Detergents As Membrane-Mimetic Media for Structural Characterization of Membrane Proteins

Detergents as membrane-mimetic media for structural characterization of membrane proteins.

David Vincent Tulumello

A thesis submitted in conformity with the requirements for the degree of Doctor of Philosophy Biochemistry University of Toronto

Detergents as membrane-mimetic media for structural characterization of membrane proteins.

David Vincent Tulumello

Doctor of Philosophy

Department of Biochemistry University of Toronto

2012

Abstract

Membrane proteins are essential cellular components, responsible for a wide variety of biological functions. In order to better understand such aspects of cell activity, researchers have pursued detailed structural analysis of this class of proteins. Because of the complexities in isolating and studying membrane proteins in their native environment, detergents are often employed as a membrane mimetic media. This thesis examines several features of transmembrane (TM) protein structure and folding in detergents through which we are able to gain insights into membrane protein folding, as well as explore the suitability of detergents as membrane-mimetic environments. We first compare the helix-helix association of a series of model TM sequences in a native bilayer to the corresponding association in a detergent environment. We find that while various classes of helix-helix interaction motifs are preserved in detergents, alterations in detergent solvation may, in turn, lead to altered association affinity. We further explore this phenomenon through investigation of the consequences of the insertion of a strongly polar residue into a TM segment. In these studies we find a correlation between sequence-dependent ii alterations in detergent solvation and predicted in vivo folding. We also extend such analyses to a variety of detergents and native TM segments, finding that native secondary structure, as it occurs in the context of a full-length protein, is generally well preserved in a variety of detergents. Finally, we assess the determinants of membrane protein folding using two- transmembrane segment constructs, in the process optimizing expression, production and characterization techniques for a diverse range of transmembrane protein sequences. Overall this thesis finds that, detergents are capable of solubilizing membrane proteins in a form suitable for in-depth structural characterization that may not be feasible in other environments. Thus, as an approximation of a native membrane, detergents are able to preserve certain features of membrane proteins such as helix-helix association and native secondary structure.

iii

Acknowledgments

It is my pleasure to thank all those who have made completing this thesis possible.

I am extremely grateful to my supervisor Dr. Charles Deber. He has been a great mentor and teacher, and has always displayed the utmost devotion to my professional growth and academic training. It has been a privilege to model myself as a scientist based on his admirable example.

Working and learning alongside all of the current and former members of the Deber lab has been a pleasure. They have always been willing and generous in sharing both their time and expertise. Through their kindness and friendship, my colleagues have made spending time in the lab an enjoyable process.

I am also thankful to the many scientists who have provided guidance throughout my development as a researcher. These include members of my supervisory committee, Dr. Alan Davidson and Dr. Igor Stagljar, collaborators from the University of Toronto Mississauga Campus, especially Dr. Scott Prosser, and former supervisors from my undergraduate studies at McMaster University, Dr. Adam Hitchcock and Dr. Graham McGibbon.

I would like to express deep gratitude to my parents, Mary and Peter. They were the first to encourage my love of science by fostering my curiosity at a young age. Instead of simply answering my many questions about how the world works, they chose to show me. In many ways, I consider them to be my very first scientific advisors.

Finally, I have been truly blessed for having the loving support of my wife, Amanda. She has continuously provided me with the strength and motivation to persevere through the hard times, and has helped me pause to celebrate the successes. She has been by my side during every phase of my graduate career, and without her, this thesis would not have been possible.

Table of Contents

Acknowledgments ...... iv

Table of Contents ...... v

List of Tables ...... x

List of Figures ...... xi

List of Appendices ...... xiii

List of Abbreviations ...... xiv

Chapter 1. Introduction...... 1

1.1 Membrane proteins...... 2

1.2 Transmembrane protein structure...... 3

1.2.1 Structure and function of β-barrel membrane proteins...... 4

1.2.2 Structural and function of α-helical membrane proteins...... 7

1.2.3 Properties of α-helical transmembrane segments...... 9

1.3 Transmembrane protein folding...... 10

1.3.1 Bilayer insertion...... 10

1.3.2 Helix – helix association...... 11

1.3.3 Helix – lipid interactions...... 13

1.3.4 Additional features of transmembrane domain assembly...... 14

1.4 Interaction of transmembrane proteins with detergents...... 15

1.4.1 Detergent properties...... 16

1.4.2 Solubilization versus stabilization: the role of detergents in membrane protein characterization...... 17

1.4.3 Overview of commonly used detergents for membrane protein characterization...... 18

1.5 Characterization of transmembrane proteins in detergents...... 22

1.5.1 General considerations...... 22

1.5.2 Detergent micelle insertion and secondary structure...... 24

1.5.3 Transmembrane protein tertiary and quaternary structure analysis...... 26

1.5.4 High resolution structure determination...... 29

1.6 Thesis hypothesis and outline...... 31

Chapter 2. SDS micelles as a membrane-mimetic environment for transmembrane segments………...... 33

2.1 Introduction...... 34

2.2 Materials and methods...... 35

2.2.1 Transmembrane segment prediction...... 35

2.2.2 Peptide synthesis and purification...... 35

2.2.3 TOXCAT assays...... 36

2.2.4 Sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) analysis...... 36

2.2.5 Circular dichroism spectroscopy...... 37

2.2.6 Tryptophan fluorescence measurements...... 37

2.2.7 Förster resonance energy transfer (FRET)...... 38

2.3 Results...... 38

2.3.1 Peptide design and TM insertion prediction...... 38

2.3.2 Oligomerization within bacterial membranes...... 40

2.3.3 Effect of polar residue substitutions on peptide oligomeric state(s) in SDS...... 41

2.3.4 Peptide characterization by circular dichroism and fluorescence spectroscopy. .. 45

2.4 Discussion...... 48

2.4.1 Structure of peptide-detergent complexes...... 49

2.4.2 The effectiveness of SDS as a ‘membrane-mimetic’...... 51

Chapter 3. Positions of polar amino acids alter interactions between transmembrane segments and detergents...... 53

3.1 Introduction...... 54

3.2 Materials and methods...... 55

3.2.1 Assessment of hydrophobicity and hydrophobic moment ...... 55

3.2.2 Peptide synthesis and purification...... 55

3.2.3 Circular dichroism spectroscopy...... 56

3.2.4 Tryptophan fluorescence measurements...... 56

3.2.5 Sodium dodecyl sulfate-polyacrylamide gel electrophoresis analysis...... 56

3.3 Results...... 57

3.3.1 Peptide design and TM insertion prediction...... 57

3.3.2 SDS-PAGE migration rates of Asn containing peptides...... 58

3.3.3 Peptide helicity in aqueous vs. detergent environments...... 60

3.3.4 Detergent binding as a dominant factor in SDS-PAGE migration...... 62

3.3.5 Tryptophan fluorescence reveals variations in SDS distribution...... 63

3.4 Discussion...... 66

3.4.1 Alteration of detergent solvation of hydrophobic peptides induced by polar residues...... 66

3.4.2 Features of SDS solvation that mimic membrane protein folding...... 69

Chapter 4. Transmembrane segment structures are similar in native proteins and detergents………...... 71

4.1 Introduction...... 72

4.2 Materials and methods...... 72

4.2.1 Transmembrane protein structural analysis...... 72

4.2.2 Peptide synthesis and purification...... 73

4.2.3 Biophysical characterization...... 74

4.2.4 Statistical analysis...... 74

4.3 Results...... 75

4.3.1 TM structure classification...... 75

4.3.2 Adoption of native secondary structure...... 77 vii

4.3.3 Interaction of TM segments with detergent micelles monitored by Trp fluorescence...... 81

4.4 Discussion...... 82

4.4.1 Transmembrane segment structure in intact proteins vs. TM peptides...... 82

4.4.2 Are detergents denaturants of membrane proteins? ...... 83

Chapter 5. De novo designed hydrophobic “hairpins” as models of membrane protein folding…………...... 85

5.1 Introduction...... 86

5.2 Materials and methods...... 87

5.2.1 Library creation...... 87

5.2.2 Colony expression immunoscreening...... 88

5.2.3 Protein expression...... 89

5.2.4 Protein extraction and purification under denaturing conditions...... 89

5.2.5 Biophysical characterization...... 90

5.2.6 SDS-PAGE migration rate ...... 90

5.3 Results...... 90

5.3.1 De novo membrane protein hairpin library design and production...... 90

5.3.2 Protein expression, and purification...... 94

5.3.3 Adoption of α-helical structure...... 96

5.3.4 Fluorescence spectroscopy indicates a degree of amino acid residue burial in both water and SDS...... 98

5.3.5 Gel migration rate variance reveals conformational differences among hairpins...... 99

5.4 Discussion...... 101

5.4.1 Methods to improve de novo membrane protein design...... 101

5.4.2 Structural features characteristic of dual transmembrane hairpin constructs. .... 102

Chapter 6. Discussion...... 105

6.1 Summary of contributions...... 107 viii

6.1.1 SDS micelles as a membrane-mimetic environment for transmembrane segments...... 107

6.1.2 Positions of polar amino acids alter interactions between transmembrane segments and detergents...... 107

6.1.3 Transmembrane segment structures are similar in native proteins and detergents...... 108

6.1.4 De novo designed hydrophobic “hairpins” as models of membrane protein folding...... 109

6.2 Detergents as membrane mimetics...... 109

6.2.1 Comparison of folding in detergents versus bilayers...... 110

6.2.2 Denaturation of membrane proteins by detergents...... 114

6.3 Future directions of investigation of membrane protein folding...... 115

6.3.1 Folding of designed membrane proteins...... 115

6.3.2 Methods to assess structural details of protein detergent complexes...... 117

6.4 Conclusions...... 122

Chapter 7. Literature Cited ...... 123

Appendices ...... 139

List of Tables

Table 1.1. Properties of commonly used detergents...... 19

Table 2.1. TOXCAT measurement of association of TM segments in the E. coli inner membrane...... 41

Table 4.1. Transmembrane segment analysis and peptide sequences...... 77

Table 5.1. Sequences of de novo designed hairpins...... 91

Table 5.2. Sequences obtained from library for further studies...... 94

List of Figures

Figure 1.1. Schematic diagram of a biological membrane...... 3

Figure 1.2. Representative structures of a β-barrel and an α-helical transmembrane protein ...... 4

Figure 1.3. Examples of β-barrel membrane protein structures...... 6

Figure 1.4. Examples of α-helical membrane protein structures ...... 8

Figure 1.5. Interactions stabilizing helix-helix contacts in membranes ...... 12

Figure 1.6. Helix and lipid conformations in membrane bilayers ...... 14

Figure 1.7. Principles of detergent solubilization of membrane proteins ...... 18

Figure 1.8. Common detergents used in membrane protein characterization...... 19

Figure 1.9. Detergents used in high resolution structure determination ...... 30

Figure 2.1. Design of model TM segments...... 39

Figure 2.2. Oligomeric state(s) of model TM segment peptides...... 42

Figure 2.3. FRET experiments on TM peptides...... 44

Figure 2.4. Circular dichroism of model TM segment peptides...... 45

Figure 2.5. Tryptophan fluorescence of TM peptides...... 47

Figure 2.6. AI5 peptides containing a polar substitution ...... 48

Figure 2.7. Peptide interactions within detergent micelles...... 51

Figure 3.1. Peptide sequence and SDS-PAGE analysis of Asn peptide variants...... 59

Figure 3.2. Circular dichroism of peptides...... 61

Figure 3.3. Sequence and SDS-PAGE analysis of additional peptide variants...... 63

Figure 3.4. Tryptophan blue shifts induced by SDS solvation, and “hydrophobic moments” of peptides...... 65

Figure 3.5. Schematic representation of detergent-peptide complexes...... 68

Figure 4.1. Classification of local residue environment in TM transporter proteins...... 76

Figure 4.2. Circular dichroism spectra of native and designed transmembrane peptides...... 78

Figure 4.3. Correlation of secondary structure of transmembrane segment peptides in detergents vs. their secondary structure in the intact protein...... 80

Figure 4.4. Fluorescence spectroscopy of native and designed transmembrane peptides...... 82

Figure 5.1. Construct for bacterial expression of model TM hairpins as fusion proteins...... 92

Figure 5.2. Purification scheme for randomized hairpin proteins...... 95

Figure 5.3. Circular dichroism of hairpins ...... 97

Figure 5.4. Fluorescence emission spectra of hairpins ...... 99

Figure 5.5. SDS-PAGE migration of hairpins...... 100

Figure 6.1. Schematic views of the transmembrane regions of selected α-helical membrane proteins...... 113

Figure 6.2. Diffusion coefficient (D) profiles of SDS and model TM segments over the course of a detergent titration...... 119

Figure 6.3. SDS saturation curves of model TM segments...... 120

xii

List of Appendices

Appendix 1: Supporting information for Chapter 2...... 139

Appendix 2: Supporting information for Chapter 3...... 142

Appendix 3: Supporting information for Chapter 4...... 144

Appendix 4: Table of oligonucleotides used in DNA library construction...... 148

Appendix 5: Detergent binding analysis of PFG-NMR diffusion studies...... 149

xiii

List of Abbreviations

ACN – acetonitrile

ANS – 1-anilino-8-naphthalene sulfonate

CAT – chloramphenicol acetyltransferase

CD – circular dichroism spectroscopy

CMC – critical micelle concentration

CHAPS – 3-[(3-cholamidopropyl)-dimethylammonio]-1-propanesulfonate

DDM – n-dodecyl-β-D-maltoside

DM – n-decyl-β-D-maltoside

DPC – n-dodecyl phosphocholine

FRET – Förster resonance energy transfer

FTIR – Fourier transform infrared spectroscopy

GpA – glycophorin A

GPCR – G protein-coupled receptor

HPLC – high performance liquid chromatography

LB – Luria broth

LDAO – lauryl dimethylamine oxide

MRE – mean residue molar ellipticity

NG – n-nonyl-β-D-gluocoside

NMR – nuclear magnetic resonance

OG – n-octyl-β-D-gluocoside

OMPG – outer membrane protein G

PAGE – polyacrylamide gel electrophoresis xiv

PDB – Protein Data Bank

PFG – pulsed field gradient

SDS – sodium dodecyl sulfate

SE-AUC – sedimentation equilibrium analytical ultracentrifugation

SEC – size exclusion chromatography

SPFO – sodium perfluorooctanoate

TDI – thiol-disulfide interchange

TB – terrific broth

TM – transmembrane vSGLT – Vibrio parahaemolyticus sodium galactose transporter

xv 1

Chapter 1. Introduction.

Contents of this chapter have been published, in part, by Tulumello D.V., and Deber C.M. (2011) Membrane protein folding in detergents in Methods in Protein Biochemistry, Harald Tschesche (Ed.), pp 23-46, Berlin; New York, NY : De Gruyter.

1.1 Membrane proteins.

Living cells are separated from the external environment by lipid membranes. These biological membranes act as efficient barriers selectively preventing - or enabling - the passage of a variety of molecules into and out of cells as well as intracellular organelles found within eukaryotic cells. This allows for the compartmentalization of biological processes - a fundamental feature of living organisms. Situated within biological membranes are proteins that to a large degree regulate both molecular and information flow across these barriers. As such, membrane proteins assume vital roles in diverse cellular processes, such as ion and solute transport, signal transduction, cellular metabolism, and membrane structural integrity.

There are two general classes of membrane associated proteins: extrinsic or peripheral membrane proteins that are attached to the surface of the membrane through interactions with the lipid head groups or other membrane bound proteins; and intrinsic, or integral membrane proteins that interact with the hydrophobic core of the membrane and span the bilayer. The latter are therefore most commonly referred to as transmembrane (TM) proteins (Fig. 1.1). TM proteins are further divided into β-barrel or α-helical membrane proteins, based upon the major secondary structural element that comprises the membrane spanning region. α-helical TM proteins, which are the most abundant and ubiquitous class of membrane proteins, are estimated to comprise nearly 30% of the human proteome (Almen et al., 2009). They have been implicated in a wide variety of diseases and disorders such as cancer, stroke, cystic fibrous, antibiotic resistance and Parkinson’s, and are the targets of over 50% of FDA-approved drugs (Overington et al., 2006).

Despite such great importance, it has been the water-soluble, globular proteins that have dominated our understanding of both protein structure and function. This discrepancy in knowledge is readily apparent in the Protein Data Bank (PDB), where currently only ~2% of the over 70,000 deposited structures are TM proteins. This lag in our understanding is largely due to the hydrophobic nature of these molecules which makes their production and purification technically challenging, as well as dictating that they must be studied in a lipid or detergent environment. The inclusion of such “membrane mimetic” environments introduces further hurdles for many biophysical characterization techniques, and in some cases may actually prevent such experiments from being feasible (particularly for systems containing lipid vesicles).

Figure 1.1. Schematic diagram of a biological membrane. The membrane consists primarily of a hydrophobic lipid bilayer (yellow), as well as membrane associated proteins. Integral membrane proteins are depicted in blue, and peripheral membrane proteins are depicted in red.

1.2 Transmembrane protein structure.

Although there remain many challenges in high resolution structure determination of membrane proteins, significant progress has been made in the past several years and the number of structures obtained has been increasing exponentially (White, 2009). From these structures, many common structural features have been determined. First, protein strands that traverse the membrane consist primarily of α-helical or β-strand secondary structure (Fig. 1.2). The formation of these secondary structures is favourable in the low dielectric environment of the membrane, as the hydrogen bonding potential of the peptide backbone is able to be fully satisfied. The formation of such H-bonds prevents the exposure of this otherwise polar functional group to the apolar lipid environment (White and Wimley, 1999). Furthermore, formation of either secondary structure results in the projection of amino acid side chains outwards where they may interact favorably with the lipid acyl chains.

Figure 1.2. Representative structures of a β-barrel and an α-helical transmembrane protein. A) Outer membrane protein G (OMPG), a beta barrel membrane protein (PDB ID: 2IWV). The membrane spanning structure is composed primarily of β-strands in which hydrogen bonds form between the amide proton of a peptide bond of one strand with the carbonyl oxygen of another, thereby satisfying the hydrogen bonding requirements of the peptide backbone. The barrel structure shown here is formed by 14 β-strands which produce a single continuous β-sheet structure that is closed by hydrogen bonding between the first and last strand. B) Rhodopsin, an α-helical membrane protein (PDB ID: 1GZM). The membrane spanning structure is composed primarily of α- helices in which hydrogen bonding occurs between the amide proton of a peptide bond and the carbonyl oxygen of the amino acid four residues earlier in the sequence. The membrane spanning structure shown here is composed of a bundle of 7 α- helices.

1.2.1 Structure and function of β-barrel membrane proteins.

β-barrel membrane proteins are present in outer membranes of Gram-negative bacteria, as well as mitochondria and chloroplasts of eukaryotic cells. They have a variety of roles such as facilitating passive diffusion, small molecule transport, enzymatic catalysis, or as toxins (Wimley, 2003; Tamm et al., 2004). There structures are typically formed by a large β-sheet composed of 8 to 22 strands (each consisting of 8-11 amino acid residues) that create a closed structure where the first and last strand are hydrogen bonded to one another. Individual strands are often connected by alternating short, tight turns (predominantly on the periplasmic side), and

5 longer flexible loops (predominantly on the extracellular side) (Wimley, 2003; Tamm et al., 2004). Alternatively, a single barrel structure may be comprised of multiple protein chains. For example, alpha-hemolysin is composed of 7 individual monomer units, each of which is a two strand “hairpin” structure (Fig 1.3D) (Song et al., 1996). Individual β-strands typically alternate between polar and hydrophobic residues (Wimley, 2002). This patterning allows for lipid facing surfaces to be composed primarily of hydrophobic residues, while residues facing the interior of the structure are primarily polar.

For many of these proteins, there is a central cavity formed by the inward facing polar residues that is filled with water. This central cavity is often found to facilitate passive diffusion. In some smaller barrel structures, there is no central pore structure formed. In these cases, the interior polar residues interact to form hydrogen bonding networks or participate in other electrostatic interactions (Tamm et al., 2004). The overall domain organization and quaternary structure of β- barrel proteins varies with examples of proteins containing additional structural features and various oligomeric states (Fig. 1.3). β-barrel protein cellular targeting, folding and insertion into bilayers require a series of chaperones and assembly machinery unique to their final cellular location much of which differs from their α-helical counterparts (Walther et al., 2009).

Figure 1.3. Examples of β-barrel membrane protein structures. Individual protein strands are coloured differently in each structure. A) Lipid A acylase PagP, a β-barrel enzyme with an additional interfacial alpha helix (PDB ID: 3GP6). B) Outer membrane protein G (OMPG), a monomeric porin (PDB ID: 2IWV). C) OprP phosphate-specific transporter, a trimeric β-barrel membrane protein where each monomer unit forms one channel (PDB ID: 2O4V). D) Alpha- hemolysin, a homo-oliogomeric bacterial toxin consisting of seven “hairpin” monomer units (PDB ID: 7AHL).

1.2.2 Structural and function of α-helical membrane proteins.

α-helical membrane proteins are found in the inner membrane of bacteria as well as eukaryotic cell and organelle membranes (Almen et al., 2009). As with β-barrel membrane proteins, they have a variety of functions and may act as cellular receptors, enzymes, and cell adhesion components. Many of these proteins allow for molecular transport across the membrane, which is achieved through either passive or active diffusion. Passive diffusion occurs through a pore or substrate-specific channel while active transport occurs against a concentration gradient and may be driven by ATP hydrolysis, substrate oxidation, or energetically favourable co-transport. The inclusion of membrane proteins in such critical cellular processes allow for a high degree of regulation. In many cases, external stimuli results in a conformational change of the membrane protein in turn leading to an alteration in cellular activity. This modification of cellular behavior may range from the import of a single, specific solute to the activation of a complex signal transduction cascade resulting in major metabolic changes such as those which occur following the activation of some G protein-coupled receptors (GPCRs).

Structurally, this class of proteins is typified by the presence of TM domains that are composed primarily of α-helices connected by extra-membranous loops. These domains vary from a single pass TM helix to higher-order multi-pass oligomeric proteins (Fig. 1.4). While in some cases there are only short loops connecting helices, for other membrane proteins extra-membranous segments form distinctly folded, water soluble domains. The oligomeric organization of this protein class includes both homo-oligomers and hetero-oligomers where some domains may themselves contain no TM segments.

Figure 1.4. Examples of α-helical membrane protein structures. Individual protein strands are coloured differently in each structure. A) Human integrin αIIbβ3, a transmembrane heterodimer (PDB ID: 2KNC). B) Phospholamban, a homopentamer found in the endoplasmic reticulum (PDB ID: 1ZLL). C) Rhodopsin, a GPCR family member (PDB ID: 1GZM). D) Vibrio parahaemolyticus sodium galactose transporter, a symporter where the channel is formed completely by one monomer (PDB ID: 3DH4). E) CorA Mg2+ transporter, which forms a pentameric channel (PDB ID: 2IUB) F) BtuCD Vitamin B12 transporter, an ABC transporter consisting of two transmembrane and two water soluble subunits (PDB ID: 1L7V).

1.2.3 Properties of α-helical transmembrane segments.

TM segments are typically comprised of 20-30 amino acids, a length ideally suited to span the width of the hydrophobic interior of a bilayer (25-30 Å). These segments have been shown to be composed primarily of hydrophobic residues, a necessity for insertion into the membrane bilayer. As such Phe, Leu, Ile, Val, Met, and Ala, account for 55% of all amino acids in TM segments (Liu et al., 2002). Less hydrophobic amino acids Ser, Thr and Gly comprise a further 22% of amino acids found in such sequences and are commonly implicated in TM helix-helix association (see Section 1.3.2 below). Within TM domains it has also been noted that conserved positions have a different average composition than non-conserved ones. In particular, Gly, Pro and Tyr are significantly more prevalent at conserved versus non-conserved positions, implying an important role for these residues (Liu et al., 2002). Although energetically unfavorable for insertion, strongly polar residues (Asp, Glu, Lys, Arg, His, Asn, Gln, and Pro) account for 9% of residues found in TM segments (Illergård et al., 2011). These residues also tend to be highly conserved, and are often of functional importance (Illergård et al., 2011).While the side chain of Pro does not, itself, contain any polar atoms, Pro typically behaves in manner similar to a strongly polar residue within the context of a TM segment due to its propensity to disrupt helix formation, thereby exposing polar atoms of the peptide backbone that would otherwise be involved in H-bonding.

It has also been shown that residue distribution along the length of TM helices is not uniform (Senes et al., 2000; Liu et al., 2002; Ulmschneider et al., 2005). In particular, both aromatic residues (Phe, Tyr and Trp) as well as positively-charged residues (Arg and Lys) are most commonly found at the border between the hydrophobic core and the interfacial region in bilayers. These observations correlate well to molecular dynamics simulations (Johansson and Lindahl, 2006; MacCallum et al., 2008) as well as in vivo partitioning studies (Hessa et al., 2007) that have demonstrated these residues to be energetically most favorable (or least unfavourable for charged residues) for insertion when positioned in this region. This energetic preference may serve to anchor a TM helix at a specific position with respect to the bilayer.

1.3 Transmembrane protein folding.

The lipid bilayer presents a unique, partially anisotropic environment that imposes constraints on protein folding not present in aqueous environments. Originally proposed in the 1990s by Popot and Engleman, the two-stage model of α-helical membrane protein folding provides a useful conceptual framework to consider the structure and assembly of these proteins (Popot and Engelman, 1990; Engelman et al., 2003). In this model, individual TM helices first insert into the interior of the lipid bilayer as independently stable, folded domains. Following insertion, individual TM segments may then associate laterally with one another to produce the final tertiary structure. Ultimately both these steps are governed by the identities of the amino acid side chains comprising the protein sequence. Although complex folding of TM domains does not necessarily occur in two distinct steps in all cases, a reduction of the membrane protein “folding problem” to this model has allowed for the relationship of protein primary sequence to folded structure to be investigated (Rath et al., 2009b; Bordag and Keller, 2010).

1.3.1 Bilayer insertion.

In order to insert into a lipid bilayer as a TM α-helix, a stretch of amino acids must be both of sufficient length and hydrophobicity. These insertion requirements have been assessed using both in vitro and in vivo models. It has been found that above a certain “threshold” value of hydrophobicity, a segment of ca. 20 residues will spontaneously insert into bilayers, adopting an α-helical structure (Liu et al., 1996; Hessa et al., 2005; White and von Heijne, 2008). In vivo bilayer insertion into the inner membrane of bacteria or the endoplasmic reticulum of eukaryotes generally occurs co-translationally, a process mediated by the translocon (White and von Heijne, 2008). The translocon machinery itself behaves as a selection apparatus where sufficiently hydrophobic helices are inserted into the lipid bilayer, while less hydrophobic sequences pass through. This selection process is thought to occur by the opening of a lateral gate into the membrane, allowing for interaction of the nascent TM segment with lipids. From there, partitioning occurs according to thermodynamic principles based largely upon hydrophobicity (Hessa et al., 2007).

In some cases, the boundaries of TM segments that are present in the final structure are not those that are initially inserted via the translocon. In such situations it is hypothesized that a more hydrophobic stretch is initially inserted and following a structural rearrangement, the final TM segment boundaries are adopted (Kauko et al., 2010). A TM helix that is less hydrophobic than would be required for independent translocation may also be inserted if it is connected by a short loop to an adjacent, suitably hydrophobic helix (Hedin et al., 2010).

1.3.2 Helix – helix association.

Following the insertion of α-helices, association may occur between either separate chains forming an oligomeric structure, or within a single multi-spanning membrane protein. In such circumstances, association occurs despite of the presence of surrounding membrane lipids which could stabilize these mostly hydrophobic segments individually. The driving force of helix-helix assembly in bilayers is largely due to several types of interactions as illustrated in Fig. 1.5. For example, van der Waals forces along complementary ‘knobs-into-holes’ type interfaces stabilize close-packing of helices (Fig. 1.5A,B). Side chain-side chain hydrogen bonding between polar residues may also mediate TM-TM association. This may occur through either a single ‘strongly polar’ residue (Asp, Glu, Asn, Gln), which in itself may be sufficient to stabilize and TM helix- helix interface (Gratkowski et al., 2001; Zhou et al., 2001), or more complex inter-helical hydrogen bond networks (Fig. 1.5C) (Dawson et al., 2002; Sal-Man et al., 2005; Call et al., 2006). Finally, Cation-pi and/or pi-pi interactions between aromatic and/or basic residues have also been shown to participate in helix-helix self-association (Fig. 1.5D) (Hong et al., 2007; Johnson et al., 2007; Sal-Man et al., 2007; Bocharov et al., 2008b).

Figure 1.5. Interactions stabilizing helix-helix contacts in membranes. A) Large Val residue ‘knobs’ (dark blue) are in van der Waals contact with the concave helix surface, as illustrated by the Gly residue ‘holes’ (light blue) of a GG4 motif at the GpA dimer interface (MacKenzie et al., 1997). B) Apolar amino acid residues of the phospholamban heptad repeat are in van der Waals contact (red line) across the helix-helix interface (Oxenoid and Chou, 2005). C) Inter-helical hydrogen bond networks between Asp-Asp and Thr-Tyr residue pairs stabilize the interface of the T-cell receptor ζ-chain dimer (Call et al., 2006). D) Edge-face cation-pi interaction between Phe residues in the ErbB2 dimer (Bocharov et al., 2008b). Figure adapted from (Rath et al., 2009b).

TM helix-helix interactions motifs typically involve one or more such interactions and can be roughly grouped based on both geometry and sequence patterning (Rath et al., 2009b). For example, right-handed packing of helix pairs are most often characterized by an i, i+4 separation of ‘small’ residues (Gly, Ala, and Ser) in the primary sequence. This motif is very well characterized in the glycophorin A (GpA) dimer, where this sequence periodicity places the two small residues on the same surface of the helix creating a shallow groove that complements the surface of a second helix (Fig. 1.5A) (MacKenzie et al., 1997). This association is stabilized by

van der Waals interactions resulting from intercalation of the Gly residue ‘holes’ and large residue (e.g. Leu, Ile, Val) ‘knobs’ (Johnson et al., 2006; Cunningham et al., 2011). This motif is alternately termed the small-xxx-small, XX4 (where X=G, A, or S), or GASRight motif and occurs for a wide range of proteins (Walters and DeGrado, 2006). Left-handed helix packing occurs most often as heptad repeats of small (Lear et al., 2004; Poulsen et al., 2009) or large residues such as Leu and Ile, with the latter exemplified by the ‘Leu zipper’ assembly of phospholamban (Fig. 1.5B) (Arkin et al., 1994; Simmerman et al., 1996). As with the small-xxx-small motif, these sequence patterns are largely stabilized by van der Waals contacts. Many of these forces and motifs have also been identified as contributing to tertiary contacts within the domains of multi-spanning membrane proteins as well (Senes et al., 2004; Walters and DeGrado, 2006; Joh et al., 2008).

1.3.3 Helix – lipid interactions.

Within the bilayer, interactions of the lipid chains, either with themselves, or with inserted TM segments, also play a role in defining TM domain structure. Because the length of a TM helix can vary, it may not always match the thickness of the hydrophobic region of the lipid bilayer. In this situation, termed “hydrophobic mismatch”, a structural rearrangement of the lipid, protein or both may occur in order to accommodate acyl chain solvation of the TM segment. Lipid ordering and/or composition surrounding the TM segment may be altered in order to decrease or increase the thickness of the bilayer at this location (Fig. 1.6B, C). If an inserted helix is longer than the thickness of the bilayer (positive mismatch), then it is also possible for the helix to tilt within the bilayer in order to accommodate all non-polar residues (Fig. 1.6D) (Killian and Nyholm, 2006). There is also the possibility of shorter hydrophobic stretches that may be accommodated if the side chain(s) of terminal polar residue(s) are able to project upwards, thereby “snorkeling” into the interfacial region (Fig. 1.6E).

Figure 1.6. Helix and lipid conformations in membrane bilayers. A) Conformation of a hydrophobic TM segment of suitable length to span the hydrophobic core of the bilayer. B) A longer TM segment which is accommodated by acyl chain extension (ordering). C) A shorter TM segment which is accommodated by acyl chain compression (disordering). D) A longer TM segment which is accommodated by helix tilting with respect to the membrane bilayer. E) A TM segment with a terminal polar residue (depicted in red) which is accommodated by side chain snorkeling out of the hydrophobic interior.

Association of helices within the membrane bilayer also necessitates a rearrangement of the packing of lipids that surround the individual helices. This alteration of lipid-lipid and lipid-helix interactions must also be taken into account when fully describing the interaction process in thermodynamic terms (White and Wimley, 1999). Indeed, the strength of association of TM segments has been correlated to the lipid exposed surface area of the individual helices (Johnson et al., 2006; Cunningham et al., 2011).

1.3.4 Additional features of transmembrane domain assembly.

Although a great deal of work has been performed to understand membrane protein folding in terms of packing of helix bundles, as the structural database of membrane proteins expands it is becoming increasingly evident that there are other features within the membrane-spanning region

that contribute to the final tertiary and quaternary structure of these proteins. Binding of prosthetic groups is one such feature, evident from some of the earliest transmembrane protein structures such as bateriorhodopsin (Grigorieff et al., 1996). Large water-filled cavities within the membrane region formed by TM helices are also present in many transporters (Abramson et al., 2003; Stroud et al., 2003). The formation of such structural features requires the sequestering of lipid from an area of the membrane where an internal aqueous cavity can form. This is presumed to occur due to the partially polar nature of the helix backbone, as well as membrane- inserted polar amino acid side chains (Engelman et al., 2003).

Another increasingly observed feature is a stretch of the polypeptide chain that penetrates into the membrane bilayer but is not a membrane spanning α-helix (Viklund et al., 2006; Yan and Luo, 2010). These stretches are commonly referred to as reentrant regions (or loops), and have been identified in roughly 10% of membrane protein structures (primarily transporters and ion channels) (Yan and Luo, 2010). For example, the aquaporin family of water channels has two such reentrant loops that each consist of an α-helix that penetrates roughly halfway into the membrane, following which the peptide chain exits the bilayer from the same side it entered via a coil structure (Murata et al., 2000). These two half helices contribute to form a selectivity filter + that excludes H3O due to the electropositive helix dipoles present in the bilayer core. Besides their function as selectivity filters, reentrant loops have also been identified as playing a role in transport processes by ligand-induced structural changes (Compton et al., 2010).

1.4 Interaction of transmembrane proteins with detergents.

As with soluble proteins, the characterization of membrane protein structure and function is often performed in vitro. In such studies, the choice of a suitable membrane mimetic environment is critical for a successful outcome. To a large degree, detergents have been used to provide a hydrophobic environment similar to a natural membrane, while imparting solubility in mixed water/detergent media to membrane proteins. This solubility, in turn, allows for their characterization by a broad range of techniques. In particular, high resolution structure determination by crystallization and solution NMR techniques, has overwhelmingly relied upon their solubilization in detergent micelles (Sanders and Sonnichsen, 2006; Privé, 2007; Carpenter et al., 2008).

Detergents represent a very broad class of molecules, with a range of biophysical characteristics, and in many cases the choice of the “correct” detergent for studies of membrane proteins is largely empirical. Although it would be extraordinarily beneficial if there were specified detergents that were always useful for particular applications, our understanding of these phenomena has not yet advanced to this level. In some cases the choice of detergent is heavily dependent upon the technique; in other circumstances, several different detergents may give similar results; and in some cases, many detergents must be investigated before a suitable one is found. It is thus helpful to understand the characteristics and limitations of detergents as to how they are applied toward studies of membrane protein structure and folding.

1.4.1 Detergent properties.

Typically a detergent is an amphiphilic molecule that contains both polar substituents (often termed head groups) and non-polar regions (most often an alkyl chain). The presence of two regions with differing polarity leads to detergent self-aggregation in water in an effort to minimize aqueous exposure of the non-polar regions. Detergents differ from most common biological lipids in that instead of the extended bilayer structures found in membranes, they typically form smaller globular assemblies termed micelles. In a micelle structure, a central core is composed primarily of the hydrophobic component of detergent monomers that are surrounded by an outer layer of the primarily hydrophilic chemical groups. Micelle structures can be described, on average, as roughly spherical particles but on an individual basis they typically have highly irregular surfaces (Bruce et al., 2002). Notably, the micellar particle differs from cells (such as erythrocytes) and phospholipid vesicles in that the micelle does not contain an aqueous interior compartment. It is this micellar form of detergent molecules that is most relevant as a membrane mimetic for membrane proteins.

Detergent micelles are most often described by an average aggregation number (or range) as well as a concentration above which this self-association into micelles occurs spontaneously [termed the critical micelle concentration (CMC)]. These two properties are dictated by both the length of the detergent alkyl chain as well as the chemical make-up (both charge and size) of the polar head groups. These features are often used to qualitatively describe how these detergents interact with membrane proteins, where small head group size, the presence of a head group charge,

and/or short alkyl tail length all increase the “harshness” of a detergent with respect to membrane protein denaturation (Privé, 2007). These features are also those that in general make detergents effective solubilizers of both membranes and membrane proteins.

1.4.2 Solubilization versus stabilization: the role of detergents in membrane protein characterization.

Detergents serve a variety of different roles in the study of membrane proteins. Perhaps the most common use of detergents is to extract these proteins from their native membrane or from insoluble aggregates produced by heterologous expression (commonly referred to as inclusion bodies) (Tate, 2001; Cunningham and Deber, 2007). In many cases detergents alone may accomplish this, although mechanical disruption (such as sonication) used in combination can often improve this process. Following this step, the solubilizing detergent may be exchanged for another one, or - from this detergent-solubilized state - membrane proteins may be exchanged into an artificial bilayer system (le Maire et al., 2000; Lin and Guidotti, 2009).

The fundamental role of detergents for structural studies is to confer aqueous solubility to the otherwise insoluble membrane proteins (Fig. 1.7). In principle, the purpose of a detergent is to coat hydrophobic regions of proteins with amphipathic detergent molecules while maintaining the native tertiary and quaternary structure of the membrane protein. However, this very ability of detergents to bind to hydrophobic regions of proteins often renders them denaturing. Thus, in soluble proteins, which are stabilized by packing of hydrophobic regions into a central core, detergents compete successfully through hydrophobic protein-detergent interactions (Imamura, 2006). The membrane-spanning domains of membrane proteins are likewise stabilized in large part by interactions of hydrophobic residues, and as such, detergents are also, to some degree, denaturing to membrane proteins. In fact, detergents have gained widespread use in membrane protein unfolding studies (Booth and Curnow, 2006; Renthal, 2006). It is this balance between solubilization versus native structure stabilization, as well as any specific characteristics required for the experimental technique being used, that is at the heart of the choice of detergent for characterization.

Figure 1.7. Principles of detergent solubilization of membrane proteins. A) Typical soluble proteins fold such that hydrophobic residues form a buried, central core and are surrounded primarily by polar residues (blue) that allow for aqueous solubility. Myoglobin (PDBID: 1MBN) is used here as an example. The central core is shown in yellow and overlaid on the remainder of the structure, shown in blue. B) A membrane protein typically has hydrophobic residues exposed to its environment, which is the hydrophobic lipid bilayer. The β-2-adrenergic receptor (PDBID: 2RH1) is used here as an example. Membrane embedded residues are shown in yellow while residues not found within the bilayer core are shown in blue. C) The functional use of detergents is to confer aqueous solubility to otherwise insoluble membrane proteins. A detergent- solubilized membrane protein will have hydrophobic regions coated with amphiphilic detergent molecules, producing a polar outer surface allowing for solubility in the heterogeneous detergent-water medium.

1.4.3 Overview of commonly used detergents for membrane protein characterization.

The chemical structures of some commonly-used detergents are shown in Figure 1.8, and a summary of their relevant characteristics is given in Table 1.1. The set described here includes acyl chain detergents that cover three broad categories: ionic, zwitterionic, and neutral. The set is composed of the two most widely-used detergents for solution nuclear magnetic resonance (NMR) studies [sodium dodecyl sulfate (SDS) and dodecyl phosphocholine (DPC)]; and the two most often used in X-ray crystallography [n-dodecyl-β-D-maltoside (DDM) and n-octyl-β- D- gluocoside (OG)], according to the Membrane Protein Databank (available online at: http://www.mpdb.tcd.ie) (Raman et al., 2006).

Figure 1.8. Common detergents used in membrane protein characterization. Structures were produced using ChemSketch (ACD Labs).

Table 1.1. Properties of commonly used detergents. Monomer Detergent mass (Da) Aggregation # CMC (mM) sodium dodecyl sulfate (SDS) 288 62-101 1.2-7.1 n-dodecyl phosphocholine (DPC) 352 50-60 1.1 n-dodecyl-β-D-maltoside (DDM) 511 110-140 0.18 n-octyl-β-D-glucoside (OG) 292 ~90 19-25 Data taken from compilation in (le Maire et al., 2000)

1.4.3.1 Anionic: sodium dodecyl sulfate (SDS).

Among the most well-characterized detergents, SDS has a 12-carbon alkyl tail, along with a negatively-charged sulfate head group. Like other charged detergents, its self-association and interaction with proteins is highly dependent upon the ionic strength of the solution (Imamura, 2006). As SDS is often considered among the ‘harshest’ of detergents commonly used, it may seem surprising that it has been used in identifying helix-helix association interfaces (Lemmon et al., 1992a; Arkin et al., 1994; Rath et al., 2006; Ng and Deber, 2010a). SDS has gained such widespread use because it is a highly effective solubilizing agent, and upon closer inspection does not appear to denature the stronger helix-helix interactions. SDS has been shown in some cases to be able to recapitulate many aspects of native tertiary/quaternary folded structure (Almeida and Opella, 1997; Howell et al., 2005; Chill et al., 2006). The anionic nature of SDS also allows for its use in polyacrylamide gel electrophoresis (PAGE) analysis - a common molecular biology technique.

Due to its well-characterized ability to denature soluble proteins (Imamura, 2006), SDS will likely denature any large extramembranous regions of membrane proteins and should not be considered a practical mimetic to preserve the overall structure of any proteins containing such domains. This renders SDS primarily useful for small membrane proteins, or fragments thereof, for which it is highly effective at producing well-solubilized, stable samples.

1.4.3.2 Zwitterionic: n-dodecyl phosphocholine (DPC).

Zwitterionic detergents are generally considered less perturbing to membrane protein structure than ionic ones, while remaining effective membrane solubilizers (le Maire et al., 2000; Privé, 2007). DPC in particular has gained increased use in structural studies due to its phosphocholine head group which itself is one of the most prevalent among phospholipids in native biological membranes. There is also evidence that the behaviour of this head group within detergent micelles is very similar to the water/membrane interface in native lipids (Beswick et al., 1999). As such, DPC is commonly used for NMR where its mono-dispersity, low aggregate number, and small particle size make it an ideal detergent (Page et al., 2006).

1.4.3.3 Neutral: n-dodecyl-β-D-maltoside (DDM).

DDM is a neutral detergent with a sugar head group, and a 12-carbon acyl tail. It is generally considered among the mildest detergents, making it an excellent initial choice for a wide range of applications. It is generally less efficient than other detergents at solubilizing biological membranes; however, membrane proteins are often exchanged into this detergent for characterization studies. DDM has minimal interactions with soluble domains, leading to its common use in the characterization of transport proteins containing such features (Newstead et al., 2008). In particular, DDM is often used in X-ray crystallography, albeit its relatively large alkyl chain length may encompass much of the protein, consequently reducing the protein- protein crystal contacts necessary for this technique (Newstead et al., 2008). Despite its minimally denaturing properties, the large aggregate size and poor mono-dispersity of DDM generally make it unsuitable for NMR studies (Sanders and Sonnichsen, 2006). n-decyl-β-D- maltoside (DM), a detergent with an identical head group as DDM but containing a 10-carbon acyl tail, may also be used in similar applications.

1.4.3.4 Short tail neutral n-octyl-β-D-gluocoside (OG).

Another detergent with a neutral head group, OG, is again considered relatively “mild”. It does, however, have a relatively shorter acyl chain length of eight carbons, which can make it somewhat more denaturing than DDM (Privé, 2007). Its smaller size may leave more surface area exposed for protein-protein contacts important for crystallization, potentially improving resolution of crystal structures (Sonoda et al., 2010). For purposes of NMR studies, OG has been shown to have poor sample stability, and is not in widespread use for this technique (Vinogradova et al., 1998). n-nonyl-β-D-glucoside (NG), a detergent with an identical head group but a nine-carbon acyl tail is also a commonly used detergent, with very similar properties to OG.

1.4.3.5 Other detergents.

Another commonly used class of detergents are the lysophospholipids which are derivatives of native phospholipids, but with only one fatty acid linked to the glycerol backbone rather than two. Lysophospholipids accordingly form micellar structures rather than bilayers. N,N- dimethyldodecylamine-N-oxide (lauryl dimethylamine oxide; LDAO), is a zwitterionic detergent that has been used in a number of crystallization studies of α-helical transmembrane proteins (Andrade et al., 2005; Ye et al., 2010). Polyoxyethylene detergents, a neutral class of detergents, are also commonly used. These detergents are typically composed of an acyl chain tail, and a head group made up of polyethylene glycol, both of which may vary in length. These detergents are named based upon their composition according to the formula CxEy, where x is the number of carbons present in the acyl tail and y is the number of ethoxy groups present in the head group. A commonly used detergent in this class is C8E4. Steroid-based detergents such as 3-[(3- cholamidopropyl)-dimethylammonio]-1-propanesulfonate (CHAPS), have also been used as effective solubilizers of membranes and membrane proteins. As a membrane mimetic environment for structural studies, these latter systems are most effective only in the presence of native lipids or other detergents (Sanders and Sonnichsen, 2006).

Fluorinated surfactants, such as sodium perfluoroctanoate (SPFO) are similar to other detergents described here except that their hydrophobic tails contain (usually exclusively) fluorine atoms in place of hydrogen atoms. For example, the structure of SPFO is CF3(CF2)6-COOH. Detergents in this category have lower affinity for hydrophobic amino acids, and are therefore thought to be less disruptive towards protein-protein interactions, thereby preserving native structure (Melnyk et al., 2002).

1.5 Characterization of transmembrane proteins in detergents. 1.5.1 General considerations.

Characterizing membrane proteins in detergents presents a host of additional considerations not encountered for soluble proteins. For example, one factor that may interfere with the performance of most spectroscopic techniques is light scattering. As the concentration of detergent increases, so does the associated light scattering and although this can be background-

subtracted in most cases, it may increase the noise in data acquisition, thereby decreasing sensitivity. For this reason it is often beneficial to have only a minimal level of detergent present. Yet if too little detergent is present, the membrane protein may not be fully saturated by detergent, potentially leading to insolubility or denaturation. One must therefore ensure that the protein is fully saturated with detergent, albeit the saturating concentration of detergent is not always clear. Membrane proteins have been shown in many cases to bind much more detergent than their soluble counterparts, and as such it is better to assume a large degree of binding (Rath et al., 2009a). It is also necessary to have a detergent concentration above the CMC to ensure the presence of micelles; in many cases, detergent monomers behave much differently than micelles themselves, a factor that can lead to membrane protein aggregation (Otzen et al., 2008; Wahlstrom et al., 2008).

In practice, for studies of membrane protein fragments consisting of 1-2 TM segments at a concentration in the μM regime, one would use an approximate 1000-fold increase in detergent - corresponding to a mM detergent concentration. For micelle aggregates with an association number on the order of 100, this represents ten times as many detergent micelles as transmembrane segments. In situations where a high concentration of peptide is necessary, such as in solution NMR where peptide concentrations are in the mM range, maintaining at least a 1:1 TM segment-to-detergent micelle ratio is recommended (Sanders and Sonnichsen, 2006).

The final peptide-to-detergent ratio is particularly important for assessing oligomerization of transmembrane segment peptides. As detergent concentrations are decreased, so is the total volume of the micellar phase, essentially increasing the apparent concentration of the TM segments in this phase. As such, it may be more reasonable to assess free energy of association of TM segments in terms of the peptide-to-detergent ratio (Fleming, 2002). Too low a concentration of detergent may also lead to some degree of co-localization of individual TM segments in the same micelle without a specific interaction between them, thereby leading to an inaccurate assessment of oligomeric state.

Another important consideration is the equilibration time required following preparation of samples. Detergent solvation generally occurs on a relatively rapid time scale (usually within minutes) from a soluble state; however, from an insoluble state this process may take much longer. The time scale between association and disassociation of TM segments is also relevant.

For some TM segments these events may occur over very long periods (Rath et al., 2009b). For techniques that require equilibration of separate fractions of TM segments, it is recommended that these fractions be mixed prior to inclusion of the final detergent environment where association occurs.

Sample stability is also an issue. For example, both NMR and, to a lesser degree, analytical ultracentrifugation (AUC) (le Maire et al., 2008) require time periods of days or longer for data collection. While lysophospholipid detergents often give high-resolution spectra, no high- resolution structures have been reported in these detergents - likely a consequence of the relatively short-term stability of the relevant samples (Page et al., 2006).

1.5.2 Detergent micelle insertion and secondary structure.

Initial characterization of detergent-solubilized states typically focuses on the adoption of the correct secondary structure, as well as insertion of TM regions into detergent micelles. Helical structure should be observed for TM domains in a native state, and loss of the correct secondary structure may be used as an indication that denaturation is occurring over time.

1.5.2.1 Circular dichroism spectroscopy (CD).

Circular dichroism (CD) is a common technique used to assess the secondary structure of membrane proteins. Helical transmembrane segments display characteristic CD spectra in the far-UV region, with minima in mean residue ellipticity [θ] at 208 and 222 nm. For individual TM segments, adoption of helical structure in the presence of detergents has been used extensively to monitor insertion into a micellar environment (Liu et al., 1996). Measurements of secondary structure are also useful for monitoring the global fold of a membrane protein (Wehbi et al., 2007).

As CD is a spectroscopic technique where minimization of light scattering is essential, the use of lower detergent concentrations is beneficial. Another strategy to improve the sensitivity of the spectral determination is to decrease the path length of the cuvette while increasing the protein concentration, thereby reducing the scattering of the detergent, while maintaining signal from the

protein. We have found that a 1 mm path length cuvette is adequate for most detergents in the mM concentration regime. In some cases, where a high CMC dictates that increased detergent concentrations be used, a 0.1 mM or smaller path length is recommended to minimize light scattering.

1.5.2.2 Fourier transform infrared spectroscopy (FTIR).

Fourier transform infrared spectroscopy (FTIR) specifically probes the frequencies of molecular vibrations. In practice, this technique is in many ways complementary to circular dichroism in that it is often used to identify the secondary structure of membrane proteins, based upon diagnostic vibration peaks of the carbonyl groups in peptide bonds. Classically, transmission FTIR of protein samples is complicated by the strong absorbance of water, which saturates the

absorption spectra. Using D2O reduces this effect; however, a very small transmission path length (>1 mm) as well as relatively high protein concentrations are needed (10 to 50 times greater than that for CD) (Baenziger and daCosta, 2008). The major advantage of FTIR versus CD is that FTIR is readily adaptable to studying membrane proteins in lipid bilayers. This is achieved by using attenuated total reflectance FTIR on a deposited bilayer containing the membrane protein of interest. This experimental setup also eliminates the absorbance of the buffer.

1.5.2.3 Tryptophan fluorescence.

The amino acid tryptophan has a fluorescence emission spectrum that is highly dependent upon the relative polarity of the local environment (Lakowicz, 2006). The emission maximum of tryptophan is typically around 350 nm when its side chain is completely exposed to an aqueous environment, but may be shifted to as low as 310 nm in purely hydrophobic solvents. A central tryptophan residue in a TM segment can thus be used to monitor insertion of this segment into a hydrophobic environment. Typically, an emission spectrum maximum of around 330 nm is indicative of a tryptophan residue in the centre of a detergent micelle. Emission spectra of TM segments in detergents may have maxima as low as 315 nm, possibly due to their self- association.

1.5.3 Transmembrane protein tertiary and quaternary structure analysis.

Although there have been recent breakthroughs in equilibrium folding and unfolding studies of membrane proteins, the field still remains very much in its infancy (Booth and Curnow, 2006; Renthal, 2006). There are outstanding issues that remain to be addressed including an understanding of the nature of unfolded states in these studies There does exist, however, a suite of classical techniques that remain in use to investigate membrane protein assembly. Specifically, much work has been done in addressing the sequence determinants of TM helix association (a Table of known helix-helix association motifs is found in Rath et al., (2009b). To this end, many techniques have been developed to directly monitor the association of TM proteins and their segments, and/or to determine assembly size from which oligomeric state(s) may be inferred. These techniques, as described below, have also been applied in some cases to determine structural characteristics such as compactness and structural homogeneity in the protein-detergent complex of monomeric transmembrane proteins.

1.5.3.1 Polyacrylamide gel electrophoresis analysis.

As mentioned above, sodium dodecyl sulfate-polyacrylamide gel electrophoresis (PAGE) is an extremely common technique in any modern protein laboratory for determination of protein molecular weights and assessment of components in a protein mixture (Reynolds and Tanford, 1970). It has been shown in certain cases that SDS-PAGE may maintain membrane protein native oligomeric states (Wegener and Jones, 1984; Heginbotham et al., 1997; Poulsen et al., 2009). SDS-PAGE is also sensitive to protein conformation (Griffith, 1972; Dunker and Kenyon, 1976), and our lab has previously demonstrated that gel migration rates may be in part be traceable to conformational changes in membrane protein structure (Therien et al., 2001; Wehbi, et al., 2008). PAGE is perhaps the simplest and quickest way to determine membrane protein oligomeric state; however it has the requirement of using an anionic detergent, typically SDS, which may denature all but the strongest TM-TM interactions. There have been some successes in substituting SPFO in place of SDS, which may allow weaker helix-helix interactions to persist (Melnyk et al., 2002). An additional consideration is that there is recent evidence that sequence- specific interactions of membrane proteins with SDS can skew the apparent molecular weight of the peptide: detergent complexes dramatically, making accurate assessment of precise

27 oligomeric state(s) more complicated (Rath et al., 2009a; Walkenhorst et al., 2009; Rath et al., 2010). Nevertheless, since SDS-PAGE does not require any protein labelling, or sophisticated instrumentation and requires limited optimization, it should be considered the initial technique of choice to assess oligomeric state.

1.5.3.2 Size exclusion chromatography (SEC).

Chromatographic techniques have long been a staple of soluble protein purification and characterization. Size exclusion chromatography (SEC) in particular has been shown to be an efficient method for determination of the presence of oligomeric states of proteins. In order to be adapted for membrane proteins, detergent must be present in the elution buffer so as to maintain a constant protein:detergent ratio throughout the course of the experiment. For this reason the chromatography column must remain stable in the presence of detergent, a requirement that has been made feasible by advances in resin chemistry. Calibration of the column is commonly performed using globular, water-soluble protein standards, which are not always practical as detergents may denature these proteins. Size exclusion should generally be treated in a qualitative manner with respect to size or oligomeric state, unless suitable standards are used. This method has also been used to screen membrane protein preparations for samples that give sharp peaks. This may indicate good “mono-dispersity” of the protein-detergent complex, which would thus be more likely to produce crystals (Sonoda et al., 2010). SEC has also been used to determine the amount of detergent bound to membrane proteins, where bound detergent is quantified following co-elution with the protein of interest (le Maire et al., 2008; Rath et al., 2009a).

1.5.3.3 Förster resonance energy transfer (FRET).

Förster resonance energy transfer (FRET) involves the non-radiative transfer of energy from an excited fluorophore (the donor) to an acceptor moiety (Lakowicz, 2006). This phenomenon only occurs when the two moieties are in proximity to one another, usually less than a few nm. The degree of FRET may be monitored either by the increase of intensity of emission of the acceptor fluorophore or a decrease in intensity of the emission of the donor fluorophore. An intrinsic

tryptophan residue may be used as one member of a FRET pair, as its fluorescence emission of around 350 nm can be used to excite various commercial fluorophores (such as dansyl). By titrating in increasing amounts of unlabeled segment, the shape of the curve will inform as to the stoichiometry of the interaction (Adair and Engelman, 1994). One main advantage of FRET is that it may be performed in the native lipid environment, and be used to give estimates of the free energy of association (You et al., 2005).

1.5.3.4 Thiol-disulfide interchange (TDI).

In thiol-disulfide interchange (TDI), the strength of association of transmembrane segments is measured by the efficiency of disulfide crosslink formation between two Cys-containing segments. In principle, disulfide bonds will form more readily if the segments are in an oligomeric form. The crosslinking is then quenched and the ratio of crosslinked species can be determined by SDS-PAGE or high performance liquid chromatography (HPLC) (Cristian et al., 2003a). The advantage of this technique is the ability to measure an association reaction under reversible conditions, and at varying peptide:detergent ratios. These conditions allow for thermodynamic parameters to be obtained. Reversibility of the TM segment crosslinking reaction is achieved by tuning the redox potential of the buffer by altering the ratio of reduced and disulfide-crosslinked glutathione present. Another advantage of this method is that it may also be used in lipid bilayers (Cristian et al., 2003b). As this method reports only on the absence or presence of dimeric crosslinks, the specific oligomeric state(s) is not readily assessed, and other techniques may be required in order to obtain thermodynamic parameters.

1.5.3.5 Sedimentation equilibrium analytical ultracentrifugation (SE- AUC).

Sedimentation equilibrium analytical ultracentrifugation (SE-AUC) provides an additional technique to determine the oligomeric state(s) of membrane proteins. In this assay, the molecular mass of the membrane protein can be directly assessed by monitoring its sedimentation in a centrifugal field. Through a ‘density matching’ technique, the mass contributed by detergent to the membrane protein: detergent complex is ignored (Reynolds and Tanford, 1976), thereby

29 providing an estimate of oligomeric state. If the association of TM segments is reversible, this method may also be used to determine thermodynamic parameters such as the equilibrium association constant (Fleming, 2008). This type of thermodynamic analysis remains subject to the long time periods required to attain equilibration (~ one day), where TM protein samples may not remain stable in certain detergents (le Maire et al., 2008). In contrast to FRET and TDI, SE- AUC is strictly applicable to only detergents, as in lipid vesicles the mass determined will be due to all molecules within a shared vesicle. One advantage of this technique is that it does not require any additional labelling of the membrane protein or segments under investigation. Alternatively, SE-AUC may also be used to provide information about the amount of detergent bound (le Maire et al., 2008).

1.5.4 High resolution structure determination.

Although the previously mentioned techniques have been used to great success in investigating a number of structural features of membrane proteins, a high resolution structure with atomic level details often provides the most complete view of membrane protein structure. Although there are now examples of high-resolution structures of membrane proteins reconstituted into lipid bilayer phases (Caffrey, 2009), or in bicelles (Bocharov et al., 2008a) the majority are obtained from detergent micelles. In X-ray crystallography a great variety of detergents have been used; with no single detergent used for more than 20% of solved structures (Fig. 1.9A). For solution NMR, DPC and SDS have dominated, accounting for an over 60% combined usage rate (Fig. 1.9B).

Figure 1.9. Detergents used in high resolution structure determination. A) Percent occurrence of the five most common detergents used during initial solubilization and / or crystallization of structures deposited in Membrane Protein Databank. B) Percent occurrence of the media (detergent or other) used during NMR data acquisition of structures deposited in the Membrane Protein Databank. Org. sol. indicates that organic solvent was used, DHPC indicates 1,2- diheptanoyl-sn-glycero-3-phosphocholine. Figure produced using statistics taken from the Membrane Protein Databank (accessed online August 21st 2011 at: http://www.mpdb.tcd.ie) (Raman et al., 2006).

1.5.4.1 X-ray crystallography.

In many cases, major screening programs with many different detergents must be used to identify the appropriate detergent and additive combinations that will produce crystals with the necessary diffraction properties to produce a membrane protein structure with the desired resolution (Faham et al., 2008). There have been, however, some efforts to identify which detergents are expected to be most successful. In one report (Sonoda et al., 2010), the authors found that there is likely a trade-off between minimally denaturing detergents, and detergents which minimally coat the protein allowing for maximal crystal contacts between the proteins themselves. The most commonly-used detergents for X-ray crystallography include DDM, OG, C8E4, and LDAO, as well as related variants. Efforts are also being made to develop new classes

of detergents specifically suitable to x-ray crystallography, such as detergents without acyl tails to reduce flexibility (Hovers et al., 2011), or detergents comprised of peptide-lipid combinations (Privé, 2009).

1.5.4.2 Solution nuclear magnetic resonance (NMR).

The large size of lipid vesicles prohibits their use for high-resolution solution NMR, and although there has been use of other media such as organic solvents, amphipathic polymers and bicelles, detergents remain the classical choice of membrane mimetic (Page et al., 2006; Sanders and Sonnichsen, 2006; Raschle et al., 2010). In general, for solution NMR studies, particle size is an important limiting factor for many conventional experiments, as larger particles will tumble slower and produce broader NMR lines. For this reason, detergents considered harsh by other methods (such as SDS or DPC) must be used as they form smaller micelles (Sanders and Sonnichsen, 2006).

1.6 Thesis hypothesis and outline.

Despite these recent advances, our understanding of the sequence-dependent structure and folding of TM proteins remains limited. To a significant degree, detergents have been successfully used to provide a hydrophobic environment similar to the membrane, while imparting aqueous solubility to the membrane proteins, thereby allowing for their characterization by a broad range of techniques. As a model system, detergents are an approximation of a biological bilayer, and as such it is important to understand both their advantages and limitations. Furthermore, while it would clearly be desirable to study membrane proteins as full-length systems, their large size, instability, and often the non-availability of milligram quantities of pure proteins, leads to the requirement of alternate strategies. The two- stage model predicts that, as an individual domain, each TM segment should retain its helical structure when excised from its native context. This model accordingly implies that the folding of fragments of TM proteins should provide information relevant to the full-length protein. Solid phase peptide synthesis has allowed for efficient production of single TM segments, while truncated membrane domains can be produced in adequate levels by heterologous protein

expression. By using a combination of these techniques, and specifically tailoring the constructs for subsequent purification and characterization, many of the hurdles involved in production of larger membrane proteins are avoided (Cunningham and Deber, 2007). Furthermore, a reductionist approach allows us to investigate the nuances of biophysical phenomena associated with minor sequence variation that might otherwise be tempered when occurring in a larger model.

In this thesis we investigate in detail how TM segments interact with detergents, with a focus on how this interaction affects TM segment association, secondary structure, and location within detergent micelle; and how the features observed in detergents compare to structure in a native membrane bilayer. In Chapter 2, the self-association of model transmembrane segments in a bacterial membrane is compared to their association in SDS micelle environments. Here we also focus on how detergent solvation alters this oligomerization event (Tulumello & Deber. Biochemistry. 2009, 48, 12096-103). Chapter 3 focuses on the specific conformation of the detergent/peptide complex using a series of model transmembrane segments, highlighting the structural differences imparted by strongly polar residues (Tulumello & Deber. Biochemistry. 2011, 50, 3928-35). Chapter 4 extends our analysis of the nature of detergent/peptide complexes by comparing how they may be correlated to their conformation in native transmembrane proteins. Here we also compare the effects of “mild” versus “harsh” detergents (Tulumello & Deber, BBA – Biomembranes, 2012, in press). Finally, Chapter 5 will describe the generation of a library of partially randomized transmembrane “hairpin” segments, along with the characterization of their residue-dependent folding in detergents (Tulumello et. al. 2012, manuscript submitted).

Chapter 2. SDS micelles as a membrane-mimetic environment for transmembrane segments.

Contents of this chapter have been published, in part, by Tulumello D.V., and Deber C.M. (2009). SDS micelles as a membrane-mimetic environment for transmembrane segments. Biochemistry. 48(51); 12096-12103.

2.1 Introduction.

Among detergents considered ‘harsh’ with respect to membrane protein denaturation are anionic detergents such as sodium dodecyl sulfate (SDS). SDS is a well-characterized denaturant of soluble protein domains, doing so primarily by binding at hydrophobic sites and inducing a largely helical structure (Imamura, 2006). This property indicates that SDS is not the detergent of choice for membrane protein crystallization - a process that necessarily relies upon retention of protein-protein contacts of soluble domains (Hunte and Michel, 2003; Privé, 2007). Charged detergents (including SDS), however, are often used in NMR studies where features of these micelles such as small size and good monodispersion are advantageous (Page et al., 2006; Sanders and Sonnichsen, 2006).

Although SDS has been used to denature polytopic membrane proteins (Renthal, 2006), it is commonly employed in the identification of native helix-helix interactions in TM segments (Lemmon et al., 1992b; Melnyk et al., 2001; Therien and Deber, 2002a; Sulistijo and MacKenzie, 2006; Gorman et al., 2008), and in some instances is able to maintain native tertiary and quaternary structure of membrane proteins that lack significant extramembranous domains (Arkin et al., 1994; Chill et al., 2006). This situation evokes questions regarding the sequence dependence and biophysical nature of the susceptibility of TM segments to denaturation vs. maintenance of native TM structure upon SDS solubilization, and indeed as to the very concept of ‘denaturation’ of membrane proteins in membrane-mimetic environments such as SDS micelles.

To examine these phenomena in a systematic manner, we investigate here the SDS detergent- solubilized states of a variety of designed hydrophobic α-helical TM segments in conjunction with selected single and multiple polar substitutions within these segments. Our findings show that (i) the local environment of individual helical surfaces within the protein-detergent complex is highly sequence-dependent; and (ii) the structural/oligomeric variations stabilized by SDS solubilization reflect those that may arise in the folded transmembrane domains of proteins.

2.2 Materials and methods. 2.2.1 Transmembrane segment prediction.

Peptide sequences were queried for prediction as TM segments using TM Finder (Deber et al., 2001), TMHMM2 (Krogh et al., 2001), Phobius (Kall et al., 2004), MemBrain (Shen and Chou, 2008), TOPPRED2 (von Heijne, 1992), SOSUI (Hirokawa et al., 1998), and TOPCONS (Bernsel et al., 2008). All programs were run using default settings. Core segment hydrophobicity was also calculated for peptide segments in the absence of the terminal Lys residues based upon the sum of individual Kyte-Doolittle parameters (Kyte and Doolittle, 1982); Liu-Deber parameters (Liu and Deber, 1999); White-Wimley ΔG values of water to octanol transfer (Wimley et al., 1996); and ΔG values of biological insertion (Hessa et al., 2007) (with position dependence).

2.2.2 Peptide synthesis and purification.

Peptides were synthesized using standard Fmoc (N-(9-fluorenyl)methoxycarbonyl) chemistry (Amblard et al., 2005) on a PAL-PEG-PS (4’-aminomethyl-3’,5’-dimethoxyphenoxyvaleric acid- poly(ethylene glycol)) polystyrene resin (Applied Biosystems) that produces an amidated C terminus upon peptide cleavage. Peptides were labeled at the N-termini by incubating the resin- bound peptides with excess label (4-dimethylaminoazobenzene-4'-sulfonyl chloride (dabsyl chloride) or 5-dimethylamino-1-naphthalenesufonyl chloride (dansyl chloride)) under basic conditions overnight. Peptides were cleaved from the resin with 2 hr incubation with 88% trifluoroacetic acid (TFA), 5% phenol, 5% ultrapure water, and 2% tri-isopropyl. Cleaved peptides were precipitated in cold ether overnight, dried, and redissolved in ultrapure water. Peptides were purified by reverse-phase high pressure liquid chromatography on a C4 preparative column (Phenomenex), using an acetonitrile/water gradient (with 0.1% TFA). Cleaved peptides were purified by reverse phase-high pressure liquid chromatography on a C4 preparative column (Phenomenex). Mass spectrometry was used to identify the molecular weight of the purified peptides. All peptides were lyophilized following purification, resuspended in ddH2O and stored in aliquots at –20 °C. Peptide concentrations were determined using amino acid analysis of SDS solubilized samples performed by the Advanced Protein Technology Centre of the Hospital for Sick Children (Toronto, ON).

2.2.3 TOXCAT assays.

The expression vector pccKan, pccGpA-WT, along with Escherichia coli strain NT326, were kindly provided by Dr. Donald Engelman, Yale University (Russ and Engelman, 1999).

Construction of pccAI10-WT (AI10) has been described elsewhere (Johnson et al., 2004).

pccAI5-WT (AI5) was produced using the same methods. Additional sequences were produced

by mutating the AI10 and AI5 construct using the QuikChange site-directed mutagenesis kit (Stratagene). The sequences of all constructs were confirmed using DNA sequencing.

Constructs transformed into NT326 cells were grown, prepared and assayed for both the expression of the total construct and reporter gene; chloramphenicol acetyltransferase (CAT) as previously described (Johnson et al., 2004). All values of CAT concentration were normalized to expression levels of both the construct and to levels of CAT concentration of glycophorin A (GpA). All measurements were performed in at least triplicate and two tailed t-tests, assuming equal variance, were performed using Excel. Constructs were also tested for insertion efficiency based on the ability to grow on M9 minimal medium plates with 0.4% maltose as the only carbon source. To achieve this NT326 cells transformed with each mutant were streaked onto the plates and incubated for 2 days at 37 °C (Russ and Engelman, 1999).

2.2.4 Sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS- PAGE) analysis.

SDS-PAGE was performed using precast 12% acrylamide NuPAGE gels in MES buffer (Invitrogen) according to manufacturer protocols. Coomassie Blue staining was used for visualization of the peptides at higher concentrations, while silver stain (Invitrogen) was used to develop gels containing lower concentrations. Apparent molecular weights were estimated based upon migration rates of Mark12 molecular weight standards (Invitrogen) using NIH Image to measure distances. Relative gel migration rates were calculated as: ([estimated MW from gel / formula MW from sequence] x 100%), and reported as a percentage of the formula MW.

2.2.5 Circular dichroism spectroscopy.

Freshly thawed aliquots of peptides were diluted to a final concentration of between 20 and 40 μM in 10 mM Tris HCl buffer (pH 8.0), with or without 34.7 mM (w/v) SDS. Spectra were recorded in a 1 mm path length cuvette on a Jasco J-810 circular dichroism spectropolarimeter. All spectra were background subtracted and converted to mean residue molar ellipticity (MRE [deg cm2 dmol-1]).

2.2.6 Tryptophan fluorescence measurements.

Freshly thawed aliquots of peptides were diluted to a final concentration of 10 μM in 10 mM Tris HCl buffer (pH 8.0), with or without 8.7 mM (w/v) SDS. Fluorescence emission spectra of peptides were measured on a Hitachi F-400 Photon Technology International C-60 fluorescence spectrometer at an excitation wavelength of 295 nm with a 2 nm slit width, and emission measured between 310 and 400 nm with a 4 nm slit width. All spectra were corrected for light scattering effects by subtraction of background, and by the correction function of FELIX software provided by the manufacturer. Measurements were recorded at a lower concentration of SDS in order to decrease light scattering effects. This strategy improved the signal-to-noise ratio, allowing for more accurate determination of emission wavelength maxima, but had no significant effect on the shape of the emission spectra (see supporting information [Appendix 1]). To confirm that a concentration of 8.7 mM SDS remained in micellar form under the conditions used, the critical micelle concentration (CMC) of the buffer system was determined by 1-anilino- 8-naphthalene sulfonate (ANS) fluorescence as previously described (Esposito et al., 1998). The CMC was found to be 5.86 with a standard error of +/- 0.11 mM, consistent with a previously reported value (obtained in the same buffer) of 5.61 +/- 0.93 mM (Esposito et al., 1998). The addition of peptide (at concentrations used in these experiments) did not appear to significantly affect the CMC (Appendix 1).

2.2.7 Förster resonance energy transfer (FRET).

FRET experiments were performed in 10 mM Tris HCl buffer (pH 8.0), with or without 34.7 mM (w/v) SDS. All samples were prepared from frozen aliquots of peptides in triplicate or quadruplicate. Separately labelled peptides were mixed from aqueous solutions prior to addition of detergent, and then allowed to equilibrate at room temperature overnight. Dansyl emission spectra were recorded between 450-650 nm upon excitation at 341 nm. All spectra were background subtracted; total fluorescence was integrated, and reported as a value normalized to the mean emission of samples containing only dansyl labelled peptide.

2.3 Results. 2.3.1 Peptide design and TM insertion prediction.

In order to systematically investigate detergent solubilization of protein TM regions, we designed two categories of TM peptide segments consisting largely of Ala and Ile residues (sequences shown in Fig. 2.1A). Both categories contain an identical surface capable, in principle, of forming helix-helix interactions in SDS through ‘knobs-into-holes’ packing as well as via small-xxx-small (AxxxA) motifs, as elucidated in previous studies (Johnson et al., 2004).

The two groups of peptides differ primarily in that the more hydrophobic segment (AI10)

contains a total of ten Ile residues, while the less hydrophobic segment (AI5) contains Ala residues instead of Ile residues at five sites that are positioned oppositely on a helical wheel to the ostensible interaction surface (Fig. 2.1B). These changes, in turn, produce differences in net hydrophobicity and amphipathicity between the two prototypic segments. Variants of these two parent peptides were also produced in which a single Asn residue was introduced at position 17. This strategy produces an additional potential mode of interaction (side chain- side chain H- bonding) in the Asn containing variants.

Figure 2.1. Design of model TM segments. A) Sequences of peptides synthesized. AI5 is numbered such that the first residue of the TM sequence is the same as that in AI10 despite differing numbers of N-terminal Lys residues. B) Helical wheel diagrams of the two parent peptides, AI10 and AI5. Note that five residues on a single surface are substituted from Ile in AI10 to Ala in AI5. The Ile-to-Asn substitution at position 17 (indicated by an asterisk) is located on the surface shared between these two peptides and is opposite of the Trp-15 residue. Residues that have previously been implicated in forming an anti-parallel helix-helix interface within the context of covalently linked segments have been underlined (Johnson et al., 2004). aLys-tags increase peptide solubility in water, and are not expected to affect peptide stoichiometry (Melnyk et al., 2003).

While all peptides were judged to be of sufficient hydrophobicity for membrane insertion based upon Kyte-Doolittle (Kyte and Doolittle, 1982) and Liu-Deber (Liu and Deber, 1999) hydrophobicity scales, the least hydrophobic sequence (AI5I17N) was deemed to moderately favor partitioning into less hydrophobic phases based upon the White-Wimley scale (Wimley et al., 1996) and the Hessa et al. biological scale (Hessa et al., 2007) (ΔG of insertion into the more hydrophobic phase was between 1.0 and 0 kcal mol-1). We noted that the inconsistent categorization among the scales of the Trp residue as being either ‘favorable’ or ‘unfavorable’ for membrane insertion produced the bulk of the variation in estimations of TM peptide

hydropath. The AI5I17N sequence was also not identified as a TM segment by two of eight TM

prediction programs [the Phobius (Kall et al., 2004) or TOPSCON (Bernsel et al., 2008) prediction programs did not predict a TM topology]. This latter sequence is thus representative of TM segments that are near the threshold for in vivo membrane insertion as dictated by the net hydrophobicity of the core sequence. We therefore expect the sets of peptides studied here to reflect the range of hydropath characteristic of natural variety of native TM segments.

2.3.2 Oligomerization within bacterial membranes.

In order to determine initially if the designed sequences were capable of insertion and/or oligomerization in a native bilayer, we utilized the TOXCAT assay (Russ and Engelman, 1999) (Table 2.1). In this assay, a TM segment is expressed as a chimeric fusion protein situated between the DNA binding domain of ToxR (a dimerization-dependent transcription factor) and the maltose binding protein (a monomeric periplasmic anchor). Upon expression of the construct, self-association of the TM segment results in the ToxR-driven activation of a reporter gene encoding chloramphenicol acetyltransferase (CAT), with the level of CAT expression being proportional to the strength of helix-helix interactions. We found that the more hydrophobic

(AI10) sequence formed an oligomer with normalized CAT expression levels around 70% that of GpA, a known strong TM dimer. The mean level of CAT expression of the less hydrophobic

sequence (AI5) could not be distinguished statistically from AI10 (p = 0.11), indicating that the helix-helix interface is largely maintained within the native bilayer following Ile-to-Ala substitutions. Importantly, the addition of Asn to the more hydrophobic sequence (AI10I17N) increased the relative level of CAT expression to ~3.5 times that of GpA - a large signal that may be due to higher order (than dimer) oligomer formation (Johnson et al., 2007). In contrast, the

corresponding inclusion of an Asn residue in AI5 prevented efficient membrane insertion of this segment (as assayed by growth with maltose as the sole carbon source [see materials and methods]), and CAT expression levels were decreased to a value of about 15% of GpA.

Table 2.1. TOXCAT measurement of association of TM segments in the E. coli inner membrane. TM Segment Normalizeda CAT concentration (standard deviation) GpA 1.00 (0.04) -TMb 0.02 (0.01)

AI10 0.72 (0.17)

AI10I17N 3.80 (0.99)

AI5 0.50 (0.11)

AI5I17N 0.16 (0.01)

aMean levels of CAT expression have been normalized to relative expression levels of individual construct and then to mean CAT expression obtained from glycophorin A (GpA). b –TM is a negative control in which the reporter construct contains a stop codon directly before the position where the putative TM segment is inserted and represents baseline expression. See “materials and methods” for assay details.

2.3.3 Effect of polar residue substitutions on peptide oligomeric state(s) in SDS.

Having determined by the TOXCAT assay that the present peptides form stable oligomeric state(s) in bacterial membranes, we performed SDS-PAGE in a neutral pH system, as shown in Fig. 2.2. Many sequences corresponding to TM segments of native membrane proteins have been shown to retain their native oligomeric structure during SDS-PAGE experiments (Wegener and Jones, 1984; Heginbotham et al., 1997; Melnyk et al., 2003). Despite sharing an identical helical face shown by TOXCAT to be competent for helix-helix interaction in vivo, we found that the migration properties on SDS-PAGE of the two categories of TM segments (loaded with

2 µg) differ significantly (Fig. 2.2, right panel). The more hydrophobic peptide (AI10) runs

primarily as an apparent dimer (apparent MW from Rf analysis is ~250% of its formula MW)

while the less hydrophobic peptide (AI5) migrates at a rate intermediate between that of the predicted monomer and dimer migration rates (~170% formula MW). The introduction of a

single Asn residue to the common surface of the more hydrophobic peptide (AI10I17N) slows the migration rate to one more consistent with a dimer-tetramer equilibrium (~360% of formula MW), implying the creation of an additional helix-helix interface in this peptide. In contrast, the inclusion of an Asn residue in the less hydrophobic peptide at this same position (AI5I17N) does

not promote higher oligomer formation, but rather increases the gel migration rate to ~140% of its formula MW. Faint bands (< 5% of the intensity of the main bands) of lower MW in both

AI10 (~120% of formula MW) and AI10I17N (~90% of formula MW) may represent monomeric forms or impurities in the sample.

We further observed that the migration rate on SDS-PAGE of two of the peptides - AI10I17N and

AI5 - displayed a strong concentration dependence, with increased migration rates at lower peptide concentrations. This observation is explicable if these peptides exist in two oligomeric states that interconvert during the time course of gel electrophoresis. In this situation only one band will be present during electrophoresis, positioned according to the average oligomeric state (Artemenko et al., 2008). A decrease in concentration will shift the equilibrium towards the smaller species resulting in a decrease in the average oligomeric state producing a faster migration rate. At much lower concentrations (50 ng of peptide loaded) (Figure 2, left panel),

both AI5 and AI5 I17N run with a similar migration rate (~150% and 140% respectively), both

appearing as monomers, while AI10 and AI10 I17N both migrate as dimers (~190% and 230%, respectively).

Figure 2.2. Oligomeric state(s) of model TM segment peptides. SDS-PAGE of peptides performed with a 12% NuPAGE precast gel. Left: 50 ng of peptide was loaded in each lane, and the gel was developed by silver staining. Right: 2 μg of peptide was loaded in each lane and the gel was stained with Coomassie blue.

Oligomerization was also assessed using Förster resonance energy transfer (FRET) experiments. Peptides were labelled with dansyl (a fluorescence donor) or dabsyl (a fluorescence acceptor) moieties. Co-localization of oppositely-labelled peptides within 30 Å (indicative of oligomerization) results in energy transfer from dansyl to dabsyl, decreasing the fluorescence

intensity of the dansyl-labelled peptide. We found that the more hydrophobic peptide (AI10) with and without an Asn displayed specific FRET in SDS buffer (Fig.2.3A), while the less

hydrophobic series (AI5) peptides did not (Fig. 2.3B), demonstrating that at the concentration used for FRET (2 to 6 μM total peptide), the latter peptides were not capable of forming an oligomer in SDS. The AI5 series of peptides was further assayed for specific FRET in aqueous conditions, as they maintained a partially helical structure in the absence of detergent (see below); however, none was observed (data not shown).

In such experiments, the degree of observed FRET is dependent upon the specific oligomeric state(s) present, the degree of dye labelling and the relative orientation of the two fluorophores. Under the experimental conditions used here, the expected decrease in fluorescence for either

parallel or antiparallel helical dimers would range between 5-22% for AI10 and 3-14% for AI10 I17N. While the assays performed here are intended to serve as a qualitative indicator of the presence of an oligomeric state, these values are consistent with the observed fluorescence decreases of 20 and 10% respectively.

Figure 2.3. FRET experiments on TM peptides. FRET was performed in SDS buffer on A) AI10, and B) AI5 peptides, respectively. The integration of the mean fluorescent signal of a minimum of three independent preparations is reported for each mixture of labelled peptides. Fluorescence is normalized to the mean of ‘donor only’ samples. Donor only samples (D) contain 1 μM donor labelled peptide + 1 μM unlabelled peptide. Donor plus acceptor mixtures (DA) contain 1 μM donor labelled + 1 μM acceptor labelled peptide. Donor plus acceptor and unlabeled mixtures (DAU) contain 1 μM donor labelled, 1 µM acceptor labelled, and 4 μM unlabelled peptide. A decrease in fluorescence intensity upon addition of acceptor labelled peptide (compare D to DA) indicates FRET between the two pairs. A subsequent increase of fluorescence upon inclusion of unlabelled peptide to the donor and acceptor mix (compare DA to DAU), indicates a decrease in FRET - implying specific oligomerization. Error bars represent one standard deviation. Two sided t-tests were performed between (D) and (DA) as well as between (DA) and (DAU) assuming equal variance. (**) indicates p < 0.01.

2.3.4 Peptide characterization by circular dichroism and fluorescence spectroscopy.

To further assess the origin of the levels of peptide oligomerization in vivo vs. detergent solubilization, we performed both circular dichroism (CD) and tryptophan fluorescence spectroscopy in aqueous buffer and in SDS. While CD spectra revealed that AI10 and AI10I17N were largely unstructured in aqueous buffer (Fig. 2.4B), the AI5 series had significant helical structure [as evidenced by strong negative peaks at 208 and 222 nm] due both to the increased Ala content [an aqueous α-helix promoter (Chou and Fasman, 1978)] and decreased Ile content [an aqueous β-sheet promoter] (Fig. 2.4A). In media containing SDS, all peptides displayed increased levels of helicity vs. aqueous phase spectra (Fig. 4), consistent with an essentially full helical structure [MRE at 222 nm near ~36,000 deg cm2 dmol-1] (Chen et al., 1974; Manning and Woody, 1991). This is indicative of an uninterrupted hydrophobic helical span as observed in NMR studies of similar TM segments in detergents (MacKenzie et al., 1997; Melnyk et al., 2003).

Figure 2.4. Circular dichroism of model TM segment peptides. Circular dichroism spectra were recorded for each of the model TM segments. Spectra of peptides in aqueous buffer are rendered in blue, while spectra obtained with addition of 34.7 mM SDS are shown in black. AI5 and AI10 spectra are depicted as solid lines, AI5 I17N and AI10 I17N spectra as dotted lines. A) CD spectra of AI5 peptides in SDS vs. aqueous buffer. B) CD spectra of AI10 peptides in SDS vs. aqueous buffer.

Tryptophan fluorescence spectroscopy demonstrated that the more hydrophobic peptides (AI10

and AI10I17N) display a large blue shift in emission wavelength (31 nm) accompanied by an increase in intensity upon addition of SDS (Fig. 2.5A), both of which are strong indicators of insertion of the central Trp residue into the core of a detergent micelle. In contrast, the less

hydrophobic peptides both have comparatively smaller blue shifts (5 nm for AI5, 14 nm for

AI5I17N) and smaller increases in fluorescence intensity (Fig. 2.5B). The observation that

AI5I17N displays a broader fluorescence emission spectrum in SDS compared to AI5 with maximum intensity at a more blue-shifted position prompted us to examine this phenomenon in

further detail. We synthesized a series of AI5 analogs in which the position of insertion of the polar Asn residue was ‘walked’ through the peptide sequence from N16 through N20. We also prepared an analog that contained Asp instead of Asn at position 17, and one that had hydropathy partially ‘restored’ by converting Ala to Ile at positions 12 and 21 while retaining Asn at position 17. The results of SDS-PAGE analysis of this library are shown in Fig. 2.6A. While there is some variation in migration positions on the gel attributable largely to varying extents of SDS binding (Rath et al., 2009a), all single Asn-analogs, and the Asp analog, migrate faster than the

AI5 parent peptide, indicating their monomeric state. Finally, the set of fluorescence spectra we obtained on this library (Fig. 2.6B) indicates that only substitutions at position 17 produce the

relatively blue-shifted spectrum (including the 2I analog), while AI5 and AI5 analogs with Asn at positions 16, 18, 19, and 20, display only a minor blue shift (ca. 1-5 nm).

Figure 2.5. Tryptophan fluorescence of TM peptides. Tryptophan fluorescence spectra were recorded for each of the model TM segments. Spectra of peptides in aqueous buffer are rendered in blue, while spectra obtained with addition of 8.7 mM SDS are shown in black. Excitation was performed at 295 nm and emission was recorded between 310 and 400 nm. AI5 and AI10 spectra are depicted as solid lines, AI5I17N and AI10I17N spectra as dotted lines. A) Fluorescence spectra of AI10 and AI10I17N peptides. B) Fluorescence spectra of AI5 and AI5I17N peptides.

Figure 2.6. AI5 peptides containing a polar substitution. A) SDS-PAGE of a series of AI5 peptides performed with a 12% NuPAGE precast gel. In each lane, 2 μg of peptides was loaded. B) Changes in Trp maximum emission wavelength of AI5 peptides in SDS buffer vs. aqueous buffer. Results with AI10 peptides are shown for comparison. Error bars represent the standard deviations of a minimum of three separate sample preparations.

2.4 Discussion.

In order to preserve the native structure of a membrane protein, a detergent must be able to promote the retention of all native TM surfaces and interfaces. The increasing number of high resolution membrane protein structures demonstrates that to a major extent, detergents have proven to be sufficient models of this situation (White, 2009). However, despite its common use throughout protein expression and purification procedures, SDS has not often been used for structural determinations (see, however, NMR studies), and is primarily viewed as a ‘harsh’ detergent and a denaturant. A greater understanding of the details of SDS interactions with a

variety of TM segments is thus essential to addressing the categorization of SDS as a TM domain denaturant vs. a suitable membrane-mimetic.

2.4.1 Structure of peptide-detergent complexes.

In the TM peptides studied herein, both the more hydrophobic (AI10 series) and less hydrophobic

(AI5 series) peptides share an identical helical face designed to be competent for helix-helix interaction. Within this overall design, we nevertheless found that the inclusion of a polar Asn residue has opposing consequences for each segment, as deduced experimentally using SDS- PAGE gels, TOXCAT assays, CD spectroscopy, and Trp and FRET fluorescence spectroscopy.

Results from these combined studies indicate that in SDS micelles, AI10 forms a strong dimer

(one locus of interaction), and upon addition of an Asn (AI10I17N) a dimer-tetramer equilibrium

is experienced during PAGE analysis (two loci of interaction). In contrast, AI5 exists in a monomer-dimer equilibrium, but perhaps most intriguingly, the addition of an Asn (AI5I17N) stabilizes the monomer form (no locus of interaction). Although both series contain much of the same potential for specific sequence-driven association, their SDS solvation states differ, in turn altering their helix-helix interactions. For the AI5 series of peptides, a case can be made that with

the reduced hydropathy of AI5 vs. AI10 - and with further reduction upon inclusion of a polar residue - the AI5 peptides reside on average much closer to the effective micelle-water interface

than do the AI10 peptides. This interpretation is supported experimentally by the observations

that (i) the AI5 series is predicted from several TM hydropathy scales to have only a moderate energetic preference between polar and nonpolar phases; (ii) AI5I17N is a monomer, indicating that Asn-17 is not available for inter-helical H-bond formation; (iii) TOXCAT assays support this largely-monomer state (a manifest consequence of inefficient membrane insertion); and (iv)

Trp blue shifts are > 30 nm for the AI10 series, indicating complete “Trp burial” in the micelle.

In contrast, blue shifts for the AI5 series range from 1-5 nm for all peptides except those with Asn/Asp at position 17 (Fig. 2.6B).

These results indicate the sequence-dependence of the overall peptide-micelle structure, and further suggest that the polar site specifically at position 17 – occurring on the face directly opposite to the Trp-15 residue (Fig. 2.1B) – apparently ‘fixes’ the Trp at a relatively micelle- buried position while the polar residue is drawn to the micelle-water interface. The broad

fluorescence spectra displayed by the three N/D-17 peptides (as in Fig. 2.6B) appear to indicate the existence of two (monomeric) states for these peptides, viz., one nearer and one further from the effective water-micelle interface; this latter suggestion is supported by the observation of a shoulder on the principal peak in size exclusion chromatography performed in SDS (not shown).

Finally, while CD spectra of AI10, AI10I17N, AI5 and AI5I17N all display comparable high helicity in SDS, we note that the Ala-rich peptides have high helicity in the aqueous phase vs. the Ile-rich peptides (Fig. 2.4), so that there is unlikely to be significant fraying of helices in any

water-based portion(s) of the AI5 segments that arise in the presence of micelles.

These overall experiments lead to the description of the structures of these peptides within the

detergent micelles illustrated schematically in Fig. 2.7. AI10 behaves according to the classical model of TM segment association within micelles, existing in a fully detergent-solvated state primarily as an oligomer (Fig. 2.7A). While SDS induces an increased helical conformation in

AI5, it apparently does so without full solvation of side chains by the acyl detergent tails. In this case, the TM segment does not adopt the classical view of an α-helix inserted into the core of a micelle (Fig. 2.7B), but instead has a surface exposed to a detergent head group or to the aqueous environment within the peptide-detergent complex (Fig. 2.7C). This results in a decrease in the strength of the protein-protein interface (evidenced by lack of dimerization during FRET [2-6 μM concentration] or during SDS-PAGE upon loading ng amounts [Fig. 2.2, left panel]) as some fraction of the segment populates the water-interactive state, preventing full acyl tail solvation of the ostensible site of oligomerization. Such a model has been previously reported in other TM segments BNIP3 (Sulistijo and MacKenzie, 2006) and GpA (Duong et al., 2007), where non- interfacial mutations that reduce hydrophobicity similarly decrease the apparent dimerization

affinity in SDS. Thus, while the AI10I17N system likely forms a tight Asn-Asn side chain-side chain H-bond that produces a new helix-helix interaction interface along with higher oligomers, the introduction of I17N into the AI5 system ultimately reduces the local hydrophobicity to the level that partial aqueous solvation becomes the most stabilizing alternative.

Figure 2.7. Peptide interactions within detergent micelles. Individual TM segments may exist in a variety of states in detergent micelles, as depicted schematically in (A)-(C). A) An oligomeric segment may have a surface with a preference to form tertiary interactions with another surface, effectively producing a protein-protein interface between interacting monomers. B) A hydrophobic segment may have a strong preference for detergent solvation and exist as a monomer fully solvated by the acyl tails of the detergent. C) A less hydrophobic surface or local sub-segment arising within a TM-like segment may have a preference for aqueous solvation, and partition to a state where the segment is partially solvated by detergent head groups and/or the surrounding aqueous environment.

2.4.2 The effectiveness of SDS as a ‘membrane-mimetic’.

The present study demonstrates that SDS can support TM helices as oligomeric species (i.e., the

AI10I17N peptide displays both the highest oligomeric state(s) in SDS (Fig. 2.2, right panel) and a correspondingly high TOXCAT signal (Table 2.1)); stabilize a micelle-solvated monomeric state; and/or support a state promoted by aqueous solvation of a local segment at the micelle- water interface. Importantly, the access to water available to a TM segment embedded in a micelle system (vs. the absence of this possibility in a lipid bilayer without major structural rearrangement) may serve as a surrogate for the in vivo structures of many membrane proteins where helix-aqueous interactions that contribute directly to protein structure and function (i.e., pores, channels) span the TM domains of the proteins (Abramson et al., 2003; Stroud et al., 2003). Thus, where close-to-threshold hydrophobicity exists for a given TM sequence, introduction of a polar mutation into a given protein- or lipid-exposed surface may induce a preference for an aqueous-exposed conformation (a change from Fig. 2.7B to Fig. 2.7C). The capacity of SDS to respond to such nuance of sequence as well as to peptide concentration mitigates against the typical view of denaturation by ‘harsh’ detergents.

Our overall findings suggest that SDS solvation of transmembrane segments - particularly with sequence-specific preservation of helix- helix interactions, and compatibility with aqueous- exposed regions - may in fact constitute a relevant biological model. The striking ability of the local SDS micellar environment to alter conformation and stoichiometry due to a single point mutation suggests that corresponding mutations in vivo can similarly impact local protein structure. The SDS-induced denaturation that is commonly observed for TM regions of proteins may in some cases be an allosteric consequence of the loss of native structure in extramembranous regions. As well, in SDS unfolding experiments there may not always be large conformational changes in TM regions that correspond to a true unfolding transition (Renthal, 2006). At least for membrane proteins with relatively short, stable extramembranous loops, our results suggest that SDS could largely preserve robust tertiary structure, and as such, serve a useful mimic of protein structure within the hydrophobic core of a biological membrane.

Chapter 3. Positions of polar amino acids alter interactions between transmembrane segments and detergents.

Contents of this chapter have been published, in part, by Tulumello D.V., and Deber C.M. (2011). Positions of polar amino acids alter interactions between transmembrane segments and detergents. Biochemistry. 50(19); 3928-3935.

3.1 Introduction.

In the previous chapter we have shown that the overall detergent – transmembrane segment complex to be dependent on the hydropathy of the segment. Such variations in how detergents interact with TM segments may alter the fine-tuning of helix-helix interactions, in turn altering overall membrane protein folding and/or susceptibility to denaturation in detergents. Yet this versatility, in turn, allows in-depth assessment of the structural consequences – both local and longer range – of introducing a point mutation in a given TM segment.

As many as 25% of polytopic membrane proteins have at least one TM segment not expected to be capable of efficient membrane insertion – due to relatively low segment hydropathy - in the absence of the remainder of the protein (Hessa et al., 2007). Furthermore, disease-phenotypic non-polar to polar mutations are common in TM domains (Partridge et al., 2004) and in some cases have been shown to prevent translocon mediated insertion of TM segments (Enquist et al., 2009). Although strongly polar residues can promote oligomerization of otherwise highly hydrophobic TM segments (Gratkowski et al., 2001; Zhou et al., 2001), little is known about the structural consequences of such residues when they occur in low hydropathy TM segments, where they would be predicted to become destabilizing towards insertion. In the present work, we have systematically investigated how detergent solvation is affected by the sequence of the spanning protein segment, focusing primarily upon the positional dependence of the introduction of a strongly polar residue (Asn) into a moderately hydrophobic TM segment. We have chosen to study the anionic detergent sodium dodecyl sulfate (SDS) due to its widespread use in identifying helix-helix association interfaces (Lemmon et al., 1992a; Arkin et al., 1994; Melnyk et al., 2001; Therien and Deber, 2002a), as well as its known sensitivity to protein sequence (Rath et al., 2009a). Our findings suggest that single polar residues in TM segments may impact fundamentally on membrane protein folding.

3.2 Materials and methods. 3.2.1 Assessment of hydrophobicity and hydrophobic moment

The relative hydrophobicities of the model TM segments were assessed based upon predicted

free energies of translocon-mediated membrane insertion (ΔGins) (available online at http://dgpred.cbr.su.se) (Hessa et al., 2005; Hessa et al., 2007), water to lipid bilayer interfacial region partitioning (ΔGwif) (Wimley and White, 1996),water to octanol partitioning (ΔGwoct) (Wimley et al., 1996), as well as an hydrophobicity scale derived by reverse phase high

performance liquid chromatography (Liu and Deber, 1998a). Hydrophobic moments (μHΦ) of the peptide sequences (not including the flanking Lys residues) were calculated from individual

amino acid hydrophobicities (Hi) by Equation (3.1) (Eisenberg et al., 1982):

2 2 1/2 μHΦ = {Σ[Hi sin(100i)] + Σ[Hi cos(100i)] } (3.1)

For the translocon mediated scale Hi values were obtained from the aforementioned prediction server, specific to the predicted transverse positioning of each residue in the central core, where 1 21 residue A11 is in the center of a 21 residue TM segment (Y to A ). For other scales Hi values were not position sensitive. The i = 0 position was chosen at the central Trp residue in the primary sequence, so that this residue projected directly onto the positive x-axis. The magnitude of the hydrophobic moment in the direction of the Trp residue was calculated by Equation (3.2):

μHΦ (Trp) = Σ [Hi cos(100i)] (3.2)

3.2.2 Peptide synthesis and purification.

Peptides were synthesized as described in Chapter 2. Peptide concentrations were determined using quantitative amino acid analysis of SDS solubilized samples performed by the Advanced Protein Technology Centre of the Hospital for Sick Children (Toronto, ON). Based upon replicate analysis of the wild type protein and select variants, as well as comparison to other concentration determination methods (UV absorbance), the estimated error in concentration determination is 10%.

3.2.3 Circular dichroism spectroscopy.

Freshly thawed aliquots of peptides were diluted to a final concentration of between 15 and 40 μM in 10 mM Tris HCl buffer (pH 8.0), with or without 34.7 mM SDS. Spectra were recorded as described in Chapter 2. For each peptide, CD spectra were obtained at a minimum of two peptide concentrations. For most peptides the standard deviation between replicate circular dichroism measurements was of the same order of magnitude or less than the estimated error in the concentration determination.

3.2.4 Tryptophan fluorescence measurements.

Freshly thawed aliquots of peptides were diluted to a final concentration of 10 µM in 10 mM Tris HCl buffer (pH 8.0), with or without 34.7 mM SDS. Additional samples were prepared by diluting those used for circular dichroism in a 1:1 ratio with buffer, and used at final concentration between 10-20 µM. Fluorescence emission spectra of peptides were measured as described in Chapter 2. The wavelength of maximum emission intensity was obtained from these spectra, and the blue shift was calculated as the difference between the average wavelength of maximum emission in aqueous buffer and SDS-containing buffer.

3.2.5 Sodium dodecyl sulfate-polyacrylamide gel electrophoresis analysis.

SDS-polyacrylamide gel electrophoresis (PAGE) was performed as described in Chapter 2. Silver stain (Invitrogen) was used to visualize proteins on all gels. 20 ng of each peptide were loaded in a total volume of 10 µL to each lane. At this concentration the wild type peptide was previously shown to be a monomer (Tulumello and Deber, 2009). Apparent molecular weights

were estimated based upon Rf analysis using Mark12 molecular standards (Invitrogen).

Migration distances were determined using NIH Image. Gel shift values (∆MWapp) were calculated as: ([estimated MW from gel / formula MW from sequence]), and normalized to the wild type peptide on each gel.

3.3 Results. 3.3.1 Peptide design and TM insertion prediction.

To address the membrane insertion requirements of prototypic TM strands in a systematic manner, we designed peptides based, in part, upon a moderately hydrophobic TM segment (Tulumello and Deber, 2009). The hydrophobic core of these peptides consists primarily of Ile and Ala residues, as well as a central Trp residue (sequence of the ‘wild type’ peptide, termed AI5 shown in Figure 3.1A). The wild type peptide is predicted to have a ΔG of translocon mediated insertion of -0.55 kcal mol-1 (Hessa et al., 2007), a ΔG of octanol partitioning of -1.4 kcal mol-1 (Wimley et al., 1996), a ΔG of POPC interfacial partitioning of -2.0 kcal mol-1 (Wimley and White, 1996), and a Liu-Deber average hydrophobicity of 1.2 (above the predicted threshold value for membrane insertion of 0.4) (Liu and Deber, 1998a). The WT peptide has been shown to be capable of insertion and dimerization in a native bilayer; it is able to adopt a helical structure in SDS micelles but does not dimerize at μM concentrations (Tulumello and Deber, 2009). In addition, the presence of Lys tags confers aqueous solubility to the peptides, allowing us to characterize their secondary structure in the absence of detergent (Melnyk et al., 2003).

In the present work, a number of variants of this peptide were produced by introducing single Asn substitutions, beginning at the central position (residue 11 of the hydrophobic core) and proceeding toward the C-terminus. The inclusion of this polar residue reduces the hydrophobicity by incremental amounts, depending upon (i) whether the Asn is substituted for an Ala or Ile residue; and (ii) the position of the substitution in the context of predicted translocon mediated insertion (see Appendix 2: Table A2.1). In many cases, this change lowers the net hydrophobicity below the predicted insertion threshold. The position of the Asn residue may further produce large differences in local hydrophobicity on one side of the helix versus the other. To mathematically quantify the asymmetry of the degree of hydrophobicity between opposite helix surfaces, we calculated the “hydrophobic moment” of each peptide using various hydrophobicity scales, as described in the materials and methods. In all cases the wild-type peptide had a relatively large hydrophobic moment (indicating one surface has greater hydrophobicity than the other), with decreased hydrophobicity in the direction of the Trp residue. Substitutions made close to position of the Trp residue about the helical axis increased the

hydrophobic moment up to ~50% without significantly changing its direction, while substitutions on the opposite side of the helix decreased the magnitude by up to ~50%, indicating that local hydrophobicity of individual surfaces was highly dependent on position of this single substitution (see also Appendix 2: Table A2.1).

3.3.2 SDS-PAGE migration rates of Asn containing peptides.

SDS-PAGE migration is commonly used to identify the monomeric molecular weight (MW) of a protein, although membrane protein oligomeric states may be retained (Wegener and Jones, 1984; Heginbotham et al., 1997; Poulsen et al., 2009). SDS-PAGE is also sensitive to protein conformation (Griffith, 1972; Dunker and Kenyon, 1976) and we have previously demonstrated that gel migration rates may, in part, be traceable to conformational changes in membrane protein structure (Therien et al., 2001; Wehbi et al., 2008). To a large degree, migration rates of soluble proteins in SDS-PAGE are determined by particle size due to the sieving effects of a polyacrylamide gel (Reynolds and Tanford, 1970; Rodbard and Chrambach, 1970). As increasing SDS binds to the segment, the mass/negative charge ratio will decrease, an effect that should, in principle, increase the gel migration rate. However, based on previous studies from our lab using a membrane protein model composed of two TM segments, increasing detergent binding is in fact correlated with decreasing gel migration, indicating that particle size is the dominant factor (Rath et al., 2009a).

When we performed SDS-PAGE migration analysis on the present series of peptides containing Asn mutations (Fig. 3.1B), all peptides migrated as a single species at a rate consistent with a monomeric state. Although these Asn-containing peptides differ at only one position – and all accordingly have very similar MWs to the wild type version – nearly all migrated significantly faster than WT peptide. The migration rates of the peptides followed two general trends: (i) the addition of Asn increased the rate of migration, with Ile-to-Asn substitutions increasing the rate more than Ala-to-Asn substitutions; and (ii) the observed increase of migration rate was more pronounced in substitutions located at or near the center of the TM helix sequence. Additionally, it has been shown that central polar residues are more deleterious to translocon mediated insertion of transmembrane segments, and therefore behave as less hydrophilic the closer they are to the helix ends (Hessa et al., 2007). Thus, both of these effects observed in gel migration

are consistent with the changes of hydrophobicity induced by these substitutions, where less hydrophobic sequences have faster gel migration as we have observed in previous studies (Rath et al., 2009a). Accordingly, the introduction of a polar residue at different positions produces similar trends in both gel migration rate and hydrophobicity changes as shown in Fig. 3.1C.

Figure 3.1. Peptide sequence and SDS-PAGE analysis of Asn peptide variants. A) Amino acid sequence of the base peptide (wild type (WT)) under investigation. The region in which Asn substitutions were made is underlined. B) Representative SDS-PAGE of WT and single Asn containing peptides. C) Plotted on the primary axis are the mean relative gel shifts (see materials and methods) compared to wild type (ΔMWapp [mut – WT]) shown as solid circles connected with a solid line. Error bars represent the standard deviation in the calculated gel shift from a minimum of four gels for this set of peptides. Plotted on the secondary axis is the difference in contribution to free energy of translocon mediated insertion between an Asn residue and the native amino acid (ΔGins [mut – WT]) at that location (open circles and dashed lines).

3.3.3 Peptide helicity in aqueous vs. detergent environments.

Circular dichroism was used to assess the adoption of α-helical secondary structure upon solubilization of members of this peptide library in either aqueous buffer or SDS. The wild type peptide contained significant helical structure in aqueous solution (Fig.3.2A, left panel). The introduction of an Asn residue at any location leads to a significant decrease in helicity (evidenced by a lower signal at 222 nm, and a shifting of the minimum near 208 to lower wavelengths, indicative of a population of ‘random coil’ structure (Gans et al., 1991), as predicted from Chou-Fasman parameters (Chou and Fasman, 1978) [shown for representative spectra in Fig. 3.2A, left panel]). In SDS, the wild type peptide as well as all variants had increased levels of helicity compared to aqueous buffer (shown for representative spectra in Fig. 3.2A, right panel). The majority of mutations contained amounts of helical secondary structure 2 -1 that were within 10% of wild type values [θ222nm = -33,000 deg cm dmol ]. Of all variants, only three had a significant decrease in helicity in SDS buffer compared to wild-type (Fig. 3.2B).

In order to make further quantitative comparisons among mutants, we measured the ratio of helicity in SDS buffer to that in aqueous buffer. This method eliminates any variation introduced by inaccuracy in concentration determination, which in many CD studies can be the largest source of error (see materials and methods). The wild type peptide has a SDS/aq. ratio of 1.7, indicating that SDS increased the helicity of this peptide by 70%. All Asn-containing variants had significantly (p < 0.05) larger ratios, ranging from 2.0 to 2.9 (Fig. 3.2C). This increase in ratio indicates that Asn was more detrimental to the formation of helical structure in aqueous buffer than in SDS, and therefore in a sense the “folding competency” of these segments is retained in SDS. An additional trend in these data is that all Ile-to-Asn mutants displayed a ratio of ~ 2, while all Ala-to-Asn mutants had larger ratios. This result indicates that removal of Ile is more detrimental to helix formation than removal of Ala in a membrane environment compared to an aqueous environment. Although Ala is predicted to be a superior helix former than Ile even in membrane mimetics (Liu and Deber, 1998b), in this situation hydrophobicity appears to be the dominant factor in helix formation. The exception to this trend was the most central Ala-to-Asn mutation at position 11, which had a ratio of 2.1 and was more similar to the Ile-to-Asn mutations in this regard.

Figure 3.2. Circular dichroism of peptides. A) CD spectra of wild type and selected variants in aqueous – left panel, or SDS containing buffer – right panel. In each panel, mutants are listed in order of increasing helicity at 222 nm. B) Mean residue ellipticity (deg cm2 dmol-1) [θ] at 222 nm of peptides in aqueous or SDS buffer. Error bars represent the combined standard deviation from both the CD measurements and concentration determination. CD measurements were preformed on between three to six samples for each peptide in each buffer condition. C) Fold increase of helicity from aqueous to SDS buffer (measured by MRE at 222 nm). Error bars represent the combined standard deviation from the CD measurements. In panels B and C, WT levels are indicated by a dashed line. Two sided t-tests were performed between WT and each peptide variant, (**) indicates p < 0.01.

3.3.4 Detergent binding as a dominant factor in SDS-PAGE migration.

While there are some apparent differences in secondary structure among the peptide Asn variants in SDS, there was only a moderate inverse correlation between gel migration rate and helical structure (R2 = 0.34, p = 0.05) (Appendix, Fig. A2.1A), indicating that changes in peptide structure alone were not a major factor in the phenomena observed. The Asn scanning results (Fig. 3.1) are thus consistent with a reduction of detergent binding due to the decreased hydrophobicity presented by polar residues. To confirm that differences in hydrophobicity are a determining factor here, we synthesized a peptide with a central Ile-to-Asp mutation (I12D); and a further I12N analog also containing two Ala-to-Ile substitutions designed to “restore” hydrophobicity (Fig. 3.3A). In addition, as a direct test of the dependence on sequence vs. composition, we synthesized a peptide variant (termed ‘scrambled’ [scr.]) that has an identical composition as wild type, but differs in sequence in a manner such that Ile residues are distributed evenly around the helix axis, as well as positioned towards the helix ends (helical wheel diagram in Fig. 3.3B). All additional peptides were found to have a similar level of helicity as WT peptide (data not shown). As shown in Fig. 3.3C, the Asp mutant migrated similarly to the Asn mutation at this position, while the more hydrophobic variant (I12N+2I) now migrated at a rate much closer to WT in accordance with its increased hydrophobicity with respect to the other polar variants. Interestingly, the scrambled version migrated at a relative rate ~ 7% slower (p < 0.01) than the wild type version, indicating the SDS-PAGE approach can detect sequence-dependent differences-in-detail in the overall detergent-peptide complex.

Taking into account all peptide variants studied, we observe a strong correlation of gel migration rate with net hydrophobicity as assessed by translocon mediated membrane insertion (Hessa et al., 2007) (R2 = 0.71, p < 0.001) (Fig. 3.3D). In contrast, the correlation between gel migration and helicity was no longer statistically significant upon inclusion of these additional variants (R 2 = 0.01, p = 0.71) (Appendix, Fig. A2.1B). We thus hypothesize that the majority of the observed variance in gel migration between peptide-detergent complexes is therefore due to differences in detergent binding as the source of variation in overall particle size.

Figure 3.3. Sequence and SDS-PAGE analysis of additional peptide variants. A) Amino acid sequences of the additional peptide variants studied, along with predicted free energy of translocon mediated membrane insertion (Hessa et al., 2007). B) Helical wheel diagram of both the wild type peptide (termed AI5 (Tulumello and Deber, 2009)) and the scrambled variant. C) Representative SDS-PAGE of these variants. D) The correlation between observed SDS-PAGE gel shift of all peptide variants (panel C, Fig. 3.1B) and predicted free energy of translocon mediated insertion is shown with correlation coefficient, R=0.84. The trend line p-value is 0.0001. Selected Asn-containing variants, as well as all segments shown in panel C, are labeled. Error bars indicate the standard deviation of gel shifts for each variant obtained for 3-8 independent gels.

3.3.5 Tryptophan fluorescence reveals variations in SDS distribution.

In previous studies of the wild type peptide, Trp fluorescence was used to demonstrate that the central Trp residue was in an aqueous environment, while the positioning of an Asn residue across from this Trp caused it to become more micelle-buried (greater blue shift) (Tulumello and Deber, 2009). In the current study, we extended this analysis to all Asn-containing variants between positions 11-21. Although all mutants effectively lowered the overall hydrophobicity, at some positions the addition of Asn resulted in a greater blue shift of the Trp residue. There is a periodicity in the pattern of Trp blue shift as the Asn is placed at various locations across the helical axis, up until position 17, after which the blue shifts become similar to wild type (Fig.

3.4A). We also calculated the magnitude of the “hydrophobic moment” vector in the direction of the Trp residue as an indicator of how ‘hydrophobic’ this face is compared to the remainder of the segment. We found that peptides that were relatively more hydrophobic on the side opposite the Trp residue (a larger value on this scale) had smaller Trp blue shifts, while those that had a more even distribution of hydrophobicity (a smaller value) had larger blue shifts (Fig. 4A). This provides evidence that there is an unequal distribution of SDS about the helix, and that this distribution may be sensitive to alteration by polar substitutions. Indeed, while the scrambled peptide variant (with a minimal “hydrophobic moment”) had a similar level of helicity as WT peptide (data not shown), it had a greater blue shift [11 nm vs. 6 nm for WT] – a value more typical of insertion into the center of a micelle (Fig. 3.4B).

Figure 3.4. Tryptophan blue shifts induced by SDS solvation, and “hydrophobic moments” of peptides. A) Patterning of blue shift (solid circles, and line), compared to “hydrophobic moment” along the axis in the direction of the Trp residue (open circles, and dashed lines). The values of blue shifts for the WT peptide and for variants between and including A11N and I15N have been previously reported (Tulumello and Deber, 2009). B) Trp emission spectra of WT (solid lines) and scrambled peptide variant (dotted lines) in aqueous solution (grey lines) or SDS-containing buffer (black lines). For each spectrum, the fluorescence intensity has been normalized to the maximum intensity in aqueous buffer.

3.4 Discussion. 3.4.1 Alteration of detergent solvation of hydrophobic peptides induced by polar residues.

The side chain of an Asn residue typifies a strongly polar functional group that would be intrinsically unfavorable for hydrophobic interactions. When we placed the Asn residue in an otherwise hydrophobic sequence (characteristic of a protein TM segment), it is evident from the biophysical analysis performed herein that this residue may have a drastic effect on how the resulting segment interacts locally with the lipid environment of a detergent. Such large variation is observed even in model TM segments of identical composition, highlighting the nuances of such differences.

While perhaps not considered classically amphipathic due to a lack of a truly hydrophilic surface, there is still a modest “hydrophobic moment” present in the TM helices studied here that dictates the relative conformation of these peptides within the detergent micelles. Specifically, if the net hydrophobicity is directed away from the Trp residue, as in the wild type or variants where the Asn is on the same surface of the helix, the Trp has a relatively small blue shift - indicating some exposure of this surface to water and/or the head group regions of the micelle. Conversely, a substitution opposite to the Trp (such as I12N and A16N) causes the further burial of this same helix surface into the center of the micelle. These findings support the notion of conformational sensitivity of a helix such as the present WT sequence that contains both a strongly hydrophobic surface and a moderately hydrophobic surface. Thus, in contrast to the situation where a helix contains a hydrophobic face and a demonstrable hydrophilic face, one could term the WT sequence as “lipopathic” due to the presence of two inherently hydrophobic faces with differing preferences to interact with lipid-like environments such as the core vs. the surface of a detergent micelle.

Although the incorporation of polar residues at most positions does not have a major effect on membrane-based helicity, they do alter the solvation of SDS in two ways:

Detergent binding is lost, likely in the area where the polar residue is located, as manifested by a faster peptide gel migration rate, indicative of a smaller particle size. This phenomenon depends primarily on hydrophobicity as most Ile-to-Asn substitutions display greater migration difference

vs. the base peptide than Ala-to-Asn substitutions. It is also noted that the introduction of an Asn residue at locations closest to the Trp residue decreases the blue shift, indicating less detergent present locally in these areas.

The distribution of SDS changes, as evidenced by the observed systematic variations in Trp blue shifts. This phenomenon is related to the “lipopathy” of the helix in that more SDS is distributed to the more hydrophobic surface, creating a situation where the helix adopts a type of “interfacial” conformation. This situation allows for the polar residue to be accommodated in a less hydrophobic environment, even if it is surrounded by hydrophobic residues.

These two effects are also potentiated by the position of the mutated residue with respect to the core of the hydrophobic sequence. The overall interpretation of the present work with respect to polar residue substitutions within a detergent-solubilized system is shown schematically in Fig. 3.5.

Figure 3.5. Schematic representation of detergent-peptide complexes. Detergent-peptide complexes are arranged from left to right according to decreasing relative particle size based upon SDS-PAGE migration rates. Ala and Ile residues are shown in blue, Trp in black and Asn in red. In this schematic, peptides that have faster gel migration rates are depicted as having less detergent bound. For WT (panel B), Trp fluorescence results show that SDS is not distributed evenly along the peptide sequence, and that the side containing the Trp along with Ala residues is less solvated than the Ile side. The role of the “lipopathic” nature of these peptides (see Discussion) is highlighted by analysis of the scrambled variant (panel A), for which the even distribution of side chain hydrophobicity allows for increased overall detergent binding, in turn giving rise to a slower migration rate for the scrambled variant vs. WT peptide despite their identical composition. Also depicted are the situations expected for a typical Ala-to-Asn mutation (A14N) (panel C) vs. a typical Ile-to-Asn mutation (I12N) (panel D). In both latter cases, less detergent is bound at the location of the Asn residue. For situations where Asn is not located on the same surface of the Trp residue (as is the case for I12N), we infer that the loss of detergent occurs locally where hydrophobicity is decreased. Additionally the large change in hydrophobic moment produced in the I12N variant produces an altered distribution of detergent in comparison to WT, where the Trp residue is now more “buried” in the micelle (panel D). Images of the TM segments were produced using PyMol (Schrödinger LLC).

We further observed that introduction of polar residues at most positions has only minor effects on membrane-based helicity, and the relatively low resolution of circular dichroism spectra limits our ability to detect minor kinks, or some local helix fraying. Interestingly the A11N variant is unique in that it has a migration rate faster than any Ile-to-Asn variant, appearing as an outlier based solely upon its hydrophobicity. It is also the least helical among the Ala-to-Asn mutants, and thus displays the smallest relative increase in helicity upon detergent solvation. Likely at this most central position, loss of hydrophobicity leads to local unfolding of the helix, as has been previously observed in other TM segments (Sulistijo and Mackenzie, 2009; Ng and Deber). The variant I19N is another example where loss of helical structure is detected by CD.

This variant has a gel migration rate somewhat greater than that of the wild type – an observation that is not directly explicable by a loss of hydrophobicity. We note, however, that substitution drastically decreases local hydrophobicity at the C-terminus, removing any strongly hydrophobic residue for an extended stretch (e.g., sequence AAANAAKKK for residues 16-24), and leads to some loss of helical structure. Given that these peptides have been Lys-tagged for synthesis, this result may arise from extant electrostatic interactions that the Lys residues have with the anionic SDS detergent. Ala-to-Asn substitutions in this region may not produce the same effect due to preservation of the most terminal Ile, and therefore a less drastic hydrophobicity change and accompanying preservation of the helical structure. These outliers indicate that detergent binding as function of hydrophobicity alone is not the sole determinant of gel migration rate, and a complex interplay of interactions is necessary to fully describe the peptide-detergent structures.

3.4.2 Features of SDS solvation that mimic membrane protein folding.

Strongly polar residues have been shown in many cases to have drastic effects on membrane protein folding, such as TM segment insertion (Hessa et al., 2007), as well as transverse positioning as observed in model bilayers (Caputo and London, 2004; Krishnakumar and London, 2007). These effects are often dependent upon the position of the residue in relation to the midpoint of the transmembrane segment sequence. The underlying sources of these effects have been observed in computational studies that show that polar residues near the ends of helices are less perturbing to membrane insertion, as the hydrophilic moieties may snorkel into the interfacial region; in contrast, polar residues positioned in more central TM regions may form H-bonds with the helix backbone, or in extreme cases (i.e., for charged residues) lead to local bilayer defects (Johansson and Lindahl, 2006; MacCallum et al., 2008). In the present work, SDS-PAGE migration rates are shown to be similarly sensitive to this positioning, and indeed there is a corresponding correlation between the predicted position-dependent in vivo insertion propensities (Fig. 3.3D). Our overall results therefore suggest that variation in sequence- dependent peptide-SDS interactions observed in biophysical processes may be similar to variation in biological TM segment insertion/ folding.

Our work also indicates that detergent systems are capable of stabilizing individual TM segments that contain both micelle- and aqueous-interactive surfaces. Such conformations have been

similarly observed in studies of antimicrobial peptides (Wimmer et al., 2006; Porcelli et al., 2008), as well as for the structures of various TM segments derived from membrane proteins (Vincent et al., 2007; Le Lan et al., 2010). In unfolding studies of full-length proteins, SDS solvated helices are often considered as the “denatured” state (Curnow and Booth, 2007; Joh et al., 2008). It is clear from our studies that polar residues can greater alter this state - a factor that must be taken into consideration for obtaining free energies of unfolding. Of equal importance is the implication from our work that where “lipopathicity” occurs in natural TM segments, a single polar mutation in an otherwise hydrophobic face can propagate potentially drastic changes in local folding patterns – perhaps most significantly in proteins that have a membrane-spanning aqueous pore or channel that may attract the polar side chain. These revelations reinforce the notion that many of the same forces driving membrane protein folding in vivo are mirrored by detergent solvation.

Chapter 4. Transmembrane segment structures are similar in native proteins and detergents

Contents of this chapter have been published, in part, by Tulumello D.V., and Deber C.M. (2012). Efficiency of detergents at maintaining membrane protein structures in their biologically relevant forms. Biochimica et Biophysica Acta – Biomembranes. 1818(5); 1351-1358.

4.1 Introduction.

As the structural database of membrane proteins has grown, it is becoming increasingly apparent that not all features of TM domains are represented within the paradigm of the two-stage model (Engelman et al., 2003; Viklund et al., 2006; Yan and Luo, 2010). For example, it has been noted that as many as one in four multi-spanning TM proteins has a membrane-embedded segment that is predicted to lack sufficient hydrophobicity for insertion (Hessa et al., 2007). Such TM segments are common in transporter proteins where segments that line internal cavities are partially exposed to an aqueous milieu, and thus to a much different environment than their more hydrophobic counterparts.

A variety of experiments have shown that native protein quaternary structure may be maintained in detergent environments (MacKenzie et al., 1997; Melnyk et al., 2001); aspects of secondary and tertiary structure may be similar in lipid bilayer versus micelle environments (Franzin et al., 2007); and the overall fold of a protein may persist in detergents (Federkeil et al., 2003). In both Chapters 2 and 3 we have observed that the interaction of model TM segments with detergents is highly sequence-dependent, and in many cases mimics the predicted in vivo folding (Tulumello and Deber, 2009, 2011b). It is thus likely that certain features (such as secondary or tertiary structure) of native TM segments persist upon solubilization as detergent complexes. However, the ability of some, if not all, detergents to preserve such features appears to differ on a case-by- case basis. To address these issues systematically, in the present work we investigate the folding of a variety of native and non-native TM segments in a selection of detergents. Overall, our results demonstrate that for these TM segments, there is a similarity in structure when these individual segments are solubilized in detergents vs. their structures as they occur within the full- length protein. We also observe that these TM segments are generally indifferent to the detergent used for solubilization, suggesting a degree of commonality among micellar environments.

4.2 Materials and methods. 4.2.1 Transmembrane protein structural analysis.

High-resolution structures of membrane proteins were downloaded from the Membrane Proteins of Known 3D Structure database (available online at http://blanco.biomol.uci.edu/mpstruc). α-

Helical TM proteins classified as a channel or transporter (symporters, antiporters included), were manually inspected for Trp residues within the TM region. TM segment boundaries were taken from the Orientations of Proteins in Membranes database (http://opm.phar.umich.edu/ - accessed Feb. 2010) (Lomize et al., 2006). Segments that contained more than one Trp residue or a Trp residue that was located within three residues from the lipid acyl chain/head group boundary were excluded from further analysis.

For protein structures that contained one or more segments that met this criteria, residue burial was calculated using the program NACCESS (Hubbard and Thornton, 1993). A probe of radius 1.88Å was used as an approximation of the radius of a methylene group of a lipid acyl chain, and analysis was preformed as previously described (Rath and Deber, 2007). A given amino acid residue that had a relative accessibility of less than 10% was classified as buried. The remaining residues were classified as having lipid exposure (outward facing) or aqueous exposure (facing an internal cavity) based upon manual inspection of the structure. For each TM segment, secondary structure assignments were made using DSSP (Kabsch and Sander, 1983; Joosten et

al., 2011), and predictions of free energy of translocon-mediated membrane insertion (ΔGins) were made using the ΔG prediction server v1.0 (http://dgpred.cbr.su.se) (Hessa et al., 2005; Hessa et al., 2007). We also calculated % helicity, adjusted for insertion by multiplying the % helicity (Equation 4.1) by the fraction of TM segment inserted (fins) (Equation 4.2).

% helicity = (# α-helical residues / total residues)*100% (4.1)

-[ΔGins/(R*298K)] -[ΔGins/(R*298K)] fins = e /(1 + e ) (4.2)

where R denotes the universal gas constant in kcal K−1 mol−1.

4.2.2 Peptide synthesis and purification.

Peptides were synthesized and purified as described in Chapter 2. Peptide concentrations were determined using based on UV absorbance at 280 nm in water using predicted molar extinction coefficients (Pace et al., 1995).

4.2.3 Biophysical characterization.

Freshly thawed aliquots of peptides were diluted to a final concentration of between 15 - 25 μM in 10 mM Tris HCl buffer (pH 8.0), with or without detergent. Detergent concentrations were chosen such that there was a minimum 1000-fold excess of detergent molecules to peptide molecules, as well a minimum of twice the critical micelle concentration. For SDS, DPC, and DDM, a detergent concentration of 20 mM was used; for OG, a concentration of 50 mM was used. Samples were incubated for a minimum of three hours prior to data acquisition. Further incubation was found to not affect either CD or fluorescence spectra. For Cys-containing peptides, spectra were also recorded in the presence of β-mercaptoethanol or dithiothreitol; these reducing agents were not found to significantly affect the magnitude or shape of the spectra other than increasing spectral noise, and as such were not included in further experiments.

CD and fluorescence spectra were recorded as described in Chapter 2. The wavelength of maximum Trp fluorescence emission intensity was obtained from these spectra, and the blue shift for each detergent was calculated as the difference between the average wavelength of maximum emission in aqueous buffer and the detergent-containing buffer.

4.2.4 Statistical analysis.

For each TM segment, the mean helicity in each detergent (measured by MRE at 222 nm -

[θ222nm]det.) was determined using a minimum of three replications. The overall mean helicity was then determined for each TM segment by taking the average helicity across detergents

( [θ222nm] TM). In order to compare the relative effects of detergents among different segments,

the helicity of each TM segment/detergent pairing was normalized to the mean helicity (in all detergents) for that TM segment (Equation 4.3).

(Δ[θ222nm]det., % ) = [(([θ222nm]det.) - ( [θ222nm] TM)) / ( [θ222nm] TM)] * 100% (4.3)

The normalized helical structure induction (Δ [θ222nm]det., %) indicates the relative amount of helical induction that occurs for each detergent. The average of this value across all TM segments ( Δ [θ222nm] det , %) was then calculated for each detergent. The same calculations were

performed using blue shifts in place of helicity to determine the average, normalized blue shift of

each detergent ( Δ λ det , %). All comparisons of biophysical data between TM segments or

detergents were made using two sided Student’s t-tests assuming equal variance using Excel (Microsoft). Linear regression (using the least squares method) was performed using Prism (GraphPad Software, Inc.).

4.3 Results. 4.3.1 TM structure classification.

Formation of TM protein structure within a lipid bilayer requires a complex network of interactions that include not only helix-helix contacts but also the formation of surfaces exposed to the surrounding environment – in many cases the hydrophobic lipid environment, but in some cases an aqueous environment (Tulumello and Deber, 2011a). In order to systematically investigate how such complex interactions might be maintained in detergent environments, we began by classifying the individual residues within proteins for which high-resolution structures are available by a novel scheme wherein a given amino acid within a TM segment is placed in one of three categories based upon its primary local environment: (i) exposed to an adjacent protein surface (viz., a helix-helix interaction); (ii) exposed to membrane lipids (the exterior surface of the protein); or (iii) exposed to an aqueous environment (such as a channel or pore of a transport protein, or a ligand or cofactor binding pocket) (Fig. 4.1A). Class (iii) is most prevalent in small molecule transporter proteins as they often have large water-filled cavities that allow for substrate access to the interior of the membrane bilayer (Fig. 4.1A). Based upon the results of this classification scheme, we then synthesized several representative TM segments from these membrane transporter proteins, which have a broad range of hydrophobicities, secondary structures, and interactive surfaces within their native structures (Fig. 4.1B; sequences in Table 4.1). Choices of the segments were further narrowed by our requirement for the inclusion of a native Trp residue as a fluorescent probe. To augment our analysis of these native segments, we performed parallel sets of experiments on two de novo designed segments (AI5 and AI10) as “controls” with contrasting hydrophobicities that have previously been shown to differ in their interactions with SDS (Tulumello and Deber, 2009).

Figure 4.1. Classification of local residue environment in TM transporter proteins. A) Schematic view of transmembrane regions showing the external surface (side view – top panel), as well as the any internal water filled cavities (open view – bottom). TM regions are shown in a space- filling representation and colored according to local environment; residues in contact with water are rendered in blue; those in contact with lipid are rendered in green; and those involved in protein-protein contacts are rendered in yellow (see materials and methods). Non-membranous regions are shown as cartoon representations in gray. For vSGLT, helix 1 is also shown in gray, as side chain atoms were not defined in the structure. Boundaries of the bilayer core are shown in black. In the open view, for clarity proteins are shown with two or more helices removed. For vSGLT, helices 4, 5 and 10 have been removed. For BtuCD, helices 3, 4 and 5 from one subunit and helices 9,10 of another have been removed. For LacY, helices 8 and 10 have been removed. For AdiC, helices 1, 5 and 8 have been removed. PDB IDs of the structures, along with the detergents used for solubilization/crystallization are given in Table 4.1. B) Surface representations of TM segments chosen for characterization. Surfaces are colored according to local environment as shown in the diagram. A cartoon representation of the backbone and a stick rendering of the Trp residue are outlined in black for each segment.

Table 4.1. Transmembrane segment analysis and peptide sequences.

ΔG % rel. PDB helix ins % exposure Protein ID # Sequence (kcal/mol) helical (Trp) Native TM segments LacYa 2V8N 3 KKKKK-Y75LLW78IITGMLVMFAPFFIFIFG96-KKKKK -2.38 46 32* KKKKK-I129LAVFW134ISLYIFVNLTSVLYLGGLAL154- vSGLTb 3DH4 4 KKKKK -1.69 92 0**

BtuCDc 1LV7 10 KK-I305GVVTATLGAPVFIW319LLLKA324-KKK 0.29 95 53* LacYa 2V8N 5 KKK-A143RMFGCVGW151ALCASIVGIMF162-KK 0.86 60 16**

AdiCd 3HQK 6 KK-I193QSTLNVTLW202SFIGVESAS211-KK 4.92 79 17** Model TM segments

AI5 ------KK-YAAAIAAIAWAIAAIAAAIAA-KKK -0.55 ------

AI10 ------KKKKK-FAIAIAIIAWAIAIIAIAIAI-KKKKK -2.58 ------

aLactose permease. Solubilized in DDM, crystallized with addition of β-D-galactopyranosyl 1- thio-β-D-galactopyranoside (Guan et al., 2007). bVibrio parahaemolyticus sodium/galactose symporter. Solubilized in n-decyl-β-D-maltoside (DM), crystallized with addition of Anzergent 3-12 (Faham et al., 2008). cBtuCD, vitamin B12 transporter. Solubilized in dodecyl-N,N-dimethylamineoxide (Locher et al., 2002). dAdiC, arginine/agmatine antiporters. Solubilized in DM (Fang et al., 2009). *Exposure to the external environment (lipid exposure). **Exposure to an internal cavity (aqueous exposure).

4.3.2 Adoption of native secondary structure.

Peptides were first assessed for formation – and extent - of helical structure in a variety of detergents using circular dichroism (CD) spectroscopy. We investigated each peptide in water as well as in four acyl chain detergents with varying head group compositions [anionic (SDS); zwitterionic, n-dodecyl phosphatidylcholine (DPC); and neutral, n-dodecyl-β-D-maltoside (DDM)] and tail lengths [8-carbon, n-octyl-β-D-glucoside (OG); and 12-carbon, (SDS, DPC, and DDM)]. The CD spectra, grouped by detergent, are shown in Fig. 4.2A; the same spectra grouped by TM segment are shown in the supporting information (Appendix 3, Fig. A3.1). in aqueous buffer, peptides typically adopted a ‘random coil’ structure indicated by a minimum near 200 nm, (Fig. 4.2A, left panel). The addition of detergent induced various degrees of helical structure (as evidenced by minima at 222 and 208 nm) depending upon the segment (Fig.4.2A).

Certain peptides (notably AdiC TM6 and LacY TM5) remained largely ‘random coil’ in all detergents (Figure 4.2A, red and light green curves, respectively). Perhaps surprisingly, even in SDS - an inducer of helical structure in water-soluble proteins (Imamura, 2006) - there is only minor helix formation in these latter TM segments.

Figure 4.2. Circular dichroism spectra of native and designed transmembrane peptides. A) CD spectra of TM segments in aqueous buffer and in the presence of various detergents as indicated in the diagram. B) Normalized helical structure induction in each detergent averaged across all TM segments (see materials and methods). C) Level of helix formation for each segment, averaged across all detergents. Model TM segments are shown in gold; TM segments that, in part, line an aqueous cavity, are shown in blue; and the remaining TM segment, which is primarily on the external surface of the protein, is shown in green. Error bars represent one standard deviation.

In general, for a given TM segment, all detergents produced similar degrees of helical induction (as assessed by the minimum at 222 nm (Fig. A3.2)), irrespective of head group composition or tail length. In order to determine if certain detergents were more efficient helix promoters, we calculated the level of helical structure induction for each TM segment in each detergent, normalized to the average for the particular segment (see materials and methods). When

79 detergents were compared in this manner, no detergent was found to produce an average helical induction significantly different from any other (Fig. 4.2B).

It should be noted that although we do not observe any statistical significance, this is in part due to the high variance (and thus large error bars) in the current dataset. The inclusion of a greater number of TM segments may result in the detection of some statistically significant differences between detergents. Nevertheless, the differences in helical induction between detergents are relatively small (at most a 10% deviation from the mean) indicating a modest effect of detergent identity on TM segment secondary structure adoption.

However, when averaged across all detergents, individual TM segments were found to have secondary structures that differed from one another (Fig. 4.2C). One interesting trend is that the TM segments with low hydrophobicity (indicated by a positive predicted free energy of insertion) adopt relatively little α-helical structure. This finding was in contrast to the other TM segments studied, which remained in a largely α-helical conformation in all detergent environments notwithstanding their excision from the remainder of the protein. In instances where two TM segments have similar hydrophobicity, the segment that is less helical in the native structure also appears as less helical in detergents. For example LacY TM3 is less helical than vSGLT TM4 in both detergents (p = 0.05) as well as in the native structure (46% versus 92% helical content in the corresponding native structure, respectively; Table 4.1). Likewise, BtuCD is largely helical in the native structure despite containing a central Pro residue, and is able to adopt greater helix formation in detergents compared to the remaining low hydrophobicity segments that contain non-helical regions in the native structure (LacY TM5 and AdiC TM6).

As the predicted insertion propensity of these segments varies with hydrophobicity, we adjusted the expected levels of helicity in the native structure for the fraction that is predicted to be inserted independently of the remainder of the protein. In doing so, we found a strong correlation between the helical content of each individual TM segment in detergents with its helicity level as it exists in the full-length protein (Fig.4.3; R2 = 0.836, p < 0.001). This relationship demonstrates that helix adoption in both proteins and peptides is contingent upon complete insertion into a hydrophobic environment, and as such is responsive to low sequence hydrophobicity.

It should be noted that additional crystal structures are available for each of the TM proteins in Fig. 4.1 that represent different quaternary compositions, contain single mutations, and/or have been obtained under various conditions (for example, in the absence or presence of a bound ligand). Using the analysis described above, we found that in some instances there are slightly different levels of helicity among the TM segments of interest, indicating nuanced differences in the corresponding TM domain conformation. These additional points are denoted by crosses in Fig. 4.3. Analysis of the average helical structure among all structures for each TM segment was found to give a significant correlation (R2 = 0.843, p < 0.001; data not shown).

Figure 4.3. Correlation of secondary structure of transmembrane segment peptides in detergents vs. their secondary structure in the intact protein. Mean residue ellipticity (MRE) is reported for each TM segment in each of the four detergents individually; some points in the diagram are not visible due to superimposition. Adjusted helicity values of each TM segment (see Table 4.1) are shown as solid circles; adjusted helicity values calculated from alternative conformation structures are shown as crosses. The alternative structures plotted are: for AdiC, PDB ID: 3LRB, 3L1L, and 3OB6; for lactose permease, 2CFQ and 2CFP; for BtuCD 2QI9; and for vSGLT 2XQ2. Error bars represent one standard deviation from the mean of a TM segment averaged across all detergents. Linear regression was performed using GraphPad Prism. The line of best fit for the circle points is shown as a solid line, and the 95% confidence intervals are shown as dashed lines.

4.3.3 Interaction of TM segments with detergent micelles monitored by Trp fluorescence.

The presence of a central tryptophan residue allows us to confirm the interaction of TM segments with detergent micelles. Peptides generally displayed a fluorescence maximum in water between 346 and 350 nm, consistent with complete exposure of the Trp residue to an aqueous environment (Fig. 4.4A, left panel); vSGLT had a fluorescence maximum in water of 341 +/- 2 nm, likely due to some secondary structure formation as observed in the CD spectra (Fig. 4.2A). The fluorescence spectra, grouped by detergent, are shown in Fig. 4A; spectra grouped by TM segment are shown in the supporting information, Fig. A3.3. In all detergents, there was an observable blue shift in the maximum of the fluorescence emission spectra for each of the peptides studied. These blue shifts (typically accompanied by an increase in emission intensity) are consistent with association of the Trp residue with detergent micelles, even in those cases where there is otherwise little change in secondary structure.

As with the CD studies, we used fluorescence spectroscopy to determine if there were any average differences in the interactions of various detergents with the TM segments. Again we found no distinction among detergents, as no detergent produced a normalized blue shift that was different than any other when compared across all segments (Fig. 4.4B). While blue shifts produced by specific TM segment/detergent pairings varied to some degree (Fig. A3.4), the only segment investigated that had a statistically different average blue shift, across all detergents, was the synthetic model TM peptide AI10. For all other segments, despite the exposure of Trp residues to varying environments within the native structure (Fig. 4.1B), there was no significant difference in the average blue shift (12-16 nm across all detergents) among the five individual native TM segments (Fig. 4.4C) – including those where the Trp residue in the native protein segment is classed as aqueous-exposed. In contrast, the AI10 segment has a very large blue shift in all detergents of between 28 to 32 nm, typical of its elevated hydrophobicity, as well as its self-association propensity in detergents (Tulumello and Deber, 2009).

Figure 4.4. Fluorescence spectroscopy of native and designed transmembrane peptides. A) Normalized fluorescence emission spectra of TM segments in aqueous buffer and in the presence of various detergents as indicated in the diagram. B) Normalized blue shift for each detergent averaged across all TM segments (see materials and methods). C) Blue shifts of each segment, averaged across all detergents. Model TM segments are shown in gray, TM segments for which the Trp residue lines an aqueous cavity in the native structure are shown in blue, segments where the Trp residue is exposed to lipid in the native structure are shown in green, and the segment where the Trp residue is buried within the protein is shown in gold.

4.4 Discussion. 4.4.1 Transmembrane segment structure in intact proteins vs. TM peptides.

Helix formation in TM segments is typically coupled with bilayer insertion and is dependent upon hydrophobicity (Liu and Deber, 1997; White and Wimley, 1999). Of the native TM segments investigated here, we note that some are predicted to be of insufficient hydrophobicity to insert independently into native bilayers (Table 4.1). In native bilayers, the translocon- mediated insertion of such segments has been hypothesized to be dependent on the remainder of the protein (Hedin et al., 2010; Kauko et al., 2010). In detergents, this situation manifests as a

partially helical state that is likely a micelle surface-interactive species, as has been observed in previous studies (Tulumello and Deber, 2009; de Foresta et al., 2010; Lawrie et al., 2010; de Foresta et al., 2011; Tulumello and Deber, 2011b). The correlation observed in Fig. 4.3 indicates that the overall secondary structures of each TM segment formed in both the protein and the detergent-solvated state are similar, and are therefore an intrinsic property of the protein sequence that is preserved in hydrophobic environments. The present results thus suggest that the two features of the peptides studied here that are the fundamental contributors to their level of helix adoption in detergents are sequence hydrophobicity and native secondary structure.

There is one outlier in the aforementioned correlation: LacY TM3 has a greater helical structure in detergents than is predicted by the observed native structure. While LacY has been shown to undergo a large conformational change during sugar translocation, all structures obtained to-date represent an inward facing conformation of this molecule (Kaback et al., 2011). In particular, TM3 has been observed to move with respect to the opposite side of the protein in the alternate conformation (Majumdar et al., 2007; Smirnova et al., 2007), and has been implicated as having a role in the transport process (Sahin-Toth et al., 1994). In this case, the conformation of the LacY TM3 segment in detergents may not correspond to that observed in the crystal structure, but to another native conformation.

4.4.2 Are detergents denaturants of membrane proteins?

Detergents are often qualitatively described as ‘harsh’ or ‘mild’ based upon their relative propensity to denature proteins. In some protein folding studies, solubilization by ‘harsh’ detergents has been used to denature the tertiary structure of multi-spanning membrane proteins.(Booth and Curnow, 2006; Renthal, 2006; Joh et al., 2008) Properties of detergents that render them denaturing include head group charge (a charged head group is more denaturing than a zwitterionic one which, in turn, is more denaturing than a neutral one), as well as acyl tail length (a shorter chain is more denaturing than a longer one) (Privé, 2007). In the present work we have observed that both ‘harsh’ (SDS and DPC) and ‘mild’ (DDM and OG) detergents interact with individual TM segments in a similar manner. This tolerance of TM structure in an assortment of hydrophobic environments may be an intrinsic property of TM segments which governs their stability in varying lipid environments (Sanders and Mittendorf, 2011). The

denaturing properties of ‘harsh’ detergents may thus be due largely to interactions with non- membranous regions of proteins where charged and zwitterionic detergent head groups provide an alternative mode of interaction with non-hydrophobic residues. Accordingly, ‘harsh’ detergents have proven to be a suitable mimetic for TM constructs with no or very limited extramembranous regions, such as those typically investigated by NMR (Page et al., 2006; Warschawski et al., 2011). While the necessary inclusion of non-hydrophobic residues in TM segments derived from transporter proteins forces these segments to sacrifice overall hydrophobicity – and, in part, stable helical insertion - for function, it has been noted that such proteins are particularly susceptible to denaturation in all but the mildest detergents (such as DDM) (Newstead et al., 2008). This situation is likely recapitulated when the corresponding TM peptides are studied in detergents, as elimination of ‘passenger’ tertiary contacts may tend to mitigate against their micelle insertion and secondary structure formation. Examples of this circumstance are the micelle complexes of LacY TM5 and AdiC TM6 that exist in a partially- inserted state. For these segments, the observed blue shifts confirm a primarily non-native state, where the central Trp residues are not water-exposed.

The characterization of individual TM segment structure in detergents may therefore represent a glimpse into the “unfolded” state of membrane proteins, where tertiary contacts between different helices are yet to form. As such, unfolding studies of membrane proteins in detergents likely probe a subset of the forces stabilizing membrane protein structure, i.e., for some TM segments only tertiary interactions are lost, while in instances where bilayer insertion is weaker due to low hydrophobicity, secondary structure of TM segments may also be lost. The parallelism of the TM segment structures observed in the native proteins vs. the corresponding TM segment peptides is perhaps not unexpected, given that the structures in each instance have, in effect, been obtained in detergent environments. Nevertheless, the consistency of structural features of TM segments observed here in a variety of micellar media does suggest that they likely correspond to the relevant biological forms that occur in each protein in its native lipid bilayer environment.

Chapter 5. De novo designed hydrophobic “hairpins” as models of membrane protein folding.

The work described in this Chapter has been aided by Rachel M. Johnson who contributed toward the initial project design. Inga Isupov contributed to the production of protein hairpins.

5.1 Introduction.

Single TM helices have been used to investigate many aspects of membrane protein folding, including threshold of insertion into membranes (Liu et al., 1996; Hessa et al., 2005), and TM helix-helix association (Gratkowski et al., 2001; Zhou et al., 2001; Rath et al., 2009b). Results from these experiments have proven to be valuable in the understanding of tertiary and quaternary folds of full-length proteins, as well in explaining in vivo observations of folding and trafficking. The ability of individual TM segments to assess membrane protein folding as a whole is a key tenet of the two stage model (Popot and Engelman, 1990, 2000). This “divide and conquer” approach to investigating tertiary membrane protein structure is further supported by experiments where multi-spanning membrane proteins are fragmented into smaller groupings of TM segments. Loop-severed or –deleted fragments of various α-helical membrane proteins such as bacteriorhodopsin (Liao et al., 1983) the muscarinic acetylcholine receptor (Schoneberg et al., 1995) and lactose permease (Zen et al., 1994) have all been shown to be capable of reassembling into native or near-native conformations.

Although investigation of membrane protein folding using individual TM helices has proven quite useful, single helices have the fundamental limitation of being unable to address the role of interconnecting loops in the folding of multi-spanning TM segments. As well, in fragmentation experiments of such proteins it is not always possible to recover full wild-type stability and/or function (Kahn et al., 1992; Allen et al., 2001). This demonstrates that in certain cases, the association of TM segments may also be governed by the interconnecting loop regions. This may be due to restriction of non-native interactions or reduction of the entropic penalty for association thereby allowing for weaker interactions to persist. Thus, in order to gain additional insights into membrane protein folding, it is necessary to consider helix-helix interactions in conjunction with the effect of the intervening loop regions. In order to address these questions, our lab and others have utilized membrane protein constructs consisting of two hydrophobic helical segments connected by a soluble loop, which we term ‘helical hairpins’, as a minimal model of TM segment helix-helix tertiary contacts (Peng et al., 1998; Johnson et al., 2004; Neumoin et al., 2009).

Partial or complete randomization of protein sequence has previously been used to generate libraries of de novo proteins (Kamtekar et al., 1993; Davidson and Sauer, 1994; Hecht et al.,

2004; Bradley et al., 2006). In such libraries, members have been identified which adopt a native-like folded state (Davidson et al., 1995; Wei et al., 2003), and/or display some functional activity (Wei and Hecht, 2004; Seelig and Szostak, 2007; Fisher et al., 2011). Studies of such un- evolved proteins have contributed greatly to our understanding of the evolution of native protein folding and structure (Schafmeister et al., 1997; Go et al., 2008; Smith and Hecht, 2011). In the present chapter we apply our expertise in the production and characterization of membrane proteins toward the creation of a library of de novo model TM helical hairpins. In our design, six amino acid residues on each helix of library members have been randomized, allowing us to explore a great diversity of hairpin sequences. In the initial work presented here, we focus on the development of expression and purification protocols that will allow for the production of milligram (mg) quantities of these hairpins. We have also performed preliminary structural characterization of several members of this library in SDS.

5.2 Materials and methods. 5.2.1 Library creation.

A library of expression plasmids containing the randomized hairpin DNA sequence was produced using two sets of partially complementary oligonucleotides (shown in Appendix 4). Each set represents one half of the final desired hairpin sequence. Each set of oligonucleotides was heated at 70 °C for 5 minutes and annealed for 5 minutes at a temperature 5 °C lower than the melting temperature of the complementary regions. The strands were extended using Sequenase DNA polymerase (Amersham) at 37 °C for 30 min. Following extension, the double stranded DNA was cleaved at both the 5’ and 3’ ends using appropriate restriction enzymes (New England Biolabs) for 2 h. The expression vector, pET32a (Novagen), was also cleaved for 2 h using the restriction enzymes NcoI and XhoI. In order to remove 3’ OH groups of the expression vector, cow intestine phosphatase was added for the final hour of this reaction. The cleaved vector and both halves of the insert were further purified using a Qiagen PCR Purification or Gel Extraction kit, and quantified by comparison to standards on an agarose gel following staining with ethidium bromide. Ligations were performed overnight, at room temperature, in a total volume of 10 μL using a molar ratio of 3:1 for the insert and vector respectively. 50 μL of XL1-Blue Supercompetent cells (Stratagene) were transformed with 2 μL

of this ligation reaction, and diluted into 1 mL of Luria broth (LB), according to the manufacturer’s protocol. 10-50 μL of this mixture was spread onto ampicillin-containing LB agar plates (ampicillin was present in all cell media at a final concentration of 100 μg/mL, unless otherwise specified), and library size was estimated by counting the resulting number of colonies. The remaining transformation was used to inoculate 6 mL of LB, grown overnight and plasmid DNA was extracted using a Qiagen miniprep kit.

The initial plasmid DNA library was further treated with the HindII restriction endonuclease in order to cleave any pET32a vector which did not contain the insert. The process of transformation, library size estimation and DNA extraction was then repeated on this sample in order to produce a final plasmid DNA library (library I). The additional HindII treatment step was omitted during the production of the second plasmid DNA library (library II) in order to produce a larger library size.

5.2.2 Colony expression immunoscreening.

1-2 μL of a 1:100 dilution of the plasmid library was used to transform BL21(DE3) cells (Novagen) according to manufacturer’s protocol. Transformation mixtures were spread onto sterile nitrocellulose filters on ampicillin-containing LB agar plates and grown at 37 °C. After 8- 10 hours, colonies were replicated onto additional filters and then grown overnight at 28 °C. Replicated filters were then transferred to fresh ampicillin containing LB agar plates that also had 100 μL of 0.1M IPTG spread onto them. Colonies were induced 3-4 hours, and lysed by exposure to chloroform for 15 minutes as well as a 3 h incubation in lysis buffer (100 mM Tris-

HCl pH=7.8, 150 mM NaCl, 5 mM MgCl2, 1.5% (w/v) BSA, 40 µg/mL lysozyme, 1 µg/mL DNAse). Filters were then washed, blocked, incubated with anti-C-Term 6xHis–HRP conjugated antibody (Invitrogen) and washed once more using buffers and incubation periods according to the manufacturer’s protocol. Detection of the antibody was achieved using ECL reagent (Amersham Biosciences). Colonies on the master plates corresponding to those colonies that produced the antibody epitope were used to inoculate 6 mL of LB containing ampicillin, and grown overnight at 37 °C with shaking. The overnight cultures were subsequently used for plasmid DNA extraction and DNA sequencing was performed by ACGT Inc. (Toronto).

5.2.3 Protein expression.

DNA plasmids of selected sequences were used to transform BL21(DE3), or BL21(DE3) Codon Plus cells (Novagen) according to the manufacturer’s protocol. Colonies resulting from this transformation were used to inoculate ~ 5 mL of either terrific broth (TB), LB, or minimal media containing ampicillin, and grown overnight at 37 °C. Overnight cultures of library members were used to inoculate the corresponding broth that was then grown at 37 °C with shaking to OD600 ~0.6. Upon reaching this density, cells were induced by adding IPTG to a final concentration of either 0.01, 0.1 mM or 1 mM. Cultures were incubated, with shaking, at 250 RPM, either overnight at 25 °C or between 1-3 hours at 37 °C. Cultures were then centrifuged at 8,000 RPM for 25 min, using a JA10 rotor in a Beckman J2-M1 floor centrifuge, and pellets were frozen and stored at - 20 °C.

5.2.4 Protein extraction and purification under denaturing conditions.

Cell pellets corresponding to 250 or 500 mL of induced cell culture were resuspended and lysed with 4 mL of lysis/wash buffer (6 M urea, 2% SDS, 10 mM imidazole, 10 mM Tris.HCl, pH=8.0) per 1 g of cell pellet. The cell lysate was then centrifuged at 9,000 RPM for 20 min using a JA-20 rotor. Following centrifugation, the resulting supernatant was purified using standard nickel affinity chromatography techniques using Ni-NTA ProBond resin according to manufacturer’s protocols (Invitrogen) using elution buffer (6 M urea, 2% Triton X-100, 200 mM imidazole, 10 mM Tris.HCl, pH=8.0). Three rounds of dialysis (a minimum of 8 h each), exchanging into a cleavage buffer (1% Triton X-100, 10 mM Tris.HCl, pH=8.0) were performed, followed by the addition of 1 µL of thrombin (Invitrogen) per mL of cell lysate, in order to cleave the solubility tag.

Detergent was removed from the protein by incubation with BioBeads (BioRad) in the presence of 20% acetonitrile (ACN) for 2 h, using the recommended detergent/bead ratio. Hairpins were further purified by reverse phase HPLC, using a Phenomenex Jupiter C4 300Ǻ semiprep column, with either a water/ACN or water/isopropanol gradient. All hairpins were lyophilized following HPLC and stored at -20 °C.

5.2.5 Biophysical characterization.

Prior to use, lyophilized samples were redissolved in 20 mM Tris.HCl pH = 8.0, with or without 2% (w/v) SDS. Concentrations of individual aliquots were determined using the BCA assay (Pierce), with standards prepared with an identical buffer. Samples were diluted to final concentrations of between 10-20 μM for circular dichroism (CD) and between 2-5 μM for Trp fluorescence. In some cases samples used for CD were diluted for subsequent use in Trp fluorescence measurements. Both CD and fluorescent spectra were obtained as described in Chapter 2. Trp fluorescence was normalized for the total concentration of the Trp residues present.

5.2.6 SDS-PAGE migration rate

SDS-polyacrylamide gel electrophoresis (PAGE) was performed using precast 4-12% acrylamide NuPAGE gels in MES buffer (Invitrogen) according to manufacturer protocols. Coomassie was used to visualize proteins on all gels. 1 µg of each peptide was loaded in a total volume of 18 µL to each lane. Apparent molecular weights were estimated based upon Rf analysis using Mark12 molecular standards (Invitrogen). Migration distances were determined using NIH Image. The increase in apparent MW (versus the formula MW) for all hairpins was calculated as: (1 - [estimated MW from gel / formula MW from sequence]) * 100 %.

5.3 Results. 5.3.1 De novo membrane protein hairpin library design and production.

Our initial goal was to develop methods to efficiently produce a series of model TM hairpins with large sequence diversity. Such a library of de novo designed proteins provides a rich source of un-evolved hydrophobic sequences, from which the fundamental requirements of TM protein folding may be addressed. The base sequence of this library was derived from a de novo designed hairpin previously produced in our lab, termed AIF, which was shown to fold into the desired tertiary structure (Johnson et al., 2004). In the present work, subsequent revision of the primary sequence of this hairpin has been made in which five Ile residues in each TM segment

have been replaced with Ala, and two Phe residues that flank each helix have been replaced with Tyr. These modifications reduce the hydrophobicity of the construct and allow for increased levels of expression. This second generation hairpin sequence has been termed AIY, and is shown along with AIF in Table 5.1.

In an initial exploratory library, we chose to randomize six residues on each model TM helix of the AIY sequence. These residues are located on the proposed interface between the two helices as determined for the first generation AIF hairpin sequence (Johnson et al., 2004). Randomization at these positions thus allows for the investigation of the sequence requirements of a tightly packed helix-helix interface. Shown in Table 5.1 are the designed sequences of members of the hairpin library (termed AIr), where the X-positions denote the location of randomized residues in individual library members. By randomizing these positions in the DNA sequence to any of the 64 codons in the genetic code, hairpin sequences are produced that may contain both hydrophobic and polar amino acids residues. The inclusion of polar residues such as Asp, Glu, Gly, His, Lys, Asn, Pro, Gln, Arg, Ser, and/or Thr, is essential in the production of native-like sequences as collectively these amino acids comprise ~30% of residues found in TM segments (Cunningham et al., 2009).

Table 5.1. Sequences of de novo designed hairpins.

First generation de novo hairpin (AIF)

KKKKKKK-FAIAIAIIAWAIAIIAIAIAI-KSPGSK-IAIAIAIIAIAWAIIAIAIAF-KKKKKKK Second generation de novo hairpin (AIY)

KKKKK-YAAAIAAIAWAIAAIAAAIAA-KSPGSK-AAIAAAIAAIAWAIAAIAAAY-KKKKK Randomized de novo hairpin library (AIr)

KKKKK-YAAXIAXIAWXIAXXAAXIAA-KSPGSK-AAIXAAXXAIXWAIXAIXAAY-KKKKK

For the purpose of protein production, putative hairpin sequences, with flanking Lys tags (included to enhance solubility), were ligated into the pET32a bacterial expression vector (Novagen). This plasmid expresses each hairpin as a fusion protein, as shown in Fig. 5.1. DNA libraries were produced on two separate occasions, the first with an approximate size of ~10,000

DNA sequences, and the second with an approximate size of ~3,000. The quality of the initial DNA library was assessed by randomly selecting 10 plasmids for sequencing. Of these 10 sequences, two did not contain a full insert, four contained frame-shifts, and two did not sequence properly. Of the remaining two that contained an error-free insert, one contained a stop codon. This is expected as the odds of incorporating a stop codon at one of the 12 randomized positions is roughly 50%.

Figure 5.1. Construct for bacterial expression of model TM hairpins as fusion proteins. The expression construct is derived from the pET32a vector, and includes a thioredoxin domain (Trx), two 6x His tags, and an S-tag. AIr library member sequences are inserted into the multiple cloning site and are comprised of two TM helices linked by a soluble loop, with two flanking 5x Lys solubility tags (sequence in Table 5.1). The thioredoxin domain as well as one of the His tags are subsequently removed by cleavage with thrombin at the location indicated.

In order to improve the quality of sequences selected from this library, we employed a colony immuno-screening procedure. In the expression vector, a C-terminal 6x His tag is produced only upon insertion of a sequence that results in a specific frame-shift. Thus, by using an antibody specific to this epitope, we are able to select from the library only those plasmids that contain inserted DNA sequences free of stop codons and frame-shift errors. Approximately 1000 colonies in total were screened in this manner, of which ~20% had positive immunogenicity toward the antibody. From this screening, 52 colonies that produced the greatest level of epitope production were selected for DNA sequencing. With a few exceptions, these sequences did indeed contain a C-terminal 6x His tag, although most did not fully correspond to the designed sequence. Many of the plasmids obtained in this manner contained only half or less of the full

hairpin sequence. Several contained multiple frame-shifts, resulting in large portions of the protein sequence differing from the design.

Of the 52 sequences obtained by screening, there were nine unique sequences, with fewer than three residues that differed at locations not intentionally randomized, that represented the original design. These sequences are summarized below in Table 5.2, along with the sequence obtained without screening (AIr 10). For each sequence, putative TM domains were predicted using TM Finder (based on experimentally derived hydrophobicity and helicity parameters – available online at http://tmfinder.research.sickkids.ca/) (Liu and Deber, 1998a; Deber et al., 2001) or the ΔG prediction server v1.0 (based upon translocon-mediated free energy of insertion – available online at http://dgpred.cbr.su.se/ ) (Hessa et al., 2005; Hessa et al., 2007). Using these programs, AIr sequences were predicted to have 0, 1 or 2 putative TM sequences. Prediction of the number of TM segments in each sequence was in good agreement between these two methods, with only three regions, among all hairpins, classified differently. Those hairpins that are predicted to have two TM segments are hereafter referred to as have putative “dual TM” toplology.

Table 5.2. Sequences obtained from library for further studies.

# predicted TMs Liu 3 1 2 Wild Type Sequence Deber ΔGins

AIY* KKKKKYAAAIAAIAWAIAAIAAAIAAKSPGSKAAIAAAIAAIAWAIAAIAAAYKKKKK 2 2 3 Sequence obtained with no screening AIr 10 KKKKKYAAVIALIAWNIAETAAVIAAKSTGSKAAISAACVAISWAIPAIKAAYKKKKK 2 2 Sequences obtained by immuno-screening3

AIr 11* KKKKKYAACIAPIAWNIARPAAPIAAKSPGSKAAISAADLSIPWVISAIPAAYKKKKK 1 0 AIr 12* KKKKKYAAGIAEIAWNIARNAARIAAKSPGSKAAILAALVAILWAILAIMAAYKKKKK 1 1 AIr 14* KKKKKYSAAIALL~RHIASGAAAIAAKSPGSKAAILAALSAIVWAINAIKAAYKKKKK 1 1 AIr 25 KKKKKYAASIAYIAWTIAGKAARIAAKSPGSKAAIFAANVAIPWAIVAIPAAYKKKKK 1 2 AIr 29 KKKKKYAASIAGIAWMIALMAAAIAAKSPGSKAAIAAASFAIRWAILAIIAAYKKKKK 2 2 AIr 31* KKKKKYAALTASIAWSIARRAANIAAKSPGSKAAIKAARTAIGWAIKAIKAAYKKKKK 0 1 AIr 36 KTKKKYAASIAIIAWLIATTAAKIAAKSPGSKAAIAAACNAILWAINAIAAAYKKKKK 2 2 AIr 49 KKKKKYAAPIAGIAWAIASEAARIAAKSPGSKAAIQAAAIAIRWAITAIQAAYKKKKK 2 2 AIr 54* KKKKKYAALIAVIAWVIAVKAANIAAKSPGSKAAIVAALMAIFWAIMAIEAAYKKKKK 2 2 1Regions predicted to be TM spanning by TM Finder, using default settings, are highlighted in grey. 2Regions predicted to be TM spanning by ΔG prediction server v1.0, using default settings, are italicized and bolded. 3Locations of positions that have been randomized are underlined. *Sequence was produced in mg quantities.

5.3.2 Protein expression, and purification.

In total, eleven sequences were obtained (including the base AIY sequence) for which protein expression optimization trials were undertaken. In these trials, we modified the following parameters: cell line, growth media, IPTG concentration, induction temperature, and induction time. Of these eleven sequences, six were found to express in amounts on the order of 1 mg/L of bacterial cell culture, an amount necessary for detailed biophysical characterization. Overall, the most successful conditions were growth using BL21 (DE3) cells, in terrific broth (TB) media, induction with 1 mM IPTG, and a 3 hr induction period at 37 °C. Among these six sequences that could be produced in sufficient quantities, two were predicted to contain two putative TM

95 segments, and thus expected to have the ideal dual TM hairpin structure (AIY and AIr 54). Further studies were performed on these sequences as well as those with zero or one predicted TM domain(s) for comparison (AIr 11, 12, 14, 31).

The expression conditions that were most successful resulted in a large portion of the expressed protein being insoluble when cells were lysed using sonication. Purification was thus achieved by using chemical denaturation (2% SDS, 6 M urea) to lyse the cells and solubilize all components. The hairpin constructs were cleaved from the thioredoxin domain and purified via Ni-NTA affinity chromatography and reverse-phase HPLC through a series of steps, as outlined in Fig. 5.2. This procedure was optimized for all hairpins, and required both the exchange, and latter removal of detergent. The only modification to the procedure necessary to purify various hairpins was the choice of gradient for the final reverse-phase HPLC purification step. A gradient of either water-to-acetronitrile or water-to-isopropanol, was chosen on a case-by-case basis in order to prevent co-elution of the hairpin with the thioredoxin domain.

Figure 5.2. Purification scheme for randomized hairpin proteins. The major buffer component(s) required for hairpin solubilization are listed for each step. See materials and methods for details.

5.3.3 Adoption of α-helical structure.

In order to observe the adoption of the desired secondary structure in a membrane-like environment, circular dichroism (CD) spectroscopy was performed on the hairpins in both an aqueous buffer and one containing sodium dodecyl sulfate (SDS). The AIr 31 sequence was not found to be water soluble, and thus was investigated only in SDS. In aqueous buffer, most hairpins displayed spectra indicative of a primarily α-helical secondary structure (Fig. 5.3A). This is consistent with the high helical propensity of the base peptides (composed largely of Ala residues) according to the Chou–Fasman scale (Chou and Fasman, 1978). For AIr 11, a random coil conformation is present (evidenced by a minima at ~200 nm), in addition to some helical structure formation. For all hairpins, SDS was found to induce increased helical structure formation (Fig. 5.3B). As a measure of the responsiveness of these hairpins to SDS solubilization we normalized the level of helicity in SDS to the level of helicity in aqueous buffer for each hairpin. This gives the relative level of increased helical structure formation induced by SDS, and was found to vary among different hairpins (Fig. 5.3C). In general, those with dual TM topology had a larger increase in helical structure content (70- 90%) versus the others (15-35%). The absolute amount of secondary structure of these former segments is consistent with the designed hairpin structure.

Figure 5.3. Circular dichroism of hairpins. A) CD spectra in aqueous buffer (10 mM Tris.HCl pH=8.0). B) CD spectra in SDS-containing buffer (34 mM SDS, 10 mM Tris.HCl pH=8.0). Spectra obtained in SDS buffer are shown with solid lines while spectra obtained in aqueous buffer are shown with dashed lines. Spectra were converted to mean residue molar ellipticity (MRE - deg cm2 dmol-1). C) Helical induction upon SDS solubilization. Reported is the% increase in helical content upon addition of SDS measured as (MRE222nm in SDS / MRE222nm in water)*100. Error bars represent the combined standard deviation.

5.3.4 Fluorescence spectroscopy indicates a degree of amino acid residue burial in both water and SDS.

The tryptophan residue displays fluorescence that is highly sensitive to the polarity of the local environment, and when exposed to water will have a maximum fluorescence emission at roughly 350 nm (Lakowicz, 2006). For all hairpins, we observe that the Trp fluorescence emission maximum is between 325-340 nm in both SDS and aqueous environments, indicating that the Trp residues are in relatively apolar local environments in all hairpins (Fig. 5.4). In aqueous environments, this local hydrophobic environment is likely produced by some tertiary structure formation, allowing the central Trp residue(s) to be partial buried by surrounding hydrophobic residues. Upon addition of SDS there is an additional blue shift of the emission maximum for AIr 12, 14 and 54, indicating the further burial of these hairpins in the presence of SDS micelles. For other hairpins there is no observable blue shift (AIY), or even a decrease in emission maxima in SDS buffer compared to aqueous buffer (AIr 11). We note, however, that in all cases the Trp emission maximum in SDS is still consistent with insertion into a detergent micelle (Liu and Deber, 1998a; Tulumello and Deber, 2011a).

Figure 5.4. Fluorescence emission spectra of hairpins. Lyophilized samples of hairpins were dissolved in either an SDS buffer (8 mM SDS, 10 mM Tris.HCl pH=8.0) or aqueous buffer (10 mM Tris.HCl pH=8.0). Spectra are the average of 2 or more samples, excited at 295 nm. Spectra obtained in SDS buffer are shown in black while spectra obtained in aqueous buffer are shown in blue. Spectra were normalized to the concentration of each sample, as well as the number of Trp residues in the sequence. Vertical dashed blue lines (at 350 nm) indicate the typical emission of Trp in water.

5.3.5 Gel migration rate variance reveals conformational differences among hairpins.

Our lab has previously demonstrated that SDS-PAGE migration rates of membrane protein hairpin structures are, in part, sensitive to conformation (Therien et al., 2001; Wehbi et al., 2007). Migration rates are also responsive to the amount of SDS bound, where segments that bind greater amounts of SDS migrate slower on the gel (Rath et al., 2009a). As such,

100

measurements of SDS-PAGE migration rates were performed in order to provide insights into any differences in conformation or amount of detergent binding among these hairpins. Indeed, we observed that the SDS PAGE migration rates of the hairpins varied greatly from one another despite having MWs within ~10% of each other (Figure 5.5A). For AIr 11, self association occurred in the absence of reducing agent, indicating disulfide bond formation under oxidizing conditions. AIr 54, despite being pure, forms multiple species on SDS-PAGE gel indicating the presence of different conformational or oligomeric states.

Overall, the monomeric form of all hairpins migrated at a rate slower than that corresponding to their formula MW (Fig. 5.5B). One interesting observation is that the increase in apparent MW for AIY, as well as the lowest band of AIr 54, was smaller in comparison to most other hairpins. These two sequences are predicted to have two putative TM segments, and are the most hydrophobic among those studied. If amount of detergent bound was the sole determinate of gel migration, these two sequences should have the highest % increase in apparent MW rate, due to their increased hydrophobicity. Instead the presence of a more compact, folded conformation is also likely a contributing factor in the migrations rates of these hairpins.

Figure 5.5. SDS-PAGE migration of hairpins. A) 4-12% SDS-PAGE NuPAGE gels of all hairpin constructs. B) % MW increase of all hairpin species. The band corresponding to AIr 11 in the absence of reducing agent is calculated using the formula MW of the dimer. Error bars represent the SD of a minimum of three replicates.

101

5.4 Discussion. 5.4.1 Methods to improve de novo membrane protein design.

In the present work we have undertaken the production and characterization of a library of diverse model hairpin sequences. By investigating a broad set of artificial proteins we are able to determine the sequence characteristics that are most compatible with these goals, as well as optimize the methods used such that they are as broadly applicable as possible.

In the randomization pattern of the current design, the number of unique protein sequences possible is ~ 1 x 1015. In practice, based upon the amount of DNA used and the stated transformation efficiency of the cells, it would be possible to generate 1 x 106 unique sequences for every ligation reaction and subsequent transformation. We find, however, that library size is greatly limited at this step, as we produced 10,000 unique sequences in the first library, and 3,000 in the second library. For the purpose of initial characterization, we produced a sufficient number of library members; however, it would be possible to increase the size of the library through various means. In particular, the transformation step can be improved through use of commercially available competent cells with increased transformation efficiency, and the ligation reaction may be improved by employing multiple steps in which only two DNA fragments are joined at a time. Ligation performed in this manner should also reduce the number of partially inserted sequences, which represented roughly 25% of the unselected library.

Another consideration is the quality of sequences contained within the plasmid library. The number of frameshifts, and associated errors in the library members (around 50%), indicate that the purity of the DNA oligonucleotides used in library construction was a limiting factor in library quality. This is a common issue in such studies as sequences with frameshifts have been found to comprise up to 80% of similarly prepared DNA libraries (Kamtekar et al., 1993; Davidson and Sauer, 1994). Purity of the DNA primers is largely a function of primer length, and a redesign in which shorter DNA primers are used would likely reduce the number of such aberrant sequences. Furthermore, in vitro selection may be applied that would allow for the screening of a much larger initial library. This pre-screening step leads to a higher quality DNA library prior to transformation into bacteria (Cho et al., 2000).

102

Screening at the colony level served two purposes: (i) it selected for sequences free of frame shifts or stop codons in the DNA sequence, and (ii) it selected constructs that could be expressed in bacteria. While selection for the first criterion was successful, selection for the second criterion was not, as levels of expression from colonies did not scale up to larger volume cultures in many cases. Although library members are 90% identical to one another in sequence, the amount of protein produced in 1 L cultures varied considerably. We note that of the seven sequences that were predicted to have dual TM topology, only two were successfully expressed. This is in contrast to the less hydrophobic sequences, of which we were able to express all four. This indicates that hydrophobicity represents a major bottleneck in the expression of these artificial proteins, as it does for native membrane proteins (Tate, 2001; Cunningham and Deber, 2007). Another contributing factor to low expression levels is that non-folded structures are more prone to degradation than folded ones (Parsell and Sauer, 1989). This has previously been shown to limit the expression of unselected sequences from similar designed libraries, and may also be occurring to some degree here (Kamtekar et al., 1993; Davidson and Sauer, 1994).

One major success in the production of this library is that we were indeed able to develop a robust purification strategy that worked for all hairpins. As was the intention of the design, sequences features that were common among all mutants (general hydrophobicity of the common residues within the hairpin domain, as well as the expression and purification tags included in the fusion construct) allowed for optimization of a single purification scheme. With only a change of solvent in the final purification step, all hairpins that expressed at desired levels could be purified using a single protocol irrespective of the identity of the randomized residues. This is essential for future production of membrane proteins from designed libraries, as purification also typically represents a challenging task in TM protein characterization.

5.4.2 Structural features characteristic of dual transmembrane hairpin constructs.

The initial biophysical analysis performed herein demonstrates a difference between hairpins with or without dual TM topology, allowing us to determine features characteristic of each class. For those sequences with a single or no predicted TM segments (AIr 11, 12, 14, 31), the relatively high levels of helicity, as well as partially buried Trp residue in water, indicate some

103

structure formation in this medium. For these segments addition of SDS induces only moderate changes in both amount of helical structure as well tryptophan fluorescence spectra. Although the dual TM hairpin constructs (AIY and AIr 54) share many features with the less hydrophobic hairpins in water, differences in structural features are apparent in SDS micelles. One defining feature of the dual TM hairpins is a large increase in helical structure upon addition of SDS. The increased helical structure is likely localized to the putative TM region(s), as regions below the threshold for membrane insertion do not adopt increased helical structure in detergent micelle environments (Tulumello and Deber, 2012). This is consistent with the SDS-induced increase of helicity in hairpins with only a single putative TM being roughly half of that observed in the dual TM hairpins. We also note that the increase in helicity in the AIY hairpin (70%) is the same increase induced in a model, single TM, peptide with an identical sequence to the individual putative TM regions in this hairpin (Tulumello and Deber, 2011b).

While some hairpins in both classes exhibited a blue shift in Trp fluorescence emission wavelength in SDS compared to water, AIr 54 displayed a large increase in fluorescence intensity. This increase in intensity was also observed for the single TM-containing, AIr 14 hairpin. This latter hairpin however, has only a single Trp residue, which is located in the region predicted to be a TM segment. Together, these results indicate that an increase in fluorescence intensity upon detergent solubilization is also a defining feature of a TM segment. The other dual TM hairpin did not display this increase, but this was consistent with the results obtained for the aforementioned single TM model (Tulumello and Deber, 2009) and likely occurs due to aberrant local detergent interaction due to the “lipopathic” nature of this helix (Tulumello and Deber, 2011b). Overall we find that, upon SDS solubilization, increased helical structure formation, as well as an increase in Trp fluorescence intensity, occurs for hairpins with dual TM topology. These features may therefore serve as structural hallmarks to verify dual TM topology in hairpin sequences.

SDS-PAGE reveals further differences in tertiary structures among hairpins, as it is a technique sensitive to both detergent binding and conformation. Despite being both highly hydrophobic and containing a positive net charge - features that contribute to increase gel migration rate (Melnyk et al., 2003; Rath et al., 2010; Tulumello and Deber, 2011b) - the dual TM hairpins migrate similarly to their formula MWs. In a CFTR TM 3/4 hairpin fragment that has been well characterized by our lab, this phenomenon is also observed (Rath et al., 2009a). Subsequent

104

mutations to the loop region of this hairpin, including those that are of identical net hydrophobicity, have been found to result in an increased apparent MW (above the formula MW) when assessed by SDS PAGE (Wehbi et al., 2007). This is thought to occur through steric interactions in the loop that disrupt helix-helix packing. As such it seems reasonable that the single TM hairpins, despite being less hydrophobic, migrate at slower rates than the dual TM versions because they are unable to adopt such a compact conformation in SDS, as they will lack the necessary helix-helix contacts. Thus, the faster gel migration rate of the dual TM hairpins is an additional indication of the formation of the predicted hairpin conformation. Overall the work described here provides an excellent framework and codifies several principles of the de novo design, selection, production and purification of model transmembrane proteins.

105

Chapter 6. Discussion.

Experiments described, in part, in this Chapter have been performed in collaboration with Rohan Alvares, Dr. R. Scott Prosser, and Dr. Peter MacDonald, from the Department of Chemistry at The University of Toronto – Mississauga campus.

106

Membrane proteins play central roles in a variety of essential biological activities. As such, a complete understanding of these processes requires knowledge of the structure and function of membrane proteins. Furthermore, misfolding or aberrant functioning of this class of proteins has been implicated in a variety of disease states such as Alzheimer’s disease, retinitis pigmentosa and cystic fibrosis (Partridge et al., 2002). Despite their importance, the study of membrane proteins remains challenging, making it necessary to continue to develop methods for the investigation of their folded structures. Many such methodological advances have relied upon the use of detergents as a membrane-mimetic environment to substitute for the native lipid bilayer environment.

In particular the generation of high-resolution structures (through X-ray crystallography and solution NMR) has relied upon identifying detergents that are both compatible with the technique, while allowing for stable membrane protein solubilization without denaturation into a non-biological state. The potential of detergents to disrupt function and/or folding of membrane proteins indicates that certain features of these proteins are not compatible with all environments. For the continued improvement of such structural characterization techniques it is essential to understand the nature of the interaction of such detergents with membrane proteins. Further elucidation of such biophysical phenomena should also aid in our understanding of the relationship of membrane protein structure in detergents to their folding in native biological systems. The work presented in this thesis has addressed such goals, and we believe is an advancement in the interpretation of membrane protein structure in detergents. The overall major insights of this thesis are discussed below:

107

6.1 Summary of contributions. 6.1.1 SDS micelles as a membrane-mimetic environment for transmembrane segments.

While sodium dodecyl sulfate (SDS) is perhaps the most widely employed micelle-forming detergent for laboratory procedures involving membrane proteins, it has generally been regarded as a ‘harsh’ detergent synonymous with membrane protein denaturation. In Chapter 2, we investigated the SDS-solubilized states of a series of model α-helical TM segments of varying Ala and Ile content in conjunction with selected single Asn polar substitutions. Using Lys- tagged peptides in a series of circular dichroism, fluorescence, TOXCAT dimerization assays, and SDS-PAGE migration experiments, we compare the folding of these segments in two different environments. Our results show that both the local environment of the individual peptide helical surfaces and the formation of oligomeric states within the SDS-peptide complex are highly sensitive to point changes in peptide sequence, particularly with respect to local segment hydrophobicity and polar residue placement. The overall findings suggest that the interaction of TM segments with SDS micelles are in some ways similar to the tertiary interactions of protein-, lipid-, and aqueous-exposed helical surfaces that arise in the folded TM domains of proteins. Certain sequence-specific characteristics of SDS-peptide complexes, such as oligomeric state and peptide-detergent interactions, may thus portend a corresponding role for similar TM sequences in the in vivo assembly of polytopic membrane proteins.

6.1.2 Positions of polar amino acids alter interactions between transmembrane segments and detergents.

Alpha-helical TM segments in membrane proteins are comprised primarily of hydrophobic amino acids that accommodate insertion into the non-polar membrane bilayer. In many such segments, however, polar residues are also present for structural or functional reasons. These latter residues impair the local favorable acyl interactions required for solvation by hydrophobic media such as phospholipids in native bilayers, or detergents used for in vitro characterization. In Chapter 3, single Asn residue substitutions (from Ile or Ala) were made successively from the center of the hydrophobic region of a model TM segment toward the C-terminus. The results demonstrate that polar residues strongly alter the nature of the interaction between TM segments

108

and the solvating detergent. Through the application of SDS-PAGE, circular dichroism spectroscopy, and tryptophan fluorescence, we observed drastic differences in the structures of the peptide-detergent complexes that contain relatively minor sequence differences. For example, the blue shift of Trp fluorescence (indicating local detergent solvation at this location) differs by as much as ~10 nm depending upon the position of a single Asn substitution in an otherwise identical segment. The overall results suggest that polar point mutations occurring in a biological membrane will elicit comparable effects, placing a significant re-folding burden on the local protein structure, and potentially leading to disease states through altered protein-lipid interactions in membrane proteins.

6.1.3 Transmembrane segment structures are similar in native proteins and detergents.

Despite their common use for structural analysis of membrane proteins, the choice of a detergent suitable for these experiments remains largely empirical. In Chapter 4 we considered the micelle-crystallized structures of lactose permease (LacY), the sodium/galactose symporter (vSGLT), the vitamin B12 transporter (BtuCD), and the arginine/agmatine antiporter (AdiC). Representative TM segments were selected from these proteins based on their relative contact(s) with water, lipid, and/or within the protein, and were synthesized as Lys-tagged peptides. Each peptide was studied by circular dichroism and fluorescence spectroscopy in water, and in the presence of the detergents sodium dodecyl sulfate (SDS, anionic); n-dodecyl phosphatidylcholine (DPC, zwitterionic); n-dodecyl-β-D-maltoside (DDM, neutral); and n-octyl- β-D-glucoside (OG, neutral, varying acyl tail length). We found that (i) the secondary structures of the TM segments were statistically indistinguishable in the four detergents studied; and (ii) a strong correlation exists between the extent of helical structure of each individual TM segment in detergents with its helicity level as it exists in the full-length protein, indicating that helix adoption is fundamentally the same in both environments. The denaturing properties of so-called ‘harsh’ detergents may thus largely be due to interactions with non-membranous regions of proteins. Given the consistency of structural features observed among these TM segments in a variety of micellar media, the overall results suggest that they likely correspond to the relevant biological form of the intact protein in its native lipid bilayer environment.

109

6.1.4 De novo designed hydrophobic “hairpins” as models of membrane protein folding.

While individual TM segments have proven useful in understanding membrane protein structure, they are unable to address the role of loops in membrane protein folding. The increased hydrophobicity and length of multi-spanning membrane proteins (versus those containing a single TM) however, makes their production and purification challenging. In Chapter 5 we utilized molecular biology cloning techniques to produce a partially randomized library of helix- loop-helix (hairpin) TM constructs. From this plasmid DNA library we selected sequences corresponding to hairpins with 0, 1 or 2 putative TM segments. Using these initial sequences we optimized both protein expression and purification protocols for application within a broad range of membrane protein hairpins. These hairpins were also characterized by CD spectroscopy, tryptophan fluorescence and SDS-PAGE. Our results indicate that hairpins composed of two TM segments have defining characteristics upon detergent solubilization such as an increase in helical structure (versus in aqueous buffer), and relatively quick gel migration rates during SDS- PAGE analysis. These findings will aid in the both the design and interpretation of the folding of de novo designed TM proteins.

6.2 Detergents as membrane mimetics.

In the work presented in this thesis, we investigate membrane protein structure primarily in detergent environments. The native environment of a lipid bilayer is in many ways different from a detergent micelle, as a bilayer is inherently more isotropic and has biophysical characteristics such as a membrane potential as well as gradients of lateral pressure and pH that are not present in micelle systems (Popot and Engelman, 2000). Gradients of dielectric constant, as well as chain ordering and packing are also present in bilayers, although it is not clear how well these features may be recapitulated in detergent micelle complexes. The sensitivity of detergent interactions to polar residue transverse position described in Chapter 3 implies that these latter gradients may indeed be mimicked in detergents to some degree. The main features of detergent micelles that most closely correspond to a bilayer are the low dielectric constant and presence of methyl and

110

ethyl chemical groups within the interior core. Overall the capability of detergents to mimic native membranes is dependent on the degree to which membrane protein structure is dictated by those features that are present in these two environments versus the sensitivity to those features that are not shared. Indeed, even less complex systems, such as hydrophobic solvent mixtures, have been successfully used in structural characterization studies indicating that in some cases only a subset of membrane bilayer features is necessary to produce a biologically relevant state (Sanders and Sonnichsen, 2006).

6.2.1 Comparison of folding in detergents versus bilayers.

Certain membrane proteins are particularly sensitive to their environment. For these proteins the aforementioned features that are found solely in natural bilayers (and not detergents) are required for the structure or function. For example, proper functioning of voltage-gated ion channels requires conformational changes intimately linked to the membrane potential (Catterall, 2011). In other circumstances specific lipid interactions are require for function (Ernst et al., 2010). Detergents would likely not be able to replace such interactions themselves, although in some cases detergents may allow for the retention of such tightly bound lipids upon solubilization from native sources (Qin et al., 2007).

Other, more general features of membrane protein folding, such as helix-helix interactions are, in some cases, well preserved in the membrane mimetic environment of detergents. The dimerization of the GpA TM domain has been studied in a variety of detergents as well as in bacterial membranes, and in model lipid bilayers. Dimerization of GpA occurs in all media through a motif containing Gly residues spaced three residues apart (termed the GxxxG or GG4 motif) (MacKenzie et al., 1997). A large number of GpA mutant variants (~50) of this dimerization interface have been produced in order to investigate sequence-dependent folding in various environments. Differences in apparent free energies of GpA mutants as measured by an α-helix association assay in the bacterial membrane (Russ and Engelman, 1999) were found to correlate well to apparent free energies obtained by SE-AUC (Fleming and Engelman, 2001), as well as to relative dimerization on SDS-PAGE (Duong et al., 2007). While the oligomerization tendency of the core GxxxG motif appears to be maintained in all environments, the relative contributions to the free energy of association of the surrounding residues vary in different

111

environments (Fisher et al., 2003; Duong et al., 2007; Hong et al., 2010; Janosi et al., 2010). Overall the results show that while the central motif itself is consistent in a variety of hydrophobic environments, overall association strength does differ to some degree in bilayers versus micelles.

In Chapter 2, we compared the association of a series of model TM segments in vivo to their association in SDS micelles. While results in both media are largely found to be in good qualitative agreement, we see that a decrease of hydrophobicity lowers relative dimerization in SDS, but not in membranes. Indeed, the correlation between in vivo association and SDS-PAGE has been found to be improved if loss of hydrophobicity is taken into account (Duong et al., 2007). This has been interpreted as partial unfolding of the helix due to disruption of insertion into the micelle, leading to an apparent loss in association strength (Lawrie et al., 2010). This discrepancy is due to the nature of TM segment insertion into these separate environments. In native lipid bilayers, insertion is mediated by the translocon. This process is largely irreversible, particularly if a TM segment is between hydrophilic domains that would be incapable of being re-translocated across the membrane (London and Shahidullah, 2009). The environment of a detergent micelle is also considerably more malleable, and potentially has smaller kinetic and or energetic barriers to interconversion between fully inserted, partially inserted, or non-inserted (aqueous-interactive) states. Such considerations are particularly important in studies of quaternary interactions (such as helix-helix dimerization), where such features make micelles more prone to changes in register of an oligomerization interface, as well as for allowing the possibility of anti-parallel interactions (Cunningham et al., 2011).

In certain circumstances such partially inserted states may limit native folding however, the exposure of a micelle-embedded TM segment to water may serve as a suitable substitute for TM segments that line water-filled cavities common to membrane protein channels and pores in vivo, as described in Chapter 4. Additionally, water exposed residues in TM segments are also present in receptor proteins where they contribute to ligand binding pockets, as well as ion channels where they help form selectivity filters (Fig. 6.1). Of the TM segments studied in Chapter 4, those that lined such water-filled cavities were of low hydrophobicity and possessed little helical structure in detergents. These segments are likely to be in a partially inserted state where significant exposure to water results in a random coil conformation as is observed for these segments in the absence of detergent. It should be noted, however, that an incomplete insertion in

112

native bilayers is also predicted for these segments if occurring in isolation from the remainder of the protein. The helical structure of these segments in the full native form is likely only realized upon stabilization by the remainder of the protein. The loss of native secondary structure in these segments is not likely a direct result of denaturation by detergents, but instead a consequence of being excised from the complete structure.

In Chapter 4, we see that the adoption of secondary structure in all TM segments studied appears to be consistent across a variety of detergents and is comparable to that observed in the native structure upon correction for bilayer insertion propensity. Furthermore, in Chapter 5 we observe that the degree of helical induction of proteins in SDS is largest for hairpins with two TM segments, indicating that the predicted native secondary structure is present. Overall the work in this thesis suggests that structural features such as α-helical secondary structure adoption and certain helix-helix association motifs may be well preserved in detergents. It is likely that the formation of these features of TM segments is primarily dependent upon placement of the segment in an apolar environment, which occurs in both membrane as well as detergents.

113

Figure 6.1. Schematic views of the transmembrane regions of selected α-helical membrane proteins. Individual residues are categorized and rendered in colors according to local environment, where yellow represents protein-exposed residues involved in helix-helix contacts, green represents lipid-exposed residues, and blue represents aqueous-exposed residues. Residues designated in each of these three categories are defined as described in Chapter 4. Residues classified as participating in helix-helix contacts were defined as having one or more atom within 3.5 Å of an amino acid residue on a separate TM segment. Shown for each structure are: a representative single TM helix derived from the overall TM domain of the protein; a view showing the external surface of the protein within the membrane (side view); the same view cutaway showing the internal structure of the protein; and a top view of the protein looking towards the membrane surface. A) Structure of the TM domain of the human β2 adrenergic receptor β2AR. TM segments are comprised predominantly of lipid-exposed and protein- exposed surfaces. Residues that contribute to the ligand binding pocket on the extracellular surface are aqueous-exposed (top view). B) Structure of LacY, a transport protein containing a significant number of residues that are aqueous-exposed in addition to lipid-exposed and protein- exposed residues. A cutaway view reveals the central hydrophilic cavity as well as residues involved in helix-helix contacts. C) Structure of the TM domain of the KcsA potassium channel that also contains a re-entrant loop and a partially-spanning helical segment. Residues from these features contribute both to the formation of the aqueous-exposed channel entrance/selectivity filter, along with helix-helix contact surfaces (both seen in the cutaway view).

114

6.2.2 Denaturation of membrane proteins by detergents.

While detergent micelles and membranes have some features that vary between them, they remain similar in one very important regard: they both present a largely hydrophobic environment into which protein sequences may fold. It should not be surprising then, that in many circumstances, detergent micelles are indeed a suitable environment for membrane protein folding. Another way of thinking about this issue is to ask the question: why should detergent micelles lead to the denaturation of TM protein structure? In order to denature water soluble, globular proteins, extreme conditions are often employed, such as drastic changes in salt concentration, the presence of a high concentration of chemical denaturant (such as guanidinium chloride or urea at concentrations up to 8 M), or heating (Dunbar et al., 1997). These effects lead to major changes in the nature of the solvent in which folding occurs, often providing competition for native protein folding interactions (such as backbone H-bonds that stabilize secondary structure) and/or by altering the structure and dynamics of water, thereby diminishing the hydrophobic effect (Bennion and Daggett, 2003). The difference in bulk properties between a bilayer and a detergent micelle do not seem as extreme in comparison to the changes in solvent properties that must occur for the unfolding of many water-soluble proteins. If it is the hydrophobic environment of the bilayer that is the main determinate of transmembrane domain folding, then the internal environment of a detergent micelle should indeed be considered similar to a native membrane bilayer.

Although there are some circumstances where specific lipids are required for function, overall native membrane protein structure can generally be accommodated in a variety of bulk lipid compositions. A native membrane protein is often required to fold in a variety of lipid compositions that may vary in response to stages of cell development or environmental factors. Lipid compositions are also different in the various sub-cellular organelles that host a membrane protein at different points during its processing and trafficking to its final membrane destination (Sanders and Mittendorf, 2011). This intrinsic robustness of membrane proteins to a variety of bilayer compositions may have been strongly selected for during evolution and consequently imparts a natural stability in detergent micelles (Sanders and Mittendorf, 2011).

Despite similar fundamental properties to bilayers, certain detergents have been used to great effect to denature membrane proteins (Booth and Curnow, 2006; Renthal, 2006; Joh et al., 2008).

115

The degree to which detergents are denaturing is not consistent however, and a variety of effects for different detergents are observed (Tulumello and Deber, 2011a). Why and how certain detergents are denaturing remains largely empirical (Privé, 2007; Newstead et al., 2008). In Chapter 4 we see that helix adoption and micelle insertion of individual TM segments is largely unaffected by the identity of the detergent. Much of the observed denaturing properties are likely due to interactions with extra-membranous regions of the proteins, or are associated with only minor direct changes in the structure of membrane protein domains themselves. Indeed, even in the “unfolded” states of membrane proteins present in detergents, there may be significant tertiary structure formation within the TM domain (Renthal, 2006). While it is clear that some detergents denature certain membrane proteins - particularly larger transport proteins that have structured cytoplasmic or extracellular domains - one should not have a biased view that many detergents are not ideal for biophysical studies. Rather, where little extra-membranous structure is present, a larger variety of detergents may be useful.

6.3 Future directions of investigation of membrane protein folding. 6.3.1 Folding of designed membrane proteins.

Compared to water-soluble proteins, the study of membrane protein folding is much less advanced. A testament to the understanding of the folding of globular water-soluble proteins is the success of many studies that have designed such proteins from first principles (Baker, 2006; Smith and Hecht, 2011). Although many designed protein sequences have been found to adopt a native-like fold, artificial proteins often form molten globule states unless specific core packing has been included in the design. These molten globular forms consist of a compact structure containing secondary structure but lack well defined, non-transient tertiary interactions (Ohgushi and Wada, 1983). For protein design studies it is thus necessary to confirm a native-like fold by demonstrating characteristics of native folded proteins such as cooperative unfolding, well- dispersed NMR spectra, and/or limited hydrogen/ deuterium exchange of buried residues (Schafmeister et al., 1997; Go et al., 2008). The ability to generate a high resolution structure by X-ray crystallography or solution NMR is also feasible in many cases, and may be used as the ultimate validation of a folded structure.

116

The intermolecular association of TM helices is among the most widely understood aspects of membrane protein folding. Accordingly, most membrane protein design efforts to date have focused on either the rational design of such quaternary folding features, or the selection of interacting TM sequences from large libraries (Ghirlanda, 2009). Much less effort has been made toward the design of polytopic membrane proteins. Such endeavors are challenging as structural hallmarks of folded polytopic membrane proteins are not as well characterized or experimentally accessible in comparison to water-soluble proteins. As such, it is difficult to assess how well native TM segment tertiary folding is recapitulated in artificial constructs, such as those described in Chapter 5. Although SDS-PAGE migration provides some indication of compactness, a more robust folding method would aid in clearer interpretation of the degree of proper TM hairpin formation. In future membrane protein design it would be beneficial to include features specifically engineered to more readily determine the presence of a folded state. For example, fluorescence tags used for FRET or pyrene excimer fluorescence may be used to further evaluate the folding of such hairpins in a variety of detergents (Therien and Deber, 2002b).

Indeed, in Chapter 5 we see that both of the hairpins with putative dual TM topology appear, to a first approximation, to have a more compact fold than those lacking two putative TM segments. This may be due to a collapsed fold but does not necessarily imply the adoption of a specific helix-helix interface TM between the adjacent segments in a hairpin structure. Extensive mutational analysis is required in order to determine the existence of such specific side chain interactions, although even such methods are confounded by differences in detergent interactions (Mulvihill and Deber, 2012). Until our understanding of membrane protein folding progresses to a point where we have a more complete thermodynamic view of both the folded and unfolded states, it may be challenging to differentiate between a collapsed fold with only transient interactions (similar to soluble protein molten globule), versus one with a stable well-defined core.

While still in its infancy, reversible membrane protein folding studies are beginning to lead to such a more complete view of membrane protein structure. For example, studies have shown that the energetic contributions of van der Waals packing are similar in the core of both soluble and membrane proteins (Joh et al., 2009). As well, kinetic analysis of folding and unfolding has been used to reveal the existence of specific transition states along the folding pathway (Curnow and

117

Booth, 2007). An important caveat to these studies is that they typically rely upon denaturation using detergents. This may be problematic as the free energy of the unfolded state may differ upon mutation due to differences in detergent interactions (Joh et al., 2008). As such a detergent- solubilized state is not completely analogous to the unfolded states observed for most water- soluble proteins. New methods that do not rely upon detergent denaturation are particularly promising, as they eliminate the consideration of detergent-protein interactions (Hong et al., 2010).

6.3.2 Methods to assess structural details of protein detergent complexes.

As shown throughout this thesis, by careful examination of detergent-peptide interactions we are able to gain valuable insights into membrane protein folding. Furthermore, an understanding of such interactions is necessary as many of the current methods to investigate the folding of membrane proteins are complicated by variance in the amount of detergent bound. In Chapter 3, we provide evidence that SDS-PAGE is correlated to the amount of detergent bound in single TM segments. By comparing the effects of single point mutations in hairpin models versus single TMs it is apparent that SDS-PAGE migration rates in more intricate systems are not solely a function of amount of detergent bound, but also conformational differences (Mulvihill and Deber, 2012). Size exclusion has also been used to assess the size of membrane protein complexes; however this technique relies on comparison to soluble protein standards, and as with SDS-PAGE, it is difficult to discern between effects related to differences in amount of detergent bound versus differences in conformation. Isothermal titration calorimetry has been used to great success for soluble proteins, but requires high concentrations of the protein, which limits the range of detergent concentrations that can be investigated (Otzen, 2011). Direct measurements of the amount of detergent bound provide good estimates, but most of these techniques are not sensitive enough to detect small differences among closely related proteins (le Maire et al., 2008; Rath et al., 2009a). Additional techniques that can be used to assess detergent binding would thus provide valuable tools in membrane protein characterization.

In this context, a current area of investigation in our lab is the development of NMR based techniques to assess structural features of TM proteins in a variety of detergents. These studies

118

are being performed in collaboration with the labs of Dr. Scott Prosser and Dr. Peter MacDonald in the Department of Chemistry, University of Toronto. We have employed pulsed field gradient (PFG) diffusion NMR spectroscopy in conjunction with CD, to monitor the interaction of detergents with model TM segments. Diffusion coefficients obtained by PFG-NMR would allow us to assess the size of any molecules with observable NMR resonances. By monitoring diffusion coefficients of both the detergent and peptide at low detergent-to-peptide ratios, just above saturating detergent concentrations, we are also able to investigate their interaction with one another. In initial studies, we have chosen to compare detergent interactions of the AI5 peptide segment (described in Chapters 2 and 3), as well as an Ile-to-Asn variant (I12N). Upon detergent saturation, we found that the AI5 I12N peptide has a smaller overall particle size (indicated by a larger diffusion coefficient - 6.5 x 10-11 m2s-1) than AI5 (diffusion coefficient of 6.3 x 10-11 m2s-1) as was observed in SDS-PAGE analysis (Chapter 3).

We also monitored the diffusion rate of both detergent and the peptides over a range of detergent concentrations where both detergent binding as well as peptide folding occurs (Fig. 6.2). There is a notable shift in the diffusion coefficient of the detergent to lower values upon the inclusion of peptides (in Fig. 6.2 compare the coefficient of the detergent in the absence [black circles] versus the presence [red and blue circles] of peptide). This shift in diffusion coefficient results from the binding of detergent monomers to the peptide and can be used to quantify the total amount of bound detergent.

119

Figure 6.2. Diffusion coefficient (D) profiles of SDS and model TM segments over the course of a detergent titration. Black circles indicate the diffusion coefficient of a sample containing SDS only. The decrease in diffusion coefficient as detergent concentration increases indicates the formation of detergent micelles. Red and blue circles indicate the diffusion coefficient of the SDS in samples that contain the AI5 or AI5 I12N TM segments, respectively. The lower diffusion coefficient of the SDS in these cases (compared to the SDS control sample) indicates binding of some detergent molecules to the TM segments. Diamonds represent the diffusion rates of the TM segments themselves (AI5 is shown in red, AI5 I17N is shown in blue). Vertical dashed lines indicate the midpoint of the induction of helical structure, as observed by CD. This folding transition is typically complete shortly after these points (within 100-200 μM). The critical micelle concentration was determined to be 1900 μM under these specific buffer conditions. In all cases, the error bars indicated 95% confidence intervals.

120

Upon quantification of detergent binding of the two aforementioned segments (see Appendix 5 for details) we observed that, upon saturation, the AI5 segment binds roughly 4 more detergent molecules per peptide molecule than the I12N variant (Figure 6.3). This implies that it is not conformational differences, but rather differences in amount of detergent binding that led to the variation in size of these two segments that is observed in both SDS-PAGE as well as NMR diffusion studies.

Figure 6.3. SDS saturation curves of model TM segments. The maximum number of SDS molecules bound to the AI5 and the I12N peptides were significantly different (p < 0.05) and were 23.0 ± 0.9 and 19.2 ± 0.9, respectively. At low SDS concentrations, resonances from the AI5 peptide were not observed, preventing this analysis from being performed in this concentration range. See Appendix 5 for calculations of amount of detergent bound from diffusion coefficients.

121

NMR thus provides an excellent alternative to evaluate changes in size in comparison to PAGE as it does not have the requirement for use with only charged detergents, nor is size estimation affected by the charge-to-mass ratio, which, in turn allows more accurate absolute comparisons. Indeed, similar investigations have also been undertaken using DPC as the solubilizing detergent, in which we have demonstrated an altered mode of binding for this zwitterionic detergent. The PFG technique also allows for the monitoring of the detergent itself, and accordingly for a highly sensitive measurement of the amount of detergent bound. The accuracy of this measurement requires the amount of detergent present to be as close to the detergent binding saturation point as possible, and therefore should be performed in conjunction with an independent assessment of this point, such as via CD spectroscopy.

Overall, these results confirm the finding in Chapter 3 that a polar residue within a TM segment has a drastic effect on its local environment, in this case within a detergent micelle. By using PFG-NMR, we are able to assess differences in both detergent particle sizes as well as amount of detergent bound in an extremely sensitive manner. This loss of ~ 4 detergent molecules upon inclusion of just a single polar residue is quite drastic and represents a loss of ~ 20 % of the total detergent bound. This may relate to the cooperative manner in which this detergent binds hydrophobic protein sequences (Imamura, 2006; Otzen, 2011). Furthermore, the acyl tail length of SDS (12 carbons) is much longer than any individual hydrophobic amino acid side chain, and thus detergent binding likely occurs via a surface consisting of multiple amino acids. A reduction in detergent affinity for a given peptide surface – such as that expected to occur upon inclusion of a strongly polar residue – would thus prevent the cooperative binding of multiple detergent molecules.

Similar major perturbations would be expected in membrane bilayers where mutation to polar residues in transmembrane domains is a common occurrence in disease states (Partridge et al., 2002). In such circumstances the interaction of the long chain acyl tails of phospholipids with the primarily hydrophobic segment are likely perturbed in a manner comparable to what occurs in detergent micelles. Being able to address the biophysical origins of such mutations with the precision that these novel techniques allow will provide a more complete understanding of the structural consequences of such disease causing mutations.

122

6.4 Conclusions.

Their cost-efficiency, wide availability, and ease of use have made detergents an indispensible tool in the investigation of membrane protein folding. While not ideally suitable for all applications, a range of detergents remain useful for many tasks if care is taken in both choosing a detergent as well as interpreting the experimental results. A detergent micelle is not a passive environment where TM segments are simply inserted as a centrally spanning helix. Instead the overall detergent-peptide complex is responsive to both the detergent identity and the sequence of the embedded TM segment. Even in comparisons of closely related sequences, the substitution of just a single residue may lead to broad changes in both detergent and protein structure. Nevertheless, responsiveness to sequence is in many cases expected to be related to native protein folding in natural membranes. With an increased understanding of the details of membrane protein detergent interactions, we will not only be better able to choose detergents for specific applications, but also to continue to use detergents to gain further in-depth insights into membrane protein folding that are not experimentally accessible in other systems.

123

Chapter 7. Literature Cited

Abramson, J., Smirnova, I., Kasho, V., Verner, G., Iwata, S., and Kaback, H. R. (2003). The lactose permease of Escherichia coli: overall structure, the sugar-binding site and the alternating access model for transport. FEBS Lett, 555, 96-101.

Adair, B. D., and Engelman, D. M. (1994). Glycophorin A helical transmembrane domains dimerize in phospholipid bilayers: a resonance energy transfer study. Biochemistry, 33, 5539-5544.

Allen, S. J., Kim, J. M., Khorana, H. G., Lu, H., and Booth, P. J. (2001). Structure and function in bacteriorhodopsin: the effect of the interhelical loops on the protein folding kinetics. J Mol Biol, 308, 423-435.

Almeida, F. C., and Opella, S. J. (1997). fd coat protein structure in membrane environments: structural dynamics of the loop between the hydrophobic trans-membrane helix and the amphipathic in-plane helix. J Mol Biol, 270, 481-495.

Almen, M. S., Nordstrom, K. J., Fredriksson, R., and Schioth, H. B. (2009). Mapping the human membrane proteome: a majority of the human membrane proteins can be classified according to function and evolutionary origin. BMC Biol, 7, 50.

Amblard, M., Fehrentz, J. A., Martinez, J., and Subra, G. (2005). Fundamentals of modern peptide synthesis. Methods Mol Biol, 298, 3-24.

Andrade, S. L., Dickmanns, A., Ficner, R., and Einsle, O. (2005). Crystal structure of the archaeal ammonium transporter Amt-1 from Archaeoglobus fulgidus. Proc Natl Acad Sci U S A, 102, 14994-14999.

Arkin, I. T., Adams, P. D., MacKenzie, K. R., Lemmon, M. A., Brunger, A. T., and Engelman, D. M. (1994). Structural organization of the pentameric transmembrane alpha-helices of phospholamban, a cardiac ion channel. EMBO J, 13, 4757-4764.

Artemenko, E. O., Egorova, N. S., Arseniev, A. S., and Feofanov, A. V. (2008). Transmembrane domain of EphA1 receptor forms dimers in membrane-like environment. Biochim Biophys Acta, 1778, 2361-2367.

Baenziger, J. E., and daCosta, C. J. (2008). Membrane protein structure and conformational change probed using Fourier transform infrared spectroscopy. In E. Pebay-Peyroula (Ed.), Biophysical analysis of membrane proteins : investigating structure and function (pp. 243-288). Weinheim: Wiley-VCH.

Baker, D. (2006). Prediction and design of macromolecular structures and interactions. Philos Trans R Soc Lond B Biol Sci, 361, 459-463.

Bennion, B. J., and Daggett, V. (2003). The molecular basis for the chemical denaturation of proteins by urea. Proc Natl Acad Sci U S A, 100, 5142-5147.

124

Bernsel, A., Viklund, H., Falk, J., Lindahl, E., von Heijne, G., and Elofsson, A. (2008). Prediction of membrane-protein topology from first principles. Proc Natl Acad Sci U S A, 105, 7177-7181.

Beswick, V., Guerois, R., Cordier-Ochsenbein, F., Coic, Y. M., Tam, H. D., Tostain, J., Noel, J. P., Sanson, A., and Neumann, J. M. (1999). Dodecylphosphocholine micelles as a membrane-like environment: new results from NMR relaxation and paramagnetic relaxation enhancement analysis. Eur Biophys J, 28, 48-58.

Bocharov, E. V., Mayzel, M. L., Volynsky, P. E., Goncharuk, M. V., Ermolyuk, Y. S., Schulga, A. A., Artemenko, E. O., Efremov, R. G., and Arseniev, A. S. (2008a). Spatial structure and pH-dependent conformational diversity of dimeric transmembrane domain of the receptor tyrosine kinase EphA1. J Biol Chem, 283, 29385-29395.

Bocharov, E. V., Mineev, K. S., Volynsky, P. E., Ermolyuk, Y. S., Tkach, E. N., Sobol, A. G., Chupin, V. V., Kirpichnikov, M. P., Efremov, R. G., and Arseniev, A. S. (2008b). Spatial structure of the dimeric transmembrane domain of the growth factor receptor ErbB2 presumably corresponding to the receptor active state. J Biol Chem, 283, 6950-6956.

Booth, P. J., and Curnow, P. (2006). Membrane proteins shape up: understanding in vitro folding. Curr Opin Struct Biol, 16, 480-488.

Bordag, N., and Keller, S. (2010). Alpha-helical transmembrane peptides: a "divide and conquer" approach to membrane proteins. Chem Phys Lipids, 163, 1-26.

Bradley, L. H., Thumfort, P. P., and Hecht, M. H. (2006). De novo proteins from binary- patterned combinatorial libraries. Methods Mol Biol, 340, 53-69.

Bruce, C. D., Berkowitz, M. L., Perera, L., and Forbes, M. D. E. (2002). Molecular dynamics simulation of sodium dodecyl sulfate micelle in water: micellar structural characteristics and counterion distribution. J Phys Chem B, 106, 3788-3793.

Caffrey, M. (2009). Crystallizing membrane proteins for structure determination: use of lipidic mesophases. Annu Rev Biophys, 38, 29-51.

Call, M. E., Schnell, J. R., Xu, C., Lutz, R. A., Chou, J. J., and Wucherpfennig, K. W. (2006). The structure of the zetazeta transmembrane dimer reveals features essential for its assembly with the T cell receptor. Cell, 127, 355-368.

Caputo, G. A., and London, E. (2004). Position and ionization state of Asp in the core of membrane-inserted alpha helices control both the equilibrium between transmembrane and nontransmembrane helix topography and transmembrane helix positioning. Biochemistry, 43, 8794-8806.

Carpenter, E. P., Beis, K., Cameron, A. D., and Iwata, S. (2008). Overcoming the challenges of membrane protein crystallography. Curr Opin Struct Biol, 18, 581-586.

Catterall, W. A. (2011). Ion channel voltage sensors: structure, function, and pathophysiology. Neuron, 67, 915-928.

125

Chen, Y. H., Yang, J. T., and Chau, K. H. (1974). Determination of the helix and beta form of proteins in aqueous solution by circular dichroism. Biochemistry, 13, 3350-3359.

Chill, J. H., Louis, J. M., Miller, C., and Bax, A. (2006). NMR study of the tetrameric KcsA potassium channel in detergent micelles. Protein Sci, 15, 684-698.

Cho, G., Keefe, A. D., Liu, R., Wilson, D. S., and Szostak, J. W. (2000). Constructing high complexity synthetic libraries of long ORFs using in vitro selection. J Mol Biol, 297, 309-319.

Chou, P. Y., and Fasman, G. D. (1978). Empirical predictions of protein conformation. Annu Rev Biochem, 47, 251-276.

Compton, E. L., Taylor, E. M., and Mindell, J. A. (2010). The 3-4 loop of an archaeal glutamate transporter homolog experiences ligand-induced structural changes and is essential for transport. Proc Natl Acad Sci U S A, 107, 12840-12845.

Cristian, L., Lear, J. D., and DeGrado, W. F. (2003a). Determination of membrane protein stability via thermodynamic coupling of folding to thiol-disulfide interchange. Protein Sci, 12, 1732-1740.

Cristian, L., Lear, J. D., and DeGrado, W. F. (2003b). Use of thiol-disulfide equilibria to measure the energetics of assembly of transmembrane helices in phospholipid bilayers. Proc Natl Acad Sci U S A, 100, 14772-14777.

Cunningham, F., and Deber, C. M. (2007). Optimizing synthesis and expression of transmembrane peptides and proteins. Methods, 41, 370-380.

Cunningham, F., Rath, A., Johnson, R. M., and Deber, C. M. (2009). Distinctions between hydrophobic helices in globular proteins and transmembrane segments as factors in protein sorting. J Biol Chem, 284, 5395-5402.

Cunningham, F., Poulsen, B. E., Ip, W., and Deber, C. M. (2011). Beta-branched residues adjacent to GG4 motifs promote the efficient association of glycophorin A transmembrane helices. Biopolymers, 96, 340-347.

Curnow, P., and Booth, P. J. (2007). Combined kinetic and thermodynamic analysis of alpha- helical membrane protein unfolding. Proc Natl Acad Sci U S A, 104, 18970-18975.

Davidson, A. R., and Sauer, R. T. (1994). Folded proteins occur frequently in libraries of random amino acid sequences. Proc Natl Acad Sci U S A, 91, 2146-2150.

Davidson, A. R., Lumb, K. J., and Sauer, R. T. (1995). Cooperatively folded proteins in random sequence libraries. Nat Struct Biol, 2, 856-864.

Dawson, J. P., Weinger, J. S., and Engelman, D. M. (2002). Motifs of serine and threonine can drive association of transmembrane helices. J Mol Biol, 316, 799-805.

126

de Foresta, B., Vincent, M., Gallay, J., and Garrigos, M. (2010). Interaction with membrane mimics of transmembrane fragments 16 and 17 from the human multidrug resistance ABC transporter 1 (hMRP1/ABCC1) and two of their tryptophan variants. Biochim Biophys Acta, 1798, 401-414. de Foresta, B., Vincent, M., Garrigos, M., and Gallay, J. (2011). Transverse and tangential orientation of predicted transmembrane fragments 4 and 10 from the human multidrug resistance protein (hMRP1/ABCC1) in membrane mimics. Eur Biophys J, 40, 1043-1060.

Deber, C. M., Wang, C., Liu, L. P., Prior, A. S., Agrawal, S., Muskat, B. L., and Cuticchia, A. J. (2001). TM Finder: a prediction program for transmembrane protein segments using a combination of hydrophobicity and nonpolar phase helicity scales. Protein Sci, 10, 212- 219.

Dunbar, J., Yennawar, H. P., Banerjee, S., Luo, J., and Farber, G. K. (1997). The effect of denaturants on protein structure. Protein Sci, 6, 1727-1733.

Dunker, A. K., and Kenyon, A. J. (1976). Mobility of sodium dodecyl sulphate - protein complexes. Biochem J, 153, 191-197.

Duong, M. T., Jaszewski, T. M., Fleming, K. G., and MacKenzie, K. R. (2007). Changes in apparent free energy of helix-helix dimerization in a biological membrane due to point mutations. J Mol Biol, 371, 422-434.

Eisenberg, D., Weiss, R. M., and Terwilliger, T. C. (1982). The helical hydrophobic moment: a measure of the amphiphilicity of a helix. Nature, 299, 371-374.

Engelman, D. M., Chen, Y., Chin, C. N., Curran, A. R., Dixon, A. M., Dupuy, A. D., Lee, A. S., Lehnert, U., Matthews, E. E., Reshetnyak, Y. K., Senes, A., and Popot, J. L. (2003). Membrane protein folding: beyond the two stage model. FEBS Lett, 555, 122-125.

Enquist, K., Fransson, M., Boekel, C., Bengtsson, I., Geiger, K., Lang, L., Pettersson, A., Johansson, S., von Heijne, G., and Nilsson, I. (2009). Membrane-integration characteristics of two ABC transporters, CFTR and P-glycoprotein. J Mol Biol, 387, 1153-1164.

Ernst, A. M., Contreras, F. X., Brugger, B., and Wieland, F. (2010). Determinants of specificity at the protein-lipid interface in membranes. FEBS Lett, 584, 1713-1720.

Esposito, C., Colicchio, P., Facchiano, A., and Ragone, R. (1998). Effect of a weak electrolyte on the critical micellar concentration of sodium dodecyl sulfate. [Article]. Journal of Colloid and Interface Science, 200, 310-312.

Faham, S., Watanabe, A., Besserer, G. M., Cascio, D., Specht, A., Hirayama, B. A., Wright, E. M., and Abramson, J. (2008). The crystal structure of a sodium galactose transporter reveals mechanistic insights into Na+/sugar symport. Science, 321, 810-814.

127

Fang, Y., Jayaram, H., Shane, T., Kolmakova-Partensky, L., Wu, F., Williams, C., Xiong, Y., and Miller, C. (2009). Structure of a prokaryotic virtual proton pump at 3.2 A resolution. Nature, 460, 1040-1043.

Federkeil, S. L., Winstone, T. L., Jickling, G., and Turner, R. J. (2003). Examination of EmrE conformational differences in various membrane mimetic environments. Biochem Cell Biol, 81, 61-70.

Fisher, L. E., Engelman, D. M., and Sturgis, J. N. (2003). Effect of detergents on the association of the glycophorin a transmembrane helix. Biophys J, 85, 3097-3105.

Fisher, M. A., McKinley, K. L., Bradley, L. H., Viola, S. R., and Hecht, M. H. (2011). De novo designed proteins from a library of artificial sequences function in Escherichia coli and enable cell growth. PLoS One, 6, e15364.

Fleming, K. G., and Engelman, D. M. (2001). Specificity in transmembrane helix-helix interactions can define a hierarchy of stability for sequence variants. Proc Natl Acad Sci U S A, 98, 14340-14344.

Fleming, K. G. (2002). Standardizing the free energy change of transmembrane helix-helix interactions. J Mol Biol, 323, 563-571.

Fleming, K. G. (2008). Determination of membrane protein molecular weight using sedimentation equilibrium analytical ultracentrifugation (Vol. 53): John Wiley & Sons, Inc.

Franzin, C. M., Teriete, P., and Marassi, F. M. (2007). Structural similarity of a membrane protein in micelles and membranes. J Am Chem Soc, 129, 8078-8079.

Gans, P. J., Lyu, P. C., Manning, M. C., Woody, R. W., and Kallenbach, N. R. (1991). The helix- coil transition in heterogeneous peptides with specific side-chain interactions: theory and comparison with CD spectral data. Biopolymers, 31, 1605-1614.

Ghirlanda, G. (2009). Design of membrane proteins: toward functional systems. Curr Opin Chem Biol, 13, 643-651.

Go, A., Kim, S., Baum, J., and Hecht, M. H. (2008). Structure and dynamics of de novo proteins from a designed superfamily of 4-helix bundles. Protein Sci, 17, 821-832.

Gorman, P. M., Kim, S., Guo, M., Melnyk, R. A., McLaurin, J., Fraser, P. E., Bowie, J. U., and Chakrabartty, A. (2008). Dimerization of the transmembrane domain of amyloid precursor proteins and familial Alzheimer's disease mutants. BMC Neurosci, 9, 17.

Gratkowski, H., Lear, J. D., and DeGrado, W. F. (2001). Polar side chains drive the association of model transmembrane peptides. Proc Natl Acad Sci U S A, 98, 880-885.

Griffith, I. P. (1972). The effect of cross-links on the mobility of proteins in dodecyl sulphate- polyacrylamide gels. Biochem J, 126, 553-560.

128

Grigorieff, N., Ceska, T. A., Downing, K. H., Baldwin, J. M., and Henderson, R. (1996). Electron-crystallographic refinement of the structure of bacteriorhodopsin. J Mol Biol, 259, 393-421.

Guan, L., Mirza, O., Verner, G., Iwata, S., and Kaback, H. R. (2007). Structural determination of wild-type lactose permease. Proc Natl Acad Sci U S A, 104, 15294-15298.

Hecht, M. H., Das, A., Go, A., Bradley, L. H., and Wei, Y. (2004). De novo proteins from designed combinatorial libraries. Protein Sci, 13, 1711-1723.

Hedin, L. E., Ojemalm, K., Bernsel, A., Hennerdal, A., Illergard, K., Enquist, K., Kauko, A., Cristobal, S., von Heijne, G., Lerch-Bader, M., Nilsson, I., and Elofsson, A. (2010). Membrane insertion of marginally hydrophobic transmembrane helices depends on sequence context. J Mol Biol, 396, 221-229.

Heginbotham, L., Odessey, E., and Miller, C. (1997). Tetrameric stoichiometry of a prokaryotic K+ channel. Biochemistry, 36, 10335-10342.

Hessa, T., Kim, H., Bihlmaier, K., Lundin, C., Boekel, J., Andersson, H., Nilsson, I., White, S. H., and von Heijne, G. (2005). Recognition of transmembrane helices by the endoplasmic reticulum translocon. Nature, 433, 377-381.

Hessa, T., Meindl-Beinker, N. M., Bernsel, A., Kim, H., Sato, Y., Lerch-Bader, M., Nilsson, I., White, S. H., and von Heijne, G. (2007). Molecular code for transmembrane-helix recognition by the Sec61 translocon. Nature, 450, 1026-1030.

Hirokawa, T., Boon-Chieng, S., and Mitaku, S. (1998). SOSUI: classification and secondary structure prediction system for membrane proteins. Bioinformatics, 14, 378-379.

Hong, H., Park, S., Jimenez, R. H., Rinehart, D., and Tamm, L. K. (2007). Role of aromatic side chains in the folding and thermodynamic stability of integral membrane proteins. J Am Chem Soc, 129, 8320-8327.

Hong, H., Blois, T. M., Cao, Z., and Bowie, J. U. (2010). Method to measure strong protein- protein interactions in lipid bilayers using a steric trap. Proc Natl Acad Sci U S A, 107, 19802-19807.

Hovers, J., Potschies, M., Polidori, A., Pucci, B., Raynal, S., Bonnete, F., Serrano-Vega, M. J., Tate, C. G., Picot, D., Pierre, Y., Popot, J. L., Nehme, R., Bidet, M., Mus-Veteau, I., Busskamp, H., Jung, K. H., Marx, A., Timmins, P. A., and Welte, W. (2011). A class of mild surfactants that keep integral membrane proteins water-soluble for functional studies and crystallization. Mol Membr Biol, 28, 171-181.

Howell, S. C., Mesleh, M. F., and Opella, S. J. (2005). NMR structure determination of a membrane protein with two transmembrane helices in micelles: MerF of the bacterial mercury detoxification system. Biochemistry, 44, 5196-5206.

Hubbard, S. J., and Thornton, J. M. (1993). NACCESS: Department of Biochemistry and Molecular Biology, University College London.

129

Hunte, C., and Michel, H. (2003). Membrane Protein Crystallization. In C. Hunte, G. Von Jagow & H. Schägger (Eds.), Membrane Protein Purification and Crystallization: A Practical Guide (2nd ed., pp. 143-160). Boston: Academic Press.

Illergård, K., Kauko, A., and Elofsson, A. (2011). Why are polar residues within the membrane core evolutionary conserved? Proteins, 79, 79-91.

Imamura, T. (Ed.) (2006) Encyclopedia of Surface and Colloid Science. New York: Taylor & Francis.

Janosi, L., Prakash, A., and Doxastakis, M. (2010). Lipid-modulated sequence-specific association of glycophorin A in membranes. Biophys J, 99, 284-292.

Joh, N. H., Min, A., Faham, S., Whitelegge, J. P., Yang, D., Woods, V. L., and Bowie, J. U. (2008). Modest stabilization by most hydrogen-bonded side-chain interactions in membrane proteins. Nature, 453, 1266-1270.

Joh, N. H., Oberai, A., Yang, D., Whitelegge, J. P., and Bowie, J. U. (2009). Similar energetic contributions of packing in the core of membrane and water-soluble proteins. J Am Chem Soc, 131, 10846-10847.

Johansson, A. C., and Lindahl, E. (2006). Amino-acid solvation structure in transmembrane helices from molecular dynamics simulations. Biophys J, 91, 4450-4463.

Johnson, R. M., Heslop, C. L., and Deber, C. M. (2004). Hydrophobic helical hairpins: design and packing interactions in membrane environments. Biochemistry, 43, 14361-14369.

Johnson, R. M., Rath, A., Melnyk, R. A., and Deber, C. M. (2006). Lipid solvation effects contribute to the affinity of Gly-xxx-Gly motif-mediated helix-helix interactions. Biochemistry, 45, 8507-8515.

Johnson, R. M., Hecht, K., and Deber, C. M. (2007). Aromatic and cation-pi interactions enhance helix-helix association in a membrane environment. Biochemistry, 46, 9208- 9214.

Joosten, R. P., te Beek, T. A., Krieger, E., Hekkelman, M. L., Hooft, R. W., Schneider, R., Sander, C., and Vriend, G. (2011). A series of PDB related databases for everyday needs. Nucleic Acids Res, 39, D411-419.

Kaback, H. R., Smirnova, I., Kasho, V., Nie, Y., and Zhou, Y. (2011). The alternating access transport mechanism in LacY. J Membr Biol, 239, 85-93.

Kabsch, W., and Sander, C. (1983). Dictionary of protein secondary structure: pattern recognition of hydrogen-bonded and geometrical features. Biopolymers, 22, 2577-2637.

Kahn, T. W., Sturtevant, J. M., and Engelman, D. M. (1992). Thermodynamic measurements of the contributions of helix-connecting loops and of retinal to the stability of bacteriorhodopsin. Biochemistry, 31, 8829-8839.

130

Kall, L., Krogh, A., and Sonnhammer, E. L. (2004). A combined transmembrane topology and signal peptide prediction method. J Mol Biol, 338, 1027-1036.

Kamtekar, S., Schiffer, J. M., Xiong, H., Babik, J. M., and Hecht, M. H. (1993). Protein design by binary patterning of polar and nonpolar amino acids. Science, 262, 1680-1685.

Kauko, A., Hedin, L. E., Thebaud, E., Cristobal, S., Elofsson, A., and von Heijne, G. (2010). Repositioning of transmembrane alpha-helices during membrane protein folding. J Mol Biol, 397, 190-201.

Killian, J. A., and Nyholm, T. K. (2006). Peptides in lipid bilayers: the power of simple models. Curr Opin Struct Biol, 16, 473-479.

Krishnakumar, S. S., and London, E. (2007). The control of transmembrane helix transverse position in membranes by hydrophilic residues. J Mol Biol, 374, 1251-1269.

Krogh, A., Larsson, B., von Heijne, G., and Sonnhammer, E. L. (2001). Predicting transmembrane protein topology with a hidden Markov model: application to complete genomes. J Mol Biol, 305, 567-580.

Kyte, J., and Doolittle, R. F. (1982). A simple method for displaying the hydropathic character of a protein. J Mol Biol, 157, 105-132.

Lakowicz, J. R. (2006). Principles of fluorescence spectroscopy (3rd ed.). New York: Springer.

Lawrie, C. M., Sulistijo, E. S., and MacKenzie, K. R. (2010). Intermonomer hydrogen bonds enhance GxxxG-driven dimerization of the BNIP3 transmembrane domain: roles for sequence context in helix-helix association in membranes. J Mol Biol, 396, 924-936.

Le Lan, C., Gallay, J., Vincent, M., Neumann, J. M., de Foresta, B., and Jamin, N. (2010). Structural and dynamic properties of juxta-membrane segments of caveolin-1 and caveolin-2 at the membrane interface. Eur Biophys J, 39, 307-325. le Maire, M., Champeil, P., and Moller, J. V. (2000). Interaction of membrane proteins and lipids with solubilizing detergents. Biochim Biophys Acta, 1508, 86-111. le Maire, M., Arnou, B., Olesen, C., Georgin, D., Ebel, C., and Moller, J. V. (2008). Gel chromatography and analytical ultracentrifugation to determine the extent of detergent binding and aggregation, and Stokes radius of membrane proteins using sarcoplasmic reticulum Ca2+-ATPase as an example. Nat Protoc, 3, 1782-1795.

Lear, J. D., Stouffer, A. L., Gratkowski, H., Nanda, V., and Degrado, W. F. (2004). Association of a model transmembrane peptide containing gly in a heptad sequence motif. Biophys J, 87, 3421-3429.

Lemmon, M. A., Flanagan, J. M., Hunt, J. F., Adair, B. D., Bormann, B. J., Dempsey, C. E., and Engelman, D. M. (1992a). Glycophorin A dimerization is driven by specific interactions between transmembrane alpha-helices. J Biol Chem, 267, 7683-7689.

131

Lemmon, M. A., Flanagan, J. M., Treutlein, H. R., Zhang, J., and Engelman, D. M. (1992b). Sequence specificity in the dimerization of transmembrane alpha-helices. Biochemistry, 31, 12719-12725.

Liao, M. J., London, E., and Khorana, H. G. (1983). Regeneration of the native bacteriorhodopsin structure from two chymotryptic fragments. J Biol Chem, 258, 9949- 9955.

Lin, S. H., and Guidotti, G. (2009). Purification of membrane proteins. Methods Enzymol, 463, 619-629.

Liu, L. P., Li, S. C., Goto, N. K., and Deber, C. M. (1996). Threshold hydrophobicity dictates helical conformations of peptides in membrane environments. Biopolymers, 39, 465-470.

Liu, L. P., and Deber, C. M. (1997). Anionic phospholipids modulate peptide insertion into membranes. Biochemistry, 36, 5476-5482.

Liu, L. P., and Deber, C. M. (1998a). Guidelines for membrane protein engineering derived from de novo designed model peptides. Biopolymers, 47, 41-62.

Liu, L. P., and Deber, C. M. (1998b). Uncoupling hydrophobicity and helicity in transmembrane segments. Alpha-helical propensities of the amino acids in non-polar environments. J Biol Chem, 273, 23645-23648.

Liu, L. P., and Deber, C. M. (1999). Combining hydrophobicity and helicity: a novel approach to membrane protein structure prediction. Bioorg Med Chem, 7, 1-7.

Liu, Y., Engelman, D. M., and Gerstein, M. (2002). Genomic analysis of membrane protein families: abundance and conserved motifs. Genome Biol, 3, 0054.0051-0054.0012.

Locher, K. P., Lee, A. T., and Rees, D. C. (2002). The E. coli BtuCD structure: a framework for ABC transporter architecture and mechanism. Science, 296, 1091-1098.

Lomize, M. A., Lomize, A. L., Pogozheva, I. D., and Mosberg, H. I. (2006). OPM: orientations of proteins in membranes database. Bioinformatics, 22, 623-625.

London, E., and Shahidullah, K. (2009). Transmembrane vs. non-transmembrane hydrophobic helix topography in model and natural membranes. Curr Opin Struct Biol, 19, 464-472.

MacCallum, J. L., Bennett, W. F., and Tieleman, D. P. (2008). Distribution of amino acids in a lipid bilayer from computer simulations. Biophys J, 94, 3393-3404.

MacKenzie, K. R., Prestegard, J. H., and Engelman, D. M. (1997). A transmembrane helix dimer: structure and implications. Science, 276, 131-133.

Majumdar, D. S., Smirnova, I., Kasho, V., Nir, E., Kong, X., Weiss, S., and Kaback, H. R. (2007). Single-molecule FRET reveals sugar-induced conformational dynamics in LacY. Proc Natl Acad Sci U S A, 104, 12640-12645.

132

Manning, M. C., and Woody, R. W. (1991). Theoretical CD studies of polypeptide helices: examination of important electronic and geometric factors. Biopolymers, 31, 569-586.

Melnyk, R. A., Partridge, A. W., and Deber, C. M. (2001). Retention of native-like oligomerization states in transmembrane segment peptides: application to the Escherichia coli aspartate receptor. Biochemistry, 40, 11106-11113.

Melnyk, R. A., Partridge, A. W., and Deber, C. M. (2002). Transmembrane domain mediated self-assembly of major coat protein subunits from Ff bacteriophage. J Mol Biol, 315, 63- 72.

Melnyk, R. A., Partridge, A. W., Yip, J., Wu, Y., Goto, N. K., and Deber, C. M. (2003). Polar residue tagging of transmembrane peptides. Biopolymers, 71, 675-685.

Mulvihill, C. M., and Deber, C. M. (2012). Structural basis for misfolding at a disease phenotypic position in CFTR: Comparison of TM3/4 helix-loop-helix constructs with TM4 peptides. Biochim Biophys Acta, 1818, 49-54.

Murata, K., Mitsuoka, K., Hirai, T., Walz, T., Agre, P., Heymann, J. B., Engel, A., and Fujiyoshi, Y. (2000). Structural determinants of water permeation through aquaporin-1. Nature, 407, 599-605.

Neumoin, A., Cohen, L. S., Arshava, B., Tantry, S., Becker, J. M., Zerbe, O., and Naider, F. (2009). Structure of a double transmembrane fragment of a G-protein-coupled receptor in micelles. Biophys J, 96, 3187-3196.

Newstead, S., Ferrandon, S., and Iwata, S. (2008). Rationalizing alpha-helical membrane protein crystallization. Protein Sci, 17, 466-472.

Ng, D. P., and Deber, C. M. (2010a). Modulation of the oligomerization of myelin proteolipid protein by transmembrane helix interaction motifs. Biochemistry, 49, 6896-6902.

Ng, D. P., and Deber, C. M. (2010b). Deletion of a terminal residue disrupts oligomerization of a transmembrane alpha-helix. Biochem Cell Biol, 88, 339-345.

Ohgushi, M., and Wada, A. (1983). 'Molten-globule state': a compact form of globular proteins with mobile side-chains. FEBS Lett, 164, 21-24.

Otzen, D. E., Nesgaard, L. W., Andersen, K. K., Hansen, J. H., Christiansen, G., Doe, H., and Sehgal, P. (2008). Aggregation of S6 in a quasi-native state by sub-micellar SDS. Biochim Biophys Acta, 1784, 400-414.

Otzen, D. E. (2011). Protein-surfactant interactions: a tale of many states. Biochim Biophys Acta, 1814, 562-591.

Overington, J. P., Al-Lazikani, B., and Hopkins, A. L. (2006). How many drug targets are there? Nat Rev Drug Discov, 5, 993-996.

133

Oxenoid, K., and Chou, J. J. (2005). The structure of phospholamban pentamer reveals a channel-like architecture in membranes. Proc Natl Acad Sci U S A, 102, 10870-10875.

Pace, C. N., Vajdos, F., Fee, L., Grimsley, G., and Gray, T. (1995). How to measure and predict the molar absorption coefficient of a protein. Protein Sci, 4, 2411-2423.

Page, R. C., Moore, J. D., Nguyen, H. B., Sharma, M., Chase, R., Gao, F. P., Mobley, C. K., Sanders, C. R., Ma, L., Sonnichsen, F. D., Lee, S., Howell, S. C., Opella, S. J., and Cross, T. A. (2006). Comprehensive evaluation of solution nuclear magnetic resonance spectroscopy sample preparation for helical integral membrane proteins. J Struct Funct Genomics, 7, 51-64.

Parsell, D. A., and Sauer, R. T. (1989). The structural stability of a protein is an important determinant of its proteolytic susceptibility in Escherichia coli. J Biol Chem, 264, 7590- 7595.

Partridge, A. W., Therien, A. G., and Deber, C. M. (2002). Polar mutations in membrane proteins as a biophysical basis for disease. Biopolymers, 66, 350-358.

Partridge, A. W., Therien, A. G., and Deber, C. M. (2004). Missense mutations in transmembrane domains of proteins: phenotypic propensity of polar residues for human disease. Proteins, 54, 648-656.

Peng, S., Liu, L. P., Emili, A. Q., and Deber, C. M. (1998). Cystic fibrosis transmembrane conductance regulator: expression and helicity of a double membrane-spanning segment. FEBS Lett, 431, 29-33.

Popot, J. L., and Engelman, D. M. (1990). Membrane protein folding and oligomerization: the two-stage model. Biochemistry, 29, 4031-4037.

Popot, J. L., and Engelman, D. M. (2000). Helical membrane protein folding, stability, and evolution. Annu Rev Biochem, 69, 881-922.

Porcelli, F., Verardi, R., Shi, L., Henzler-Wildman, K. A., Ramamoorthy, A., and Veglia, G. (2008). NMR structure of the cathelicidin-derived human antimicrobial peptide LL-37 in dodecylphosphocholine micelles. Biochemistry, 47, 5565-5572.

Poulsen, B. E., Rath, A., and Deber, C. M. (2009). The assembly motif of a bacterial small multidrug resistance protein. J Biol Chem, 284, 9870-9875.

Privé, G. G. (2007). Detergents for the stabilization and crystallization of membrane proteins. Methods, 41, 388-397.

Privé, G. G. (2009). Lipopeptide detergents for membrane protein studies. Curr Opin Struct Biol, 19, 379-385.

Qin, L., Sharpe, M. A., Garavito, R. M., and Ferguson-Miller, S. (2007). Conserved lipid- binding sites in membrane proteins: a focus on cytochrome c oxidase. Curr Opin Struct Biol, 17, 444-450.

134

Raman, P., Cherezov, V., and Caffrey, M. (2006). The Membrane Protein Data Bank. Cell Mol Life Sci, 63, 36-51.

Raschle, T., Hiller, S., Etzkorn, M., and Wagner, G. (2010). Nonmicellar systems for solution NMR spectroscopy of membrane proteins. Curr Opin Struct Biol, 20, 471-479.

Rath, A., Melnyk, R. A., and Deber, C. M. (2006). Evidence for assembly of small multidrug resistance proteins by a "two-faced" transmembrane helix. J Biol Chem, 281, 15546- 15553.

Rath, A., and Deber, C. M. (2007). Membrane protein assembly patterns reflect selection for non-proliferative structures. FEBS Lett, 581, 1335-1341.

Rath, A., Glibowicka, M., Nadeau, V. G., Chen, G., and Deber, C. M. (2009a). Detergent binding explains anomalous SDS-PAGE migration of membrane proteins. Proc Natl Acad Sci U S A, 106, 1760-1765.

Rath, A., Tulumello, D. V., and Deber, C. M. (2009b). Peptide models of membrane protein folding. Biochemistry, 48, 3036-3045.

Rath, A., Nadeau, V. G., Poulsen, B. E., Ng, D. P., and Deber, C. M. (2010). Novel hydrophobic standards for membrane protein molecular weight determinations via sodium dodecyl sulfate-polyacrylamide gel electrophoresis. Biochemistry, 49, 10589-10591.

Renthal, R. (2006). An unfolding story of helical transmembrane proteins. Biochemistry, 45, 14559-14566.

Reynolds, J. A., and Tanford, C. (1970). Binding of dodecyl sulfate to proteins at high binding ratios. Possible implications for the state of proteins in biological membranes. Proc Natl Acad Sci U S A, 66, 1002-1007.

Reynolds, J. A., and Tanford, C. (1976). Determination of molecular weight of the protein moiety in protein-detergent complexes without direct knowledge of detergent binding. Proc Natl Acad Sci U S A, 73, 4467-4470.

Rodbard, D., and Chrambach, A. (1970). Unified theory for gel electrophoresis and gel filtration. Proc Natl Acad Sci U S A, 65, 970-977.

Russ, W. P., and Engelman, D. M. (1999). TOXCAT: a measure of transmembrane helix association in a biological membrane. Proc Natl Acad Sci U S A, 96, 863-868.

Sahin-Toth, M., Frillingos, S., Bibi, E., Gonzalez, A., and Kaback, H. R. (1994). The role of transmembrane domain III in the lactose permease of Escherichia coli. Protein Sci, 3, 2302-2310.

Sal-Man, N., Gerber, D., and Shai, Y. (2005). The identification of a minimal dimerization motif QXXS that enables homo- and hetero-association of transmembrane helices in vivo. J Biol Chem, 280, 27449-27457.

135

Sal-Man, N., Gerber, D., Bloch, I., and Shai, Y. (2007). Specificity in transmembrane helix-helix interactions mediated by aromatic residues. J Biol Chem, 282, 19753-19761.

Sanders, C. R., and Sonnichsen, F. (2006). Solution NMR of membrane proteins: practice and challenges. Magn Reson Chem, 44, S24-40.

Sanders, C. R., and Mittendorf, K. F. (2011). Tolerance to changes in membrane lipid composition as a selected trait of membrane proteins. Biochemistry, 50, 7858-7867.

Schafmeister, C. E., LaPorte, S. L., Miercke, L. J., and Stroud, R. M. (1997). A designed four helix bundle protein with native-like structure. Nat Struct Biol, 4, 1039-1046.

Schoneberg, T., Liu, J., and Wess, J. (1995). Plasma membrane localization and functional rescue of truncated forms of a G protein-coupled receptor. J Biol Chem, 270, 18000- 18006.

Seelig, B., and Szostak, J. W. (2007). Selection and evolution of enzymes from a partially randomized non-catalytic scaffold. Nature, 448, 828-831.

Senes, A., Gerstein, M., and Engelman, D. M. (2000). Statistical analysis of amino acid patterns in transmembrane helices: the GxxxG motif occurs frequently and in association with beta-branched residues at neighboring positions. J Mol Biol, 296, 921-936.

Senes, A., Engel, D. E., and DeGrado, W. F. (2004). Folding of helical membrane proteins: the role of polar, GxxxG-like and proline motifs. Curr Opin Struct Biol, 14, 465-479.

Shen, H., and Chou, J. J. (2008). MemBrain: improving the accuracy of predicting transmembrane helices. PLoS ONE, 3, e2399.

Simmerman, H. K., Kobayashi, Y. M., Autry, J. M., and Jones, L. R. (1996). A leucine zipper stabilizes the pentameric membrane domain of phospholamban and forms a coiled-coil pore structure. J Biol Chem, 271, 5941-5946.

Smirnova, I., Kasho, V., Choe, J. Y., Altenbach, C., Hubbell, W. L., and Kaback, H. R. (2007). Sugar binding induces an outward facing conformation of LacY. Proc Natl Acad Sci U S A, 104, 16504-16509.

Smith, B. A., and Hecht, M. H. (2011). Novel proteins: from fold to function. Curr Opin Chem Biol, 15, 421-426.

Song, L., Hobaugh, M. R., Shustak, C., Cheley, S., Bayley, H., and Gouaux, J. E. (1996). Structure of staphylococcal alpha-hemolysin, a heptameric transmembrane pore. Science, 274, 1859-1866.

Sonoda, Y., Cameron, A., Newstead, S., Omote, H., Moriyama, Y., Kasahara, M., Iwata, S., and Drew, D. (2010). Tricks of the trade used to accelerate high-resolution structure determination of membrane proteins. FEBS Lett, 584, 2539-2547.

136

Stroud, R. M., Miercke, L. J., O'Connell, J., Khademi, S., Lee, J. K., Remis, J., Harries, W., Robles, Y., and Akhavan, D. (2003). Glycerol facilitator GlpF and the associated aquaporin family of channels. Curr Opin Struct Biol, 13, 424-431.

Sulistijo, E. S., and MacKenzie, K. R. (2006). Sequence dependence of BNIP3 transmembrane domain dimerization implicates side-chain hydrogen bonding and a tandem GxxxG motif in specific helix-helix interactions. J Mol Biol, 364, 974-990.

Sulistijo, E. S., and Mackenzie, K. R. (2009). Structural basis for dimerization of the BNIP3 transmembrane domain. Biochemistry, 48, 5106-5120.

Tamm, L. K., Hong, H., and Liang, B. (2004). Folding and assembly of beta-barrel membrane proteins. Biochim Biophys Acta, 1666, 250-263.

Tate, C. G. (2001). Overexpression of mammalian integral membrane proteins for structural studies. FEBS Lett, 504, 94-98.

Therien, A. G., Grant, F. E., and Deber, C. M. (2001). Interhelical hydrogen bonds in the CFTR membrane domain. Nat Struct Biol, 8, 597-601.

Therien, A. G., and Deber, C. M. (2002a). Oligomerization of a peptide derived from the transmembrane region of the sodium pump gamma subunit: effect of the pathological mutation G41R. J Mol Biol, 322, 583-550.

Therien, A. G., and Deber, C. M. (2002b). Interhelical packing in detergent micelles. Folding of a cystic fibrosis transmembrane conductance regulator construct. J Biol Chem, 277, 6067- 6072.

Tulumello, D. V., and Deber, C. M. (2009). SDS micelles as a membrane-mimetic environment for transmembrane segments. Biochemistry, 48, 12096-12103.

Tulumello, D. V., and Deber, C. M. (2011a). Membrane protein folding in detergents. In H. Tschesche (Ed.), Methods in Protein Biochemistry (pp. 23-46). Berlin: De Gruyter.

Tulumello, D. V., and Deber, C. M. (2011b). Positions of polar amino acids alter interactions between transmembrane segments and detergents. Biochemistry, 50, 3928-3935.

Tulumello, D. V., and Deber, C. M. (2012). Efficiency of detergents at maintaining membrane protein structures in their biologically relevant forms. Biochim Biophys Acta, 1818, 1351- 1358.

Ulmschneider, M. B., Sansom, M. S., and Di Nola, A. (2005). Properties of integral membrane protein structures: derivation of an implicit membrane potential. Proteins, 59, 252-265.

Viklund, H., Granseth, E., and Elofsson, A. (2006). Structural classification and prediction of reentrant regions in alpha-helical transmembrane proteins: application to complete genomes. J Mol Biol, 361, 591-603.

137

Vincent, M., Gallay, J., Jamin, N., Garrigos, M., and de Foresta, B. (2007). The predicted transmembrane fragment 17 of the human multidrug resistance protein 1 (MRP1) behaves as an interfacial helix in membrane mimics. Biochim Biophys Acta, 1768, 538- 552.

Vinogradova, O., Sonnichsen, F., and Sanders, C. R., 2nd. (1998). On choosing a detergent for solution NMR studies of membrane proteins. J Biomol NMR, 11, 381-386.

von Heijne, G. (1992). Membrane protein structure prediction. Hydrophobicity analysis and the positive-inside rule. J Mol Biol, 225, 487-494.

Wahlstrom, A., Hugonin, L., Peralvarez-Marin, A., Jarvet, J., and Graslund, A. (2008). Secondary structure conversions of Alzheimer's Abeta(1-40) peptide induced by membrane-mimicking detergents. FEBS J, 275, 5117-5128.

Walkenhorst, W. F., Merzlyakov, M., Hristova, K., and Wimley, W. C. (2009). Polar residues in transmembrane helices can decrease electrophoretic mobility in polyacrylamide gels without causing helix dimerization. Biochim Biophys Acta, 1788, 1321-1331.

Walters, R. F., and DeGrado, W. F. (2006). Helix-packing motifs in membrane proteins. Proc Natl Acad Sci U S A, 103, 13658-13663.

Walther, D. M., Rapaport, D., and Tommassen, J. (2009). Biogenesis of beta-barrel membrane proteins in bacteria and eukaryotes: evolutionary conservation and divergence. Cell Mol Life Sci, 66, 2789-2804.

Warschawski, D. E., Arnold, A. A., Beaugrand, M., Gravel, A., Chartrand, E., and Marcotte, I. (2011). Choosing membrane mimetics for NMR structural studies of transmembrane proteins. Biochim Biophys Acta, 1808, 1957-1974.

Wegener, A. D., and Jones, L. R. (1984). Phosphorylation-induced mobility shift in phospholamban in sodium dodecyl sulfate-polyacrylamide gels. Evidence for a protein structure consisting of multiple identical phosphorylatable subunits. J Biol Chem, 259, 1834-1841.

Wehbi, H., Rath, A., Glibowicka, M., and Deber, C. M. (2007). Role of the extracellular loop in the folding of a CFTR transmembrane helical hairpin. Biochemistry, 46, 7099-7106.

Wehbi, H., Gasmi-Seabrook, G., Choi, M. Y., and Deber, C. M. (2008). Positional dependence of non-native polar mutations on folding of CFTR helical hairpins. Biochim Biophys Acta, 1778, 79-87.

Wei, Y., Kim, S., Fela, D., Baum, J., and Hecht, M. H. (2003). Solution structure of a de novo protein from a designed combinatorial library. Proc Natl Acad Sci U S A, 100, 13270- 13273.

Wei, Y., and Hecht, M. H. (2004). Enzyme-like proteins from an unselected library of designed amino acid sequences. Protein Eng Des Sel, 17, 67-75.

138

White, S. H., and Wimley, W. C. (1999). Membrane protein folding and stability: physical principles. Annu Rev Biophys Biomol Struct, 28, 319-365.

White, S. H., and von Heijne, G. (2008). How translocons select transmembrane helices. Annu Rev Biophys, 37, 23-42.

White, S. H. (2009). Biophysical dissection of membrane proteins. Nature, 459, 344-346.

Wimley, W. C., Creamer, T. P., and White, S. H. (1996). Solvation energies of amino acid side chains and backbone in a family of host-guest pentapeptides. Biochemistry, 35, 5109- 5124.

Wimley, W. C., and White, S. H. (1996). Experimentally determined hydrophobicity scale for proteins at membrane interfaces. Nat Struct Biol, 3, 842-848.

Wimley, W. C. (2002). Toward genomic identification of beta-barrel membrane proteins: composition and architecture of known structures. Protein Sci, 11, 301-312.

Wimley, W. C. (2003). The versatile beta-barrel membrane protein. Curr Opin Struct Biol, 13, 404-411.

Wimmer, R., Andersen, K. K., Vad, B., Davidsen, M., Molgaard, S., Nesgaard, L. W., Kristensen, H. H., and Otzen, D. E. (2006). Versatile interactions of the antimicrobial peptide novispirin with detergents and lipids. Biochemistry, 45, 481-497.

Yan, C., and Luo, J. (2010). An analysis of reentrant loops. Protein J, 29, 350-354.

Ye, S., Li, Y., and Jiang, Y. (2010). Novel insights into K+ selectivity from high-resolution structures of an open K+ channel pore. Nat Struct Mol Biol, 17, 1019-1023.

You, M., Li, E., Wimley, W. C., and Hristova, K. (2005). Forster resonance energy transfer in liposomes: measurements of transmembrane helix dimerization in the native bilayer environment. Anal Biochem, 340, 154-164.

Zen, K. H., McKenna, E., Bibi, E., Hardy, D., and Kaback, H. R. (1994). Expression of lactose permease in contiguous fragments as a probe for membrane-spanning domains. Biochemistry, 33, 8198-8206.

Zhou, F. X., Merianos, H. J., Brunger, A. T., and Engelman, D. M. (2001). Polar residues drive association of polyleucine transmembrane helices. Proc Natl Acad Sci U S A, 98, 2250- 2255.

139

Appendices Appendix 1: Supporting information for Chapter 2.

Figure A1.1 Representative individual Trp fluorescent spectra obtained in 34.7 mM (open

circles) or 8.7 mM SDS (closed circles) are shown for (A) AI10I17N (top two curves) and AI10

(bottom two curves) as well as (B) AI5I17N (top two curves) and AI5 (bottom two curves). For ease of comparison spectra obtain in 34.7 mM SDS have been normalized to those obtained in 8.7 mM and offset by 0.3 (panel A) or 0.4 (panel B) arbitrary fluorescence units. All spectra were background subtracted and emission corrected. In general, spectra obtained at the lower SDS concentration had the same shape as those obtained at higher concentration but had less noise due to decrease light scattering from the micelles. This allows for both a clearer observation of any fine structure in the spectra as well as a more accurate determination of emission maxima wavelength (used to generate Fig. 2.6A).

140

Critical micelle concentration determination.

Solutions of 10 mM Tris HCl (pH 8.0), 5 mM 1-anilino-8-naphthalene sulfonate (ANS), with and without 20 mM SDS were mixed to create a series of 400 μL samples that varied between an SDS concentration of 0-20 mM. ANS emission spectra were recorded on a Hitachi F-400 Photon Technology International C-60 fluorescence spectrometer in a 10 mm excitation, 1mm emission pathlength cuvette. Samples were excited at 370 nm using 5 nm slit widths and fluorescence emission was recorded between 450 and 550 nm using 5 nm slit widths. Fluorescent intensity at 500 nm was then plotted as a function of SDS concentration. CMC determination was also

performed in the presence of 10 μM of either the most hydrophobic (AI10) or least hydrophobic

(AI5I17N) peptide. In order to minimize the amount of peptide used for these experiments, fluorescence of a 2 mL solution of the aqueous buffer including 10 μM of peptide was first recorded in a 10 mm excitation/emission pathlength cuvette. The SDS concentration of this sample was then titrated with an identical sample containing 40 μM SDS. The total volume as well as concentration of peptide remained constant throughout the titration. Fluorescence spectra of these titrations were recorded following a minimum equilibration period of 8 minutes (with stirring). In order to determine the CMC, linear regression was used to calculate the equation of best-fit lines describing the SDS concentration dependence of fluorescence before and after the change in slope (indicative of the CMC). The range of linear regions was chosen to maximize the correlation coefficient. The CMC was calculated as the intercept of these two lines. In the absence of peptide there was no dependence of fluorescence upon SDS concentration prior to the CMC. In this case the mean fluorescence of all points prior to the increase was used to calculate the CMC. The standard error of the slope and intercept of the linear regression lines were obtained using R. These values were used to estimate the error of CMC determination through error propagation.

141

Figure A1.2. CMC determination by ANS fluorescence. Points used in the fitting of linear regions are shown as solid circles, while those points not used in fitting are shown as open circles. (A) The CMC obtained in the absence of peptide was determined to be 5.86 mM with an estimated error of 0.11 mM. (B) The CMC obtained in the presence of 10 μM of AI10 (the most hydrophobic peptide) was determined to be 5.63 mM with an estimated error of +/- 0.11mM. (C) The CMC obtained in the presence of 10 μM of AI5 I17N (the least hydrophobic peptide) was determined to be +/- 5.65 mM with an estimated error of +/- 0.89 mM. A decrease in ANS fluorescence upon increasing sub-CMC SDS concentrations is likely a result of ANS binding to the hydrophobic peptides, and subsequent displacement by SDS.

142

Appendix 2: Supporting information for Chapter 3.

Table A2.1. Hydrophobicity of model TM segments

ΔGwoct 2 ΔGwoct ΔGwif LD ΔGwoct 3 ΔGwif 4 ΔGins Polar LD ΔGwif ΔGins ΔGins 1 μHΦ μHΦ μHΦ residue LD μHΦ (kcal μHΦ (kcal μHΦ μHΦ (kcal -1 (kcal -1 (kcal position μ mol ) -1 mol ) -1 (deg) HΦ (deg) mol-1) (kcal mol ) mol ) (deg) (deg) mol-1)

WT -- 1.49 0.76 1 -1.4 5.83 -3 -1.96 1.23 -5 -0.55 2.11 -1

A11N 0 1.30 0.75 16 -1.05 5.76 0 -1.71 1.19 6 0.67 2.20 30

I12N +1 1.10 0.41 -17 0.57 4.09 -14 -1.23 0.65 -34 0.87 0.80 -57

A13N +2 1.30 0.86 -10 -1.05 6.03 -6 -1.71 1.39 -14 0.83 2.90 -22

A14N +3 1.30 0.91 9 -1.05 6.09 -1 -1.71 1.42 2 0.84 3.05 12

I15N +4 1.10 0.53 30 0.57 4.41 12 -1.23 0.76 28 0.77 1.31 45

A16N +5 1.30 0.68 -13 -1.05 5.68 -6 -1.71 1.15 -16 0.4 1.87 -31

A17N +6 1.30 0.93 -3 -1.05 6.17 -4 -1.71 1.48 -8 0.64 3.02 -7

A18N +7 1.30 0.81 14. -1.05 5.88 0 -1.71 1.28 6 0.39 2.39 19

I19N +8 1.10 0.37 2 0.57 3.87 -5 -1.23 0.51 -12 0.18 1.11 -4

A20N +9 1.30 0.81 -12 -1.05 5.92 -6 -1.71 1.32 -15 0.21 2.35 -18

A21N +10 1.30 0.86 -10 -1.05 6.16 -2 -1.71 1.46 -1 0.21 2.69 3

I12D5 +1 1.16 0.46 -12 3.36 2.37 -55 -0.42 0.67 -109 1.32 0.83 -88

I12N +2I +1 1.50 0.44 -72 -2.7 3.94 -60 -2.19 1.09 -94 0.06 1.41 -83

Scr. -- 1.49 0.17 -01 -1.4 1.66 -165 -1.96 0.98 8 -0.73 0.39 155 1Average hydrophobicity based upon the Liu-Deber (LD) scale (a value above 0.4 is predicted to be capable of membrane insertion) (Liu and Deber, 1998a). 2 Predicted free energy of water to lipid bilayer interfacial region partitioning (ΔGwif) (Wimley and White, 1996). 3 Predicted free energy of water to octanol partitioning (ΔGwoct) (Wimley et al., 1996). 4 Predicted free energies of translocon-mediated membrane insertion (ΔGins) (Hessa et al., 2007). 5Aspartic acid side chain is assumed to be in deprotonated form.

143

Figure A2.1. A) The correlation between SDS-PAGE gel shift of Asn containing peptide variants and change in helical secondary structure (assessed by circular dichroism) versus WT, p = 0.046. B) The same correlation, also including additional peptide variants described in Figure 3, p = 0.71.

144

Appendix 3: Supporting information for Chapter 4.

Figure A3.1. Circular dichroism spectra of transmembrane segments under investigation. Spectra are shown for TM segment peptides in aqueous buffer (blue); OG (gold); DDM (green); DPC (red); SDS (black). All spectra are the average of a minimum of three independent experiments.

145

Figure A3.2. Circular dichroism of transmembrane segments under investigation. Shown are MRE values representing a minimum of three replicates for each TM segment/detergent pairing. Error bars represent one standard deviation from the mean. Statistical differences were assessed using two-sided Students t-tests assuming equal variance; *indicates p <0.05, **indicates p < 0.01.

146

Figure A3.3. Tryptophan fluorescence of transmembrane segments under investigation. Spectra are shown for TM segments in aqueous buffer (blue); OG (gold); DDM (green); DPC (red); SDS (black). All spectra are the average of a minimum of three independent experiments, and have been normalized to the maximum emission in water.

147

Figure A3.4. Tryptophan fluorescence blue shifts of transmembrane segments under investigation. Shown are average blue shifts representing a minimum of three replicates for each TM segment/detergent pairing. Error bars represent one standard deviation from the mean. Statistical differences were assessed using two-sided Students t-tests assuming equal variance. *indicates p <0.05, **indicates p < 0.01.

148

Appendix 4: Table of oligonucleotides used in DNA library construction.

Table A4.1: Oligonucleotides used in library construction

Reaction 1 Forward Primer - NcoI restriction site

5’ GGG GGG CC ATG GCT AAA AAG AAG AAG AAA TAC GCG GCT NNN ATT GCT NNN ATT GCG TGG NNN ATC GCA 3’

Reaction 1 Reverse Primer - BamHI restriction site

5’ GGC CTT GGA TCC TGG CGA TTT AGC TGC AAT NNN TGC CGC NNN NN TGC GAT NNN CCA CGC AAT 3’

Reaction 2 Forward Primer - BamHI restriction site

5’ TCG CCA GGA TCC AAG GCC GCT ATC NNN GCA GCT NNN NNN GCC ATC NNN TGG GCA ATT NNN GCT ATC 3’

Reaction 2 Reverse Primer - XhoI restriction site

5’ CCC CCC CTC GAG TTT CTT TTT CTT TTT GTA TGC AGC NNN GAT AGC NNN AAT 3’

N denotes sites where the oligonucleotides were synthesized with equimolar ratios of all four bases incorporated. Sites of complementarity are underlined while restriction endonuclease sites are in bold.

149

Appendix 5: Detergent binding analysis of PFG-NMR diffusion studies.

By monitoring both the diffusion coefficient of the detergent and peptide simultaneously we are able to quantify the amount of detergent bound. By assuming all detergent molecules are in fast exchange with one another we can express the observed diffusion coefficient of the detergent

(DDet.obs) as a weighted average of the diffusion coefficients of the free detergent monomer

(DDet.free), self-bound detergent aggregates (DDet.det) and peptide bound (DDet.peptide) detergent fractions as follows,

DDet.obs = DDet.free ([Detergent]free/[Detergent]T) + DDet.det ([Detergent]det/[Detergent]T) +

DDet.peptide ([Detergent]peptide/[Detergent]T)

Equation A5.1

The diffusion coefficient of the bound detergent should be equivalent to that of the peptide itself, while the diffusion coefficient of the free (monomer) state can be approximated from the initial titration points of the detergent control. However, the diffusion coefficients of the detergent molecules associating with one another are not directly measurable, and are expected to be dependent on detergent concentration even prior to the CMC due to the formation of pre- micellular aggregates. This ternary system is difficult to fit but can be simplified to a binary one if we assume the contribution of the free and detergent bound states (i.e. the non-peptide bound states) is reflected by the diffusion coefficient of the detergent control at an equivalent concentration,

DDet.obs = DDet.np ([Detergent]np/[Detergent]T) + DDet.peptide ([Detergent]peptide/[Detergent]T)

Equation A5.2

where DDet.np is the diffusion coefficient of the non-peptide bound detergent. Using this equation, the concentrations were determined iteratively by continuously refining the value of

DDet.np through minimization the difference in concentration of non-peptide bound detergent from the control curve versus that determined from the equation. The ratio of the concentration

150 of peptide-bound detergent to the concentration of peptide then yielded the number of number of detergents bound to the peptide allowing for the construction of a peptide saturation plot. The latter points in the saturation curve were used to estimate the maximum amount of detergent bound to the peptide.

151

Figure 1.5 has been adapted from Rath, A., Tulumello, D. V., and Deber, C. M. (2009b). Peptide models of membrane protein folding. Biochemistry, 48, 3036-3045. Copyright 2009, American Chemical Society.

Chapter 2 has been, in part, adapted with permission from Tulumello D.V., and Deber C.M. (2009). SDS micelles as a membrane-mimetic environment for transmembrane segments. Biochemistry. 48(51); 12096-12103. Copyright 2009, American Chemical Society.

Chapter 3 has been, in part, adapted with permission from Tulumello D.V., and Deber C.M. (2011). Positions of polar amino acids alter interactions between transmembrane segments and detergents. Biochemistry. 50(19); 3928-3935. Copyright 2011, American Chemical Society.

Chapter 4 has been, in part, adapted with permission from Tulumello D.V., and Deber C.M. (2012). Efficiency of detergents at maintaining membrane protein structures in their biologically relevant forms. Biochimica et Biophysica Acta – Biomembranes, 1818(5), 1351-1358. Copyright 2012, Elsevier.