called native conditions. Under non-native conditions, alternative folding pathwayscan be accessed that sometimes result in non-native structures with new functional proper-ties, though more often the result is unfolding, i.e. the total loss of structure. Such changes onthe molecular level can lead to systemic malfunction and disease in the organism, prominentexamples are the pathological prion and amyloid proteins. In the case of pathogens, knowl-edge of the function of the pathogenic enzymes can open ways to interfere with infection,or to cure the disease. In the quest to explore and understand the functions and propertiesof proteins, alternative structures have been discovered that are based on non-native folds ofknown sequences, e.g. the molten globule. Modified enzymes are specifically engineered forapplications in the molecular biology lab, the food, pharmaceutical and other industries.

Some physico-chemical conditions that are known to affect folding are temperature, pH,solvent or salt concentration, and complex environments and interfaces like membranes, li-

1

1.1 Fold — un-fold

posomes or chaperones. Solvents that are typically used for biochemical unfolding in the lab-oratory are alcohols, acidic, basic or alkaline solvents, and their mixtures. These have beenwidely studied in chemical experiments, and by physical methods such as spectroscopy, e.g.X-ray diffraction, circular dichroism (CD), flourescence, UV-visual (UV-VIS), Fourier trans-form infra-red (FTIR), or nuclear magnetic resonance (NMR) spectroscopy. Computationalmodelling is a further method of investigating the macroscopic properties of a compound,e.g. in Monte Carlo or molecular dynamics (MD) simulations.

This thesis studies the onset of structural changes in two native protein structures inMD simulation, addressing the following questions:

QI. Is urea a suitable denaturant for the partial unfolding of bacterial cytochrome c-550?

QII. Does the protonation of histidine residues trigger the restructuring of a pH-dependentviral fusion protein?

Partial unfolding of a bacterial peroxidase by urea

Cytochrome c is an integral component of the electron transport chain in mitochondrial andbacterial membrane systems. It is a redox-active protein carrying a heme group, the cat-alytic centre of the protein. In addition, c-type cytochromes show peroxidase activity in thepresence of hydrogen peroxide H2O2 when the heme-iron is in the ferric state Fe3+ [George1953; Harbury & Loach 1960; Ubbink et al. 1992];

In contrast to essential peroxidases, C-type cytochromes are very stable and remainhighly soluble even under conditions of extreme heat, acidity and basicity. At the sametime, the peroxidase activity of C-type cytochromes is approximately 1000-fold lower thanthe activity of, for instance, microperoxidases. This low activity was related to the limitedaccessibility of the heme pocket for peroxide substrates [Diederix et al. 2001, 2002a,b].

In experiment, the partial unfolding of bacterial cytochrome c lead to an approximately1000-fold increase in activity, suggesting an expansion of the heme pocket that allowed theperoxide substrate to access the active site more easily [Diederix et al. 2004, 2002a; Worrallet al. 2005a]. If the conformation of this peroxidase could be easily and reversibly manip-ulated, the activity could be enhanced and regenerated multiple times. In this thesis MDsimulations were used to examine the effect of the denaturant urea on the structure of a bac-terial cytochrome c-550, in order to understand the relationship between partial unfoldingand peroxidase activity.

2

1 Introduction

Low-pH dependent activation of a viral fusion protein

Another physico-chemical condition that is known to influence protein folding and structureis pH. Some proteins require acidic pH to obtain a specific functional fold. Such proteinsare found in a number of viruses that require low pH to infect the host cell, among them theInfluenza viruses and members of the Alphavirus and Flavivirus genera. Dengue virus is aflavivirus and the most common mosquito-borne virus that infects humans. It causes denguefever and dengue haemorrhagic fever, that can lead to death. During the last two decadesdengue has emerged globally as a major public health concern [CDC 2008; WHO 2008]. Nodrugs for the treatment of the dengue diseases or vaccines to prevent infection by the denguevirus have yet been found.

The flaviviral envelope is covered with proteins that are thought to mediate the fusion ofthe viral and the cell membrane that allows the viral RNA to access and infect the cell. Thehost cell takes up the virus through endocytosis, enclosing the virus in an endosome. There,increasing acidification triggers the fusion of the virus with the endosomal membrane. LowpH usually denatures and potentially deactivates proteins, but in the case of the flaviviral en-velope protein low pH is believed to trigger a large-scale conformational change that leads toits activation for fusion [Schibli & Weissenhorn 2004; Zimmerberg et al. 1993]. A numberof models have been proposed [Gibbons et al. 2003; Harrison 2008; Helenius 1995; Poum-bourios et al. 1999], but the exact mechanism of viral protein-mediated membrane fusion arestill unknown.

However, several essential aspects of viral fusion are known that could be of particularinterest in the design of antiviral drugs. For instance i) the structural change on exposure tolow pH is irreversible, and ii) activation is quickly followed by inactivation. In other words,it is possible to prevent infection by prematurely activating the viral fusion protein beforethe virus comes into contact with the host cell [Heinz 2003; Hurrelbrink & McMinn 2001;Mandl 2005; Mandl et al. 2001; McMinn 1997].

A particularly interesting aspect of low-pH-dependent viral fusion is that the pH of fu-sion is similar to the pKa of protonation of histidine in water (pKa = 6.0). This motivatedthe hypothesis that the protonation of histidine residues might trigger the activation of theviral fusion protein [Kampmann et al. 2006]. This principle of the so-called histidine-switchmight present the link between the pH-dependency of the fusion process and a specific struc-tural mechanism of the viral envelope protein.

MD simulations of a flaviviral envelope protein were performed to elucidate the effectof histidine protonation on the conformation of a low-pH-dependent viral fusion protein. Theaim was to determine i) what effect histidine protonation has on the structure of the protein;and ii) whether these effects are related to the conformational change of the envelope protein

3

1.2 Molecular dynamics simulation

observed in experiment after acidification.


Molecular dynamics (MD) simulation is a computational method that is used to obtain adetailed view of the dynamic behaviour of molecules. This method has proven especiallyuseful in the modelling of complex biomolecular systems such as proteins, membranes ornucleic acids. In an MD simulation a deterministic time series of configurations is generatedof a compound or a system of interest, in a temporal and spatial resolution that are usuallyunattainable in real macroscopic experiments.

MD simulation is based on physical interactions on the molecular level, which definesthe accuracy of the method. Therefore the frequency of molecular vibrations sets the upperlimit for the iterative time step of computation, which lies in the order of femto-seconds.The time-scale of a process and the size of a system that can be simulated are limited by thecomputer time available, i.e. the computational resources in real time, and by the speed withwhich the calculations are processed. For the simulation of a protein in atomistic detail, asperformed in the studies presented in this thesis, several 100 nano-seconds of data could begenerated, which compares to the time-scale of the motions of a loop, or the relative motionsbetween two domains.

In an MD simulation the classical trajectory of a molecule is modelled by solving New-ton’s equations of motion [Newton 1678]

F = ma (1.1)

= m v (1.2)

for the force F, acceleration a, mass m and velocity v. Using the leap-frog algorithm, thevelocity v and position r of a particle are updated at time step intervals of ∆t according to

v(t +∆t2

) = v(t− ∆t2

)+Fm

∆t (1.3)

r(t +∆t) = r(t)+v(t +∆t2

)∆t (1.4)

which is based on the Verlet algorithm [Verlet 1967]. The particles have physical-chemicalproperties, e.g. mass and charge, and their interactions are determined by a force field V (r):

F =−∂V∂ r

(1.5)

4

1 Introduction

The particles are modelled as Lennard-Jones particles, i.e. impenetrable spheres that interactvia an attractive van der Waals potential. Chemical bonds are described by harmonic poten-tials. Interactions between charged particles are modelled via a classical coulombic electro-static potential acting between point charges. In order to lower the number of computations,non-bonded pair-interactions beyond a distance cut-off are neglected, and the long rangeelectrostatic interactions in a dielectric are approximated by a reaction field. The degreesof freedom between hydrogens that are bound to heavy atoms are considered insignificantfor the dynamics of the protein and are integrated in the so-called united-atom scheme tofurther save on the number of computations. The system is set in a periodic box to eliminateboundary artefacts, and the box dimensions are chosen such that the solute does not interactwith its periodic image. The number of particles N, the pressure p and the temperature T arekept constant.

This level of description is used to generate the classical trajectory of an isothermic-isobaric (NPT) ensemble of chemically inert molecules. It is considered suitable for themodelling of thermodynamic equilibrium processes like conformational changes and molec-ular motions. Although the results of this iterative method are in principle deterministic,the interactions in a large molecule are too complex for an analytic prediction of the states.Therefore MD simulation presents an extremely helpful tool for the study of molecules.

While the underlying models are empirical, MD simulations have successfully repro-duced experimental results [Daura et al. 1999; Güntert et al. 1997; Laasonen et al. 1993;MacKerell et al. 1998; Marrink et al. 2004; Marszalek et al. 1999; Oostenbrink et al. 2004;Rappe & Goddard III 1991; Srinivasan et al. 1998; Verlet 1967]. Often the MD trajectoryof one simulated molecular specimen is integrated to infer the thermodynamic properties ofa statistical ensemble; then the trajectory is assumed to be ergodic [von Neumann 1932]. Inthe evaluation of a simulation it is assumed that ergodicity is approximated through moreextensive sampling, e.g. in longer trajectories. Due to the limited accuracy of floating-pointoperations, the time-series obtained is usually chaotic. In practice this turns into an advan-tage with respect to the sampling of the thermodynamic states, as a chaotic trajectory willnot be caught in a periodic loop but continue to explore more phase-space with time.

While ergodicity is relevant for the thermodynamic characterisation of a system in equi-librium, many questions addressed by MD simulations do not aim at a complete thermody-namic description and may therefore neglect the ergodicity issue. For instance the simu-lations undertaken for this thesis investigate systems that are not in equilibrium. Here thequestion is whether a specific state is reached while the system equilibrates, and a positiveanswer presents a proof of concept. The above question QI is answered when the desiredstate is obtained, i.e. partially unfolded cytochrome c. QII is answered when any significantdeviation from the initial state occurs, i.e. an altered conformation of the sE protein.

5


Apart from providing results on the atomistic, single molecule level, an important ad-vantage of simulations in general is that conditions can be applied which may be dangerousor difficult to control in experiment. In addition, simulations can help find hypotheses, andthereby reduce the number of animal or human experiments, thus addressing important ethi-cal issues in bio-medical research and drug development. Therefore simulations are a usefuland indispensable alternative to handling real compounds or processes that can be hazardousto the experimentalist and the environment.

6

Partial unfolding ofbacterial cytochrome c byurea 22.1 Introduction

Cytochrome c is a phylogenetically “ancient” enzyme that is found throughout the eukaryoticand prokaryotic kingdoms. It is an integral component of the electron transport chain in mito-chondrial and bacterial membrane systems and therein essential for the generation of energyin the cell. Like many redox-active proteins c-type cytochromes carry a prosthetic group,an iron ion-coordinating heme group, which is covalently bound to two cysteine residuesand presents the catalytic centre of the protein (Fig.2.1). Cytochrome c-550 (cytc) was iso-lated from the rhodobacterium Paracoccus versutus (the genus is also known under the nameThiobacillus) [Lommen et al. 1990]; the suffix “c-550” derives from the heme type C andthe visual absorption peak at a wavelength of 550 nm. Cytc has been studied extensively,and was classified as a class I cytochrome c like the archetypal mitochondrial cytochromes cand many bacterial cytochromes [Diederix et al. 2001].

In addition to their role in electron transport, c-type cytochromes also show peroxidaseactivity in the presence of hydrogen peroxide H2O2 when the heme-iron is in the ferric stateFe3+ [George 1953; Harbury & Loach 1960; Ubbink et al. 1992]; hence they are also re-ferred to as peroxidase mimics. Peroxidases are enzymes that contain a heme group andefficiently catalyse the oxidation of substrates using the environmentally innocuous oxidanthydrogen peroxide [Diederix et al. 2001]. Whereas essential1 peroxidases are generally veryinstable, C-type cytochromes are stable and remain highly soluble even under conditions of

1 An essential peroxidase is an enzyme with the primary activity of a peroxidase.

9

2.1 Introduction

Figure 2.1 Cytochrome c-550 (cytc) from Paracoccus versutus. Cross-eyed stereo image of theX-ray crystallographic structure of cytc. The heme group, Cys15 and Cys18 are depicted in stickrepresentation. The secondary structure shown was assigned using the STRIDE software [Frishman& Argos 1995; Humphrey et al. 1996]. Colour legend: purple � α-helix, blue � 310-helix, yellow �β-sheet, cyan � turn, white � loop, orange � heme-iron.

extreme heat, acidity and basicity. Furthermore the covalent linkage of the heme prostheticgroup to the protein matrix prevents dissociation of the catalytic moiety from the protein.[Diederix et al. 2001] As a consequence the potential to use c-type cytochromes as perox-idases in commercial applications has attracted much interest. Wild-type ferricytochromec-550 has a second order rate constant for the reaction of H2O2 of k = 22.8± 0.4 M−1s−1

for hydrogen peroxide concentrations [H2O2]< 100mM [Worrall et al. 2005a], which is≈ 1000-fold lower than the activity of microperoxidases [Diederix et al. 2001; Harbury &Loach 1960]. Canters and co-workers have related this low activity to the low accessibilityof the heme pocket for potential peroxidic substrates which are too large to access the cat-alytic centre within the protein [Diederix et al. 2001, 2002a,b]. They found the peroxidaseactivity of cytc unfolded in guanidinium hydrochloride to be increased≈ 1200-fold, and thatthis increase correlated well with unfolding, indicating better accessibility of the active siteto the substrate [Diederix et al. 2002a]. From this they concluded that in partially unfoldedcytc the conformation of the protein matrix around the heme pocket was expanded, allowingperoxide substrates to access the active site more easily and to be catalysed more readily[Diederix et al. 2004, 2002a; Worrall et al. 2005a].

The structure of cytc from P. versutus was solved by X-ray crystallography (Fig. 2.1)[Worrall et al. 2005b]. As is common to all c-type cytochromes the heme group is bound tocytc by the Cys-Xaa-Xaa-Cys-His (CXXCH) heme binding sequence motif, where the sidechain thiol groups of Cys15 and Cys18 form thioether linkages with the two vinyl groups ofthe heme [Ambler 1963; Worrall et al. 2005b]. The Fe(III) ion is hexa-coordinated by thefour pyrroline N-atoms and two protein ligands, the Nτ-atom of the His19 side chain and the

10

2 Unfolding of cytochrome c-550 in urea

Sδ-atom of Met100 [Dickerson et al. 1971; Gadsby et al. 1987][IUPAC 1975]. The alkalineunfolding of cytochromes c has been studied extensively, and multiple structural species havebeen found to occur at high pH ' 9.5 [Brautigan et al. 1977; Hong & Dixon 1989; Theorell& Åkesson 1941]. Various studies have also found evidence that the dissociation of the axialheme ligand Met100 precedes unfolding under alkaline conditions [Diederix et al. 2002b;Hoang et al. 2003; Worrall et al. 2005b], and that the methionine ligand is exchanged witha lysine [Brautigan et al. 1977; Gadsby et al. 1987]. Also increased peroxidase activity waslinked to the loss of the heme-iron ligands, which would otherwise cover the heme-iron andblock the access to solvent [Worrall et al. 2005a]. After removal of the Met-ligand the ironwould be free to bind a peroxide anion at the vacated coordination site [Dumortier et al.1999]. In addition to the coordination by a lysine residue a large structural rearrangement ofthe ligand loop was observed during the unfolding process [Rosell et al. 1998; Winkler 2004;Worrall et al. 2005b]. Therefore specific rearrangements of the ligand loop could result inthe enlargement of the heme cavity.

If cytc could be stabilised in an expanded conformation, it could be used as an industrialperoxidase [Diederix et al. 2001]. If furthermore the conformation of cytc could be easilyand reversibly manipulated, the peroxidase activity could be regenerated multiple times andcytc applied as a recyclable industrial peroxidase. In the reference experiments of this study,guanidinium hydrochloride (Gdn·HCl, C(NH2) +

3 ·Cl– ) was used to unfold cytochrome c-550[Diederix et al. 2002a,b; Worrall et al. 2005a,b]. In hydrochloric acid guanidinium forms thecationic hydrochloric salt C(NH2) +

3 with a pKa of 12.5, which is thought to denature proteinsby longer-range electrostatic interactions [Camilloni et al. 2008]. However, in common MDsimulation setups, which use cut-off truncation of coulombic interactions to limit computa-tional costs, electrostatic interactions take a very long time to converge [Baker et al. 1999;Bergdorf et al. 2003]. Ionic solvents are therefore less suitable for simulations of intrinsi-cally slow processes like protein unfolding. For this reason neutral urea ((H2N)2CO) wasused for the in silico solvation of cytc.

Urea is widely-used as a chemical denaturant for the unfolding of proteins, and hasbeen used for the denaturation of cytochrome c [Ahmad et al. 1996; Bhuyan 2002; Brunoriet al. 2003; Creighton 1979; Elöve et al. 1994; Fedurco et al. 2004; Khoshtariya et al. 2006;Latypov et al. 2006]. While protein denaturation by urea has been studied and documentedextensively [Bolen & Rose 2008; Ibarra-Molero et al. 2001; McCarney et al. 2005], the exactmechanism by which urea induces unfolding is still not known. Direct interactions withthe protein as well as indirect effects via interactions with the co-solvent water have beenproposed [Makhatadze & Privalov 1992; Timasheff 1993]. One experimental study directlycompared urea with guanidinium with respect to interactions with the protein, where ureawas found to form hydrogen bonds directly with the peptide group but guanidinium chloridenot [Lim et al. 2009]. Different urea models have been parametrised for use with different

11

2.1 Introduction

MD force fields [Caballero-Herrera & Nilsson 2006; Chitra & Smith 2002; Smith et al.2004a] and urea has been the subject of numerous MD simulation studies [Åstrand et al.1994; Idrissi et al. 2000; Stumpe & Grubmüller 2007a,b; Trzesniak et al. 2004; Wallqvistet al. 1998], including being used as a denaturant for proteins [Bennion & Daggett 2003;Caballero-Herrera et al. 2005; Schiffer et al. 1995; Smith et al. 2004b; Smith 2004; Stumpe& Grubmüller 2008; Tirado-Rives et al. 1997; Tobi et al. 2003; Weerasinghe & Smith 2003].For reviews of the experimental and theoretical studies of urea and protein denaturation Irefer the reader to Beck et al. [2007]; Dobson et al. [1998]; Graziano [2002]; Idrissi [2005];McCarney et al. [2005]; Tran & Pappu [2006].

In an experimental study it was found that in 6 M urea cytochrome c unfolded to amolten-globule state in which the global conformation was strongly altered throughout; at thesame time a rather compact and native-like tertiary structure was retained with a “chemical-like” equilibrium between the native-like and non-native local folding of the metal coordi-nation [Khoshtariya et al. 2006]. Different types of denaturing conditions have been used tostudy the unfolding of c-type cytochromes, e.g. thermal, alkaline, acidic or chaotropic de-naturation [Bhuyan & Kumar 2002; Kimura et al. 2007; Kumar et al. 2006; Víglaský et al.2000; Wain et al. 2001; Winkler 2004]. Easily applicable and reversible conditions for theunfolding of P. versutus cytc were tested in the form of basic solvents and detergents whichare commonly used to unfold proteins [Diederix et al. 2004, 2002b]. The reason for employ-ing MD simulation to study cytc in urea was to observe possible structural changes of cytcin atomistic detail, in particular in the region of the heme cavity. MD simulation has beenused to study c-type cytochromes in the past [Arteca et al. 2001; Cai et al. 1992; Collinset al. 1992; Kieseritzky et al. 2006; La Penna et al. 2007; Mao et al. 1999a, 2001], also toinvestigate thermal unfolding [Mao et al. 1999b; Prabhakaran et al. 2004; Roccatano et al.2003].

This chapter presents molecular dynamics (MD) simulations of cytc that were per-formed to determine the structure and stability of cytc during unfolding by the chaotropicdenaturant urea. The simulations were used to evaluate the conformational stability of theprotein at three different urea concentrations, 0, 6 M and 10 M. The questions asked were:

12


i) Is it possible to simulate the urea-induced unfolding of cytochrome c-550? and

ii) Can urea stabilise the enzyme in a partially unfolded state?

Positive answers would suggest that urea, and possibly other denaturants, can be used toreversibly manipulate the catalytic activity of cytochrome c-550 peroxidase and recycle theenzyme for application as an environmentally friendly, “green” reduction catalyst.

2.2 Methods

Systems simulated

The starting structure for the simulations was taken from the X-ray crystal structure 2BGV[Worrall et al. 2005b] of ferricytochrome c-550 from Paracoccus versutus. In this structureresidues Glu2–Pro120, the backbone nitrogen atom of Asp121 and the heme group were re-solved. To complete the structure residues Asp121 and Ala122 were modelled. The hemegroup was linked to the protein by covalent bonds to the sulphur atoms of Cys15 and Cys18.This was done to allow for the loss of the iron-sulphur interaction observed experimentallyduring unfolding under alkaline conditions or associated with high concentrations of de-naturant [Hoang et al. 2003; Worrall et al. 2005b]. Coordination of the heme-iron by thenitrogen atoms of the pyrroline groups and the His19.Nτ-atom was modelled by covalentbonds. As the objective of this study was to probe the unfolding of cytc, coordination of theferric heme-iron Fe(III) with Met100 was modelled using a combination of Lennard-Jonesand coulombic interactions. His19 and His118 were both singly-protonated on the Nπ-atom.2

The side chains of Arg and Lys residues were protonated and Asp and Glu side chains depro-tonated. Water molecules present in the crystal structure were removed. The structure wasplaced in a triclinic, periodic box with a distance of at least 1.1 nm between the protein andthe sides of the box. The protein was then solvated with SPC-water [Berendsen et al. 1981]in combination with a flexible neutral urea model [Duffy et al. 1993; Smith et al. 2004a].Four systems containing different proportions of water and urea molecules corresponding to0, 6 M and 10 M urea were simulated. Five sodium ions Na+ were added to neutralise thenet charge of the system. Table 2.1 lists the compositions and simulation times of all thesystems simulated.

2 The nitrogen atoms of the imidazole ring of histidine are denoted by pros (“near”), abbreviated π, and tele(“far”), abbreviated τ, to show their position relative to that of the side chain [IUPAC 1975].

13

2.2 Methods

Table 2.1 The simulation systems of cytc in water and urea.

[urea] Nmoleculesb

Natomsc A% d t

/M a water urea /[ns] e

water MD1 0 4883 0 15854 0 230water MD2 0 4883 0 15854 0 150water MD3 0 4883 0 15854 0 2306 M urea 6.0 3367 500 15306 33.1 23010 M urea MD1 9.7 2508 791 15057 51.3 17010 M urea MD2 9.7 4064 1282 23653 51.3 70

a molar concentration of urea, g-mol/L [Weast et al. 1986]b number of moleculesc total number of atoms in the systemd anhydrous solute weight per cent of urea, g solute/100 g solution [Weast et al. 1986]e simulation time

Simulation parameters

All simulations and energy minimisations were performed using the GROMACS packageversion 3.1.4 [van der Spoel et al. 2005] in conjunction with the GROMOS96 force fieldversion 43A1 [van Gunsteren et al. 1996]. The temperature (T = 300 K) and pressure (p =1 bar) were held constant by weak coupling to an external bath [Berendsen et al. 1984].A twin-range cutoff of 0.8 nm and 1.4 nm was used in conjunction with a reaction field(εRF = 54) to correct for the truncation of electrostatic interactions beyond the long-rangecutoff [Tironi et al. 1995]. Interactions within the short-range cutoff were updated at everytime step. Longer-range interactions, together with the pair list, were updated every 5 steps.Bonds were constrained by applying the LINCS algorithm [Hess et al. 1997]. The integrationtime step was 2 fs. Initial atom velocities were assigned from a Maxwell distribution at T =300 K. Each system was first energy minimised, then simulated for 100 ps to equilibrate thesolvent, where the positions of all heavy atoms in the protein were harmonically restrainedfor 50 ps, and the heavy atoms of only the backbone for another 50 ps, using a force constantof 500 kJ mol−1nm−1.

Three MD simulations were run of the system containing only water and the protein,starting from different sets of initial velocities at 0 ps. One simulation each of the 6 M ureasystem and the two 10 M urea systems were run.

14


Analysis

Root mean square deviation (RMSD) The RMSD of the positions of the backbone atomswas determined after performing a least squares fit of each configuration in the trajectory to(a subset of) the backbone atoms of the starting crystal structure. Short term fluctuationswithin 1 ns were smoothed by performing a 500 point-running average.

Radius of gyration The mass-weighted radius of gyration Rg of a structure is related tothe distances of all atoms to the centre of mass (COM) of the structure and was calculatedaccording to

Rg =

∑i‖ri‖2mi

∑i

mi

1/2

(2.1)

where ri is the position of atom i with respect to the COM of the structure, and mi the massof atom i.

Secondary structure The secondary structure was determined based on the DSSP defini-tions of secondary structure [Kabsch & Sander 1983].

Molecular contacts Atoms the centres of which are separated by less than or equal to0.6 nm were considered to be in contact. The number of contacts between the solventmolecules, the heme group and the protein were determined for configurations sampled every200 ps.

Solvent accessible surface (SAS) SAS areas of each protein residue were estimated us-ing the algorithm of Eisenhaber et al. [1995], with the diameter of the probe sphere being0.14 nm. The SAS areas per residue were determined by summing the SAS areas per atom.

Radial distribution function (RDF) The cumulative number (CN) of atoms (e.g. solvent)with respect to a reference point or group of atoms (e.g. the protein) was calculated by in-tegrating the number of atoms (particles) N as a function of the distance r to the referencegroup. The density of particles ρ(r) at a given distance was obtained by normalising CN(r)against the probe volume. For a one-particle distribution as described here, the radial dis-tribution function (RDF) is defined as the ratio of the density ρ(r) at a given distance with

15

2.3 Results

respect to the average density ρ of the bulk

g(r) := ρ(r)/ρ (2.2)

The RDF g(r) is then related to the CN according to

CN(R) :=∫ R

0ρg(r)4πr2 dr = NR . (2.3)

From equation (2.3) the RDF g(r) of a (sub)set of N particles from a total set S can benumerically approximated as

g(r) ≈ ∆Nr VS43π(r3− (r′)3)NS

, r = r′+∆r (2.4)

with NS the total number of particles in S, VS the volume taken up by the particles in S, and∆r the thickness of the spherical shells used for determining ∆Nr = CN(r)−CN(r′).

In this study, equation (2.4) was used for calculating a) the RDF of the atoms of cytcwith respect to the heme-iron, and b) the RDF of water and urea atoms around cytc. In a) theset S contained all heavy atoms of the protein, in b) S comprised all urea and water atomscombined.

2.3 Results

Changes in secondary and tertiary structure

In order to investigate the effect of urea on the structural stability of cytc the positionalRMSD was analysed with respect to the initial structure. Figure 2.2 shows the RMSD ofthe backbone atoms of the protein as a function of simulation time for the simulations inwater, 6 M urea and 10 M urea. As can be seen in Figure 2.2A there was an initial rapidincrease in the backbone RMSD of cytc during all of the simulations, irrespective of the ureaconcentration. Among the multiple simulations in water (Fig. 2.2A left panel), the RMSDvalues that are reached vary between 0.3 and 0.5 nm. The simulations in 6 M and 10 M ureareached RMSD values of up to 0.55 nm after simulation times of 200 ns.

To determine whether the structural deviations observed were due to the overall unfold-ing of the structure, or whether they were dominated by less structured parts of the protein,e.g. the loop regions, the backbone RMSD of only the helical regions was determined whilemaintaining the same fit to all backbone atoms as above. The helix RMSD, plotted in Fig-

16


Figure 2.2 Root mean square deviation (RMSD) of the backbone atoms of cytc during MDsimulation in water and urea. The RMSD was calculated with respect to the initial structure ofthe simulations (X-ray crystal structure 2BGV), of A) the entire backbone, and B) the backboneof the helices. Left panels: RMSD of cytc simulations in water, simulation 1 (black line), 2 (darkgrey), and 3 (light grey); middle: in 6 M urea; right: in 10 M urea, simulation 1 (black), and 2(dark grey).

ure 2.2B, had significantly lower values in all simulations, indicating that deviations of theextra-helical regions dominate the RMSD. Nevertheless the RMSD values were high withthe helix RMSD in simulation MD3 in water reaching values comparable to those obtainedin 6 M and 10 M urea. This raised the question of whether cytc was unstable in water usingthe GROMOS96 force field 43A1. Visual inspection of the trajectories however indicatedthat the protein maintained a high degree of structural integrity even in urea, suggesting thatthe high RMSD reflected changes in the relative positions of secondary structure elementsas opposed to loss of structure per se.

The trajectories were analysed in terms of secondary structure to elucidate the extentto which the initial structure was maintained Secondary structure assignments were madebased on the DSSP definitions. For the DSSP analysis sets of five configurations from eachtrajectory were sampled between 142 to 150 ns of simulation. The predominant secondary

17

2.3 ResultsTa

ble2.2

Second

arystructureof

cytc

inwa

terandurea.

Thesecond

arystructureof

each

resid

uein

theX-raycrystals

tructureof

cytc

(2BG

V[W

orrallet

al.20

05b])t

hatw

asused

astheinitia

lcon

�guration

,and

inlat

ercon�

guratio

nsfro

mtheM

Dsim

ulation

sinwa

tera

ndurea

wasd

etermined

accordingto

theDS

SPde�n

ition

s.Th

evalues

listedrepresenta

consensuso

ver�

vecon�

guratio

nstakenevery2ns

from

thetra

jectory

intheinterval

142�

150ns,e

xceptf

ortheshorters

imulation

�10M

urea

2�(*)e

very

1ns

intheinterval

63�6

7ns.Lette

rcod

e:∼

-rando

mcoil,

E-β

-sheet,B

-β-brid

ge,S

-bend,

T-t

urn,

H-α

-helix,

G-3

10-helix;

helix

andturn

structures

�H�,�G

�and

�T�a

rehigh

lighted

indark

grey,b

etastructures

�E�a

nd�B�in

light

grey.

18


Figure 2.3 Cumulative number (CN) of the radial distribution function (RDF), and RDF of the cytcatoms with respect to the heme-iron. A) The CN(r) was determined by summing the number ofheavy atoms of cytc that were located within a distance r of the heme-iron. The analysis was basedon con�gurations sampled every 0.2 ns between 130 and 150 ns. Legend: line colours accordingto solvent type: black - water (line styles continuous, dashed, dotted for MD1, 2, 3 respectively),grey - urea (circles for 6 M urea, and squares, diamonds for 10 M urea MD1, 2 respectively). B)Di�erence ∆CNwater with respect to the control CN. The control CN was obtained by averagingthe CN curves from the simulations 1, 2 and 3 in water. C and D) RDF of all heavy atoms of cytcwith respect to the heme-iron. The data from the simulations in water were averaged (triangles).Graphs C and D contain the same data but shown on di�erent scales.

structure assignments for each residue in the configurations sampled are given in Table 2.2along with the secondary structure of the X-ray crystal structure 2BGV. It can be seen thatin water all five alpha-helices remained stable after 150 ns of simulation, whereas at bothurea concentrations alpha-helix 1 and alpha-helix 5 were partially unfolded. Most of theβ-sheet and β-strand of the initial cytc conformation remained intact in all the simulations,with some slight shifts along the primary sequence.

19

2.3 Results

Heme accessibility

The packing of the protein matrix around the heme was examined as a means of quantifyingstructural changes within the protein related to the accessibility of the heme to the solventand substrates. To measure the accessibility of the heme group the distribution of distancesof the heavy atoms of the protein from the heme-iron was analysed. This was done bycalculating the cumulative numbers (CN) of the radial distribution function (RDF) of proteinatoms with respect to the heme-iron. The CN values for each simulation were averaged overconfigurations sampled every 0.2 ns from 130 to 150 ns and are plotted in Figure 2.3A. Ascan be seen all curves have a similar sigmoidal shape with the curves shifted towards largerdistances r with increasing urea concentration. The shifts of the curves show that in ureathe protein atoms on average shifted further away from the heme group than in water. Theslope of the curve also decreased with increasing urea concentration and the distribution ofprotein atoms around the heme iron in urea was less dense than in water. The three CNcurves of the simulations in water were averaged to serve as a control. The protein packingin each simulation was compared to the control by calculating the difference ∆CN betweenthe average of the three CN curves in water (control) and the CN in 6 M and in 10 M urea.Figure 2.3B clearly shows that the difference in the number of protein atoms between the CNand the control was approximately twice as large in 10 M urea as in 6 M urea. The differenceplots furthermore show that in comparison with the average density in water the decrease inprotein density in urea was most pronounced at a radius of approximately 1.7 nm around theheme group. The radial distribution functions (RDF) calculated from the CNs are plotted inFigure 1.2C and D. These show that in urea the density of protein atoms within ≈ 1.7 nm ofthe heme-iron was lower than in water, and that this was compensated by higher densities atdistances ' 1.7 nm.

The radius of gyration Rg is another structural property that gives an estimate of thesize of a molecule. The Rg calculated for the configurations extracted at 150 ns of simulationwere similar to the initial configuration, varying by less than 1 % (data not shown).

Another measure of the accessibility of the heme group within the protein matrix is thenumber of contacts with solvent molecules and with the protein environment. Atoms thatlay within 0.6 nm of the reference group were considered to be in contact. Solvent contactswith the heme group give a measure of (solvent) accessibility, protein contacts a measure ofocclusion (from the solvent). In Figure 2.4 the numbers of solvent and protein contacts withthe heme group are plotted as a function of simulation time. From panels A and B it canbe seen that the number of contacts between the heme group and solvent increased in a ureaconcentration-dependent manner while the number of contacts with the protein decreased. Inline with the increasing number of solvent contacts an increase in solvent accessible surfacearea (SAS) of the heme group was also observed (data not shown). A more detailed analysis

20


Figure 2.4 Protein environ-ment and solvation of theheme group in cytc. Num-ber of protein and solvent con-tacts with the heme groupduring MD simulations of cytcin water and in urea. Con-�gurations every 200 ps wereanalysed and the data curvessmoothed by 10 ns-runningaverages. A) Contacts withcytc; B) contacts with the sol-vent. Legend: no symbols� water, simulation 1 (blackline), 2 (grey), 3 (dotted); cir-cles � 6 M urea; squares �10 M urea simulation 1; dia-monds � 10 M urea simulation2.

of solvent interactions with the protein are presented in the following section.

The largest changes in the number of protein and solvent contacts with the heme groupwere observed for simulation 2 in 10 M urea (Fig. 2.4). The number of solvent contactsincreased by approximately twice as much after 45 ns compared to the value reached in theother 10 M urea simulation 1 after 170 ns. Configurations of cytc at these time points in the10 M urea simulations are shown in Figures 2.5 and 2.6. As can be seen in Figure 2.5 A, insimulation 1 the protein matrix remained tightly packed around the heme group, whereas insimulation 2, shown in panel B, the protein matrix opened, exposing the heme to the solvent.In panels B and C it can also be seen that where in simulation 2 (pink) a cavity is formed,the configuration from simulation 1 (blue) and the initial structure (green tube) both enclosethe heme group to both sides of the heme plane. In panel D the protruding cartoon traces ofthe configurations in 10 M urea illustrate that in particular helix 1 and several loops deviatefrom the initial configuration (green surface).

Figure 2.6 shows the backbone configurations from the simulations in 10 M urea super-imposed upon the initial structure. It can be seen that in simulation 1 the tertiary structureoverall deviated more from the initial structure than in simulation 2 (Fig. 2.6 A and B re-spectively), in particular helix 1, and the loop regions adjacent to helix 1 and helix 2. At thesame time the middle section of this loop and the loop between helix 5 and helix 6 remainedclose to either side of the heme in simulation 1. In simulation 2 on the other hand, these loop

21

2.3 Results

Figure 2.5 Legend see next page.

22


Figure 2.5 (previous page) Opening of the cytc heme cavity in 10 M urea. A�D) Superposition ofcon�gurations from the simulations in 10 M urea upon the initial structure (green): after 170 ns ofsimulation 1 (light blue), and after 42 ns of simulation 2 (pink) in 10 M urea. A and B show cross-eyed stereo images. In A) the con�guration from simulation 1 is shown as the solvent accessiblesurface (SAS), and the con�guration from simulation 2 in the secondary structure representationof VMD [Humphrey et al. 1996]; in B) the representations are vice versa. C) The con�gurationsof simulations 1 and 2 are both shown as surfaces. In A�C the initial structure is shown in thesecondary structure representation of VMD. D) The con�gurations of simulations 1 and 2 areboth shown in the secondary structure representation of VMD and the initial structure is shownas a surface (green). The heme groups (cyan) belong to the respective structures shown as SAS.Con�gurations were �tted onto the backbone structure of the helices in the initial con�guration ofthe simulations.

Figure 2.6 Conformation of cytc in 10 M urea. Superposition of con�gurations from the simula-tions in 10 M urea upon the initial structure: A) after 170 ns of simulation 1 (blue), and B) after42 ns of simulation 2 (pink) in 10 M urea, the heme group is coloured blue. Protein structuresare shown in the cartoon representation of VMD [Humphrey et al. 1996], the heme group in stickrepresentation. Con�gurations were �tted onto the backbone structure of the helices in the initialcon�guration. In the initial structure the heme group is coloured by atom type and the cartoontrace according to the secondary structure: purple � α-helix, yellow � β-sheet, cyan � turn, white� loop.

sections and also the domain around helix 1 shifted away from the heme. These shifts arealso clearly visible in Figure 2.5D.

Protein-solvent interactions

Interactions between the different solvents and cytc were analysed with the aim of findingindications in regard to how urea affects the protein structure. This question was approachedby comparing the distributions of water and urea. To this end the radial distribution functions(RDF) of solvent molecules around the protein were determined and are plotted in Figure 2.7.Comparing the RDF of water with the RDF of urea in the simulations in urea, it can be seen

23

2.3 Results

Figure 2.7 Cumulative number (CN) of the radial distribution function (RDF), and RDF of solventspecies with respect to cytc. A) The CN(r) of the atom density of water and of urea with respect tocytc was determined by summing the number of solvent atoms that were located within a distancer of the protein, in con�gurations at 150 ns. Legend: line colours indicate the solvent composition:black � water simulations 1�3, dark grey � 6 M urea, light grey � 10 M urea simulations 1 & 2; linestyles indicate the solvent species analysed: continuous lines � water, dashed lines � urea. B�C)The RDF(r) of the atom density derived from the CN given in panel A, on two di�erent lengthscales; legend as in A.

24


that a shell of urea molecules coated the protein, displacing the water molecules away fromthe protein. Interactions between the sodium ions and the protein decreased with increasingurea concentration (data not shown).

Heme ligands

Experimentally it has been shown that specific interactions between the protein and the hemegroup correlate with either the native or (partially) unfolded cytc [Brautigan et al. 1977;Diederix et al. 2002b; Dumortier et al. 1999; Gadsby et al. 1987; Hoang et al. 2003; Worrallet al. 2005b]. Here specific interactions were used as indicators of whether partially unfoldedstates of cytc occurred in the simulations.

Figure 2.8 Hydrogen bonding between Lys99 and the heme group. Occurrence plot for thehydrogen bond between the Lys99 Nζ-atom and one of the carboxyl substituents CγδOδ

2 of theheme group. The occurrence of the hydrogen bond was determined for con�gurations sampledevery 0.02 ns and is plotted as circles against the simulation time. Black vertical lines indicate theending times of the shorter simulations. Graphs 1�6 from top to bottom refer to simulations: 1)water simulation 1, 2) water 2, 3) water 3, 4) 6 M urea, 5) 10 M urea 1, 6) 10 M urea 2.

One specific interaction is the coordination between Met100 and the heme-iron. Lossof this interaction was previously shown to be associated with alkaline unfolding [Hoanget al. 2003; Worrall et al. 2005b]. To determine whether structural rearrangements duringthe simulation lead to the loss of the Met100 heme interaction, the distance between theMet100.Sδ-atom and the heme-iron was monitored. In all simulations the initial distance of

25

2.4 Discussion

0.29 nm was lost within the first picoseconds of MD after which the values fluctuated aroundan average distance of 0.42 ±0.04 nm (data not shown).

The unfolding of cytc has also been reported to lead to interactions between the heme-iron and one or more lysine residues [Rosell et al. 1998; Worrall et al. 2005b]. Thereforethe distances between all lysine residues and the heme-iron in the simulations were mon-itored. Under all solvent conditions the lysines remained at least 0.66 nm away from theheme-iron and remained at similar distances throughout the simulations (data not shown).Furthermore a number of hydrogen bonds between other functional groups in the heme andlysine residues were detected. An analysis of the correlation between the hydrogen bond-ing propensity and the concentration of urea indicated that a hydrogen bond between theLys99 side chain and one of the heme carboxyl groups CγδO2 was anti-correlated with thepresence of urea. Figure 2.8 contains an existence plot for this hydrogen bond showing thatthe interaction was found in all simulations in water, but only at the beginning in simulation2 in 10 M urea and at isolated time points in the other simulations with urea. For signifi-cant proportions of the simulations with urea the distances between these two groups werelarger than 0.4 nm (data not shown) and therefore well above the hydrogen bonding distance.

2.4 Discussion

The analysis of the secondary and tertiary structure of bacterial cytochrome c-550 during thesimulations in water and urea (Tab. 2.2) showed that at all urea concentrations the proteinremained intact during 150 ns. The persistence of residual structure after denaturation in ureawas already reported for other proteins [McCarney et al. 2005; Shortle & Ackerman 2001].The high RMSD values that were observed nonetheless (Fig. 2.2) were due to unstructuredregions such as the highly fluctuating termini and loop regions, and slight shifts of the helicesrelative to each other (data not shown). The partial unfolding of the N- and C-terminal helicesin both 6 M and 10 M urea were likely to be due to the higher mobility and solvent exposurethat are commonly observed for protein termini and which resulted in an increase in solventinteractions that propagated along the sequence and destabilised the adjacent helices.

The distribution of protein atoms around the heme-iron was used as an indicator for thepacking of the protein (Fig. 2.3). The wider distributions of the protein atoms in urea indi-cated that the molecule occupied a larger volume. This could be regarded as a tertiary struc-tural effect, where secondary structure elements remained intact but moved apart slightly.Less density in packing implied an increased solvation of the protein, with urea and/or watermolecules invading the protein structure and driving secondary structure elements apart. Thiswas confirmed by a decrease in protein-internal interactions in the simulations with urea.

26


In 6 M urea the number of contacts between the heme group and the solvent molecules(Fig. 2.4) took intermediate values between what was observed for water and for 10 Murea. The changes in the SAS area of the heme group (data not shown) and the RDF cu-mulative number (CN) of protein atoms around the heme-iron were also correlated with theurea concentration (Fig. 2.3). This suggests that the structural changes observed were urea-dependent.

With the secondary structural elements staying largely intact (Tab. 2.2) and a medium-high RMS deviation from the initial backbone conformation (Fig. 2.2), the partially unfoldedprotein remained stable during 150 ns in 10 M urea. However, solvent interactions with theprotein and heme in 10 M urea were still increasing towards the end of the 150 ns trajectory(Fig. 2.4). This suggests that the protein structure and/or solvent distribution had yet to reachequilibrium, and even more extensive unfolding might be observed if the simulations wereextended. From experiment it is known that high concentrations of urea fully denature andunfold proteins. In an MD simulation with 10 M urea, i.e. at approximately the highest con-centration typically used in protein denaturing experiments, a protein would be expected tofully unfold. Thermal and chemical unfolding have been determined experimentally to takeplace on the microsecond time scale [Huang & Oas 1995; Mayor et al. 2000], e.g. the coldshock protein CspB, which is known to be an extremely fast-folding protein, was reportedto unfold at a rate of about 100 s−1 in 8 M urea [Schindler & Schmid 1996]. Accordinglythe unfolding of cytc in urea can be expected to require more time than was accessed in thisstudy. The reason simulation 1 in 10 M urea was stopped after 150 ns was that as a result ofthe expansion of the unfolding protein, there were increasing interactions with its periodicimage.

Despite the apparent stability of cytc in urea, structural changes were nevertheless evi-dent from the MD simulations. In particular the protein environment surrounding the hemegroup and solvent interactions with the heme were affected (Fig. 2.4), suggesting a urea-dependent widening of the heme cavity, as proposed by Canters et al. from experimentalobservations [Diederix et al. 2002a; Worrall et al. 2005a]. Particularly simulation 2 in 10 Murea displayed a notable widening of the heme cavity (Fig. 2.5). In this simulation, and alsoto a lesser extent in the other simulations with urea, a decrease in the number of contactsbetween the protein and the heme group indicated a progressive separation of the proteinmatrix from the heme (Fig. 2.4). This allowed more urea molecules to access the cavity, aswas indicated by the increasing interactions of urea molecules with the heme group (data notshown). The configurations from the simulations in 10 M urea in Figure 2.5 clearly showthat in simulation 2 the heme cavity opened, exposing the heme group to the bulk solvent. Insimulation 2 the displacement of the loop containing the heme ligand Met100 away from theheme iron (Fig. 2.6) corresponded to one of the structural rearrangements observed in alka-line denatured cytc experimentally [Rosell et al. 1998; Winkler 2004; Worrall et al. 2005b].

27

2.4 Discussion

Interestingly the backbone conformation of the structure with the open heme cavity in thissimulation was quite similar to the initial structure. This preservation of structure incorpo-rating an exposed catalytic centre is a very promising prospect for an enhanced peroxidasethat can be readily regenerated. In contrast to what was observed for 10 M urea the numberof contacts of the heme group remained constant in 6 M urea (Fig. 2.4), suggesting that at thelower urea concentration the system had reached equilibrium and that the structure of cytcwas stable. Whether urea had a direct effect on the conformation of the protein as a solvent,resulting in an expansion of the cavity by mechanical refolding, or whether specifically theheme group interacted preferentially with urea, leading to a coating of the heme group withurea that displaced the protein matrix, cannot be determined from the present simulations.

The urea molecules preferentially associated with the protein, forming a shell aroundcytc, while the water molecules were displaced away from the protein (Fig. 2.7). This resultis in line with recent experimental data [Lim et al. 2009]. There was a very small accumula-tion of water molecules close to the protein, which probably reflects the internal hydration ofthe protein. A comparison of the RDF of water molecules around cytc in water and in ureashowed that in urea even the first shell of water molecules was displaced by urea molecules(Fig. 2.7B). In the first shell of solvation the clustering of urea molecules around cytc wasonly slightly weaker in 6 M urea than in 10 M urea, suggesting a strong interaction betweenthe protein and urea. The enhanced local density of urea around the protein could result fromhydrophobic interactions between the protein and urea, as proposed in the literature as a pos-sible mechanism for protein denaturation by urea [Kamoun 1988; Nozaki & Tanford 1963;Smith et al. 2004b]. The degree of clustering observed in the two systems in 10 M urea wasless pronounced in the smaller system in simulation 1 than in the larger system in simulation2. This was most likely an artefact due to the size of the system and the greater number ofurea molecules available to cluster around the protein in the larger system. The decreasedinteractions of the sodium ions with the protein in urea was presumably a consequence of thedisplacement of water and the higher solubility of sodium ions in water [Serjeant & Dempsey1979].

The Met100.Sδ-atom was uncharged in the present simulation setup. Thus no interac-tion with, or coordination of, the heme-iron was expected. Under alkaline conditions lysineresidues were proposed to act as a sixth ligand for the central iron atom and replace the me-thionine ligand during the unfolding of cytc [Gadsby et al. 1987; Russell et al. 2000; Ubbinket al. 1994]. In the simulations a large distance was maintained between the lysine residuesand the heme-iron in all cases (data not shown). A significant conformational change wouldbe necessary to bring a lysine close enough to the heme-iron to enable them to interact. Al-though a structural rearrangement of the ligand loop was observed in simulation 2 in 10 Murea, it did not bring Lys99 close to the heme-iron (Fig. 2.6). However, the native hydrogenbond between the Lys99 side chain and one of the heme carboxyl groups was disrupted in all

28


the simulations with urea (Fig. 2.8). Disruption of this hydrogen bond is required in orderfor the hinge region between α-helices 4 and 5 (Ser94–Asn107 in Paracoccus versutus cytc)to shift and Lys99 to displace the heme-iron ligand Met100, a phylogenetically conservedmotion which was shown experimentally in alkaline unfolded cytc [Dumortier et al. 1999].Removal of the Lys99 side chain away from the heme results in increased solvent exposureof the heme [Worrall et al. 2005b]. In the simulations this hydrogen bond proved to beextremely urea-sensitive to an extent that it almost did not form at all in two out of threesimulations with urea.

Considering i) the expansion of the heme cavity, ii) the coating of the protein by urea,and iii) the displacement of the protein matrix away from the heme group, urea appeared tohave a loosening effect on the tertiary structure of the protein which subsequently lead to awidening of the heme pocket.

2.5 Conclusion

Bacterial cytochrome c-550 was shown in MD simulations to partially unfold in solutionwith urea. Partial unfolding occurred on a time scale of 100 ns, and the extent of unfoldingdepended on the urea concentration. The partial unfolding was associated with increasedaccess of the solvent to the prosthetic heme group in the cavity of the protein matrix, anda decrease in interactions between the heme group and the protein matrix. In 10 M urea, asignificant widening of the heme cavity was observed, exposing the heme group and lettingthe solvent molecules access the catalytic site more easily, as proposed by Diederix et al.[2002b].

In the systems simulated containing urea, a solvation shell of urea molecules formedaround the cytc, displacing the water molecules away from the protein. The number of in-teractions between the solvent and the protein in urea clearly increased compared to thesimulations in water. At the same time the hydrophobic interactions within the protein di-minished, which has been previously proposed as an effect of urea [Nozaki & Tanford 1963].Enhancement of the solubility of the protein is believed to be one of the main mechanismssuggested for protein denaturation by urea [Kamoun 1988], and the simulations here provideevidence for this mechanism.

Progressive unfolding was observed in 10 M urea, while in 6 M urea and in neat wa-ter the protein conformation remained stable and the interactions between the solvent, theprotein and the heme group remained constant within 230 ns of simulation. The density ofthe protein decreased in a urea concentration-dependent manner while the tertiary structure

29

2.5 Conclusion

remained intact. Only isolated terminal regions of two helices unfolded, similarly followinga urea concentration-dependent trend, which suggests that the extent of unfolding can bemodulated by the urea concentration. Therefore solvation in urea promises to be a suitablemeans for the partial and reversible unfolding of cytochrome c-550 peroxidase, as a recy-clable, switchable enzyme for the environmentally safe reduction of peroxides.

30

Introduction to flaviviralenvelope glycoprotein E 3

During the last two decades, dengue fever and dengue haemorrhagic fever have emergedglobally as a major public health concern [CDC 2008; WHO 2008]. Both conditions arecaused by infection with the dengue virus, a flavivirus that is now endemic in more than 100countries. Both the disease and its vector are prevalent in Africa, the Americas, Australia,India, the Pacific and South-East Asia. Dengue fever is the most common mosquito-borneviral disease for humans, and the vector of transmission is the day-active mosquito Aedesaegypti. The life-cycle of the dengue virus is completed in humans, where the outcome canbe deadly, especially in children and adolescents [CDC 2008]. Currently there are no anti-viral drugs available that can be used to treat or inhibit infection by the dengue virus [WHO2008].

The dengue virus belongs to the Flaviviridae family, that also contains the viruses caus-ing tick-borne encephalitis (TBE), Japanese encephalitis, yellow fever and West Nile fever.Like all flaviviruses, the dengue virus is organised as an enveloped particle that containsthe viral RNA. In all enveloped viruses, the nucleic acid is surrounded by a membrane thatis covered entirely by membrane-anchored viral envelope proteins [Schibli & Weissenhorn2004; Söllner 2004]. As a first step towards infection, the entire viral particle is taken up intoan endosome through endocytosis by the host cell [Marsh 1984].

A central event in the invasion of a host cell by an enveloped virus is the fusion of theviral and the cell membrane. The flaviviral envelope consists of 90 dimers of the membrane-anchored envelope glycoprotein E, that cover the surface of the viral membrane [Stiasny &

31

3 Introduction

Figure 3.1 Schematic diagram of a model of the �aviviral membrane fusion process, proposed byHeinz et al. [Stiasny et al. 2005]. A) The E protein homo-dimer in its native state at the surface ofthe mature virion (lower bilayer). B) Low-pH-induced dissociation of the E dimer and interaction ofthe fusion peptide loop (orange dot) with the endosomal membrane (upper bilayer). C) Formationof an E trimer (including the �ipping back of domain III (blue)) proceeding via still unde�nedintermediates. D) Formation of the �nal post-fusion conformation through interactions of thestem-anchor region with domain II (yellow), leading to the juxtaposition of the fusion peptideloops and the membrane anchors (green cylinders) in the fused membrane. Original image byStiasny et al. [2005].

Heinz 2006]. Each identical subunit of the flaviviral E protein contains about 500 residues.As in other low-pH-dependent viruses, the conformation of the envelope protein is pH-dependent [Modis et al. 2004; Skehel et al. 1982; Zhang et al. 2004]. Specifically, theacidification of the endosomal pH is believed to trigger a large-scale conformational changeof the E protein [Schibli & Weissenhorn 2004; Zimmerberg et al. 1993]. This in turn is asso-ciated with the fusion of the viral membrane with the endosomal membrane of the host cell[Gollins & Porterfield 1986; Helenius et al. 1980; Stiasny & Heinz 2006], allowing the viralRNA to access and infect the cell through the fusion stalk. Although the exact mechanismsof viral protein-mediated membrane fusion are not known, a number of models have beenproposed [Gibbons et al. 2003; Harrison 2008; Helenius 1995; Poumbourios et al. 1999].Figure 3.1 shows a schematic of one model of flaviviral membrane fusion induced by thelow-pH dependent conformational changes of the E protein [Stiasny et al. 2005].

The primary differences between the pre- and the post-fusion conformation of the E pro-tein are generally considered to be the irreversible, oligomeric rearrangement from dimersto trimers [Allison et al. 1995; Bressanelli et al. 2004], and the displacement of domain IIIrelative to domains I and II by 3.3 nm [Bressanelli et al. 2004; Modis et al. 2003, 2004; Reyet al. 1995]. The conformation-specific position of domain III is characterised by associatedchanges of the domain interfaces; X-ray crystal structures of the pre- and post-fusion confor-mation are shown in Figure 3.2. One essential aspect of viral fusion is the irreversibility ofthe structural changes of the fusion protein following exposure to low pH, ultimately result-ing in an inactive fusion protein. This suggests that the fusion-active form of the E proteinis an unstable intermediate of the low-pH-dependent conformational change. In flavivirusesso-called fusion inactivation, the premature activation of the fusion protein, prevents the

32

III Histidine protonation in pH-dependent viral fusion proteins

Figure 3.2 X-ray crystalstructures of the dengueviral type 2 envelope proteinsoluble ectodomain sE, A)in the dimeric, pre-fusionconformation, and B) in thetrimeric, post-fusion confor-mation (PDB IDs: 1OANand 1OK8 respectively). Theidentical rotamers are dividedinto three domains, dI (red),dII (yellow/green) and dIII(blue) [Rey et al. 1995]. TheMMDB database identi�edthe elongated part of domainII in the post-fusion confor-mation as a fourth homologydomain (green) [Chen et al.2003; Wang et al. 2007].

infection of the host cell [Heinz 2003; Hurrelbrink & McMinn 2001; Mandl 2005; Mandlet al. 2001; McMinn 1997]. Therefore an understanding of the process of the activation offlaviviral envelope glycoproteins, in particular of the factors that trigger the conformationalchange, would be an important step towards finding ways to manipulate this critical step inthe infection process, e.g. in the design of an anti-flaviviral drug.

From X-ray crystallographic and cryo-electron-microscopic structures it is known thatthe class II viral envelope proteins from flaviviruses and alphaviruses are very similar in thestructural organisation of the domains and subunits [Bressanelli et al. 2004; Gibbons et al.2004; Kuhn et al. 2002; Modis et al. 2004; Zhang et al. 2003, 2002]. The pH threshold forfusion depends on the specific virus. For various flaviviruses the pH of fusion was determinedto be pH ≈ 6.6 [Stiasny & Heinz 2006]. In tick-borne encephalitis virus (TBEV) the solubleectodomain of E was found to be activated by acidic pH < 6.5 [Allison et al. 1995]. The factthat the pH of fusion is similar to the pKa of protonation of histidine in water (pKa = 6.0)strongly suggests that the process is triggered by the protonation of one or more histidineresidues [Kampmann et al. 2006]. A number of histidine residues in the E protein dimer arelocated at the domain and subunit interface, which implies that they play a significant role inthe structural stability of the dimer. Figure 3.3 shows the positions of the histidines in the Edimer.

33

3 Introduction

Figure 3.3 The pre-fusion conformation of the mature DEN2 sE protein (1OKE [Modis et al.2003]) and the positions of the histidines in the dimer. The histidine residues are represented asspheres. A) View onto the inner envelope surface (in situ facing the viral membrane), B) rotated by90◦ and C) top view onto the outer envelope surface. The solvent-accessible surface was renderedfor one half of the structure, the other half is shown in cartoon representation. Colour legendaccording to domains: red � domains dI; yellow and orange � dII of the two subunits respectively;blue � dIII; the domains of one subunit of the dimer are labelled dI-III, of the other subunit dI'�dIII'.

34

III Histidine protonation in pH-dependent viral fusion proteins

The structural similarity and the low-pH dependency of the envelope proteins through-out the Flavivirus family and in other virus families strongly suggest a common fusion acti-vation mechanism. The following chapters of this thesis present the details of this hypothesis,and the molecular modelling studies that were undertaken to test it. Chapter 4 introduces thehistidine-switch hypothesis and reviews the different modelling approaches undertaken totest this hypothesis. Chapter 5 presents the molecular modelling and MD simulation studyof the dengue viral type 2 envelope glycoprotein soluble ectodomain sE undertaken to in-vestigate the effect of histidine protonation on a pH-dependent fusion protein. Chapter 6introduces a model for the fusion activation mechanism of flaviviral envelope proteins thatis based on the results of the MD simulation study.

35

The role of histidineresidues inlow-pH-mediated viralmembrane fusion 4This chapter was adapted from the publications

Kampmann, T., D. S. Mueller, A. E. Mark, P. R. Young & B. Kobe. The roleof histidine residues in low-pH-mediated viral membrane fusion. Structure 14(10),1481�1487 (2006)

and

Mueller, D. S., T. Kampmann, R. Yennamalli, P. R. Young, B. Kobe & A. E. Mark.Histidine protonation and the activation of viral fusion proteins. Biochem. Soc. T.36(Part 1), 43�45 (2008).

In the article by Kampmann et al. [2006] I was responsible for the molecular dynamicssimulations of the dengue viral type 2 sE protein and wrote the corresponding parts of thearticle. I also contributed substantially to the structural analyses of the sequence alignmentsand the homology models. Among the studies reviewed in Mueller et al. [2008], written bymyself, I was responsible for the simulations of the wild type sE protein.

37

4.1 Introduction

4.1 Introduction

Membrane fusion is an essential step during the entry of enveloped viruses into their hostcell. Depending on the type of virus, membrane fusion is initiated by various mechanisms,including receptor binding (e.g. HIV-1), changes in pH (e.g. Influenza virus and Denguevirus) or a combination of both (e.g. Avian sarcoma virus and Leukosis virus), with themechanism of fusion being related to the individual viral life cycle. Membrane fusion is afundamental biological process that occurs within a wide range of organisms and processes.Often this process is promoted or catalysed by so-called fusion proteins that attach to themembranes, thus prolonging the contact period and possibly enhancing interactions betweenthe membranes. Both viral and eukaryotic fusion proteins show a high degree of structuralsimilarity [Söllner 2004]. These findings raise the questions:

i) What are the underlying mechanisms that trigger fusion?

ii) Are these mechanisms common to one or more entire classes of viral fusion proteins?

Before the fusion event, viral surface proteins that drive fusion adopt a “meta-stable”conformation, the so-called “pre-fusion” conformation. After a specific regulatory eventsuch as receptor binding or a change in pH, they undergo a series of structural transitions,ultimately leading to the stable “post-fusion” conformation. All viral fusion proteins containa hydrophobic segment referred to as the “fusion peptide”, which is in most cases initiallyburied within the pre-fusion form. However, once the conformational change is triggered,the fusion peptide is exposed and can associate with the membrane of the host cell. In thistransition phase the protein is anchored simultaneously in the viral envelope and the host cellmembrane, and further conformational changes drive the two membranes to fuse [Harrison2005; Schibli & Weissenhorn 2004].

Two major classes of viral membrane fusion proteins have been characterised:

Class I fusion proteins are found in the envelopes of viruses belonging to the Coronaviridae,Filoviridae, Arenaviridae, Orthomyxoviridae, Paramyxoviridae and Retroviridae fam-ilies [Colman & Lawrence 2003; Earp et al. 2004; Poumbourios et al. 1999]. Theinfluenza virus hemagglutinin (HA) is the archetypical class I fusion protein that hasbeen studied the most (Fig. 4.1A). In the pre-fusion conformation HA forms trimericspikes on the virion surface [Wilson et al. 1981]. During fusion the proteins rearrange,forming post-fusion hairpin structures in which the C-terminal membrane anchor andthe fusion peptide are juxtaposed at the same end of a helical, rod-like structure [Bul-lough et al. 1994].

38

4 Hypothesis: Histidine protonation triggers low-pH-mediated viral fusion

Class II fusion proteins such as those encoded by the viruses of the Togaviridae and Flaviviri-dae families have a very different molecular architecture than the class I proteins (Fig.4.1) [Allison et al. 2001; Kielian 2006; Lee et al. 1997; Lescar et al. 2001; Modis et al.2003; Rey et al. 1995]. They have three domains that are rich in β-structure, and theirfusion peptide is located within an internal loop. While the fusion proteins from dif-ferent viruses in this class share a remarkably similar tertiary structure, the processingof the fusion proteins differ. Flavivirus E, and alphavirus E1 fusion proteins initiallyassemble as heterodimers, involving a companion viral surface protein that acts as afolding chaperone, prM in flaviviruses, and PE2 in alphaviruses. Maturation of theviral particle requires proteolytic cleavage of the chaperone by furin, leading to the re-arrangement of the fusion protein as an E homodimer in flaviviruses, and as an E1/E2heterodimer in alphaviruses, where E2 derives from PE2. Once activated the dimerrearranges further into fusion protein homotrimers, which expose the fusion peptide atthe tip of a rod-like structure [Bressanelli et al. 2004; Gibbons et al. 2004; Modis et al.2004].

Although the details of the conformational transitions in class I and class II proteinsdiffer, they nevertheless share common mechanistic features. Notably, in both cases theprotein folds back onto itself, causing the two membrane attachment points to be locatedclose together in the post-fusion conformation, which in turn facilitates membrane fusion[Bressanelli et al. 2004; Bullough et al. 1994; Fass et al. 1996; Gibbons et al. 2004; Jardetzky& Lamb 2004; Kielian & Rey 2006; Kobe et al. 1999; Modis et al. 2004; Roche et al. 2006;Schibli & Weissenhorn 2004] (Figure 4.1).

Histidine is the only amino acid the protonation state of which changes near the pH offusion. In solution in water, histidine is predominantly uncharged at neutral pH, and doublyprotonated and positively charged at pH 6 and below. However, the effective pKa of a spe-cific histidine ultimately depends on its local environment. Figure 4.2 illustrates the effectof protein-internal interactions of histidine that significantly differ before and after histidineprotonation. In the pre-fusion state (left) the histidine residue is located in the vicinity ofpositively charged residues (+ symbols). At low pH (middle) the histidine residue is dou-bly protonated. This will favour electrostatic interactions with negatively charged groups,which may lead to the formation of new salt bridges (right) and thereby facilitate irreversiblerefolding. Hydrogen bonds involving histidine as the acceptor will also be perturbed uponprotonation, due to repulsion from the donor. There are many examples of histidine pro-tonation triggering structural changes at low pH [Nordlund et al. 2003], including changesinduced within the endosome [Lazar et al. 2003]. Therefore, histidine residues are expectedto play a critical role in the process of viral membrane fusion [Bressanelli et al. 2004; Chenet al. 1998; Da Poian et al. 2005; Roussel et al. 2006; Stevens et al. 2004]. Questions re-

39

4.1 Introduction

Figu

re4.1Th

estructures

ofA)

thein�u

enza

virus

hemagglu

tinin

(HA)

,and

ofB)

thedeng

uevir

alenvelop

eprotein

E,in

therespectiv

epre-fusio

n(le

ft)andpo

st-fu

sion(righ

t)conformation

s.Th

erib

bons

representt

hebackbo

neconformation

,the

beadsind

icate

selec

tedresid

ues.

Onesubu

nitin

each

protein

complex

ishigh

lighted

orange

andcyan

again

sttheothers

ubun

itsin

grey.C

olour

legend:

magenta

�histidines,b

lue�po

sitive

lycharged,

red�negativ

elycharged;

inA:

orange

�HA

1,cyan

�HA

2;in

B:o

rang

e�do

main

sIandII,

cyan

�do

main

III.

40


Figure 4.2 Schematic diagram of the interactions of a histidine residue in a viral fusion protein inthe pre-fusion and the post-fusion state.

garding the nature of this role remain, specifically:

iii) Are the initial steps in viral membrane fusion simply due to the general effects of anincrease in surface charge?

iv) Or is fusion triggered by the protonation of one or more critical histidine residues?

This chapter presents an extension of previous structural analyses [Bressanelli et al. 2004;Chen et al. 1998; Roussel et al. 2006; Stevens et al. 2004], to consider the role that individualhistidine residues play, and whether these are similar in class I and class II viral fusionproteins.

In sequence alignments of viral fusion proteins from class I and II, shown in Table 4.1, asmall number of highly conserved histidine residues were identified that lie in key structurallocations. Based on this it is proposed here that these specific histidine residues and theirinteraction partners play a significant role in initiating the structural transition leading toviral fusion, employing similar triggering mechanisms commonly involving :

1. the collocation of histidines and positively charged residues in the pre-fusion structure,and

2. the protonation of these histidines, which promotes their rearrangement to form saltbridges with specific negatively charged residues in the post-fusion structure.

In this chapter the experimental evidence for this hypothesis is reviewed.

41

4.2 Activation of class I fusion proteins

4.2 Activation of class I fusion proteins

The prototypical class I fusion protein is the influenza virus protein HA. The X-ray crystalstructures of the three different forms of the protein, the precursor form, the pre-fusion form,and a proteolytic fragment of the post-fusion form, have been determined [Bullough et al.1994; Chen et al. 1998; Skehel & Wiley 2000; Wilson et al. 1981]. In the crystal struc-ture of the pre-fusion form of HA, HA1 residue His184 and HA2 residues His106, His142and His159 are located in the vicinity of positively charged residues; their locations withinthe structure are highlighted in Figure 4.1A. Not only are these histidine residues them-selves highly conserved among the 16 different influenza serotypes, but so are a number ofneighbouring residues in the pre-fusion structure. His142 forms salt bridges with Asp86 andAsp90. In the post-fusion conformation both side chain nitrogen atoms of His106 may actas donors in hydrogen bonds with neighbouring glutamic acid residues (Figure 4.1A). Al-though His106 is not strictly conserved, the sequence alignments shown in Table 4.1 suggestthat HA2 residue His111 may substitute for the role of His106 in some serotypes. Note thatin the crystal structures only fragments of the viral fusion proteins are resolved. In the crys-tal structure of the post-fusion form His184 of HA1 and His159 of HA2 were not resolved,therefore their locations and environments are not known.

The role of histidines and their protonation may be viewed in two contexts:

1. In their neutral, singly protonated form, histidines interact strongly with the posi-tively charged residues found in their vicinity in the pre-fusion form, usually throughhydrogen-bonds. This will effectively lock the structure into the pre-fusion form, untilthe histidines become doubly-protonated. The proximity to positively charged residueswill make protonation more difficult by lowering the effective pKa value of the histi-dine and increasing the initial barrier to activation.

2. Once the histidines are doubly-protonated, their newly acquired charge will interactwith the environment and lead to the formation of new salt bridges with specificnegatively-charged residues. As illustrated in Figure 4.2, the formation of such saltbridges will stabilise the doubly protonated form, resulting in an increase in the pKa ofthe histidine, and possibly making the change effectively irreversible [Tanford 1970;Warwicker 1989, 1992].

Previously it was proposed that histidine residues play important roles in the structural tran-sitions of HA [Chen et al. 1998; Stevens et al. 2004]. An antigenic selection experiment ofHA identified His142 in HA2 [Nakajima et al. 2007] as a critical residue for the binding, anda computational analysis identified regions surrounding several other histidines, as energet-ically critical for the overall stability of the protein [Isin et al. 2002]. In HA1 the mutation

42


of His17 to glutamine or arginine resulted in an increase in the pH of fusion [Daniels et al.1985]. Note that whereas His17 is only partially conserved, it is located in the vicinity ofseveral other conserved histidines at positions 18 and 38 in HA1 and at 106 and 111 in HA2,suggesting that these residues may regulate the structural transitions cooperatively.

4.3 Activation of class II fusion proteins

The prototypical class II fusion protein is the dengue virus E protein (Fig. 4.1B). In thestructure of the pre-fusion dimer [Modis et al. 2003], the highly conserved histidine residuesHis244, His261 and His317 were located in the vicinity of positively charged residues, high-lighted in Figure 4.1B. His261 and His317 form conserved salt bridges in the post-fusionstructure. His244 does not form a salt bridge in the post-fusion structure of the dengue vi-ral E protein, but in the tick borne encephalitis (TBE) viral E protein the equivalent residueHis248 does form a salt bridge with Asp253. These histidine residues are located at molec-ular interfaces, His317 at the interface between domains I and III, and His244 and His261 atthe dimer interface — at interfaces which undergo extensive reorientation during conversionto the post-fusion structure. As can be seen in the sequence alignments in Table 4.1B, theresidues that make up the local environment of these three histidines are also conserved in awide range of flaviviruses.

In the alphaviral Semliki Forest virus (SFV) envelope protein E1, residues His3 andHis125 also fit the pattern observed for the proposed key histidines in the fusion proteinsfrom influenza viruses and flaviviruses. It is possible that other histidines play similar rolesin fusion activation, but their identification has to wait until more complete structures becomeavailable (e.g. the pre-fusion form of the E1-E2 complex of SFV). In SFV the substitutionHis230Ala abrogated membrane fusion [Chanel-Vos & Kielian 2004]. While this residue isstructurally analogous to His244 in the dengue virus E protein, its role in fusion was not clearfrom the available structures of the SFV E1 protein. Residues His146 and His323 in the TBEvirus, which were found to be equivalent to the respective residues His144 and His317 in thedengue virus E protein, have been proposed as allowing the breakage of domain I/domain IIIcontacts at the initiation of the conformational changes leading to fusion [Bressanelli et al.2004]. The environment of His144 in the dengue virus E protein is consistent with thatof other presumably important histidines discussed above; however, this residue formed ahydrogen bond with a negatively charged residue already in the pre-fusion structure.

Critical interactions such as the salt bridge Arg9-Glu368 could be experimentally probedin model systems through site-directed mutagenesis of the proposed residue partners, fol-lowed by examination of the fusion phenotype. The pre-fusion and the post-fusion structureof the dengue virus E protein suggest that the Arg9–Glu368 salt bridge may act as a “linch-

43

4.3 Activation of class II fusion proteinsTa

ble4.1Se

quence

align

ments

ofA)

in�u

enza

hemagglu

tinin

HA,a

ndB)

�aviv

irale

nvelo

peprotein

sEsequ

ences.

Fort

heleg

endandtableBseenext

page.

44


Table4.1continued:

Sequ

ence

align

ments

ofA)

in�u

enza

hemagglu

tinin

HA,a

ndB)

�aviv

irale

nvelo

peprotein

sEsequ

ences.

Thesequ

encesw

ere

align

edin

their

entirety,

thetablelists

selec

tedresid

ueslocatedwithin

0.5nm

ofthesid

echain

nitro

genatom

sof

selec

tedhistidines.A)

Sequ

ence

align

ments

ofHA

1andHA

2sequ

encesfrom

16in�u

enza

virus

serotypes[Ph

ipps

etal.

2004

].Se

quencesfor

which

anX-raycrystalstru

cturehasb

een

solve

dareindicatedby

their

Protein

Data

Bank

ID:1

HGJ,

from

theA/

Aichi/2

/196

8(H

3N2)

isolat

e[Sautere

tal.19

92];1T

I8,t

urkeyH7

[Russell

etal.

2004

];1JSM

,avia

nH5

[Haet

al.20

02];1R

VT,1

930-sw

ine/hu

man

H1[Gam

blin

etal.

2004

];1JSD

,swine

H9[Haet

al.20

02].

Thelas

trow

indicatestherespectiv

esubu

nit,

HA1or

HA2,

oftheheterodimer.B)

Sequ

ence

align

mentof

twelv

e�a

vivira

lEprotein

sequ

ences:

from

thefour

deng

uevir

usserotypes,

DEN1

,DEN

2,DE

N3andDE

N4,a

ndclo

sely

relat

ed�a

viviru

ses,

WestN

ilevir

us(W

N),S

t.Lo

uisenceph

alitis

virus

(SLE

),yellow

feverv

irus(

YF),

MurrayVa

lleyenceph

alitis

virus

(MVE

),Ku

njin

virus

(KUN

),tic

k-bo

rneenceph

alitis

virus

(TBE

),andJapanese

enceph

alitis

virus

(JE).X

-raycrystalstru

ctures

ofE:

1OAN

,DEN

2[M

odiset

al.20

03];1U

ZG,D

EN3[M

odiset

al.20

05];1S

VB,T

BE[Rey

etal.

1995

].Th

elas

trow

indicatest

herespectiv

esubu

nit,Aor

B,of

theho

mod

imer.C

olour

legendforr

esidue

types:

magenta

�histidines;red

�negativ

elycharged;

blue

�po

sitive

lycharged;

orange

�po

lar;b

lack�no

n-po

lar;g

reen

�arom

atic.

45

4.4 Environment of histidine residues

pin”, maintaining the structure in the pre-fusion state.1 The conclusion is that once thisinteraction is lost, the conformational transition from pre- to post-fusion state occurs spon-taneously. It also has to be noted that the process is complex and aided by additional factors;for example liposomes are required in addition to low pH for the transition of dengue andTBE E protein ectodomains into the trimeric state [Modis et al. 2004; Stiasny et al. 2002]).

4.4 Environment of histidine residues

The propensity of specific histidines to be doubly protonated depends strongly on the localenvironment, including the proximity of proton-donating groups and the degree of hydropho-bic shielding from polar solvents [Tanford 1970; Warwicker 1989, 1992]. This provides ameans by which the protonation of critical histidines (and hence the triggering process) canbe modulated. While making protonation more difficult, the presence of positive chargeswill provide a repulsive force driving conformational change, once the histidine side chainis doubly protonated. An analogous model has been proposed for the mechanism of vi-ral uncoating [Warwicker 1989, 1992]. The degree of accessibility to the solvent will alsomodulate the effect of neighbouring positive charges on the pKa. In general, the less thehistidine side chain is accessible to the solvent, the stronger will be the effect of the proteinenvironment.

To determine whether the protonation of histidines could be the critical step in the struc-tural transition from the pre- to the post-fusion conformation, molecular dynamics (MD)simulations were performed of the pre-fusion dimer of the dengue virus sE protein. Twosystems were simulated, in which the histidine residues were either all singly- or all doubly-protonated, labelled HIS0 and HIS+ respectively. The results of these simulations are pre-sented in Chapter 5. To further examine the specific roles of His244, His261 and His317,simulations were performed in which combinations of these residues were selectively pro-tonated Ragothaman & Kobe [2006–7]. To test the mechanical side-chain function of thesehistidines, another series of simulations of a triple mutant was performed in which these his-tidine residues were mutated to alanine Ragothaman & Kobe [2006–7]. It was found thatwhen only a subset of histidines were either protonated or mutated to alanine, no significantstructural changes were observed within 20 ns of simulation. While this result does not in-dicate whether the specific residues identified can induce the conformational change, it wasmost likely due to the limited timescale of the simulation.

1 linch-pin: “a pin passed through the end of an axle-tree to keep the wheel in its place” [Oxford EnglishDictionary, Second Edition 1989].

46


4.5 Conclusion

The evidence presented in this chapter suggests that in a wide range of pH-activated viralfusion proteins, the initial conformational changes associated with the transition from thepre-fusion to the post-fusion form are triggered by the protonation of a small number ofconserved histidine residues. This conclusion is based on the analysis of structures of class Iand class II viral fusion proteins, and implies that in both classes the activation of the fusionprotein is triggered by an analogous molecular mechanism. Specifically, the model proposedhere stands in contrast to a general mechanism where conformational change is induced byan increase in surface charge. In the pre-fusion form of the protein, the specific histidines arelocated adjacent to positively charged residues. Protonation of these histidines would leadto their expulsion and the subsequent formation of specific salt bridges, which could be thereason that the transformation is essentially irreversible. The molecular surfaces involvedin the corresponding structural rearrangements leading to fusion are highly conserved andmight thus provide a suitable common target for the design of antivirals that could be activeagainst a diverse range of pathogenic viruses. The essential role of histidine residues mayeven extend to proteins such as the glycoprotein G from Vesicular stomatitis virus, a memberof the Rhabdoviridae family, although the structural transition in this protein is reversible[Carneiro et al. 2003]. The structure of the low-pH form determined very recently suggeststhat upon deprotonation a number of residues including histidines would have a destabilisingeffect [Roche et al. 2006].

As many of the viruses for which membrane fusion is induced at low pH are signifi-cant pathogens, the surfaces associated with the histidine-triggered structural rearrangementsmight represent important potential target sites for antiviral compound design [Hoffman et al.1997; Luo et al. 1997]. Indeed the conserved triggering mechanism of viral membrane fu-sion may open the possibility for the design of antiviral compounds with a broad spectrumof efficacy.

47

The effect of histidineprotonation on thedengue viral envelopeprotein ectodomain sE 5Chapter 4 introduced the hypothesis that histidine protonation is responsible for triggeringthe conformational change of pH-dependent viral fusion proteins. To test this hypothesis,molecular dynamics (MD) simulations were performed of the pre-fusion form of the dengueviral type 2 (DEN2) envelope glycoprotein soluble ectodomain sE dimer. The results of thesesimulations are presented in this chapter. Furthermore, the structures available of the DEN2E protein were analysed to define the conformational change in terms of interactions betweenresidues, domains and the subunits. In the simulations significant structural changes occurredthat are related to the conformational change to post-fusion state of the protein. Most notableare the disruption of the salt bridge Arg9-Glu368, the partial separation of the dimer, and anincrease in the exposure of the fusion peptide. The results indicate that protonation of thehistidine residues is sufficient to induce specific structural changes in the sE ectodomain ofthe E protein otherwise observed at low pH. The conclusion from the results presented inthis chapter is that the protonation of specific histidine residues induces the activation offlaviviral E protein.

5.1 Introduction

The conformational change associated with fusion activation involves the disassembly of theE protein homodimer, the refolding and reorganisation of the subunits on the viral membrane,

49

5.1 Introduction

and the subsequent irreversible assembly of three monomers into a homotrimer that extendsas a spike from the viral surface (Fig. 3.2B) [Allison et al. 1995; Modis et al. 2004; Wengler& Wengler 1989]. One subunit is approximately 500 residues large. Two identical subunitsform a dimer, the pre-fusion state of the flaviviral E protein (Fig. 3.2A) [Stiasny & Heinz2006]. The soluble ectodomain sE is a large fragment of the E glycoprotein and containsdomains I–III. In the dengue virus type 2 (DEN2) sE comprises residues 1–394 of E. Indengue virus and in tick-borne encephalitis virus (TBEV) the dissociation of the sE dimer isreversible [Allison et al. 1995; Modis et al. 2004; Stiasny et al. 1996]. The pre-fusion dimerwas characterised as a meta-stable fold [Allison et al. 1999; Bressanelli et al. 2004], whereasthe trimeric complex of the post-fusion state was shown to be stable and resistant to changesin pH in the range pH 5–9 [Modis et al. 2004].

Based on the structural similarities and the common pH-dependency in all flaviviruses,alphaviruses and influenza viruses known, the envelope proteins of these viruses are ex-pected to show similar structural responses to acidification [Kampmann et al. 2006; Strauss& Strauss 2001]. The fact that pH-induced fusion occurs close to the pKa of histidine pro-tonation in water strongly suggests that the process is triggered by the protonation of oneor more histidines. The local electrostatic changes after histidine protonation might lead tothe structural rearrangements that result in the disassembly of the E protein dimer and theformation of the trimer [Kampmann et al. 2006]. In the TBE viral E protein the protonationof two strictly conserved histidines, His146 and His323, was proposed to be an importantfactor in the destabilisation of the domain I–III interface [Bressanelli et al. 2004]. These areequivalent to His144 and His317 in dengue virus. The optimal pH for fusion in DEN2 wasdetermined by the titration of “fusion from within” (FFWI) infected mosquito cells, whichindicated a pKa of 6.9 (Fig. 5.1) [Guirakhoo et al. 1993]. In another study of cell fusionby DEN2, the optimal pH of fusion ranged between pH 5.0 and 6.5 (±0.25) [Randolph &Stollar 1990].

Low pH also plays a role in the enzymatic cleavage of the flaviviral precursor membraneprotein envelope protein heterodimer prM-E in the trans-Golgi network (TGN), which is anessential step in the maturation of the virus before it is released from the host cell (Fig. 5.2)[Allison et al. 1995; Li et al. 2008]. The prM-E dimer contains the envelope protein E incomplex with the precursor membrane protein prM. In the pH-dependent step required forthe interaction with the furin protease, the prM-E protein rearranges on the viral surfacereversibly from trimers to dimers [Allison et al. 1995; Stadler et al. 1997]. In the denguevirus this conformational change occurs at pH 6.0 [Yu et al. 2008], suggesting that, again,the protonation of histidine residues may play a role. This is followed by the transportfrom the TGN to the extracellular matrix inside secretory granules, in which the virus isexposed to even lower pH ≈ 5.7 [Paroutis et al. 2004]. After the cleavage by furin thepr peptide remains bound to the fusion peptide, where it is thought to protect segments of

50

5 MD of the dengue viral E protein after histidine protonation

Figure 5.1 Titration curves of A) histidine protonation, and B) fusion from within mosquitocells infected with dengue virus type 2. B) The fusion index FI was determined by calculatingFI = 1− (number of cells/number of nuclei). The number of cells and number of cell nuclei werecounted in �ve microscopic �elds. Original image in A) from Nelson & Cox [2000] and in B) byGuirakhoo et al. [1993].

51

5.1 Introduction

Figure 5.2 Model of the �avivirus maturation pathway. A) The conformational changes of thevirus particles in the secretory pathway. Immature particles bud into the endoplasmic reticulum(ER) as spiky virions and are transported through Golgi into the trans-Golgi network (TGN), whereacidi�cation induces a conformational change of the virion. Furin cleavage takes place in the TGN,and pr remains associated until the virion is released to the extracellular milieu. The approximateluminal pH values of the speci�ed cellular compartment are indicated [Paroutis et al. 2004]. B)Con�guration of the glycoproteins on the surface of the virion during maturation. The structureof the E protein in the secretory pathway is largely unchanged, except for movements at the hingebetween domains I and II. In contrast, the oligomerisation states of the glycoproteins are criticallydependent on pH. The fusion loops (residues 98�111) are indicated by red stars. Original imageand legend by Yu et al. [2008].

the E protein from becoming protonated, thereby preventing a conformational change andpremature membrane fusion. Two X-ray crystal structures of the DEN2 prM-E heterodimerhave been solved [Li et al. 2008], providing insight into the interactions and interfaces of thecomplex in the immature virus (Fig. 5.2B). Specifically, in the low pH of the Golgi complexthe pr peptide is thought to protect some segments of the E protein from protonation. [Yuet al. 2008] Therefore shielding of a pH-sensitive residue of E by the pr peptide may indicatethat the protonation of this residue triggers the activation of the protein. The last step in thematuration of the E protein is the dissociation of the pr peptide from the dimeric E proteinand requires neutral pH [Yu et al. 2008]. Until the pH is raised the pr peptide remains boundto the fusion peptide. Premature cleavage of the prM-E complex by furin was indeed relatedto fusion inactivation [Guirakhoo et al. 1992; Holbrook et al. 2001; Pryor et al. 1998; Yuet al. 2008]. Fusion inactivation of the envelope protein was observed in a number of virusfamilies: in the flaviviruses dengue and Murray Valley encephalitis virus, the alphavirus

52


Figure 5.3 Diagrams of the ectodomain and the membrane domains of the dengue viral envelopeproteins E and M. The volume occupied by the ectodomain of an E monomer is pink (domainI), yellow (domain II) and blue (domain III). The blue and orange cylinders are the stem andanchor helices of E and M, respectively. Helices of the E stem region are identi�ed by E-H1, E-H2,the transmembrane anchor region of E by E-T1, E-T2, the stem region of M by M-H and thetransmembrane anchor region by M-T1, M-T2. CS represents the conserved sequence betweenE-H1 and E-H2. a) Side view. b) Top view of the membrane domain helices with the superimposedE ectodomain homodimer. Original image by Zhang et al. [2003].

Semliki Forest virus, and influenza virus [Doms et al. 1985; Guirakhoo et al. 1992; Kielian& Helenius 1985].

Six X-ray crystal structures of the DEN2 sE ectodomain have been solved: three pre-fusion structures (PDB IDs: 1OAN, 1OKE [Modis et al. 2003], 1TG8 [Zhang et al. 2004]),one post-fusion structure (1OK8 [Modis et al. 2004]), and two structures of the prM-Ehetero-complex at neutral and at low pH (3C6E, 3C5X [Li et al. 2008]). Crystal struc-tures have been solved of other flaviviral E proteins, from dengue viral serotypes 1 and 3,TBEV, West Nile virus and yellow fever virus. Entire viral envelopes have been modelledby cryo-electron microscopy (cryo-EM). One of these models includes the structures of themembrane domains of the E and the M protein (1P58 [Zhang et al. 2003], Fig. 5.3).

In this study the interactions of the histidine residues and their micro-environments inthe DEN2 envelope protein ectodomain sE were analysed, to determine whether the proto-nation of specific histidines triggers the low-pH-dependent conformational change. Previoussequence alignments indicated that some of the histidine residues in the DEN2 protein arefully conserved (see Chapter 4 [Kampmann et al. 2006]). To determine the structural con-servation of the less conserved histidine residues and their environments, homology modelsof other flaviviral E protein sequences were analysed.

Experiments lack the structural and temporal resolution necessary to investigate the role

53

5.2 Methods

of specific residues in the pH-dependent activation of a protein. Here, molecular dynamics(MD) simulations of the DEN2 envelope protein soluble ectodomain sE were performed tomodel the effect of histidine protonation on a mature flaviviral envelope protein in dynamicand atomic detail. X-ray crystal and cryo-EM structures of the DEN2 E protein were an-alysed to characterise the pre-fusion and post-fusion conformations of the protein, to serveas references for the evaluation of the simulations. The pKa values of the histidine residueswere calculated for various conformations and configurations from the simulations to deter-mine the protonation states of sE. The simulations of the pre-fusion dimer were analysedfor changes in specific interactions in the protein, to elucidate the initial stages in the pro-tonation-dependent activation. A series of events were associated with histidine protonation.I propose specific functions for a number of histidine residues. In particular His144 andHis317 were identified as critical residues in the low-pH-dependent fusion activation of fla-viviral fusion proteins.

5.2 Methods

MD simulations

The starting structure for the MD simulations was taken from the X-ray crystal structure1OKE [Modis et al. 2003] of the pre-fusion form of the dengue viral type 2 envelope proteinectodomain sE. Non-protein compounds (i.e. detergents, sugars and water) present in thecrystal structure were not included in the simulation model. Two systems were simulated,one at neutral pH and one at slightly acidic pH. The two pH conditions “neutral” and “slightlyacidic” were mimicked by altering the protonation states of the histidine residues of theprotein. In the neutral system all the histidine residues were uncharged, i.e. singly-proton-ated at the Nτ-atom1, also termed the Nε-H tautomeric form [IUPAC 1975], while for theslightly acidic condition all the histidine residues were charged, i.e. doubly-protonated. Inthe following the two systems will be referred to as HIS0 (all histidines singly-protonated)and HIS+ (all histidines doubly-protonated) respectively. In both the systems the arginine,lysine, aspartic and glutamic acid residues were charged.

The structure of the protein was placed in a rectangular, periodic box at a distance of atleast 0.9 nm from the sides of the box, and solvated with 25 969 water molecules. Initiallythe two systems, HIS0 and HIS+, contained approximately 86 000 atoms. Each systemwas examined periodically to ensure that the rotation of the protein, or the conformational

1 The nitrogen atoms of the imidazole ring of histidine are denoted by pros (“near”), abbreviated π, and tele(“far”), abbreviated τ, to indicate their position relative to the side chain [IUPAC 1975].

54


changes did not lead to interactions between the periodic images of the protein. When anyatom of the protein came within 1.4 nm of a periodic image of the protein, the box dimensionswere increased to restore the minimal distance of 0.9 nm to the box wall, and water moleculeswere added to the system to fill the new volume. The largest system was solvated with100 064 water molecules and contained ≈ 308 000 atoms in total.

The simulations were performed using the GROMACS software package version 3.2[Lindahl et al. 2001]. The protein molecules were described using the GROMOS96 forcefield version 43A1 [van Gunsteren et al. 1996], in which the hydrogen atoms of aliphaticgroups are treated as united atoms with the carbon atoms to which they are covalentlybound. For the water molecules the simple point charge, SPC, model was used [Berend-sen et al. 1981]. The equations of motion were integrated by use of the leapfrog algorithmwith a time step of 2 femtoseconds. Non-bonded interactions were evaluated by use of atwin-range cutoff scheme. Lennard-Jones interactions and coulombic interactions within theshort-range cutoff of 0.8 nm were evaluated every step, while longer-range coulombic inter-actions of 0.8–1.4 nm were updated every ten steps together with the pair list. The truncationof electrostatic interactions beyond the long-range cutoff was corrected by use of a reac-tion field correction, with a relative dielectric permittivity of εRF = 80 [Tironi et al. 1995].Bond lengths within the protein were constrained using the LINCS algorithm [Hess et al.1997]. To maintain constant temperature and constant pressure, a Berendsen thermostat anda Berendsen barostat were applied [Berendsen et al. 1984]. The protein and water were cou-pled independently to an external heat bath of 300 K with a relaxation time of 0.1 ps. Thepressure was maintained at 1 bar with a relaxation time of 0.8 ps and an isothermal compress-ibility of 4.5×10−5 bar−1. Configurations for analysis were sampled every 2 picoseconds.

Analysis

Root mean square deviation (RMSD) The positional RMSD of a configuration of theprotein was determined after performing a least squares fit to a subset of the protein atoms,e.g. the backbone atoms, of the reference structure, e.g. the initial configuration s = r(0).The RMSD is defined as

RMSD(t) =

√1N

N

∑i

[ri(t)− si

]2

with ri(t) the position of atom i at time t, and N the number of atoms analysed. RMSD valueswere determined for the dimer and for every subunit (i.e. the monomeric units) and domain.Short term fluctuations within 1 ns were smoothed by a 500 point-running average.

The positional RMSDx(t) of each Cα-atom x with respect to its initial position sx =rx(0) was calculated to identify regions of different mobility, e.g. hinge modules, or domain

55

5.2 Methods

rotations. Here, rx(t) is the position of atom x at time t and N = 1. Single configurationsat 60 ns in HIS0 and at 70 ns in HIS+ were analysed, to select the low frequency, globalstructural modes. The subunits were analysed separately. To include the degree of freedombetween domain II of a subunit and domain III of the respective other subunit, the structuralunits analysed contained a subunit and domain III of the respective other subunit.

Solvent accessible surface area (SASA) SAS areas were calculated for every atom in theprotein using the algorithm of Eisenhaber et al. [1995], with a probe sphere 0.14 nm indiameter. The SASA per residue was obtained by summation of the SASA of the atoms.

Molecular contacts All pairs of atoms the centres of which were 0.6 nm or less apart wereconsidered to be in contact. As one atom can form multiple contacts, in addition the netnumber of atoms involved in interfacial contacts, termed “contact atoms”, were counted toquantify the extent of an interface. The number of contacts and the number of contact atomswere determined between the domains of a subunit and between the subunits of the dimer, inconfigurations sampled every 100 ps.

Eletrostatic interactions Coulombic interactions between atoms X and Y of two aminoacid side chains were determined for interaction distances ≤ 0.5 nm, with X the heavy atomof the functional group of the side chain of any aspartic acid, glutamic acid, lysine, arginineor histidine residue, and Y the heavy atom of the functional group of the side chain of anycharged or polar residue or the main chain. An attractive coulombic interaction between theside chains of charged residues was considered a salt bridge if the interaction distance was< 0.4 nm [Barlow & Thornton 1983; Elcock 1998].

Hydrogen bonds were identified based on geometric criteria [Baker & Hubbard 1984],considering any side chain or main chain donor or acceptor groups. The criteria were 1) amaximum distance of 0.25 nm between the hydrogen atom and the acceptor atom, and 2) aminimum of 90◦ for the donor–hydrogen–acceptor angle. The hydrogenated nitrogen atomsof any amide, amine or imidazole group, and the oxygen atoms of any hydroxyl group wereconsidered potential hydrogen donors. The nitrogen atoms of any tertiary amine or non-hydrogenated imidazole group, and the oxygen atoms of any carboxyl or carbonyl groupwere considered potential hydrogen acceptors. The donors and acceptors are given in theformat R.a, with R the residue type, and a the atom type according to the IUPAC nomen-clature of α-amino acids, e.g. N-2 and O-1 for functional groups of the main chain [IUPAC1975]. When the donor or acceptor atom is not named explicitly, the side chain donor oracceptor atom is implied.

56


Secondary structure The protein secondary structure was determined according to theDSSP definitions [Kabsch & Sander 1983].

Spatially conserved amino acid substitution To determine if an unconserved histidineresidue was spatially or functionally conserved by a compensatory mutation in a similarmicro-environment, the position of the mutation was examined in homologous protein struc-tures to see whether it compared to the position and the structural relevance of the originalhistidine. Any residue within a radius of 0.5 nm of the side chain of a histidine residue wasconsidered part of that histidine’s environment. At a distance of 0.5 nm the coulombic forcebetween two charged atoms such as an oxygen and a nitrogen atom is about 38 % of themaximal force at van der Waals contact distance. The X-ray crystal structure 1OKE of thepre-fusion form of the DEN2 sE protein served as the structural template for the homologymodels of the micro-environments in eleven other flaviviral envelope proteins. These modelswere obtained by substituting a specific residue in the structure according to the sequencealignment with the DEN2 sE protein (gi|31615787) [Kampmann et al. 2006].

Calculation of the pKaKaKa of titratable residues The pKa values of titratable groups in struc-tures and configurations of the DEN2 E protein were calculated by use of a method based ona Poisson-Boltzmann approach to classical continuum electrostatics [Bashford & Karplus1990; Gordon et al. 2005a]. The electrostatic potential was estimated with the Poissonequation. The probability of a particular protonation state was determined by integratinga Boltzmann-weighted sum over all the possible protonation states at different pH values.The apparent pKa was calculated for histidines and for arginine, lysine, aspartic acid, glu-tamic acid, and tyrosine residues positioned within 1.5 nm of any histidine residue, in thecrystal structures of the DEN2 sE protein and in configurations from the simulations of thesE protein. The calculations were processed using the automated H++ tool, accessed via aweb server [Gordon et al. 2005b]. The calculation parameters were: solvent dielectric con-stant εo = 80, protein-internal dielectric constant εi = 20, pH 6.5, and saline concentrationc = 0.15 M.

5.3 Results

The pre-fusion and the post-fusion conformation

X-ray crystal structures of the pre-fusion and the post-fusion form of the DEN2 E proteinwere analysed to determine the specific interactions that characterise the two conformations.

57

5.3 Results

Specifically salt bridges and hydrogen bonds across the subunit interfaces in the dimer andthe trimer and at the interface of domain III were analysed and examined for the involvementof histidine residues.

Subunit interfaces in the dimer and the trimer

Two crystal structures of the pre-fusion sE dimer were examined (PDB IDs: 1OKE, 1OAN[Modis et al. 2003]). The number of contacts between the subunits was determined to detectthe possible disassembly of the subunits. In the crystal structures 1OKE and 1OAN thenumber of contacts up to a distance of 0.6 nm between the subunits was 2 305 and 2 054respectively. Despite the symmetry of the subunits in the homodimer, the number of inter-subunit contacts varied by up to 8.2 % between some corresponding domains of the twosubunits in 1OKE; in 1OAN the agreement between the subunits was better. The number ofcontact atoms was analysed to determine the size of the subunit interface. The number ofcontact atoms in one subunit was ≈ 350 in 1OKE and ≈ 300 in 1OAN.

The subunit interfaces of the two crystal structures were analysed for specific electro-static interactions. Because the subunits of the dimer are rotamers arranged anti-parallel andcyclic-symmetrical, the interactions could occur pairwise. Table 5.1 lists the inter-molecularsalt bridges and hydrogen bonding interactions across the subunit interface of the pre-fusiondimer.2 Four pairs of hydrogen bonds and one pair of salt bridges were found across thesubunit interface in both pre-fusion crystal structures 1OAN and 1OKE. A fifth pair of hy-drogen bonds and a second pair of salt bridges was present only in 1OKE, thus some localinteractions differed between the two crystal structures. In the pre-fusion structures only onehistidine, His261, interacted across the subunit interface.

One crystal structure was available of the post-fusion sE trimer (PDB ID: 1OK8 [Modiset al. 2004]). In this structure the subunits lie parallel to each other, resulting in an asym-metrical subunit interface. Seven salt bridges and fourteen hydrogen bonds interacted acrosseach trimer interface. Arg9 in particular was a central component in a network of salt bridgesand hydrogen bonds. The Trp20 residues of all three subunits interacted closely with eachother, forming a ring at the base region of the trimeric complex formed by the domains III.Only one of the histidines, His317, interacted across the subunit interface, and its backboneformed a hydrogen bond with Gln167.Nε. His209 was located at the subunit interface, closeto domain III, and interacted with Glu133 of another subunit and with the C-terminus Lys394.Due to the oligomeric rearrangements of the conformational change, the subunit interactionsare expected to differ between the dimer and the trimer. Indeed, Asp98 was the only residueinvolved in specific electrostatic interactions across the subunit interfaces in both the dimer

2 As this study investigates the effect of histidine protonation on the pre-fusion structure of the E protein,the focus is on this structure and only selected data are presented for the post-fusion structure.

58


Table 5.1 Hydrogen bonds (H) and salt bridges (SB) across the subunit interface and the domainIII interface of the dengue viral sE protein dimer, in the pre-fusion crystal structures 1OAN and1OKE [Modis et al. 2003]. The donor and acceptor atoms are given in the format 〈residue〉.〈atomtype〉.

X-ray diffr. MD simulation

HIS0 HIS+

subunit A subunit B subunit A subunit B

donor acceptor 1OAN 1OKE i ii i ii i ii i ii

subunit interface

Ser7.N-2 Asp98.Oδ H H 66 72 87 78 56 75 77 35Ser7.Oγ Asp98.Oδ H 87 82 71 52 69 77 78 47Lys110.Nζ Ser7.Oγ H H 2 0 12 8 8 0 3 8Lys241.Nζ Glu269.Oε SB 21 3 47 10 3 3 11 4Lys246.Nζ Glu44.Oε SB SB 19 0 0 0 1 0 0 0Ser255.N-2 Glu257.Oε H H 46 10 88 68 8 0 87 16His261.Nπ Leu253.O-1 H H n.a. n.a. n.a. n.a. 1 0 1 0

domain III interface

Arg9.Nω Glu368.Oε SB SB 98 99 100 100 22 2 30 4Thr40.Oγ Ile352.O-1 H H 98 93 99 99 63 89 33 0Trp299.N-2 Gly296.O-1 H H 32 28 11 23 18 28 20 1His317.Nτ Ser7.O-1 H H 4 0 27 48 0 0 4 11Lys334.Nζ Gly296.O-1 H H 8 21 9 27 3 3 0 0Arg350.Oω Glu13.Oε SB SB 48 36 39 43 21 18 0 0Thr353.Oγ His144.Nπ H H 18 0 3 0 0 0 0 0Asn355.Nδ Lys295.O-1 H H 8 0 18 5 2 0 0 0Asn366.N-2 Gly146.O-1 H H 2 0 1 0 94 15 28 0

and the trimer: Ser7-Asp98 in the dimer, and the forked main chain-side chain hydrogenbonds Thr76.N-2–Asp98–Gly78.N-2 in the trimer complex. The salt bridge Asp98-Lys110was present in both conformations, stabilising the base of the fusion loop (residues 98–111).

In the viral envelope the E dimers are closely packed, therefore interactions betweenneighbouring dimers might play a role in the conformational change. To determine whetherhistidines were involved in such interactions, the cryo-EM structure of the mature envelopeassembly of pre-fusion E protein was analysed (PDB-ID: 1THD [Zhang et al. 2004]). In thisstructure His346 was the only histidine located at the interface between two neighbouringdimers. However, His346 may be located at three different interfaces depending on theposition of the respective dimer within the icosahedral assembly of the envelope. Therefore

59

5.3 Results

His346 does not play a distinctive role at the inter-dimer interface.

Domain III interface

Due to the relocation of domain III and the different subunit arrangements in the dimer andthe trimer, the pre-fusion-specific interactions at the interface of domain III are broken duringthe conformational change to the post-fusion structure (Fig. 3.2) [Modis et al. 2004]. Thepre-fusion crystal structures were analysed for electrostatic interactions between domain IIIand domains I and II to determine the conformation-specific interactions at the domain IIIinterface. The interactions are listed in Table 5.1. Only domain I of the same subunit formedspecific salt bridge and hydrogen bonding interactions with domain III (Tab. 5.1). In the pre-fusion structure there were two salt bridges, Arg9-Glu368 and Arg350-Glu13, at the domainI–III interface. Two histidine residues, His144 of domain I and His317 of domain III, formedspecific interactions across the domain I-III interface (Tab. 5.1). In the post-fusion structureHis317 was buried against domain I of a neighbouring subunit and His144 was exposed onthe outer surface of the trimer. Thr353 is buried between domains I and III, and coordinatesinteractions between these domains in both conformations. In the pre-fusion conformationThr353 interacted with His144 and Glu368, and in the post-fusion conformation with Glu26and His282.

The interaction profiles of His317 of domain III were conformation-dependent, and theinteractions of the pre-fusion structure were replaced by other interactions in the post-fusionstructure. In the pre-fusion structure Ser7 of domain I formed a junction between domainsI, III and the other subunit of the dimer, via hydrogen bonds with His317 and with Asp98of the fusion loop of the other subunit (Tab. 5.1). Both these interactions are lost duringthe conformational change. In the post-fusion structure His282 and His317 were located atthe domain I–III interface and both interacted with Glu368. His317 formed a domain III-internal salt bridge with Glu368, replacing the pre-fusion salt bridge Arg9-Glu368. Thus inboth conformations Glu368 formed key salt bridges that define the tertiary structure. In thepost-fusion trimer Arg9 becomes part of the subunit interface and interacts with Glu172 ofanother subunit.

His144 and Asp42 formed hydrogen bonds in both the pre-fusion and the post-fusioncrystal structures. Although in the pre-fusion structure the side chain of His144 was only0.32 nm away from Glu368.Oε, no hydrogen bond formed. However, a hydrogen bond orsalt bridge interaction His144-Glu368 was observed in the crystal structures of the prM-E complex. The absence of the hydrogen bond His144-Glu368 in the pre-fusion crystalstructures could be an artefact related to the pH 8.5 at which the X-ray diffraction data forthe pre-fusion structures 1OAN and 1OKE were collected.

60


The salt bridge Arg350-Glu13 at the domain III–I interface of the pre-fusion structure isdisrupted during the conformational change. In the post-fusion structure Arg350 and Glu13formed new salt bridges, Arg350-Glu136 and Lys291-Glu13. Arg350-Glu136 interacts be-tween domain III and domain I of two different subunits of the trimer. Lys291-Glu13 is anintra-molecular salt bridge within domain I.

pKaaa and solvent accessibility of the histidine residues

The local pKa of the titratable residues in the pre-fusion crystal structures of the sE ectodomainwere calculated to determine their protonation state in the pre-fusion form of sE (Fig. 3.2A).The results for the histidine residues were averaged over the two subunits of the dimer andare listed in Table 5.2. Most of the histidine residues had slightly acidic pKa values, as low aspKa = 5 for His27. Only His94, His149 and His346 had neutral or slightly basic pKa. Con-sidering the pH 8.5 of the X-ray diffraction experiment, all the histidines in the pre-fusioncrystal structures were singly-protonated. His94 was the only histidine residue that had aslightly basic pKa in both pre-fusion crystal structures, and may be doubly-protonated at pH

Table 5.2 Predicted pK a values for the protonation of the histidine residues in the crystal structures1OKE and 1OAN and in the MD simulations of the sE protein dimer in the pre-fusion conformation.The numbers in brackets indicate the standard deviation.

X-ray diffr.a MD simulation

resi- 1OAN 1OKE HIS0 t ≤ 0b HIS0 ∼ 60 nsc HIS+ ∼ 70 nsd

due subunit A subunit B subunit A subunit B

H27 5.2 5.0 5.2 (0.5) 6.6 (0.5) 6.1 (0.8) 6.9 (0.2) 9.3 (0.2)H94 7.8 7.4 6.8 (0.2) 6.2 (0.2) 6.4 (0.4) 7.2 (0.8) 8.1 (0.2)H144 6.6 5.2 4.5 (0.8) 5.7 (0.8) 5.4 (0.7) 9.1 (1.2) 9.7 (0.2)H149 6.6 7.5 6.6 (0.5) 5.5 (0.7) 6.8 (0.2) 6.4 (0.8) 9.5 (1.4)H158 5.6 5.6 5.7 (0.3) 5.7 (0.4) 5.0 (0.9) 9.3 (0.8) 8.5 (1.6)H209 6.4 6.3 6.4 (0.1) 5.8 (0.4) 5.8 (0.5) 8.5 (0.2) 8.4 (0.3)H244 5.1 5.3 5.7 (0.5) 6.3 (0.3) 6.1 (0.7) 7.5 (0.6) 7.8 (0.4)H261 5.5 6.1 4.7 (0.5) 5.2 (0.8) 5.8 (0.6) 8.3 (0.1) 6.9 (0.2)H282 6.5 6.8 6.5 (0.2) 6.1 (0.7) 6.6 (0.3) 8.1 (0.2) 7.1 (0.2)H317 6.2 6.0 5.0 (1.0) 6.5 (0.6) 6.1 (1.3) 8.2 (0.4) 6.5 (0.6)H346 7.0 6.8 6.8 (0.0) 6.3 (0.4) 6.8 (0.1) 8.4 (0.2) 7.2 (0.5)

Values indicate the average of: a the two subunits of the dimer; b three configurations fromthe preparatory MD equilibration of the side chains; c three configurations from simulationHIS0 sampled at random between 50 and 60 ns; d three configurations from simulation HIS+sampled at random between 60 and 70 ns.

61

5.3 Results

7. His27, His144, His158, His244 and His261 had pKa values below the pKa = 6.0 of histi-dine in water. Most values varied between the two crystal structures, with differences up to1.4 pK-units. His27, His144 and His261 were buried and had the greatest desolvation effect.Non-titratable groups in the protein micro-environment, i.e. polar side chain and backbonegroups, contributed significantly to the pKa of His261. The micro-environments of His149,His158, His209, His282 and His317 in the ectodomain of E were slightly hydrophilic andslightly desolvated. The environments of the other histidines were more hydrophilic andcontributed to increases in the pKa of His94, His144 and His346 and to decreases in the pKa

of His27, His244 and His261. Particularly large contributions to the shifts in pKa were dueto the interactions of His94 with Glu79; of His144 with Asp42, Glu368 and the backbone ofThr353; and of His261 with Lys204. Interactions between histidine residues that contributedto their pKa were found among His27, His244 and His282, among His144, His149 andHis158, between His317 and His144 and His244 respectively, and between the two His261in the dimer.

The environment of the histidines was analysed in the cryo-EM structure of the full-length E protein (PDB-ID: 1P58 [Zhang et al. 2003]), which in addition contains the mem-brane domain of the M protein, to determine the interactions with the membrane domainsof E and M (Fig. 5.3). This structure was the only structure available containing the Cα-trace of the membrane domain of E and a fragment of the M protein (PDB-ID: 1P58 [Zhanget al. 2003]). In this structure the Cα-atom of His209 was located 1.4 nm from the Cα-atoms of Arg38 and His39 of M, and His282.Cα 1.0 nm from Asp421.Cα of the stem regionof E. Electrostatic interactions among these residues and with the polar head groups of themembrane lipids might contribute to the local pKa of His209 and His282. Because only theCα-trace of the M-E complex was known, the pKa of His209 and His282 in situ could not becalculated.

The pKa were calculated for the trimeric complex of the post-fusion structure to de-termine the protonation state of the histidine residues after the conformational change. Thecalculation was performed for pH 7, corresponding to the pH in the X-ray diffraction exper-iment of the post-fusion crystal structure 1OK8. In the trimer the pKa of all the histidineresidues except His94 and His209 were acidic and therefore similar to the pKa in the pre-fusion structures. His94 had a pKa ≈ 7.2 and the pKa of His209 had shifted to 8.0. Fromthese results His94 and His209 were predicted to be predominantly doubly-protonated andall the other histidines singly-protonated in the post-fusion crystal structure at pH 7. A com-parison between the pre-fusion and the post-fusion crystal structures showed that only themicro-environments of His94 and His346 were similar in both conformations.

The solvent accessible surface area (SASA) of the histidine residues were analysed inthe pre-fusion and the post-fusion crystal structures to determine whether the conformational

62


change led to changes in solvent exposure; the results are listed in Table 5.3. The SASA perresidue was calculated as the SASA of the side chain atoms. For a fully solvated histidinethis SASA was≈ 1.9 nm2. The average SASA per residue in the pre-fusion crystal structureswas 0.52 nm2 and in the trimeric model of the post-fusion crystal structure 0.47 nm2. Thevalues calculated indicate that in the pre-fusion crystal structure His144 and His261 wereburied, His27, His94, His158, His244 and His317 were partly solvent exposed, and His149,His209, His282 and His346 were exposed. In the post-fusion structure His144, His244 andHis261 became exposed, the solvent exposure of His94 increased by more than 10 %, His209became partly buried, and His282 and His317 fully buried. His144 in the post-fusion crystalstructure was the most exposed histidine residue. The correlation between the SASA and thepKa (Tab. 5.2) in the pre-fusion crystal structures was low, with correlation coefficients of0.21 for 1OAN and 0.54 for 1OKE. Tables 5.2 and 5.3 also include the results for the MDsimulations, which are described in section 5.3 on page 73.

A model for the pH-dependent kinetics of cell fusion was obtained by fitting the datapoints from the fusion index curve FI(pH) in Guirakhoo et al. [1993] (square symbols inFigure 5.5A) to the logistic function

P(pH) = G

(1− 1

1+ exp(− kG(pH−pK)

)( GP(pK) −1

))

(5.1)

using a least squares approach, with the maximum ratio of fused cells G = 0.96 and themidpoint of the curve P(pK) = G/2 = 0.48. The fusion index FI was determined by thenumber of cell fusions

FI = 1− number of cellsnumber of nuclei

(5.2)

[Guirakhoo et al. 1993]. The time-independent3 constant for the pH-dependency of cellfusion k = 6.5 pH−1 and the midpoint of the curve pK = 6.9 (RMSD(P) = 0.035) weredetermined by fitting P (eq. 5.1) to the experimental data FI; the function is plotted in Figure5.5A. The curve indicates that fusion from within (FFWI) mosquito cells by the dengue virustype 2 begins at pH ≈ 7.7 (0.5 % fusion) and maximum cell fusion is reached at pH 6.0.

The titration curves of all the titratable residues were analysed to determine the pHrange in which the side chains are protonated. Titration curves were determined for threeconfigurations of the E dimer, selected at random from the preparatory simulation of HIS0,in which position restraints were applied to the backbone atoms. The curves for the histidine

3 The cell fusion index was determined by counting the number of fused cells 2 hours after acidification at37◦C. As fusion in wild type dengue virus occurs in the order of seconds, the resulting kinetic constant can beconsidered quasi time-independent.

63

5.3 Results

Table5.3So

lvent

accessible

surfa

ceareas(SAS

A,in

nm2 )

ofthehistidinesid

echain

sin

thepre-fusio

ncrystal

structures

1OKE

and1O

AN,a

nddu

ringtheM

Dsim

ulation

softhe

sEprotein

inthep

re-fu

sionconformation

.The

numbers

inbrackets

indicate

thestandard

devia

tion.

X-r

aydi

ffr.

MD

sim

ulat

ionc

resi

-1O

KE

a1O

AN

a1O

K8b

HIS

0H

IS+

due

subu

nitA

subu

nitB

subu

nitA

subu

nitB

H27

0.57

0.48

0.65

(0.1

0)0.

30(0

.10)

0.53

(0.1

1)0.

57(0

.20)

0.57

(0.0

9)H

940.

680.

660.

91(0

.10)

1.00

(0.1

3)0.

76(0

.20)

0.82

(0.1

3)0.

65(0

.12)

H14

40.

200.

221.

22(0

.04)

0.17

(0.0

8)0.

08(0

.05)

0.10

(0.2

1)0.

11(0

.06)

H14

91.

151.

09n.

r.d0.

80(0

.16)

1.06

(0.1

1)1.

10(0

.21)

0.78

(0.1

9)H

158

0.83

0.78

n.r.d

0.86

(0.5

0)0.

97(0

.18)

0.80

(0.2

6)0.

67(0

.14)

H20

91.

191.

010.

68(0

.10)

0.66

(0.1

2)0.

67(0

.10)

0.68

(0.1

7)0.

67(0

.11)

H24

40.

740.

981.

11(0

.07)

1.04

(0.2

8)0.

84(0

.17)

1.40

(0.2

8)0.

67(0

.29)

H26

10.

320.

280.

92(0

.07)

0.34

(0.1

0)0.

44(0

.12)

0.76

(0.1

9)0.

82(0

.17)

H28

20.

981.

100.

15(0

.02)

0.74

(0.1

3)0.

82(0

.11)

0.89

(0.1

7)0.

65(0

.13)

H31

70.

830.

780.

26(0

.04)

0.91

(0.1

2)0.

70(0

.14)

1.26

(0.2

3)0.

75(0

.16)

H34

61.

061.

181.

08(0

.14)

0.79

(0.1

6)1.

09(0

.12)

1.09

(0.1

4)1.

12(0

.14)

Val

ues

wer

eav

erag

edov

erth

esu

buni

tsof

ath

epr

e-fu

sion

dim

er,b

the

post

-fus

ion

trim

er,c

over

Nco

nfigu

-ra

tions

sam

pled

ever

y1

nsfr

omth

etr

ajec

tory

(HIS

0:N

=61

,HIS

+:N

=65

).d

Res

idue

was

notr

esol

ved

inth

ecr

ysta

lstr

uctu

re.

64


Figure 5.4 Titration curves predicted for the histidine residues in the DEN2 sE protein, calculatedusing the H++ software. The curves were calculated for three con�gurations of the E dimersampled at random from the preparatory simulation of the system HIS0 (t ≤ 0). The values wereaveraged over the subunits of the dimer and over the three con�gurations.

residues, averaged over the three configurations, are plotted in Figure 5.4. The protonationstate of all the aspartic acid, glutamic acid, arginine, lysine and tyrosine residues remainedconstant at pH values between 5.7 and 8.1, i.e. throughout the pH range in which the fusionindex increases (Fig. 5.5 and data not shown). Only the protonation of the histidine residuesand the termini overlapped with the pH range of fusion (Fig. 5.4 and data not shown). Thetitration curves of His27, His144, His244 and His317 averaged over the two subunits startedat pH 7.8–7.6. The titration of His94, His149, His158, His209, His282 and His346 startedat pH 8.4–7.8, i.e. above the pH of cell fusion. The titration of His261 began at the slightlyacidic pH 6.8, suggesting that His261 does not trigger fusion activation. In HIS0 after 60 nsof simulation the pH values at which His27, His94 and His317 start to become protonatedshifted to significantly greater values at pH 8.8, 7.6 and 8.2 respectively.

Based on the pH range of protonation, His27, His144 and His317 were consideredpotential candidates for triggering the conformational change. The titration curves of His27,His144 and His317 were multiplied together to estimate the population of E protein, E∗, in

65

5.3 Results

Figure 5.5 Model of low-pH-dependent, dengue viral cell fusion triggered by histidine protonationin the E dimer. A) The experimental fusion index FI (squares) determined previously by Guirakhooet al. [1993], the titration curves of His27, His144 and His317 and their products ∏P (trianglesand circles respectively). B) The logarithmic functions lnP of the curves in A. The �tted functionsare plotted as solid lines, grey symbols indicate the data points selected for the �t. The titrationcurves of the histidines in the E dimer were calculated using the H++ software [Bashford & Karplus1990; Gordon et al. 2005b].

66


which these three histidines i were simultaneously protonated, y = ∏i

Pi. A logistic curve PE∗

was fitted to the resulting data set with the fixed parameters G = 1 and P(pK) = 0.5. Theparameters k = 1.5 and pK = 3.6 were derived from the fit (RMSD(P) = 0.016). The pH-dependent rates of the fusion index FI and the population E∗ were compared by calculatingdPF/dPE∗ of the fitted curves, based on the first-order linear differential equation for thederivative of P,

dPdpH

= kP(G−P) , P 6= 0 (5.3)

The result was a bell-shaped curve with dPFI/dPE∗(7.8) = 6±3 at pH = 7.8, where the fusionindex curve begins, and a maximum dPFI/dPE∗(6.9) = 124± 18 at pH = 6.9. I.e. the ratiobetween cell fusion and the rate of activation of the E protein increased up to pH ≈ pK, thendropped again.

The membrane domains of E and M

In the polyprotein of the envelope glycoprotein the M protein is located N-terminal to theE protein. A cellular protease cleaves M off E, and the C-terminal fragment of M remainsanchored in the membrane [Li et al. 2008]. In the cryo-EM structure of pre-fusion E and M,1P58, His209 and His282 are located at the interface between the ectodomain and the mem-brane domains of M and E (Fig. 5.3). In the post-fusion structure of sE, 1OK8, the proximityof His209 to the C-terminus and to clefts in the outer surface of the sE trimer suggest thatthe linker to the C-terminal membrane domain may be close to His209 and interact with thehistidine.

The prM-E complex

Two X-ray crystal structures of the DEN2 prM-E heterodimer complex were available, crys-tallised at pH 7 and at pH 5.5 respectively, in which the attachment of prM to E was preservedthrough recombinant linking (PDB IDs: 3C6E and 3C5X [Li et al. 2008]). Furthermore, twocryo-EM structures of the immature DEN2 envelope assembly of prM-E were available, onein the trimeric state and one in the low-pH-dependent dimeric state at pH 6.0 (PDB-IDs:1TGE [Zhang et al. 2004] and 3C6R [Yu et al. 2008]). In the trimeric state the subunitsare arranged in a tripod formation. The dimeric state has the same subunit arrangement asthe mature dimer. All four of these structures contained the ectodomain sE of the envelope

67

5.3 Results

protein. In the crystal structures and in 3C6R, additionally the 81 N-terminal residues ofthe 166 amino acid sequence of prM were resolved. In the heterodimer, the prM fragmentis positioned on top and around one side of the distal end of domain II of E and interactswith the elongated region of domain II and the fusion peptide. One salt bridge attaches prto domain II, between His244 of domain II and Asp63 of pr (Fig. 5.6) [Li et al. 2008]. Inthe conserved domain superfamily pfam01570, Asp63 was conserved in all 10 representativeflaviviral sequences [Marchler-Bauer et al. 2009]. The crystal structure 3C5X was comparedwith the cryo-EM structure 3C6R in a superposition of domains I and III. In the superposi-tion a positional overlap was observed between the loop E0F0 of the cryo-EM structure andthe artificial N-terminal peptide linker of the recombinant protein in the crystal structure.

The pKa of the histidine residues in the crystal structures (PDB-IDs: 3C6E and 3C5X[Li et al. 2008]) were analysed to estimate the effect of the binding of pr on the proton-ation state of the histidines. The pKa calculated for the single hetero-dimeric units are listedin Table 5.4. Based on the predicted pKa, in the crystal structure of prM-E at neutral pHHis94, His144, His244, His282 and His346 were doubly-protonated and the other histidinessingly-protonated. In the crystal structure of prM-E at pH 5.5 all the histidines were doubly-protonated. The pKa of His244 was 8.2, mainly due to the salt bridge His244.Nπ-Asp63,with further contributions from interactions with Asp98 and Asp245 of E, and Asp40, Tyr51,Glu62, Asp65 and Tyr77 of pr. The predicted pKa of His27, His144, His158, His209 andHis282 varied by more than 0.5 pK-units between the two crystal structures. These structureswere obtained from crystals of monomeric E protein and therefore lack the environment ofthe native oligomeric state, i.e. of the trimeric form [prM-E]3 and of the dimeric form [prM-E]2 respectively [Yu et al. 2008; Zhang et al. 2004]. To determine the effect of interactionswithin the E dimer, the pKa of the histidine residues were calculated in a single subunit ofE taken from the crystal structure of the mature pre-fusion sE dimer 1OAN (sE2, ∆E) andcompared with the values for the dimer (Tab. 5.2). The difference in pKa, ∆pKa(∆E), islisted in Table 5.4. For some of the histidines this difference was large, e.g. one pK unit forHis27. The correlation coefficient between the difference in pKa and the distance to the othersubunit was 0.56.

The assemblies of four prM-E molecules in the crystallographic space groups in thecrystal structures of the prM-E hetero-dimer had no relation to the geometry of the nativedimer in the viral envelope. Therefore a model of the native assembly of two prM-E het-erodimers, [prM-E]2, was generated to gain insight into the interactions of the pr peptidewith the E dimer. The model was generated by superimposing the crystal structures of themature dimer and the prM-E complex onto the cryo-EM structure of the immature prM-Edimer at low pH (PDB-IDs: 1OAN [Modis et al. 2003], 3C5X [Li et al. 2008] and 3C6R [Yuet al. 2008]). The structure 1OAN was superimposed onto domains I and III of one subunit,A, and the structure 3C5X onto each domain II of the subunits of the dimer, A and B. The

68


Figure 5.6 Histidine interactions in the dimeric hybrid model of [prM-E]2. All the histidines andGlu60, Glu62 and Asp63 of pr are highlighted in stick representation. Proposed interactions withHis149 and His158 and the salt bridge His244-Asp63 are highlighted by green sticks. For a clearerview only the conserved loop 58�63 of pr is shown opaque, the rest of pr is transparent. LoopE0F0 (144�158) of E is highlighted by a thicker tube. Domain colours: red � domain dI, yellow �dII, blue � dIII, purple � pr peptide; dII' denotes domain II of the opposite subunit.

69

5.3 Results

Table 5.4 Predicted pKa values for the protonation of the histidines in the prM-E complex andvarious oligomeric states of the mature sE protein (sE and sE2). Numbers in bold indicate the pKa

that predict double protonation to be predominant at the respective pH. dpr is the distance betweenthe imidazole nitrogens and the side chains of charged residues of the pr peptide in the dimericmodel [prM-E]2. ∆pKa(∆E) is the di�erence in pKa between the dimer sE2 and the monomer sE.

3C6E 3C5X dpr 1OANpH 7 pH 5.5 / [nm] pH 8.5

prM-E [prM-E]2 sE2 sE ∆pKa(∆E)

H27 6.4 5.7 1.4 5.2 6.4 1.2H94 8.0 8.2 1.7 7.8 7.8 0.0H144 8.5 7.6 1.0 6.6 6.4 −0.2H149 5.8 5.6 0.6 6.6 6.8 0.2H158 5.5 6.6 0.4 5.6 5.8 0.2H209 6.8 7.5 2.7 6.4 6.4 0.0H244 8.2 8.2 0.3 5.1 5.7 0.6H261 6.6 6.7 1.3 5.5 6.0 0.5H282 7.4 6.7 1.9 6.5 6.7 0.2H317 6.5 6.7 0.8 6.2 6.5 0.3H346 7.9 7.5 3.4 7.0 7.1 0.1

RMSD of the Cα-atoms was 0.11 nm for 1OAN and 0.03 nm for 3C5X. From the superpo-sitions three structural fragments, fragment (I+III)A containing domains I and III of 1OAN,and fragments IIA and (II+pr)B containing domain II and pr of 3C5X, were combined in afull-residue model of the dimer [prM-E]2, shown in Figure 5.6. In this model, steric clasheswere observed between Trp101 and Phe108 of the fusion loop and Ala313 and Glu314 ofdomain III, and between the loop 58–62 of pr and Gly5, Met6 and the loop 148–154 ofdomain I. The imidazole nitrogens of His27, His144, His149, His158, His261 and His317were located less than 1.5 nm away from the side chains of charged residues of pr, thereforeinteractions with pr may contribute significantly to the pKa of these histidines (Tab. 5.4). Inparticular the proximity of His149 and His158 to Glu60 and Glu62 of the conserved polarloop 57–65 of pr suggests significant interactions at low pH. Like Asp63, Asp61 was fullyconserved in all ten sequences of the domain superfamily pfam01570. The structure of theC-terminal sequence of prM, residues 82–130, had yet to be resolved, therefore it remainsto be seen how it might interact with E, for instance with His149 and His158. The completesequence of prM contained six histidine residues, but only His2 and His11 were resolvedin the crystal structures. In the dimeric model the side chains of these two histidines werelocated 0.8 nm away from the E protein and therefore do not interact directly with E.

In the superposition of the crystal structure of the pre-fusion dimer, 1OAN, on the cryo-EM structure of the [prM-E]2 dimer, 3C6R, shown in Figure 5.7, it can be seen that after

70


Figure 5.7 Cryo-EM structure of the prM-E heterodimer at pH 6.0 (red, yellow and blue) super-imposed on the crystal structure of the mature dimer (cyan) (PDB-IDs: 3C6R [Yu et al. 2008] and1OAN [Modis et al. 2003]). The red arrow indicates the shift of the loop of His244 (thick tubesegments), framed in the image by a red box. The transparent blue line between His244 and theother subunit indicates a hydrogen bond with the backbone of His27. Green sticks indicate thesalt bridge His244-Asp63 and proposed interactions between Arg2, His149, His158 of E and Glu60and Glu62 of pr. Colour legend: red � domain dI, yellow � dII, blue � dIII, transparent violet tube� pr peptide; green spheres � histidine residues.

71

5.3 Results

the dissociation of pr (violet) the loop of His244 (yellow) moves towards the other subunit(cyan). Charged residues or histidines located close to pr in this model were Arg2, His149and His158. Some possible interactions between these residues and pr are highlighted inFigure 5.7 with green sticks. The SASA of the His244 side chain in the crystal structures atneutral and at low pH was 0.87 nm2 and 0.91 nm2 respectively, i.e. in both structures His244was solvent-exposed.

Conservation of the histidine residues and their micro-environments

The sE protein sequences from the four dengue viral serotypes showed 60–76 % sequenceidentity, with 8 of the 16 histidine positions fully conserved. The Conserved Domain Database(CDD) contains superfamilies of conserved domain models determined from multiple se-quence alignments of related proteins. These proteins may span a variety of organisms andmust contain at least one 3D structure [Marchler-Bauer et al. 2009; Marchler-Bauer & Bryant2004]. At the time this study was performed the two superfamilies of the flaviviral glyco-protein ectodomain, pfam00869 and pfam02832, contained 38 and 18 sequences respec-tively and included 3D structures from dengue virus serotypes 1–4, TBE virus, West Nilevirus, yellow fever virus, Langat virus, Omsk hemorrhagic fever virus, Japanese encephalitisvirus, Kunjin virus, St. Louis encephalitis virus, Murray Valley encephalitis virus and tick-borne Powassan virus. The domains conserved in these two superfamilies are the central anddimerisation domain (domains I and II) and the immunoglobulin-like domain (domain III),respectively. Table 5.5 lists the conservation rates of the histidine residues of the DEN2 sEprotein within the conserved superfamilies. The sequence alignments showed that His144,His244 and His282 were conserved in 36–37 of 38 sequences; only in one sequence, of theTamana bat virus, were all three of these histidines absent. His317 was conserved in all 18sequences of the superfamily pfam02832. Glu368, which forms a salt bridge with His317 inthe post-fusion crystal structure and interacted with His144 during the simulation HIS+, wasalso conserved in all 18 representative sequences.

The micro-environments of the unconserved histidine residues were compared in ho-mology models of the sequences of the two respective superfamilies, to determine whetheran unconserved histidine was replaced by a mutation to histidine in a similar location. For se-quences in which a histidine of DEN2 was not conserved, alternative histidines were soughtin the corresponding homology model within a radius of 0.57 nm of the side chain of theoriginal histidine. The homology structures were based entirely on the pre-fusion crystalstructure of the DEN2 sE protein, 1OAN, without further refinement.

In the following, the position numbers refer to the homologous positions in DEN2,

72


Table 5.5 Conservation of the histidine residues in the �aviviral glycoprotein domain superfamiliespfam00869 and pfam02832 of the ectodomain sE [Marchler-Bauer et al. 2009; Marchler-Bauer &Bryant 2004]. c indicates the number of sequences out of 38 or 18, respectively, that contained ahistidine at a position equivalent to the DEN2 virus. Numbers in italics indicate that compensatingmutations increase the conservation rate to more than 50 % of the sequences comprising thesuperfamily.

H27 H94 H144 H149 H158 H209 H244 H261 H282 H317 H346

c 13 2 37 5 8 26 36 17 37 18 2c/38 / [%] 34 5 97 13 21 68 97 45 97c/18 / [%] 100 11

i.e. the numbering of the DEN2 sequence is transferred to the aligned sequences. In 24 se-quences histidines at positions 28, 278 or 279 complemented the loss of His27. While His94was barely conserved, in 27 sequences residues 93 and 94 were both either lysine or arginine,and in 8 other sequences one of these residues was either histidine or glutamine. In addition,a di-glutamate Glu84-85 was conserved in 23 sequences, possibly conserving the interactionwith His94 observed in the pre-fusion crystal structures. A cluster of histidine mutations wasfound in the poorly conserved loop E0F0 in positions 148, 149, 151, 153, 156 and 158. In19 sequences this loop contained one histidine, and in 5 other sequences two histidines. In35 sequences the absence of a histidine in position 209 was complemented by a histidine inposition 214 in a similar location in the homology structure. Histidines in positions 204, 205,261, 262 and 276–279 were complementary to each other in 28 sequences and were all lo-cated close to the hinge region between domains I and II. His346 was poorly conserved, butin 6 other sequences a histidine was found either in position 342 or 347, which are in similarlocations on the loop ED of domain III. Among the residues that interacted with histidinesin the MD simulations of the E protein (see section 5.3 on page 73), a negatively chargedresidue was conserved at position 79, and a charged or polar residue at position 195.

MD simulation of the sE protein

Electrostatic interactions

MD simulations of the dengue viral sE protein dimer were performed to determine the ef-fect of double-protonation of the histidine residues on the structure of the protein. Thesimulations were analysed for salt bridges and hydrogen bonds, to determine whether thetwo protonation conditions “neutral” and “slightly acidic” led to characteristic electrostaticinteraction profiles. The interactions were compared with the specific interactions in the

73

5.3 Results

crystal structures, that characterise the pH-dependent conformations (Tab. 5.1), to determinewhether the simulations reproduced any pre-fusion- or post-fusion-specific interactions. Theinteraction propensity in a simulation was determined as the percentage of configurationssampled in which the interaction occurred. The propensities of selected salt bridge and hy-drogen bonding interactions in the simulations are listed in Table 5.6.

In the simulation HIS0 only three hydrogen bonds of the seven interactions at the sub-unit interface of the pre-fusion crystal structures were preserved, the forked Ser7-Asp98-Ser7.N-2 and Ser255.N-2-Glu257.Oε (Tab. 5.1 and Tab. 5.6). The other four interaction pairsat the pre-fusion subunit interface were largely lost, including the two pairs of salt bridges,Lys241-Glu269 and Lys246-Glu44. Because in HIS0 the histidines were unprotonated at theNπ-atom the hydrogen bond His261.Nπ-Leu253.O-1 could not persist. At the domain I–IIIinterface only three of the eight pre-fusion-specific interaction pairs were preserved in HIS0,the salt bridge Arg9-Glu368 and the hydrogen bonds Thr40-Ile352.O-1 and Asn366.N-2-Gly146.O-1. An interaction was considered significantly protonation-dependent if the dif-ference in propensity between the simulations HIS0 and HIS+ was larger than 50 %. Furtherinteractions with significantly higher propensities in HIS0 than in HIS+ were Val15.N-2-Trp20.O-1, His27.Nτ-Ile46.O-1, Lys123-Glu257, Met196.N-2-Asp192.O-1 and Thr319-Glu368 (Tab. 5.6). The singly-protonated histidines in HIS0 not only acted as donors butalso as acceptors, enabling hydrogen bonds such as Gln211.N-2-His209.Nπ. The interac-tion His144.Nτ-Asp42 was found in HIS0 and in HIS+, with a slightly lower propensity inHIS0. This interaction was analysed as an example of a stable hydrogen bond, the interactiondistance is plotted in Figure 5.8.

Two pre-fusion-specific interactions of Glu368 were disrupted in HIS+, the salt bridgeArg9-Glu368 and the hydrogen bond Thr319-Glu368 (Tab. 5.6). These interactions arebroken during the conformational change, thus the simulation HIS0 reproduced these in-teractions and HIS+ reproduced their disruption after protonation. In HIS+ they were re-placed by the interactions His144.Nπ-Glu368 and Thr353-Glu368. Interactions that had highpropensities in the simulation HIS+ only were the histidine interactions His27.Nτ-Glu44,the double salt bridges Glu79-His94.Nτ-Glu85, Glu192-Nτ.His209.Nπ-Glu195 and Asp10-Nτ.His282.Nπ-Glu26, His144.Nπ-Glu368, His149.Nπ/τ-Glu148, His158.Nτ-Glu147,His158.Nτ-Asp154, His261.Nπ-Glu269 and His317.Nπ-Glu314, the inter-domain, backbonehydrogen bonds Thr48.N-2-Gly275.O-1, Leu191.N-2-Thr280.O-1, Leu277.N-2-Ile46.O-1,Gln316.N-2-Gly5.O-1 and Ile357.N-2-Ser145.O-1, and Thr353-Glu368 (Tab. 5.6). Almostall of these interactions formed during the simulation HIS+, with the exception of His94-Glu85, His158-Glu147 and His282-Glu26, which were already present in the pre-fusioncrystal structure. His209-Glu195 was the only post-fusion-specific salt bridge that formedin the simulations, with a higher propensity in the simulation HIS+ than in HIS0. The saltbridge His144.Nπ-Glu368 at the domain I–III interface occurred only in HIS+, when the

74


histidines were doubly-protonated and both imidazole nitrogen atoms could act as donors.The only other specific inter-domain interaction that occurred in the simulation HIS+ butnot in HIS0 was the backbone hydrogen bond Leu277.N-2-Ile46.O-1, joining the β-barrel ofdomain I to the hinge region between domains I and II.

Among the electrostatic interactions present in the pre-fusion crystal structures thatwere disrupted in both simulations were the interactions Arg2-Glu44 (data not shown),Lys246-Glu44 and His261.Nπ-Leu253.O-1 at the subunit interface, and Thr353-His144.Nπ

at the domain III interface (Tab. 5.1). In the simulation HIS+ some of these interactionswere replaced by salt bridges with histidine residues, His27-Glu44, His144-Glu368, His244-Glu44 and His261-Glu269. In addition interactions formed in the simulations that wereabsent in the crystal structures. Specific interactions that occurred in both subunits in thesimulation HIS0 but not in the pre-fusion crystal structures were His144.N-2-Thr40.O-1,Arg323-Glu148.O-1 at the domain III–I interface and Gln211.N-2-His209.Nπ (Tab. 5.6 and

Figure 5.8 A stable hydrogen bond, His144τ-Asp42.Oδ during the MD simulation HIS0. A) Min-imum distance between the imidazole nitrogen atoms of His144 and the carboxyl oxygen atomsof Asp42, in the two subunits (black and grey, respectively). B) Existence plots of the hydrogenbond in the two subunits.

75

5.3 ResultsTa

ble5.6Sa

ltbridges(SB)

andhydrog

enbo

nds(H

)in

thecrystalstru

ctures

1OKE

,1OA

Nand1O

K8[M

odiset

al.20

03,

2004

],andin

theM

Dsim

ulation

ofthed

engu

evira

lsEprotein

.The

occurre

nceo

faninteraction

isindicatedas

thep

ercentage

ofcon�

guratio

nssampled

from

asim

ulation

,inwh

ichtheinteraction

wasfoun

d,andwe

reanaly

sedfore

achsubu

nit,

Aand

B,in

twoseparate

simulation

intervals

,i)0

�30ns,a

ndii)

30�6

0ns.Prop

ensities

ofat

least

75%

areprintedin

bold.�n.r.�

indicatest

hatinthecrystalstru

cturethedo

noro

racceptorw

asno

tresolv

ed;�n.a.�no

tapp

licable.

aAr

abicnu

mbers

indicate

interaction

swith

inthedo

main

ofthat

number,roman

numbers

indicate

interaction

sbetwe

entherespectiv

edo

main

s.

dom

ain

X-r

aydi

ffr.

MD

sim

ulat

ion

orpr

e-po

st-

HIS

0H

IS+

subu

nit

fusi

onfu

sion

subu

nitA

subu

nitB

subu

nitA

subu

nitB

dono

rac

cept

orin

terf

acea

1OA

N1O

KE

1OK

8i

iii

iii

iii

ii

Ser7

.Oγ

Asp

98.O

δA

-BH

8782

7152

6977

7847

Ser7

.N-2

Asp

98.O

δA

-BH

H66

7287

7856

7577

35A

rg9.

Nω

Glu

368.

Oε

I–II

ISB

SB98

9910

010

022

230

4V

al15

.N-2

Trp2

0.O

-11

9593

9191

20

00

His

27.N

τG

lu44

.Oε

10

00

01

685

99H

is27

.Nτ

Ile4

6.O

-11

6565

4842

00

00

Thr

40.O

γIl

e352

.O-1

I–II

IH

H98

9399

9963

8933

0T

hr48

.N-2

Gly

275.

O-1

I-II

00

00

071

8576

His

94.N

π/τ

Glu

79.O

ε2

43

38

9687

100

99H

is94

.Nτ

Glu

85.O

ε2

SBSB

SB1

00

142

099

96Ly

s123

.Nζ

Glu

257.

Oε

2SB

SB9

4260

8312

23

2H

is14

4.N

τA

sp42

.Oδ

1H

HH

9175

9276

100

100

100

100

His

144.

Nπ

Glu

368.

Oε

I–II

In.

a.n.

a.n.

a.n.

a.10

010

010

010

0H

is14

9.N

π/τ

Glu

148.

Oε

1n.

r.1

40

067

7080

63H

is14

9.N

π/τ

Asp

154.

Oε

1n.

r.0

20

00

02

73

cont

inue

don

next

page

76


Tabl

e5.

6co

ntin

ued

X-r

aydi

ffr.

MD

sim

ulat

ion

dom

ain

pre-

post

-H

IS0

HIS

+

orfu

sion

fusi

onsu

buni

tAsu

buni

tBsu

buni

tAsu

buni

tB

dono

rac

cept

orin

terf

ace

a1O

AN

1OK

E1O

K8

iii

iii

iii

iii

His

158.

Nτ

Glu

147.

Oε

1H

Hn.

r.12

60

081

9793

70H

is15

8.N

τA

sp15

4.O

δ1

n.r.

00

00

00

4710

0L

eu19

1.N

-2T

hr28

0.O

-1II

-I0

017

160

047

84M

et19

6.N

-2A

sp19

2.O

-12

H61

7920

8457

117

2H

is20

9.N

τA

sp19

2.O

δ2

00

60

610

00

0H

is20

9.N

π/τ

Glu

195.

Oε

2SB

2930

1731

9810

098

99G

ln21

1.N

-2H

is20

9.N

π2

6781

6678

n.a.

n.a.

n.a.

n.a.

His

244.

Nπ

Glu

44.O

εA

-Bn.

a.n.

a.n.

a.n.

a.0

032

65H

is26

1.N

τG

lu25

7.O

ε2

10

20

00

2465

His

261.

Nπ

Glu

269.

Oε

2n.

a.n.

a.n.

a.n.

a.52

100

20

Leu

277.

N-2

Ile4

6.O

-1II

-I0

10

029

5291

94H

is28

2.N

τA

sp10

.Oδ

10

00

00

075

99H

is28

2.N

πG

lu26

.Oε

1SB

SBn.

a.n.

a.n.

a.n.

a.87

100

4458

His

282.

Nτ

Glu

26.O

ε1

SBSB

6452

5959

120

70

Gln

316.

N-2

Gly

5.O

-1II

I-I

00

10

70

4276

His

317.

Nπ

Glu

314.

Oε

3n.

a.n.

a.n.

a.n.

a.51

920

0T

hr31

9.O

γG

lu36

8.O

ε3

HH

H10

010

010

010

00

07

43T

hr35

3.O

γG

lu36

8.O

ε0

00

198

9081

18Il

e357

.N-2

Ser1

45.O

-1II

I-I

n.r.

127

00

7981

912

77

5.3 Results

data not shown).

Number of contacts at the subunit and domain interfaces

The number of contacts between the two subunits and between all the domains of the dimerin the simulations were monitored to determine whether the subunits of the dimer dissociatedand whether the extents of the domain interfaces changed. In the simulation HIS0 the numberof contacts between the monomers of the dimer increased from 2 500 to 2 730, i.e. by 9 %within 60 ns, whereas in the simulation HIS+ the number of monomer-monomer contactsremained largely unchanged, with a minor decrease by 2 % from 2 310 to 2 260.

Table 5.7 Number of contacts and number of contact atoms at the subunit interface of the sEprotein dimer, in the initial con�guration, the crystal structure 1OKE, after solvent and side chainequilibration of the systems (t = 0) and after 60�70 ns of simulation.

domain, t = 0 60–70 ns

subunit 1OKE HIS0 HIS+ HIS0 HIS+

contacts

A/B 2305 2500 2310 2730 2260

contact atoms

IA 74 85 95 82 40B 74 92 84 70 27

IIA 202 221 211 198 206B 207 228 203 193 187

IIIA 67 79 69 97 81B 65 74 80 76 70

A 347 385 375 377 327B 350 394 367 339 284Σ 697 779 742 716 611

Compared to the initial number of contacts between the subunits in the crystal structure1OKE, during the preparatory equilibration (t ≤ 0) and the production simulation taken to-gether these contacts increased by 18 % in HIS0, and decreased by 2 % in HIS+. The numberof contact atoms increased in HIS0 by 3 % and in HIS+ it decreased by 12 %. In HIS+ thiswas mainly due to a decrease in the number of contact atoms of domains I by 46 % and 64 %respectively. The change in the number of contacts during the simulations with respect to theinitial value in the crystal structure was tested for outliers using Dixon’s Q test, to determinewhether extreme values might indicate significant changes. However, the extreme valuescan only be rejected from the corresponding sets with less than 90 % confidence and do notqualify as outliers, i.e. as significantly extreme values.

78


Figure 5.9 Histidine protonation-dependent conformational change of the sE dimer after 70 ns ofMD simulation. Cross-eyed stereo images of A) the secondary structure, and B) the surface of sE.The subunits are highlighted in green and pink respectively.

Structural changes during the MD simulations

To determine the structural changes of the dengue viral sE protein dimer during the sim-ulations HIS0 and HIS+, configurations of the protein after 60–70 ns of simulation weresuperimposed on the initial configuration of the simulations, the pre-fusion crystal structure1OKE. Local structural changes were expected in both simulations due to the strong inter-actions that can result from the addition or subtraction of a whole charge when changingthe protonation state of a residue. In both simulations the protein remained dimeric and thebasic structural elements such as the domain organisation and the secondary structure weremaintained well. This is illustrated in Figure 5.9, which shows the configuration of HIS+after 70 ns of simulation.

The positional root mean square deviation (RMSD) with respect to the pre-fusion andthe post-fusion crystal structures as a function of time was calculated for configurationsfrom the simulations to evaluate the structural changes with respect to the two conforma-tions. Figure 5.10 shows a plot of the positional RMSD of the backbone atoms of the dimerwith respect to the initial structure during the simulations HIS0 and HIS+. From this plot itcan be seen that the deviation from the initial structure was much larger in HIS+, in whichthe RMSD reached values up to 1 nm, than in the case of HIS0, in which the maximumRMSD was ≈ 0.7 nm. During the preparatory equilibration of HIS+, in which position re-straints were applied to the backbone atoms of the protein, the RMSD of the backbone atomsalready changed by 0.16 nm. For a protein backbone the RMSD values in both simulations

79

5.3 Results

Figure 5.10 Positional root mean square deviation (RMSD) of the sE protein dimer during theMD simulations HIS0 (black) and HIS+ (grey), of all backbone atoms (solid lines), and of thebackbone atoms of domains I and II only (dotted lines). Data were sampled every 2 ps, and thecurves were smoothed by a 1 ns-running average.

are considered large and may indicate unfolding. However, the fact that the secondary struc-ture remained largely intact in both simulations suggests that the large RMSD values reflectinter-domain motions as opposed to unfolding (Fig. 5.9). In simulations of large, extendedmolecules such as the E protein, small variations in the relative positions of the domains canlead to large differences in the relative positions of groups located in distal or peripheral partsof the structure. To estimate the impact of the distal parts on the average structural deviationof the dimer, the RMSD was calculated for the backbone atoms of the central domains I andII only, i.e. excluding the two distal domains III. The difference between the RMSD exclud-ing (Fig. 5.10 dotted lines) and including (solid lines) domains III reflects the contributionfrom the domains III to the overall RMSD of the dimer. Towards the end of both simulationsthe overall RMSD of the dimer was about 0.1–0.2 nm larger than the RMSD excluding do-mains III. The shift of the center of mass of the backbone atoms of domain III relative to thebackbone atoms of domains I and II after 60 ns of simulation was calculated to determine ifdomain III relocated. In the simulation HIS0 this shift was 0.4 nm and 0.8 nm for the twosubunits respectively, and in HIS+ 0.6 nm and 0.4 nm respectively.

The RMSD of backbone atoms was determined for the individual domains to examine

80


the structural change within the domains. The domain definitions were based on the pre-fusion structure of the E protein, as specified in the Entrez Molecular Modeling Database(MMDB), where domains are identified by compactness [Chen et al. 2003]. According tothe MMDB the 3D domains of the pre-fusion structure of the E protein ectodomain containedthe following segments of the protein sequence: domain I – residues 1–52, 133–190, 278–296; domain II – residues 53–132, 191–277; domain III – residues 297–394. The MMDBalso identified a fourth domain in the post-fusion crystal structure, containing residues 68–120 in the elongated part of domain II (coloured green in Fig. 3.2). Figures 3.2 and 3.3 showthe sE protein with the domains highlighted by different colours.

Table 5.8 Positional RMSD of the backbone atoms of the domains of the sE protein in the MDsimulations, with respect to the pre-fusion and the post-fusion crystal structures, 1OKE and 1OK8.The values were averaged over data sampled every 2 ps between 30 and 60 ns. The numbers inbrackets indicate the standard deviation.

HIS0 HIS+

domain(s) subunit A subunit B subunit A subunit B

pre-fusion, 1OKE

I ∪ II 0.30 (0.032) 0.43 (0.057) 0.45 (0.053) 0.43 (0.049)I 0.21 (0.013) 0.27 (0.017) 0.25 (0.008) 0.30 (0.014)II 0.25 (0.015) 0.28 (0.019) 0.33 (0.025) 0.29 (0.027)III 0.19 (0.011) 0.16 (0.015) 0.19 (0.009) 0.18 (0.015)

post-fusion, 1OK8

I ∪ II 0.55 (0.024) 0.64 (0.043) 0.67 (0.053) 0.73 (0.034)I 0.54 (0.009) 0.54 (0.008) 0.60 (0.009) 0.65 (0.017)II 0.28 (0.011) 0.31 (0.020) 0.32 (0.023) 0.32 (0.029)III 0.20 (0.010) 0.18 (0.012) 0.19 (0.010) 0.17 (0.012)

Table 5.8 lists the RMSD of the backbone of each domain of the sE protein dimer withrespect to the pre-fusion and the post-fusion crystal structures, averaged over configurationssampled every 2 ps from the second half of a respective simulation, t ≥ 30 ns. The RMSDvalues of the individual domains were smaller than the RMSD of the dimer (Tab. 5.8, Fig.5.10). In both simulations domain III was the domain most similar to the pre-fusion and thepost-fusion structure. In both simulations domain I remained more similar to the pre-fusionstructure than the post-fusion structure. This was also true for the structure of domains I andII combined (I ∪ II). In the simulation HIS+ domain I deviated further from the post-fusionstructure than in HIS0. Domain II and domain III both deviated similarly from the pre-fusionand the post-fusion conformations and also similarly in both simulations.

81

5.3 Results

Figure 5.11 Positional root mean square deviation (RMSDx) of the Cα-atoms after 60�70 nsof simulation of the sE dimer, in the simulation HIS0 (top; grey plot in the bottom panel) andHIS+ (bottom, black). The area between the RMSDx plots of the two subunits was �lled in black.Negative residue numbers refer to domain III of the respective other subunit. The patterned barbetween the graphs indicates the extent of the domains along the sequence.

The positional RMSDx of the Cα-atoms was calculated to identify regions with variousmobilities, that determine the structural motions during the simulations. The RMSDx is plot-ted in Figure 5.11; the negative residue numbers −97–0 refer to domain III of the respectiveother subunit. Peak regions indicate regions of high mobility such as domain hinges. In bothsimulations there were regions with various mobilities. In HIS+ (bottom panel) the devia-tions were on average larger than in HIS0 (top). The covariance between the RMSDx of thetwo subunits was 1 % in HIS0, and 10 % in HIS+, meaning that in HIS+ the mobilities ofspecific residues became more similar in the two subunits. Some of the peaks coincide withthe domain boundaries specified in the Molecular Modeling Database (MMDB).4 An addi-

4 See page 81 for the domain boundaries.

82


tional hinge region was identified at residues 83 and 229, between the base part and the distalpart of domain II. Some of the peaks are assigned to residues located on opposite sides of adomain structure, indicating rotations of these domains, e.g. in HIS0 residues 17 and 155 ofdomain I (Fig. 5.11 top). Similarly the data from the simulation HIS+ indicate rotations ofall three domains (bottom panel). Some of the peaks expected at the proposed hinge regionsonly appeared in the simulation HIS+, in particular at residues 52, 133, 191 and 276 of thehinge region between domains I and II.

pKaaa and solvent accessibility of the histidine residues

The pKa of the histidine residues were calculated for configurations from the simulations,to determine whether the structural changes observed led to changes in the local pKa; theresults are listed in Table 5.2 on page 61. In both simulations the pKa shifted away fromthe values observed in the pre-fusion crystal structures. For some of the histidines, shifts inpKa were already observed during the preparatory equilibration of the side chains, in whichposition restraints were applied to the backbone atoms (column “HIS0 t ≤ 0” in Table 5.2).

Three configurations from the simulation HIS+, sampled at random after 60–70 ns,provided a set of structures of sE for determining the effect of the protonation of the histidineresidues on the protein. In these configurations the pKa values of all the histidine residueswere higher than in the simulation HIS0, and on average also higher than in the crystalstructures of the sE dimer and trimer (Tab. 5.2 and data not shown). This was expected tosome extent as a consequence of the protonation of the histidines, as the local environmentwill adapt to stabilise the positive charge. However, major changes by several pK-units tovalues pKa ' 8 for some of the histidines indicate that the structural changes increase thelikelihood of double protonation by several orders of magnitude and that a more basic pHthan before is required in order for these histidines to become deprotonated again. In otherwords, once these histidines were exposed to lower pH, structural changes stabilized thedoubly-protonated state and made the protonation of these histidines irreversible. In threeconfigurations from the simulation HIS0, sampled at random after 50–60 ns, the pKa of thehistidine residues were on average below 7.

The initial low pKa of His144 (Tab. 5.2) was mainly due to high desolvation and theinteraction with Arg9; further contributions were interactions with His149 and His317. Afterprotonation of the histidines these interactions were lost, and in particular the interaction withGlu368 contributed instead to the large increase in the pKa of His144 in HIS+. After 70 nsof simulation in HIS+, the pKa of His144 had increased by 2.3 and 3.1 pK units in the twosubunits respectively. In HIS+, the pKa of His244, His261 and His282 also displayed largeshifts to higher values, mainly due to interactions with Asp154 and Asp249, with Glu257and Glu269, and with Asp10 and Glu26, respectively.

83

5.3 Results

During the simulations the SASA of all the histidine residues except His209 and His261remained similar to the SASA in the pre-fusion crystal structures (Tab. 5.3, p. 64). In bothsimulations the SASA of His209 was 40 % smaller than in the pre-fusion crystal structures.The SASA of His261 remained similar in the simulation HIS0 but increased significantlyfrom 0.3 nm2 to 0.8 nm2 in HIS+.

Restructuring of the E dimer

The configurations of the dimer in the simulations HIS0 and HIS+ after 60–70 ns (Fig.5.12 red, yellow, green, dark blue) were superimposed upon all the backbone atoms of theinitial configuration (cyan) to visualise the structural changes. In the superpositions shownin Figure 5.12, the emergence of one molecular surface over the other reflects the structuraldifferences between the configurations. The grey plane in the figure indicates where the viralmembrane would be located in situ.

In both simulations the changes in the subunits seem symmetrical. The superpositionin Figure 5.12A shows that during the simulation HIS0, domains I and III (red and bluerespectively) bent towards the side which would face the viral membrane in situ (grey plane),whereas domains II (yellow/green) curved slightly away from that side. Compared to thesuperposition of the configuration from the simulation HIS+, shown in panel B, the visibilityof the surfaces of the configurations from HIS0 and the initial structure in panel A appearsmore homogeneous, indicating that in HIS0 the structure remained more similar to the initialstructure. In contrast the superposition of the configuration from the simulation HIS+ (panelB) illustrates a global structural change of the dimer on the domain level. From panel B itcan be seen that domain I (red) and the base region of domain II (yellow) buckled towardsthe viral membrane plane (grey plane), while the extended region, i.e. the distal part, ofdomain II (green) and domain III (dark blue) rose in the opposite direction, away from themembrane plane. Taken together, these motions resulted in shearing motions between thesubunits. Figure 5.12C shows the superposition of the configurations from the simulationsHIS0 and HIS+ (dark grey and other colours, respectively). This superposition compares thestructural changes in the two simulations and suggests that the domain motions in the twosimulations are converse. While in both simulations the fusion peptide (orange spheres) rosevisibly (A, B, respectively), the effect was more pronounced in HIS+ (C) and correlates withthe motion of domain III away from the viral membrane plane.

Each subunit in the configuration from the simulation HIS+ after 70 ns was superim-posed on domain I of the initial structure (cyan) to determine the motions of the domains rel-ative to each other; the superpositions are shown in Figure 5.13. In one of the subunits (panelA) a large hinging motion was observed between domains I (red) and II (yellow, green). Inthe other subunit (panel B) a different and considerably smaller motion around this hinge

84


Figu

re5.12

Con�

guratio

nsof

theEprotein

after6

0ns

ofMD

simulation

inthecase

ofHIS0

(AandC)

,and

after7

0ns

ofsim

ulation

inthecase

ofHIS+

(BandC)

.A�B

)The

con�

guratio

nswe

re�ttedon

tothebackbo

neof

thepre-fusio

ncrystalstru

cture1O

AN(cyan;

cf.F

ig.3.3,

p.34

).C)

Thecon�

guratio

nfro

msim

ulation

HIS+

�ttedon

tothecon�

guratio

nfro

msim

ulation

HIS0

(darkgrey).

Top:

view

oftheou

ters

urface,m

iddle:

side

view,

botto

m:v

iewof

thesurfa

cethat

wouldface

thevir

almem

branein

situ.

Thegrey

planeindicatesw

here

thevir

almem

branewo

uldbe

locatedin

situ.

Domain

colou

rschem

e:red�do

main

dI,y

ellow

/green

�dII,blue

�dIII;

thefusio

npeptideiscolou

redorange.

85

5.3 Results

Figure 5.13 Conformational change of thesubunits of the sE dimer in the simulationHIS+ after 70 ns. The con�guration of eachof the two subunits (panel A and B, re-spectively) was �tted onto the backbone ofdomain I in the initial con�guration 1OKE(cyan) and rotated into the same view. C)Perspective view of the same �t as in B butshowing the whole dimer. The domains inthe con�guration after 70 ns are coloured:red � domain dI, yellow/green � dII, blue �dIII.

86


Figure 5.14 Contact regions between the subunits in the sE dimer. A) Subunit of the crystalstructure 1OAN, B) subunit of the crystal structure 1OK8, and C) the two subunits (left and rightrespectively) in the con�guration from the simulation HIS+ after 70 ns, rotated into the sameorientation as in A and B. In A and C bottom row, atoms located within 0.6 nm of the oppositesubunit are highlighted by atom type: white � hydrophobic, green � polar, red � acidic, blue �basic. In B and C top row, atoms are coloured according to the structure in which they are foundwithin 0.6 nm of the opposite subunit: blue/violet � 1OAN, red/violet � con�guration from HIS+.For a clearer view in panel B, domain III is shown in ribbon representation (green).

region resulted in domain II to bend in the opposite direction, following the “downward”motion of domains I and III (red and blue) of the former subunit (panels A, C). From thestructure of the dimer it can be seen that the subunits rotated towards each other, approxi-mately on two planes folding along the length of the dimer towards the surface that wouldface the viral membrane in situ. In this rotation the α-helices A of the two subunits foldedtowards each other and aligned. The RMSD with respect to the post-fusion structure wascalculated to estimate whether the hinge motions between domain I and the other two do-mains approached the post-fusion conformation. First the trajectory of the backbone atomsof one subunit was fitted to the post-fusion structure, then a translational fit to domain I ofthe post-fusion structure was performed on the fitted trajectory and the RMSD calculatedfor each domain. This RMSD between the pre-fusion and the post-fusion crystal structurewas 1.0 nm for domain II and 2.2 nm for domain III. In the simulation HIS+ the RMSD re-mained approximately the same for both subunits, 1.0–1.3 nm for domain II and 2.2–2.3 nm

87

5.3 Results

for domain III, indicating that the domain arrangement did not converge to the post-fusionstructure.

The contacts between the subunits of the dimer before and after the conformationalchange were compared to determine changes in the interface and the arrangement of the sub-units. Figure 5.14 highlights the positions of the subunit contacts in the pre-fusion structure(A), and the positions of the same atoms in the post-fusion structure (B) and in the configura-tion from the simulation HIS+ (C top row, blue). In the post-fusion structure (B) the positionsof these atoms (blue) signify a rotation of domain II relative to domain I. In the simulationHIS+ a similar effect is observed in both subunits (Fig. 5.14C top row), where some of thepre-fusion-specific contacts were broken (blue) and new contacts formed (red); some of theinitial contacts were maintained (violet). The distribution of broken and new contacts indi-cates that the subunits, in particular domains II, rotated outwards and away from the subunitinterface. Whereas the contacts in the pre-fusion structure were distributed along the lengthof the subunit (A, C top blue), in the simulation the contacts condensed into three patches (Ctop violet and red): in the indentation between domains I and III, in the dimerisation region,and in the region of the fusion peptide. In the bottom row of panel C it can be seen thatalmost all the new contacts were hydrophobic (white) or polar (green).

Figure 5.15 Histidine residues (sticks) at the protein surface of the sE dimer during the simulationHIS+. Domain colour scheme: red � domain dI, yellow/orange � dII, blue � dIII.

As a result of the twisting motion of the dimer and the separation of the subunits be-tween domains I and II, the quarternary structure of the dimer adopted the topology of atwisted figure 8 (Fig. 5.9B), in which the contact regions (Fig. 5.14) form the waist and closethe loops of the figure 8. In one half of the dimer His27, His209, His244, His261 and His282

88


lined the opening between the subunits that formed after the protonation of the histidines(Fig. 5.15). In the other half of the dimer, the interface separated less, and only His27, His94and His244 were exposed at the surface around the opening.

5.4 Discussion

The fact that in the crystal structure of the post-fusion trimer there were about three times asmany salt bridges and hydrogen bonds connecting two subunits than in the crystal structure ofthe pre-fusion dimer, may explain the high stability of the trimer and the irreversibility of thetrimer formation. The inter-subunit salt bridges in the trimer had donors other than histidineresidues, which are stable over a wider pH range than salt bridges involving histidines. Itis important to note that for the post-fusion crystal structure 1OK8, trimeric sE protein wascrystallised at pH 7–8 and the X-ray diffraction data was collected at pH 7 [Modis et al.2004]. Despite the irreversible nature of the formation of the trimer [Modis et al. 2004], theuse of neutral or slightly basic pH instead of acidic pH may have led to changes in localinteractions, in particular in salt bridges involving histidines, which form only under acidicconditions. For example His261 and Glu269 (Tab. 5.6), or His282 and Glu368 did not formsalt bridges in the post-fusion structure, which may be experimental artefacts resulting fromthe neutral pH.

From the comparison between the protein configurations of the two simulations it isevident that the protonation of the histidine residues led to significant structural changes(Fig. 5.12). Changes were observed in the overall structural arrangement of the domains(Figs. 5.12, 5.13, 5.15) and in specific salt bridge and hydrogen bonding interactions (Tab.5.6). However, major molecular motions such as the relocation of domain III take place on amicro- to millisecond timescale [Mayor et al. 2000], therefore the protein structure was notexpected to converge to the post-fusion conformation in the simulations presented here. Asexpected the protonation of the histidine residues led to alterations in local interaction net-works (Tab. 5.6). His27, His144, His244 and His261 were involved in different salt bridgesand hydrogen bonds depending on the protonation state of the histidines, and the interactionsare therefore considered to be pH-dependent.

The simulations reproduced several salt bridges and hydrogen bonds present in the crys-tal structures, most notably the pre-fusion-specific Arg9-Glu368 in HIS0, and His144.Nτ -Asp42 (Tabs. 5.1, 5.6). Furthermore they confirmed the hinge regions between the domains(Fig. 5.11), predicted previously from domain compactness [Chen et al. 2003] and from mu-tations that compromise mechanical functions of the envelope protein [Modis et al. 2004].A further hinge was identified between the base part of domain II and the distal part, which

89

5.4 Discussion

contains the fusion loop (Fig. 5.11). Several dimer-specific interactions from the pre-fusioncrystal structures were lost in the simulations, for instance the inter-molecular salt bridgeLys246-Glu44 (Tab. 5.1). The fact that in both subunits in the control simulation HIS0 theinteraction Arg2-Glu44 replaced this salt bridge may suggest that it was an artefact of thenon-native conditions in the crystallographic experiment.

The pre-fusion subunit interface

The disruption of interactions between the subunits by the doubly-protonated His27, His244and His261 in HIS+ (Tab. 5.6) suggests that the protonation of these histidines destabilisesthe dimer. Whereas the salt bridge Lys246-Glu44 and hydrogen bond Ser255-Glu257 brokein both simulations, the formation of alternative salt bridges by His27, His244 and His261was protonation-dependent and occurred only in HIS+. For instance the salt bridge His27-Glu44 formed in one of the subunits and displaced the inter-subunit salt bridge Lys246-Glu44. Although some of these new interactions were observed only sporadically duringthe short simulation, or in only one of the subunits, they demonstrated nonetheless howcompetitive interactions with doubly-protonated histidine might facilitate dimer disassembly.

One hypothesis of how protonation could trigger the disassembly of the dimer wasby repulsion between positively charged groups across the subunit interface, e.g. Lys122,Lys123, Lys241 and His261. However, this is unlikely, as the interaction with a positivecharge would lower the pKa and lead to deprotonation of the histidine. In that case theprotonation of the histidine would require a preliminary structural change and dissociationof the interface. Rather, the competitive attraction of negative charges may lead to the releaseand displacement of a competing interaction partner, as was observed with His27 and His261,and thereby trigger initial structural changes in the conformational change.

His261

The simulation HIS+ predicted the formation of the salt bridge His261-Glu269 after proto-nation. This interaction was not present in any of the crystal structures (Tab. 5.6). The inter-subunit interactions His261-Leu253 and Lys241-Glu269 of the pre-fusion structure weredisrupted in both simulations, but the alternative interaction His261-Glu269 seems to beprotonation-dependent, as it formed only in HIS+. The location of this salt bridge closeto the hinge region between domains I and II suggests that it may play a role in the hingemotion. None of the crystal structures indicated the orientation of the α-helix B (residues256–264) that was observed in HIS+. However, “significant rearrangements” around the klhairpin (residues 270–279), which is also located close to the hinge region, were describedas part of the conformational change [Modis et al. 2004]. The kl hairpin is pulled apart in

90


the conformational change, and a number of mutations in the kl hairpin were reported thataffect the pH threshold. Therefore the kl hairpin was proposed as a target region for inter-fering with the conformational change to inhibit flavivirus entry Modis et al. [2003, 2004].The subunit in which the salt bridge His261-Glu269 formed also showed more pronouncedhinging between domains I and II than the other subunit (Fig. 5.13A). This correlation maysuggest that the salt bridge promotes the hinging between domains I and II. Furthermore, thesalt bridge might stabilise strand k of the hairpin against the α-helix B and thereby promotethe pulling apart of the kl hairpin. Although His261 and Glu269 did not form a salt bridgein the post-fusion crystal structure, the side chain positions and the secondary structure sug-gest that the formation of this salt bridge is possible. At the pH of the post-fusion crystalstructure, pH 7–8, His261 may have been singly-protonated and not able to form a stable saltbridge.

Number of subunit contacts

The increase in the number of contact atoms during the solvent and side chain equilibrationof HIS0 was distributed over all the domains (Tab. 5.7). This and the increase in the numberof contacts indicate a general growth of the interface between the subunits as a result of theequilibration. In contrast, during the equilibration of HIS+ the number of contacts barelychanged, indicating that the structural changes that already took place during equilibrationresulted in a smaller interface than in HIS0. As the increase of the RMSD of the backboneatoms during the equilibration of HIS+ showed, the interactions between the histidines andthe micro-environment after double-protonation were strong enough to change the backbonestructure despite the positional restraints on the backbone, demonstrating the meta-stabilityof the structure. In the simulation HIS+ the large decreases in the subunit contacts of domainsI reflect the partial separation of the subunits, which is clearly visible inside the loops of thefigure 8-shaped protein (Fig. 5.9). This suggests that the protonation-dependent disassemblyof the dimer begins at the interface between domains I and II. The partial separation wasrapid, whereas the complete dissociation of a protein this large is expected to require muchmore time.

The increase in the number of contacts during the simulation HIS0 correlates with therearrangement of domains III (Fig. 5.12). Furthermore, this increase and the simultaneousdecrease in the number of contact atoms (Tab. 5.7) indicate compaction of the atoms in thecontact regions. The compaction of particles increases the number of contacts, i.e. the moredense a cluster of particles is, the more dense the network of possible interactions becomes.The effect was observed in HIS0 and in HIS+ and reflects the equilibration of the structure.The fact that in simulation HIS+ the number of contacts remained similar despite the partialseparation of the subunits again indicates compaction of the interacting atoms.

91

5.4 Discussion

The interface of domain III

As domain III remained close to its initial position in both simulations, no post-fusion-specific interactions were expected at the domain I–III interface (Fig. 5.12). His144, Thr353and Glu368 were part of a network of interactions at this interface (Tab. 5.6 and data notshown). The interactions in this network that formed only in the simulation HIS+, in partic-ular the salt bridge His144-Glu368, may represent intermediate interactions that occur onlyafter histidine protonation, at the onset of the conformational change. His317 was involvedin a network of interactions at the interface between domains I and III and the fusion loopof the other subunit. In the post-fusion crystal structure both His282 and His317 were inclose proximity to the critical residue Glu368. Although the distance between His282 andGlu368 slightly exceeded the cutoff used for determining salt bridge interactions, this wasthe only attractive interaction in the micro-environment of His282, suggesting that at lowpH Glu368 forms a salt bridge both with His317 and with His282. The salt bridge His282-Glu368 would attach domain III to domain I, analogous to the attachment via Arg9-Glu368in the pre-fusion structure. Therefore Glu368 may be the key residue for the attachmentof dIII to dI In both conformations. The absence of the salt bridge His282-Glu368 in thepost-fusion crystal structure may be an artefact due to single-protonation of His282 at theexperimental pH 7.

His144

In the simulation HIS+ multiple stable interactions with His144 at the domain I–III inter-face were observed. Despite the initial interaction distance of 0.42 nm, a stable salt bridgeHis144-Glu368 formed in both subunits immediately after protonation. No correspondinghydrogen bond was observed in HIS0, indicating a strict protonation-dependency. In thesimulations the interactions of His144 and Glu368 showed a distinct dependency on the pro-tonation state of the histidines (Tab. 5.6). The high pKa = 9.5 (SD 0.9) of His144 in HIS+70 ns after protonation illustrates the strong response through interactions with the environ-ment and presents a reciprocal measure for the sensitivity of the micro-environment to theprotonation of His144.

Like all the non-bonded interactions at the domain I–III interface in the pre-fusion struc-ture, the hydrogen bond Thr353-His144 is disrupted during the relocation of domain III atlow pH. In the simulation HIS+ this disruption was directly facilitated by the double-proton-ation of His144. HIS+ also reproduced the disruption of the salt bridge Arg9-Glu368, thatis necessary for the relocation of domain III (Tab. 5.6). Although the replacing salt bridgeHis144-Glu368 still links domain III to domain I, it may give domain III more freedom ofmotion compared to Arg9-Glu368. The salt bridge His144-Glu368 formed rapidly, even

92


before the disruption of Arg9-Glu368, leading to the transient formation of a double saltbridge. The double salt bridge Arg9-Glu368-His144 is also present in both crystal structuresof prM-E, i.e. at neutral and at low pH. The conclusion is that pr in the prM-E complexpreserves the salt bridge Arg9-Glu368, whereas double-protonation of His144 at low pHin the absence of pr leads to the disruption of Arg9-Glu368, as predicted in the simulationHIS+. Furthermore, the model of the dimeric [prM-E]2 complex predicts increases in thepKa of Arg9 and His144 (Tab. 5.4 and data not shown) through interactions with Glu60,Glu62 and Asp63 of pr (Fig. 5.6), stabilising the double salt bridge. When pr dissociates,the decrease in pKa and the competitive interaction of Asp42 may facilitate the disruption ofHis144-Glu368, which may otherwise competitively disrupt Arg9-Glu368.

The slightly acidic pKa calculated for His144 in the pre-fusion crystal structures (Tab.5.2) predict His144 to be singly-protonated at the pH 8.5 of the X-ray diffraction experi-ment [Modis et al. 2003]. The hydrogen bond complex Thr353.Oγ–Nπ.His144.Nτ–Asp42furthermore indicates single-protonation at the Nτ-atom (Tab. 5.6). Furthermore, the largedifference between the pKa values of His144 in the simulations HIS0 and HIS+ indicates thatthe micro-environment adapts to the respective protonation state (Tab. 5.2). These findingsstrongly suggest that His144 has a switch function. The low pKa ≈ 4.5 after side chain equi-libration in HIS0 suggests that this switch is not very pH-sensitive and requires a relativelylow pH to become charged. The high pKa ≈ 9.4 after protonation and equilibration suggeststhat the switch is irreversible.

The hydrogen bond His144-Asp42 is present in the pre-fusion and the post-fusion con-formation and was maintained in simulation HIS0 and HIS+ (Tab. 5.6). It restricts the posi-tion of His144 within the polar and charged micro-environment at the domain I–III interface.Although a corresponding salt bridge was expected in the crystal structure of prM-E at lowpH, 3C5X, it was absent. This absence may be an artefact due to interactions with the artifi-cial N-terminal linker between pr and E. The loop E0F0, C-terminal to His144, interacts withdomain III in the mature dimer, and presumably with pr in the immature dimer. Thereforethe interaction His144-Asp42 might be important for the stability of domain I, e.g. duringthe dissociation of the pr peptide, dimer disassembly, and the relocation of domain III.

His317

The full conservation of His317 (Tab. 5.5) and evidence from mutagenesis experiments [Fritzet al. 2008] indicate that His317 has an essential function in the low-pH-dependent fusionactivation of the E protein. The post-fusion salt bridge His317-Glu368 is likely to be thisfunction. Based on the crystal structure conformations it is assumed that His317 competeswith Arg9 for interaction with Glu368. In the simulation HIS+ a further interaction part-ner of Glu368 was His144 (Tab. 5.6). The solvent accessibility and the micro-environment

93

5.4 Discussion

determine the kinetics of protonation, and the SASA and the pKa predict that His317 is pro-tonated at a higher pH than His144 (Tabs. 5.2 and 5.3). The salt bridge His144-Glu368 (Tab.5.6) might allow domain III more freedom of motion than the salt bridge Arg9-Glu368, butit does not permit the complete relocation of domain III associated with the formation ofHis317-Glu368. Thus His144-Glu368 may explain the decrease in fusion activity of his-tidine mutants in an experimental study of the closely related tick-borne encephalitis virus(TBEV) E protein. In TBEV the fusion activity of E critically depends on His323 [Fritzet al. 2008], which is equivalent to His317 in dengue virus type 2 (DEN2). This criticaldependence may be due to the competing salt bridge equivalent to His144-Glu368 in DEN2,His146-Glu373, that would arrest domain III in an intermediate position in the absence ofHis323. If domain III shifts to such a position after the disruption of Arg9-Glu373 and theformation of His146-Glu373, that might explain the residual exposure of the fusion pep-tide and the lower trimerisation activity observed in the His323 mutant. Unfortunately noHis146 mutants of TBEV E could be generated to test this hypothesis [Fritz et al. 2008]. In-stead, MD simulations of the E dimer with doubly-protonated His317 and singly-protonatedHis144 may clarify whether the protonation of His144 is the inhibiting factor.

His323 in the TBEV E protein was shown to be essential for the pH-dependent displayof the fusion peptide, enabling it to attach to liposomes [Fritz et al. 2008]. However, despitethe presence of His323, the H248N-H287A double mutant of the TBEV E protein showedonly residual trimerisation and no fusion activity [Fritz et al. 2008]. This suggests thatthe protonation of His323 triggers only the display of the fusion peptide, but not the fullconformational change to the post-fusion conformation, and that this requires His248 orHis287, which are equivalent to His244 and His282 in DEN2.

The membrane domains

His209 and His282

The conservation of His209 and His282 suggests that these histidines play important func-tional roles. Their location at the interface between the ectodomain and the membrane do-mains of the M and the E protein furthermore suggests that the interaction with the membranedomains may be these functional roles. Considering the possible interactions with the mem-brane domains of E and M or even with the membrane, the salt bridge His209-Glu195 inthe post-fusion structure and in the simulations could be artefacts due to the absence of themembrane domains in the truncated ectodomain sE. While the structure of M is known onlyfor the membrane domain fragment, the linker to the N-terminus of E is thought to run un-derneath the dimerisation region of domain II of E, adjacent to the viral membrane [Modiset al. 2004; Zhang et al. 2003], and might interact with His209 and His282. However, the

94


micro-environments and the protonation state of His209 and His282 in situ the viral envelopecannot be determined until the atomistic structure of the M protein is resolved and more isknown on the interaction of the viral membrane with the protein. The location proposed herefor His209, close to the linker to the membrane domain of E in the post-fusion structure,is in agreement with the putative position of the C-terminal stem region of E, in the groovebetween the domains II of two subunits in the trimer [Modis et al. 2004].

Intra-domain interactions

The intra-domain salt bridge His209.Nτ-Glu195 was the only post-fusion-specific salt bridgeobserved in the simulations after histidine protonation (Tab. 5.6). However, due to the possi-bility of interactions of His209 and Glu195 with the membrane domains of M and E or eventhe membrane in situ, this salt bridge could be an artefact of the truncation of these domainsfrom the soluble ectodomain of E. In the simulation HIS+ intra-domain interactions withHis27 and His261 replaced inter-subunit interactions of the initial structure.

His94

In the simulation HIS+, a double salt bridge Glu79-His94-Glu85 formed in the elongated partof domain II (Tab. 5.6). Due to the distribution of the charge between the imidazole-nitrogenatoms of His94, the interactions are expected to be weaker than in a canonical salt bridge.The salt bridges attach the loop bc and the π-helix bc to β-strand c, thereby stabilising thestructure of domain II. This could be important when the fusion loop is pulled away from theother subunit during the low-pH dependent conformational change, or during fusion.

According to the calculated pKa of 7.6, His94 was already doubly-protonated in thepre-fusion crystal structures (Tab. 5.2). The decrease to pKa ≈ 6.8 after the side chain equili-bration of HIS0 was probably due to the singly-protonated state of His94 and the loss of thedouble salt bridge with Glu79 and Glu85. The conservation of a histidine or a basic residueat position 94 further supports the proposition that His94 is doubly-protonated at neutral pH.The conservation of Glu79 and Glu85 indicates that the double salt bridge-link is also con-served. The conclusion is that His94 has the same structural function at neutral and at acidicpH and therefore in both conformations of the E protein, and that it is not a pH-sensor in thelow-pH-dependent conformational change.

95

5.4 Discussion

pKa and solvent accessibility of the histidine residues

The pKa and the titration behaviour of a specific histidine depend on the solvent accessi-bility of the micro-environment and interactions with polar and charged groups [Bashford &Karplus 1990]. For instance, a slightly buried and/or slightly hydrophilic micro-environmentcontaining few titratable residues and no direct interaction with an acidic residue can inducea slight increase in the pKa of a histidine to a value that matches the pH of fusion, pH 6.9.A residue will act as a pH-sensor by forming new interactions in response to protonation.In addition, protonation may trigger a conformational change, after which the protonatedstate may form new, favourable interactions. If these increase the pKa, the probability ofthe residue to become deprotonated again will diminish and the residue functions as a pH-dependent switch.

As expected due to the similarity of the two pre-fusion crystal structures, the pKa valuesof the respective histidines were also similar (Tab. 5.2). The slight differences in pKa betweenthe two structures and even between the homomeric subunits of one structure, demonstratethe sensitivity of the electrostatic potential to local variations in close-range contacts [Gordonet al. 2005b]. Ideally the same force field should be used for the equilibration of the con-figurations before calculating the pKa, but the differences between the force fields, AMBER[Hornak et al. 2006] and GROMOS96 [Oostenbrink et al. 2004], were assumed to be neg-ligible due to the averaging effect in the integration of all the interactions of the proteinmicro-environment.

Based on the pKa calculations, His94 is the only histidine residue predicted to bedoubly-protonated in the pre-fusion crystal structures of sE (Tab. 5.2). In addition, the pKa

of His94 remained similar in the simulation HIS+, indicating that the micro-environment didnot respond to the protonation and that this histidine and its micro-environment do not actas a switch upon protonation. The decrease of the pKa of the singly-protonated His94 inthe simulation HIS0 indicates a loss of attractive interactions, which implies structural im-portance for the doubly-protonated form of His94. In contrast, the large changes in the pKa

values of His144 and His149 between the singly- and the doubly-protonated state, even dur-ing side chain equilibration, suggest that the environments of these histidines are particularlysensitive to the protonation of the respective histidine. His144 and His149 are located at thedomain I–III interface, thus the pKa shifts may be associated with the minor domain rear-rangements of the protein, which shows that changes in merely local interactions at low pHcan already prevent reprotonation. A large shift in pKa in only one of the subunits, such asfor His149, was considered sufficient to validate the sensitivity of the environment. The con-formational changes in HIS+ after 70 ns of simulation also led to large increases in the pKa

of His27, His158 and His317, that suggest significant changes in the micro-environments(data not shown) of these histidines.

96


The relatively low pH range of the titration curve of His261 and the fact that His261is buried in the subunit interface suggest that double-protonation of this residue requires aninitial conformational change that leads to solvent exposure. Thus His261 may not becomeprotonated until the subunits separate at low pH. A low pKa was not always connected toprotection from solvent, as the low correlation between the pKa (Tab. 5.2) and the SASA(Tab. 5.3) shows. For instance His282 and His317 were solvent-exposed in the pre-fusionstructure and exhibited large pKa shifts in HIS+ after protonation, demonstrating that solvent-exposure does not prevent a residue from acting as a pH-sensitive switch as long as there areresidues in the environment that act as effectors of the protonation.

The fusion index curve for fusion from within mosquito cells [Guirakhoo et al. 1993] isshaped sigmoidal. The decrease of the slope with lower pH values is presumably due to thedecrease in the number of cells as fusion progresses. Among all the titratable groups in thesE protein only the histidine residues and the termini are protonated in the pH range of fusion(Fig. 5.5 and data not shown), which strongly supports the hypothesis that the protonationof histidines is critical for the activation of the E protein for fusion. In the titration curves ofthe histidines the mid 88 % quantiles spanned on average 3.4 pH-units (Fig. 5.4). Therefore,some degree of protonation is expected at pH values 1.5–2 units above the respective pK,which is close to the upper pH threshold at which fusion is observed. Assuming that the ini-tial pH-dependent rate of the fusion index curve at higher pH values is directly proportionalto the rate of activation of the E protein, and that the histidine switch hypothesis applies,then the activation rate would be proportional to the probability of the critical histidines tobe doubly-protonated. The protonation of the critical histidines would directly determine theinitial pH-dependent rate of cell fusion. The onsets of the protonation of His27, His144 andHis317 coincide with the fusion index curve and therefore fit the protonation behaviour ofresidues that trigger the fusion activation of E. Therefore His27, His144 and His317 are con-sidered as candidates for critical histidines that trigger the low-pH-dependent conformationalchange of the E protein. The more acidic pH range of the titration of His261 suggests thatthe protonation of His261 does not trigger fusion activation. His94, His149, His158, His209,His282 and His346 start to become protonated at higher pH values, which suggests that theyalso do not trigger the activation, although the protonation of these histidines may still benecessary for the conformational change. This relates to a study of histidine mutants of theTBEV E protein, in which various trimerisation activities were observed [Fritz et al. 2008],which suggest that the histidine residues in question have distinct effects on the conformationof the protein.

The titration curves were calculated from the pre-fusion structure and therefore describethe static pKa that can be assumed for the protein before acidification only (Fig. 5.5). How-ever, the pKa can dynamically change, as proposed for His261, for instance after an initialstructural change following the protonation of putative critical histidines. The shift of the

97

5.4 Discussion

titration curves of His27 and His317 to higher pH values in HIS0 after 60 ns of simulationwas presumably due to the slight rearrangements of the domains. The mobility of the do-mains may be an artefact due to the solubilisation of the protein and the absence of a viralmembrane. At the same time the robustness of the pKa of His144 in HIS0 was probably dueto its occlusion in the protein and protection from solvent, and makes this histidine a primecandidate for a critical residue in the triggering of the conformational change. Despite theslight shift of domain III observed in the simulation HIS0, which may affect the environmentand the solvation of His144 at the domain I–III interface, the titration curve of His144 shiftedonly slightly, from pH 7.8 to pH 8.0. Assuming that more than one histidine is critical fortriggering the conformational change, the protonation of the histidine with the lowest pKa

would present the limiting factor for fusion activation and determine the pH of fusion, re-gardless of any increase in the pKa of another critical residue. The large standard deviationsof the pKa of His27, His144 and His317 in HIS0 (Tab. 5.2) suggest that the interactions withthe respective micro-environment are dynamic when these histidines are singly-protonated.In the fusion from within (FFWI) experiment by Guirakhoo et al. [1993] the infected cellsexpressed large amounts of E protein. Even if the protonation of a critical histidine was ef-fectively sustained in only a small part of the population, given the large amount of E proteinper cell this may result in enough activated E protein to trigger cell fusion.

The prM-E complex

One difficulty in determining which residues in the dengue E protein are critical for thelow-pH-dependent fusion activation is that low pH also plays a role in the maturation of theviral envelope [Allison et al. 1995; Guirakhoo et al. 1991]. The micro-environment of thehistidine residues may vary depending on the stage in the life-cycle of the virus. This isexpected in particular for His244, which lies at the interface between pr and E, due to thedissociation of the pr peptide during maturation [Li et al. 2008].

The crystal structure 3C5X was solved at the physiologically relevant pH 5.5 [Li et al.2008]. Thus the structure is expected to reflect the interactions that occur in the trans-Golgi network (TGN), e.g. the conserved salt bridge His244-Asp63. The His248Asn mutantshowed full fusion activity, suggesting that a hydrogen bond at this location may be sufficientfor the attachment of pr. The His287Ala mutant also showed full fusion activity, however,the activity was lost in the His248Asn-His287Ala double mutant, suggesting that the simul-taneous absence of His248 and His287 in TBEV E, and correspondingly His244 and His282in DEN2, has a non-local effect on the structure that impairs function [Fritz et al. 2008].The comparatively small increase in the pKa of His244 in the simulation HIS+ suggests thatthe double-protonation of His244 does not lead to the formation of significant protonation-dependent interactions. This would mean that the protonation of His244 does not trigger the

98


low-pH-dependent conformational change of E and therefore does not require protection bythe pr peptide. Rather, the function of His244 at low pH might solely be the attachment ofpr. If that is the case, the binding of pr might prevent the conformational change by stericinhibition and lock the structure in the pre-fusion conformation. In the dimeric model of theprM-E complex the prM fragment interacts not only with domain II, but also with the othersubunit and forms part of the micro-environments of His149 and His158.

Conservation of the histidine residues and their micro-environments

The high conservation rates of His94, His144, His244, His282 and His317 suggest that theycarry essential functions in the flaviviral envelope proteins. Mutations of histidine residuesHis27, His149, His158, His209 and His261 were compensated by replacements in similarlocations. His346 was the only histidine for which no or few replacements were found inthe homology structure of the dimer. The high variability in the positions of histidines inthe loop E0F0 may be compensated by the flexibility of the loop. The proximity to the prpeptide suggests that the loop E0F0 might interact with pr, possibly with parts of pr of whichthe structure is not known. If the pr peptide contains mutations that are complementary tothe mutations in the loop, this may predict such interactions. Considering the conserva-tion rates among flaviviral E proteins in addition to the titration curves (section 5.4, p. 96),His144, His317 and possibly His27 are potential candidates for pH-dependent triggers of theconformational change of the E protein.

Restructuring and dimer disassembly

Comparing the configurations of sE in the simulations HIS0 and HIS+ it can clearly be seenthat double-protonation of the histidine residues in the pre-fusion structure of the E proteinled to significant and systematic, non-random structural changes of the dimer, and to thepartial dissociation of the subunits. The immediate onset of the structural change and its pro-gression for the duration of the simulation strongly suggest that an energetically favourableprocess was triggered from a meta-stable state. The results show that histidine protonationitself has a notable and fast effect even on a large structure such as the dengue viral envelopeprotein ectodomain (88 kDa). The hinge motion between domains I and III and the expo-sure of the fusion peptide in HIS+ were observed consistently in both subunits. Importantly,in the control simulation HIS0 the domains III moved in the opposite direction to that inHIS+, which further suggests that in HIS+ this motion is a response to the protonation of thehistidine residues.

Previously Stiasny et al. [2007] proposed the structure of alkaline unfolded E protein

99

5.4 Discussion

as a model for an intermediate of the pH-dependent conformational change. However, theacid-base titration of a protein is a monotone function of pH, therefore alkaline unfoldingand acidic unfolding are distinct and diverging processes. Compared to acidic pH, basic pH10 as applied in the experiment [Stiasny et al. 2007] disrupts a different set of interactions,which may lead to the formation of again other interactions. Both the disruption and theformation of interactions depend on the specific pKa of the interaction partners. As the failureto trimerise already suggests [Stiasny et al. 2007], alkaline unfolding is unlikely to lead to anintermediate in the activation of the E protein for fusion. Furthermore, the different trends inthe structural changes observed in the simulations HIS0 and HIS+ imply that already a smallpH gradient may lead to significantly different if not opposite effects on the structure of theE protein.

The various structures of the E protein from the immature virion, the mature virionand in the post-fusion state indicate hinging between domains II and I of varying degrees[Zhang et al. 2004]. The simulation HIS+ confirmed a rotation of domain II, in concert witha buckling motion involving all three domains of a subunit and a rising motion of domainII. A pH-dependent hinge at the domain I–II interface had been described previously, basedon mutations between domains I and II that alter the pH threshold of fusion [Modis et al.2003]. This and other hinge regions were evident in the simulation HIS+ as regions of greatermobility (Fig. 5.11). Harrison and co-workers proposed that the dissociation of the E dimerallowed domains I and II to move relative to one another [Modis et al. 2004]. However, in thesimulation HIS+ the bending between domains I and II did not require the dissociation of thedimer. Buckling and crimping motions of the individual domains were observed within thesubunits and between the two subunits (Fig. 5.12). The individual domains maintained theircompactness and structure (Tab. 5.8), representing the stable mechanical modules of thesemotions. The bending motion between domain III and the fusion loop and the exposure ofthe fusion loop suggest a mechanism for the release of the fusion loop.

Due to the packing of the domains and the subunits in the dimer, cooperativity is ex-pected between the subunits, and similarly among the packed dimers in the viral envelope.Indeed, the structural rearrangements in the simulation HIS+ are symmetrical and corre-spond to the symmetry of the dimer, which suggests cooperativity between the domains andsubunits. Based on this symmetry and the shearing motions observed between the subunitsin HIS+, a model is proposed for the restructuring of the E dimer after histidine protonation:The hinge region between domains I and II of one subunit and the hinge region between thebase and elongated regions of the other subunit lie opposite each other and bend in oppositedirections, which leads to the dissociation of the subunits at these regions. The twisting ofthe dimer into a figure 8 is a first consequence of this effect. The subunit interface rotatesfrom charged to predominantly hydrophobic groups, which exposes the charged groups andthereby enhances the solvation – and dissociation – of the subunits. As the hinge angles are

100


roughly orthogonal to the membrane plane, these motions would not require more lateralspace on the viral surface. In addition the contraction at the hinges leads to a decrease in thelength of the dimer. These effects may solve a problem suggested previously by Bressanelliet al. [2004], that the surface of the virion does not provide enough lateral space for the re-arrangements of the monomers.

5.5 Summary and conclusion

The MD simulations in this study showed that double-protonation of the histidine residuesin the pre-fusion dimer of the dengue viral type 2 envelope protein ectodomain sE triggersspecific structural changes that are related to the low-pH-dependent conformational changeof the protein. The specific interactions that characterise the pre-fusion and the post-fusionconformation of the sE dimer were determined and compared in the prM-E complex andin configurations from the simulations of the sE dimer. In the simulations double-proton-ation of the histidines was sufficient to induce the disruption of the pre-fusion-specific saltbridge Arg9-Glu368, which is a critical step for the pH-dependent rearrangement of domainIII. A detailed structural analysis and the theoretical estimation of the protonation statesof the histidine residues enabled the prediction if and how the histidine residues and theirmicro-environments respond to the pH of fusion. In the pre-fusion structure the pH at whichHis27, His144 and His317 start to become protonated is also the pH threshold at which fu-sion starts. Therefore the protonation of these histidines could be the events that trigger theconformational change of the E protein. His27, His144, His317 and also His261 qualified aspH-sensors. The major increases in the pKa of His144 and His261 due to structural changesin their micro-environments after protonation indicate that the protonation of these histidinesis effectively irreversible, and may explain the irreversibility of the conformational change.The protection of His144 and His261 from solvent allows these histidines to discriminate be-tween small pH fluctuations and sustained acidification as found in the endosome. ThereforeHis144 and His261 may be critical for preserving the meta-stable conformation until the pHis sufficiently low.

The conservation and the conformation-dependent interactions suggest that His144 andHis317 are critical for the activation of the dengue viral fusion protein. A mechanism wasproposed for the release of the fusion loop from domain I, based on the competitive interac-tions of the doubly-protonated His144 and His317 with Glu368. Both His144 and His317are predicted to disrupt Arg9-Glu368 after protonation, but leading to different intermediatestructures. The attachment of the pr pro-peptide to the mature prM-E complex is predictedto be the essential function of the conserved His244. His94 was the only histidine predicted

101

5.5 Summary and conclusion

to be doubly-protonated in the pre-fusion structure at neutral pH and may play an importantrole in the stabilisation of domain II.

The simulations provide detailed insight into initial steps in the protonation-dependentrestructuring of the dengue viral envelope protein ectodomain. Histidine protonation trig-gered the contraction of the subunits of the sE dimer, the partial dissociation of the subunitinterface and the systematic restructuring into a twisted figure 8 conformation. His244 andHis261 formed various interactions at the subunit interface. The simulations confirm therotation of domain II and a hinge motion between domains II and I, which were previouslyproposed in experimental studies. The salt bridge His261-Glu269 is proposed to promote thehinge motion between domains I and II. Buckling motions were observed in the subunits ofthe dimer, that resulted in the projection of domains I towards the membrane and in shearingbetween the subunits.

This simulation study demonstrates that histidine protonation efficiently triggers con-formational change. The effectiveness seen here and the generality of the principle of pro-tonation at low pH suggest that it may also apply to other low-pH-dependent viral fusionproteins. Whether the structural changes of the ectodomain sE observed here predict anintermediate of the pH-dependent conformational change of the E protein remains to be val-idated experimentally. Nonetheless the importance of the protonation state of the histidinesfor the structure of the sE protein and the systematic effect on both subunits of the dimersuggest that these structural changes model initial steps in the activation of the sE dimer. Inaddition, the results of this study provide new targets for further experimental and theoreticalinvestigations of flaviviral E protein.

102

Model for the activationof flaviviral fusionproteins 66.1 Introduction

The envelope protein E of a flavivirus is composed of a soluble ectodomain and a C-terminal,membrane-associated domain, the so-called stem-anchor region (Fig. 5.3). The ectodomainis a large structure of approximately 400 residues, suspended on the surface of the viral mem-brane (Fig. 3.2A). The truncated and solubilised fragment sE contains only the ectodomain.The stem-anchor region, or membrane domain, consists of approximately 100 residues. Itssecondary structure was predicted to contain two membrane-peripheral helices, the stem he-lices H1 and H2, and two trans-membrane helices, the anchor helices T1 and T2, that areconnected by flexible linkers [Zhang et al. 2003]. The flaviviral envelope proteins are struc-turally and functionally similar to the envelope proteins of the alphaviruses. The two groupsare therefore collectively referred to as class II viral fusion proteins.

At low pH the E protein dimers dissociate into monomers and assemble into trimers[Allison et al. 1995; Modis et al. 2004]. The E trimer and the sE trimer both are more stablethan the corresponding dimers. Furthermore the E trimer is more temperature-stable than thesE trimer [Stiasny et al. 2005]. To date the exact mechanism by which flaviviral E proteinsinduce fusion is not known, nor the atomic structure of the stem-anchor region. Thereforeit remains to be shown what the specific function of the stem-anchor region is, and whetherit plays an essential role in the fusion mechanism, other than attaching the E protein to theviral membrane.

The existing models of class II viral fusion assume that the active form of E is trimeric

103

6.1 Introduction

[Harrison 2008; Jardetzky & Lamb 2004], in analogy to the active, trimeric forms of alphavi-ral and influenza viral fusion proteins.1 However, there is no experimental evidence that inflaviviruses trimerisation occurs before fusion, and therefore it is unclear whether trimeri-sation is a requirement for the activation of flaviviral fusion proteins. In fusion assays oftick-borne encephalitis virus (TBEV), dengue virus (DEN2) and their respective recombi-nant subviral particles, pre-exposure to low pH resulted in the loss of fusion activity [Corveret al. 2000; Stiasny et al. 2002]. Due to the rapid and irreversible inactivation of flaviviral Eprotein after exposure to low pH, the post-fusion crystal structures of flaviviral sE proteins(PDB-IDs: 1OK8 [Modis et al. 2004], 1URZ [Bressanelli et al. 2004])2 most likely representthe inactive form of the E protein ectodomain.

The kinetics of fusion were found to differ significantly between influenza virus, thealphavirus Semliki Forest virus (SFV), and the flavivirus TBEV. From this it was concludedthat the fusion active state of the TBE viral E protein may fundamentally differ from the ac-tive states of the envelope proteins in influenza and Semliki Forest virus [Corver et al. 2000].Furthermore, in TBEV the kinetics of inactivation are slow compared to the kinetics of fu-sion, which suggests that the inactivation may involve additional conformational changes inthe E protein [Corver et al. 2000]. The trimerisation step would be a good candidate for aslow, concluding step in the conformational change of the E protein. Thus the notion thatinactivation results from trimerisation fits the kinetics of the E protein.

The inactivity of the trimer might be due to: i) the inability of the fusion peptide toinsert into the target membrane, due to crowding among the three fusion loops, resultingin restricted mobility, unfavourable positioning with respect to the lipid head groups, andrepulsion from the membrane; ii) the simple fact that the trimer is held upright on the viralsurface by symmetry constraints and neighbouring trimers, which act as spacers between theviral and target membranes and thereby prevent fusion. Point ii is in line with a proposalby Corver et al. [2000], that “the fast rate of fusion and the low activation energy of this[fusion] process [...] may be a consequence of the flat orientation of the E protein [on theviral surface], which, by lying parallel to the interacting membranes, would facilitate theestablishment of direct molecular contact between them.” They concluded that the trimerformed under low-pH conditions may represent a final fusion-inactive structure, and that aminor conformational change may be sufficient for fusion [Corver et al. 2000]. Based onthese considerations, this chapter investigates the minimal structural requirements for theconformational change of flaviviral E protein.

The exact structural dynamics of flaviviral E protein and any associated effects on

1 Although influenza viral envelope proteins were classified as class I viral fusion proteins, they have similarrequirements for fusion and are also activated by low pH [White et al. 2008].

2 The X-ray crystal structures available of flaviviral E proteins in the post-fusion conformation were ob-tained from crystals of the soluble ectodomain sE.

104

6 Fusion mechanism of class II viral fusion proteins

Figure 6.1 Models of the mature dimer of the full-length E protein at neutral and at low pH.A) Superposition of the Cα-atoms of the cryo-EM structure 1P58 (cyan tubes) onto the crystalstructure 1OAN [Modis et al. 2003; Zhang et al. 2003]. Red � domain dI, yellow � dII, blue � dIII.B) Perspective view of the protein rotated by 20 degrees around a vertical axis compared with Aor C. C) Cross-eyed stereo image. The model in B and C was generated by superposition of thecon�guration of the protonated sE dimer from the simulation HIS+ after 70 ns onto the backboneatoms of the pre-fusion crystal structure of the ectodomain, 1OAN (A). The two subunits arecoloured yellow and pink respectively.

the viral or the host membrane are unknown. However, some general functions of fusion-mediating proteins are known and include: the local bending of the membranes in order tominimise inter-membrane repulsion; to induce local dehydration at the membrane contact;to promote stalk formation by generating bilayer stress [Chernomordik & Kozlov 2003]. Infact, simulation studies of membranes [Leontiadou et al. 2007; Marrink & Mark 2004] andof bilayer fusion [Noguchi & Takasu 2001] support the so-called stalk-pore hypothesis ofmembrane fusion [Zimmerberg & Chernomordik 1999], in which during the initial stage offusion the contacting leaflets of the fusing membranes connect and form a stalk.

The propensity for two membranes to fuse depends on the spontaneous curvature of themembranes, which in turn depends on the lipid composition [Chernomordik & Kozlov 2003].In experiments with liposomes the absence of specific lipids impaired viral fusion [Lee et al.2008; Stiasny & Heinz 2004; Stiasny et al. 2003]. Experimental [Fujii et al. 1993; Murataet al. 1991] and simulation studies [Pécheur et al. 1999] have shown that the interactionwith small amphiphilic peptides can alter the spontaneous curvature of a membrane and leadto fusion. Larger proteins also induce curvature in membranes, for example the so-called

105

6.2 Methods

BAR domains (bin, amphiphysin and Rvs, members of the amphiphysin protein family)[Peter et al. 2004; Saarikangas et al. 2009], which bind preferentially to curved membranesand are found in proteins that regulate the remodeling of membranes, e.g. the shaping ofcompartments and organelles, or the recycling of synaptic vesicles [Gallop & McMahon2005; Peter et al. 2004; Saarikangas et al. 2009]. MD simulations predicted that the N-BARdomain of amphiphysin induces a mean curvature of 0.15 nm−1 [Ayton et al. 2007].

In MD simulations of the DEN2 sE dimer double-protonation of all the histidine residuesled to a conformational change (see Chapter 5), while the protein remained dimeric. Specif-ically, hinge motions between the domains resulted in domain I to project towards where themembrane would be located in situ (Fig. 6.1) and predict interactions between the proteinand the viral membrane. In the current chapter new models of intermediate conformations ofthe E protein are introduced, some of which include the membrane domain. The conforma-tional change observed during the MD simulation of the protonated sE dimer is evaluated interms of interactions between the ectodomain and the stem-anchor region. The implicationsof these interactions for the activation of flaviviral E proteins are discussed. A fusion modelis proposed, in which fusion occurs before trimerisation. This fusion model is supplementedby a structural model of an intermediate after acidification, that was generated by MD simu-lation.

6.2 Methods

Sequence variability analysis The sequence variability of the full-length E protein wasdetermined for the 28 flaviviral E protein sequences of the conserved domain (CD) super-family pfam00869 that also code for the membrane domain of E [Marchler-Bauer et al.2009]. The sequence variability was determined as the Shannon entropy [PVS; Shannon1948]

H(l) =−20

∑r=1

f (r, l)log2 f (r, l) (6.1)

The sequence similarity diagram was generated using the WebLogo web server [Crooks2009; Crooks et al. 2004; Schneider & Stephens 1990].

The relative frequency f of a residue type r at position l and the measure for sequenceinformation Rseq determine the height L of the letter of the residue type r in the WebLogodiagram

L(l) = f (r, l)Rseq(l) (6.2)

and is given in [bits per residue] [Schneider & Stephens 1990]. Rseq is derived from the corre-

106


sponding Shannon entropy H and a correction term e(n) for small sample sizes n [Schneideret al. 1986]

Rseq(l) = E(Hnr)−H(l) (6.3)

= log220− (H(l)+ e(n)) (6.4)

(n = 28 ⇒ e(n) ≈ 0.5). Hnr is the uncertainty of obtaining a particular combination of nresidues of types r

Hnr =−20

∑r=1

(nr

n

)log2

(nr

n

)(6.5)

where n =20∑r

nr is the number of sequences without gaps, and Pr is the general frequency of

residue type r [Schneider et al. 1986], which is simplified to equiprobability for all residuetypes, i.e. Pr = 1

20 ∀r [Schneider & Stephens 1990]. The average of the uncertainty Hnr isgiven by the expectation value

E(Hnr) = ∑all nr

PnrHnr (6.6)

weighted by the probability Pnr of obtaining the effective combination:

Pnr =n!

20∏r

nr!

20

∏r

Pnrr (6.7)

Atomistic model of the full-length E protein An atomistic, full-residue model of themembrane domain was generated from the Cα-trace of a cryo-EM structure (PDB-ID: 1P58[Zhang et al. 2003]) using the Geno3D web server [Combet et al. 2002]. Atomistic struc-tures of the ectodomain of the E dimer were obtained from X-ray crystal structures (PDB-IDs: 1OAN and 1OKE [Modis et al. 2003]) and from the configuration of an MD simulation(Chapter chapter 5). These were combined with the model of the membrane domain, bysuperposition onto the cryo-EM structure of the full-length E dimer, 1P58.

107

6.3 Results

6.3 Results

The low-pH-dependent conformational change of the E protein

The pre-fusion and the post-fusion conformation of the DEN2 E protein ectodomain werecompared to determine the differences in the tertiary structure that result from the conforma-tional change. In order to visualise the rearrangement of the domains, the crystal structures ofthe backbone in the two conformations of sE were superimposed (PDB-IDs: 1OAN [Modiset al. 2003], 1OK8 [Modis et al. 2004]); the two superpositions are shown in Figure 6.2A.In the left panel of the figure the superposition of domain I, the so-called organising domainof E, shows that domains II and III both move in similar directions relative to domain I.Accordingly, in the superposition of domains II and III, shown in the right panel, the mostobvious difference is the position of domain I. In the following these two superpositions arecalled “SI” and “SII+III”, after the respective domains used for the fits.

In the superposition SII+III the main difference between the two conformations is theposition of domain I, which appears to be released or ejected from between domains II andIII during the conformational change. Figure 6.2B shows a model of a possible intermediateconformation (orange) after the release of domain I. This model is a hybrid of the two super-imposed structures, assembled from domain I of the post-fusion structure and domains II andIII of the pre-fusion structure (orange). In the figure the pre-fusion structure (yellow), thepost-fusion structure (red) and the hybrid model (orange) are compared in various dimericassemblies. These were assembled by superposition onto the subunits of the pre-fusion dimer(Fig. 3.2A p. 33).

Another noticeable difference between the pre-fusion and the post-fusion structure inSII+III (Fig. 6.2A) is the position of domain II (yellow). Domain II is shifted by approxi-mately 9 nm along the long principal axis of the dimer towards domain I (red). Based onthis finding, the hybrid model was modified in a second modelling step (Fig. 6.3), wheredomain II (orange) was shifted by 9.0 nm (green), towards domain I (red). As can be seenin the figure, this shift eliminates the steric clash between the fusion loop (blue) and domainIII (orange), that is still present in the opposite homomer where no shift was applied. Inaddition, the shift separates the helices in the central region of the subunit interface.

In order to evaluate the fits of the superpositions (Fig. 6.2A), the RMSD between thesuperimposed structures was calculated. The RMSD of all the backbone atoms was 2.1 nmin SI and 1.3 nm in SII+III. The shift of the center of mass of domain III between the super-imposed structures was 3.4 nm in SI and in SII+III it was 1.2 nm, i.e. less than half of theshift in SI. The shift of the C-terminus was 4.0 nm in SI and 2.1 nm in SII+III; the C-terminusis located on the surface of domain III. In terms of these three criteria, SII+III matched the

108


Figure 6.2 Model of the transition between the pre-fusion and the post-fusion conformation ofthe E protein (PDB-ID: 1OAN [Modis et al. 2003], 1OK8 [Modis et al. 2004]). A) Superpositionsof the crystal structures, on domain I (SI, left), and on domains II and III (SII+III, right). Domaincolouring scheme: red � domain dI; yellow � base region of dII; green: fusion region of dII; blue:dIII. B) The pre-fusion dimer of E (1OAN, top left) and three dimeric models that combine thepre-fusion form of E (yellow), the hybrid model (orange) and the post-fusion form of E (red).The hybrid model is a combination of the pre-fusion structures of domains dII and dIII and thepost-fusion structure of dI from the superposition SII+III (A). In each panel, clockwise from topleft, the two subunits of the dimer are transformed successively from the pre-fusion form to thehybrid model to the post-fusion form.

conformations more closely than SI and therefore seems the better choice for modelling thetransition between the two conformations.

Interaction between the ectodomain and the stem-anchor region

An atomistic model of the full-length E dimer was obtained by superposition of the pre-fusion crystal structure of the E protein dimer and of an atomistic model of the membranedomain onto the cryo-EM structure of the mature E protein. The resulting structure is shownin Figures 6.1A and 6.4. In the cryo-EM structure residues 395–399, which link the C-

109

6.3 Results

Figure 6.3 Conformational hybrid model of the pre-fusion E protein dimer, based on the super-position SII+III. View of the envelope exterior. In addition to the model introduced above (Fig.6.2B orange), in one subunit of the dimer domain II (dII' green, blue) was shifted by 9 nm towardsdomain dI (dI'). Orange/grey: pre-fusion structure; blue: fusion loop.

terminus of the ectodomain to the C-terminal membrane domain, were not resolved. TheRMSD between the Cα-atoms of 1OAN and 1P58 was 0.5 nm. The conservation of the full-length E protein was analysed to predict conserved interactions at the interface between theectodomain and the stem-anchor region. The conserved domain database CDD identified 38representative sequences in the flaviviral glycoprotein superfamily (pfam00869 [Marchler-Bauer et al. 2009; Marchler-Bauer & Bryant 2004]). Twenty-eight of these sequences alsocode for the stem anchor region and were analysed in terms of sequence variability; the resultis presented as a WebLogo diagram in Figure 6.5.

For the prediction of salt bridges between the ectodomain and the membrane domainand of hydrophobic interactions between the protein and the viral membrane, the conser-vation of charged, or large and hydrophobic residues at the protein surface was analysed.Figure 6.6 indicates the degree of conservation of the residues at the surface of the E protein:Colours indicate full conservation, lighter shades of grey higher conservation, and darkershades less conservation. The surface of the ectodomain facing the viral membrane (Fig.6.6C) was generally more conserved than the external surface. Figure 6.6C shows the mem-brane domain rotated by 180 ◦ around a vertical axis into an open-book presentation along-side the ectodomain. On the surface of the ectodomain, Asp10, Phe11, Glu26, Lys110,His244, His282 and His317 were fully conserved. These residues cluster around the inter-faces between domains I, II’ and III, opposite the stem helices (Fig. 6.4). Trp212, Trp220and Phe240, which are located on the surface at the hinge region between the base and theelongated part of domain II (Fig. 6.2, yellow and green respectively) were fully conserved.In the stem helices and the CS linker, Arg407, Arg411, Asp417, Trp420, Asp421, Phe422,Lys434, His437 and Phe448 were conserved. In the atomistic model of the stem-anchor re-gion, Asp421 and His437 formed a salt bridge between the CS linker and the stem helix H2.

110


Figure 6.4 The membrane-facing surface of the full-length E protein: the ectodomain (surface)and the Cα-trace of the membrane domain (blue tube). From left to right: The structure isrotated in steps of 30 ◦ around a vertical axis, from the view of the surface facing the membraneto the side view. Red boxes highlight the extensive groove on the surface of domain I. The grey-scale colour gradient indicates the conservation of the residues (cf. Fig. 6.6), white correspondsto full conservation. Domain colouring scheme: rose � domain dI, yellow/orange � dII of the tworespective subunits, blue � dIII.

The CS linker is highly conserved [Zhang et al. 2003] (Fig. 6.5) and contains a sequence ofsmall residues 423–427, GSLGG.

In Figure 6.6C it can be seen that the stem helix H1 and the CS linker align with highlyconserved residues on the surfaces of domains I and III. The conserved, charged residuesArg411 and Asp417 of H1 align with the conserved, complementary charged residues Asp10,Glu26 and His282 on domain I (Fig. 6.6). In Figure 6.4 it can furthermore be seen that theseconserved residues are located along the interfaces of domains I (rose) and III (blue) withdomain II’ (orange), the fusion domain of the opposite subunit. On the surface of domainI (rose) a long groove extends around approximately one third of the circumference of thedomain; in Figure 6.4 red boxes highlight the groove. The groove begins approximatelyopposite the C-terminus of stem helix H1 and leads along a series of highly conserved hy-drophobic residues, Val24, Cys185 and Cys285, across the β-sheet BIH.

Structural intermediate in the conformational change of the full-lengthE protein

An MD simulation was performed to test the effect of double-protonation of all the histidineresidues on the structure of the ectodomain. The stem-anchor region was modelled into theconfiguration of the sE dimer after 70 ns of simulation to determine whether the conforma-tional changes observed were relevant to interactions with the membrane domain of E. Themodel was generated by superposition onto the cryo-EM structure of domains II and III ofthe dimer, in line with the model SII+III established above (p. 108). The model is shown in

111

6.3 Results

Figu

re6.5Se

quence

similarityam

ong28

�aviv

iralE

protein

s.Th

eheigh

tofa

lettera

taparticular

position

isprop

ortio

naltothefre

quency

ofthe

corre

spon

ding

resid

uetype

amon

gthealign

edsequ

encesa

ndto

thesequ

ence

inform

ation

inthealign

menta

ndisgiv

enin

bits

[Crooks2

009;

Croo

kset

al.20

04;S

chneider

&Stephens

1990

].Th

eresid

uenu

mberin

gisbasedon

thedeng

uevir

altype

2Eprotein

,gi|316

1578

7.Th

elet

ters

arecolou

red

accordingto

theresid

uetype:r

ed�acidic,

blue

�basic

,orang

e�histidine,

green�po

lar,rose�hydrop

hobic,

black�lar

geandhydrop

hobic.

112


Figure 6.6 Surface of the ectodomain and the membrane domain of the pre-fusion E protein.The three views A�C of the ectodomain result from rotations by 90 ◦ around a vertical axis.A) View of the envelope exterior surface. B) Side view. C) The surface of the ectodomain facingthe membrane, and the membrane domain of one subunit are shown side-by-side in open-bookpresentation, i.e. the membrane domain was rotated by 180 ◦ around a vertical axis. In themembrane domain only the charged and the large hydrophobic side chains are shown in atomicdetail, the predicted α-helices are shown as cylinders. The grey-scale colour gradient indicates theShannon entropy H :]0;3.441]→]white;black] as a measure of the conservation of the residues,lighter shades correspond to higher conservation. Fully conserved residues (H = 0) are colouredaccording to residue type: red � acidic, blue � basic, green � polar, yellow � hydrophobic.

Figures 6.1B–C, in which can be seen that in both subunits the domains I arch out towards therespective membrane domain. Furthermore, structural elements of domains I and II extendbetween the stem helices and the anchor helices respectively.

The steric overlaps in the superposition, between domain I and the membrane domain,are due to the deformation of the protein and the bias of the superposition by the referencedomains dII and dIII. In order to alleviate these overlaps, the positions of the membranedomains were modelled individually for each subunit, as shown in Figure 6.7. The internalstructures of the stem-anchor region were preserved. In both subunits of this model, domainI interacts with the CS linker and separates the stem helices. Due to the amphipathic natureof the stem helices [Zhang et al. 2003], they are expected to remain suspended in the headgroup region of the membrane. The stem helix H1 would be constrained by the C-terminusof the ectodomain (purple beads), while the stem helix H2 may move at the end of the flexi-ble CS linker.

113

6.3 Results

Figure 6.7 previous page Model of an intermediate from the MD simulation of the E protein afterhistidine protonation, interacting with the membrane domain. The con�guration of the ectodomainsE (surface) after 70 ns of MD simulation was superimposed onto domains II and III of the cryo-EM structure of the E dimer. A) Side view. B) Cross-eyed stereo view of that side of the Eprotein that faces the membrane. In B the rigid structures of the two stem-anchor regions (bluetubes) of the dimer were modelled individually to �t the con�guration of the respective subunit ofthe ectodomain. Fully conserved residues are represented by beads coloured according to residuetype: red � acidic, blue � basic, green � polar, white � hydrophobic; C-terminus � purple. Domaincolouring scheme: rose � domain dI, yellow/orange � dII, blue � dIII.

114


6.4 Discussion

The low-pH-dependent conformational change of the E protein

Until now the conformational change from the pre-fusion to the post-fusion structure hascommonly been described as a motion of domain III relative to the other two domains (Figs.3.1, 6.2A left) [Harrison 2008; Modis et al. 2004]. While this is true, one must consider thatthe process occurs in the envelope environment and possibly involves the viral membrane.Here, an alternative view on the conformational change is introduced (Figs. 6.2, 6.3), thatconsiders the restrictive membrane environment and other experimental findings. The newmodel of the conformational change presented here is based on the same structures as theprevious model, but uses a different set of atoms as reference for the domain motions. Thiscondenses the conformational change to a motion of domain I, which is released from itsmeta-stable position between domains II and III (SII+III, Fig. 6.2A right).

From structural comparisons similar to the superposition SI (Fig. 6.2A left) it was previ-ously concluded that the conformational change required more lateral space than is availablein the packed envelope. In search of a solution to this problem it was even proposed that a ra-dial expansion of the viral membrane might accommodate the conformational change at lowpH [Bressanelli et al. 2004], but this seems unlikely from a physical and chemical perspec-tive. In contrast the superposition onto domains II and III, SII+III (Fig. 6.2A right), showsthat the post-fusion structure fits into the initial flat arrangement of the pre-fusion dimer,where it would occupy a similar area on the viral membrane as the pre-fusion structure (Fig.6.2B). Therefore no additional space is needed in this model of the conformational changeof the subunit, for the release of domain I.

The structural rearrangements in the new model are less dramatic than in the previousmodel and may therefore occur on a faster time scale. This stands in agreement with theextraordinarily fast rate of fusion from which Corver et al. [2000] inferred that the TBE viralE protein was active in a flat orientation and that the activation may therefore require only aminimal conformational change. The new model presented here reconciles a flat orientationwith the complete conformational change concluded from the crystal structures (excludingthe oligomeric rearrangement into trimers; Figs. 6.2, 6.3). Even though the protein may notremain in a perfectly flat orientation during the entire fusion process, the new model showsthat the domain rearrangements could, in theory, be completed to a large degree in the flatdimer.

The new model furthermore suggests that the dissociation of the dimer and the spikingof the envelope might occur after most of the domain rearrangements have already takenplace. This matches the kinetics of the E protein, which suggest that after the fast activation

115

6.4 Discussion

a slow conformational change follows that leads to inactivation [Corver et al. 2000]. Theassembly of the trimer is likely to be a time-consuming process and may therefore correspondto the slow inactivation step. Fusion would occur in between the fast and the slow step,after an initial conformational change into a yet unknown intermediate state. A model forthe active intermediate conformation of flaviviral E protein, based on the results of MDsimulations, (Chapter chapter 5) is proposed in the following sections.

Interaction between the ectodomain and the stem-anchor region

In the superposition of the intermediate configuration of sE, obtained from the simulation ofsE after histidine protonation, onto the structure of the full-length E protein, the juxtapositionof the deformation of domain I with the membrane domain of the cryo-EM structure isstriking (Figs. 6.1B–C, 6.7). The strip of highly conserved residues on domains I and IIIis directly opposite the stem helix H1 in the pre-fusion conformation and strongly suggeststhat this is an important interface (Fig. 6.4). The alignment of H1 with the interface betweendomains I and III and the other subunit suggests that interactions between H1 and domains Iand III may play a role in the dissociation of the fusion domain.

The interactions with the stem helices H1 and H2 that were predicted from the interme-diate configuration suggest that the deformation of domain I observed during the simulationmight affect the positions of the helices (Fig. 6.7). Specifically, domain I might interact withthe CS linker connecting the helices and induce a separation between H1 and H2 to eitherside of the dimer. The proximity of the extended groove on the surface of domain I (Fig. 6.4)suggests that the small residues of the linker might run through this groove and hold the stemhelices H1 and H2 in place. This would involve the disruption of the salt bridge between thelinker and H2 (Fig. 6.6). His282 of domain I is located close to this salt bridge, thereforedouble-protonation of His282 might facilitate the disruption of the salt bridge.

Stiasny et al. [2007] claim to have generated an intermediate of the fusion pathwayof a flaviviral E protein by applying alkaline pH, which led to the reversible dissociationof the icosahedral protein envelope. However, the protein was unable to mediate fusion[Stiasny et al. 2007], whereas activation by low pH rapidly leads to fusion. This impliesthat at high pH the conformational change follows a different pathway than at low pH. Moreimportantly, the reversibility of the alkaline state indicates that this state lay outside theactivation pathway, as this is an irreversible process. In contrast, the structural changesobserved in the MD simulations in this thesis were triggered by the protonation of histidineresidues, i.e. by the modelling of a specifically low-pH dependent effect that would notoccur at alkaline pH. Therefore the intermediate configuration from the simulation is likelyto represent an activated intermediate of the sE protein.

116


Model for flaviviral membrane fusion mediated by dimeric E protein

The MD simulations of the sE ectodomain were performed in the absence of the membranedomain, therefore one might expect the structural changes to differ in the full-length E pro-tein in situ the membrane. However, little force is required to bend a membrane [Zimmerberg& Chernomordik 1999], therefore the membrane might follow the structural changes of theprotein. The structural changes observed in the simulation of the sE ectodomain after pro-tonation predict increases in the local curvatures of the viral and the host membrane due tointeractions between the ectodomain and the membranes at low pH.

The fully conserved residues on the protein surface facing the viral membrane in situ(Fig. 6.6) suggest specific interactions with the stem-anchor region and with the viral mem-brane. The highly conserved tryptophans and phenylalanines that are found on both surfacesof domain II might interact with the viral and with the target membrane. The hinge motionbetween domains I and II in the simulation resulted in a curved conformation of the sE dimer(Fig. 6.7A). In situ, interactions between domain I in this conformation, the stem-anchor re-gion and the viral membrane might promote the bending of the viral membrane.

One requirement for fusion in the so-called stalk-pore hypothesis are out-of-plane ther-mal fluctuations of the bilayers [Zimmerberg & Chernomordik 1999]. In addition, low pHhas been reported to result in leakage from liposomes [Drummond et al. 2000], which sug-gests that low pH increases the thermal fluctuations of the membrane and may thereby pro-mote low-pH-dependent fusion. It is thought that the lipid composition bends the membraneand thereby controls the local concentration and the activity of specific proteins in the mem-brane [Chernomordik & Kozlov 2003]. Reciprocally, a higher protein concentration, as givenon the viral envelope surface, may in turn determine the local lipid composition and mem-brane bending. In flaviviruses, the stem-anchor regions of both the E and the M protein3

might facilitate the bending of the viral membrane. The insertion of the fusion peptide mighthave a similar effect on the target membrane. The membrane domain of M is positionedunderneath the base region of domain II (Fig. 5.3). Strong deformations were observed indomain II during the simulation, that may affect the membrane underneath.

In the intermediate configuration from the simulation the fusion peptide was fully ex-posed (Fig. 5.12B). Exposure of the fusion peptide is required for it to attach to the targetmembrane, thus the structure fulfills this essential prerequisite for fusion. During the sim-ulation the domains II lifted slightly and exposed the fusion peptide, while the protein re-mained dimeric and close to its initial flat conformation. In this conformation the dimericprotein might already bind to a target membrane. The buckling motions of the protein do-mains during the simulation resulted in a conformation in which the “underside”, i.e. the

3 The M protein is cleaved off the N-terminus of the E protein during the maturation of the viral particleand remains in the viral membrane [Li et al. 2008; Zhang et al. 2003]

117

6.4 Discussion

membrane-facing side, of the fusion region of domain II (orange and yellow respectively)became almost level with the “upside” of domain I (rose) of the respective other subunit(Fig. 6.7A). With a slight increase in the amplitude of the buckling motion beyond what wasobserved during the 70 ns of simulation, attachment of the membranes to the protein couldbring the viral membrane, attached to domain II (yellow or orange), into close proximitywith the target membrane, attached to domain I (rose) of the other subunit. When the sub-units then separate, the two membranes could come into contact between the subunits (Fig.6.7B, 5.9) and form a stalk.

Figure 6.8 shows a schematic of the proposed mechanism. Panels A1–3 show crosssections of the viral and the target membrane interacting with the E protein; the sections runparallel to the long principal axis of the E dimer. The black dotted lines indicate a furtherset of cross sections at 90 ◦, shown in panels B1–3. The membrane domain of E is impliedand anchors the ectodomain to the viral membrane. In the figure the E protein ectodomainis sandwiched between the viral and the target membrane (A1–2, B1–2). The constructionof the fusion intermediate (A2) was based on the configuration of the ectodomain sE after70 ns of MD simulation in water. The domain motions observed during the simulation (A2)were extended to obtain the model of a hypothetical active intermediate of the E proteinectodomain (A3, B3). The display of the membranes focuses at two different depths relativeto the viewer, indicated by transparent green and pink slabs (A1–3) or boxes (B1–3) respec-tively. The small inset at the top shows the position of these slabs in bird’s eye view of theviral surface.

The proposed mechanism is as follows: 1) In the pre-fusion conformation at neutral pH,the E protein ectodomain lies flat on the surface of the viral membrane, depicted in Figure 6.8as a bar beneath the protein; the upper bar represents the target membrane (A1, B1). 2) Thestructural changes facilitate the insertion of the fusion loops into the target membrane andthe bending of the membranes. In the two subunits the structural changes of the dimer andthe associated local deformations of the membranes (green and pink slab) move in oppositedirections. 3) The E protein in the putative active intermediate conformation (A3, B3). Themembranes are brought together between the subunits of the dimer, where they merge andform a stalk, labelled in the figure with a *-symbol. Alternatively the membranes mergebetween two adjacent dimers, as illustrated by a second stalk in the left part of B3.

In the cryo-EM structure of the mature pre-fusion E protein, the transmembrane helicesT1 and T2 of E are located directly underneath the holes between the subunits of the dimer.Thus the protein structure presents no barrier against the possible bending of the viral mem-brane towards the target membrane by T1 and T2. The fully conserved Phe448, located inthe middle of the linker between stem helix H2 and anchor helix T1, might insert into thetarget membrane and stabilise the putative activated conformation of the protein-membrane

118


Figure 6.8 Schematic of the model proposed for �aviviral membrane fusion mediated by dimericE protein. The dimeric E protein ectodomain is sandwiched in between the viral and the targetmembrane. A/B1) The protein and the membranes are in a �at conformation. A/B2) Structuralchanges of the protein promote the bending of the membranes. A/B3) Peak deformations inadjacent sections of the membranes make contact and form fusion stalks, labelled with *-symbols.A) Cross sections through the viral and the target membrane, parallel to the long principal axis ofthe E dimer. The membranes are displayed in two slabs, coloured green and pink respectively, andde�ned by the membrane normal and the orientations of the subunits of the dimer as depicted inthe small inset at the top. The green slab lies in front of the pink slab. The model in A2 wasconstructed using the con�guration of the ectodomain sE after 70 ns of MD simulation with allhistidines doubly-protonated. B) Cross sections of the models as indicated by black dotted lines inA, perpendicular to the long principal axis of the protein. The transparent green and pink boxescorrespond to the membrane slabs shown in A. Filled ovals and circles represent the domains ofthe E dimer ectodomain: red � domain dI, yellow/orange � dII, blue � dIII.

119

6.4 Discussion

complex shown in Figure 6.8AB3. Alternatively, this insertion might perturb the membranesand promote the merger of the membranes.

Based on structural similarities between the fusion proteins of flaviviruses and al-phaviruses it was suggested that the proteins are related [Kielian & Rey 2006; Strauss &Strauss 2001]. As there is no sequence identity between flaviviral and alphaviral envelopeproteins, the relationship, if any, is distant and the split from a putative common ancestormust have occurred a long time ago. One common functional property among all the low-pHdependent viral envelope proteins known is trimerisation. For instance influenza hemagglu-tinin and the active forms of alphaviral envelope proteins and class III viral fusion proteinsare all trimeric [Backovic & Jardetzky 2009; Gibbons et al. 2004; Weis et al. 1990]. Basedon this common property it was proposed that trimerisation is essential for fusion [Earp et al.2004]. This said, the sequence of events during fusion is not known, and it has been sug-gested that flaviviral fusion might occur in the absence of trimer formation [Stiasny et al.2007]. From the simultaneous absence of fusion and trimerisation in mutants [Fritz et al.2008] it cannot be concluded that fusion requires trimerisation, only that they are correlated.Moreover, a trimeric form of the E protein that mediates fusion implies that there are twodifferent trimeric states, i.e. one active and one inactive. However, only an inactive trimerconformation is known.

In the life cycle of enveloped viruses there are some common environmental constraintswhich present similar evolutionary constraints on the structures and may have led indepen-dently to the parallel evolution of similar structures and fusion mechanisms. Such constraintsare for instance the membrane environment and topology before and after fusion, the cellcompartment during the biogenesis of the envelope, or the assembly of the viral particle. Inthe model proposed here for flaviviral fusion, the E protein binds to the target membrane andmediates fusion in the dimeric form, and trimerisation is a consequence of the separation ofthe dimers, but is not relevant for fusion. This hypothetical mechanism of fusion might haveevolved only recently, from an ancestor in which trimerisation was necessary for fusion.Then the flaviviral fusion protein might still display trimerisation after fusion, as a struc-tural atavism from the ancestral protein. Clearly, the model for flaviviral membrane fusionproposed here remains to be validated experimentally. This could be done by cross-linkingexperiments, e.g. with extended linkers that allow the buckling of the protein and the partialseparation of the dimer, but prevent trimerisation.

120


6.5 Conclusion

A new model is proposed for the low-pH-dependent conformational change of the flaviviralenvelope protein ectodomain, in which domain I is released from its meta-stable positionbetween domains II and III. This model is based on a specific fit of crystallographic data andhas the advantage that less space is required for the conformational change than previouslyassumed.

MD simulations of the dengue viral envelope protein ectodomain sE in the pre-fusion,dimeric conformation led to protonation-dependent structural changes that predict interac-tions between domain I of the ectodomain and the membrane domain of E. This configurationis proposed as an activated intermediate of the ectodomain at low pH and was used for mod-elling the interactions with the membrane domain and the membrane environment. A modelfor the mechanism of flaviviral membrane fusion is proposed, in which the deformations ofthe ectodomain promote bending of the viral and the target membrane and lead to the mergerof the membranes and stalk formation. This model differs from previous models in that theactive conformation of the E protein is dimeric and sandwiched in between the viral and thetarget membrane in a relatively flat orientation. Importantly, in the mechanism proposed theviral and the host membrane remain close to each other.

Both the new model of the conformational change and the new mechanism of fusionfit the kinetics of flaviviral membrane fusion, which differ significantly from the kinetics offusion in other viruses including the alphaviruses. One important conclusion is a differentsequence of conformational changes than previously assumed, where the trimerisation stepcomes after fusion and leads to the inactivation of the protein.

121

Bibliography

Åstrand, P. O., A. Wallqvist & G. Karlström. Molecular dynamics simulations of 2 M aque-ous urea solutions. J. Phys. Chem. 98(33), 8224–8233 (1994)

Ahmad, Z., S. Yadav, F. Ahmad & N. Z. Khan. Effects of salts of alkali earth metals andcalcium chloride on the stability of cytochrome c and myoglobin. Biochim. Biophys. Acta1294(1), 63–71 (1996)

Allison, S. L., J. Schalich, K. Stiasny, C. W. Mandl & F. X. Heinz. Mutational evidencefor an internal fusion peptide in flavivirus envelope protein E. J. Virol. 75(9), 4268–4275(2001)

Allison, S. L., J. Schalich, K. Stiasny, C. W. Mandl, C. Kunz & F. X. Heinz. Oligomericrearrangement of tick-borne encephalitis virus envelope proteins induced by an acidic pH.J. Virol. 69(2), 695–700 (1995)

Allison, S. L., K. Stiasny, K. Stadler, C. W. Mandl & F. X. Heinz. Mapping of functionalelements in the stem-anchor region of tick-borne encephalitis virus envelope protein E. J.Virol. 73(7), 5605–5612 (1999)

Ambler, R. P. The amino acid sequence of Pseudomonas cytochrome c-551. Biochem. J. 89,349–380 (1963)

Arteca, G. A., C. T. Reimann & O. Tapia. Proteins in vacuo: Denaturing and folding mecha-nisms studied with computer-simulated molecular dynamics. Mass Spectrom. Rev. 20(6),402–422 (2001)

Ayton, G. S., P. D. Blood & G. A. Voth. Membrane remodeling from N-BAR domain inter-actions: Insights from multi-scale simulation. Biophys. J. 92(10), 3595–3602 (2007)

Backovic, M. & T. S. Jardetzky. Class III viral membrane fusion proteins. Curr. Opin. Struct.Biol. 19(2), 189–196 (2009)

Baker, E. N. & R. E. Hubbard. Hydrogen bonding in globular proteins. Prog. Biophys. Mol.Biol. 44(2), 97–179 (1984)

Baker, N. A., P. H. Hünenberger & A. J. McCammon. Polarization around an ion in adielectric continuum with truncated electrostatic interactions. J. Chem. Phys. 110(22),10679–10692 (1999)

123

Barlow, D. J. & J. M. Thornton. Ion-pairs in proteins. J. Mol. Biol. 168(4), 867–885 (1983)

Bashford, D. & M. Karplus. pKa’s of ionizable groups in proteins: atomic detail from acontinuum electrostatic model. Biochemistry 29(44), 10219–10225 (1990)

Beck, D. A., B. J. Bennion, D. O. Alonso & V. Daggett. Simulations of macromolecules inprotective and denaturing osmolytes: Properties of mixed solvent systems and their effectson water and protein structure and dynamics. In: Osmosensing and Osmosignaling, vol.428 of Methods Enzymol., pp. 373–396. Elsevier Academic Press Inc. (2007)

Bennion, B. J. & V. Daggett. The molecular basis for the chemical denaturation of proteinsby urea. Proc. Natl. Acad. Sci. U. S. A. 100(9), 5142–5147 (2003)

Berendsen, H. J. C., J. P. M. Postma, A. DiNola & J. R. Haak. Molecular dynamics withcoupling to an external bath. J. Chem. Phys. 81(8), 3684–3690 (1984)

Berendsen, H. J. C., J. P. M. Postma, W. F. van Gunsteren & J. Hermans. Interaction modelsfor water in relation to protein hydration. In: Pullman, B., ed., Intermolecular Forces, pp.331–342. Reidel, Dordrecht (1981)

Bergdorf, M., C. Peter & P. H. Hünenberger. Influence of cut-off truncation and artificialperiodicity of electrostatic interactions in molecular simulations of solvated ions: A con-tinuum electrostatics study. J. Chem. Phys. 119(17), 9129–9144 (2003)

Bhuyan, A. K. Protein stabilization by urea and guanidine hydrochloride. Biochemistry41(45), 13386–13394 (2002)

Bhuyan, A. K. & R. Kumar. Kinetic barriers to the folding of horse cytochrome c in thereduced state. Biochemistry 41(42), 12821–12834 (2002)

Bolen, W. D. & G. D. Rose. Structure and energetics of the hydrogen-bonded backbone inprotein folding. Ann. Rev. Biochem. 77(1), 339–362 (2008)

Brautigan, D. L., B. A. Feinberg, B. M. Hoffman, E. Margoliash, J. Preisach & W. E. Blum-berg. Multiple low spin forms of the cytochrome c ferrihemochrome. EPR spectra ofvarious eukaryotic and prokaryotic cytochromes c. J. Biol. Chem. 252(2), 574–582 (1977)

Bressanelli, S., K. Stiasny, S. L. Allison, E. A. Stura, S. Duquerroy, J. Lescar, F. X. Heinz &F. A. Rey. Structure of a flavivirus envelope glycoprotein in its low-pH-induced membranefusion conformation. EMBO J. 23(4), 728–738 (2004)

Brunori, M., M. G. Bigotti, F. Cutruzzolà, S. Gianni & C. Travaglini-Allocatelli. Cytochromec551 as a model system for protein folding. Biophys. Chem. 100(1-3), 409–419 (2003)

Bullough, P. A., F. M. Hughson, J. J. Skehel & D. C. Wiley. Structure of influenza haemag-glutinin at the pH of membrane fusion. Nature 371(6492), 37–43 (1994)

Caballero-Herrera, A. & L. Nilsson. Urea parametrization for molecular dynamics simula-tions. J. Mol. Struct. 758(2-3), 139–148 (2006)

124

Caballero-Herrera, A., K. Nordstrand, K. D. Berndt & L. Nilsson. Effect of urea on peptideconformation in water: Molecular dynamics and experimental characterization. Biophys.J. 89(2), 842–857 (2005)

Cai, M., E. G. Bradford & R. Timkovich. Investigation of the solution conformation ofcytochrome c-551 from pseudomonas stutzeri. Biochemistry 31(36), 8603–8612 (1992)

Camilloni, C., A. Guerini Rocco, I. Eberini, E. Gianazza, R. A. Broglia & G. Tiana. Ureaand guanidinium chloride denature protein L in different ways in molecular dynamicssimulations. Biophys. J. 94(12), 4654–4661 (2008)

Carneiro, F. A., F. Stauffer, C. S. Lima, M. A. Juliano, L. Juliano & A. T. Da Poian. Mem-brane fusion induced by vesicular stomatitis virus depends on histidine protonation. J.Biol. Chem. 278(16), 13789–13794 (2003)

CDC. Dengue Fever Fact Sheet – CDC Division of Vector-Borne Infectious Diseases(DVBID) (2008). URL http://www.cdc.gov/NCIDOD/DVBID/dengue. Modificationdate 7 May 2008

Chanel-Vos, C. & M. Kielian. A conserved histidine in the ij loop of the Semliki Forest virusE1 protein plays an important role in membrane fusion. J. Virol. 78(24), 13543–13552(2004)

Chen, J., J. B. Anderson, C. Deweese-Scott, N. D. Fedorova, L. Y. Geer, S. He, D. I. Hurwitz,J. D. Jackson, A. R. Jacobs, C. J. Lanczycki, C. A. Liebert, C. Liu, T. Madej, A. Marchler-Bauer, G. H. Marchler, R. Mazumder, A. N. Nikolskaya, B. S. Rao, A. R. Panchenko,B. A. Shoemaker, V. Simonyan, J. S. Song, P. A. Thiessen, S. Vasudevan, Y. Wang, R. A.Yamashita, J. J. Yin & S. H. Bryant. MMDB: Entrez’s 3D-structure database. Nucl. AcidsRes. 31(1), 474–477 (2003)

Chen, J., K. H. Lee, D. A. Steinhauer, D. J. Stevens, J. J. Skehel & D. C. Wiley. Structureof the hemagglutinin precursor cleavage site, a determinant of influenza pathogenicity andthe origin of the labile conformation. Cell 95(3), 409–417 (1998)

Chernomordik, L. V. & M. M. Kozlov. Protein-lipid interplay in fusion and fission of bio-logical membranes. Ann. Rev. Biochem. 72(1), 175–207 (2003)

Chitra, R. & P. E. Smith. Molecular association in solution: A Kirkwood-Buff anal-ysis of sodium chloride, ammonium sulfate, guanidinium chloride, urea, and 2,2,2-trifluoroethanol in water. J. Phys. Chem. B 106(6), 1491–1500 (2002)

Collins, J. R., P. Du & G. H. Loew. Molecular dynamics simulations of the resting and hy-drogen peroxide-bound states of cytochrome c peroxidase. Biochemistry 31(45), 11166–11174 (1992)

Colman, P. M. & M. C. Lawrence. The structural biology of type I viral membrane fusion.Nat. Rev. Mol. Cell Biol. 4(4), 309–319 (2003)

Combet, C., M. Jambon, G. Deléage & C. Geourjon. Geno3D: automatic comparative molec-ular modelling of protein. Bioinformatics 18(1), 213–214 (2002)

125

Corver, J., A. Ortiza, S. L. Allison, J. Schalich, F. X. Heinz & J. Wilschut. Membrane fusionactivity of tick-borne encephalitis virus and recombinant subviral particles in a liposomalmodel system. Virology 269(1), 37–46 (2000)

Creighton, T. Electrophoretic analysis of the unfolding of proteins by urea. J. Mol. Biol.129(2), 235–264 (1979)

Crooks, G. E. WebLogo - Create Sequence Logos (2009). URL http://weblogo.berkeley.edu. Modification date 9 September 2009

Crooks, G. E., G. Hon, J.-M. Chandonia & S. E. Brenner. WebLogo: A Sequence LogoGenerator. Genome Res. 14(6), 1188–1190 (2004)

Da Poian, A. T., F. A. Carneiro & F. Stauffer. Viral membrane fusion: Is glycoprotein G ofrhabdoviruses a representative of a new class of viral fusion proteins? Brazilian J. Med.Biol. Res. 38(6), 813–823 (2005)

Daniels, R. S., J. C. Downie, A. J. Hay, M. Knossow, J. J. Skehel, M. L. Wang & D. C. Wiley.Fusion mutants of the influenza virus hemagglutinin glycoprotein. Cell 40(2), 431–439(1985)

Daura, X., K. Gademann, B. Jaun, D. Seebach, W. F. van Gunsteren & A. E. Mark. Peptidefolding: When simulation meets experiment. Angew. Chem.-Int. Edit. 38(1-2), 236–240(1999)

Dickerson, R. E., T. Takano, D. Eisenberg, O. B. Kallai, L. Samson, A. Cooper & E. Mar-goliash. Ferricytochrome c. I. general features of the horse and bonito proteins at 2.8 åresolution. J. Biol. Chem. 246(5), 1511–1535 (1971)

Diederix, R. E. M., S. Busson, M. Ubbink & G. W. Canters. Increase of the peroxidaseactivity of cytochrome c-550 by the interaction with detergents. J. Mol. Catal. B-Enzym.27(2-3), 75–82 (2004)

Diederix, R. E. M., M. Ubbink & G. W. Canters. The peroxidase activity of cytochromec-550 from Paracoccus versutus. Eur. J. Biochem. 268(15), 4207–4216 (2001)

Diederix, R. E. M., M. Ubbink & G. W. Canters. Effect of the protein matrix of cytochromec in suppressing the inherent peroxidase activity of its heme prosthetic group. Chem-BioChem 3(1), 110–112 (2002a)

Diederix, R. E. M., M. Ubbink & G. W. Canters. Peroxidase activity as a tool for studyingthe folding of c-type cytochromes. Biochemistry 41(43), 13067–13077 (2002b)

Dobson, C. M., A. Sali & K. M. Protein folding: a perspective from theory and experiment.Angew. Chem.-Int. Edit. 37(7), 868–893 (1998)

Doms, R. W., A. Helenius & J. White. Membrane fusion activity of the influenza virushemagglutinin. The low pH-induced conformational change. J. Biol. Chem. 260(5), 2973–2981 (1985)

Drummond, D. C., M. Zignani & J.-C. Leroux. Current status of pH-sensitive liposomes indrug delivery. Prog. Lipid Res. 39(5), 409–460 (2000)

126

Duffy, E. M., D. L. Severance & W. L. Jorgensen. Urea – potential functions, log-P, andfree-energy of hydration. Isr. J. Chem. 33(3), 323–330 (1993)

Dumortier, C., T. E. Meyer & M. A. Cusanovich. Protein dynamics: Imidazole binding toclass I C-type cytochromes. Arch. Biochem. Biophys. 371(2), 142–148 (1999)

Earp, L., S. Delos, H. Park & J. White. In: Membrane Trafficking in Viral Replication, vol.285 of Curr. Top. Microbiol. Immunol., chap. The many mechanisms of viral membranefusion proteins, pp. 25–66. Springer Berlin Heidelberg (2004)

Eisenhaber, F., P. Lijnzaad, P. Argos, C. Sander & M. Scharf. The double cubic latticemethod: Efficient approaches to numerical integration of surface area and volume and todot surface contouring of molecular assemblies. J. Comput. Chem. 16(3), 273–284 (1995)

Elcock, A. H. The stability of salt bridges at high temperatures: implications for hyperther-mophilic proteins. J. Mol. Biol. 284(2), 489–502 (1998)

Elöve, G. A., A. K. Bhuyan & H. Roder. Kinetic mechanism of cytochrome c folding:Involvement of the heme and its ligands. Biochemistry 33(22), 6925–6935 (1994)

Fass, D., S. C. Harrison & P. S. Kim. Retrovirus envelope domain at 1.7 Å resolution. Nat.Struct. Mol. Biol. 3(5), 465–469 (1996)

Fedurco, M., J. Augustynski, C. Indiani, G. Smulevich, M. Antalík, M. Bánó, E. Sedlák,M. Glascock & J. Dawson. The heme iron coordination of unfolded ferric and ferrous cy-tochrome c in neutral and acidic urea solutions. Spectroscopic and electrochemical studies.BBA-Proteins Proteomics 1703(1), 31–41 (2004)

Frishman, D. & P. Argos. Knowledge-based protein secondary structure assignment. Pro-teins 23(4), 566–579 (1995)

Fritz, R., K. Stiasny & F. X. Heinz. Identification of specific histidines as pH sensors inflavivirus membrane fusion. J. Cell Biol. 183(2), 353–361 (2008)

Fujii, G., M. E. Selsted & D. Eisenberg. Defensins promote fusion and lysis of negativelycharged membranes. Protein Sci. 2(8), 1301–1312 (1993)

Gadsby, P. M., J. Peterson, N. Foote, C. Greenwood & A. J. Thomson. Identification of theligand-exchange process in the alkaline transition of horse heart cytochrome c. Biochem.J. 246(1), 43–54 (1987)

Gallop, J. L. & H. T. McMahon. BAR domains and membrane curvature: bringing yourcurves to the BAR. In: Lipids, Rafts and Traffic, vol. 72 of Biochem. Soc. Symp., pp.223–231 (2005)

Gamblin, S. J., L. F. Haire, R. J. Russell, D. J. Stevens, B. Xiao, Y. Ha, N. Vasisht, D. A.Steinhauer, R. S. Daniels, A. Elliot, D. C. Wiley & J. J. Skehel. The structure and receptorbinding properties of the 1918 influenza hemagglutinin. Science 303(5665), 1838–1842(2004)

127

George, P. The chemical nature of the second hydrogen peroxide compound formed bycytochrome c peroxidase and horseradish peroxidase. 1. Titration with reducing agents.Biochem. J. 54(2), 267–276 (1953)

Gibbons, D. L., I. Erk, B. Reilly, J. Navaza, M. Kielian, F. A. Rey & J. Lepault. Visualiza-tion of the target-membrane-inserted fusion protein of Semliki Forest virus by combinedelectron microscopy and crystallography. Cell 114(5), 573–583 (2003)

Gibbons, D. L., M. C. Vaney, A. Roussel, A. Vigouroux, B. Reilly, J. Lepault, M. Kielian &F. A. Rey. Conformational change and protein-protein interactions of the fusion protein ofSemliki Forest virus. Nature 427(6972), 320–325 (2004)

Gollins, S. W. & J. S. Porterfield. pH-dependent fusion between the flavivirus West Nile andliposomal model membranes. J. Gen. Virol. 67(1), 157–166 (1986)

Gordon, J. C., J. B. Myers, T. Folta, V. Shoja, L. S. Heath & A. Onufriev. H++: A serverfor estimating pKas and adding missing hydrogens to macromolecules. Nucleic Acids Res.33(Suppl. 2), W368–W371 (2005a)

Gordon, J. C., J. B. Myers, T. Folta, V. Shoja, L. S. Heath & A. Onufriev. H++: A serverfor estimating pKas and adding missing hydrogens to macromolecules (2005b). URLhttp://biophysics.cs.vt.edu/H++

Graziano, G. Size and temperature dependence of hydrocarbon solubility in concentratedaqueous solutions of urea and guanidine hydrochloride. Canadian Journal of Chemistry80(4), 388–400 (2002)

Guirakhoo, F., R. A. Bolin & J. T. Roehrig. The Murray Valley encephalitis virus prM proteinconfers acid resistance to virus particles and alters the expression of epitopes within theR2 domain of E glycoprotein. Virology 191(2), 921–931 (1992)

Guirakhoo, F., F. X. Heinz, C. W. Mandl, H. Holzmann & C. Kunz. Fusion activity of fla-viviruses: comparison of mature and immature (prM-containing) tick-borne encephalitisvirions. J. Gen. Virol. 72(6), 1323–1329 (1991)

Guirakhoo, F., A. Hunt, J. Lewis & J. Roehrig. Selection and partial characterization ofdengue 2 virus mutants that induce fusion at elevated pH. Virology 194(1), 219–223 (1993)

Güntert, P., C. Mumenthaler & K. Wüthrich. Torsion angle dynamics for NMR structurecalculation with the new program DYANA. J. Mol. Biol. 273(1), 283–298 (1997)

Ha, Y., D. J. Stevens, J. J. Skehel & D. C. Wiley. H5 avian and H9 swine influenza virushaemagglutinin structures: possible origin of influenza subtypes. EMBO J. 21(5), 865–875(2002)

Harbury, H. A. & P. A. Loach. Interaction of nitrogenous ligands with heme peptides frommammalian cytochrome c*. J. Biol. Chem. 235(12), 3646–3653 (1960)

Harrison, S. C. Mechanism of membrane fusion by viral envelope proteins. In: Virus Struc-ture and Assembly, vol. 64 of Adv. Virus Res., pp. 231–261. Elsevier Academic Press Inc(2005)

128

Harrison, S. C. Viral membrane fusion. Nature Struct. Mol. Biol. 15(7), 690–698 (2008)

Heinz, F. X. Molecular aspects of tbe virus research. Vaccine 21(Suppl. 1), S3–S10 (2003)

Helenius, A. Alphavirus and flavivirus glycoproteins: Structures and functions. Cell 81(5),651–653 (1995)

Helenius, A., J. Kartenbeck, K. Simons & E. Fries. On the entry of Semliki Forest virus intoBHK-21 cells. J. Cell Biol. 84(2), 404–420 (1980)

Hess, B., H. Bekker, H. J. C. Berendsen & J. G. E. M. Fraaije. LINCS: A linear constraintsolver for molecular simulations. J. Comput. Chem. 18(12), 1463–1472 (1997)

Hoang, L., H. Maity, M. M. Krishna, Y. Lin & W. S. Englander. Folding units govern thecytochrome c alkaline transition. J. Mol. Biol. 331(1), 37–43 (2003)

Hoffman, L. R., I. D. Kuntz & J. M. White. Structure-based identification of an inducerof the low-pH conformational change in the influenza virus hemagglutinin: irreversibleinhibition of infectivity. J. Virol. 71(11), 8808–8820 (1997)

Holbrook, M. R., H. Ni, R. E. Shope & A. D. Barrett. Amino acid substitution(s) in the stem-anchor region of Langat virus envelope protein attenuates mouse neurovirulence. Virology286(1), 54–61 (2001)

Hong, X. & D. W. Dixon. NMR study of the alkaline isomerization of ferricytochrome c.FEBS Lett. 246(1-2), 105–108 (1989)

Hornak, V., R. Abel, A. Okur, B. Strockbine, A. Roitberg & C. Simmerling. Comparison ofmultiple Amber force fields and development of improved protein backbone parameters.Proteins 65(3), 712–725 (2006)

Huang, G. S. & T. G. Oas. Submillisecond folding of monomeric lambda repressor. Proc.Natl. Acad. Sci. U. S. A. 92(15), 6878–6882 (1995)

Humphrey, W., A. Dalke & K. Schulten. VMD: Visual molecular dynamics. J. Mol. Graph.14(1), 33–38 (1996)

Hurrelbrink, R. J. & P. C. McMinn. Attenuation of Murray Valley encephalitis virus by site-directed mutagenesis of the hinge and putative receptor-binding regions of the envelopeprotein. J. Virol. 75(16), 7692–7702 (2001)

Ibarra-Molero, B., G. I. Makhatadze & C. R. Matthews. Mapping the energy surface for thefolding reaction of the coiled-coil peptide GCN4-p1. Biochemistry 40(3), 719–731 (2001)

Idrissi, A. Molecular structure and dynamics of liquids: aqueous urea solutions. Spectroc.Acta Pt. A-Molec. Biomolec. Spectr. 61(1-2), 1–17 (2005)

Idrissi, A., F. Sokolic & A. Perera. A molecular dynamics study of the urea/water mixture.J. Chem. Phys. 112(21), 9479–9488 (2000)

Isin, B., P. Doruker & I. Bahar. Functional motions of influenza virus hemagglutinin: Astructure-based analytical approach. Biophys. J. 82(2), 569–581 (2002)

129

IUPAC Commission on the Nomenclature of Organic Chemistry and IUPAC-IUB Commis-sion on Biochemical Nomenclature. Nomenclature of α-amino acids. (recommendations,1974). Biochemistry 14(2), 449–462 (1975)

Jardetzky, T. S. & R. A. Lamb. A class act. Nature 427(6972), 307–308 (2004)

Kabsch, W. & C. Sander. Dictionary of protein secondary structure: Pattern recognition ofhydrogen-bonded and geometrical features. Biopolymers 22(12), 2577–2637 (1983)

Kamoun, P. P. Denaturation of globular proteins by urea: Breakdown of hydrogen or hy-drophobic bonds? Trends Biochem. Sci. 13(11), 424–425 (1988)

Kampmann, T., D. S. Mueller, A. E. Mark, P. R. Young & B. Kobe. The role of histidineresidues in low-pH-mediated viral membrane fusion. Structure 14(10), 1481–1487 (2006)

Khoshtariya, D. E., T. D. Dolidze, S. Seifert, D. Sarauli, G. Lee & R. van Eldik. Kinetic,thermodynamic, and mechanistic patterns for free (unbound) cytochrome c at Au/SAMjunctions: Impact of electronic coupling, hydrostatic pressure, and stabilizing/denaturingadditives. Chem.-Eur. J. 12(27), 7041–7056 (2006)

Kielian, M. Class II virus membrane fusion proteins. Virology 344(1), 38–47 (2006)

Kielian, M. & A. Helenius. pH-induced alterations in the fusogenic spike protein of SemlikiForest virus. J. Cell Biol. 101(6), 2284–2291 (1985)

Kielian, M. & F. A. Rey. Virus membrane-fusion proteins: More than one way to make ahairpin. Nat. Rev. Microbiol. 4(1), 67–76 (2006)

Kieseritzky, G., G. Morra & E.-W. Knapp. Stability and fluctuations of amide hydrogenbonds in a bacterial cytochrome c: a molecular dynamics study. J. Biol. Inorg. Chem.11(1), 26–40 (2006)

Kimura, T., J. C. Lee, H. B. Gray & J. R. Winkler. Site-specific collapse dynamics guide theformation of the cytochrome c’ four-helix bundle. Proc. Natl. Acad. Sci. U. S. A. 104(1),117–122 (2007)

Kobe, B., R. J. Center, B. E. Kemp & P. Poumbourios. Crystal structure of human T cellleukemia virus type 1 gp21 ectodomain crystallized as a maltose-binding protein chimerareveals structural evolution of retroviral transmembrane proteins. Proc. Natl. Acad. Sci.U. S. A. 96(8), 4319–4324 (1999)

Kuhn, R. J., W. Zhang, M. G. Rossmann, S. V. Pletnev, J. Corver, E. Lenches, C. T. Jones,S. Mukhopadhyay, P. R. Chipman, E. G. Strauss, T. S. Baker & J. H. Strauss. Structure ofdengue virus. Cell 108(5), 717–725 (2002)

Kumar, R., P. N. Prabhu, K. D. Rao & A. K. Bhuyan. The alkali molten globule state ofhorse ferricytochrome c: Observation of cold denaturation. J. Mol. Biol. 364(3), 483–495(2006)

La Penna, G., S. Furlan & L. Banci. Molecular statistics of cytochrome c: structural plasticityand molecular environment. J. Biol. Inorg. Chem. 12(2), 180–193 (2007)

130

Laasonen, K., M. Sprik, M. Parrinello & R. Car. “ab initio” liquid water. J. Chem. Phys.99(11), 9080–9089 (1993)

Latypov, R. F., H. Cheng, N. A. Roder, J. Zhang & H. Roder. Structural characterization ofan equilibrium unfolding intermediate in cytochrome c. J. Mol. Biol. 357(3), 1009–1025(2006)

Lazar, G. A., S. A. Marshall, J. J. Plecs, S. L. Mayo & J. R. Desjarlais. Designing proteinsfor therapeutic applications. Curr. Opin. Struct. Biol. 13(4), 513–518 (2003)

Lee, C. J., H. R. Lin, C. L. Liao & Y. L. Lin. Cholesterol effectively blocks entry of flavivirus.J. Virol. 82(13), 6470–6480 (2008)

Lee, E., R. C. Weir & L. Dalgarno. Changes in the dengue virus major envelope protein onpassaging and their localization on the three-dimensional structure of the protein. Virology232(2), 281–290 (1997)

Leontiadou, H., A. E. Mark & S.-J. Marrink. Ion transport across transmembrane pores.Biophys. J. 92(12), 4209–4215 (2007)

Lescar, J., A. Roussel, M. W. Wien, J. Navaza, S. D. Fuller, G. Wengler, G. Wengler &F. A. Rey. The fusion glycoprotein shell of Semliki Forest virus: An icosahedral assemblyprimed for fusogenic activation at endosomal pH. Cell 105(1), 137–148 (2001)

Levinthal, C. Are there pathways for protein folding? J. Chim. Phys.-Chim. Biol. 65(1),44–45 (1968)

Li, L., S.-M. Lok, I.-M. Yu, Y. Zhang, R. J. Kuhn, J. Chen & M. G. Rossmann. The fla-vivirus precursor membrane-envelope protein complex: structure and maturation. Science319(5871), 1830–1834 (2008)

Lim, W. K., J. Rösgen & S. W. Englander. Urea, but not guanidinium, destabilizes proteinsby forming hydrogen bonds to the peptide group. Proc. Natl. Acad. Sci. U. S. A. 106(8),2595–2600 (2009)

Lindahl, E., B. Hess & D. van der Spoel. Gromacs 3.0: A package for molecular simulationand trajectory analysis. J. Mol. Model. 7(8), 306–317 (2001)

Lommen, A., A. Ratsma, N. Bijlsma, G. W. Canters, J. E. van Wielink, J. Frank & J. vanBeeumen. Isolation and characterization of cytochrome c550 from the methylamine-oxidizing electron-transport chain of Thiobacillus versutus. Eur. J. Biochem. 192(3), 653–661 (1990)

Luo, G., A. Torri, W. E. Harte, S. Danetz, C. Cianci, L. Tiley, S. Day, D. Mullaney, K. L. Yu,C. Ouellet, P. Dextraze, N. Meanwell, R. Colonno & M. Krystal. Molecular mechanismunderlying the action of a novel fusion inhibitor of influenza A virus. J. Virol. 71(5),4062–4070 (1997)

MacKerell, A., D. Bashford, M. Bellott, R. Dunbrack, J. D. Evanseck, M. J. Field, S. Fis-cher, J. Gao, H. Guo, S. Ha, D. Joseph-McCarthy, L. Kuchnir, K. Kuczera, F. T. K.Lau, C. Mattos, S. Michnick, T. Ngo, D. T. Nguyen, B. Prodhom, W. Reiher, B. Roux,

131

M. Schlenkrich, J. C. Smith, R. Stote, J. Straub, M. Watanabe, J. Wiórkiewicz-Kuczera,D. Yin & M. Karplus. All-atom empirical potential for molecular modeling and dynamicsstudies of proteins. J. Phys. Chem. B 102(18), 3586–3616 (1998)

Makhatadze, G. I. & P. L. Privalov. Protein interactions with urea and guanidinium chloride:A calorimetric study. J. Mol. Biol. 226(2), 491–505 (1992)

Mandl, C. W. Steps of the tick-borne encephalitis virus replication cycle that affect neu-ropathogenesis. Virus Res. 111(2), 161–174 (2005)

Mandl, C. W., H. Kroschewski, S. L. Allison, R. Kofler, H. Holzmann, T. Meixner & F. X.Heinz. Adaptation of tick-borne encephalitis virus to BHK-21 cells results in the formationof multiple heparan sulfate binding sites in the envelope protein and attenuation in vivo.J. Virol. 75(12), 5627–5637 (2001)

Mao, Y., M. A. Ratner & M. F. Jarrold. Molecular dynamics simulations of the charge-induced unfolding and refolding of unsolvated cytochrome c. J. Phys. Chem. B 103(45),10017–10021 (1999a)

Mao, Y., M. A. Ratner & M. F. Jarrold. Molecular dynamics simulations of the rehydrationof folded and unfolded cytochrome c ions in the vapor phase. J. Am. Chem. Soc. 123(27),6503–6507 (2001)

Mao, Y., J. Woenckhaus, J. Kolafa, M. A. Ratner & M. F. Jarrold. Thermal unfolding ofunsolvated cytochrome c: Experiment and molecular dynamics simulations. J. Am. Chem.Soc. 121(12), 2712–2721 (1999b)

Marchler-Bauer, A., J. B. Anderson, F. Chitsaz, M. K. Derbyshire, C. Deweese-Scott, J. H.Fong, L. Y. Geer, R. C. Geer, N. R. Gonzales, M. Gwadz, S. He, D. I. Hurwitz, J. D. Jack-son, Z. Ke, C. J. Lanczycki, C. A. Liebert, C. Liu, F. Lu, S. Lu, G. H. Marchler, M. Mul-lokandov, J. S. Song, A. Tasneem, N. Thanki, R. A. Yamashita, D. Zhang, N. Zhang &S. H. Bryant. CDD: specific functional annotation with the Conserved Domain Database.Nucl. Acids Res. 37(Database issue), D205–210 (2009)

Marchler-Bauer, A. & S. H. Bryant. CD-Search: protein domain annotations on the fly.Nucleic Acids Res. 32(Web Server issue), W327–331 (2004)

Marrink, S. J., A. H. de Vries & A. E. Mark. Coarse grained model for semiquantitative lipidsimulations. J. Phys. Chem. B 108(2), 750–760 (2004)

Marrink, S.-J. & A. E. Mark. Molecular view of hexagonal phase formation in phospholipidmembranes. Biophys. J. 87(6), 3894–3900 (2004)

Marsh, M. The entry of enveloped viruses into cells by endocytosis. Biochem. J. 218(1),1–10 (1984)

Marszalek, P. E., H. Lu, H. Li, M. Carrion-Vazquez, A. F. Oberhauser, K. Schulten & J. M.Fernandez. Mechanical unfolding intermediates in titin modules. Nature 402(6757), 100–103 (1999)

132

Mayor, U., C. M. Johnson, V. Daggett & A. R. Fersht. Protein folding and unfolding inmicroseconds to nanoseconds by experiment and simulation. Proc. Natl. Acad. Sci. U. S.A. 97(25), 13518–13522 (2000)

McCarney, E. R., J. E. Kohn & K. W. Plaxco. Is there or isn’t there? The case for (andagainst) residual structure in chemically denatured proteins. Crit. Rev. Biochem. Mol.Biol. 40(4), 181–189 (2005)

McMinn, P. C. The molecular basis of virulence of the encephalitogenic flaviviruses. J. Gen.Virol. 78, 2711–2722 (1997)

Modis, Y., S. Ogata, D. Clements & S. C. Harrison. A ligand-binding pocket in the denguevirus envelope glycoprotein. Proc. Natl. Acad. Sci. U. S. A. 100(12), 6986–6991 (2003)

Modis, Y., S. Ogata, D. Clements & S. C. Harrison. Structure of the dengue virus envelopeprotein after membrane fusion. Nature 427(6972), 313–319 (2004)

Modis, Y., S. Ogata, D. Clements & S. C. Harrison. Variable surface epitopes in the crystalstructure of dengue virus type 3 envelope glycoprotein. J. Virol. 79(2), 1223–1231 (2005)

Mueller, D. S., T. Kampmann, R. Yennamalli, P. R. Young, B. Kobe & A. E. Mark. Histidineprotonation and the activation of viral fusion proteins. Biochem. Soc. T. 36(Part 1), 43–45(2008)

Murata, M., S. Kagiwada, S. Takahashi & S. Ohnishi. Membrane fusion induced by mutualinteraction of the two charge-reversed amphiphilic peptides at neutral pH. J. Biol. Chem.266(22), 14353–14358 (1991)

Nakajima, S., K. Nakajima, E. Nobusawa, J. Zhao, S. Tanaka & K. Fukuzawa. Comparisonof epitope structures of H3HAs through protein modeling of influenza A virus hemag-glutinin: Mechanism for selection of antigenic variants in the presence of a monoclonalantibody. Microbiol. Immunol. 51(12), 1179–1187 (2007)

Nelson, D. L. & M. M. Cox, eds. Lehninger Principles of Biochemistry. W. H. Freeman, 3rded. (2000). URL http://www.biochem.arizona.edu/ classes/bioc462/462a/-NOTES/Amino_Acids/amino_acids.htm

Newton, I. Philosophiæ Naturalis Principia Mathematica (1678)

Noguchi, H. & M. Takasu. Fusion pathways of vesicles: A Brownian dynamics simulation.J. Chem. Phys. 115(20), 9547–9551 (2001)

Nordlund, H. R., V. P. Hytönen, O. H. Laitinen, S. T. H. Uotila, E. A. Niskanen, J. Savolainen,E. Porkka & M. S. Kulomaa. Introduction of histidine residues into avidin subunit inter-faces allows pH-dependent regulation of quaternary structure and biotin binding. FEBSLett. 555(3), 449–454 (2003)

Nozaki, Y. & C. Tanford. Solubility of amino acids and related compounds in aqueous ureasolutions. J. Biol. Chem. 238(12), 4074–4081 (1963)

133

Oostenbrink, C., A. Villa, A. E. Mark & W. F. Van Gunsteren. A biomolecular force fieldbased on the free enthalpy of hydration and solvation: The GROMOS force-field parame-ter sets 53A5 and 53A6. J. Comp. Chem. 25(13), 1656–1676 (2004)

Oxford English Dictionary, Second Edition. entry on “linch-pin”. (1989). URLhttp://dictionary.oed.com/cgi/entry/50133548

Paroutis, P., Touret & S. Grinstein. The pH of the secretory pathway: Measurement, deter-minants, and regulation. Physiology 19(4), 207–215 (2004)

Pécheur, E.-I., J. Sainte-Marie, A. Bienvenüe & D. Hoekstra. Lipid headgroup spacing andpeptide penetration, but not peptide oligomerization, modulate peptide-induced fusion.Biochemistry 38(1), 364–373 (1999)

Peter, B. J., H. M. Kent, I. G. Mills, Y. Vallis, P. J. G. Butler, P. R. Evans & H. T. McMa-hon. BAR domains as sensors of membrane curvature: The amphiphysin BAR structure.Science 303(5657), 495–499 (2004)

Phipps, L. P., S. C. Essen & I. H. Brown. Genetic subtyping of influenza A viruses usingRT-PCR with a single set of primers based on conserved sequences within the HA2 codingregion. J. Virol. Methods 122(1), 119–122 (2004)

Poumbourios, P., R. J. Center, K. A. Wilson, B. E. Kemp & B. Kobe. Evolutionary conser-vation of the membrane fusion machine. IUBMB Life 48(2), 151–156 (1999)

Prabhakaran, M., S. H. Gursahani, C. S. Verma, R. Garduno-Juarez & V. Renugopalakr-ishnan. Cytochrome c: the effect of temperature and pressure from molecular dynamicssimulations. J. Phys. Chem. Solids 65(8-9), 1615–1622 (2004)

Pryor, M. J., R. C. Gualano, B. Lin, A. D. Davidson & P. J. Wright. Growth restriction ofdengue virus type 2 by site-specific mutagenesis of virus-encoded glycoproteins. J. Gen.Virol. 79(11), 2631–2639 (1998)

PVS. PVS: Protein Variability Server (2009). URL http://imed.med.ucm. es/PVS/.Modification date 12 October 2009

Ragothaman, Y. M. & B. Kobe. personal communication (2006–7). Cen-tre for Computational Biology and Bioinformatics, School of Informa-tion Technology, Jawaharlal Nehru University, New Delhi - 110067, India.email 〈[email protected]〉

Randolph, V. B. & V. Stollar. Low pH-induced cell fusion in flavivirus-infected Aedes Al-bopictus cell cultures. J. Gen. Virol. 71(8), 1845–1850 (1990)

Rappe, A. K. & W. A. Goddard III. Charge equilibration for molecular dynamics simulations.J. Phys. Chem. 95(8), 3358–3363 (1991)

Rey, F. A., F. X. Heinz, C. Mandl, C. Kunz & S. C. Harrison. The envelope glycoproteinfrom tick-borne encephalitis virus at 2 Å resolution. Nature 375(6529), 291–290 (1995)

134

Roccatano, D., I. Daidone, M.-A. Ceruso, C. Bossa & A. Di Nola. Selective excitation ofnative fluctuations during thermal unfolding simulations: Horse heart cytochrome c as acase study. Biophys. J. 84(3), 1876–1883 (2003)

Roche, S., S. Bressanelli, F. A. Rey & Y. Gaudin. Crystal structure of the low-pH form ofthe vesicular stomatitis virus glycoprotein G. Science 313(5784), 187–191 (2006)

Rosell, F. I., J. C. Ferrer & G. A. Mauk. Proton-linked protein conformational switching:Definition of the alkaline conformational transition of yeast iso-1-ferricytochrome c. J.Amer. Chem. Soc. 120(44), 11234–11245 (1998)

Roussel, A., J. Lescar, M. C. Vaney, G. Wengler, G. Wengler & F. A. Rey. Structure and in-teractions at the viral surface of the envelope protein E1 of Semliki Forest virus. Structure14(1), 75–86 (2006)

Russell, B. S., R. Melenkivitz & K. L. Bren. NMR investigation of ferricytochrome c un-folding: Detection of an equilibrium unfolding intermediate and residual structure in thedenatured state. Proc. Natl. Acad. Sci. U. S. A. 97(15), 8312–8317 (2000)

Russell, R. J., S. J. Gamblin, L. F. Haire, D. J. Stevens, B. Xiao, Y. Ha & J. J. Skehel. H1and H7 influenza haemagglutinin structures extend a structural classification of haemag-glutinin subtypes. Virology 325(2), 287–296 (2004)

Saarikangas, J., H. Zhao, A. Pykäläinen, P. Laurinmäki, P. K. Mattila, P. K. J. Kinnunen, S. J.Butcher & P. Lappalainen. Molecular mechanisms of membrane deformation by I-BARdomain proteins. Curr. Biol. 19(2), 95–107 (2009)

Sauter, N. K., J. E. Hanson, G. D. Glick, J. H. Brown, R. L. Crowther, S. J. Park, J. J. Skehel& D. C. Wiley. Binding of influenza virus hemagglutinin to analogs of its cell-surfacereceptor, sialic acid: Analysis by proton nuclear magnetic resonance spectroscopy andX-ray crystallography. Biochemistry 31(40), 9609–9621 (1992)

Schibli, D. J. & W. Weissenhorn. Class I and class II viral fusion protein structures revealsimilar principles in membrane fusion (Review). Mol. Membr. Biol. 21(6), 361–371 (2004)

Schiffer, C. A., V. Dötsch, K. Wüthrich & W. F. van Gunsteren. Exploring the role of thesolvent in the denaturation of a protein: A molecular dynamics study of the DNA bindingdomain of the 434 repressor. Biochemistry 34(46), 15057–15067 (1995)

Schindler, T. & F. X. Schmid. Thermodynamic properties of an extremely rapid proteinfolding reaction. Biochemistry 35(51), 16833–16842 (1996)

Schneider, T. D. & R. M. Stephens. Sequence logos: a new way to display consensus se-quences. Nucl. Acids Res. 18(20), 6097–6100 (1990)

Schneider, T. D., G. D. Stormo, L. Gold & A. Ehrenfeucht. Information content of bindingsites on nucleotide sequences. J. Mol. Biol. 188(3), 415–431 (1986)

Serjeant, E. P. & B. Dempsey. Ionisation constants of organic acids in aqueous solution,vol. 23 of IUPAC chemical data series. Pergamon press: Oxford (1979)

135

Shannon, C. E. A mathematical theory of communication. Bell Systems Tech. J. 27(3),379–423 (1948)

Shortle, D. & M. S. Ackerman. Persistence of native-like topology in a denatured protein in8 M urea. Science 293(5529), 487–489 (2001)

Skehel, J. J., P. M. Bayley, E. B. Brown, S. R. Martin, M. D. Waterfield, J. M. White, I. A.Wilson & D. C. Wiley. Changes in the conformation of influenza virus hemagglutinin atthe pH optimum of virus-mediated membrane fusion. Proc. Natl. Acad. Sci. U. S. A. 79(4),968–972 (1982)

Skehel, J. J. & D. C. Wiley. Receptor binding and membrane fusion in virus entry: Theinfluenza hemagglutinin. Annu. Rev. Biochem. 69, 531–569 (2000)

Smith, L. J., H. J. C. Berendsen & W. F. van Gunsteren. Computer simulation of urea-watermixtures: A test of force field parameters for use in biomolecular simulation. J. Phys.Chem. B 108(3), 1065–1071 (2004a)

Smith, L. J., R. M. Jones & W. F. van Gunsteren. Characterization of the denaturation ofhuman α-lactalbumin in urea by molecular dynamics simulations. Proteins 58(2), 439–449 (2004b)

Smith, P. E. Cosolvent interactions with biomolecules: Relating computer simulation datato experimental thermodynamic data. J. Phys. Chem. B 108(48), 18716–18724 (2004)

Söllner, T. H. Intracellular and viral membrane fusion: A uniting mechanism. Curr. Opin.Cell Biol. 16(4), 429–435 (2004)

Srinivasan, J., T. E. Cheatham III, P. Cieplak, P. A. Kollman & D. A. Case. Continuumsolvent studies of the stability of DNA, RNA, and phosphoramidate-DNA helices. J. Am.Chem. Soc. 120(37), 9401–9409 (1998)

Stadler, K., S. L. Allison, J. Schalich & F. X. Heinz. Proteolytic activation of tick-borneencephalitis virus by furin. J. Virol. 71(11), 8475–8481 (1997)

Stevens, J., A. L. Corper, C. F. Basler, J. K. Taubenberger, P. Palese & I. A. Wilson. Structureof the uncleaved human H1 hemagglutinin from the extinct 1918 influenza virus. Science303(5665), 1866–1870 (2004)

Stiasny, K., S. L. Allison, A. Marchler-Bauer, C. Kunz & F. X. Heinz. Structural require-ments for low-pH-induced rearrangements in the envelope glycoprotein of tick-borne en-cephalitis virus. J. Virol. 70(11), 8142–8147 (1996)

Stiasny, K., S. L. Allison, J. Schalich & F. X. Heinz. Membrane interactions of the tick-borneencephalitis virus fusion protein E at low pH. J. Virol. 76(8), 3784–3790 (2002)

Stiasny, K. & F. X. Heinz. Effect of membrane curvature-modifying lipids on membranefusion by tick-borne encephalitis virus. J. Virol. 78(16), 8536–8542 (2004)

Stiasny, K. & F. X. Heinz. Flavivirus membrane fusion. J. Gen. Virol. 87(10), 2755–2766(2006)

136

Stiasny, K., C. Koessl & F. X. Heinz. Involvement of lipids in different steps of the flavivirusfusion mechanism. J. Virol. 77(14), 7856–7862 (2003)

Stiasny, K., C. Kössl & F. X. Heinz. Differences in the postfusion conformations of full-length and truncated class II fusion protein E of tick-borne encephalitis virus. J. Virol.79(10), 6511–6515 (2005)

Stiasny, K., C. Kössl, J. Lepault, F. A. Rey & F. X. Heinz. Characterization of a structuralintermediate of flavivirus membrane fusion. PLoS Pathog. 3(2), 191–199 (2007)

Strauss, J. H. & E. G. Strauss. Virus evolution: How does an enveloped virus make a regularstructure? Cell 105(1), 5–8 (2001)

Stumpe, M. C. & H. Grubmüller. Aqueous urea solutions: Structure, energetics, and ureaaggregation. J. Phys. Chem. B 111(22), 6220–6228 (2007a)

Stumpe, M. C. & H. Grubmüller. Interaction of urea with amino acids: Implications forurea-induced protein denaturation. J. Am. Chem. Soc. 129(51), 16126–16131 (2007b)

Stumpe, M. C. & H. Grubmüller. Polar or apolar — the role of polarity for urea-inducedprotein denaturation. PLoS Comput Biol 4(11), 1–10 (2008)

Tanford, C. Protein Denaturation Part C. Theoretical Models for The Mechanism of Denat-uration, vol. 24 of Advances in Protein Chemistry, pp. 1–95 (1970)

Theorell, H. & A. Åkesson. Studies on cytochrome c. III. Titration curves. J. Am. Chem.Soc. 63(7), 1818–1820 (1941)

Timasheff, S. N. The control of protein stability and association by weak interactions withwater: How do solvents affect these processes? Annu. Rev. Biophys. Biomol. Struct. 22(1),67–97 (1993)

Tirado-Rives, J., M. Orozco & W. L. Jorgensen. Molecular dynamics simulations of theunfolding of barnase in water and 8 M aqueous urea. Biochemistry 36(24), 7313–7329(1997)

Tironi, I. G., R. Sperb, P. E. Smith & W. F. van Gunsteren. A generalized reaction fieldmethod for molecular dynamics simulations. J. Chem. Phys. 102(13), 5451–5459 (1995)

Tobi, D., R. Elber & D. Thirumalai. The dominant interaction between peptide and ureais electrostatic in nature: A molecular dynamics simulation study. Biopolymers 68(3),359–369 (2003)

Tran, H. T. & R. V. Pappu. Toward an accurate theoretical framework for describing en-sembles for proteins under strongly denaturing conditions. Biophys. J. 91(5), 1868–1886(2006)

Trzesniak, D., N. F. A. van der Vegt & W. F. van Gunsteren. Computer simulation studieson the solvation of aliphatic hydrocarbons in 6.9 M aqueous urea solution. Phys. Chem.Chem. Phys. 6(4), 697–702 (2004)

137

Ubbink, M., J. Van Beeumen & G. W. Canters. Cytochrome c550 from Thiobacillus versutus:Cloning, expression in Escherichia coli, and purification of the heterologous holoprotein.J. Bacteriol. 174(11), 3707–3714 (1992)

Ubbink, M., G. C. M. Warmerdam, A. P. Campos, M. Teixeira & G. W. Canters. The role oflysine 99 of Thiobacillus versutus cytochrome c-550 in the alkaline transition. FEBS Lett.351(1), 100–104 (1994)

van der Spoel, D., E. Lindahl, B. Hess, G. Groenhof, A. E. Mark & H. J. C. Berendsen.GROMACS: Fast, flexible, and free. J. Comput. Chem. 26(16), 1701–1718 (2005)

van Gunsteren, W. F., S. R. Billeter, A. A. Eising, P. H. Hünenberger, P. Krüger, A. E. Mark,W. R. Scott & I. G. Tironi. Biomolecular Simulation: The GROMOS96 Manual and UserGuide. Hochschulverlag AG an der ETH Zürich, Zürich (1996)

Verlet, L. Computer “experiments” on classical fluids. I. Thermodynamical properties ofLennard-Jones molecules. Phys. Rev. 159(1), 98–103 (1967)

Víglaský, V., M. Antalík, J. Bagel’ová, Z. Tomori & D. Podhradský. Heat-induced confor-mational transition of cytochrome c observed by temperature gradient gel electrophoresisat acidic pH. Electrophoresis 21(5), 850–858 (2000)

von Neumann, J. Physical applications of the ergodic hypothesis. Proc. Natl. Acad. Sci. U.S. A. 18(1), 263–266 (1932)

Wain, R., T. A. Pertinhez, E. J. Tomlinson, L. Hong, C. M. Dobson, S. J. Ferguson & L. J.Smith. The cytochrome c fold can be attained from a compact apo state by occupancy ofa nascent heme binding site. J. Biol. Chem. 276(49), 45813–45817 (2001)

Wallqvist, A., D. G. Covell & D. Thirumalai. Hydrophobic interactions in aqueous ureasolutions with implications for the mechanism of protein denaturation. J. Am. Chem. Soc.120(2), 427–428 (1998)

Wang, Y., K. J. Addess, J. Chen, Y. Geer, Lewis, J. He, S. He, S. Lu, T. Madej, A. Marchler-Bauer, A. Thiessen, Paul, N. Zhang & H. Bryant, Stephen. MMDB: annotating proteinsequences with Entrez’s 3D-structure database. Nucleic Acids Res. 35(Suppl. 1), D298–D300 (2007)

Warwicker, J. A theoretical study of the acidification of the rhinovirus capsid. FEBS Lett.257(2), 403–407 (1989)

Warwicker, J. Model for the differential stabilities of rhinovirus and poliovirus to mild acidicpH, based on electrostatics calculations. J. Mol. Biol. 223(1), 247–257 (1992)

Weast, R. C., M. J. Astle & W. H. Beyer, eds. CRC Handbook of Chemistry and Physics,chap. Concentrative properties of aqueous solutions: Conversion tables, pp. D–266. CRCPress, Inc., Boca Raton, Florida, 67th ed. (1986)

Weerasinghe, S. & P. E. Smith. Cavity formation and preferential interactions in urea solu-tions: Dependence on urea aggregation. J. Chem. Phys. 118(13), 5901–5910 (2003)

138

Weis, W. I., A. T. Brünger, J. J. Skehel & D. C. Wiley. Refinement of the influenza virushemagglutinin by simulated annealing. J. Mol. Biol. 212(4), 737–761 (1990)

Wengler, G. & G. Wengler. Cell-associated West Nile flavivirus is covered with E+pre-Mprotein heterodimers which are destroyed and reorganized by proteolytic cleavage duringvirus release. J. Virol. 63(6), 2521–2526 (1989)

White, J. M., S. E. Delos, M. Brecher & K. Schornberg. Structures and mechanisms of viralmembrane fusion proteins: Multiple variations on a common theme. Crit. Rev. Biochem.Mol. Biol. 43(3), 189–219 (2008)

WHO. WHO | Dengue. (2008). URL http://www.who.int/topics/dengue/ en. Mod-ification date 23 May 2008

Wilson, I. A., J. J. Skehel & D. C. Wiley. Structure of the haemagglutinin membrane glyco-protein of Influenza virus at 3 Å resolution. Nature 289(5796), 366–373 (1981)

Winkler, J. R. Cytochrome c folding dynamics. Curr. Opin. Chem. Biol. 8(2), 169–174(2004)

Worrall, J. A. R., R. E. M. Diederix, M. Prudêncio, C. E. Lowe, S. Ciofi-Baffoni, M. Ubbink& G. W. Canters. The effects of ligand exchange and mobility on the peroxidase activityof a bacterial cytochrome c upon unfolding. Chem. Bio. Chem. 6(4), 747–758 (2005a)

Worrall, J. A. R., A.-M. M. van Roon, M. Ubbink & G. W. Canters. The effect of replacingthe axial methionine ligand with a lysine residue in cytochrome c-550 from Paracoccusversutus assessed by X-ray crystallography and unfolding. FEBS J. 272(10), 2441–2455(2005b)

Yu, I.-M., W. Zhang, H. A. Holdaway, L. Li, V. A. Kostyuchenko, P. R. Chipman, R. J. Kuhn,M. G. Rossmann & J. Chen. Structure of the immature dengue virus at low pH primesproteolytic maturation. Science 319(5871), 1834–1837 (2008)

Zhang, W., P. R. Chipman, J. Corver, P. R. Johnson, Y. Zhang, S. Mukhopadhyay, T. S.Baker, J. H. Strauss, M. G. Rossmann & R. J. Kuhn. Visualization of membrane proteindomains by cryo-electron microscopy of dengue virus. Nat. Struct. Biol. 10(11), 907–912(2003)

Zhang, W., S. Mukhopadhyay, S. V. Pletnev, T. S. Baker, R. J. Kuhn & M. G. Rossmann.Placement of the structural proteins in Sindbis virus. J. Virol. 76(22), 11645–11658 (2002)

Zhang, Y., W. Zhang, S. Ogata, D. Clements, J. H. Strauss, T. S. Baker, R. J. Kuhn & M. G.Rossmann. Conformational changes of the flavivirus E glycoprotein. Structure 12(9),1607–1618 (2004)

Zimmerberg, J. & L. V. Chernomordik. Membrane fusion. Adv. Drug Deliv. Rev. 38(3),197–205 (1999)

Zimmerberg, J., S. S. Vogel & L. V. Chernomordik. Mechanisms of membrane-fusion. Annu.Rev. Biophys. Biomolec. Struct. 22, 433–466 (1993)

139

Outlook

For me the work on this thesis has certainly raised further questions. I hope to be able topursue research on the topic of viral fusion, not only because it is such an interesting topic.I want to understand how viruses infect humans and animals, and most importantly how wecan prevent viral infection. This may enable us to prevent illness and save lives. Beyondthe need to survive and to live and travel in risk areas for viral diseases, the understandingof viral infection mechanisms may also lead to new technologies. We might for exampledevelop drug delivery systems that reach targets inside patient cells in similar ways as viruseswhen they infect the host cell.

I intend to run more simulations of the dengue viral E protein to test the new hypothesesput forward in this thesis. The plan is to add one or two membranes, i.e. lipid bilayers, to thesimulation system and to simulate the interactions between the protein and the membranes.The effects that the proteins might have on the membranes may offer new insights into howthe E protein mediates membrane fusion. In fact, for many viral proteins the mechanisms bywhich they mediate fusion are not known and difficult to investigate experimentally, mainlybecause membranes are difficult to study. Membrane structure and dynamics are good exam-ples of topics that can be studied more easily with computer simulations than in real exper-iments. What we learn from simulations can in turn be used to generate new hypotheses forexperiments, in other words, simulations can give clues on what to look for in experiments.In the case of the dengue viral E protein, the model for fusion proposed in this thesis couldbe tested by attaching the two subunits of the protein with a flexible linker that allows someconformational change, but not the separation into individual subunits. If the protein medi-ates fusion in the dimeric form, i.e. consisting of two subunits, then fusion will occur despitethe linker, but no trimers will form, as this would require three separate subunits.

Further questions related to viral fusion are: Given the mixture of lipids in viral mem-branes, how are the different lipid species distributed, homogeneously or do they form clus-ters? How do the transmembrane helices of the envelope proteins affect the viral membrane?

141

So far I have simulated the ectodomain of the E protein, which is truncated from the mem-brane domain. The membrane domain is made up of transmembrane helices that run throughthe viral membrane and anchor the ectodomain to the membrane. The M protein is another,smaller envelope protein. It also has a membrane domain that contains transmembrane he-lices. I suspect that the transmembrane helices of both the E and the M protein organisethe structure of the viral membrane, and I would like to investigate this in simulations ofthe membrane domains with a membrane. The membrane domains might even affect theectodomain of the E protein, which was proposed in the hypothesis on the flaviviral fusionmechanism in the final chapter of this thesis. This could be tested by simulating a systemthat combines all the components of the viral envelope: the E and the M protein and the viralmembrane.

Eventually any simulation result needs to be verified experimentally. New techniqueslike fluorescent imaging are being developed for membrane experiments, that may offer in-sights into the effects of protein/membrane interactions on membrane structure and dynam-ics.

142

SummaryIn my dissertation I present two studies in computational chemistry. The first examines thepartial unfolding of an enzyme in urea. The enzyme cytochrome c reacts with peroxides, butnormally this “peroxidase” property is weak. However, the solvation in urea leads to a 1000-fold increase of the peroxidase activity. Urea loosens the structure of the protein and thusmakes the active site inside the protein more accessible for the peroxides, leading to a higherreaction rate. My simulations of cytochrome c in urea showed the partial opening of thereactive site. At the same time the overall structure of the protein remained intact, allowingit to return to its original, less active state when the concentration of urea is lowered. Thismay enable the recycling, stable storage and multiple activation of cytochrome c peroxidasefor industrial application.

The second study investigates the structural changes of a viral protein under acidicconditions. Many viruses are enveloped hollow particles covered with proteins. At acidicpH these proteins transform and enable the virus to merge with the host cell and infectthe host. Histidine is a small component of proteins that is sensitive to the acidic pH atwhich the infection takes place. My simulations of a viral envelope protein showed that thelocal changes of the histidines at acidic pH are capable of triggering irreversible and globalrearrangements in the protein. These may be important for the merger of the virus with thehost cell and therefore for infection.

143

Samenvattingvan de proefschrift De effecten van ureum en van pH op de structuur van eiwitten on-derzocht met moleculaire dynamica simulatie

In mijn proefschrift presenteer ik twee studies in de computationele chemie. De eerste gaatover het partieel ontvouwen van een enzym in ureum. Het enzym cytochroom c reageert metperoxiden, maar normaal is deze “peroxidase”-eigenschap zwak. Oplossing in ureum leidtechter tot een 1000-vouwdige verhoging van de peroxidaseactiviteit. Het ureum maakt destructuur van het eiwit losser en zo het actieve centrum binnen het eiwit toegankelijker voorde peroxiden. Dit leidt tot een hogere reactiesnelheid. Mijn simulaties van cytochroom c inureum toonden het gedeeltelijk openen van het actieve centrum. Hierbij bleef de algemenestructuur van het eiwit intact, zo dat de structuur in zijn originele, minder actieve staat kanterugkeren als de concentratie van de ureum verminderd wordt. Dit maakt recycling, stabielbewaren en meervoudig activering van cytochrome c peroxidase voor industriële toepassingmogelijk.

De tweede studie betreft de structurele veranderingen van een viraal eiwit onder zureomstandigheden. Veel virussen zijn omhulde holle deeltjes, die met eiwitten bedekt zijn. Bijzure pH veranderen deze eiwitten en laten het virus met de gastheercel fuseren en de gastheerinfecteren. Histidine is een klein bestanddeel van eiwitten, dat gevoelig is voor de zure pHwaarbij infectie plaatsvindt. Mijn simulaties van een viraal enveloppe-eiwit lieten zien datde lokale veranderingen van histidines bij zure pH onomkeerbare en globale herschikkingenin het eiwit teweegbrengen. Deze kunnen voor de fusie van de virus met de gastheercel endaarom voor de infectie belangrijk zijn.

144

Zusammenfassungder Dissertation Die Wirkung von Harnstoff und von pH auf die Struktur von Proteinenuntersucht mit Molekulardynamik-Simulation

In meiner Dissertation präsentiere ich zwei Studien aus der Computerchemie. Die erste Stu-die behandelt die teilweise Entfaltung eines Enzyms in Harnstoff. Das Enzym Cytochrom creagiert mit Peroxiden, normalerweise ist diese “Peroxidase”-Eigenschaft jedoch sehrschwach. Die Lösung in Harnstoff wiederum steigert die Peroxidase-Aktivität um ein 1000-faches. Denn Harnstoff lockert die Struktur des Proteins und macht dadurch das aktive Zen-trum für die Peroxide zugänglicher, was zu einer Steigerung der Reaktionsrate führt. Inmeinen Simulationen des Cytochrom c in Harnstoff konnte ich die teilweise Öffnung des ak-tiven Zentrums beobachten. Dabei blieb die Struktur des Proteins insgesamt intakt, so dassdas Protein bei Verringerung der Harnstoffkonzentration wieder in seinen ursprünglichen,weniger aktiven Zustand zurückkehren kann. Dies ermöglicht das Recycling, die stabileLagerung und die mehrfache Aktivierung der Cytochrom c-Peroxidase für die industrielleAnwendung.

Die zweite Studie untersucht die Strukturänderungen eines Virusproteins bei niedrigempH, also unter sauren Bedingungen. Viele Viren stellen hohle Teilchen dar, deren Hüllemit Proteinen bedeckt ist. Diese Proteine verändern sich bei niedrigem pH und bewirkendie Vereinigung des Virus mit der Wirtszelle und damit die Infektion des Wirts. Histidinist ein kleiner Bestandteil von Proteinen, der empfindlich auf pH-Änderungen reagiert wiesie bei der Infektion stattfinden. In meinen Simulationen des Virushüllproteins kann mansehen, dass schon lokale Veränderungen der Histidine bei niedrigem pH irreversible undglobale Veränderungen des Proteins auslösen. Diese könnten für die Fusion des Virus mitder Wirtszelle von Bedeutung sein und damit wichtig für die Infektion.

145

Acknowledgments

I would like to thank everyone who has supported me during my PhD study and in writingthis thesis:

the Groningen Biomolecular Sciences and Biotechnology Institute (GBB) of the Universityof Groningen for giving me the opportunity to conduct this PhD research; the Faculty ofBiological and Chemical Sciences of the University of Queensland for sponsoring my visitsto Brisbane;

my colleagues at the University of Groningen: Herman Berendsen, MonicaBulacu-Cioceanu, Burçin, Klaas Dijkstra, Santi Esteban-Martín, Hao Fan, Marc Fuhrmans,Nicu Goga, Gerrit Groenhof, Marlon Hinner, Djurre de Jong, Volker Knecht, HariLeontiadou, César Lopez Bautista, Martti Louhivuori, Siewert-Jan Marrink, Pieter van derMeulen, Meike Müller-Trimbusch, Frans Mulder, Alia Oktaviani, Jolanda Oldengarm,Renee Otten, Xavier Periole, Gilles Pieffet, Aldo Rampioni, Hilda Riemens, JelgerRisselada, Andrzej Rzepiela, Lars Schäfer, Durba Sengupta, Magda Siwko, Gabi Solomon,Kamil Tamiola, Alex de Vries, Tsjerk Wassenaar, Katy Wood, Ying Xue, Serge Yefimov ofthe MD group and the NMR group; Ria Broer, Bauke Dijkstra, Ria Duurkens, BertPoolman, Liesbeth Veenhoff;

my colleagues at the University of Queensland: Henning Avenhaus, Matt Breeze, HuijunChen, Itamar Kass, Alpesh Malde, Laura Marshall, Kim Nguyen, In-Keun Oh, ChantelPotter, Maria Ratajczak, Sophie Turner of the MD group; Thomas Huber, ChristopheSchmitz, Elizabeth Skippington, Mitchell Stanton-Cook of the Biomolecular Modellinggroup; Marlies Hankel, Seth Olsen, Pierre Tran at the Centre for Computational MolecularScience; Thorsten Kampmann, Bostjan Kobe, Ragothaman Yennamalli, Paul Young at theSchool of Chemistry and Molecular Biosciences; Doune Macdonald, Alastair McEwan;

147

my PhD advisor Alan Mark, Catherine Mark;

Jürgen Schlitter, Ruhr-University Bochum; Ragav Kannan, Martin Zacharias, TechnischeUniversität München; Jolande van Gunsteren, Wilfred van Gunsteren, Chris Oostenbrink;Madeleine Kittner, MPI of Colloids and Interfaces, Potsdam; Martin Stumpe, formerly MPIfor Biophysical Chemistry, Göttingen; Lars Heinke, Gleb Solomentsev; BiochemicalSociety, CECAM, WISENET Women in Science Enquiry Network;

my brother Sebastian Müller;

Adele Donnermeyer, Amélie Richeux, Axel Pawellek (1972–2007), Daniela Süß, EvaNagel, Mark “Jetten” Donnermeyer, Jörn Güldenhaupt, Lena Haug, Maike Boldt-Schäfer,Manfred Thon, Mario Franz, Meik Kösters, Nagat Karroum, Nadja Schmidt, PaulSchneider, Regine Scheder, Stefan Schäfer, Stefanie Stueber, Susanne Dicke, Tanja Kösters,Tom Thelen;

Alfredo Nantes, Ian Leiper, Julie Scopelitis, Megan Seydel, Alice Broos, Amy Cupitt, AmyMiao, Claire Doyle, Clelia Murabito, Dan McHugh, Darren McDonnell, Gayle Davies,Huyen Nguyen, Ileah Saad, Jon Davies, Marie Guinoiseau, Nicholas John, Noel Firkin,Sirisopha Vongthevan, Sivan Macover;

Anette Riebel, Christin, Christine Vogel, David Brehme, Dineke, Dirk Best, DuskoJovanovic, Esther Van Straten, Luuk, Gijs Bekenkamp, Giuseppe Cimo, Giuseppe Papari,Golnar Karimian, Herfita Agustiandari, Jan de Groot, Jelena Jevtic, Jelly Kuiters, JohnMcKean, Marta Palomo Reixach, Martin Smith, Maxi Meissner, Mehreen Mahmud,Monika Hanisch, Parisa Noorishad, Pourya Khosropanah, Sadia Bari, Sandra Klompmaker,Stephen Bourke, Tamara, Vaiva Petrikaite, Valeriu Tudose; Wietske, Douwe, Jelina &Martijn;

sweet and wise Merel, my furry muse, who appeared miraculously and looked after mewhile I wrote this thesis;

and any friends that I may have forgotten to mention;

148

Arachis hypogaea, BBC 6 Music, Couchsurfing, Lindenstraße, Mentha x piperita,Morrissey, New Order, Northern Exposure, the sea, the sun;

and whoever I forgot to mention whom I would like to thank.

next page: Merel, May 2010

149

CV

Daniela was born in Lünen, a city in the “Ruhrgebiet”, which is an area in Germany namedafter the river “Ruhr” and famous for its history in coal-mining. Shortly after she moved withher parents to a small village in Hesse, Germany, where her grandmother lived and agricul-ture and rural landscapes dominate the picture. The Müllers continued to move and landedin Hong Kong, then a colony under the British crown. There Daniela and her brother went tothe German Swiss International School and were exposed to a mix of Chinese, German andvarious other expat cultures. Hong Kong became the home place Daniela feels most attachedto.

Daniela was musically educated and learned the violin since the age of four. However,at the age of 13 she decided that she had enough of the violin and the piano and becameobsessed with ballet. In her childhood she devoured books and wanted to become a doctor,first a general medical practitioner then a surgeon, a pilot, a horse breeder and vet, an experi-mental scientist and finally a professional dancer. After seven years in the far east the familymoved back to Germany, returning to the Ruhr-area. There Daniela was deeply impressed bythe industrial culture, the heritage from the old days of coal-mining, and the unique characterof the people: a blend of working class mentality, influences of various immigrant nationali-ties and urban life. It was the heyday of grunge, which became her youth subculture. Danielagraduated from high school without a clue of what to study, as her interests were varied, andmany of her teachers recommended her to pursue their respective subject.

Luckily, the German system of higher education at that time offered endless opportuni-ties to try almost any subject — and Daniela used that opportunity to find the ideal (combina-tion of) subjects. She enrolled for Landscape Ecology, Media and Communication studies,Film and Television studies, English, Philosophy, Art History, Geography and finally Biol-ogy. Although she felt that Physics and for some time even Mathematics would have beenthe better choice, she stuck to Biology and finished her degree. Today she is glad that shechose this beautiful subject. Since then Daniela has been focusing on structural biology,using computers of different sizes. She loves plants, nature, art, culture, outdoor activities,diving, food, cooking and film. She was recently adopted by a cat. There is no telling inwhich country she will eventually settle down. Daniela wants to continue being a scientistand learn more about nature, biology and life, which is how this book came about.

151

the effects of urea and of ph on protein structure · the effects of urea and of ph on protein...

Documents