KINETIC STUDIES OF THE FOLDING OF THE MEMBRANE

Amjad Farooq

Department of Biochemistry Imperial College London

A THESIS SUBMITTED TO THE UNIVERSITY OF LONDON FOR THE DEGREE OF DOCTOR OF PHILOSOPHY

1997

ABSTRACT

Stop-flow fluorescence and photodiode array spectroscopy, with a millisecond time resolution, are used here to study the refolding kinetics of the membrane protein bacteriorhodopsin. Refolding is initiated by mixing partially denatured apoprotein bacterioopsin in SDS with DMPC/CHAPS micelles. The kinetics of micelle mixing can be differentiated from the kinetics of protein folding. The results indicate that the apoprotein bacterioopsin can be refolded in the micelles to a state with a native-like secondary structure. The is not required for the formation of this molten globule-like state. Further, apoprotein refolding is observed to conform to biphasic kinetics with lifetimes in the second range. This observation is taken to imply the involvement of parallel branched pathways in the refolding of the apoprotein bacterioopsin to a molten globule-like state. The mechanism of retinal binding to the apoprotein bacterioopsin is also elucidated. Two bimolecular reactions involving retinal and bacterioopsin are detected. This implies that retinal binds in a parallel fashion to the apoprotein bacterioopsin. The product of these bimolecular reactions is a spectroscopically distinct intermediate that absorbs maximally around 430 nm. The formation of this 430-nm intermediate occurs with lifetimes in the order of hundreds of milliseconds. The formation of a molten globule-like state of apoprotein is thus evidently rate-limiting for the subsequent binding of retinal. The oscillator strength of the retinal absorption band in free retinal and in the 430-nm intermediate appears to be the same. This finding is interpreted as an evidence that the retinal chromophore is non-covalently bound to the apoprotein within the 430-nm intermediate. A free energy change of - 30 kJmol-1 is determined for the formation of the 430-nm intermediate. The decay of the 430-nm intermediate into the native protein bacteriorhodopsin occurs via at least three parallel alternative routes with lifetimes ranging from several seconds to several minutes. This process is characterized by the formation of a Schiff base between the aldehyde group of retinal and the e-amino group of Lys216. A kinetic model is suggested for the folding of bacteriorhodopsin from a partially denatured state in SDS. The results of this study should have profound implications on the mechanisms of folding of within biological membranes as well as in water.

ACKNOWLEDGMENTS

I express my sincere thanks to Jim Barber for a critical appraisal of the work, for support and advice, and for the highly comforting and productive time that I had in his laboratory whilst preparing this manuscript.

I am grateful to Steve Matthews for his support and encouragement, and for the many exciting and stimulating discussions.

I owe a special thanks to Jeremy Baum for his advise and support.

I thank Paula Booth for her practical help with the stop-flow spectroscopy, and Martin Bell for providing the kinetic analysis software Look.

I express my warm-hearted gratitude to Bee Khoo and Maria Conte for their friendship and company that I enjoyed meticulously.

I offer my profound thanks to my family and, in particular, my sister Nahid, uncle Fazal and, parents Fatima and Mohammed for their understanding and appreciation.

Finally, I wish to record a very special tribute to Peter Fryer (Bradford College) for it is a credit to him above all else that I have had the privilege to study for a PhD.

This work was funded by the BBSRC.

Amjad Farooq London 1997

CONTENTS

Title 1

Abstract 3

Acknowledgments 4

Abbreviations 6

Symbols 8

List of Tables 9

List of Figures 10

Chapter 1: Introduction 15

Chapter 2: Materials and Methods 65

Chapter 3: Equilibrium Studies of Bacteriorhodopsin Refolding 91

Chapter 4: Kinetics of Bacterioopsin Refolding 111

Chapter 5: Kinetics of Retinal Binding to Bacterioopsin 129

Chapter 6: Kinetics of Retinal Binding at Various Times in the Refolding of Bacterioopsin 171

Chapter 7: Dependence of Refolding Kinetics of Bacteriorhodopsin on Retinal Concentration 189

Chapter 8: Conclusions 215

References 235

ABBREVIATIONS

ANS 8-anilino-1-naphthalenesulphonate

ATP adenosine triphosphate

ATPase adenosine triphosphatase

AU absorbance unit bO bacterioopsin bO/SDS partially denatured bO in SDS bO¢ partially refolded bO in DMPC/CHAPS bR bacteriorhodopsin

BPTI bovine pancreatic trypsin inhibitor

BSE bovine spongiform encephalopathy

CD circular dichroism

CHAPS 3-[(3-cholamidopropyl)dimethylammonio]-1-propanesulphonate

CJD creutfeld-jacob disease

Csp cold shock protein

DHFR dihydrofolate reductase

DHPC L-a-1,2-dihexanoylphosphatidylcholine

DMPC L-a-1,2-dimyristoylphosphatidylcholine

DNA deoxyribonucleic acid

DNAse deoxyribonuclease

ER endoplasmic reticulum

FKBP FK506 binding protein

FTIR fourier transform infra-red spectroscopy

FWHM full width half maximum (of an absorption or fluorescence band)

I intermediate (partially folded state)

LHCII -harvesting complex II mRNA messenger RNA

N native state

NMR nuclear magnetic resonance

Omp outer membrane protein

PAGE polyacrylamide gel electrophoresis

PDA photodiode array spectroscopy

PM purple membrane

R all-trans-retinal

RNA ribonucleic acid

RNAse ribonuclease

SDS sodium dodecylsulphate

SRP signal recognition particle

SVD singular value decomposition

2D two-dimensional

3D Three-dimensional

TEMED N,N,N¢,N¢-tetramethylethylenediamine

Tris Tris(hydroxymethyl)aminomethane

U unfolded state

UV ultraviolet

SYMBOLS

a experimentally observed amplitude e extinction coefficient k intrinsic rate constant

K equilibrium constant l wavelength n experimentally observed rate constant

R universal molar gas constant t lifetime t time

T absolute temperature

LIST OF TABLES

Table 3.1. Comparison of chromophore absorbance and protein fluorescence maxima and their FWHM for bacteriorhodopsin refolding at pH 6 and pH 8 102

Table 3.2. Dependence of bacteriorhodopsin refolding yield on pH and the mode of retinal addition 103

Table 4.1. Comparison of experimentally observed rate constants and their amplitudes for fluorescence and absorbance refolding measurements of bacterioopsin 121

Table 5.1. Comparison of experimentally observed rate constants and their amplitudes obtained from fluorescence data for the refolding of bacterioopsin in the absence of retinal and for the reaction of retinal with bacterioopsin 152

Table 5.2. Relative oscillator strength of the retinal absorption band during the course of the reaction of retinal with bacterioopsin 154

Table 5.3. Comparison of experimentally observed rate constants obtained from fluorescence and absorbance data for the reaction of retinal with bacterioopsin 156

Table 7.1. Intrinsic rate constants, equilibrium constants and free energy changes for the phases associated with reactions involving retinal and bacterioopsin 206

Table 8.1. Folding intermediates of bacteriorhodopsin: Comparison of the various nomenclatures used in the previous studies and this thesis 223

LIST OF FIGURES

Fig 1.1. A schematic diagram illustrating the concept of protein folding 18

Fig 1.2. Simplified free energy profiles of protein folding 21

Fig 1.3. Simplified energy landscapes of protein folding 28

Fig 1.4. Simplified funnels of protein folding 29

Fig 1.5. Cis-trans isomerization about proline imide bond 30

Fig 1.6. Integral membrane protein topology 33

Fig 1.7. The two-stage model of membrane protein folding 37

Fig 1.8. A simplified diagram showing a typical halobacterium 40

Fig 1.9. Chemical structure of a typical membrane lipid of halobacterium 41

Fig 1.10. Chemical structure of retinal and retinal-protein Schiff base 43

Fig 1.11. A ribbon diagram of bacteriorhodopsin 46

Fig 1.12. A diagram of bacteriorhodopsin showing amino acids lining the retinal binding pocket 47

Fig 1.13. All-trans and 13-cis configurations of retinal in bacteriorhodopsin 48

Fig 1.14. Resonance structures of the protonated Schiff base of retinal 50

Fig 1.15. Photocycle of bacteriorhodopsin 51

Fig 1.16. A general strategy for bacteriorhodopsin refolding in vitro 53

Fig 1.17. Chemical structures of SDS, CHAPS and DMPC 54

Fig 1.18. A model for the structure of mixed DMPC/CHAPS micelles 56

Fig 2.1. Absorption spectrum of all-trans-retinal in ethanol at room temperature 67

Fig 2.2. Absorption spectra of purple membrane isolated from Halobacterium salinarium and delipidated bacterioopsin 71

Fig 2.3. SDS-PAGE analysis of purple membrane and delipidated bacterioopsin 73

Fig 2.4. Stop-flow mixing circuit 78

Fig 2.5. The Xenon lamp intensity profile 80

Fig 3.1. Dependence of refolding yield of bacteriorhodopsin on bO preparation (pH 6) 94

Fig 3.2. Dependence of refolding yield of bacteriorhodopsin on the mode of retinal addition (pH 6) 97

Fig 3.3. Dependence of refolding yield of bacteriorhodopsin on the mode of retinal addition (pH 8) 100

Fig 4.1. Addition of DMPC/CHAPS to bO/SDS 114

Fig 4.2. Addition of DMPC/CHAPS to SDS 118

Fig 5.1. Addition of retinal/DMPC/CHAPS to bO/SDS 134

Fig 5.2. Addition of retinal/DMPC/CHAPS to bO/DMPC/CHAPS 144

Fig 6.1. Addition of retinal at various times in the refolding of bO 174

Fig 7.1. Bacteriorhodopsin refolding under limiting amounts of retinal 192

Fig 7.2. Bacteriorhodopsin refolding in excess retinal 198

Fig 8.1. A paradigm for bacteriorhodopsin folding 217

Fig 8.2. Energetics of Bacteriorhodopsin folding 221

IN THE NAME OF ALLAH THE MOST BENEFICENT, THE MOST MERCIFUL

THE ALL KNOWING, THE ALL SEEING

Chapter 1

INTRODUCTION

1.1 Opening Remarks 16

1.1 Protein Folding 18 1.2.1 What is Protein Folding? 18 1.2.2 Why Study Protein Folding? 19 1.2.3 Protein Folding Problem 20 1.2.4 Experimental Studies of Protein Folding 22 1.2.5 Classical Models of Protein Folding 24 1.2.6 New View of Protein Folding 26 1.2.7 Proline Isomerization 30

1.3 Membrane Proteins 32 1.3.1 Technical Problems of Membrane Proteins 32 1.3.2 Structural Diversity of Membrane Proteins 33 1.3.3 Experimental Studies of Membrane Protein Folding 35 1.3.4 Two-Stage Model of Membrane Protein Folding 36

1.4 Bacteriorhodopsin 40 1.4.1 Halobacteria 40 1.4.2 Retinal Proteins 43 1.4.3 Structure of Bacteriorhodopsin 45 1.4.4 Dark-Light Adaptation of Bacteriorhodopsin 48 1.4.5 Shift of Bacteriorhodopsin 49 1.4.6 Photocycle of Bacteriorhodopsin 51 1.4.7 Refolding Studies of Bacteriorhodopsin 53 1.4.8 Models of Bacteriorhodopsin Folding 57

1.5 Summary 61

16 Chapter 1

1.1 Opening Remarks

The research presented in this thesis sets out to elucidate the mechanisms of protein folding. How proteins fold remains a mystery and, in particular, in the case of membrane proteins. The intriguing feature of the work undertaken here is that the main focus is on the folding of proteins within the hydrophobic environment of biological membranes. The photoactive protein bacteriorhodopsin is employed here as a model system for studying membrane protein folding. Bacteriorhodopsin is a small integral membrane protein that acts as a proton pump in the cell membrane of halobacteria. Its seven-transmembrane helical topology is characteristic of many cellular proteins and receptors. A retinal chromophore is covalently bound to the protein within the seven-helical bundle. The biological activity of bacteriorhodopsin is dependent upon the photoisomerization of its retinal chromophore. This is also a feature of many proteins involved in cellular in that their activity is dependent upon the binding of a small ligand molecule. Bacteriorhodopsin thus shares many structural and functional characteristics of a wide variety of proteins. A number of features make bacteriorhodopsin an attractive candidate for studying membrane protein folding. Three-dimensional structure of bacteriorhodopsin is known to near-atomic resolution. Bacteriorhodopsin can be quantitatively refolded from a denatured state in mixed phospholipid/detergent micelles. The refolding assay is simple and based on the recovery of a native-like chromophore absorption band. More specifically, the emphasis of this work is on studying the refolding kinetics of bacteriorhodopsin and, in particular, to elucidate the mechanism of retinal binding. This is achieved by using rapid stop-flow mixing method, with a millisecond time resolution, to initiate bacteriorhodopsin refolding in mixed DMPC/CHAPS micelles from a partially denatured state in SDS. Refolding is monitored by intrinsic protein fluorescence and retinal absorbance. Bacteriorhodopsin is the first integral membrane protein for which refolding kinetics are being studied in detail. This work should unequivocally provide an insight into the mechanisms of membrane protein folding and further stimulus for research in this important area of protein biochemistry. In order to fully appreciate the impact of this study on protein folding research, a brief review of the literature is imperative. This chapter therefore effectively kicks off

Introduction 17 with a section on protein folding. The scope of the problem is highlighted. The experimental and theoretical strategies employed in this area are outlined. The current views held on the folding mechanisms of water-soluble proteins are discussed, and where necessary, evidence from specific model systems is presented. The next section concentrates on membrane proteins. The structural and functional diversity that exists among this class of hydrophobic proteins is described. The reasons that make this area an experimentalist’s nightmare are mentioned. How membrane proteins might fold is discussed. A separate section is devoted to bacteriorhodopsin. This section begins with the biology of halobacteria and retinal proteins to put bacteriorhodopsin into a broader context. Next the structure and function of bacteriorhodopsin is discussed. What little is known about bacteriorhodopsin folding is described. Finally, a summary is provided to sum up this chapter and to introduce the subsequent chapters in this thesis.

18 Chapter 1

1.2 Protein Folding

1.2.1 What is Protein Folding? Proteins are involved in important biological functions in the form of enzymes, transport and signalling molecules, and have a structural role in life. To be biologically active, proteins must take up a specific 3D shape. The mechanism by which a mRNA molecule is synthesized from a DNA template is called “transcription”. Subsequent synthesis of a polypeptide chain by ribosomes using mRNA as a template is referred to as “translation”. The mechanism by which a polypeptide chain acquires a biologically active 3D structure has come to be known as “protein folding” (Fig 1.1). The mechanisms of

H2N

H N 2 ?

COOH COOH

unfolded native

Fig 1.1. A schematic diagram illustrating the concept of protein folding.

transfer of information in transcription and translation are well understood, at least in outline. In contrast, the mechanisms of protein folding that involve transfer of information from one dimension to three dimensions remain largely elusive. Although reports that proteins could be refolded after denaturation have been appearing since the 1930s (Anson 1945), it was not until the early 1960s that the elegant

Introduction 19 experiments of Christian Anfinson on RNAse A put protein folding on the map of scientific research (Anfinson et al 1961). Little progress was made in the field of protein folding over the next two decades. The advent of modern molecular biology techniques in the 1980s rekindled further interest in this imporatnt area of protein biochemistry. Only recently, however, significant experimental and theoretical advances have been made in understanding the mechanisms of protein folding. With the development of stop-flow NMR (Balbach et al 1995, Hoeltzli & Frieden 1996, Liu et al 1996) and methods for initiating protein refolding on a submillisecond scale (Jones et al 1993, Williams et al 1996, Eaton et al 1997), there is now real optimism for unravelling the folding code.

1.2.2 Why Study Protein Folding? Protein folding, like many other disciplines in biochemistry, was originally pursued on pure academic grounds. The relevance of protein folding studies to other biological processes such as protein synthesis, degradation and transport in the body cannot be overemphasized. The main impetus for protein folding research, however, comes from its potential to combat disease. Mutations within the polypeptide sequence may cause incorrect folding of a protein. This phenomenon is termed “protein misfolding”. Protein misfolding is the cause of a number of potentially fatal diseases including cancer. It has also been linked to amyloid and prion disease. Amyloid disease results from misfolding of a protein into fibrils (hundreds of nm in length) that have a characteristic cross-b repeat structure (Lansbury et al 1995, Kelly 1996). Deposition of amyloid fibrils in vital organs and tissues is usually fatal by the fifth decade. Alzheimer’s disease is a classical example of amyloid that leads to dementia and ultimate death. Recently, Ile56Thr and Asp67His variants of human lysozyme have been reported to be involved in amyloid disease (Pepys et al 1993). Prion disease involves conversion of a protein conformation from a-helical to b-sheet secondary structure as a result of specific point or insertional mutations within the prion protein sequence (Pan et al 1993, Baldwin et al 1995, Harrison et al 1997). Examples of prion disease include neurodegenerative disorders such as Creutzfeld-Jacob disease (CJD), bovine spongiform encephalopathy (BSE) and scrapie in sheep.

20 Chapter 1

Mutations within serpins (serine protease inhibitors) have been shown to be associated with protein conformational changes that can lead to disease (Sifers 1995). Night blindness (Fahmy et al 1996) and retinitis pigmentosa (Rao & Oprian 1996), degeneration of , are also attributed to a set of mutations within the human protein sequence. Another interesting example of protein misfolding linked to disease is the connective tissue protein collagen, where mutations within the protein sequence are responsible for osteogenesis imperfecta and Ehlers-Danlos syndrome (Prockop & Kivirikko 1995). Osteogenesis imperfecta is a disease where bones become brittle and fragile, and Ehlers-Danlos syndrome is associated with the rupture of vascular arteries. Protein folding studies can thus be justified on the grounds that they hold promise for rectifying the cause of protein misfolding. A better understanding of protein folding should lead to rational development of therapeutic approaches for the treatment of disease. Protein folding studies should also have profound implications in biotechnology, where refolding of recombinant proteins poses considerable practical difficulties (Pain 1987, Hlodan et al 1991). Expression of foreign genes in bacterial hosts often results in the production of polypeptides that spontaneously coalesce to form aggregates called “inclusion bodies” (Marston 1986). Recovery of functionally active proteins from these inclusion bodies is a pre-requisite for their subsequent medicinal use. A large number of cellular proteins including insulin and interleukin-2 have been successfully expressed in bacterial cells, although the fraction of the expressed protein that can be refolded is far from optimal in many cases (Williams et al 1982, Devos et al 1983).

1.2.3 Protein Folding Problem Many proteins can refold in vitro after complete denaturation of their tertiary structure. This suggests that the 3D shape of a protein must be determined by its primary sequence (Anfinson 1973). How the information stored in a protein’s primary sequence is used to construct its tertiary structure is referred to as the “protein folding problem”. Solution of this problem requires a knowledge of the kinetics and thermodynamics of the folding process (Fig 1.2). The kinetics involve the measurement of the rates of all the steps from

Introduction 21 the unfolded to the folded state. The thermodynamics are concerned with the stability of the native state and any intermediates relative to the unfolded state. Characterization of all the species involved in the folding process in terms of their structure at an atomic level is also an obligatory requirement. Ideally, one must also understand the physico-chemical basis of why a particular given

(a) ts

U Free energy N

(b) tsN tsI

U I

N

Reaction coordinate

Fig 1.2. Simplified free energy profiles of protein folding. (a) Protein folding without accumulation of intermediates. The free energy maximum represents the transition state (ts), and the free energy minimum the native state (N). The unfolded state is denoted by U. (b) Protein folding with a detectable intermediate (I). The free energy maxima indicate the transition states: one for the formation of the intermediate from the unfolded protein (tsI), and one for the decay of this intermediate into the native protein (tsN).

22 Chapter 1 sequence leads to the construction of a particular 3D structure, instead of some other. The ultimate goal in protein folding research is to be able to predict the 3D structure of a protein given its amino acid sequence. Although many proteins can refold in vitro spontaneously, it is now widely recognized that other cellular proteins termed “chaperones” may assist folding in vivo (Martin & Hartl 1993, Frydman et al 1994, Rassow & Pfanner 1996, Martin & Hartl 1997). The role that these chaperones play is a subsidiary one and is mainly concerned with increasing the efficiency of folding. These chaperones interact with the unfolded polypeptide chains and increase the yield of productive folding by suppressing off-pathway reactions. The physiological significance of this role can be appreciated given the high concentrations of proteins and other macromolecules in the cell, which would strongly favour protein aggregation. One of the well-studied chaperones is the GroEL, which is found within the bacterial cytosol. GroEL belongs to a family of heat shock proteins which mediate folding in ATP-dependent reactions (Clark 1996, Hartl & Martin 1995). GroEL is a porous cylinder composed of two heptameric toroids of protein subunits that are stacked back-to-back (Braig et al 1994). The central cavity of GroEL is thought to provide an ideal environment in which a polypeptide chain can fold without interference from other cellular activities (Itzhaki et al 1995b, Mayhew et al 1996).

1.2.4 Experimental Studies of Protein Folding Equilibrium studies only provide information on the denatured, partially folded or native states of a protein. In order to obtain information on the transient states or intermediates in protein folding, time- resolved measurements are necessary. These typically involve rapid-mixing techniques such as stop-flow in conjunction with spectroscopy. Optical probes such as fluorescence and absorption provide information on the environment of aromatic side-chains during folding. Circular dichroism (CD) can be used to monitor secondary structure formation (far-UV region) as well as changes in the environment of aromatic residues (near-UV region). Examples of cases in which extrinsic probes have been used to follow folding are also well documented. An interesting example is the hydrophobic fluorescent dye

Introduction 23

ANS, which is often used to detect the formation of non-polar surfaces during protein folding. An increase in the hydrophobicity of the surrounding environment leads to an enhancement in the fluorescence of ANS and vice versa. The appearance of non-polar surfaces during folding is believed to result from the formation of a compact state with native secondary structure but fluctuating tertiary interactions (Ohgushi et al 1983, Ikeguchi et al 1986, Baldwin 1990, Varley et al 1993). This state has been termed the “molten globule” (Ptitsyn et al 1990). The conventional stop-flow methods at best have a time resolution of a few milliseconds. However, many fundamental processes in protein folding occur over a much faster time scale. These include the initial polypeptide collapse and formation of secondary structure elements such as a-helices and b-sheets which are thought to occur on a time scale of ns to ms (Creighton 1992, Williams et al 1996). Techniques, however, have been recently developed to study protein folding in the submillisecond region. These are based on the initiation of folding by ultrarapid mixing and optical triggering methods (Eaton et al 1996, Plaxco & Dobson 1996, Eaton et al 1997). Using these techniques, the refolding studies of a number of proteins including cytochrome c (Pascher et al 1996, Takahashi et al 1997, Yeh et al 1997), apomyoglobin (Ballew et al 1996a, Ballew et al 1996b), barstar (Nolting et al 1995, Nolting et al 1997) and RNAse A (Phillips et al 1995) have been carried out with microsecond to nanosecond time resolution. The conformational probes based on fluorescence, absorption or CD only measure the average behaviour of the protein and hence provide no information at the atomic level. The experimental data are usually approximated to one or more kinetic phases. Each kinetic phase is described by a single exponential term with an observed rate and an amplitude. The parameters of the exponential term do not correspond to any realistic species but represent an average of all the microscopic states. An exponential term essentially implies a first-order process. To what extent protein folding reactions conform to this process in reality is questionable. Methods, however, have been developed which can provide information at an atomic level during protein folding. These include hydrogen exchange pulse labelling and protein engineering techniques. In hydrogen exchange pulse labelling, the exchange of the backbone and sidechain amide proton with deuterium is measured using 2D NMR

24 Chapter 1 spectroscopy (Schmid & Baldwin 1979, Kim & Baldwin 1980, Roder 1989, Englander et al 1996). If the exchange is carried out at various times in the refolding using a stop-flow method, it is possible to obtain detailed structural information. This approach has been used to characterize partially structured intermediates in terms of amides involved in hydrogen-bonded interactions in a number of proteins including RNAse A (Udgaonkar & Baldwin 1990) and lysozyme (Radford et al 1992). In protein engineering studies, formation of sidechain interactions is followed during folding. The strategy is based upon substitution of an amino acid residue by another (with minimal disturbance to the protein structure) so as to remove a desired sidechain interaction. The effect of such a substitution on the stability of intermediate and ground states (unfolded and the native) of the mutant protein relative to that of the wild type can provide information regarding the stage at which a particular interaction is formed during folding (Matouschek et al 1989, Matouschek et al 1990, Fersht et al 1992, Fersht 1994, Fersht 1995a). Such experiments have had tremendous success in mapping out the structures of transition states and intermediates in the folding of chymotrypsin inhibitor-2 (Otzen et al 1994, Itzhaki et al 1995a) and barnase (Serrano et al 1992, Matouchek et al 1992, Fersht 1993). Thus, unlike hydrogen exchange method which largely measures backbone secondary structure, protein engineering has the power to provide information on almost all sidechain tertiary interactions. To what extent the substitution of an amino acid residue by another deviates the protein structutre is however a controversial issue in protein engineering studies. More recently, stop-flow NMR has been developed to obtain high resolution structural information directly during protein folding, although the time resolution of this method is at best seconds (Balbach et al 1995, Hoeltzli & Frieden 1996, Liu et al 1996).

1.2.5 Classical Models of Protein Folding How does a protein fold to its native conformation? In the late 1960s, Cyrus Levinthal argued that if a protein were to fold to its native structure by random search through all its possible conformations, the time required would be on an astronomical scale (Levinthal 1968). This is the result of the rather large conformational space available to

Introduction 25 the fully unfolded polypeptide chain. The ideal unfolded protein is the random coil, in which rotations about backbone and sidechain atoms can generate a very large number of conformations (Ramachandran & Sasisekharan 1968). The odds of finding the native fold among all possible conformations are thus huge. In practice, however, most proteins fold to their native conformation in less than a second. There must therefore exist a mechanism that allows a protein to access only a small number of its possible conformations and reach the native state on a biologically meaningful time scale. A number of theoretical models have been proposed to account for this so-called “Levinthal’s paradox” (Wetlaufer 1973, Dill 1985, Harrison & Durbin 1985, Zwanzig et al 1992). The “framework” model postulated that secondary structure elements are formed first from an unfolded protein and subsequently associate to form the tertiary structure (Baldwin 1989, Kim & Baldwin 1990, Karplus & Weaver 1994). Indeed time-resolved CD studies indicate that the change in the far-UV signal, which monitors secondary structure formation, is predominantly complete within the first few milliseconds of refolding. In contrast the changes in the near-UV signal, which monitores tertiary structure formation, occurs over a much slower time scale (Labhardt 1986, Gilmanshin & Ptitsyn 1987, Kuwajima et al 1987, Elove et al 1992). These observations lend compelling support to the view that secondary structure formation precedes tertiary interactions. The framework mechanism inherently invokes the involvement of intermediates during folding. Intermediates have indeed been detected in many proteins including barnase (Bycroft et al 1990, Kippen & Fersht 1995), RNAse A (Udgaonkar & Baldwin 1988, Udgaonkar & Baldwin 1990, Udgaonkar & Baldwin 1995), ubiquitin (Khorasanizadeh et al 1996), cytochrome c (Roder et al 1988) and BPTI (Creighton & Goldenberg 1984, Creighton 1990). The framework model also predicts that the secondary structure elements would be expected to be stable in isolation from the remainder of the protein structure. This, for example, has been observed in the case of N-terminal a-helix of RNAse A (Kim & Baldwin 1984). The “nucleation” model proposed that some region within the unfolded polypeptide chain would initially form native secondary structure that would act as a starting point (nucleus) for the subsequent formation of tertiary structure (Wetlaufer

26 Chapter 1

1973, Wetlaufer 1990). No nuleation sites have been found in proteins. Recent reports indicate that these may be formed by interactions between residues that are distant in sequence, unlike the classical view in which the local residues were thought to constitute the folding nucleus. In this respect, the nucleus may be a much more expanded structure maintained by secondary as well as tertiary interactions and the need for folding intermediates may not be a strict requirement in nucleation mechanisms (Fersht 1997). The “hydrophobic-collapse” model hypothesized that the unfolded polypeptide chain undergoes a rapid collapse to a state that shields hydrophobic groups from the aqueous solvent and the subsequent tertiary interactions would propagate as a result of this compactness (Dill 1985, Gregoret & Cohen 1991). The collapsed state is now thought to be the molten globule. Experimental support for the hydrophobic-collapse mechanisms in protein folding comes from a number of proteins including a- lactalbumin (Ikeguchi et al 1986, Ptitsyn et al 1990), b-lactamase (Thomas et al 1983), lysozyme (Chaffotte et al 1992, Morozova et al 1995) and DHFR (Garvey et al 1989, Christensen & Pain 1991) in which a molten globule is detected. Indeed, the molten globule state is thought to be a general intermediate in protein folding (Ptitsyn et al 1990).

1.2.6 New View of Protein Folding The classical models of protein folding presented above imply the involvement of specific intermediates and folding pathways. Recent experimental and theoretical studies of protein folding however suggest that intermediates may not be necessary, at least in the case of small proteins. The experimental observations are that some proteins fold via a simple two-state process with no accumulation of intermediates, implying that the formation of secondary and tertiary structure occurs in parallel (Miranker & Dobson 1996, Fersht 1997). Examples include chymotrypsin inhibitor-2 (Jackson & Fersht 1991, Otzen et al 1994, Fersht 1995a), the cold-shock protein CspB (Schindler et al 1995), monomeric l repressor (Huang & Oas 1995a, Huang & Oas 1995b) and acyl-coenzyme A binding protein (Kragelund et al 1995). Although intermediates have been detected in the

Introduction 27 folding of lysozyme (Kiehhaber 1995) and cytochrome c (Sosnik et al 1994, Sosnik et al 1996), these proteins have also been shown to employ an alternative two-state mechanism. The theoretical studies of protein folding are based on the representation of the polypeptide chain as a string of beads on a three-dimensional cubic lattice (Karplus & Shakhnovich 1992, Sali et al 1994a, Karplus & Sali 1995, Shakhnovich 1997). These models ignore sidechain interactions but satisfy other criteria of real proteins such as closeness of two residues which may be distant in sequence. The idea is to find the native fold among all possible conformations based on the fulfilment of thermodynamic and kinetic requirements. The thermodynamic requirement is that the native state is the most stable with a global free energy minimum. The kinetic requirement is that folding is achieved on a biologically meaningful time scale. Such computer simulations are supportive of two-state mechanisms in protein folding, although one must take great precaution in their interpretation as these models are based on short peptides and small proteins (Sali et al 1994b, Gutin et al 1995). As a result of recent theoretical and experimental findings, a new view on protein folding has emerged (Baldwin 1994, Baldwin 1995, Dill & Chan 1997). The new view considers protein folding as ensembles or distributions of conformations rather than specific structures or intermediates. Emphasis is placed upon multiple folding routes rather than specific sequential pathways. The new ideas are based on “energy landscapes” and “folding funnels” (Bryngelson et al 1995, Wolynes et al 1995, Onuchic et al 1995). An energy landscape is the free energy of a protein as a function of its conformational space (entropy) (Fig 1.3). The conformational entropy undergoes a massive reduction as the protein refolds from the unfolded to the native state. The native conformation lies at the global energy minimum. A folding funnel is the representation of an energy landscape in multidimensions (Fig 1.4). The vertical axis in a folding funnel is the free energy and the many lateral axis represent the conformational space. The relevance of the funnel to the folding of real proteins arises from the fact that the unfolded state is highly heterogenous representing an ensemble of conformations rather than a specific structure. Such an ensemble of conformations is a subset of the unfolded state U which is represented by the rim of the funnel. The many conformations in U are envisaged to fold via independent routes to reach the global

28 Chapter 1

U Free energy

N

U

I I

N

Conformational entropy

Fig 1.3. Simplified energy landscapes of protein folding. Protein folding without any intermediates (top) and with a detectable intermediate (bottom). energy minimum (the bottom of the funnel) corresponding to the native state N. This concept is analogous to the way water trickles down a funnel - it always comes out at the same point at the bottom of the funnel regardless of where it starts at the top. The surface of the funnel may be smooth such that the unfolded molecules reach the bottom of the funnel directly without any significant obstacles. This situation would be characteristic of proteins in which no ephemeral intermediates are observed (two- state kinetics). Alternatively the surface of the funnel may be rough and bumpy, giving rise to the valleys, in the case of other proteins. The unfolded molecules will have to overcome these barriers on their way to the bottom of the funnel. In their quest for such a journey,

Introduction 29

U Free energy

N Conformational entropy

Conformational entropy U

I I

N

Fig 1.4. Simplified funnels of protein folding. Protein folding without any intermediates (top) and with a detectable intermediate (bottom). some molecules may become trapped in the valleys and may have to take an alternative route or perhaps travel back to the top of the funnel and start all over again. The bumpiness of the folding funnel would thus explain the role of intermediates in some proteins (multi-state kinetics). These statistical mechanical ideas thus invoke thinking about protein folding as a distribution of structures from the unfolded to the native state rather than being confined to a well-defined pathway, as it was originally envisaged in Levinthal’s paradox (Levinthal 1968).evan

30 Chapter 1 e 1.2.7 Proline Isomerization The folding kinetics of nearly all proteins studied contain fast phases, in the millisecond to second time range, and slow phases, in the several seconds to several minutes time range. Although the chemical events that accompany fast phases remain obscure, one of the reasons for slow refolding reactions is believed to be cis-trans isomerization about proline imide bonds (Brandts et al 1975, Nall 1994) (Fig 1.5). In an unfolded polypeptide chain, the proline imide bond exists in a mixture of native and non- native conformers. Upon refolding, those proline imide bonds that are in

polypeptide + N trans H O O polypeptide

polypeptide O

polypeptide + N cis H O

Fig 1.5. Cis-trans isomerization about proline imide bond.

Introduction 31 a non-native conformation must isomerize to the native conformation before the polypeptide chain can refold. The cis-trans isomerization about proline imide bond is intrinsically a slow process with a lifetime of about a minute (Brandts et al 1975). Proline isomerization has therefore been assigned to the slow phases observed in the refolding reactions of a number of proteins including RNAse A (Mayr et al 1996, Houry & Scheraga 1996), RNAse T1 (Odefey et al 1995) and FKBP (Veeraraghavan et al 1996).

32 Chapter 1

1.3 Membrane Proteins

1.3.1 Technical Problems of Membrane Proteins Membrane proteins play vital cellular roles ranging from cell communication and transport of molecules across membranes to the synthesis of ATP. Yet, very little is known about the structure, stability and folding of membrane proteins relative to their water-soluble counterparts. This is in part due to the technical problems associated with membrane proteins. The hydrophobic nature of this class of proteins makes their isolation and purification an extremely difficult task. Expression of membrane proteins in bacterial or yeast hosts is a tricky pursuit - expressed protein may often reside in the host cell membrane rather than being secreted into the external medium. Moreover, membrane proteins tend to be highly prone to irreversible denaturation and aggregation. Isolation of membrane proteins in large quantities for structural studies is almost next to impossible. Although 2D crystals can be obtained for a number of membrane proteins (Kuhlbrandt et al 1992, Kuhlbrandt et al 1994, Henderson et al 1990, Grigorief et al 1996), it is extremely notorious to obtain 3D crystals for higher resolution structure determination using electron and X-ray diffraction methods (vide infra). More recently, a priori approaches such as single-particle analysis have been developed which hold a great deal of promise for predicting the 3D structures of membrane proteins in the future (Henderson 1995). Solubilization and reconstitution of membrane proteins often requires a detergent or a lipid system. These procedures limit the application of a number of techniques for their biophysical study. Thus, in the presence of detergents or lipids, membrane proteins reorient too slowly for multi- dimensional solution NMR methods to be effective. Solid-state NMR methods, however, have been developed to overcome some of these technical problems (Cross & Opella 1994). The fact that, in some membrane proteins, the lipid bilayer can be substituted by organic solvents such as chloroform has led to the application of solution NMR in high resolution structure determination of membrane proteins. This method, for example, was successfully used in the structural study of the c subunit of the membrane domain (F0) of ATPase (Moody et al 1987). Whether the protein structure in the organic solvent is the same as that in the lipid bilayer remains to be addressed. Further, as in the case of water-soluble

Introduction 33 proteins, NMR methods are only applicable to small membrane proteins. Despite these practical difficulties and the fact that investigators in general are reluctant to enter the tough world of membrane proteins, significant advances have been witnessed in this important area over the last few years.

1.3.2 Structural Diversity of Membrane Proteins Membrane proteins occur within the hydrophobic milieu of the lipid bilayer. They may traverse the bilayer once or several times (Fig 1.6). The regions of the protein that are immersed in the interior of the bilayer are predominantly hydrophobic and contain few polar or charged residues. In fact, a membrane protein can be identified from its amino acid sequence alone by the presence of uninterrupted stretches of 20-25 non-polar residues. The portions of the protein that interact with the external aqueous environment are rich in polar or charged amino acid residues. Examples of proteins which traverse the bilayer once include the erythrocyte membrane protein glycophorin, cytochrome f, and many receptors involved in cellular

(a) (b)

Fig 1.6. Integral membrane protein topology. (a) single-transmembrane domain, and (b) multi- transmembrane domain.

34 Chapter 1 signal transduction. Examples of proteins which traverse the bilayer several times include transmembrane channels such as glucose transporter, ion pumps such as bacteriorhodopsin and virtually all G proteins. These so-called integral membrane proteins are distinguished from peripheral membrane proteins in that the latter only interact with the surface of the lipid bilayer via ionic interactions, or in some cases the protein may be anchored to the bilayer by a covalent linkage. Examples of membrane proteins only partially inserted into the bilayer have not yet been found. Structure per se may not hold all the clues to the function but it remains central to the understanding of many fundamental properties of a protein including its stability and folding. Although atomic resolution structures of a large number of water-soluble proteins are available, the same argument cannot be made for membrane proteins. Unlike NMR methods which have become a routine with water-soluble proteins, crystallography reigns supreme in the quest for membrane protein structure determination. The first moderately high resolution structure obtained for an integral membrane protein was that of bacteriorhodopsin in the 1970s (Henderson & Unwin 1975). To date, high resolution structures of only a handful of membrane proteins are known. These include bacterial photosynthetic reaction centre (Deisenhofer et al 1985, Allen et al 1987, Komiya et al 1988, Chang et al 1991, Ermler et al 1994), bacterial light-harvesting complex (McDermott et al 1995, Karrasch et al 1995, Fyfe & Cogdell 1996), plant light-harvesting antenna complex (Kuhlbrandt et al 1994), photosystem I (Krauss et al 1993, Chitnis et al 1995, Fromme 1996, Jansson et al 1996, Krauss et al 1996), cytochrome c oxidase (Iwata et al 1995, Tsukihara et al 1995, Tsukihara et al 1996, Ostermeier et al 1996), cytochrome bc complex (Xia et al 1997), prostaglandin H synthase (Picot et al

1994, Picot & Garavito 1994) and the peripheral domain (F1) of ATPase (Abraham et al 1994, Fillingame 1996). The transmembrane regions in all of these proteins are in the form of an a-helix, although it is not clear whether this is also true in the case of ATPase. High resolution structural studies of bacterial porins indicate that their transmembrane parts consist entirely of b-sheets that are exquisitely arranged in an antiparallel fashion to form a sixteen- stranded b-barrel (Weiss et al 1991, Weiss &

Introduction 35

Schulz 1992, Cowan et al 1992, Kreusch et al 1994, Cowan et al 1995, Schulz 1996). It is, however, possible that membrane proteins with mixed ab transmembrane secondary structures exist in nature. Indeed, preliminary structural studies on acetylcholine receptor suggest that the protein contains a bundle of five a-helices surrounded by an outer rim of three-stranded antiparallel b-sheets (Unwin 1993, Unwin 1995). In general, transmembrane motifs based on a-helix tend to be more hydrophobic than b- sheets. Thus unlike helical integral membrane proteins, b-barrel integral membrane proteins are usually soluble in water when unfolded.

1.3.3 Experimental Studies of Membrane Protein Folding Membrane protein folding is becoming an area of increasing significance. Folding studies of membrane proteins, for example, have an important relevance to cellular processes such as protein translocation across membranes and membrane biogenesis (Wickner et al 1979, Blobel 1980, Wickner et al 1988, Cowan & Rosenbusch 1994, Borel & Simon 1996). These studies also hold the potential to the development of better approaches for drug delivery to specific tissues (Bychkova et al 1988, Wickner et al 1991, Schatz & Dobberstein 1996). Unlike water-soluble proteins, however, little is known about the folding or insertion of proteins into membranes (Jahnig 1983). In vitro refolding has been achieved for only a few integral membrane proteins including bacteriorhodopsin (Huang et al 1980, Huang et al 1981, London & Khorana 1982), plant light-harvesting complex II (Plumley & Schmidt 1987, Paulsen et al 1990, Kohorn & Auchincloss 1991), bacterial porins (Dornmair et al 1990, Eisele & Rosenbusch 1990, Surrey & Jahnig 1992), diacylglycerol kinase (Sanders et al 1996) and the M13 procoat protein (Soekarjo et al 1996). Although these integral membrane proteins can refold spontaneously in vitro, the situation in vivo may be different. As in the case of water-soluble proteins, chaperonin-like proteins are thought to be involved in directing membrane protein assembly in vivo (Lecker et al 1990, Engelman 1996). Many membrane proteins are synthesized in vivo with an N-terminal leader sequence termed “signal peptide”. The signal peptide is typically about 20 amino acids in length and consists of a hydrophobic core flanked by basic amino acid residues.

36 Chapter 1

The signal peptide in conjunction with signal recognition particle (SRP) aids insertion of proteins into the membrane of endoplasmic reticulum (ER). This process occurs concurrently with the synthesis of polypeptide chain by the ribosomes on the surface of ER. SRP is a large multi-subunit protein within the ER membrane. The signal peptide is cleaved off from the remainder of the protein once insertion into the membrane is complete. Subsequently the protein may be transported from the ER membrane in a vesicular vehicle to the cell membrane and other destinations. In addition to SRP, a number of other proteins have been identified that may also be involved in the assembly and translocation of proteins across membranes (Schatz & Dobberstein 1996). The main problem with studying the kinetics of folding of membrane proteins arises from their insolubility in water. As a result, in the folding studies of membrane proteins, the starting state is a detergent-solubilized protein in which some secondary structure may be present. Bacterial porins, however, are an exceptional case in that they can be solubilized in water upon denaturation with urea. This is a consequence of the low hydrophobicity of b-barrel relative to an a-helix. To date, folding kinetics of only four integral membrane proteins have been studied. These include bacteriorhodopsin (Booth et al 1995), plant light-harvesting complex II (Booth & Paulsen 1996) and the porins OmpA (Surrey & Jahnig 1995, Kleinschmidt & Tamm 1996) and OmpF (Surrey et al 1996). The refolding strategy of these integral membrane proteins is based upon mixing the SDS-solubilized or urea- denatured protein with lipid-based micelles or vesicles. Such studies have led to observation of multiphasic kinetics, implying that intermediates are likely to be involved in the folding of membrane proteins.

1.3.4 Two-Stage Model of Membrane Protein Folding The constraints imposed by the lipid bilayer matrix lead to a reduction in the conformational space available to a polypeptide chain (von Heijne & Manoil 1990, Donnelly et al 1993, Tocanne et al 1994). Indeed, an a-helix is observed to be the dominant structural motif in majority of the membrane proteins studied. Further, the cleavage of a polypeptide chain into two or more fragments corresponding to transmembrane regions has no observable effect on the tertiary organization of many polytopic membrane proteins within the lipid bilayer. These considerations led to the

Introduction 37 proposition of a two-stage model for the folding of helical integral membrane proteins (Popot & Engelman 1990, Popot 1993, Lemmon & Engelman 1994) (Fig 1.7). The first stage involves the formation of helices across the hydrophobic region of the lipid bilayer independently of each other. In the second stage, the transbilayer helices interact with one another to generate the tertiary and quarternary structure of the protein. This model parallels the framework model postulated for the folding of water-soluble proteins, in which the formation of secondary structure elements is envisaged to precede tertiary interactions (Baldwin 1989, Kim & Baldwin 1990, Karplus & Weaver 1994). Transmembrane helices are much more stable than their water-soluble counterparts. This difference in stability primarily results from the fact that whereas

formation of Stage I transmembrane helices

association Stage II of helices

Fig 1.7. The two-stage model of membrane protein folding.

38 Chapter 1 helices in water-soluble proteins have the option of hydrogen bonding with the solvent, the same is not true in the case of transmembrane helices. The energetic cost of breaking an hydrogen bond within the hydrophobic environment of the membrane is much greater than that in water (Allen 1975, Li & Deber 1994, Honig & Yang 1995). In fact, an a-helix in water is only marginally more stable than the unfolded polypeptide, whereas the formation of a transmembrane helix is concomitant with the release of free energy of hundreds of kJ per mole (Engelman & Steitz 1981, Lemmon & Engelman 1994). The pronounced stability of an a-helix within the membrane has led to the view that transmembrane helices could be considered as autonomous folding domains (Popot & Engelman 1990, Popot 1993). The mechanism by which transmembrane helices are formed is, however, not clear. It is possible that a polypeptide may first form an helix in the aqueous environment prior to insertion into the membrane or alternatively, helix formation may occur within the membrane following insertion of a polypeptide (Lemmon & Engelman 1994). It is also conceivable that insertion and transmembrane helix formation may be coupled. Whatever the exact mechanism of transmembrane helix formation, the driving force responsible for this process is the hydrophobic effect - the exclusion of hydrophobic residues from the surrounding aqueous environment, or the exclusion of hydrophilic residues from the surrounding hydrophobic environment of the lipid bilayer. Helix association within the membrane is thought to be mainly governed by interhelical interactions (Lemmon & Engelman 1994, Lemmon et al 1994). These may include electrostatic or van der Waals forces. Hydrogen bonding between polar residues or formation of salt bridges between charged residues on adjacent helices could, for example, drive helix association within the membrane (Engelman 1982, Honig & Hubbel 1984). Indeed, several polar residues are present in the transmembrane helices of bacteriorhodopsin (Henderson et al 1990, Grigorieff et al 1996) and the photosynthetic reaction centres (Diesenhofer et al 1985, Allen et al 1987, Komiya et al 1988), and a salt bridge between Glu and Arg on adjacent helices is observed in the atomic model of plant light- harvesting antenna complex II (Kuhlbrandt et al 1994). Other factors such as ligand binding, folding of interhelical loops and bilayer rearrangement are thought to contribute little to the close packing of transmembrane helices.

Introduction 39

This two-stage model is purely based on thermodynamic arguments and in no way does it have any implications on the folding mechanism of membrane proteins. Although demonstration that a number of proteins including bacteriorhodopsin (Liao et al 1983, Popot et al 1987, Kahn & Engelman 1992), lactose permease (Wrubel et al 1990, Bibi & Kaback 1990, Zen et al 1994), the b-adrenergic receptor (Kobilka et al 1988), a voltage-dependent Na+ channel (Stuhmer et al 1989), the yeast a-factor transporter STE6 (Berkower & Michaelis 1991) and adenylate cyclase (Tang et al 1991) can be refolded from their fragments corresponding to transmembrane regions is a compelling evidence in support of the two-stage model, it is possible that some membrane proteins may not conform to this mechanism given their diversity. Further advances in high resolution structural studies of membrane proteins and an increase in the knowledge of membrane protein folding should provide a stern test for the generality of the two- stage model.

40 Chapter 1

1.4 Bacteriorhodopsin

1.4.1 Halobacteria Bacteriorhodopsin was discovered as the sole protein of the purple membrane of Halobacterium salinarium (formerly H halobium) in the early 1970s (Osterhelt & Stoeckenius 1971). H salinarium belongs to the group halobacteria (halophiles) which also includes the strains H cutirubrum, H trapanicum and H marismortui. The halobacteria (DasSarma & Fleischmann 1995) and the two other major groups methanogens (Sowers & Schreier 1995) and the thermophiles (Robb & Place 1995) are prokaryotes that belong to the newly recognized kingdom of archaebacteria (Woese & Fox 1977). The archaebacteria are highly heterogenous and phylogenetically distinct from the eubacteria. Some of their distinguishing properties include membrane lipids that contain ether-linked instead of the ester-linked hydrocarbons to the glycerol moiety that are found in all other biological systems. The cell wall of archaebacteria surrounding the cell (plasma) membrane is chemically distinct from that of eubacterial walls. Other differences are found in the transcriptional and translational apparatus of archaebacteria and eubacteria (Stanier et al 1987). Halobacteria are typically polarly-flagellated rod-shaped cells (Fig 1.8). The flagellum confers motility upon the cell. Their cell wall consists mainly of glycoprotein

gas vacuole

chromosome cell membrane

cell wall ribosome ..

flagellum

Fig 1.8. A simplified diagram showing a typical halobacterium.

Introduction 41 and, unlike eubacterial walls, no lipids, peptidoglycans or teichoic acids have been found (Stoeckenius & Rowen 1967, Stoeckenius & Kunau 1968). The lipids in the cell membrane of halobacteria consist of phytane chains linked to the glycerol moiety via ether linkage (Kates et al 1965, Kates et al 1982) (Fig 1.9). With the exception of membranes surrounding the gas vacuoles, no other intracellular membranes are present in halobacteria.

O

OPO -2 O 3

phytane chain glycerol

Fig 1.9. Chemical structure of a typical membrane lipid of halobacterium. The phosphate group may be substituted by a glycan.

Halobacteria require saturating salt concentrations (4-5 M) for optimal growth. Their cells lyse upon exposure to low salt concentrations or water. Moreover, their biochemical machinery functions effectively only at high salt concentrations. The primary source of energy for halobacteria is respiration, in which generation of a proton gradient across the cell membrane is used to drive ATP synthesis. Under aerobic conditions, the cell membrane contains the respiratory apparatus as well as carotenoids, which give the membrane its characteristic red colour. The carotenoids protect the cell from photochemical damage by absorption of high intensity light. Under conditions of low oxygen, however, halobacteria synthesize a modified cell membrane - the synthesis of bacteriorhodopsin is induced within certain regions of the cell (red) membrane, referred to as the “purple membrane”. Purple membrane is laid

42 Chapter 1 in patches of about 0.5 mm in diameter and can account for up to half the total membrane area. Bacteriorhodopsin within the purple membrane is organized into a two-dimensional hexagonal lattice, with each unit of the hexagon containing a trimer of bacteriorhodopsin molecules (Blaurock 1982), and accounts for nearly 75 % of the total membrane weight. In response to light, bacteriorhodopsin pumps protons across the cell membrane. The resuting thermodynamic potential is utilized by an ATPase to synthesize ATP. Bacteriorhodopsin thus generates an alternative source of ATP under conditions when oxygen tension is low in the environment. Halobacteria, however, cannot grow under complete anaerobic conditions. This is because there is a strict requirement for the biosynthesis of its chromophore retinal from the precursor b-carotene. The switch to purple membrane synthesis thus occurs only if the partial pressure of oxygen is low enough to induce the synthesis of bacteriorhodopsin but high enough to allow retinal biosynthesis. In addition to bacteriorhodopsin, the synthesis of two other proteins is also induced within the cell membrane of halobacteria in response to low oxygen tension. These include (Stoeckenius & Bogomolni 1982, Stoeckenius 1985, Lanyi 1986, Oesterhelt & Tittor 1989) and sensory rhodopsin I (Bogomolni & Spudich 1982, Perazzona et al 1996). Halorhodopsin acts as a light-driven chloride pump whereas the sensory rhodopsin I allows the halobacterial cell to move to areas of optimal light intensity under conditions where the distribution of light is uneven. The chloride transport is normally driven by the proton gradient during respiration but it is apparently more efficient to couple the light energy directly to chloride transport rather than use the proton gradient generated by bacteriorhodopsin. Yet, another protein is constitutively synthesized within the cell membrane of halobacteria. This so-called sensory rhodopsin II (Wolff et al 1986, Marwan & Oesterhelt 1987) protects halobacterial cells from harmful radiation under conditions where oxygen supply is not limiting and respiration is operative. Sensory rhodopsin II functions by causing halobacterial cells to move away from areas of high intensity light and thus behaves in an antagonistic manner to sensory rhodopsin I. In brief, the four proteins bacteriorhodopsin, halorhodopsin, sensory rhodopsin I and II act in a concerted fashion to allow halobacteria to use light energy efficiently.

Introduction 43

1.4.2 Retinal Proteins Bacteriorhodopsin, halorhodopsin, sensory I and II belong to a family of proteins in which retinal is covalently attached to the e-amino group of a Lys residue within the apoprotein (opsin) via a Schiff base (Fig 1.10). Other members of this family include the mammalian visual rhodopsin (Hargrave et al 1993, Schertler 1993) and the visual pigments of other and invertebrates (Junge & Kakitani 1986, Yoshizawa & Kuwata 1991, Pepe & Cugnoli 1992, Mathies & Kakitani 1992, Hao & Fong 1996). Retinal proteins are small to medium size integral membrane proteins (20-

O

retinal

H N 2 Lys

opsin

N

opsin retinal-protein Schiff base

Fig 1.10. Chemical structures of retinal and retinal-protein Schiff base.

44 Chapter 1

50 kD) with a seven-helical transmembrane topology, which is also characteristic of -coupled receptors (Donnelly & Findlay 1994). The biological activity of retinal proteins is dependent upon the isomerization of retinal chromophore in response to light. In the mammalian visual rhodopsin found within the rod cells of retina, for example, the photoisomerization of 11-cis retinal to all-trans retinal induces a conformational change within the opsin that in turn interacts with the G protein leading to its activation (Birge 1990, Hargrave et al 1993, Rao & Oprian 1996). Activation of transducin initiates an enzymatic cascade that results in the closure of cation channels leading to hyperpolarization of the membrane and subsequent neuronal signalling. This activity is terminated by the dissociation of rhodopsin-transducin complex by the action of a number of other enzymes. In order for rhodopsin to perform another cycle, retinal has to be reisomerized to its initial state. Retinal however cannot be reisomerized whilst bound to the protein. Instead all-trans retinal is released from the retinal binding pocket by hydrolysis of the retinal-protein Schiff base, and replaced by 11-cis-retinal. The binding of a small ligand molecule, albeit noncovalently, is also responsible for the activation of other G protein-coupled receptors (Savarese & Fraser 1992, Rao & Oprian 1996), which mediate a large number of cellular signals. Rhodopsin is the only known G protein-coupled receptor in which the ligand is covalently bonded.

1.4.3 Structure of Bacteriorhodopsin Bacteriorhodopsin consists of a single polypeptide chain of 248 amino acid residues (referred to as bacterioopsin) to which a retinal is covalently attached to Lys216 via a Schiff base (Bayley et al 1981, Gerber & Khorana 1982). In vivo synthesis of bacteriorhodopsin occurs as a precursor with an N- terminal leader sequence of 13 amino acids and an aspartic acid at the C-terminus, both of which are removed upon insertion into the membrane (Seehra & Khorana 1984). Bacteriorhodopsin contains all but cysteine and histidine residues of the 20 naturally occuring amino acids. Nearly 70 % of amino acid residues in bacteriorhodopsin are hydrophobic.

Bacteriorhodopsin is a small integral membrane protein (Mr ~ 26 kD) with its C-terminus on the cytoplasmic side and the N-terminus on the outside of the cell

Introduction 45 membrane (Gerber et al 1977, Zingsheim et al 1978). Like visual rhodopsin, bacteriorhodopsin traverses the membrane seven times in the form of a-helices, which are designated A through to G from the N- to the C-terminus (Henderson & Unwin 1975, Engelman et al 1980, Engelman et al 1982, Renthal et al 1987). The Schiff base between retinal and the protein is found within the helix G. Electron cryo-microscopy of 2D crystals of purple membrane has led to the determination of a high resolution structure for bacteriorhodopsin with a resolution of 3.5 Å in the plane of the membrane but lower in the perpendicular direction (Henderson et al 1990, Grigorieff et al 1996) (Fig 1.11). In this model, the beginning and the end points of transmembrane a-helices are accurate to within one residue and it has been suggested that many helices begin with one or two residues in the conformation of a 310 helix at the N-terminus. The interhelical loops are thought to be highly disordered structures. More recently, high resolution structural studies of the 2D crystals of the purple membrane down to 3.0 Å have been achieved (Kimura et al 1997). In this study, the loops have been shown to be in a b conformation. A salt bridge is envisaged between Arg134, which is buried in the membrane near the extracellular surface of helix E, and Glu194 within the loop FG. Unlike the structure of plant light- harvesting complex II (Kuhlbrandt et al 1994), there is as yet no evidence for transmembrane salt bridges in bacteriorhodopsin. The transmembrane seven-helix bundle lies roughly perpendicular to the plane of the membrane although the helices B, C and F are kinked due to the presence of a Pro residue in each of these helices (there are 11 Pro residues in total). The kinks are suggested to play important structural role in the close packing of retinal and in the formation of the proton channel (Mogi et al 1989, Grigorieff et al 1996). Helix G also appears to be slightly bent near Lys216. The other three helices are generally straight. Bacteriorhodopsin contains 32 aromatic amino acids of which there are 13 Phe, 11 Tyr and 8 Trp residues. These are mainly located within six of the seven transmembrane helices. Helix D contains no aromatic residues. There are 34 charged amino acid residues in bacteriorhodopsin, ten of which are positioned within five of the seven transmembrane helices although nearly all helices begin or end with a charged residue. There are no charged residues within the core regions of helices A and E. The angle between the polyene sidechain of the retinal chromophore and the membrane

46 Chapter 1

COOH

H2N

Fig 1.11. A ribbon diagram of bacteriorhodopsin (Grigorieff et al 1996). The amino acid residues from 7 through to 227, for which the atomic coordinates are available, are coloured according to their chemical properties. The various colours denote Leu, Val and Ile (green), Asp and Glu (bright red), Lys and Arg (blue), Ser and Thr (orange), Phe and Tyr (mid blue), Asn and Gln (cyan), Gly (light grey), Ala (dark grey), Met (yellow), Trp (pink) and Pro (flesh). The retinal chromophore is denoted by “sticks” within the seven-helix bundle.

Introduction 47 normal appears to be about 70o on the extracellular side. At least 12 amino acid residues, contributed by all seven transmembrane helices, constitute the so-called retinal binding pocket within the protein (Henderson et al 1990, Feng et al 1994, Han & Smith 1995, Grigorieff et al 1996) (Fig 1.12). These include four Trp, three Tyr, three Asp and two Met residues.

Fig 1.12. A diagram of bacteriorhodopsin showing amino acids lining the retinal binding pocket (Grigorieff et al 1996). The residues indicated are four Trp (pink), three Tyr (mid blue), three Asp (bright red) and two Met (yellow). The retinal chromophore is shown in black in the center of the pocket.

48 Chapter 1

1.4.4 Dark-Light Adaptation of Bacteriorhodopsin Bacteriorhodopsin exhibits the phenomenon of dark-light adaptation (Oesterhelt et al 1973). In the dark, bacteriorhodopsin contains retinal in a mixture of all-trans and 13-cis configurations (Fig 1.13). Upon illumination, however, this so-called dark-adapted

O

all-trans

O

13-cis

Fig 1.13. All-trans and 13-cis configurations of retinal in bacteriorhodopsin.

form converts to the light-adapted form in which retinal is solely in the 13-cis configuration. The optical properties of the dark- and light-adapted forms of bacteriorhodopsin are distinct. Thus, for example, the absorption maximum of the dark-adapted form is slightly blue shifted relative to that of the light-adapted form which absorbs maximally around 570 nm (Braiman et al 1987, Popot et al 1987).

Introduction 49

1.4.5 Opsin Shift of Bacteriorhodopsin The light absorbing properties of bacteriorhodopsin in the visible region are due to its chromophore retinal, the main absorption transition of which lies at 380 nm although a number of smaller transitions have also been observed in the UV region (Becker et al 1971, Schaffer et al 1974). Although bacteriorhodopsin absorbs maximally at 560 nm in the visible region, the free protonated Schiff base of retinal has an absorption maximum of 440 nm in a solvent such as methanol (Pitt et al 1955, Maeda et al 1984, Liu et al 1993). The change in the absorption maximum of the free protonated Schiff base from 440 nm to 560 nm in the native chromophore is known as the “opsin shift” (Nakanishi et al 1980, Beppu & Kakitani 1994, Hu et al 1994). The change in the absorption maximum of retinal in the free protonated Schiff base and when bound to the apoprotein bacterioopsin is explained in terms of the p electron delocalization (Honig & Ebrey 1982). In the protonated form of the Schiff base, the positive charge on nitrogen becomes delocalized throughout the p electron system of the polyene chain of retinal (Fig 1.14). Such electron delocalization is believed to induce a red shift in the absorption maximum of polyenes like retinal (Labhart 1957). Any mechanism that increases the stability of the delocalized structures (delocalization of the positive charge on nitrogen) would lead to an increase in the absorption maximum. Theoretical calculations indicate that a free protonated Schiff base of retinal should absorb around 600 nm (Blatz & Mohler 1975, Honig et al 1976). Yet, a solution of free protonated Schiff base (such as chloride of retinylbutylimine in methanol) absorbs around 440 nm. This difference in theoretical and the experimental value is thought to arise from the interaction of a counterion (chloride) with the positive charge on the nitrogen. The effect of a counterion is to prevent the delocalization of positive charge on the nitrogen. This leads to a blue shift in the absorption maximum of a free protonated Schiff base relative to that expected in theory. Within bacterioopsin, the negatively charged residues Asp85 and/or Asp212 (Marti et al 1991, Needleman et al 1991) act as counterions to the retinal protonated Schiff base. However, polar or negatively charged amino acid residues near the cyclohexene ring of retinal are thought to act as counterions to stabilize the positive charge away from the nitrogen (Honig et al 1976, Honig et al 1979). According to this

50 Chapter 1

+ H N

R

H N +

R

enhanced p electron delocalization

Fig 1.14. Resonance structures of the protonated Schiff base of retinal. The structures in which the positive charge is delocalized away from the nitrogen induce a red shift in the absorption maximum of retinal. R is an alkyl group. so-called external point charge model (Nakanishi et al 1980), the stabilization of the positive charge near the cyclohexene ring induces a red shift in the absorption maximum of the free protonated Schiff base from 440 nm to 560 nm. The counterions responsible for the stabilization of positive charge near the cyclohexene ring of retinal are believed to be Trp and Tyr residues found within the retinal binding pocket (Beppu & Kakitani 1994). Asp115 is found to be near the cyclohexene ring of retinal and thus has been suggested to be the most likely charged residue that may act as a counterion to the delocalized positive charge on nitrogen (Henderson et al 1990). In addition to protein-chromophore interactions, a number of other factors are also believed to contribute to the opsin shift. These include exciton coupling between aromatic amino acid residues and retinal (Ebrey et al 1977, Kakitani et al 1985) as well as the twisting of the conjugated bonds of the chromophore retinal (van der Steen et al 1986).

Introduction 51

1.4.6 Photocycle of Bacteriorhodopsin Bacteriorhodopsin acts as a light-driven proton pump in halobacteria (Stoeckenius et al 1979, Khorana 1988, Krebs & Khorana 1993). Upon absorption of a photon of light, the light-adapted form of bacteriorhodopsin undergoes a series of reactions (Lozier et al 1975, Sharkov et al 1985, Polland et al 1986b). This so-called “photocycle” involves a number of spectroscopically distinct intermediates that are arbitrarily assigned J through to O (Mathies et al 1991, Oesterhelt et al 1992, Rothschild 1992, Lanyi 1993) (Fig 1.15).

bR hn

J610

O640

K590 extracellular side cytoplasmic side

N 520 L550

H+ M H+ 410

cell membrane

Fig 1.15. Photocycle of bacteriorhodopsin. The subscripts to the letters J through to O, representing photocycle intermediates, indicate wavelength maxima in nm. The light-adapted form of bacteriorhodopsin (bR) absorbs maximally around 570 nm.

52 Chapter 1

The first step in the photocycle involves photoisomerization of retinal, covalently attached to Lys216 (helix G) via a Schiff base, from all-trans to 13-cis configuration (Braiman & Mathies 1982). This induces a series of conformational changes within the protein such that the protonated Schiff base between retinal and Lys216 releases its proton directly to Asp85 (helix C) in the L to M step (Braiman et al 1988, Mogi et al 1988, Gerwert et al 1990). Asp85 is approximately 4 Å from the Schiff base on the extracellular side. It serves as a counterion to the Schiff base as well as a proton acceptor (Henderson et al 1990). The precise mechanism by which the proton is transferred from Asp85 to the extracellular side is not known, although a priori the involvement of a number of other acidic residues has been suggested (Zimanyi et al 1992, Brown et al 1995, Grigorieff et al 1996). Asp85 remains protonated until the last step of the photocycle. The reprotonation of the Schiff base in the M to N step is thought to occur by the transfer of a proton from Asp96 (helix C) (Fodor et al 1988, Gerwert et al 1990). Asp96 is about 12 Å from the Schiff base on the cytoplasmic side and thus a direct proton transfer is ruled out (Henderson et al 1990). The Schiff base reprotonation must therefore occur either via the intermediacy of other groups or alternatively, some sort of conformational change within the protein may be involved that draws the Schiff base and Asp96 closer together. Asp96 reprotonation (Zimanyi et al 1993) and retinal reisomerization to all-trans configuration (Smith et al 1983) are believed to occur during the N to O step. Finally, the deprotonation of Asp85 (Pfefferle et al 1991) returns bR to its ground state, with the net result of the transfer of a single proton from the cytoplasmic to the extracellular side of the membrane. A similar photocycle for the activity of halorhodopsin is also proposed (Stoeckenius 1985, Lanyi 1986, Oesterhelt & Tittor 1989). Unlike rhodopsin whose function is dependent upon the activation of a G protein (Birge 1990, Hargrave et al 1993, Khorana 1993, Rao & Oprian 1996), the halobacterial ion pumps bacteriorhodopsin and halorhodopsin do not involve interaction with other proteins during their photocycles.

Introduction 53

1.4.7 Refolding Studies of Bacteriorhodopsin Bacteriorhodopsin shares the honour of not only being the first integral membrane protein for which a moderately high resolution structure was obtained (Henderson & Unwin 1975), but it also became the first integral membrane protein for which in vitro refolding was established (Huang et al 1980). Bacteriorhodopsin can be refolded in a variety of detergent and lipid-based systems. Typically, the refolding strategy of bacteriorhodopsin involves solubilization of the apoprotein bacterioopsin in SDS after removal of endogenous lipids and retinal from the purple membrane (Fig 1.16).

purple membrane

delipidation and removal of retinal

delipidated bO

solubilization in SDS

bO/SDS (partial secondary structure)

add phospholipid/detergent micelles

bO/micelles (full native secondary structure)

add all-trans retinal

bR

Fig 1.16. A general strategy for bacteriorhodopsin refolding in vitro.

54 Chapter 1

Bacterioopsin denatured in SDS contains about 60 % of the secondary structure found in the native state of bacteriorhodopsin (London & Khorana 1982). Addition of mixed phospholipid/detergent micelles, such as DMPC/CHAPS, to partially denatured bacterioopsin in SDS restores full native secondary structure. Subsequent addition of all-trans retinal to this mixture regenerates a native-like chromophore with an absorption maximum at 560 nm (Huang et al 1981, London & Khorana 1982, Bayley et al 1982). Fig 1.17 shows the chemical structures of SDS, CHAPS and DMPC. Bacteriorhodopsin can also be refolded from a completely denatured state. Transfer of delipidated bacterioopsin directly into formic acid or trifluoroacetic acid

- OSO3

SDS

O + - OH N N SO H 3

HO OH CHAPS

O

O + O O O P N - O O O

DMPC

Fig 1.17. Chemical structures of SDS, CHAPS and DMPC.

Introduction 55 completely denatures the protein as demonstrated by far-UV CD and NMR (London & Khorana 1982). Neutralization of acid and dialysis against a solution of SDS restores some secondary structure in bacterioopsin. Subsequent addition of mixed phospholipid/detergent micelles and all-trans retinal leads to reappearance of a native-like chromophore (London & Khorana 1982). Bacteriorhodopsin regenerated in mixed phospholipid/detergent micelles exhibits the phenomenon of dark-light adaptation (Braiman et al 1987). Unlike oligomeric arrangement in the purple membrane, bacteriorhodopsin exists as a monomer in the mixed phospholipid/detergent micelles (Brouillette et al 1989). Near-UV CD indicates that the tertiary structure of bacteriorhodopsin in micelles is similar to that in the purple membrane (Brouillette et al 1989). Upon removal of detergent by dialysis, bacteriorhodopsin can be incorporated into phospholipid vesicles (liposomes) that are fully active in light-driven proton translocation (Huang et al 1981). The orientation of bacteriorhodopsin in phospholipid vesicles is, however, opposite to that found in the native purple membrane (Huang et al 1980). Although mixed phospholipid/detergent micelles are the most commonly used system, there is no absolute requirement of lipid for in vitro refolding of bacteriorhodopsin. Bacteriorhodopsin, for example, can be refolded from partially denatured bacterioopsin in SDS by the addition of all-trans- retinal in Triton X-100 or octylglucoside (London & Khorana 1982, Barnett et al 1996). The vesicles or liposomes consist of an enclosed lipid bilayer. That is, there is an inside and an outside. In contrast, lipid, detergent or mixed lipid/detergent bilayers that are not enclosed (open sheets) are referred to as micelles. Monolayers of lipid, detergent or mixed lipid/detergent forming an enclosed structure are also referred to as micelles. A model for the structure of mixed DMPC/CHAPS micelles based on light scattering measurements has been suggested (Mazer et al 1980). In this model, the micelles are believed to consist of a mixed bilayer disk which is surrounded on its rim with zwitterionic detergent CHAPS (Fig 1.18). Electron microscopy is confirmitive of this model giving a diameter of about 10 nm for the bilayer disk (Brouillette et al 1989). Bacteriorhodopsin can also be refolded to a native-like chromophore from two chymotryptic fragments corresponding to helices A-B and C-G in mixed phospholipid/detergent micelles (Liao et al 1983, Popot et al 1987). The product is

56 Chapter 1

(a) 3D view

(b) Longitudinal view DMPC

CHAPS

Fig 1.18. A model for the structure of mixed DMPC/CHAPS micelles. active in light-driven proton pumping and its moderately high resolution X-ray structure is indistinguishable from that of intact bacteriorhodopsin and the purple membrane (Popot et al 1986). Reports have also appeared indicating the regeneration of bacteriorhodopsin from fragments corresponding to helices A-E and F-G as well as from helices A-E and C-G (Liao et al 1984, Kataoka et al 1992). Similar attempts to refold bacteriorhodopsin from the chymotryptic five-helix fragment (helices C-G) and chemically synthseized peptides corresponding to helices A and B have also been successful (Kahn & Engelman 1992). Although bacteriorhodopsin cannot be refolded from fragments corresponding to individual helices, peptides corresponding to five (A, B, C, D and E) of the seven helices have been shown to be capable of independently forming transmembrane helices in phospholipid bilayers (Hunt et al 1993). Peptides corresponding to helices F and G, however, do not form transmembrane helices per se.

Introduction 57

The effect of amino acid substitutions within the protein sequence on the refolding of bacteriorhodopsin have also been investigated. Of particular interest are the observations that point or deletion mutations within the interhelical loops generally tend to be innocuous (Khorana 1988, Gilles-Gonzalez et al 1991). In contrast, single amino acid substitutions within the transmebrane helical regions are usually lethal. Thus, substitution of a number of transmembrane residues regenerates a chromophore that is not only functionally inactive but has its absorption maximum shifted up to 100 nm from that of the native chromophore (Khorana 1988, Mogi et al 1988, Marinetti et al 1989, Mogi et al 1989, Stern & Khorana 1989).

1.4.8 Models of Bacteriorhodopsin Folding Bacteriorhodopsin regeneration in mixed phospholipid/detergent micelles is well established. However, little is known about the kinetics of this process. One of the earliest studies on the folding mechanism of bacteriorhodopsin involved reconstitution of bacteriorhodopsin from apomembrane and retinal at lower temperatures (Schreckenbach et al 1977, Schreckenbach et al 1978a). Apomembrane is the purple membrane from which endogenous retinal has been removed by treatment with hydroxylamine under mild illumination (bleaching). The apomembrane is essentially a largely folded protein with a native-like secondary structure but loose tertiary interactions. In this inaugural study, a simple consecutive three-step model

R

apomembrane < s 400-nm < s 430/460-nm min bR chromophore chromophore

was proposed for the binding of retinal to the apoprotein. The 400-nm chromophore contains retinal non-covalently bound to the protein. The polyene chain and the cyclohexene ring of retinal are co-planar within this intermediate. The formation of the

58 Chapter 1

430/460-nm chromophore is not understood in molecular terms but this species also contains retinal non-covalently bound to the protein. The decay of the 430/460-nm species into bacteriorhodopsin is the overall rate-limiting step of the process and is characterized by the formation of a Schiff base between the aldehyde group of retinal and the e-amino group of Lys216. The kinetics of retinal binding to bacterioopsin within micelles have also been studied (London & Khorana 1982). These investigators proposed a simple two-step reaction scheme

R

> s min bO bR

in which the species bO¢ is a state of bacterioopsin with a native-like secondary structure. The binding of retinal to this state results in the formation of bacteriorhodopsin in which retinal and the protein are covalently bound via a Schiff base. The refolding of bacteriorhodopsin in DMPC/CHAPS micelles from a denatured state in SDS (bO), using stop-flow fluorescence spectroscopy, was first reported in the mid 1990s (Booth et al 1995). In this study, the kinetics of both the apoprotein refolding and retinal binding were investigated. A simple sequential scheme

R

< s > s < s min bO I1 IO IR bR

was invoked. In this scheme, the intermediates I1 and IO represent apoprotein refolding. The apoprotein refolding was suggested to be rate-limiting for the subsequent binding of

Introduction 59 retinal. Retinal is thought to be non-covalently bound to the protein within the species IR. The decay of this intermediate into bacteriorhodopsin is characterized by the formation of a Schiff base between retinal and the protein, which is the overall rate-limiting step of the process. More recently another scheme for bacteriorhodopsin folding, partly based on the results presented in this thesis, has appeared (Booth et al 1996). This scheme

R

< s > s < s min min bO I1 I2 IR I3 bR

is essentially similar to the previous one (Booth et al 1995), except that another putative intermediate I3 is thought to be involved in the decay of the IR intermediate into bacteriorhodopsin. The intermediate I1 and I2 in this scheme are analogous to the intermediate I1 and IO, respectively, described in the previous study (Booth et al 1995). One common theme to all of the models of bacteriorhodopsin folding presented above is that they are purely based on sequential steps. None of these models however exclude the possibility for parallel branched pathways. Other reports on bacteriorhodopsin refolding kinetics have also appeared recently. A far-UV CD study suggested that a small fraction (15 %) of helix formation during bacteriorhodopsin refolding was observed to occur over a time scale of tens of seconds (Riley et al 1997). This is extremely interesting in view of the notion that helices are formed over a time scale of ns to ms in water-soluble proteins (Creighton 1992, Williams et al 1996). Refolding of bacteriorhodopsin, however, occurs in a lipid bilayer milieu. It was therefore suggested that although helix formatiom may be intrinsically a fast process, the constraints imposed on the protein by the lipid bilayer may apparently slow this process. In fact, a model was postulated to account for these observations. It was proposed that majority (85 %) of helix formation would occur upon insertion of the protein into the lipid bilayer. Such helix formation would predominantly take place within the core regions of the transbilayer helices. The ends of helices would not be

60 Chapter 1 initially in a helical conformation because the interhelical loops would be stretched. However, subsequent association of helices over a time scale of tens of seconds would bring the helices together and this would allow the helix assembly at the ends of helices at the bilayer interfacial region. The refolding kinetics of bacteriorhodopsin are also observed to depend on the rigidity of the lipid bilayer as well as on the pH of the aqueous environment (Booth et al 1997). Thus, the refolding studies on bacteriorhodopsin are only in their early stage. Further work clearly needs to be undertaken in order to gain a better understanding of the folding of this and other membrane proteins.

Introduction 61

1.5 Summary

A large body of data exists on the folding of water-soluble proteins. However, little is known about the folding of membrane proteins. Membrane protein folding studies should provide an insight into the mechanisms of protein synthesis and assembly into membranes, protein translocation across membranes and should have relevance to the development of better ways of drug delivery to diseased tissues. The small integral membrane protein bacteriorhodopsin from Halobacterium salinarium, where it acts as a light-driven proton pump, is used here as a prototype for studying membrane protein folding. A number of features make bacteriorhodopsin a candidate par excellance for such purposes. Bacteriorhodopsin can be easily purified and stored in deep freeze for years without any observable change in its activity. The near-atomic resolution structure for bacteriorhodopsin is available. Bacteriorhodopsin can be easily refolded from a denatured state in mixed phospholipid/detergent micelles and refolding can be assayed by the appearance of a native-like chromophore absorption band. Folding in micelles should be similar to the situation in vivo. The overall structural architecture of micelles appears to resemble that of a biological membrane and both contain amphipathic molecules surrounded by an aqueous environment. There is therefore little reason to doubt the applicability of the principles of folding established in vitro to folding in vivo. This thesis is divided into eight chapters. In this inaugural chapter, the status quo of protein folding research has been outlined and, in particular, the reasons that make membrane proteins such a strenuous exercise to follow have been discussed. Despite these difficulties, a lot can be learned about membrane protein folding by the application of simple classical stop-flow mixing methods in conjunction with spectroscopy. Chapter 2 gives a detailed account of the major practical strategies and techniques used in this study. Methods used to purify bacterioopsin from the purple membrane are described. The technical aspects of stop-flow fluorescence and photodiode array spectroscopy are outlined. The theoretical background to data analysis is provided. In chapters 3 through to 7, the results of this study are presented and their implications on the folding mechanism of bacteriorhodopsin, in light of what is known about the folding of other proteins, are discussed.

62 Chapter 1

Chapter 3 is concerned with the refolding of bacteriorhodopsin in DMPC/CHAPS micelles under equilibrium conditions. How the refolding yield of bacteriorhodopsin depends upon the pH and the mode of retinal addition, that is whether bacterioopsin refolding is initiated in the presence or absence of retinal, is demonstrated. The remaining chapters 4 through to 7 are concerned with the time-resolved measurements of bacteriorhodopsin refolding in DMPC/CHAPS micelles using stop-flow fluorescence and absorption spectroscopy, with millisecond time-resolution. Chapter 4 presents the refolding kinetics of the apoprotein bacterioopsin in the absence of retinal. Refolding is initiated by mixing denatured bacterioopsin in SDS with DMPC/CHAPS micelles in a stop-flow cuvette. Although intrinsic protein fluorescence is a highly sensitive probe for providing qualitative information on the environment of aromatic residues during refolding, it cannot differentiate whether the change in the environment is due to structural changes within the protein or the micelles. Taking advantage of the light scattering properties of the micelles, the structural changes within the micelles per se are followed upon mixing SDS with DMPC/CHAPS. This should allow a distinction to be made between kinetics of micelle mixing and bacterioopsin refolding. Chapter 5 demonstrates the kinetics of retinal binding to the apoprotein bacterioopsin. Binding of retinal to bacterioopsin is studied in two ways: retinal is added such that the apoprotein refolding is initiated in its presence, or alternatively, retinal is added after the apoprotein has been allowed to pre- equilibrate with the DMPC/CHAPS micelles. Binding of retinal to bacterioopsin is also monitored by measuring changes in the retinal absorption band using a photodiode array. The refolding of the apoprotein bacterioopsin appears to be rate-limiting for the subsequent binding of retinal. A spectroscopically distinct intermediate is observed in the folding of bacteriorhodopsin. The oscillator strength of the retinal absorption band is exploited to probe the chemistry of retinal during the course of the reaction of retinal with bacterioopsin. Chapter 6 examines the effect of adding retinal at various times in the refolding of bacterioopsin. It is shown that there are at least two rate-limiting steps in the refolding of the apoprotein bacterioopsin for the subsequent binding of retinal. This finding raises the intriguing possibility of alternative parallel retinal binding steps.

Introduction 63

Chapter 7 explores the effect of varying retinal concentration on the refolding kinetics of bacteriorhodopsin. Two bimolecular reactions involving retinal and bacterioopsin are observed. This demonstrates strongly that retinal binds in a parallel fashion to the apoprotein. How thermodynamic information can be obtained on the folding intermediates of bacteriorhodopsin is demonstrated. Finally in Chapter 8, the results and discussions from the previous chapters are brought together to provide a unified view on the possible folding mechanism of bacteriorhodopsin. The implications of this study on the folding mechanisms of other proteins are highlighted. How the findings of this study should open up new ways for gaining further information on bacteriorhodopsin folding and other membrane proteins are discussed. Some of the work presented in this thesis has been previously reported (Booth et al 1996, Booth et al 1997, Booth & Farooq 1997).

Chapter 2

MATERIALS AND METHODS

2.1 Materials 66 2.1.1 Chemical Reagents 66 2.1.2 Phosphate Buffer Preparation 66 2.1.3 Micelle Preparation 66 2.1.4 Retinal Preparation 66

2.2 Isolation and Purification of Bacterioopsin 68 2.2.1 Cell Growth and Harvesting 68 2.2.2 Isolation of Purple Membrane 69 2.2.3 Sucrose Density Gradient Centrifugation 69 2.2.4 Delipidation of Purple Membrane 70 2.2.5 Phosphorus Assay 72 2.2.6 SDS-PAGE 72

2.3 Equilibrium Measurements 74 2.3.1 Addition of Retinal/DMPC/CHAPS to bO/SDS 74 2.3.2 Addition of Retinal/DMPC/CHAPS to bO/DMPC/CHAPS 74 2.3.3 Steady-State Spectroscopy 75

2.4 Kinetic Measurements 76 2.4.1 Addition of Retinal/DMPC/CHAPS to bO/SDS 76 2.4.2 Addition of Retinal/DMPC/CHAPS to bO/DMPC/CHAPS 76 2.4.3 Addition of Retinal at Various Times in the Refolding of bO 77 2.4.4 Stop-Flow Spectroscopy 77 2.4.5 Photodiode Array Spectroscopy 79

2.5 Data Analysis 81 2.5.1 Errors 81 2.5.2 Marquardt Algorithm 81 2.5.3 Exponential Kinetics 82 2.5.4 Global Analysis 83 2.5.5 Pseudo-First-Order Kinetics 86 2.5.6 Singular Value Decomposition 88 2.5.7 Scattering Analysis 88 2.5.8 Gaussian Decomposition 89

66 Chapter 2

2.1 Materials

2.1.1 Chemical Reagents DMPC was purchased from Avanti (Alabama, USA) and CHAPS from Calbiochem (Nottingham, UK). All-trans retinal (vitamin A aldehyde), SDS (electrophoresis grade), DNAse I (bovine pancreas) and phosphorus standard solution were supplied by Sigma (Dorset, UK). Oxoid peptone was obtained from Unipath (Basingstoke, UK). All other reagents and chemicals were of analytical grade.

2.1.2 Phosphate Buffer Preparation Phosphate buffer was prepared by mixing 0.5 M sodium dihydrogen phosphate with 0.5 M disodium hydrogen phosphate at an appropriate ratio to give the desired pH. The buffer was stored at room temperature in 0.025 % (w/v) sodium azide for up to six months.

2.1.3 Micelle Preparation Mixed DMPC/CHAPS micelles were prepared by stirring 2 % (w/v) DMPC in 50 mM sodium phosphate buffer for 2 h at room temperature. CHAPS was then added to 2 % (w/v) final concentration and the mixture sonicated in a bath sonicator for 30 min. Micelles were stored in 0.025 % (w/v) sodium azide at 4 °C for up to a month. Stored micelles were usually resonicated prior to use.

2.1.4 Retinal Preparation Retinal was dissolved in ethanol (pre-equilibrated with argon). The concentration was determined from absorbance at 380 nm using an extinction coefficient of 42 800 cm-1M-1 (Rehorek & Heyn 1979). Retinal was stored at -70 °C under argon for up to a month. The absorption spectrum of all-trans-retinal in ethanol is shown in Fig 2.1.

Materials and Methods 67

0.4

0.3

0Absorbance .2

0.1

0.0

250 300 350 400 450 500 550 600 650 700 750 Wavelength / nm

Fig 2.1. Absorption spectrum of all-trans retinal in ethanol at room temperature.

68 Chapter 2

2.2 Isolation and Purification of Bacterioopsin

2.2.1 Cell Growth and Harvesting The method used to isolate purple membrane from Halobacterium salinarium was essentially that employed previously (Oesterhelt & Stoeckenius 1974). H salinarium (strain S9) was maintained on plates of solid media [basal salt containing 1 % (w/v) peptone and 2 % (w/v) agar] and replated every three months to minimize risk of mutation. Basal salt was 4 M sodium chloride, 100 mM magnesium sulphate, 25 mM potassium chloride and 10 mM sodium citrate. After incubation at 39 °C in the dark for 2 weeks, the plates were stored at 4 °C. A loopful of bright purple colonies from these plates was used to inoculate a 50-ml peptone media [basal salt containing 1 % (w/v) peptone] in a 250-ml conical flask. The flask was then loosely capped with cotton wool, to minimize water evaporation but allow limited air supply, and shaken gently at 100 rpm in the dark at 39 °C on a rotary shaker. Under these conditions, a limited supply of oxygen ensures optimal synthesis of purple membrane. Strain S9 has the advantage that illumination of cultures is not necessary due to the quasi-constitutive expression of bacteriorhodopsin. The growth of cells was monitored spectrophotometrically by measuring cell density at 660 nm.

After 1 week incubation (A660 ~ 0.5), the late-log phase culture was used to inoculate a 1-litre peptone media in a 3-litre conical flask. After a further 1 week incubation as above (A660 ~ 1), the culture was harvested just prior to reaching the stationary phase. This was achieved by centrifuging the culture at 10 000 rpm for 20 min in a Beckman centrifuge (Model J2-21) at 4 °C using JA10 rotor. After decanting away the supernatant, the dark-purple pellet was resuspended in as little volume as possible of basal salt - typically 50 ml for every litre of bacterial culture. DNAse I was added at a final concentration of 10 mgml-1 to the cell paste which was subsequently dialyzed overnight in a cellulose acetate dialysis tubing against 0.1 M sodium chloride. This leads to cell lysis and the cell membrane is fragmented into sheets of approximately 1-2 mm in diameter. Treatment with DNAse I reduces the viscosity of the solution. Dialysis also results in the extrusion of cell debris and organelles into the external medium. The membrane sheets are retained within the dialysis bag.

Materials and Methods 69

2.2.2 Isolation of Purple Membrane The dialysate was centrifuged at 30 000 rpm for 30 min in a Beckman ultracentrifuge (Model L8-70M) at 4 °C using 45 Ti rotor. The supernatant was decanted away and the pellet resuspended in water. After homogenizing in a glass mortar using a stainless steel pestle with PTFE head (Merck, Lutterworth, UK), the solution was centrifuged as above. This step was repeated three times or until the supernatant was nearly colourless. This procedure removes most of the contaminants including the carotenoids and the red membrane. The resulting pellet, containing purple membrane suspension, was redissolved in as little volume as possible of water containing 0.025 % (w/v) sodium azide to prevent microbial growth and stored at 4 °C. Under aerobic conditions, the cell membrane of halobacteria is referred to as the red membrane because the carotenoids are evenly distributed throughout the membrane. The carotenoids give the membrane its characteristic red colour. However, halobacteria are grown here under low oxygen tension and these conditions favour the synthesis of bacteriorhodopsin within certain regions of the cell membrane. These regions are referred to as the purple membrane because they soley contain bacteriorhodopsin, which gives these regions their purple colour. The cell membrane of halobacteria grown here thus consists of a red membrane interspersed with patches of up to 0.5 mm in diameter of purple membrane. These purple patches can account for up to half the total area of the cell membrane under optimal growth conditions.

2.2.3 Sucrose Density Gradient Centrifugation This procedure was carried out to remove any traces of red membrane from purple membrane. A combination of step and continuous sucrose gradient was used with slight modifications to that described previously (Oesterhelt & Stoeckenius 1974). A centrifuge tube (25 ml) was layered in the bottom with 5 ml of a 60 % (w/v) sucrose solution and then a continuous gradient was built on top of this with 45 and 25 % (w/w) sucrose solutions using a gradient former. The density of the gradient decreased from the bottom of the tube to the top. The purple membrane suspension (2-5 ml) was layered on top of this gradient which was then centrifuged at 25 000 rpm overnight in a Beckman ultracentrifuge (Model L8-70M) at 4 °C using SW28 rotor. The purple

70 Chapter 2 membrane band was recovered, washed with water and then centrifuged at 30 000 rpm for 30 min in a Beckman ultracentrifuge using 45 Ti rotor. This step was repeated twice more with water. The purple membrane was finally resuspended in as little volume of water as possible. Sodium azide was added to 0.025 % (w/v) and the suspension stored at 4 °C. The red membrane band was only rarely seen at this stage of the preparation and was usually found just above the purple band. Protein concentration of the purple membrane suspension was measured from absorbance at 560 nm using an extinction coefficient of 54 000 cm-1M-1 (Oesterhelt & Stoeckenius 1974). Typical yields of purple membrane were 20 mg protein per litre culture of bacteria.

2.2.4 Delipidation of Purple Membrane Purple membrane was delipidated according to the method used previously (Braiman et al 1987), with slight modifications. Purple membrane suspension (20 mgml-1 protein) was mixed with Solvent A [chloroform, methanol and triethylamine at a volume ratio of 100:100:1] in a glass centrifuge tube at a 1:10 volume ratio to denature the protein. This was accompanied by a loss of purple colour. Hydroxylamine was then added in excess at a final concentration of 1 mM to the denatured protein suspension to cleave the retinylidene-protein Schiff linkage. After vortexing, phase separation was brought about by the addition of an equal volume of 0.1 M sodium phosphate buffer (pH 6). The mixture was centrifuged at 2 000 rpm (8 000 g) for 20 min in a bench top MSE centrifuge to separate the aqueous phase from the organic phase. The latter contains membrane lipids and retinal oxime. The protein was recovered as a pellet formed at the aqueous-organic interface. The pellet was washed in excess water and recovered by centrifugation as above. This step was repeated twice more. The protein was redissolved in solvent A and phase partitioning was repeated twice more to remove as much endogenous membrane lipids as possible. The protein was finally washed several times with water to completely remove organic solvent and then redissolved in 1 % (w/v) SDS. After centrifuging down undissolved matter, the clear supernatant was removed and protein concentration was measured from absorbance at 280 nm using an extinction coefficient of 66 000 cm-1M-1 (London & Khorana 1982). The protein was aliquoted and stored at -20 °C. Typically 50 % of

Materials and Methods 71 protein was recovered upon delipidation. The absorption spectra of purple membrane and delipidated bacterioopsin are compared in Fig 2.2.

0.5

0.4

0.3 Absorbance 0.2

PM 0.1

bO 0.0 250 300 350 400 450 500 550 600 650 700 750 Wavelength / nm

Fig 2.2. Absorption spectra of purple membrane (PM) isolated from Halobacterium salinarium and delipidated bacterioopsin (bO). Purple membrane is an aqueous suspension and delipidated bacterioopsin is in 1 % (w/v) SDS.

72 Chapter 2

2.2.5 Phosphorus Assay Phosphorus assays were carried out to determine the extent to which endogenous phospholipids were removed upon delipidation of the purple membrane. The method is based on precise determination of inorganic phosphate released upon the acid hydrolysis of phospholipids. The procedure was carried out, with slight modifications, according to the method described previously (Ames 1966, Kates 1972). Purple membrane [10 ml of a 0.5 mM protein in water] and delipidated bacterioopsin [0.5 ml of a 50 mM protein in 1 % (w/v) SDS] were added to separate boiling tubes. After drying at 100 °C, 0.26 ml of a 60 % (v/v) perchloric acid was added to each tube. The tubes were vortexed, capped with a tinfoil film and then heated at 180 °C for 30 min to hydrolyse phospholipids. After cooling to room temperature, 0.92 ml of water, 0.4 ml of a 5 % (w/v) ascorbic acid and 0.4 ml of a 1.25 % (w/v) ammonium molybdate were added to each tube. Tubes were vortexed, capped with a tinfoil film to minimize solvent evaporation and then heated at 100 °C for 5 min to assist the formation of a coloured complex between ammonium molybdate and inorganic phosphate. The absorbance of the coloured complex was measured at 800 nm. To measure the content of inorganic phosphate released, a calibration curve was obtained using known quantities of phosphate. A phosphorus standard solution (Sigma) was added (0-150 ml) to several boiling tubes and treated as above. The amount of inorganic phosphate was plotted against absorbance at 800 nm. From this plot, the phosphate content in the purple membrane and delipidated bacterioopsin was determined. Typically over 95 % endogenous phospholipids were removed upon delipidation of purple membrane, as reported previously (Braiman et al 1987). Although phospholipids are not the only lipid constituent of the cell membrane of H salinarium, such an analysis shows that the delipidated preparations of bacterioopsin are likely to be largely free of endogenous lipid.

2.2.6 SDS-PAGE SDS-PAGE analysis of purple membrane and bacterioopsin was carried out using a mini gel (Hoefer, San Francisco, USA) based on the standard procedure (Laemmli 1970). A 12 % separating gel [Solvent B containing 12 % (w/v) acrylamide/bis-

Materials and Methods 73 acrylamide and 0.375 M Tris-HCl (pH 8.8)] and a 4 % stacking gel [Solvent B containing 4 % (w/v) acrylamide/bis-acrylamide and 0.125 M Tris-HCl (pH 6.8)] were used. Solvent B was 0.1 % (w/v) SDS, 0.1 % (w/v) TEMED and 0.05 % (w/v) ammonium persulphate. The acrylamide to bis-acrylamide ratio was 37:1 (w/w). Delipidated bacterioopsin, purple membrane and protein standard [SDS-7 (Sigma)] were mixed with an equal volume of Loading Buffer [60 mM Tris-HCl (pH 6.8), 10 % (v/v) glycerol, 2 % (w/v) SDS, 5 % (v/v) 2-b-mercaptoethanol and 0.01 % (w/v) bromophenol blue]. After heating at 95 °C for 5 min to denature proteins, 5-10 mg of each sample was loaded onto the gel pre-equilibrated with Running Buffer [1.5 % (w/v) glycine, 0.3 % (w/v) Tris and 0.1 % (w/v) SDS, pH 8.3]. The gel was then run at 100 V for 1 h and then stained in Solvent C [water, methanol and acetic acid at a volume ratio of 5:4:1] containing 0.1 % (w/v) Coomassie blue (Sigma) for 1 h. After destaining overnight in Solvent C, a photograph of the gel was taken. SDS-PAGE analysis of purple membrane and delipidated bacterioopsin are shown in Fig 2.3. In each case, a single protein band with an apparent molecular mass of ~ 22 kD is observed. This value is somewhat lower than the actual value of 26 kD. This is most likely due to the hydrophobic nature of bacterioopsin which accounts for its higher mobility relative to water-soluble protein markers, as observed previously (Braiman et al 1987).

_ PM bO

66 45 36 29 24 20 direction of electrophoresis 14 +

Fig 2.3. SDS-PAGE analysis of purple membrane (PM) and delipidated bacterioopsin (bO). The outer lanes contain water-soluble protein markers with their molecular masses indicated in kD. All samples were boiled in SDS prior to loading (see text).

74 Chapter 2

2.3 Equilibrium Measurements

2.3.1 Addition of Retinal/DMPC/CHAPS to bO/SDS Addition of retinal in DMPC/CHAPS to bO partially denatured in SDS (bO/SDS) was carried out as described previously (Huang et al 1981, London & Khorana 1982). Briefly, 8 mM bO in 0.2 % (w/v) SDS and 50 mM sodium phosphate buffer (pH 6) were mixed with an equal volume of 2 % (w/v) DMPC / 2 % (w/v) CHAPS micelles in 50 mM sodium phosphate buffer (pH 6) containing all-trans retinal. The final volume of the mixture was 1 ml. Retinal was added in 1-5 ml ethanol using an Hamilton syringe [final ethanol concentration < 0.5 % (v/v)]. Final bO concentration was constant at 4 mM and that of retinal was varied between 0 and 8 mM. The final concentrations of other reagents were 0.1 % (w/v) SDS, 1 % (w/v) DMPC, 1 % (w/v) CHAPS and 50 mM sodium phosphate buffer. Experiments were also carried out at pH 8. All samples were incubated overnight in the dark at 22.0 ± 0.1 oC. Absorption and fluorescence spectra were subsequently recorded in the dark or dim red light at 22.0 ± 0.1 oC. The final pH of the reaction mixture at this temperature was within ± 0.1 of the value indicated.

2.3.2 Addition of Retinal/DMPC/CHAPS to bO/DMPC/CHAPS Addition of retinal in DMPC/CHAPS to bO pre-equilibrated with DMPC/CHAPS micelles (bO/DMPC/CHAPS) was performed essentially as described previously (London & Khorana 1982). All- trans retinal in 1 % (w/v) DMPC / 1 % (w/v) CHAPS / 0.1 % (w/v) SDS micelles and 50 mM sodium phosphate buffer (pH 6) was mixed with an equal volume of bO/DMPC/CHAPS. The final volume of the mixture was 1 ml. bO/DMPC/CHAPS was prepared by mixing 16 mM bO in 0.2 % (w/v) SDS and 50 mM sodium phosphate buffer (pH 6) with an equal volume of 2 % (w/v) DMPC / 2 % (w/v) CHAPS micelles in 50 mM sodium phosphate buffer (pH 6) and allowed to equilibrate for 30 min. Retinal was added in 1-5 ml ethanol using an Hamilton syringe [final ethanol concentration < 0.5 % (v/v)]. Final bO concentration was constant at 4 mM and that of retinal was varied between 0 and 8 mM. The final concentrations of other reagents were 0.1 % (w/v) SDS, 1 % (w/v) DMPC, 1 % (w/v) CHAPS and 50 mM sodium phosphate buffer. Experiments were also carried out at pH 8. All samples were incubated overnight

Materials and Methods 75 in the dark at 22.0 ± 0.1 oC. Absorption and fluorescence spectra were subsequently recorded in the dark or dim red light at 22.0 ± 0.1 oC. The final pH of the reaction mixture at this temperature was within ± 0.1 of the value indicated.

2.3.3 Steady-State Spectroscopy Absorption spectra were recorded between 250-750 nm on a UVIKON 930, aminco DW2000 or CARY 1G spectrophotometer using a 2-nm bandwidth and a quartz cuvette with a 1 cm pathlength. The absorbance maximum of regenerated bacteriorhodopsin was plotted against retinal concentration. From the initial slope of these plots, molar extinction coefficients of bacteriorhodopsin at the absorbance maximum were calculated (Rehorek & Heyn 1979, London & Khorana 1982). The yield of refolding was determined from the ratio of bacteriorhodopsin concentration to the initial bacterioopsin concentration. Bacterioopsin concentration was determined in 0.2 % (w/v) SDS from absorbance at 280 nm, prior to addition of DMPC/CHAPS micelles, using an extinction coefficient of 66 000 cm-1M-1 (London & Khorana 1982). The concentration of regenerated bacteriorhodopsin was determined from absorbance at 555 nm using an extinction coefficient of 55 300 cm-1M-1 (pH 6) or at 548 nm using an extinction coefficient of 53 500 cm-1M-1 (pH 8). Both of these extinction coefficients were calculated here (see Chapter 3). Fluorescence spectra were recorded between 300-500 nm on a Perkin Elmer LS50 spectrometer, with excitation and emission bandwidths of 2.5 nm and using a quartz fluorescence cuvette. The excitation wavelength was 290 nm. Relative fluorescence yield was plotted against retinal concentration. Relative fluorescence yield was calculated by integrating under the fluorescence band from 300-500 nm.

76 Chapter 24

2.4 Kinetic Measurements

2.4.1 Addition of Retinal/DMPC/CHAPS to bO/SDS Addition of retinal in DMPC/CHAPS to bO partially denatured in SDS (bO/SDS) was carried out, as described previously (Booth et al 1995). 4 mM bO in 0.2 % (w/v) SDS and 50 mM sodium phosphate buffer (pH 6) were mixed, in the stop-flow cuvette, with an equal volume of 2 % (w/v) DMPC / 2 % (w/v) CHAPS micelles in 50 mM sodium phosphate buffer (pH 6) containing retinal [final ethanol concentration < 0.5 % (v/v)]. The final volume of the reaction mixture was 100 ml. Final bO concentration was 2 mM and that of retinal was varied between 0 and 2 mM. The final concentrations of all other reagents were 0.1 % (w/v) SDS, 1 % (w/v) DMPC, 1 % (w/v) CHAPS and 50 mM sodium phosphate buffer. All measurements were made at 22.0 ± 0.1 °C in the dark or dim red light. The final pH of the reaction mixture at this temperature was 6.0 ± 0.1. All experiments were carried out on bO preparations with refolding yields of 95 % or more at pH 6.

2.4.2 Addition of Retinal/DMPC/CHAPS to bO/DMPC/CHAPS Addition of retinal in DMPC/CHAPS to bO pre-equilibrated with DMPC/CHAPS micelles (bO/DMPC/CHAPS) was carried out, as described previously (Booth et al 1995). All-trans-retinal in 1 % (w/v) DMPC / 1 % (w/v) CHAPS / 0.1 % (w/v) SDS micelles and 50 mM sodium phosphate buffer (pH 6) were mixed, in the stop-flow cuvette, with an equal volume of bO/DMPC/CHAPS [final ethanol concentration < 0.5 % (v/v)]. The final volume of the reaction mixture was 100 ml. bO/DMPC/CHAPS was prepared by mixing manually 8 mM bO in 0.2 % (w/v) SDS and 50 mM sodium phosphate buffer (pH 6) with an equal volume of 2 % (w/v) DMPC / 2 % (w/v) CHAPS micelles in 50 mM sodium phosphate buffer (pH 6) and allowed to equilibrate for 30 min. Final bO concentration was 2 mM and that of retinal was varied between 0 and 32 mM. The final concentrations of all other reagents were 0.1 % (w/v) SDS, 1 % (w/v) DMPC, 1 % (w/v) CHAPS and 50 mM sodium phosphate buffer. All measurements were made at 22.0 ± 0.1 °C in the dark or dim red light. The final pH of the reaction mixture at this temperature was 6.0 ± 0.1. All experiments were carried out on bO preparations with refolding yields of 95 % or more at pH 6.

Materials and Methods 77

2.4.3 Addition of Retinal at Various Times in the Refolding of bO Addition of retinal at various times in the refolding of bacterioopsin was carried out as follows. 8 mM bO in 0.2 % (w/v) SDS and 50 mM sodium phosphate buffer (pH 6) were mixed, in the stop-flow, with an equal volume of 2 % (w/v) DMPC / 2 % (w/v) CHAPS micelles in 50 mM sodium phosphate buffer (pH 6). After variable refolding times this mixture was rapidly mixed, in the stop-flow cuvette, with an equal volume of 1 % (w/v) DMPC / 1 % (w/v) CHAPS / 0.1 % (w/v) SDS micelles in 50 mM sodium phosphate buffer (pH 6) containing all-trans retinal [final ethanol concentration < 0.5 % (v/v)]. The final volume of the reaction mixture was 200 ml. The refolding time was varied from 0.1 to 1000 s. The final concentrations of bO and retinal were 2 mM. The final concentrations of all other reagents were 0.1 % (w/v) SDS, 1 % (w/v) DMPC, 1 % (w/v) CHAPS and 50 mM sodium phosphate buffer. All measurements were made at 22.0 ± 0.1 °C in the dark or dim red light. The final pH of the reaction mixture at this temperature was 6.0 ± 0.1. All experiments were carried out on bO preparations with refolding yields of 95 % or more at pH 6.

2.4.4 Stop-Flow Spectroscopy Time-resolved studies were performed using an Applied Photophysics SX.17MV stop-flow spectrometer (Leatherhead, UK), with a deadtime of ~ 1.4 ms. Deadtime is the time between mixing solutions and detection. The detection usually begins a few milliseconds after the initial mixing of solutions. Fig 2.4 shows a schematic diagram of the stop-flow mixing circuit. The single-mixing (addition of retinal to bO/SDS and bO/DMPC/CHAPS) and sequential-mixing (addition of retinal at various times in the refolding of bO) procedures are depicted. For fluorescence studies, excitation wavelength was at 290 nm (1 nm bandwidth) and emission was collected, with an emission photomultiplier tube, above 305 nm using a cut-off filter. For absorbance experiments at single wavelengths, data were recorded by a photomultiplier tube using an excitation bandwidth of 1 nm and a pathlength of 1 cm. A 150-W Xenon arc lamp was used as a light source. The alignment of the lamp was checked at the beginning of absorbance measurements to ensure stability. A thermostatic waterbath was used to maintain the temperature at 22.0 ± 0.1

78 Chapter 2

stop syringe

cuvette delay loop

F C B A

simple or flush ram pre-mix ram

Fig 2.4. Stop-flow mixing circuit. The single-mixing method involves mixing solutions in the cuvette using syringes C and F, and the simple ram. The circuit connecting syringes A and B is by-passed during this operation. In the sequential-mixing method, the solutions are pre-mixed in the delay loop using syringes A and B, and the pre-mix ram. After a certain delay time, the flush ram drives solutions from syringes C and F. The displacement of pre-mixed solution in the delay loop by solution F allows its mixing in the cuvette with the solution C. The rams are used to drive the syringes with a gas pressure of 9 bar.

Materials and Methods 79

°C. Data were acquired over different time scales, ranging from 50 ms to 1000 s. For each time scale, 4000 data points were collected. Data acquisition was also carried out over a split time scale, with data points halved for each interval. To increase the signal-to-noise ratio, the signal was electronically filtered before recording. Typically, the filtering time was not exceeded more than 1/10000th the time over which data were recorded. Two, four or eight transients were also averaged to improve the signal-to- noise ratio.

2.4.5 Photodiode Array Spectroscopy Time-resolved absorption spectra were recorded using the stop-flow spectrometer described above, equipped with a photodiode array detector. This detector comprises a linear 256 element diode array covering the range 305-1100 nm, with a diode wavelength separation of about 3.3 nm and a deadtime of ~ 2-3 ms. Water was used as a reference for all experiments. An excitation bandwidth of 1 nm and a pathlength of 1 cm were used for all experiments. The alignment of Xenon arc lamp, used as a light source, was checked at the beginning of each experiment. Time-resolved absorption spectra were collected over different time scales, ranging from 0.3 to 1000 s, over the wavelength range 305-750 nm or 350-750 nm. For each time scale 100-400 spectra were recorded. To improve the signal-to-noise ratio below 400 nm, where the intensity of the Xenon lamp is lower (Fig 2.5), data over the wavelength range 305-750 nm were collected in two parts. Light intensity is a function of integration time (time between each scan or re-reading the same diode) and thus can be varied by varying the integration time. Data over the wavelength range 305-400 nm were taken using an integration time of about 100 ms, whilst data over the wavelength range 400-750 were taken using an integration time of about 10 ms. An integration time of more than about 10 ms saturates the diode in the region 500-600 nm. This is clearly not desirable. The idea is to use the maximum light intensity without saturating the diode. However, an integration time of about 10 ms results in a very small light output in the region below 400 nm. The only alternative therefore is to acquire data in two parts as described above. The two sets of data in the range 305-400 nm and 400-750 nm were

80 Chapter 2

4

3

2

Light intensity 1

0 300 400 500 600 700 800 900 1000 1100 Wavelength / nm

Fig 2.5. The Xenon lamp intensity profile.

normalized to any changes in light intensity incurred during the experiment. The signal- to-noise ratio for data collected over shorter time scales of 0.3 to 5 s cannot be improved in the region 305-400 nm. This is because the integration time is also a function of the time scale over which the data is collected. Increasing the integration time, to improve the light intensity, whilst keeping the time scale short are mutually exclusive. Thus for experiments where data collection over shorter time scales was essential, data were recorded in one part over the wavelength range 350-750 nm using an integration time of 10 ms.

Materials and Methods 81

2.5 Data Analysis

2.5.1 Errors Errors for all experimentally determined parameters were calculated from measurements on three different bO preparations, unless otherwise stated. Several bO preparations differ in that they are delipidated separately but the source of purple membrane may be the same. The value for each parameter quoted is the mean value. The extent to which the true value of the parameter might deviate from this mean value yo is the standard deviation s given by the relationship

1/2 s = {[å (yi - yo)] / (N - 1)} where yi is the value of a parameter at a measurement i and N is the total number of measurements. All errors are given to one standard deviation. That is, there is 68 % chance that if the experiment were to be repeated the value of that parameter will lie within this range (Robyt & White 1987).

2.5.2 Marquardt Algorithm Marquardt algorithm was used to fit theoretical functions to the experimental data (Marquardt 1963, Bevington & Robinson 1992). This is a method of least-squares and is based on minimizing the sum of 2 the squares of the deviation of the data yi from the fitting function fi. That is, to minimize (yi - fi) . When this sum is minimum, a line of best fit of the theoretical function to the experimental data is obtained. The quality of fits was assessed by plots of residuals, that is, the difference between the experimental and the theoretical curve (yi - fi). The choice of a particular theoretical function for a particular experimental curve was based on the inspection of the quality of the residuals. When there was only a marginal improvement in the quality of the residuals between two theoretical functions over the experimental curve, the function with the least complexity was considered sufficient to describe the experimental data. Other factors, such as analysis of data separately over different time scales, were also taken into account.

82 Chapter 2

2.5.3 Exponential Kinetics For a quantitative analysis, approximation of time-resolved experimental data to a sum of exponentials is de rigueur. Fit of raw data to a specific model is fraught with danger. One can never know with certainity that this model is the correct description of the data. Thus a more accurate method of data analysis is to fit the experimental curve to a function that does not conform to any particular model. That is, it is model-independent. A sum of exponentials is such a function. An exponential describes an irreversible first-order reaction (vide infra). Thus a sum of exponentials denotes, in theory, independent first-order reactions. These reactions may be sequential or parallel. Thus the models

A B C

B A A C C B can all be described by a sum of two exponentials. It is this very nature of exponentials that renders them applicable to the mechanistic studies of a wide variety of systems. The only drawback of exponentials is that it is assumed a priori that all the elementary steps in a system can be described by first-order kinetics. This is not always the case however and, in particular, in the case of ligand-binding proteins. The binding of a ligand to a protein effectively obeys second-order kinetics. In brief, it is a common practice to approximate protein folding reactions to a sum of exponentials. Time-resolved data S[t] were fit to sums of exponentials using the function

S[t] = å {ai exp(- ni t)} + d [1]

where ai is the observed amplitude and ni the observed rate constant of a kinetic phase i, and d is the baseline offset. Experimentally observed rate constants and their amplitudes were calculated by iterative reconvolution based on Marquardt algorithm. Data were fit simultaneously over different time scales using kinetic softwares Look (Imperial

Materials and Methods 83

College, UK), GraFit (Imperial College, UK) and/or Origin (Microcal, USA). The amplitude of an exponential phase relates to the difference in the optical property, such as absorbance or fluorescence, of the two interconverting states. For a simple one-step reaction such as

k+1 A B k -1 the rate of decay of A into B with first-order kinetics is given by the equation

dA[t] / dt = - n A[t] [2] where A[t] is the concentration of A at time t and n is the experimentally observed rate constant related to the intrinsic rate constants k+1 and k-1 by the relationship

n = k+1 + k-1 [3]

Integrating eq [2] with respect to t shows that the concentration of A decreases exponentially with time

A[t] = Ao exp(- n t) [4] where Ao is the initial concentration of A. The lifetime (t) of a reaction is defined as the time taken for A to decay to 1/e of its initial concentration. From eq [2] it follows that

t = 1/n [5]

That is, the lifetime is the reciprocal of the experimentally observed rate constant.

2.5.4 Global Analysis In a global analysis, time-resolved data as a function of another physical property such as wavelength are analyzed simultaneously. The rate constants are held constant for all the data curves but their amplitudes are allowed to vary. This procedure thus provides a global measure of the rate constant of a particular kinetic phase. Its significance is especially realized in view of the fact that exponentials are not strictly independent variables. Thus the rate constant of one exponential is to some extent dependent on the

84 Chapter 2 magnitude of the rate constant of another exponential. This interdependence of rate constants is minimized in a global analysis. Time-resolved absorption spectra S[t,l] were globally fit to sums of exponentials using the function

S[t,l] = å {ai[l] exp(- ni t)} + d[l] [6] where ai[l] is the observed wavelength-dependent amplitude and ni is the observed rate constant of a kinetic phase i, and d[l] is the baseline offset. The parameters ai[l], d[l] and ni were calculated by iterative reconvolution based on Marquardt algorithm using the kinetic software Look (Imperial College, UK). From the Beer-Lambert law, eq [6] can be rewritten as

S[t,l] = å ei[l] ci[t] [7] where ei[l] is the extinction coefficint of a species i at wavelength l and ci[t] is the concentration of a species i at time t. For a simple one-step reaction involving the decay of species A into B, eqs [6] and [7] become

F[t,l] = a[l] exp(- n t) + d[l] [8]

F[t,l] = å eo[l] A[t] + ef[l] B[t] [9] where A[t] and B[t] are the concentrations of A and B at time t respectively, and eo[l] and ef[l] are the extinction coefficients of A and B respectively. Assuming that the initial concentration of A is 1, then the concentrations of A and B are given by

A[t] = exp(- nt) [10] B[t] = 1 - exp(- nt) [11]

Substituting eqs [10] and [11] into eq [9] we have

F[t,l] = {eo[l] - ef[l] } exp(- nt) + ef[l] [12]

Equating pre- and non-exponential terms of eqs [8] and [12] gives

a[l] = eo[l] - ef[l] [13] d[l] = ef[l] [14]

Adding eqs [13] and [14] gives

eo[l] = a[l] + d[l] [15]

Materials and Methods 85

A number of important conclusions for the reaction of retinal with bacterioopsin can be drawn from ab initio calculations presented above. Eq [13] implies that the wavelength-dependent amplitude of a kinetic phase is the difference between the absorption spectra of the decaying and the non-decaying species. The change in the oscillator strength of the retinal absorption band is thus calculated from the wavelength-dependent amplitude of a kinetic phase. Since oscillator strength is proportional to the area under an absorption spectrum (Gilbert & Baggot 1991), it follows that the ratio of the area under the positive band to that of the negative band in a wavelength-dependent amplitude (difference spectrum) is equal to the oscillator strength of the decaying species relative to the non-decaying species. Oscillator strength is a dimensionless parameter. It is a measure of how strong a molecule absorbs in response to electromagnetic radiation. The oscillator strength of a molecule is dependent upon a number of factors such as molecular geometry, solvent polarity and interaction with other chemical groups, and can thus be used to probe its chemistry. Eq [14] shows that the kinetic absorption spectrum of the non-decaying (final) species is equal to the baseline offset. That is, the kinetic absorption spectrum of bacteriorhodopsin is determined from the baseline offset. Eq [15] indicates that the kinetic absorption spectrum of the decaying (starting) species is equal to the sum of the amplitude of a kinetic phase and the baseline offset. In the case of a multi- exponential, the sum of the amplitudes of all kinetic phases and the baseline offset should give the spectrum of the starting species. The assumption made here is that exponentials can be treated as independent variables. The absorption spectrum of retinal was thus determined by summing up the amplitudes of all kinetic phases and the baseline offset. The reaction of retinal and bacterioopsin proceeds via a spectroscopically distinct intermediate absorbing maximally around 430 nm. The kinetic absorption spectrum of this 430-nm intermediate was calculated by subtracting the wavelength-dependent amplitude of the kinetic phase, corresponding to the decay of retinal into the intermediate, from the kinetic absorption spectrum of retinal.

86 Chapter 2

2.5.5 Pseudo-First-Order Kinetics Free energy changes accompanying the bimolecular steps involving retinal and bacterioopsin were determined from experiments in which the concentration of retinal used was in excess over that of bacterioopsin. Retinal (R) binds to a state of bacterioopsin with a native-like secondary structure in the micelles (bO¢). The product of this reaction is the 430-nm intermediate (I430). Consider this reaction

k +R I bO¢ + R 430 k -R where k+R is the second-order intrinsic association rate constant and k-R is the first-order intrinsic dissociation rate constant. The equilibrium constant K is given by

K = [I430¥] / {[bO¢¥] [R¥]} [16] where [bO¢¥], [R¥] and [I430¥] are the concentrations of the species at equilibrium. At equilibrium, the rate of the forward reaction is equal to that of the backward reaction, that is

k+R [bO¢¥] [R¥] = k-R [I430¥] [17]

From eqs [16] and [17] we have

K = k+R / k-R [18] If retinal is in large excess over bacterioopsin, that is

[Ro] » [bO¢o] where [Ro] and [bO¢o] are the initial concentrations of the species, then the concentration of retinal practically remains constant throughout the reaction. Eq [17] thus becomes

k+R [bO¢¥] [R] = k-R [I430¥] [19] where [R] is the concentration of retinal at time t. Since k+R [R] is a constant in eq [19], we can define it as

k¢ = k+R [R] [20] From Eqs [19] and [20] we have

k¢ [bO¢¥] = k-R [I430¥] [21]

Materials and Methods 87

The concentration of bO¢ at equilibrium is given by

[bO¢¥] = [bO¢o] - [I430¥] [22]

Rearranging eqs [21] and [22] gives

[bO¢o] = {(k-R / k¢) [I430¥]} + [I430¥] [23]

The rate equation for the formation of the 430-nm intermediate is given by

d[I430] / dt = k¢ [bO¢] - k-R [I430] [24] where [bO¢] and [I430] are the concentrations of the species at time t. The concentration of bO¢ at time t is given by

[bO¢] = [bO¢o] - [I430] [25]

From eqs [23] and [25] we have

[bO¢] = {(k-R / k¢) [I430¥]} + [I430¥] - [I430] [26]

Eliminating [bO¢] from eqs [24] and [26] gives

d[I430] / dt = (k-R + k¢) {[I430¥] - [I430]} [27]

Eq [27] has the same form as a first-order reaction shown in eq [2]. The experimentally observed first- order rate constant nR in eq [27] is given by

nR = k-R + k¢ [28]

From eqs [20] and [28] we have

nR = k+R [R] + k-R [29]

A plot of nR versus [R] should thus yield a straight line with a slope of k+R and a y-intercept of k-R. The equilibrium constant (K) for the bimolecular steps involving retinal and bacterioopsin can then be calculated using eq [18]. Free energy change (DG) for these bimolecular reactions were determined from the equation DG = - RTlnK [30] where R is the universal molar gas constant and T is the absolute temperature.

88 Chapter 2

2.5.6 Singular Value Decomposition Time-resolved absorption spectra for the reaction of retinal with bacterioopsin were subjected to singular value decomposition (SVD) analysis in order to identify the number of independently absorbing species in the reaction of bacterioopsin with retinal. The SVD algorithm (Henry & Hofrichter 1992) decomposes a data matrix Y (time ´ wavelength) containing a set of time-dependent spectra into a matrix product

Y = USVT where columns of V contain time-dependent amplitudes that are related to the time-dependent concentration profiles of physical intermediates, the columns of U contain wavelength-dependent amplitudes (basis spectra) that are related to the absolute spectra of physical intermediates (VT is the transpose of V) and the diagonal elements of S constitute singular values, that is, weighting terms needed to normalize elements in U and V to the corresponding elements in Y. The result of a typical SVD analysis consists of a series of wavelength-dependent amplitudes, weighted by their singular values, and time-dependent amplitudes. Typically, a plot of singular values falls from the first and largest value and plateaus out at the value characteristic of the number of spectrally distinct species. SVD analysis thus gives a model-free indication of the number of independently absorbing species in a reaction.

2.5.7 Scattering Analysis Light scattering L[l] by the micelles was analyzed by a Rayleigh function

L[l] = (b / l4) + d where b is a constant related to Rayleigh factor, d is the baseline offset and l is the wavelength. Fitting was accomplished by iterative reconvolution based on Marquardt algorithm using Origin (Microcal, USA).

Materials and Methods 89

2.5.8 Gaussian Decomposition The kinetic absorption spectrum calculated for the 430-nm intermediate, from a global analysis of time- resolved absorption spectra for the reaction of retinal with bacterioopsin, includes the contribution of retinal because of the transient nature of the intermediate. To determine the absolute absorption spectrum of the 430-nm intermediate in order to obtain its absorbance maximum, the calculated spectrum of the intermediate e[l] was decomposed into two absorption components using Gaussian and Rayleigh functions simultaneously

1/2 2 4 e[l] = å {ai / si(2p) } exp{-1/2 [(l - li) / si] } + (b / l ) + d

where ai, si and li are respectively the area, 0.425 times the full width half maximum and absorbance maximum of a Gaussian peak i. l is the wavelength, b is a constant related to the Rayleigh factor and d is the baseline offset. Fitting was achieved by iterative reconvolution based on Marquardt algorithm using Origin (Microcal, USA). The inclusion of Rayleigh function improves fitting by separating the contribution of light scattering by the micelles from retinal absorption. The decomposition was carried out as follows. First, the kinetic absorption spectrum of retinal was fitted to a single Gaussian and a Rayleigh function using the above function (i = retinal) with all parameters allowed to vary. The values obtained for sretinal and lretinal, that is 0.425 times the full width half maximum and the absorbance maximum of retinal, were subsequently locked during the fitting of the kinetic spectrum of the 430-nm intermediate to the above function (i = retinal, intermediate). This procedure decomposes the kinetic spectrum of the 430-nm intermediate into two absorption components: one for free retinal and one for the intermediate. The absorbance maximum of the intermediate can then be obtained from its absolute absorption spectrum.

Chapter 3

EQUILIBRIUM STUDIES OF BACTERIORHODOPSIN REFOLDING

3.1 Overview 92

3.2 Results 93 3.2.1 Bacteriorhodopsin Refolding at pH 6 93 3.2.2 Bacteriorhodopsin Refolding at pH 8 99

3.3 Discussion 104

92 Chapter 3

3.1 Overview

Bacteriorhodopsin can be refolded in mixed phospholipid/detergent micelles from a partially denatured state in SDS or from a completely denatured state in an organic solvent such as formic acid or trifluoroacetic acid (Huang et al 1980, Huang et al 1981). The extent of refolding can be quantified from the appearance of a native-like chromophore absorption band (London & Khorana 1982, Braiman et al 1987, Brouillette et al 1989). This band results from a shift in the absorption maximum of retinal from 380 nm in free retinal to 560 nm in the native chromophore. The extent of refolding of bacteriorhodopsin is highly sensitive to pH of the aqueous environment. The refolding yield of bacteriorhodopsin as a function of pH consists of a bell-shaped profile with a pH optimum of around 6 (Liao et al 1983). The refolding yield is defined as

concentration of native protein bacteriorhodopsin refolding yield = 100 % concentration of apoprotein bacterioopsin

The refolding yield of bacteriorhodopsin can thus be quantified if the molar extinction coefficients of both the bacterioopsin and bacteriorhodopsin are known. In this chapter, steady-state absorption and fluorescence spectroscopy are used to measure the refolding yields of bacteriorhodopsin from several bacterioopsin preparations in mixed DMPC/CHAPS micelles from a denatured state in SDS at pH 6 and pH 8. A knowledge of the yield of refolding is crucial for a comprehensive interpretation of refolding kinetics of bacteriorhodopsin. Bacterioopsin preparations with a low refolding yield are undesirable for kinetic studies. The refolding of bacteriorhodopsin is assayed by the recovery of a native-like chromophore absorption band upon the addition of retinal to bacterioopsin. The extent of the recovery of this band provides a quantitative measure of the extent of refolding of bacteriorhodopsin. In addition, it is shown that the quenching of intrinsic protein fluorescence upon the binding of retinal to bacterioopsin also provides a sensitive measure of bacteriorhodopsin refolding. How the refolding yield depends on the pH and the mode of addition of retinal, that is whether refolding of bacterioopsin is initiated in the presence of retinal or whether retinal is added after bacterioopsin has been allowed to pre-equilibrate with the micelles, will be described.

Equilibrium Studies of Bacteriorhodopsin Refolding 93

3.2 Results

3.2.1 Bacteriorhodopsin Refolding at pH 6

Bacteriorhodopsin was regenerated in DMPC/CHAPS micelles at a range of retinal concentrations in order to produce a plot of chromophore absorbance versus retinal concentration. It should be possible to obtain a value for the molar extinction coefficient of bacteriorhodopsin from this plot (Rehorek & Heyn 1979). A molar extinction coefficient for bacterioopsin denatured in SDS has been calculated previously (London & Khorana 1982). Thus using these extinction coefficients, the extent of bacteriorhodopsin refolding can be determined. Note that, retinal does not bind to bO partially denatured in SDS (bO/SDS). Fig 3.1a shows absorption spectra for bacteriorhodopsin refolding at pH 6 for the addition of retinal/DMPC/CHAPS to bO/SDS. The spectra were recorded at various retinal concentrations from 0- 8 mM. The protein concentration was constant at 4 mM. Only some of these spectra are shown. The absorption spectral changes observed are characteristic of the formation of a native-like chromophore of bacteriorhodopsin (Huang et al 1981), with an absorption band centred at 555 nm. The absorption band seen at 380 nm is that of free retinal when it is added in excess. Fig 3.1b shows fluorescence spectra for bacteriorhodopsin refolding at pH 6 for the addition of retinal/DMPC/CHAPS to bO/SDS. The spectra were recorded at various retinal concentrations from 0- 8 mM. The protein concentration was constant at 4 mM. Only some of these spectra are shown. As can be seen, binding of retinal to bacterioopsin leads to quenching of protein fluorescence. This is demonstrated by a reduction in the intensity of protein fluorescence band in the presence of retinal. Fig 3.1c shows a plot of chromophore absorbance at 555 nm versus retinal concentration. The chromophore absorbance increases linearly with increasing retinal concentration up to about 3.5 mM, which is only slightly less than the bacterioopsin concentration of 4 mM. Thus, in line with previous reports (Huang et al 1981, London & Khorana 1982, Braiman et al 1987), nearly all bacterioopsin binds retinal to regenerate bacteriorhodopsin. From the initial slope of this plot, a molar extinction coefficient of 55 300 ± 800 cm-1M-1 can be calculated for bacteriorhodopsin in DMPC/CHAPS/SDS

94 Chapter 3

Fig 3.1 (opposite). Dependence of refolding yield of bacteriorhodopsin on bO preparation. Retinal in DMPC/CHAPS micelles was manually mixed with an equal volume of bO/SDS, at a range of retinal concentrations. After overnight incubation in the dark, absorption and fluorescence spectra were measured. Fluorescence spectra were obtained using an excitation wavelength of 290 nm and emission was measured between 300-500 nm. The final protein concentration was constant at 4 mM.

(a) Absorption spectra, in the order of increasing absorbance at 555 and/or 380 nm, correspond to final retinal concentrations of 0, 0.5, 1.0, 2.0, 3.0, 4.0, 6.0 and 8.0 mM, respectively.

(b) Fluorescence spectra, in the order of decreasing fluorescence intensity, correspond to final retinal concentrations of 0, 0.5, 1.0, 2.0, 3.0, 4.0, 6.0 and 8.0 mM, respectively.

(c) The chromophore absorbance at 555 nm ( ) and fluorescence yield (·) as a function of retinal concentration are shown. Fluorescence yield was calculated by integrating under the fluorescence band from 300-500 nm. Each data point is an average of five experiments on different bO preparations. Error bars are given to one standard deviation.

Equilibrium Studies of Bacteriorhodopsin Refolding 95 0.4 a

0.3

0.2 Absorbance

0.1

0.0 250 300 350 400 450 500 550 600 650 700 750 Wavelength / nm

b 0.06

0.04

0.02

0Relative fluorescence / arbitrary units .00 300 350 400 450 500 Wavelength / nm

0.25 c

4 0.20

0.15 3

0.10 2 Absorbance at 555 nm

0.05 1 Relative fluorescence yield / arbitrary units 0.00 0 0 1 2 3 4 5 6 7 8 Retinal concentratiomnM /

96 Chapter 3 micelles at 555 nm (pH 6). This value is in good agreement with that of 56 600 cm-1M-1 at 558 nm (pH 6) which has been previously reported for this refolding system but with slightly different ratios of protein, DMPC, CHAPS and SDS (Brouillette et al 1989). Micelles containing native bacteriorhodopsin will be referred hereinafter as DMPC/CHAPS instead of DMPC/CHAPS/SDS for simplicity. This is because SDS probably plays no active role once the protein has been refolded in the micelles. SDS can be done away with almost completely without any observable change in the refolding yield of bacteriorhodopsin. The yield of refolding of bacteriorhodopsin was determined from the ratio of the concentration of regenerated bacteriorhodopsin to the initial bacterioopsin concentration. The concentration of regenerated bacteriorhodopsin was measured at 555 nm using an extinction coefficient of 55 300 cm-1M- 1 calculated here at pH 6. The concentration of bacterioopsin was measured in SDS before adding micelles at 280 nm using an extinction coefficient of 66 000 cm-1M-1 (London & Khorana 1982). An average of five different bO preparations gave a refolding yield of 88 ± 16 %. Several bO preparations differ in that they are delipidated separately but the source of purple membrane may be the same. Also shown in Fig 3.1c, is the change in intrinsic protein fluorescence as a function of retinal concentration. Plotted is the fluorescence yield versus retinal concentration. Fluorescence yield was calculated by integrating under the fluorescence band from 300-500 nm. Fluorescence yield decreases linearly with increasing retinal concentration and thus mirrors the change in chromophore absorbance. Another way in which bacteriorhodopsin can be regenerated is to add retinal to bO pre- equilibrated with the micelles (bO/DMPC/CHAPS). It is well established that the apoprotein bacterioopsin can refold to a state with a native-like secondary structure in mixed phospholipid/detergent micelles (London & Khorana 1982, Riley et al 1997). This is therefore an important experiment in that it is possible that the apoprotein could undergo irreversible aggregation or misfolding in the micelles in the absence of retinal. If this phenomenon is prevalent here, it would be reflected in the overall refolding yield of bacteriorhodopsin. Absorption and fluorescence spectra of regenerated bacteriorhodopsin at pH 6 for the addition of retinal/DMPC/CHAPS to bO/DMPC/CHAPS were recorded at a range of retinal concentrations as above (not

Equilibrium Studies of Bacteriorhodopsin Refolding 97

0.25

4 0.20

3 0.15

0.10 2 Absorbance at 555 nm

0.05 Relative fluorescence / arbitrary units 1

0.00

0 0 1 2 3 4 5 6 7 8 Retinal concentratiomnM /

Fig 3.2. Dependence of refolding yield of bacteriorhodopsin on the mode of retinal addition (pH 6). Retinal in DMPC/CHAPS micelles was manually mixed with an equal volume of bO/SDS or bO/DMPC/CHAPS, at a range of retinal concentrations. bO/DMPC/CHAPS was prepared by mixing manually bO/SDS with an equal volume of DMPC/CHAPS and the mixture allowed to equilibrate for 30 min. After overnight incubation in the dark, absorption and fluorescence spectra were measured as described in the legend to Fig 3.1. The final protein concentration was constant at 4 mM.

The chromophore absorbance ( ) and fluorescence yield (·) for the addition of retinal/DMPC/CHAPS to bO/SDS and, chromophore absorbance ( ) and fluorescence yield (O) for the addition of retinal/DMPC/CHAPS to bO/DMPC/CHAPS are shown. Each data point is an average of three experiments on the same bO preparation. Error bars are shown to one standard deviation.

98 Chapter 3 shown). There was no observable change in the absorption or fluorescence maxima from those shown in Figs 3.1a and 3.1b. Fig 3.2 shows changes in chromophore absorbance and protein fluorescence as a function of retinal concentration for bacteriorhodopsin regeneration for the addition of retinal/DMPC/CHAPS to bO/SDS and bO/DMPC/CHAPS. As can be seen, there is no difference in the overall extent of bacteriorhodopsin refolding whether bacterioopsin refolding is initiated in the presence of retinal or whether retinal is added after bacterioopsin has been allowed to pre-equilibrate with the micelles.

Equilibrium Studies of Bacteriorhodopsin Refolding 99

3.2.2 Bacteriorhodopsin Refolding at pH 8

Refolding yields of bacteriorhodopsin were also measured at pH 8. It is known that some fraction of protein undergoes irreversible misfolding at pH 8 (London & Khorana 1982). It would however be interesting to see what effect the mode of addition of retinal has on the overall refolding yield of bacteriorhodopsin at this pH. Absorption and fluorescence spectra for bacteriorhodopsin refolding at pH 8 for the addition of retinal/DMPC/CHAPS to bO/SDS and bO/DMPC/CHAPS were recorded at a range of retinal concentrations as above (not shown). There were some notable changes in the spectral properties of bacteriorhodopsin regenerated here compared to those shown in Figs 3.1a and 3.1b. The chromophore absorbance was centred at 548 nm. A comparison of the spectral parameters of bacteriorhodopsin refolding at pH 6 and pH 8 is given in Table 3.1. Fig 3.3 shows the change in chromophore absorbance as a function of retinal concentration for bacteriorhodopsin refolding at pH 8 for the two modes of retinal addition (addition of retinal/DMPC/CHAPS to bO/SDS or bO/DMPC/CHAPS). The chromophore absorbance increases linearly up to a retinal concentration of about 2.5 mM for the addition of retinal/DMPC/CHAPS to bO/SDS. In contrast, a linear increase in the chromophore absorbance for the addition of retinal/DMPC/CHAPS to bO/DMPC/CHAPS is only observed up to about 1 mM retinal. The initial slopes corresponding to a change in chromophore absorbance in both experiments are the same. Also shown in Fig 3.3, is the dependence of fluorescence yield on retinal concentration for the two modes of retinal addition. The change in intrinsic protein fluorescence mirrors the change in chromophore absorbance, as observed previously. From Fig 3.3, it is clearly evident that the extent of bacteriorhodopsin refolding at pH 8 is dependent upon whether bacterioopsin refolding is initiated in the presence of retinal or whether retinal is added after bacterioopsin has been allowed to equilibrate with the micelles. From the initial slope of the plot of chromophore absorbance versus retinal concentration for the addition of retinal/DMPC/CHAPS to bO/SDS, a value of 53 500 ± 700 cm-1M-1 is calculated for the extinction coefficient of bacteriorhodopsin in DMPC/CHAPS micelles at 548 nm (pH 8). This value is of course the same for the plot of chromophore absorbance versus retinal concentration for the addition of

100 Chapter 3

0.16 4

0.14

0.12 3 0.10

0.08

Absorbance at 548 nm 2 0.06

0.04

1 Relative fluorescence / arbitrary units 0.02

0.00

0 0 1 2 3 4 5 6 7 8 Retinal concentratiomnM /

Fig 3.3. Dependence of refolding yield of bacteriorhodopsin on the mode of retinal addition (pH 8). Retinal in DMPC/CHAPS micelles was manually mixed with an equal volume of bO/SDS or bO/DMPC/CHAPS, at a range of retinal concentrations. bO/DMPC/CHAPS was prepared by mixing manually bO/SDS with an equal volume of DMPC/CHAPS and the mixture allowed to equilibrate for 30 min. After overnight incubation, absorption and fluorescence spectra were measured as described in the legend to Fig 3.1. The final protein concentration was constant at 4 mM.

The chromophore absorbance ( ) and fluorescence yield (·) for the addition of retinal/DMPC/CHAPS to bO/SDS and, chromophore absorbance ( ) and fluorescence yield (O) for the addition of retinal/DMPC/CHAPS to bO/DMPC/CHAPS are shown. Each data point is an average of three experiments on the same bO preparation (same as the one used for experiments described in Fig 3.2). Error bars are shown to one standard deviation.

Equilibrium Studies of Bacteriorhodopsin Refolding 101 retinal/DMPC/CHAPS to bO/DMPC/CHAPS. A value of 48 000 cm-1M-1 has been previously reported for bacteriorhodopsin in DMPC/cholate micelles at pH 8 (London & Khorana 1982). Using this extinction coefficient, the refolding yield for bacteriorhodopsin at pH 8 can be calculated. Table 3.2 compares the dependence of refolding yields of bacteriorhodopsin on pH and the mode of addition of retinal.

102 Chapter 3

Table 3.1. Comparison of chromophore absorbance and protein fluorescence maxima and their FWHM for bacteriorhodopsin refolding at pH 6 and pH 8.

Chromophore absorbance Protein fluorescence

pH lmax / nm FWHM / nm lmax / nm FWHM / nm

6 555 ± 1 110 ± 2 337 ± 1 56 ± 1

8 548 ± 2 122 ± 2 340 ± 1 65 ± 1

Footnotes to Table 3.1

The values shown for the various spectral parameters are for the addition of retinal/DMPC/CHAPS to bO/SDS. Similar values were obtained for the addition of retinal/DMPC/CHAPS to bO/DMPC/CHAPS.

The values quoted for the various parameters were independent of bO preparation and retinal concentration.

Errors were calculated from three experiments on the same bO preparation at a protein to retinal molar ratio of 1:1. All errors are given to one standard deviation.

Equilibrium Studies of Bacteriorhodopsin Refolding 103

Table 3.2. Dependence of bacteriorhodopsin refolding yield on pH and the mode of retinal addition.

Refolding yield / % Addition of Addition of pH retinal/DMPC/CHAPS retinal/DMPC/CHAPS to bO/SDS to bO/DMPC/CHAPS

6 98 ± 2 98 ± 1

8 66 ± 3 51 ± 2

Footnotes to Table 3.2

Refolding yield was calculated from the ratio of the concentration of regenerated bacteriorhodopsin to bacterioopsin concentration (see text).

Errors were calculated from three experiments on the same bO preparation at a protein to retinal molar ratio of 1:1. All errors are given to one standard deviation.

104 Chapter 3

3.3 Discussion

Bacteriorhodopsin regenerated in mixed phospholipid/detergent micelles exhibits all the photoreactions characteristic of the native protein in the purple membrane (Braiman et al 1987). The binding of retinal to bacterioopsin is accompanied by a change in the absorption maximum of retinal from 380 nm to around 550 nm. This change in the absorption maximum of retinal is characteristic of the formation of a native-like chromophore of bacteriorhodopsin (Huang et al 1981, London & Khorana 1982). Retinal is covalently bound to Lys216 via a Schiff base within the apoprotein bacterioopsin (Bayley et al 1981, Lemke & Oesterhelt 1981, Mullen et al 1981). The Schiff base between retinal and the protein is not directly responsible for the red shift observed in the absorption maximum of retinal within the native chromophore (Schaffer et al 1974). Protonation of a Schiff base between retinal and an amine however appears to shift the absorption maximum of retinal to about 440 nm (Pitt et al 1955, Liu et al 1993). It is believed that a number of interactions between the protonated Schiff base of retinal and the amino acid residues within the protein are responsible for the red shift observed in the absorption maximum of retinal within the native chromophore (Nakanishi et al 1980, Beppu & Kakitani 1994, Hu et al 1994). The fact that both the chromophore absorbance and protein fluorescence quenching increase linearly with increasing retinal concentration implies that they provide convenient assays of retinal binding to bacterioopsin. The mechanism by which retinal quenches protein fluorescence is thought to be due to energy transfer from Trp residues to retinal (Polland et al 1986a). There are at least four Trp and three Tyr residues within the retinal binding pocket (Henderson et al 1990, Grigorieff et al 1996). The fact that tyrosine fluorescence in bacteriorhodopsin is fully quenched by Trp residues rules out the possibility that energy transfer from Tyr residues to retinal may also be responsible for the quenching of protein fluorescence (Kalisky et al 1981). The extent of recovery of native-like chromophore absorption band on regeneration of bacteriorhodopsin provides a convenient measure of the refolding yield of bacteriorhodopsin (Oesterhelt & Schuhmann 1974, Huang et al 1981). The linear region of the chromophore absorbance versus retinal concentration plots is indicative of the fact that every retinal which is added binds. Bacteriorhodopsin concentration in this linear region is thus equivalent to the concentration of added retinal, assuming a 1:1

Equilibrium Studies of Bacteriorhodopsin Refolding 105 protein to retinal stoichiometry. From the Beer-Lambert law, the slope of this linear region is thus equivalent to the molar extinction coefficient of regenerated bacteriorhodopsin (Rehorek & Heyn 1979). The refolding yield is a ratio of the concentration of regenerated bacteriorhodopsin to bacterioopsin concentration. The accuracy of the refolding yield depends on the accuracy with which the extinction coefficients and bacterioopsin concentration can be determined. The accuracy of the extinction coefficients determined here for bacteriorhodopsin at pH 6 and pH 8 is dependent on the accuracy with which the concentration of retinal can be measured. The concentration of free retinal in ethanol was calculated using an extinction coefficient of 42 800 cm-1M-1 at 380 nm (Rehorek & Heyn 1979). An accurate knowledge of bacterioopsin concentration is not required for the determination of extinction coefficients although it is essential for the determination of refolding yield. The concentration of bacterioopsin was measured in SDS using an extinction coefficient of 66 000 cm-1M-1 (London & Khorana 1982). The refolding yields calculated here for bacteriorhodopsin under different conditions provide some interesting clues to protein folding in general. At pH 6, an average of five different bO preparations gave a refolding yield of 88 ± 16 %, in agreement with previous studies (Huang et al 1981, London & Khorana 1982, Braiman et al 1987). Although the exact nature of this spread is not known, it probably results from the delipidation of purple membrane. Refolding yields of up to nearly 100 % were observed at pH 6 for bO preparations where during delipidation the protein coalesced into a single layer at the organic-aqueous interface and was thus easily recovered. However during some delipidations, the protein did not coalesce well at the aqueous-organic interface and this resulted in lower refolding yields of about 70 %. An alternative explanation for the spread observed in the refolding yields of bacteriorhodopsin at pH 6 is that for bO preparations where during SDS-solubilization, the delipidated protein took less than two hours to dissolve generally gave higher refolding yields than those in which the delipidated bO took overnight. Freeze-thaw cycles did not appear to affect the refolding yields of bacteriorhodopsin. Far UV-CD studies on bacterioopsin pre-equilibrated with DMPC/cholate micelles (London & Khorana 1982) and DMPC/DHPC micelles (Riley et al 1997) indicate that the apoprotein ontains a full native secondary structure that is

106 Chapter 3 indistinguishable from that in the native chromophore. It is therefore possible that this state of bO in the micelles represents a partially refolded protein, and that retinal is not necessary for initiating bacterioopsin refolding. This partially refolded state of bO is not to be confused with the refolding yields of bacteriorhodopsin which correspond to the native chromophore (Table 3.2). The refolding yield describes the percentage of the apoprotein that can be transformed into the native protein upon the addition of retinal, whereas the partially refolded state of bO in the micelles indicates that it structurally resembles the native protein. In terms of the refolding assay based on chromophore absorbance for the recovery of the native chromophore, the partially refolded state of bO in the phospholipid/detergent micelles or partially denatured bO in SDS represent zero yield of refolding. Far UV-CD studies on bO pre-equilibrated with DMPC/CHAPS micelles cannot be carried out as the zwitterionic detergent CHAPS absorbs strongly in the UV region. It is however very likely that bO pre-equilibrated with DMPC/CHAPS micelles also reflects a state with a native secondary structure. This state of bO with a native-like secondary structure in the micelles will be referred hereinafter as bO¢. Addition of retinal/DMPC/CHAPS to bO/SDS (initiation of bacterioopsin refolding in the presence of retinal) or bO/DMPC/CHAPS (partially refolded apoprotein) has no observable effect on the refolding yield of bacteriorhodopsin at pH 6. This implies that bacterioopsin does not irreversibly misfold or aggregate upon refolding in the absence of retinal. A model that accommodates these observations is

R

bO bR

where bO represents the denatured state of bacterioopsin in SDS, bO¢ is a partially refolded state of bacterioopsin in the micelles to which retinal apparently binds to quantitatively regenerate the native chromophore bacteriorhodopsin (bR). Note that, the state bO¢ is envisaged here as an intermediate in bacteriorhodopsin folding. This is a bona fide argument for retinal does not appear to bind to the partially denatured state of

Equilibrium Studies of Bacteriorhodopsin Refolding 107 bO in SDS. Only upon the addition of DMPC/CHAPS micelles is the binding of retinal observed. This combined with the fact that the secondary structure of bO in the micelles is indistinguishable from that of the native protein strongly implies an intermediary role of the state bO¢ in bacteriorhodopsin folding. Only kinetic studies of bacteriorhodopsin refolding can however confirm this speculation. It is conceivable that bO misfolds in the micelles in the absence of retinal. These misfolded species however must be able to reconvert to the state bO¢ in order for retinal to bind and result in quantitative regeneration of bacteriorhodopsin that is observed here at pH 6. The refolding yield of bacteriorhodopsin is typically 30 % lower at pH 8 relative to pH 6 (Table 3.2), in agreement with previous reports (London & Khorana 1982, Liao et al 1983). The refolding of bacteriorhodopsin thus appears to be pH-dependent. This implies that the ionization state of the charged amino acid residues within the protein and the amphiphiles within the micelles must therefore be critical in the folding process. The effect of raising pH from 6 to 8 may lead to an increase in the negative charge on the surface of the micelles as well as within the protein. This would, of course, result in the electrostatic repulsion between the micelles and the protein molecules. The insertion of protein into the micelles must therefore become energetically less favourable. An increase in the pH of the aqueous environment may thus simply cause a shift in the equilibrium in the direction of the partially denatured protein in SDS. This could explain a reduction in the refolding yield of bacteriorhodopsin observed at pH 8 relative to pH 6. It is conceivable that an increase in pH induces bacterioopsin misfolding or aggregation, or somehow leads to off-pathway reactions instead of productive folding. Indeed, the FWHM of both the chromophore absorption band and the protein fluorescence band become pronounced at pH 8 relative to pH 6 (Table 3.1), implying that an increase in pH leads to an increase in the heterogeneity of bacteriorhodopsin and that of the denatured state in SDS. A slight increase in the fluorescence maximum of bacteriorhodopsin at pH 8 relative to pH 6 suggests that the aromatic residues are overall in a less hydrophobic environment at the higher pH. This is understandable in view of the fact that an increase in pH may confer additional negative charge on the protein and the micelles. Similar arguments may also account for a blue shift observed in the chromophore absorbance maximum at pH 8 relative to that at pH 6, which may be

108 Chapter 3 explained by a change in the polarity of the retinal binding pocket upon increasing pH of the aqueous environment. The refolding yield of bacteriorhodopsin at pH 8 is dependent upon the mode of addition of retinal. A decrease of about 15 % in the refolding yield is observed if retinal is added to partially refolded bacterioopsin (bO¢) instead of denatured bacterioopsin in SDS (Table 3.2). This implies that in the absence of retinal some fraction of apoprotein becomes irreversibly misfolded and cannot subsequently bind retinal (a behaviour not encountered at pH 6). This fraction of misfolded apoprotein cannot be recovered even when retinal is added in excess. Thus although retinal may not be strictly required for initiating bacterioopsin refolding, it clearly appears to aid productive folding over protein misfolding. This action of retinal could be explained by a shift in the equilibrium in the direction of productive folding relative to off-pathway reactions during apoprotein refolding. A model that is consistent with these results is

bO¢

bO R bR

bO*

where bO¢ represents a state of bacterioopsin with a native-like secondary structure in the micelles to which retinal binds, and bO* an irreversibly misfolded state of bacterioopsin in the micelles which cannot bind retinal. The probability of the formation of bO¢ relative to bO* is the same in the absence of retinal (refolding yield of bacteriorhodopsin at pH 8 is about 50 % if retinal is added to bO pre-equilibrated with the micelles). Initiation of bacterioopsin refolding in the presence of retinal would however shift the equilibrium in the direction of bO¢ by about 15 %. Preliminary experiments indicate that the species bO* can be induced to enter the productive pathway upon switching the pH of the aqueous solution to 6. That is, the

Equilibrium Studies of Bacteriorhodopsin Refolding 109 fraction of the protein that is irreversibly misfolded and cannot bind retinal at pH 8 will bind retinal to regenerate bacteriorhodopsin at pH 6. Misfolded protein aggregates are of interest as studies on these may provide an insight into the mechanisms that interfere with folding reactions (Creighton 1994, Sosnik et al 1994, Shortle 1996). The foregoing argument may lead one to speculate that misfolding of bacteriorhodopsin at pH 8 may also occur in vivo. One might therefore conclude that it would be beneficial for halobacteria to maintain their cytoplasmic pH around 6. Halobacteria are found in natural brines where environmental conditions can become extreme (Stanier et al 1987). In particular when oxygen tension in the external environment is low, bacteriorhodopsin is the only system that is capable of generating ATP that is critical for the viability of the halobacterial cell (Stoeckenius et al 1979). Under low oxygen tension, misfolding of bacteriorhodopsin must be a highly forbidden rule. One may thus suspect that halobacteria employ mechanisms that ensure that an optimal cytoplasmic pH is maintained regardless of the external conditions. In practice, however, this may not necessarily be the case. It is now widely recognized that chaperones play a critical role in making protein folding an highly efficient process in vivo (Martin & Hartl 1993, Rassow & Pfanner 1996, Martin & Hartl 1997). It is thus possible that the involvement of chaperones in vivo may prevent bacteriorhodopsin misfolding at pH 8. Of course, no chaperones associated with bacteriorhodopsin folding have yet been discovered. A chaperonin-like protein has been found in the periplasm that is thought to assist the folding of bacterial porins OmpA and OmpF (Lazar & Kolter 1996). In summary, the results presented here indicate that bacteriorhodopsin can be quantitatively refolded at pH 6. The refolding yield at pH 6 is independent of the mode of addition of retinal, that is, whether retinal is added at the beginning or after bacterioopsin has been allowed to equilibrate with the micelles. The knowledge of the refolding yield is important for the interpretation of kinetic data on the refolding studies of bacteriorhodopsin. Time-resolved studies require assignment of kinetic phases to folding events. Such assignments can be made easier if the protein can be quantitatively refolded, as possibilities for protein misfolding or aggregation can be excluded.

Chapter 4

KINETICS OF BACTERIOOPSIN REFOLDING

4.1 Overview 112

4.2 Results 113 4.2.1 Addition of DMPC/CHAPS to bO/SDS 113 4.2.2 Addition of DMPC/CHAPS to SDS 117

4.3 Discussion 122

112 Chapter 4

4.1 Overview

Several lines of evidence exist in support of the notion that retinal is not required for initiating the folding of the apoprotein bacterioopsin. Far-UV CD studies indicate that bacterioopsin denatured in SDS contains nearly 60 % helices relative to the native protein bacteriorhodopsin (London & Khorana 1982, Riley et al 1997). To what extent these helices represent native secondary structure elements remains arguable. The helical content of bacterioopsin pre-equilibrated with phospholipid/detergent micelles is indistinguishable from that of the native chromophore (London & Khorana 1982). This implies that bacterioopsin in phospholipid/detergent micelles represents a state with a native-like secondary structure. This finding is corroborated by FTIR spectroscopy in which the removal of retinal from bacteriorhodopsin in the purple membrane has a little effect on the secondary structure of the apoprotein bacterioopsin (Cladera et al 1996). In the previous chapter it has indeed been shown that the binding of retinal to bacterioopsin only occurs upon the transfer of bO/SDS to DMPC/CHAPS micelles, suggesting that retinal probably binds to a partially refolded state of bacterioopsin in the micelles. It would thus be interesting to study the kinetics of how partially denatured bacterioopsin in SDS is transformed into a state with a complete native secondary structure in mixed phospholipid/detergent micelles. Refolding of bacterioopsin in DMPC/CHAPS micelles has been previously investigated using stop-flow fluorescence (Booth et al 1995). In this report, bacterioopsin refolding was shown to be mutiphasic. However, protein fluorescence is a non-specific probe of protein folding in that it provides a qualitative measure of the polarity of the surrounding environment of aromatic residues. The observed change in protein fluorescence could well result from a change in the architecture of micelles rather than a change in protein structure. Thus, structural changes within the micelles need to be characterized in order for a comprehensive understanding of the refolding kinetics of bacterioopsin. In this chapter, the refolding of bacterioopsin in DMPC/CHAPS micelles is investigated using stop-flow fluorescence and absorption spectroscopy. Protein fluorescence is used to probe the overall changes in the surrounding environment of aromatic residues during folding. The change in absorbance, resulting from the scattering of light by the micelles, is used to specifically differentiate structural changes within the micelles from protein folding events.

Kinetics of Bacterioopsin Refolding 113

4.2 Results

4.2.1 Addition of DMPC/CHAPS to bO/SDS

Kinetics of mixing bO partially denatured in SDS (bO/SDS) with DMPC/CHAPS micelles should provide some infromation regarding the mechanism of the formation of a state of bO with a native-like secondary structure in the micelles. Fig 4.1a shows changes in intrinsic protein fluorescence observed upon the addition of DMPC/CHAPS to bO/SDS. Data were collected over various time scales between 0.1 to 1000 s. Only two such time scales are shown. Time-resolved data S[t] shown over the two time scales were simultaneously fit to a sum of four exponentials using the function

S[t] = å {ai exp(- ni t)} + d [4.1]

where ai and ni are respectively the amplitude and the observed rate constant of an exponential phase i,

-1 -1 -1 and d is the baseline offset. The rate constants obtained were 220 s (n1), 3 s (n2), 0.13 s (n3) and

-1 0.031 s (n4). Table 4.1 lists the amplitudes of these rate constants. Figs 4.1b and 4.1c respectively show the residuals obtained for the fit of a sum of three and four exponentials to the data shown in Fig 4.1a.

The phase n1 is accompanied by an initial rise in fluorescence and the subsequent decrease in fluorescence represents the phase n2. The phases n3 and n4 are accompanied by an overall rise in fluorescenece that follows the fluorescence changes associated with the phases n1 and n2. Data were also analyzed separately over different time scales. Data over time periods of 0-0.25 s and 0.25-250 s were best described by a two-exponential in each case. On the basis of these observations together with substantial improvement in the quality of the residuals shown in Fig 4.1c, the choice of a four- exponential to describe the data in Fig 4.1a is justified.

As can be seen from Fig 4.1a, nearly 90 % of the amplitude associated with the phase n1 occurs within the deadtime of the measurement. Thus, the rate constant associated with n1 cannot be accurately determined and the value given here is only approximate. The change in fluorescence that cannot be recorded during the deadtime amounts to about 60 % of the total fluorescence change observed upon adding bO/SDS

114 Chapter 4

Fig 4.1 (opposite). Addition of DMPC/CHAPS to bO/SDS. DMPC/CHAPS micelles were mixed, in the stop-flow cuvette, with an equal volume of bO/SDS. Refolding was monitored by intrinsic protein fluorescence.

(a) Data collected over first 250 s after initial mixing are shown. The time scale is split over 0-0.25 and 0.25-250 s. Data shown over 0-0.25 s are an average of eight transients taken over 0.5 s. Data shown over 0.25-250 s are an average of two transients taken over 1000 s. For each full scale, 4000 data points were collected. The data shown over the two time scales were simultaneously fit to a sum of four exponentials. The fit is indicated by a solid line. Note that the fluorescence levels of bO denatured in SDS (bO/SDS) and bO pre-equilibrated with the micelles for 30 min (bO/DMPC/CHAPS) are also shown.

(b) A plot of residuals for the fit of a sum of three exponentials to the data shown in (a).

(c) A plot of residuals for the fit of a sum of four exponentials to the data shown in (a).

(d) Data were collected over a full time scale of 1000 s with the light shutter open throughout the experiment (as above) and with the light shutter opened just before the end of the experiment. Data shown are an average of two transients in each case.

Kinetics of Bacterioopsin Refolding 115

3.5 a bO/DMPC/CHAPS

3.0 n 4 n 2.5 3 n n1 2 2.0

1.5 bO/SDS Relative fluorescence / V

1.0 0.00 0.05 0.10 0.15 0.20 0.25 50 100 150 200 250

b 0.03

0.00 Residuals -0.03 0.00 0.05 0.10 0.15 0.20 0.25 50 100 150 200 250

c 0.03

0.00 Residuals -0.03 0.00 0.05 0.10 0.15 0.20 0.25 50 100 150 200 250

d

3.0

2.5

shutter close Relative fluorescence / V shutter open

2.0 0 200 400 600 800 1000

Time / s

116 Chapter 4 to DMPC/CHAPS micelles. The rate constants resolved here are in good agreement with those reported previously for the refolding of bO in DMPC/CHAPS micelles (Booth et al 1995). In this previous study, the phases n3 and n4 approximated to a single exponential. The fluorescence measurements were carried out here using a higher light intensity, resulting in improved signal-to-noise ratio, and thus an additional exponential was required to fit the data here. An earlier account of this work has also appeared previously (Booth et al 1996). There was a slight downward drift in fluorescence beyond about 250 s after initial mixing of bO/SDS with DMPC/CHAPS micelles (Fig 4.1d). When this experiment was repeated with the light shutter closed initially but opened just before the end of the experiment, this downward drift in fluorescence disappeared (Fig 4.1d). Further, the extent of this downward drift was dependent on the intensity of excitation light. When this intensity was halved, a similar reduction in the downward fluorescence drift was observed. Conversely, increasing the light intensity led to a corresponding increase in the downward drift in fluorescence (data not shown). These experiments thus demonstrate that the slow downward drift in protein fluorescence is very probably due to photobleaching of aromatic residues rather than any further structural or conformational change in the protein or due to diffusion of sample from the stop-flow cuvette. Photobleaching of aromatic residues is a common phenomenon observed in protein fluorescence measurements (Walmsley et al 1994).

Kinetics of Bacterioopsin Refolding 117

4.2.2 Addition of DMPC/CHAPS to SDS

The refolding strategy of the apoprotein bacterioopsin essentially involves mixing SDS with DMPC/CHAPS micelles. Thus, it is important to differentiate between micelle mixing and protein folding events. The fact that the micelles scatter light can be used to provide a sensitive measure of the structural changes occuring within the micelles per se during bacterioopsin refolding. Fig 4.2a shows time-resolved changes in absorption spectra between 350-750 nm upon mixing SDS with DMPC/CHAPS. The spectra were measured using a photodiode array. The change in absorbance results from the extent to which the micelles scatter light at various stages in the mixing. The light scattering is therefore clearly a function of the structural architecture of the micelles. Data were recorded over various time scales between 0.3 and 1000 s. Only the shortest time scale is shown. The time-resolved spectra S[t,l] over the time scale shown were fit to a sum of two exponentials in a global analysis using the function

S[t,l] = å {ai[l] exp(- ni t)} + d[l] [4.2]

where ai[l] and ni are respectively the wavelength-dependent amplitude and the observed rate constant of an exponential phase i, and d[l] is the wavelength-dependent baseline offset. The rate constants

-1 -1 obtained were 710 s (n1) and 90 s (n2). Fig 4.2b shows a plot of residuals obtained for the fit of these spectra to a sum of two exponentials in a global analysis. The scattering of light by the micelles could be described by a Rayleigh function. That is, the time-resolved spectra shown in Fig 4.2a conformed to Rayleigh scattering. Fig 4.2c shows the fit of a Rayleigh function to three such spectra. Rayleigh scattering is a property of molecules smaller in size than the wavelength of light. This would be expected a priori for DMPC/CHAPS micelles which are believed to be about 10 nm in diameter (Brouillette et al 1989). To accurately determine the rate constants associated with changes in light scattering upon mixing SDS with DMPC/CHAPS micelles, a photomultiplier tube was used to collect a higher density of data at single wavelengths over shorter time scales (0.1-0.5 s). A photodiode array only allows collection of a maximum of 100 data points

118 Chapter 4

Fig 4.2 (opposite). Addition of DMPC/CHAPS to SDS. DMPC/CHAPS micelles were mixed, in the stop-flow cuvette, with an equal volume of SDS in phosphate buffer. Mixing was monitored by changes in light scattering (absorbance).

(a) Time-resolved absorption spectra between 350-750 nm were recorded using a photodiode array. A total of 100 spectra were collected over a full time scale of 0.3 s, only first 0.25 s are shown.

(b) A plot of residuals obtained upon globally analysing the data shown in (a) to a sum of two exponentials.

(c) Time-resolved absorption spectra in (a) were fit to a Rayleigh function and fits (solid lines) for three spectra at times of 2, 10 and 100 ms after the initial mixing are shown.

(d) Changes in absorbance at 380 nm monitored using a photomultiplier tube. Data were collected using a split time scale of 0-0.05 s and 0.05-0.25 s. For each interval, 2000 data points were recorded. The solid line shows the fit of the data to a sum of two exponentials.

Kinetics of Bacterioopsin Refolding 119

0.06 0.03 a b 0.04 0.00 0.02 Residuals Absorbance 0.00 -0.03 400 400 500 500 600 600

700 Wavelength / nm 700 0.00 0.05 0.10 0.15 0.20 0.25 0.00 0.05 0.10 0.15 0.20 0.25 Wavelength / nm Time / s Time / s

0.06 c 0.06 d

0.05 0.05

0.04 0.04

n1 Absorbance Absorbance 0.03 2 ms 0.03

n2

0.02 0.02 100 ms

0.01 0.01 10 ms

0.00 0.00 400 500 600 700 0.00 0.05 0.10 0.15 0.20 0.25

Wavelength / nm Time / s

120 Chapter 4 per wavelength over a time scale of about 300 ms. Data on time scales shorter than this cannot be collected using a photodiode array. Fig 4.2d shows changes in absorbance at 380 nm upon mixing SDS with DMPC/CHAPS micelles using a photomultiplier tube as the detector. Data were also collected over slower time scales (1-1000 s). The data shown in Fig 4.2d were fit to a sum of two exponentials

-1 -1 using the eq [4.1]. The rate constants obtained were 850 s (n1) and 70 s (n2). The amplitudes of these rate constants are shown in Table 4.1. The phase n1 is accompanied by an initial decrease in light scattering (absorbance) and the subsequent increase in light scattering represents the phase n2. Note that, these values are in good agreement with those obtained from a global analysis of time-resolved spectra. There was no further change in light scattering over slow time scale upon micelle mixing (data not shown). Further, the changes in light scattering associated with micelle mixing were independent of whether bO was included or not. Table 4.1 compares fluorescence and absorbance measurements for bacterioopsin refolding and micelle mixing. The value for the rate constant of phase n1 from absorbance measurements is only approximate since it coincides with the deadtime of the measurement. The above results have been reported recently (Booth & Farooq 1997). In this study, the values quoted for the absorbance phases n1 and n2 were less accurate. The data reported were analyzed over separate time scales. Thus, the absorbance changes associated with the phases n1 and n2 were analyzed per se to monoexponentials. This procedure carries potential risks and is less reliable. The discrepancy between the values reported for the rate constants of the absorbance phases n1 and n2 previously (Booth & Farooq 1997) and those reported here thus arises from the method of analysis used.

Kinetics of Bacterioopsin Refolding 121

Table 4.1. Comparison of experimentally observed rate constants (n) and their amplitudes (a) for fluorescence and absorbance refolding measurements on bacterioopsin.

Fluorescence Absorbance (Light scattering)

Phase n / s-1 a / V n / s-1 a / AU Assignment (1) Micelle mixing

n1 220 -0.19 850 0.025 ± 30 ± 0.03 ± 200 ± 0.007 (2) Protein misfolding

(3) Formation of the core

n2 3 0.06 70 -0.002 regions of transmembrane ± 2 ± 0.03 ± 30 ± 0.001 helices

n3 0.13 -0.44 * * ± 0.02 ± 0.05 (1) Helix rearrangement and association

(2) Helix formation at the

n4 0.031 -0.22 * * bilayer interfacial regions ± 0.002 ± 0.04

Footnotes to Table 4.1

* Phase NOT detected.

Values for n1 and n2 for absorbance were determined from single wavelength measurements at 380 nm in the absence of bO (inclusion of bO in SDS prior to mixing with DMPC/CHAPS micelles gave similar values).

Errors were calculated from measurements on three different bO preparations. All errors are given to one standard deviation.

122 Chapter 4

4.3 Discussion

This study demonstrates the use of fluorescence and absorption spectroscopy to probe structural changes occuring within the protein and the micelles during bacterioopsin refolding to a state with a full native secondary structure. The starting state of bacterioopsin here is not a completely unfolded structure but a state that apparently retains about 60 % of helices even after denaturation with SDS (London & Khorana 1982, Riley et al 1997). Time-resolved measurements on the refolding of bO from a completely denatured state are not as yet possible. Although organic solvents such as formate can lead to a complete denaturation of bacterioopsin, subsequent refolding can only occur upon dialysing away these solvents (Huang et al 1981). From a practical point of view, it does not really matter whether the kinetics of bacterioopsin refolding are measured from a completely or partially denatured state. The time-resolution of the stop- flow method is no less than a millisecond, whereas secondary structure formation occurs over a time scale of ms to ns (Creighton 1992, Williams et al 1996). Thus secondary structure formation would predominantly occur within the deadtime of the measurement. It is this consideration that makes the issue of whether the starting state of bO is completely or partially denatured irrelevant in terms of the refolding kinetics of bacterioopsin, which are primarily concerned with the rearrangement of pre-formed secondary structure elements. Although methods for initiating protein refolding on a submillisecond scale have been developed (Jones et al 1993, Eaton et al 1997), these are not applicable to complex systems such as membrane proteins. The refolding kinetics of bacterioopsin in DMPC/CHAPS micelles can be kinetically resolved into four distinct phases n1, n2, n3 and n4 with lifetimes ranging from a few milliseconds to tens of seconds (Table 4.1). The phase n1 was previously assigned to the mixing of micelles and the other phases to apoprotein folding, although the phases n3 and n4 could not be distinguished in this report (Booth et al 1995). The light scattering measurements on the mixing of SDS with DMPC/CHAPS however indicate that structural changes within the micelles occur over the time scale over which the fast phases n1 and n2 in the fluorescence refolding experiments are observed. The fact that the rate constants of the phases n1 and n2 measured from light scattering (micelle

Kinetics of Bacterioopsin Refolding 123 mixing) are somewhat higher than those obtained from fluorescence experiments (micelle mixing and bacterioopsin refolding) implies that structural changes within the micelles alone cannot account for these phases. The most straightforward interpretation is that the phases n1 and n2 result from structural changes occuring within both the micelles and the protein. This would of course be a bona fide case in that it is difficult to imagine how the protein could remain silent during the first few milliseconds in the mixing of bO/SDS with DMPC/CHAPS. It is interesting to note that a step in a few millisecond range and a step in the tens of second range have been respectively assigned to the binding and insertion of a 25-residue leader peptide of subunit IV of cytochrome c oxidase with phospholipid membranes (Golding et al 1996). A step in a few millisecond range has also been assigned to the mixing of SDS with octylglucoside micelles in the refolding studies of the plant light-harvesting complex II (Booth & Paulsen 1996). The detailed description of the molecular events occuring within the micelles or the protein during the phases n1 and n2 is not possible at this stage. In an analogy with water-soluble proteins, where helix formation has been observed to occur within a few milliseconds of initiating refolding (Matouchek et al 1990, Radford et al 1992, Elove et al 1992, Matthews 1993, Roder & Colon 1997), it would be reasonable to suggest that these phases may accompany the formation of transbilayer helices. Although helix formation is intrinsically very fast (Creighton 1992, Williams et al 1996), other rate- limiting factors may apparently slow down this process. In the case of membrane proteins, the constraints imposed upon the protein by the lipid bilayer matrix could apparently slow down transbilayer helix formation.

It is also possible that the phases n1 and n2 accompany the formation of non-native structure within bacterioopsin in response to the structural changes occuring within the micelles. Studies indicate that CHAPS can induce non-native b-secondary structure formation in bacterioopsin in the absence of DMPC (Sugiyama & Mukohata 1996). The formation of non-native secondary structure has been observed during the first few milliseconds in the refolding of a-lactoglobulin, which later reassembles into the native protein (Hamada et al 1996). Preliminary experiments indicate that the phases n1 and n2 become more pronounced when bacterioopsin refolding is carried out at pH 8, where the refolding yield of bacteriorhodopsin is nearly 30 % lower in contrast to the

124 Chapter 4 quantitative regeneration at pH 6 (Chapter 3). In fact under alkaline conditions (pH 10- 12), where bacteriorhodopsin cannot be regenerated, the refolding kinetics of bacterioopsin only comprise of the phases n1 and n2 (preliminary observations). It is thus tempting to suggest that these phases may represent protein misfolding. Refolding experiments on the bacterial porin OmpA also detect a misfolded state that is formed within less than a second (Surrey & Jahnig 1995). The possibility that non-native interactions persist in the denatured state of bO in SDS cannot be ruled out. Thus, the fast phases n1 and n2 could also correspond to the disruption of these non-native interactions in the denatured state before real folding can begin.

One can also speculate about the nature of molecular events involved in the slow phases n3 and n4. The two-stage model of membrane protein folding postulates that the formation of transbilayer helices precedes helix association (Popot & Engelman 1990). Thus keeping in line with this paradigm, it is possible that these phases may involve the association and rearrangement of preformed transbilayer helices. Indeed, helix-helix interactions are thought to be the dominant force in the stabilization of transbilayer helices (Lemmon & Engelman 1994). The state with a native-like secondary structure in the micelles (bO¢) would therefore not only contain native secondary structure but tertiary interactions may also pursue. The state bO¢ would thus be analogous to the molten globule state of water-soluble proteins (Ptitsyn et al 1990). The classical test for a molten globule is the binding of the hydrophobic dye ANS to the protein during folding. ANS however is not suitable in the case of membrane protein folding as it can also bind to the membrane lipids. Molten globule-like states are thought to be involved in the insertion of the pore-forming domain of colicin A (van der Goot et al 1991) and the bacterial porin OmpA (Surrey & Jahnig 1995). Cytochrome c is also thought to resemble a molten globule-like state under conditions simulating those near the membrane surface (Bychkova et al 1996). Time-resolved far-UV CD studies (with an experimental deadtime of about 20 s) indicate that nearly half of the total change in the signal upon mixing bO/SDS with DMPC/DHPC micelles is observed with a rate constant similar to that of phase n4 (Riley et al 1997). This implies that some helix formation persists over a slow time scale. Note that, bacterioopsin denatured in SDS contains 60 % of the total helix content of the native protein. The total change in the CD signal observed in this report thus

Kinetics of Bacterioopsin Refolding 125 corresponds to the remaining 40 % helices that are formed upon the transfer of bO/SDS into DMPC/DHPC micelles. It was suggested that this slow helix formation corresponded to the ends of helices at the bilayer interfacial region. The authors argued that the helix formation at the ends of transmembrane segments would be rate-limited by the association of helices. The slow association of transmembrane helices would bring the interhelical loops closer together which in turn would facilitate the ends of helices to take up helical conformation. The overall folding of bacterioopsin is relatively slow compared to the elementary processes of water-soluble proteins (Kim & Baldwin 1990, Matthews 1993). One attractive explanation for this behaviour is the slow cis-trans proline isomerization (Brandts et al 1975, Nall 1994). Note that, there are three proline residues within the transmembrane helices of bacteriorhodopsin. Alternatively, the high viscosity and the anisotropically ordered structure of the micelles may apparently slow the folding of bacterioopsin. Lateral diffusion of the solvent molecules is two to three orders of magnitude slower in a membrane than in water and translocation of residues across the micelles by simple diffusion may be even more restricted. Bacterioopsin refolding kinetics have been shown to be dependent on the bilayer bending rigidity (Booth et al 1997). Protein folding in lipid bilayers is also significantly slower than in mixed phospholipid/detergent micelles observed here (Surrey & Jahnig 1995, Surrey et al 1996, Kleinschmidt & Tamm 1996). The multiexponential kinetics observed for bacterioopsin refolding from a partially denatured state in SDS (bO) to a state with a native-like secondary structure (bO¢) can account for a number of possible kinetic schemes. The simplest scheme that can accommodate these observations would be a sequential model

k1 k2 k3 k4 bO I1 I2 I3 I4

in which the intrinsic rate constants k1-k4 equate to the corresponding experimentally observed rate constants n1-n4. In this model, the apoprotein intermediate I4 structurally

126 Chapter 4

resembles the equilibrium state bO¢. The notation I4 is used here to describe this state so as to emphasize the point that this state is not the final native conformation but a probable intermediate on the pathway to the native bacteriorhodopsin. The state bO is converted to the apoprotein intermediate I4 via putative intermediates I1, I2 and I3. It is now widely recognized that although intermediates may not be necessary in the case of small proteins, for others they may be obligatory (Creighton 1990, Kim & Baldwin 1990, Matthews 1993, Ptitsyn 1995).

It is however conceivable that the intermediates I1 and I2 may not be intrinsic to the formation of the intermediate I4, for the reasons mentioned above. That is, these intermediates lie off-pathway to bacterioopsin refolding. A model that is consistent with these considerations is

I 1 k 1 k3 k4 bO I3 I4

k 2 I2

in which the formation of the intermediate I4 occurs via the intermediate I3. The species I1 and I2 are envisage here to play no active role in bacterioopsin refolding, and probably have a transient existence. Theoretical studies of protein folding, largely based on statistical mechanics and Monte Carlo simulations, indicate that folding of proteins can be achieved without the involvement of intermediates (Karplus & Shakhnovich 1992, Sali et al 1994a, Karplus & Sali 1995, Shakhnovich 1997). Indeed, several water-soluble proteins have been observed to fold via a simple two-state process with no accumulation of intermediates. Examples include chymotrypsin inhibitor-2 (Jackson & Fersht 1991, Otzen et al 1994, Fersht 1995), the cold-shock protein CspB (Schindler et al 1995), monomeric l repressor (Huang & Oas 1995) and acyl-coenzyme A binding protein (Kragelund et al 1995a). Further, folding is envisaged as a distribution of structures from the unfolded to

Kinetics of Bacterioopsin Refolding 127 the final state rather than being confined to a well defined pathway in the new view of protein folding (Baldwin 1995, Dill & Chan 1997). Thus, in line with such ideas, an equally plausible model that can also explain the above observations on bacterioopsin refolding is

I I 3 1 k k 3 1 bO

k 2 k 4 I2 I4

in which all the species I1, I2, I3 and I4 lie on parallel pathways. These species could represent the substates within the state bO¢. In this regard, the state bO¢ may consist of a heterogeneous mixture of several species instead of a single conformation. This latter model is more attractive because the denatured bO is likely to be highly heterogenous. One major source of heterogeneity in denatured bO arises from cis-trans isomerization about proline imide bonds (Brandts et al 1975). Thus, polypeptide chains in the denatured state of bO with all native proline isomers will refold faster than those with some or all non-native isomers. The various kinetic phases may therefore represent parallel rather than sequential steps. The fact that kinetic partitioning during protein folding can yield multiple final states is a well-known phenomenon (Sinclair et al 1994). In summary, the data presented above suggest that the kinetics of bacterioopsin refolding can be differentiated from the kinetics of micelle mixing. The formation of the state of bacterioopsin with a native-like secondary structure in the micelles may be a simple two-state process. To what extent the various kinetic species observed here are involved in the productive folding of bacteriorhodopsin may become clearer when the refolding of bacterioopsin is carried out in the presence of its retinal chromophore (the subject of next three chapters).

Chapter 5

KINETICS OF RETINAL BINDING TO BACTERIOOPSIN

5.1 Overview 130

5.2 Results 131 5.2.1 Addition of Retinal/DMPC/CHAPS to bO/SDS 131 5.2.2 Addition of Retinal/DMPC/CHAPS to bO/DMPC/CHAPS 142

5.2 Discussion 158

130 Chapter 5

5.1 Overview

The functionally active bacteriorhodopsin contains retinal covalently bound to Lys216 via a Schiff base (Bayley et al 1981, Lemke & Oesterhelt 1981). Photoisomerization of retinal triggers a sequence of events, involving conformational changes within the apoprotein bacterioopsin, which ultimately results in the transport of a proton from the cytoplasmic to the extracellular side of the halobactetrial cell (Khorana 1988, Birge 1990). Thus the role that the retinal chromophore plays in the activity of bacteriorhodopsin cannot be overemphasized, but to what extent retinal dictates the folding of bacteriorhodopsin remains largely elusive. It is believed that bacterioopsin can refold to a state with a native-like secondary structure in the absence of retinal (Huang et al 1981, London & Khorana 1982). This salient observation has led to the speculation that retinal may not be involved in initiating the folding of bacterioopsin. Although regeneration of bacteriorhodopsin in mixed phospholipid/detergent micelles is well established (London & Khorana 1982, Brouillette et al 1989), little is known about the kinetics of this process. The reconstitution studies of bacteriorhodopsin from apomembrane and retinal at lower temperatures propose the involvement of spectroscopically distinct intermediates (Schreckenbach et al 1977). More recently, stop-flow fluorescence has been employed to study the refolding kinetics of bacteriorhodopsin (Booth et al 1995). In this report, three sequential stages were identified: The formation of a state of bacterioopsin with a native-like secondary structure, the binding of retinal to this state resulting in the formation of a non-covalent complex between retinal and bacterioopsin, and finally the decay of this complex into bacteriorhodopsin. In this chapter, stop-flow fluorescence is used to study in detail the binding of retinal to bacterioopsin. To what extent the refolding kinetics of bacterioopsin are affected by retinal will be closely examined. Photodiode array spectroscopy is also used to measure changes in the retinal absorption band during the course of the reaction of retinal with bacterioopsin. It is hoped that the presence or absence of a single isobestic point in the time-resolved spectra should reveal whether the formation of bacteriorhodopsin occurs directly or via an intermediate from retinal and bacterioopsin. How the oscillator strength of the retinal absorption band can be used to monitor the formation of the Schiff base between retinal and bacterioopsin is demonstrated.

Kinetics of Retinal Binding to Bacterioopsin 131

5.2 Results

5.2.1 Addition of Retinal/DMPC/CHAPS to bO/SDS

Retinal does not bind to bacterioopsin denatured in SDS (bO/SDS). Only in the presence of DMPC/CHAPS micelles can the binding of retinal be observed. Binding of retinal to bacterioopsin can be monitored by intrinsic protein fluorescence and retinal absorbance. Fig 5.1a shows changes in intrinsic protein fluorescence upon the addition of retinal/DMPC/CHAPS to bO/SDS. The binding of retinal to bacterioopsin leads to fluorescence quenching. This is the result of energy transfer from Trp residues within the retinal binding pocket to retinal (Polland et al 1986a). Data were collected over various time scales from 0.1 to 1000 s. Only two of such time scales are shown. Time-resolved data S[t] shown over the two time scales were simultaneously fit to a sum of six exponentials using the function

S[t] = å {ai exp(- ni t)} + d [5.1]

where ai and ni are respectively the amplitude and the observed rate constant of an exponential phase i,

-1 -1 -1 and d is the baseline offset. The rate constants obtained were 250 s (n1), 5 s (n2), 0.14 s (n3), 0.042

-1 -1 -1 s (n4), 0.009 s (n8) and 0.0015 s (n9). The amplitudes of these exponentials are given in Table 5.1. Figs 5.1b and 5.1c show the residuals obtained upon the fit of a sum of five and six exponentials to the data shown in Fig 5.1a, respectively. Data were also analyzed separately on different time scales. The values obtained for the various rate constants generally agreed with those indicated above.

The two fastest rate constants (n1 and n2) are associated with the initial rise and the subsequent small decrease in fluorescence observed over the first 0.25 s in the mixing. These changes in fluorescence are also observed upon the addition of DMPC/CHAPS to bO/SDS (see Chapter 4). The overall decay in intrinsic protein fluorescence observed over the time period 0.25-1000 s approximates to four exponentials. The rate constants of two of these exponentials are similar in magnitude to the two slower kinetic phases (n3 and n4) observed upon the addition of DMPC/CHAPS

132 Chapter 5

to bO/SDS (see Chapter 4). These are thus denoted n3 and n4 here as well. The other two exponentials

(n8 and n9) result from the binding of retinal to the apoprotein. The reasons for subscripting these 8 and 9, and not 5 and 6, will become clearer later. The choice of a six-exponential function to describe the data in Fig 5.1a is further justified on the grounds that addition of an extra exponential substantially improves the quality of the residuals shown in Figs 5.1b and 5.1c. The above results are in global agreement with those reported previously (Booth et al 1996). The only notable difference being that in the previous study, no distinction was made between the rate constants n8 and n9. Furhter, the above results hold superiority in that simultaneous analysis over different time scales is much more reliable than analysis over different time scales. The latter was the case in the previous study. A photodiode array was also used to time-resolve changes in the retinal absorption band between 305-750 nm upon the addition of retinal/DMPC/CHAPS to bO/SDS (Fig 5.1d). The binding of retinal to bacterioopsin results in the loss of retinal absorption band centred at about 380 nm with a concomitant appearance of an absorption band centred at 560 nm. This band is characteristic of the formation of a native-like chromophore of bacteriorhodopsin (Huang et al 1981, London & Khorana 1982). Time- resolved spectra were recorded over various time scales from 0.3 to 1000 s. Only two such time scales are shown. The time-resolved spectra S[t,l] over the two time scales shown were globally analyzed to a sum of three exponentials using the function

S[t,l] = å {ai[l] exp(- ni t)} + d[l] [5.2]

where ai[l] and ni are respectively the wavelength-dependent amplitude and the observed rate constant of an exponential phase i, and d[l] is the wavelength-dependent baseline offset. The rate constants

-1 -1 -1 obtained were 0.11 s (n3,4), 0.010 s (n8) and 0.0019 s (n9).

The phases n3 and n4 observed in fluorescence experiments cannot be distinguished when binding of retinal to bacterioopsin is monitored by absorbance. These phases are thus collectively denoted by n3,4 in absorbance measurements. The

Kinetics of Retinal Binding to Bacterioopsin 133

changes in absorbance (light scattering) corresponding to the initial micelle mixing (phases n1 and n2, see Chapter 4) over the first 100 ms have been omitted from this analysis. Fig 5.1e shows the residuals obtained upon the fit of a sum of three exponentials to the data shown in Fig 5.1d. Analysis of time- resolved spectra over separate time scales gave similar results to those indicated above. To clearly observe the decay of the retinal absorption band into bacteriorhodopsin, an overlay of time-resolved spectra is shown in Fig 5.1f. As can be seen, the absence of an isobestic point in the region 400-450 nm unequivocally demonstrates that the formation of bacteriorhodopsin occurs via at least one spectroscopically distinct intermediate in this region (vide infra). To determine the number of spectroscopically distinct species involved in the reaction of bacterioopsin and retinal, time-resolved spectra shown in Fig 5.1d were analyzed by SVD algorithm. This analysis provides a model-free indication of the number of spectrally distinct components in a data (Henry & Hofrichter 1992). The results of such an analysis are shown in Fig 5.1g, where the singular value appears to plateau out at the third spectrally distinct component. This implies that there are only three spectrally distinct species in the reaction of retinal with bacterioopsin. That is, there is only one spectroscopically distinct intermediate involved in the formation of bacteriorhodopsin.

Fig 5.1h shows the wavelength-dependent amplitudes of the three exponentials (n3,4, n8 and n9) fitted in a global analysis to the time-resolved spectra in Fig 5.1d. Such amplitudes are essentially a difference between the absorption spectra of the decaying and the non-decaying species (see Chapter 2). Thus, a positive amplitude indicates a decrease in absorbance and a negative amplitude an increase in absorbance. The amplitude of the phase n3,4 consists of a positive band at 380 nm and a negative band centred at about 430 nm, indicating the decay of retinal into the spectroscopically distinct intermediate.

A small positive band in the region 500-600 nm is also observed in the amplitude of the phase n3,4. The physical basis of this band is unclear. The amplitudes of the phases n8 and n9 consist of positive bands centred at around 400 nm and negative bands centred at about 560 nm, indicating the decay of retinal and the intermediate into the native chromophore. Note that, the absorbance maxima observed in the wavelength-dependent amplitudes only equate to the absorbance maxima of the real species if their absorption

134 Chapter 5

Fig 5.1 (pp 135-138). Addition of retinal/DMPC/CHAPS to bO/SDS. Retinal in DMPC/CHAPS micelles was mixed, in the stop-flow cuvette, with an equal volume of bO/SDS at a final protein to retinal molar ratio of 1:1. Refolding was monitored by intrinsic protein fluorescence and retinal absorbance.

(a) Refolding monitored by intrinsic protein fluorescence. The time scale is split between two periods of 0-0.25 s and 0.25-1000 s. Data shown over 0-0.25 s are an average of eight transients collected over a full scale of 0.5 s. Data shown over 0.25-1000 s are an average of two transients collected over a full scale of 1000 s. For each interval, 4000 data points were recorded. Data shown over the two time scales were simultaneously fit to a sum of six exponentials. The fit is indicated by a solid line. (b) Residuals obtained upon the fit of a sum of five exponentials to the data in (a). (c) Residuals obtained upon the fit of a sum of six exponentials to the data shown in (a).

(d) Refolding monitored by retinal absorbance using a photodiode array. Time-resolved spectra between 305-750 nm shown over two periods of 0-60 s and 60-1000 s were collected over full scales of 63 s and 1000 s respectively. For each period, 400 spectra were recorded. The time-resolved spectra shown over the two time scales were simultaneously fit to a sum of three exponentials in a global analysis using Look.

(e) Residuals obtained upon the fit of a sum of three exponentials to the data in (d).

(f) An overlay of the time-resolved absorption spectra shown in (d). The spectra in the order of decreasing absorbance at 380 nm (increasing at 560 nm) correspond to times of 0.1, 1, 2, 5, 10, 20, 50, 100, 200, 500 and 1000 s in the refolding of bacteriorhodopsin.

(g) SVD analysis of the time-resolved absorption spectra shown in (d).

(h) Wavelength-dependent amplitudes of the three exponentials n3,4, n8 and n9.

(i) Kinetic absorption spectra of retinal, the intermediate and bacteriorhodopsin.

(j) Gaussian decomposition of the kinetic absorption spectrum of retinal. The solid line indicates the fit of eq [5.3] (i = retinal) to the kinetic spectrum of retinal (open circles). The decomposed components are denoted by dotted lines.

(k) Gaussian decomposition of the kinetic absorption spectrum of the intermediate. The solid line indicates the fit of eq [5.3] (i = retinal, intermediate) to kinetic spectrum of the intermediate (open circles). The decomposed components are denoted by dotted lines.

(l) Absorbance changes at 380, 430 and 560 nm obtained from the time-resolved spectra shown in (d). The solid lines show the fit of a three-exponential to the data shown in (d). (m) Residuals obtained at 430 nm upon the fit of a two-exponential to the data in (d).

(n) Residuals obtained at 430 nm upon the fit of a three-exponential to the data in (d).

Kinetics of Retinal Binding to Bacterioopsin 135

3.5 a bO/DMPC/CHAPS

3.0

2.5

2.0

1.5 bO/SDS Relative fluorescence / V 1.0

0.5

0.0 0.00 0.05 0.10 0.15 0.20 0.25 200 400 600 800 1000

b 0.02

0.00

Residuals -0.02

0.00 0.05 0.10 0.15 0.20 0.25 200 400 600 800 1000

c 0.02

0.00

Residuals -0.02

0.00 0.05 0.10 0.15 0.20 0.25 200 400 600 800 1000

Time / s

136 Chapter 5

d 0.12 0.08

0.04

0.00 Absorbance 400 500 600 700 0 20 40 60 500 1000 Wavelength / nm Time / s e 0.02

0.00 Residuals -0.02 400 500 600 700 0 20 40 60 500 1000 Wavelength / nm Time / s

0.12 f

0.08

Absorbance 0.04

0.00 300 350 400 450 500 550 600 650 700 750 Wavelength / nm

g 15

10

5 Singular value

0

1 2 3 4 5 6 7 8

Spectral component

Kinetics of Retinal Binding to Bacterioopsin 137

Wavenumber / (106 nm)-1 3000 2500 2000 1500 h i

bacteriorhodopsin 0.02 0.12 retinal n8 0.10 0.00

0.08 n3,4 -0.02 Absorbance 0.06 Absorbance difference -0.04 intermediate 0.04

-0.06 n 9 0.02

-0.08 0.00 300 400 500 600 700 300 400 500 600 700

Wavelength / nm Wavelength / nm

0.12 j 0.12 k

0.10 0.10

0.08 0.08

0.06 0.06

0.04 0.04 Absorbance Absorbance

0.02 0.02

0.00 0.00

300 400 500 600 700 300 400 500 600 700 Wavelength / nm Wavelength / nm

138 Chapter 5

0.12 l

0.10

380 nm 0.08

0.06 430 nm Absorbance

0.04

560 nm 0.02

0.00 0 10 20 30 40 50 60 200 400 600 800 1000

m 0.002

0.000 Residuals -0.002 0 10 20 30 40 50 60 200 400 600 800 1000

n 0.002

0.000 Residuals -0.002 0 10 20 30 40 50 60 200 400 600 800 1000

Time / s

Kinetics of Retinal Binding to Bacterioopsin 139 spectra do not overlap. However, the spectroscopically distinct intermediate observed in the folding reaction of bacteriorhodopsin appears to absorb in the same region as retinal (400-450 nm). Thus one would not expect the absorbance maximum of the negative band of the amplitude of the phase n3,4 to correspond to the actual absorbance maximum of the spectroscopically distinct intermediate. The transient nature of this intermediate also means that the absorbance maxima of the positive bands of the phases n8 and n9 do not necessarily represent the decay of retinal into the native chromophore. The absorbance maxima of the positive bands of the phases n8 and n9 would however be expected to lie anywhere between the absorbance maximum of 380 nm of free retinal and that of the intermediate. The absorbance maximum of the intermediate would be expected to lie in the region 400-450 nm. That is, the region where no single isobestic point is observed. A more quantitative insight into the binding of retinal to the apoprotein bacterioopsin would not be fraught with danger. The area under a wavelength-dependent amplitude reflects the difference in the area under the absorption spectra of the decaying and the non-decaying species (see Chapter 2). This opens up the possibility of measuring the change in the oscillator strength of the retinal absorption band of a particular phase. Of importance is the ratio of the area under the negative band to the area under the positive band. As can be seen from Fig 5.1h, the area under the positive band is about the same as that under the negative band of the amplitude of the phase n3,4. This indicates that the oscillator strength of the retinal absorption band does not change, within the limits of the experimental technique, upon the decay of retinal into the intermediate. In contrast, the areas under the negative bands of amplitudes of the phases n8 and n9 are much greater than the areas under their positive bands. This implies that the oscillator strength of the retinal absorption band increases upon the decay of retinal and the intermediate into the native chromophore. Table 5.2 lists the change in the oscillator strength of the retinal absorption band for the various kinetic phases. Although a spectroscopically distinct intermediate has been unequivocally identified in the folding reaction of bacteriorhodopsin, its absorption spectrum remains to be delineated. In order to do this, one requires kinetic absorption spectra of retinal and bacteriorhodopsin. The kinetic spectrum of bacteriorhodopsin (the non-decaying species) equates to the wavelength-dependent baseline offset d[l] in eq [5.2] (see

140 Chapter 5

Chapter 2). The kinetic spectrum of retinal can be obtained by adding the wavelength- dependent amplitudes of the phases n3,4, n8 and n9 to the kinetic spectrum of bacteriorhodopsin. The kinetic spectrum of the intermediate can then be calculated by subtracting the amplitude of the phase n3,4 from the kinetic spectrum of retinal. Fig 5.1i shows the kinetic absorption spectra calculated for retinal, the intermediate and bacteriorhodopsin. The fact that the spectroscopically distinct intermediate is transitory means that the kinetic absorption spectrum of this intermediate shown in Fig 5.1i includes the contribution of retinal. The absorbance maximum of such a kinetic spectrum therefore does not equate to the actual absorbance maximum of the intermediate. To calculate the absorbance maximum of the intermediate, the contribution of retinal from the kinetic spectrum of the intermediate needs to be removed. In order to do this, the kinetic spectrum of the intermediate must be decomposed into Gaussian components corresponding to free retinal and the intermediate. The decomposition was carried out as follows. First, the kinetic spectrum of retinal e[l] was fitted to the function

1/2 2 4 e[l] = å {ai / si(2p) } exp{-1/2 [(l - li) / si] } + (b / l ) + d [5.3]

where ai, si and li are respectively the area, 0.425 times the full width half maximum and absorbance maximum of a Gaussian peak i (i = retinal). l is the wavelength, b is a constant related to the Rayleigh factor and d is the baseline offset. Fig 5.1j shows the fit and the decomposed Gaussian and Rayleigh components. The Gaussian component is the absolute spectrum of free retinal.

The values obtained for sretinal and lretinal, that is 0.425 times the full width half maximum and the absorbance maximum of retinal, were subsequently locked during the fitting of the kinetic spectrum of the intermediate to eq [5.3] (i = retinal, intermediate). Fig 5.1k shows the fit and the decomposed Gaussian and Rayleigh components. The Gaussian components correspond to the absolute spectra of free retinal and the intermediate. The absorbance maximum of the intermediate was found to be 430 nm. The inclusion of the l-4 component was necessary to obtain a better fit of the data to the above functions. That is, this has the effect of removing the contribution of light

Kinetics of Retinal Binding to Bacterioopsin 141 scattering from the absorbance of retinal. Note also that data were fitted over the range 350-750 nm in order to improve the quality of the fit. Fig 5.1l shows the absorbance changes at single wavelengths of 380, 430 and 560 nm. These were derived from the time-resolved spectra shown in Fig 5.1d. As can be seen, the decay of the retinal absorption band at 380 nm is initially matched by a rise in absorbance at 430 nm, corresponding to the formation of the intermediate. During this period, the absorbance at 560 nm effectively remains constant. This so-called lag period is indicative of the decay of a reactant into a product via an intermediate. The final rise in the absorbance at 560 nm is coincident with the decay in the absorbance at 430 nm, implying that the intermediate decays directly into bacteriorhodopsin. Figs 5.1m and 5.1n show the residuals obtained at 430 nm upon fit of a sum of two and three exponentials in a global analysis to the time-resolved spectra shown in Fig 5.1d.

142 Chapter 5

5.2.2 Addition of Retinal/DMPC/CHAPS to bO/DMPC/CHAPS

An alternative way of studying retinal binding to bacterioopsin is to add retinal after bO has been allowed to equilibrate with the micelles (bO/DMPC/CHAPS). This procedure should remove the slow rate-limiting step(s) in bacterioopsin refolding, and thus would allow identification of fast phases associated with retinal binding. Fig 5.2a shows changes in intrinsic protein fluorescence upon the addition of retinal/DMPC/CHAPS to bO/DMPC/CHAPS. bO/DMPC/CHAPS was prepared by mixing manually bO/SDS with an equal volume of DMPC/CHAPS and the mixture allowed to equilibrate for 30 min. Data were collected over various time scales from 0.1 to 1000 s. Only two such time scales are shown. Data shown over the two time scales were simultaneously fit to a sum of five exponentials using the eq

-1 -1 -1 -1 - [5.1]. The rate constants obtained were 5.4 s (n5), 1.5 s (n6), 0.21 s (n7), 0.011 s (n8) and 0.0027 s

1 (n9). The amplitudes of these exponentials are given in Table 5.1. Table 5.1 also compares the experimentally observed rate constants of these exponentials with those obtained upon the addition of retinal/DMPC/CHAPS to bO/SDS and for bacterioopsin refolding in the absence of retinal (addition of DMPC/CHAPS to bO/SDS, see Chapter 4). Figs 5.2b and 5.2c show the residuals obtained upon fit of the data shown in Fig 5.1a to a sum of four and five exponentials, respectively.

The two slowest rate constants n8 and n9 were also obtained upon refolding bacterioopsin in the presence of retinal (addition of retinal/DMPC/CHAPS to bO/SDS). Retinal here is however added after bacterioopsin refolding has reached completion. The rate constants n5, n6 and n7 thus result from the binding of retinal to bacterioopsin with fast lifetimes. These rate constants could not be resolved upon the addition of retinal/DMPC/CHAPS to bO/SDS as a result of slow rate-limiting step(s) in bacterioopsin refolding. Data were also analyzed over separate time scales. Data over 1 s full time scale was best described by a three-exponential. Data over 1000 s full time scale was also best described by a three-exponential. The fastest phase derived from this scale was most likely the same as the slowest phase observed on the 1 s full scale. These observations, together with an improvement in the quality of the residuals obtained with a five-exponential over a four-exponential (Figs 5.2b and 5.2c), support the choice of the former function to describe the data in Fig 5.2d. The above results

Kinetics of Retinal Binding to Bacterioopsin 143 have been reported previously (Booth et al 1996). In this previous study, the data were analyzed over separate time scales. Only four exponentials were considered sufficient to describe the data. Thus, simultaneous analysis of data over different time scales is a much more reliable method. Fig 5.2d shows time-resolved absorption spectra between 305-750 nm obtained upon the addition of retinal/DMPC/CHAPS to bO/DMPC/CHAPS. As with the addition of retinal/DMPC/CHAPS to bO/SDS (Fig 5.1d), the appearance of a native-like absorption band centred at 560 nm is observed at the expense of the retinal absorption band at 380 nm. Time-resolved spectra were recorded over various time scales from 0.3 to 1000 s. Only two such time scales are shown. The time- resolved spectra S[t,l] shown over the two time scales were simultaneously fit to a sum of four

-1 -1 exponentials in a global analysis using eq [5.2]. The rate constants obtained were 2.2 s (n5,6), 0.13 s

-1 -1 (n7), 0.013 s (n8) and 0.0032 s (n9). The phases n5 and n6 observed in fluorescence experiments cannot be distinguished when retinal binding is monitored by absorbance. These phases are collectively denoted by n5,6 in absorbance measurements. The mixing of SDS/DMPC/CHAPS with SDS/DMPC/CHAPS in the presence of bO and/or retinal produced no observable changes in absorbance (light scattering), implying that there were no structural changes associated with micelle mixing in this type of experiment (data not shown). Thus, all the kinetic phases observed upon mixing retinal/DMPC/CHAPS with bO/DMPC/CHAPS most likely represent protein folding events. Fig 5.2e shows the residuals obtained upon the fit of a sum of four exponentials to the time-resolved spectra shown in Fig 5.2d. A comparison of the experimentally observed rate constants obtained from fluorescence and absorbance measurements for the reaction of retinal with bacterioopsin is presented in Table 5.3. As with the addition of retinal/DMPC/CHAPS to bO/SDS (Fig 5.1f), the lack of a single isobestic point in the time-resolved spectra is indicative of the formation of bacteriorhodopsin via at least one spectroscopically distinct intermediate (Fig 5.2f). The SVD analysis of the time-resolved spectra presented in Fig 5.2g is likewise suggestive of the presence of only three spectroscopically distinct species in the reaction involving retinal and bacterioopsin.

Fig 5.2h shows the wavelength-dependent amplitudes of the four exponentials n5,6, n7, n8 and n9 fitted in a global analysis to the time-resolved spectra in Fig 5.2d. The

144 Chapter 5

Fig 5.2 (pp 145-148). Addition of retinal/DMPC/CHAPS to bO/DMPC/CHAPS. Retinal in DMPC/CHAPS micelles was mixed, in the stop-flow cuvette, with an equal volume of bO/DMPC/CHAPS at a final protein to retinal molar ratio of 1:1. bO/DMPC/CHAPS was prepared by mixing manually bO/SDS with an equal volume of DMPC/CHAPS and the mixture allowed to equilibrate for 30 min. Refolding was monitored by intrinsic protein fluorescence and retinal absorbance.

(a) Refolding monitored by intrinsic protein fluorescence. The time scale is split between two periods of 0-10 s and 10-1000 s. Data were collected over two periods of 0-10 s and 10-1000 s and, represent an average of two transients. For each period, 2000 data points were recorded. Data shown over the two time scales were simultaneously fit to a sum of five exponentials. The fit is indicated by a solid line. (b) Residuals obtained upon the fit of a sum of four exponentials to the data in (a). (c) Residuals obtained upon the fit of a sum of five exponentials to the data in (a). (d) Refolding monitored by retinal absorbance using a photodiode array. Time-resolved spectra between 305-750 nm shown over periods of 0-15 s and 15-1000 s were collected over full scales of 16 s (200 spectra) and 1000 s (400 spectra) respectively. The time-resolved spectra shown over the two time scales were simultaneously fit to a sum of four exponentials in a global analysis using the kinetic software Look.

(e) Residuals obtained upon the fit of a four-exponential to the data shown in (d).

(f) An overlay of the time-resolved spectra shown in (d). The spectra in the order of decreasing absorbance at 380 nm (increasing at 560 nm) correspond to times of 0.1, 0.2, 0.5, 1, 2, 5, 10, 20, 50, 100, 200, 500 and 1000 s in the refolding of bacteriorhodopsin.

(g) SVD analysis of the time-resolved absorption spectra shown in (d).

(h) Wavelength-dependent amplitudes of the four kinetic phases n5,6, n7, n8 and n9.

(i) Kinetic absorption spectra of retinal, the intermediate and bacteriorhodopsin.

(j) Gaussian decomposition of the kinetic absorption spectrum of retinal. The solid line indicates the fit of eq [5.3] (i = retinal) to the kinetic spectrum of retinal (open circles). The decomposed components are denoted by dotted lines.

(k) Gaussian decomposition of the kinetic absorption spectrum of the intermediate. The solid line indicates the fit of eq [5.3] (i = retinal, intermediate) to kinetic spectrum of the intermediate (open circles). The decomposed components are denoted by dotted lines.

(l) Absorbance changes at 380, 430 and 560 nm obtained from the time-resolved spectra shown in (d). The solid lines show the fit of a four-exponential to the data shown in (d).

(m) Residuals obtained at 430 nm upon the fit of a three-exponential to the data in (d).

(n) Residuals obtained at 430 nm upon the fit of a four-exponential to the data in (d).

Kinetics of Retinal Binding to Bacterioopsin 145

3.5 a bO/DMPC/CHAPS

3.0

2.5

2.0

1.5

Relative fluorescence / V 1.0

0.5

0.0 0 2 4 6 8 10 200 400 600 800 1000

b 0.02

0.00

Residuals -0.02

0 2 4 6 8 10 200 400 600 800 1000

c 0.02

0.00

Residuals -0.02

0 2 4 6 8 10 200 400 600 800 1000

Time / s

146 Chapter 5

d 0.12 0.08

0.04

0.00 Absorbance 400 500 600 700 0 5 10 15 500 1000 Wavelength / nm Time / s e 0.02

0.00 Residuals -0.02 400 500 600 700 0 5 10 15 500 1000 Wavelength / nm Time / s

0.12 f

0.08

0.04 Absorbance

0.00 300 350 400 450 500 550 600 650 700 750 Wavelength / nm

g 15

10

5 Singular value

0

1 2 3 4 5 6 7 8

Spectral component

Kinetics of Retinal Binding to Bacterioopsin 147

Wavenumber / (106 nm)-1 3000 2500 2000 1500 h i 0.12 0.02 retinal n8

bacteriorhodopsin n 7 0.10 0.00

0.08

-0.02 n5,6

0.06 absorbance -0.04 intermediate Absorbance difference 0.04

n -0.06 9 0.02

-0.08 0.00 300 400 500 600 700 300 400 500 600 700 Wavelength / nm Wavelength / nm

j k

0.12 0.10

0.10 0.08

0.08 0.06

0.06

0.04 Absorbance 0.04 absorbance

0.02 0.02

0.00 0.00

300 400 500 600 700 300 400 500 600 700 Wavelength / nm Wavelength / nm

148 Chapter 5

0.12 l

0.10 380 nm

0.08

430 nm 0.06 Absorbance

0.04

0.02 560 nm

0.00 0 5 10 15 200 400 600 800 1000

m 0.003

0.000 Residuals -0.003 0 5 10 15 200 400 600 800 1000

n 0.003

0.000 Residuals -0.003 0 5 10 15 200 400 600 800 1000

Time / s

Kinetics of Retinal Binding to Bacterioopsin 149

amplitude of the phase n5,6 consists of a positive band centred at 380 nm and a negative band centred at

430 nm, indicating the decay of retinal into the intermediate. The amplitude of the phase n5,6 thus resembles the amplitude of the phase n3,4, obtained upon the addition of retinal/DMPC/CHAPS to bO/SDS (Fig 5.1h), except that it lacks a small positive band observed in the region 500-600 nm for the amplitude of the phase n3,4. Thus, as suspected previously, this small positive band is unlikely to be associated with any protein folding event. One likely cause of this band may be the misfit of data over the wavelength range 500-600 nm during the refolding of bacterioopsin in the presence of retinal (addition of retinal/DMPC/CHAPS to bO/SDS). Under these conditions, the slow refolding of bacterioopsin results in the apparently slow rate of formation of the 430-nm intermediate. As a result, during this period the absorbance in the region 500-600 nm remains practically constant (the lag period). However, fit of data to a sum of three exponentials in a global analysis in this region over this period shows a small decrease resulting in poor residuals.

As with the amplitude of the phase n3,4, there is practically no change in the oscillator strength of the retinal absorption band during the phase n5,6, as judged by comparing the areas under the negative and positive bands. Note that, the formation of the spectroscopically distinct intermediate in the folding reaction of bacteriorhodopsin is accompanied by the phase n5,6 and not the phase n3,4. The phase n3,4 is associated with bacterioopsin refolding prior to retinal binding. The fact that the formation of the intermediate is observed with a rate constant of the magnitude of the phase n3,4 upon the addition of retinal/DMPC/CHAPS to bO/SDS arises because bacterioopsin refolding is rate-limiting for the subsequent binding of retinal. Thus when this rate-limiting step is removed, retinal is observed to bind to the apoprotein with a rate constant of the magnitude of the phase n5,6.

The amplitudes of the phases n8 and n9 consist of positive bands centred around 400 nm and negative bands centred around 560 nm, and thus exhibit similarities to the amplitudes obtained for these phases upon the addition of retinal/DMPC/CHAPS to bO/SDS (Fig 5.1h). The amplitude of the phase n7 consists of a positive band centred around 430 nm and a small negative band around 520 nm. This phase thus reflects the decay of the 430-nm intermediate into a 520-nm species. This observation indicates the

150 Chapter 5 presence of more than one absorbing species in the native chromophore of bacteriorhodopsin. Alternatively, such blue shift observed in the native chromophore may be a consequence of the fact that the phase n7 could not be properly resolved from the kinetic phases n8 and n9. At first sight, one may conclude that the retinal absorption band experiences a substantial reduction in its oscillator strength during the phase n7 but this may not necessarily be the case for the reasons given above. Table 5.2 compares the change in the oscillator strength of the retinal absorption band during the course of the reaction of retinal with bacterioopsin for the addition of retinal/DMPC/CHAPS to bO/SDS and bO/DMPC/CHAPS. The kinetic absorption spectrum obtained for the intermediate is shown in Fig 5.2i, along with those of retinal and bacteriorhodopsin. These kinetic spectra were calculated from the wavelength- dependent amplitudes of the phases n5,6, n7, n8 and n9, and the baseline offset in eq [5.2] (vide supra). In order to obtain an absolute spectrum of the intermediate, the kinetic spectra of retinal and the intermediate were decomposed as described in the previous section using eq [5.3]. Fig 5.2j shows the decomposition of the kinetic spectrum of retinal into the absolute spectrum of free retinal. The decomposition of the kinetic spectrum of the intermediate into the absolute spectra of free retinal and the intermediate is indicated in Fig 5.2k. As before, the absorbance maximum of the intermediate was found to be 430 nm. Fig 5.3l shows the absorbance changes at single wavelengths of 380, 430 and 560 nm. These were derived from the time-resolved spectra shown in Fig 5.2d. These data are consistent with the idea that the 430-nm intermediate is involved in the formation of bacteriorhodopsin. The only difference between these data and those presented in Fig 5.1l is that the apparent rate of the formation of the 430- nm intermediate here is an order of magnitude greater. This is understandable as retinal is added after bacterioopsin refolding has reached completion. Otherwise, the rate of the formation of the native chromophore from the 430-nm intermediate is the same in both sets of data shown in Figs 5.1l and 5.2l. The residuals obtained at 430 nm upon fit of a sum of three and four exponentials to the time- resolved spectra shown in Fig 5.2d in a global analysis are shown in Figs 5.2m and 5.2n, respectively. Although at first sight, no significant improvement in the quality of residuals is apparent for a four- exponential function over

Kinetics of Retinal Binding to Bacterioopsin 151 the three-exponential function but a closer examination shows that this is not true over the first 4 s in the folding reaction of bacteriorhodopsin. That is, addition of an extra exponential is required to obtain a good fit of the data over the first 4 s. Note that, the corresponding fluorescence measurements necessitate a sum of five exponentials (Fig 5.2a). It thus should not be surprising that a four-exponential is required to optimally describe the time-resolved spectra shown in Fig 5.2d.

152 Chapter 5

Table 5.1. Comparison of experimentally observed rate constants (n) and their amplitudes (a) obtained from fluorescence data for the refolding of bacterioopsin in the absence of retinal (addition of DMPC/CHAPS to bO/SDS, Chapter 4) and for the reaction of retinal with bacterioopsin.

Addition of Addition of Addition of DMPC/CHAPS to retinal/DMPC/CHAPS retinal/DMPC/CHAPS bO/SDS to bO/SDS to bO/DMPC/CHAPS Phase n / s-1 a / V n / s-1 a / V n / s-1 a / V

n1 220 -0.19 250 -0.18 * * ± 30 ± 0.03 ± 20 ± 0.02

n2 3 0.06 5 -0.07 * * ± 2 ± 0.03 ± 2 ± 0.03

n3 0.13 -0.44 0.14 0.81 * * ± 0.02 ± 0.05 ± 0.03 ± 0.17

n4 0.031 -0.22 0.042 0.75 * * ± 0.002 ± 0.04 ± 0.005 ± 0.11

n5 * * * * 5.4 0.65 ± 2.5 ± 0.12

n6 * * * * 1.5 0.91 ± 0.7 ± 0.21

n7 * * * * 0.21 0.63 ± 0.09 ± 0.25

n8 * * 0.009 0.55 0.011 0.35 ± 0.003 ± 0.23 ± 0.002 ± 0.17

n9 * * 0.0015 0.32 0.0027 0.42 ± 0.0008 ± 0.15 ± 0.0011 ± 0.25

Kinetics of Retinal Binding to Bacterioopsin 153

Footnotes to Table 5.1

* Phase Not detected.

Errors were calculated from measurements on three different bO preparations. All errors are given to one standard deviation.

Only phases n1-n4 are detected during apoprotein refolding (see Chapter 4). The remaining five phases n5-n9 are associated with retinal binding to bacterioopsin.

The phases n5, n6 and n7 cannot be detected when bacterioopsin refolding is initiated in the presence of retinal (addition of retinal/DMPC/CHAPS to bO/SDS). This is because the rate constants associated with these phases are an order of magnitude smaller than the rate of bacterioopsin of refolding. The binding of retinal thus reflects the rate of bacterioopsin refolding in this type of experiment.

The phases n1-n4 cannot be detected when retinal is added to bO pre-equilibrated with the micelles (addition of retinal/DMPC/CHAPS to bO/DMPC/CHAPS). This is because bacterioopsin refolding is allowed to reach completion before the addition of retinal. The phases n1-n4 that are associated with bacterioopsin refolding would be thus not present in this type of experiment.

154 Chapter 5

Table 5.2. Relative oscillator strength of the retinal absorption band during the course of the reaction of retinal with bacterioopsin.

Relative oscillator strength Addition of Addition of Nature of Phase retinal/DMPC/CHAPS retinal/DMPC/CHAPS retinal binding to bO/SDS to bO/DMPC/CHAPS

n3,4 1.07 * non-covalent ± 0.17

n5,6 * 1.15 non-covalent ± 0.11

n7 * 0.27 covalent ± 0.09

n8 2.23 2.03 covalent ± 0.19 ± 0.21

n9 7.55 6.41 covalent ± 0.71 ± 0.45

Kinetics of Retinal Binding to Bacterioopsin 155

Footnotes to Table 5.2

* Phase NOT detected.

Errors were calculated from measurements on three different bO preparations. All errors are given to one standard deviation.

The relative oscillator strength is the ratio of the area under the negative band to the area under the positive band of the wavelength-dependent amplitude of a kinetic phase (Figs 5.1k and 5.2k). The area under each band was calculated in dimensions of nm-1 (proportional to the energy of an electronic transition), that is from the plots of amplitudes on an abscissa linear in wavenumber.

The phases n3,4 and n5,6 can each be distinguished into two single kinetic phases in fluorescence experiments. Both of these phases are associated with the formation of the 430-nm intermediate. The rate constant associated with the phase n5,6 reflects the apparent rate constant obtained from absorbance experiments for the formation of the 430-nm intermediate. The rate constant associated with the phase n3,4 reflects the apparent rate constant obtained from absorbance experiments for bacterioopsin refolding. This is because bacterioopsin refolding is rate-limiting for the subsequent binding of retinal, with the phase n3,4 being an order of magnitude slower than the phase n5,6.

156 Chapter 5

Table 5.3. Comparison of experimentally observed rate constants (n) obtained from fluorescence and absorbance data for the reaction of retinal with bacterioopsin.

Experimentally observed rate constant / s-1 Addition of Addition of retinal/DMPC/CHAPS retinal/DMPC/CHAPS to bO/SDS to bO/DMPC/CHAPS Phase Fluorescence Absorbance Fluorescence Absorbance Assignment (1) Micelle mixing

n1 250 850 * * ± 20 ± 200 (2) Protein misfolding

(3) Formation of

n2 5 70 * * transmembrane ± 2 ± 30 helices within the core regions (1) Helix rearr-

n3 0.14 * * angement and ± 0.03 0.11 association ± 0.04 (2) Helix form-

n4 0.042 * * ation at the bilayer ± 0.005 interfacial regions

n5 * * 5.4 2.2 Non-covalent ± 2.5 ± 1.2 binding of retinal

n6 * * 1.5 ± 0.7 (1) Ring-chain

n7 0.21 0.13 planarization of * * ± 0.09 ± 0.06 retinal

(2) Covalent bind-

n8 0.009 0.010 0.011 0.013 ing of retinal ± 0.003 ± 0.002 ± 0.002 ± 0.001

(3) Formation of

n9 0.0015 0.0019 0.0027 0.0032 bacteriorhodopsin ± 0.0008 ± 0.0017 ± 0.0012 ± 0.0021

Kinetics of Retinal Binding to Bacterioopsin 157

Footnotes to Table 5.3

* Phase Not detected.

Errors were calculated from measurements on three different bO preparations. All errors are given to one standard deviation.

The phases n1 and n2 indicated for absorbance were obtained upon mixing SDS with DMPC/CHAPS (Chapter 4). No change in light scattering was observed upon mixing SDS/DMPC/CHAPS with

SDS/DMPC/CHAPS. Inclusion of bO and/or retinal has no effect on the phases n1 and n2.

The phases n3 and n4, and n5 and n6 cannot be distinguished in absorbance experiments (denoted by n3,4 and n5,6, respectively).

158 Chapter 5

5.3 Discussion

The results presented here provide an intriguing insight into the mechanism of retinal binding to bacterioopsin. Binding of retinal can be divided into fast and slow phases with lifetimes ranging from hundreds of milliseconds to several minutes. Retinal appears to have a little effect on the kinetics of bacterioopsin refolding. Thus the phases n1, n2, n3 and n4, which have been previously assigned to bacterioopsin refolding (addition of DMPC/CHAPS to bO/SDS, see Chapter 4), can also be resolved when bacterioopsin refolding is initiated in the presence of retinal (addition of retinal/DMPC/CHAPS to bO/SDS, Table 5.1). The phases n1 and n2 are accompanied here by similar changes in fluorescence to those observed upon the refolding of bacterioopsin in the absence of retinal (see Chapter 4), and their rate constants are remarkably similar in both experiments (Table 5.1).

The phases n3 and n4, although associated with bacterioopsin refolding, here reflect the binding of retinal to the apoprotein. This is because bacterioopsin refolding is apparently rate-limiting for the subsequent binding of retinal. When bacterioopsin is pre-equilibrated with the micelles (bO/DMPC/CHAPS), so as to remove this rate-limiting barrier, retinal can be observed to bind with rate constants (phases n5 and n6) that are an order of magnitude greater than the rate constants associated with the phases n3 and n4. Thus although the phases n3 and n4 appear to reflect retinal binding here but they in actual fact are associated with apoprotein refolding. The quenching of protein fluorescence upon the binding of retinal (Polland et al 1986a), however, means that the phases n3 and n4 are accompanied here by a decrease in fluorescence instead of a rise observed upon the refolding of bacterioopsin in the absence of retinal (see Chapter 4). Thus in line with previous suggestions (London & Khorana 1982, Booth et al 1995), retinal does not appear to initiate bacterioopsin refolding. Retinal probably binds to a state of bacterioopsin with a native-like secondary structure (bO¢). This idea is consistent with the two-stage model of membrane protein folding (Popot & Engelman 1990), in which the transmembrane helices are considered as autonomous folding domains. The state bO¢ may be envisaged as an obligatory apoprotein intermediate on the pathway leading to bacteriorhodopsin formation. Because of the multiphasic kinetics observed in bacterioopsin refolding, the possibility that the state bO¢ represents an ensemble of structures rather than a well-defined structure cannot be excluded. The

Kinetics of Retinal Binding to Bacterioopsin 159 foregoing argument invokes the possibility of retinal binding to more than one species within the state bO¢.

Of the several kinetic phases associated with retinal binding, the fast phases n5, n6 and n7 cannot be resolved when the binding of retinal is studied in parallel with the refolding of bacterioopsin (addition of retinal/DMPC/CHAPS to bO/SDS, Table 5.1). The fact that the bacterioopsin refolding is rate- limiting for the subsequent binding of retinal means that these fast phases would not expected to be apparent in this experiment. Only by removing this rate-limiting barrier, do the fast phases associated with retinal binding become apparent (addition of retinal/DMPC/CHAPS to bO/DMPC/CHAPS, Table 5.1). The values for the rate constants obtained from fluorescence experiments here are broadly in agreement with those reported earlier (Booth et al 1995). Because of the higher signal-to-noise ratio obtained here, as a result of higher light intensity level used, a number of additional kinetic phases have been resolved upon the binding of retinal to bacterioopsin. The rate constants for the binding of retinal to bacterioopsin obtained from absorbance measurements, where changes in retinal absorption band are monitored using a photodiode array, are in an excellent agreement with those obtained from fluorescence experiments (Table 5.3). Because of the lower sensitivity of absorbance relative to fluorescence, certain phases observed in fluorescence experiments cannot be distinguished when the binding of retinal is monitored by absorbance. Thus phases n3 and n4, and phases n5 and n6 are observed as single kinetic events (denoted by n3,4 and n5,6, respectively) in absorbance experiments. The fluorescence data shown in Table 5.1 has been reported previously (Booth et al 1996). The values quoted here and those reported earlier are broadly in agreement. In this previous study, however, time-resolved fluorescence data were analyzed separately over different time scales. The procedure used here to determine the rate constants and their amplitudes employs simultaneous analysis of data collected over different time scales. This latter method of analysis is of course more reliable. Thus, any minor discrepancies between the data shown in table 5.1 and those reported previously (Booth et al 1996) are understandable. The lack of an isobestic point in the time-resolved absorption spectra (Figs 5.1f and 5.2f) together with the SVD analysis (Figs 5.1g and 5.2g) provide an unequivocal

160 Chapter 5 demonstration of the formation of bacteriorhodopsin via a spectroscopically distinct intermediate. An isobestic point is the wavelength at which the extinction coefficients of two species are the same. The presence of a single isobestic point implies that the extinction coefficient does not change at this wavelength during the course of the reaction. This scenario can only be explained if a reactant decays directly into a product without the involvement of an intermediate. It is however conceivable that the intermediate may have no absorbance at this wavelength. Thus, conversely the lack of a single isobestic point indicates that the decay of a reactant into a product occurs via at least one intermediate. The lack of a single isobestic point observed here does not exclude the possibility that, in addition to the involvement of an intermediate, the formation of bacteriorhodopsin directly from bacterioopsin and retinal is also feasible. Indeed the wavelength-dependent amplitudes of the phases n8 and n9 (Figs 5.1h and 5.2h) indicate that the decay of retinal directly into bacteriorhodopsin is possible, at least in physical terms. Further experiments are thus necessary to resolve this issue (see Chapter 7). The change in retinal absorbance maximum from 380 nm in free retinal in the micelles to 430 nm in the intermediate is attributed to a change in the polarity of the surrounding environment (solvent- induced shift). The formation of a covalent bond between retinal and Lys216 is unlikely to be responsible for this observed shift (vide infra). The absorbance maxima of retinal and retinal Schiff base have been shown to be very similar in methylpentane (Schaffer et al 1974). The absorbance maxima of retinal in hexane and chloroform have been reported to be 370 and 390 nm, respectively (Alex et al 1992). Further, addition of trifluoroacetic acid in excess to retinal in hexane or chloroform led to a red shift in the absorbance maximum of retinal of up to 50 nm. The major factor responsible for this red shift was suggested to be the hydrogen bonding between the aldehyde group of retinal and the carboxylic group of trifluoroacetic acid. The retinal binding pocket within the protein contains hydrophobic, polar as well as charged amino acid residues (Henderson et al 1990, Grigorieff et al 1996, Kimura et al 1997). It therefore seems reasonable that hydrogen bonding between the aldehyde group of retinal and the polar or charged residues within the retinal binding pocket may in part be responsible for the red shift observed in the absorbance maximum of retinal from 380 nm in free retinal to 430 nm in the intermediate.

Kinetics of Retinal Binding to Bacterioopsin 161

The oscillator strength of the retinal absorption band is exploited here to identify the step(s) involving the formation of the Schiff base between the e-amino sidechain moiety of Lys216 and the aldehyde group of retinal. The oscillator strength is a measure of how strong a molecule absorbs in response to electromagnetic radiation or, in other words, it is the strength of an electronic transition between the ground state and the excited state. From a practical point of view, the area under an absorption band is proportional to the oscillator strength of a molecule (Gilbert & Baggot 1991). Since the energy of an electronic transition is proportional to l-1, the area under an absorption band must be expressed in dimensions of l-1. The study undertaken here involves measuring a change rather than an absolute value for the oscillator strength of the retinal absorption band. The change in the oscillator strength can be determined from the area under the wavelength-dependent amplitude of a particular kinetic phase (see Chapter 2). One assumption made here is that the wavelength-dependent amplitude of a kinetic phase equates to the difference spectrum between the decaying and the non-decaying species. It can be mathematically shown that this is so in the case of a mono-exponential function (see Chapter 2). However in the case of a multi-exponential function used here, the wavelength-dependent amplitudes would approximate to the difference spectra of the decaying and the non-decaying species given that the kinetic phases under consideration (n3,4, n5,6, n7, n8, and n9) can be well-resolved. This is indeed the case except for the phase n7 which does not appear to be well-resolved here. One thus should take precaution in the interpretation of the change in the oscillator strength that is observed during the phase n7 (Table 5.2). Further, it should be stressed that the approximation made here in determining the change in the oscillator of the retinal absorption band is reasonable and within the accuracy of the experimental technique. Such ad hoc approximation is necessary. The difference spectra cannot be directly calculated, unless a specific sequential step model is invoked for the various kinetic phases. This however does not appear to be the case. Rather branched pathways appear plausible. The method used here to determine the change in the oscillator strength of the retinal absorption band thus does not rely on any model. A change in the oscillator strength of the retinal absorption band may result due to a change in the polarity of the surrounding environment, due to a change in the

162 Chapter 5 geometry of the retinal molecule, due to Schiff base formation between the e-amino group of Lys216 and the aldehyde group of retinal, or due to electrostatic interactions between retinal Schiff base and the amino acid residues within the protein binding site. The oscillator strength of the retinal absorption band in free retinal and in the 430-nm intermediate appears to be the same (Figs 5.1i and 5.2i, Table 5.2). This implies that the oscillator strength of the retinal absorption band is independent of the polarity of the solvent. This salient observation is consistent with a previous study showing that the oscillator strength of retinal-like polyenes is relatively insensitive to solvent environment (Myers & Birge 1980). This raises the intriguing question of whether retinal is non-covalently or covalently bound to the protein within the 430-nm intermediate. If the latter possibility was accepted then one would expect a change in the oscillator strength of the retinal absorption band in the 430-nm intermediate. This conclusion is derived from an important observation that the cyclohexene ring and the polyene sidechain of retinal need to be coplanar in order for the subsequent Schiff base formation between retinal and the protein (Schreckenbach et al 1978a, Schreckenbach et al 1978b, Gartner & Oesterhelt 1988). The planarization of the retinal molecule would be expected to cause a change in the oscillator strength, which is highly sensitive to the conformation of retinal-like polyenes (Honig et al 1980). The most straightforward interpretation therefore is that retinal is non-covalently bound to the protein within the 430-nm intermediate. The formation of Schiff base may also alter the oscillator strength of the retinal absorption band but this is not a firm evidence per se. The oscillator strength of retinal and retinal Schiff base has been shown to be similar in methylpentane (Schaffer et al 1974). The binding of retinol (the alcohol analogue of retinal) to retinal binding pocket within bacterioopsin has been observed to occur with a lifetime of about 1 s (Schreckenbach et al 1977, Booth et al 1995). This is remarkably similar to the lifetime of the phase n5,6 that accompanies the formation of the 430-nm intermediate (Table 5.3). The fact that retinol cannot undergo covalent linkage to bacterioopsin lends further support for the assignment of the phase n5,6 to the non-covalent binding of retinal to the apoprotein. An intermediate with double-peaked absorbance maximum has been observed in the regeneration of bacteriorhodopsin from apomembrane (purple

Kinetics of Retinal Binding to Bacterioopsin 163 membrane minus retinal) and retinal at lower temperatures (Schreckenbach et al 1977). Retinal was also shown to be non-covalently bound to the protein within this 430/460-nm intermediate. In fact, the spectral properties of the bacteriorhodopsin mutant Lys216Ala resemble this 430/460-nm species (Schweiger et al 1994). These observations provide compelling support for the assignment of the 430- nm intermediate observed here to a non-covalent complex between retinal and the protein. The double-peaked absorption spectrum of the 430/460-nm intermediate results from the planarization of the cyclohexene ring and the polyene sidechain of retinal (Schreckenbach et al 1977, Schweiger et al 1994). The absolute absorption spectrum of the 430-nm intermediate however cannot be measured directly because of its transient nature in the refolding reaction of bacteriorhodopsin. It is however likely that the absorption spectrum of the 430-nm consists of a single peak, since retinal is envisaged to be in a non-planarized conformation within this intermediate. The oscillator strength of the retinal absorption band would be expected to differ from that in the free retinal if retinal assumed a planarized confromation within the 430-nm intermediate. This is however clearly not the case here. A species absorbing around 400 nm has also been observed in the regeneration studies of bacteriorhodopsin from apomembrane and retinal (Schreckenbach et al 1977, Schreckenbach et al 1978a). Retinal is thought to be planarized and non-covalently bound to the protein within this species. The proton pumping photocycle of bacteriorhodopsin also proceeds through a 410-nm species, in which retinal is covalently bound to the protein via a non-protonated Schiff base (Lanyi 1993). A species with an absorption maximum around 440 nm has been observed in the denaturation of purple membrane with SDS (Konishi & Packer 1977, London & Khorana 1982). This 440-nm species was shown to be the protonated Schiff base, in which interactions between retinal and the protein have largely been lost. The conclusion that is reached here is that retinal has a non-planar conformation and is non-covalently bound to the protein within the 430-nm intermediate. The decay of the 430-nm intermediate into the native chromophore can take from seconds

(phase n7) to minutes (phases n8 and n9), and is accompanied by a change in the absorbance maximum of retinal (Figs 5.1i and 5.2i, Table 5.2). The change in the absorbance maximum of retinal from 430 nm in the intermediate to around 560 nm in

164 Chapter 5 the native chromophore is attributed to the formation of a protonated Schiff base between retinal and the protein. A protonated retinal Schiff base absorbs around 440 nm (Pitt et al 1955, Liu et al 1993). It is thought that various interactions between retinal protonated Schiff base and the amino acid residues within the apoprotein shift the absorbance maximum of retinal from 440 nm in the protonated schiff base to around 560 nm in the native chromophore (Nakanishi et al 1980, Beppu & Kakitani 1994, Hu et al 1994). This phenomenon is called the opsin shift.

The native chromophore formed via the phase n7 has its absorption maximum slightly blue- shifted to around 520 nm. The possibility that micellar bacteriorhodopsin is a composite of two absorbing species cannot therefore be excluded. Mutiple absorbing species in regenerated bacteriorhodopsin have been observed previously (Liao et al 1983). The proton pumping photocycle of bacteriorhodopsin also proceeds through a 520-nm species (Mathies et al 1991, Oesterhelt et al 1992, Rothschild 1992, Lanyi 1993). It is however unlikely that the 520-nm species observed here represents the 520-nm species of the photocycle, since the latter has a lifetime of no more than a few milliseconds. The blue shift observed in the absorption maximum of a fraction of the native chromophore formed via the phase n7 could also result from the inability to resolve the phase n7 properly from the phases n8 and n9 (vide infra). The change in the oscillator strength of the retinal absorption band upon the decay of the 430-nm into the native chromophore is likely to be partly due to the ring-chain planarization of retinal. The oscillator strength is highly sensitive to the conformation of retinal-like polyenes (Honig et al 1980). The ring-chain planarization of retinal is a strict pre-requisite for the subsequent Schiff base formation, since retinal isomers that cannot coplanarize fail to regenerate bacteriorhodopsin (Schreckenbach 1978a). Interactions between the protonated Schiff base and the protein may also influence the oscillator strength of the retinal absorption band in the native chromophore (Birge et al 1987, Birge et al 1988). There is generally an increase in the oscillator strength of the retinal absorption band upon the decay of the 430-nm intermediate into the native chromophore (phases n8 and n9, Table 5.2). The decay of the 430-nm intermediate into the native chromophore via the phase n7 however occurs with a decrease in the oscillator strength. This may not

Kinetics of Retinal Binding to Bacterioopsin 165 necessarily be the case in reality, and could result from the fact that this phase could not be well- resolved from the phases n8 and n9. Photodiode array studies on the formation of the 430-nm intermediate have been recently reported (Booth & Farooq 1997). In this report, the decay of this intermediate into bacteriorhodopsin was not fully investigated. Further, a simple consecutive two-step scheme, modelling the formation of the 430-nm intermediate from retinal and bacterioopsin and its subsequent decay into bacteriorhodopsin, was used to calculate the difference spectra between retinal and the intermediate, and the intermediate and bacteriorhodopsin. The oscillator strength of the retinal absorption band in free retinal and the 430- nm intermediate was found to be the same, whereas the decay of the 430-nm intermediate into bacteriorhodopsin resulted in the increase in the oscillator strength of at least two-fold on the basis of the difference spectra calculated using the two-step scheme. The method used here provides a model- independent measure of the oscillator strength of the retinal absorption band during bacteriorhodopsin folding. Although this method is slightly less accurate than that used in the previous study (Booth & Farooq 1997), the simple two-step scheme is hardly an accurate representation of bacteriorhodopsin folding. The folding events in the minute range for bacteriorhodopsin folding may appear slow compared to the folding of water-soluble proteins (Kim & Baldwin 1990, Matthews 1993). Lifetimes in the minute range, however, have also been observed during the folding of bacterial porins (Surrey & Jahnig 1995, Surrey et al 1996, Kleinschmidt & Tamm 1996) and the plant light-harvesting complex II (Booth & Paulsen 1996). Some water-soluble ligand-binding proteins are also observed to fold in the minute range (Chiba et al 1994). Formation of certain covalent linkages between thiol groups of cysteine residues in the folding of water-soluble proteins RNAse A (Creighton 1980, Talluri et al 1994), RNAse T1 (Pace & Creighton 1986) and BPTI (Creighton 1981, Freedman 1995) has also been observed to occur with a lifetime in the minute range. The fact that a fraction of bacteriorhodopsin appears to reach the native state in only a few seconds (phase n7) whereas the remainder takes minutes (phases n8 and n9) suggests that the possibilities for incorrect retinal binding may be reserved. Thus a fraction of the 430-nm intermediate may contain retinal in the native configuration. This

166 Chapter 5 fraction may thus quickly decay into bacteriorhodopsin with a lifetime of a few seconds. In contrast, some fraction of the 430-nm intermediate may contain retinal in a non-native configuration. These molecules would thus be expected to undergo some rearrangement before the appearance of the native chromophore. The rearrangement may take several minutes. Under optimal conditions, cytochrome c folding occurs on a submillisecond time scale (Takahashi et al 1997, Yeh et al 1997). In oxidized cytochrome c, however, misligation of the haem to histidines transiently traps misfolded structures and markedly slows folding into almost a second regime (Sosnik et al 1994, Elove et al 1994, Sosnik et al 1996). It is conceivable that the decay of the 430-nm intermediate into bacteriorhodopsin via the slower steps (phases n8 and n9) is under thermodynamic control, whilst the kinetic factors are largely responsible for the faster step (phase n7). Thus the fraction of the native chromophore formed via the phase n7 from the 430-nm intermediate may be kinetically allowed but may not necessarily produce the thermodynamically most stable species, whereas the phases n8 and n9 may accompany the formation of the thermodynamically most stable species that lie on the global free energy minimum. In this respect, the fraction of the native chromophore formed via the phase n7 may slowly convert to the thermodynamically more stable species. On the basis of above results, one can invoke several kinetic schemes for the folding of bacteriorhodopsin (bR) from a partially denatured state in SDS (bO) and retinal (R). One such model is

k k 8 I 5 1 k 1 k3 k4 k7 R I bR

bO I3 I4 430

k 2 k 6 I2 k 9

in which the intrinsic rate constants k1-k9 equate to the experimentally observed rate constants n1-n9. In this model, the apoprotein intermediates I3 and I4 lie on a sequential

Kinetics of Retinal Binding to Bacterioopsin 167 pathway. That is, the apoprotein state with a native-like secondary structure (bO¢) is formed from the denatured state bO via an intermediate. The kinetic intermediate I4 is structurally similar to the equilibrium species bO¢.

Retinal is envisaged to bind to the apoprotein intermediate I4 in a parallel fashion to form the

430-nm intermediate (I430), which then subsequently decays into bacteriorhodopsin. Alternatively bacteriorhodopsin may be formed directly, via two kinetically distinct pathways, from the reaction between the intermediate I4 and retinal without the involvement of the 430-nm intermediate. Thus, according to this model, the 430-nm intermediate is not absolutely necessary in the folding of bacteriorhodopsin. The intrinsic rate constants (corresponding to the observed phases n5 and n6) could however equally denote consecutive steps. It is also possible that the phases n5 and n6 represent a single retinal-binding step. The point being made however is that the above model can account for all the kinetic data presented here. It is however equally conceivable that the formation of the state bO¢ occurs directly from the state bO without accumulation of any intermediates. That is, the apoprotein intermediates I3 and I4 are on parallel pathways. The model that accommodates these considerations is

k 8 k 5 I I3 1 k k 3 1 k bO 7 R I430 bR

k 2 k 4

I2 I4 k 6 k 9

in which the intermediates I3 and I4 decay into the 430-nm intermediate independently of each other upon the binding of retinal. In this model, both the kinetic intermediates I3 and I4 would structurally resemble the equilibrium state bO¢. The state bO¢ may

168 Chapter 5 therefore be a heterogenous mixture of several species rather than a specific conformation. Yet, an alternative model can be invoked. It is only a speculation that the 430-nm intermediate may not be absolutely necessary in the folding of bacteriorhodopsin. Thus the model

I1 I3 k k 3 k k8 1 5 k bO R 7 I430 bR k 2 k 4 I2 I4 k 6 k9

accommodates the notion that the conversion of the apoprotein intermediates I3 and I4 into bacteriorhodopsin must occur via the 430-nm intermediate. Note that, none of the models discussed above suggest an intrinsic role for the species I1 and I2 in the folding of bacteriorhodopsin. These species very probably represent off-pathway products, which may only have a transient existence and are unlikely to contribute to productive folding. One common theme among all three models presented above is that they invoke parallel routes rather than a specific pathway for the conversion of the partially denatured state bO into the native state. The recent consensus on the mechanisms of protein folding is that parallel pathways are a rule rather than an exception. The new view of protein folding suggests that folding can be considered as a distribution of structures from the unfolded to the native state rather than being confined to a well- defined pathway (Baldwin 1995, Dill & Chan 1997). In summary, the formation of a state of bacterioopsin with a native-like secondary structure appears to be rate-limiting for the subsequent binding of retinal. The binding of retinal to this state results in the formation of a spectroscopically distinct intermediate. This intermediate absorbs maximally around 430 nm. Retinal appears to be non-covalently bound to the apoprotein within this 430-nm intermediate. The

Kinetics of Retinal Binding to Bacterioopsin 169 formation of the final native state bacteriorhodopsin can apparently occur in three distinct pathways. One of these pathways involves the formation of bacteriorhodopsin from the 430-nm intermediate over a time scale of seconds. The other two pathways involve the formation of bacteriorhodopsin either directly from retinal and bacterioopsin or via the 430-nm intermediate. The formation of bacteriorhodopsin via these pathways can take minutes. The formation of bacteriorhodopsin is accompanied by the formation of a Schiff base between retinal and the apoprotein. Addition of retinal at various times in the refolding of bacterioopsin should allow further elucidation of the steps involving retinal binding and the role of apoprotein intermediates in the folding of bacteriorhodopsin (the subject of next chapter).

Chapter 6

KINETICS OF RETINAL BINDING AT VARIOUS TIMES IN THE REFOLDING OF BACTERIOOPSIN

6.1 Overview 172

6.2 Results 173

6.3 Discussion 182

172 Chapter 6

6.1 Overview

In the previous chapter, it has been shown that refolding of the apoprotein bacterioopsin is rate-limiting for the subsequent binding of retinal. Thus, in line with previous reports (London & Khorana 1982, Booth et al 1995), it was suggested that retinal probably binds to a state of bacterioopsin with a native- like secondary structure (bO¢). The multiphasic nature of the refolding kinetics of the apoprotein led to the intriguing speculation that retinal could bind in a parallel fashion to the state bO¢. That is, the state bO¢ is a composite of kinetically distinct substates rather than a single species. The binding of haem to the water-soluble protein cytochrome c is thought to proceed via multiple parallel steps (Elove et al 1994). This situation is however slightly different in that whereas retinal binds to a well-defined single site Lys216 within the apoprotein, haem has several coordination sites. Thus the parallel binding of haem could be explained by the order in which it coordinates to its axial ligands. The parallel binding of retinal, if it indeed prevails, must result from its binding to kinetically distinct apoprotein species. Folding studies on ligand-binding proteins are lacking compared to their non-ligand-binding counterparts. The membrane protein plant light-harvesting complex II has a strict requirement for its cofactors chlorophylls and xanthophylls to initiate its refolding (Booth & Paulsen 1996). Multiphasic kinetic are observed for the refolding of this protein from a partially denatured state in SDS. Whether the binding of cofactors to the plant light-harvesting complex II occurs in parallel is not known. In this chapter, stop-flow fluorescence and absorption spectroscopy are used to study the effect of adding retinal at various times in the refolding of bacterioopsin from a partially denatured state in SDS. It is hoped that these studies should further lead to elucidation of retinal binding step(s) and reveal further information regarding the nature of the rate-limiting step(s) in apoprotein refolding. If retinal does indeed bind to the apoprotein in a parallel fashion then one would expect to see the dependence of more than one kinetic phase, associated with bacteriorhodopsin refolding, on the time of addition of retinal. That is, there should be at least two independent rate-limiting steps in the refolding of the apoprotein bacterioopsin. In brief, this study should not only add to our knowledge of the mechanisms of ligand binding but should also provide a stern test of the idea that retinal does not initiate apoprotein folding.

Kinetics of Retinal Binding at Various Times in the Refolding of Bacterioopsin 173

6.2 Results

The effect of adding retinal to partially denatured bO in SDS and bO/DMPC/CHAPS micelles has been examined in the previous chapter. Yet, another way of studying retinal binding is to add retinal at various times in the refolding of bacterioopsin from a partially denatured state in SDS to a state with a native-like secondary structure in DMPC/CHAPS. This can be achieved by sequential-mixing stop-flow method. Retinal binding can be monitored by intrinsic protein fluorescence and, retinal absorbance at single wavelengths. Photodiode array spectroscopy however cannot be used in conjunction with the sequential-mixing method. Fig 6.1a shows changes in intrinsic protein fluorescence for the addition of retinal at various times between 0.1-1000 s in the refolding of bacterioopsin. Data at only certain times of addition of retinal are shown. Data were also collected over various times scales from 1 to 1000 s. Only two such time scales are shown. The initial rate of fluorescence quenching, as a result of energy transfer from Trp residues to retinal (Polland et al 1986a), apparently remains the same for up to a time of addition of retinal of about 1 s. There is little change in the initial fluorescence level during this period. The initial rate of fluorescence quenching then progressively increases with increasing the time of addition of retinal to 100 s. A similar increase in the initial fluorescence level recorded is also observed as a result of an increase in the extent of bacterioopsin refolding, which is characterized by a rise in fluorescence, with increasing the time of addition of retinal. Finally, the initial rate of fluorescence quenching plateaus out beyond a time of addition of retinal of 100 s. The above observations are thus consistent with the idea that bacterioopsin refolding is rate-limiting for the subsequent binding of retinal (see Chapter 5). Fig 6.1b shows changes in absorbance at single wavelengths of 380, 430 and 560 nm for the addition of retinal at various times between 0.1-1000 s in the refolding of bacterioopsin. Data at only certain times of addition of retinal are shown. Data were also collected over various times scales from 1 to 1000 s. Only two such time scales are shown. Note that, these wavelengths correspond to the absorbance maxima of the retinal chromophore, the spectroscopically distinct intermediate (see Chapter 5) and the native

174 Chapter 6

Fig 6.1 (pp 175-178). Addition of retinal at various times in the refolding of bO. DMPC/CHAPS was mixed, in the stop-flow, with an equal volume of bO/SDS. Retinal/DMPC/CHAPS was added at various times in the refolding, at a final protein to retinal molar ratio of 1:1. Retinal binding was monitored by intrinsic protein fluorescence and retinal absorbance.

(a) Retinal binding monitored by intrinsic protein fluorescence. Data at times of addition of retinal of 0.1, 1, 5, 20, 100 and 1000 s are shown. Data shown were collected over two time periods of 0-50 s and 50-1000 s using a split time scale, and represent an average of two transients. For each interval, 2000 data points were recorded. Data up to a time of addition of retinal of 1 s effectively required a sum of four exponentials, except that an additional exponential was necessary for data recorded at a time of addition of retinal of 0.1 s. This additional exponential accounts for the phase n2, which is prominent over the first 0.5 s in the mixing of bO/SDS with DMPC/CHAPS. Data at all other times of addition of retinal between 5-1000 s required a sum of five exponentials. Data shown over the two time periods were analyzed simultaneously using the kinetic analysis software Look. The fit is indicated by a solid line. Residuals are also shown.

(b) Retinal binding monitored by retinal absorbance using a photomultiplier tube. Absorbance changes at single wavelengths of 380, 430 and 560 nm are shown at times of addition of retinal of 0.1, 1, 5, 20, 100 and 1000 s. Data shown were collected over two time periods of 0-50 s and 50-1000 s using a split time scale, and represent an average of two transients. For each interval, 2000 data points were recorded. Data at all three wavelengths were globally analyzed to a sum of exponentials using Look. Data up to a time of addition of retinal of 1 s effectively required a sum of three exponentials, except that an additional exponential was necessary for data recorded at a time of addition of retinal of 0.1 s.

This additional exponential accounts for the phase n2, which is prominent over the first 0.5 s in the mixing of bO/SDS with DMPC/CHAPS. Data at all other times of addition of retinal between 5-1000 s required a sum of four exponentials. Data shown over the two time periods were analyzed simultaneously. The fit is indicated by a solid line. Residuals at 430 nm are shown.

(c) The dependence of rate constants of exponentials (rates a, b, g, d and e) on the time of addition of retinal for fluorescence measurements shown in (a). Each data point represents an average of three measurements on different bO preparations. Error bars are given to one standard deviation.

(d) The dependence of rate constants of exponentials (rates ab, g, d and e) on the time of addition of retinal for absorbance measurements shown in (b). Each data point represents an average of three measurements on different bO preparations. Error bars are given to one standard deviation.

Kinetics of Retinal Binding at Various Times in the Refolding of Bacterioopsin 175 a

3 0.1 s 2 1

0 0.02 0.00 -0.02 0 25 50 500 1000

3 1 s 2 1

0 0.02 0.00 -0.02 0 25 50 500 1000

3 5 s 2

1

0 0.02 0.00 -0.02 0 25 50 500 1000

3 20 s 2 1

0 0.02 0.00 -0.02 0 25 50 500 1000

3 100 s 2 1

0 0.02 0.00 -0.02 0 25 50 500 1000

3 1000 s 2 1

0 0.02 0.00 -0.02

0 25 50 500 1000 Residuals Relative fluorescence / V Time / s

176 Chapter 6

b 0.15 0.1 s 0.10 380 nm

0.05 430 nm 560 nm 0.00 0.001 0.000 -0.001 0 25 50 500 1000 0.15 1 s 0.10

0.05

0.00 0.001 0.000 -0.001 0 25 50 500 1000 0.15 5 s 0.10

0.05

0.00 0.001 0.000 -0.001 0 25 50 500 1000 0.15 20 s 0.10

0.05

0.00 0.001 0.000 -0.001 0 25 50 500 1000 0.15 100 s 0.10

0.05

0.00 0.001 0.000 -0.001 0 25 50 500 1000 0.15 1000 s 0.10

0.05 Absorbance 0.00 0.001 0.000 -0.001

0 25 50 500 1000 Residuals

Time / s

Kinetics of Retinal Binding at Various Times in the Refolding of Bacterioopsin 177

c

5 4 rate a 3 2 1 0 0.0 0.5 1.0 50 100 500 1000

1.5 rate b 1.0

-1 0.5

0.0 0.0 0.5 1.0 50 100 500 1000

0.50 rate g

0.25

0.00 0.0 0.5 1.0 50 100 500 1000

0.015 rate d

Experimentally observed rate constant / s 0.010

0.005

0.000 0.0 0.5 1.0 50 100 500 1000

0.0050 rate e

0.0025

0.0000 0.0 0.5 1.0 50 100 500 1000

Time of addition of retinal / s

178 Chapter 6

d

2.0 rate ab 1.5

1.0

0.5

0.0 0.0 0.5 1.0 50 100 500 1000

0.2 -1 rate g

0.1

0.0 0.0 0.5 1.0 50 100 500 1000

0.015 rate d 0.010

0.005 Experimentally observed rate constant / s

0.000 0.0 0.5 1.0 50 100 500 1000

0.0050 rate e

0.0025

0.0000 0.0 0.5 1.0 50 100 500 1000

Time of addition of retinal / s

Kinetics of Retinal Binding at Various Times in the Refolding of Bacterioopsin 179 protein bacteriorhodopsin. The initial rate of decay of absorbance at 380 nm is mirrored by the initial rise in absorbance at 430 nm at all times of addition of retinal. This reflects the formation of the 430-nm intermediate from retinal and bacterioopsin. Little change in absorbance at 560 nm, characteristic of a native-like chromophore of bacteriorhodopsin, is observed initially. This so-called lag period is indicative of the formation of bacteriorhodopsin via an intermediate. The rise in absorbance at 560 nm is matched by a similar decay of absorbance at 430 nm at all times of addition of retinal. This observation further strengthens the notion that an intermdiate is involved in the refolding of bacteriorhodopsin from retinal and bacterioopsin. Note that, the initial absorbance levels recorded at 380, 430 and 560 nm as a function of the time of addition of retinal remain the same. This is because bacterioopsin refolding, unlike fluorescence, is not associated with a change in absorbance. Such absorbance measurements at single wavelengths complement the fluorescence data above and photodiode array data presented in the previous chapter. To gain a more quantitative insight into the mechanism of retinal binding, the dependence of the various kinetic phases on the time of addition of retinal needs to be determined. This strategy should lead to identification of rate-limiting step(s) associated with retinal binding to the apoprotein much more clearly. Time-resolved fluorescence data S[t] shown over the two time scales (Fig 6.1a) were simultaneously fitted to a sum of exponentials using the function

S[t] = å {ai exp(- ni t)} + d [6.1]

where ai and ni are respectively the amplitude and the observed rate constant of an exponential phase i, and d is the baseline offset. Data up to a time of addition of retinal of 1 s effectively required a sum of four exponentials, except that an additional exponential was necessary for data recorded at a time of addition of retinal of 0.1 s. This additional exponential accounts for the phase n2, which is prominent over the first 0.5 s in the mixing of bO/SDS with DMPC/CHAPS. Data at all other times of addition of retinal between 5-1000 s required a sum of five exponentials. The dependence of rate constants of exponentials on the time of addition of retinal for fluorescence measurements is shown in Fig 6.1c. As can be seen, the rate constants of

180 Chapter 6 only two exponentials increase with increasing the time of addition of retinal from 1 s to 100 s (rates a and b). As the time of addition of retinal is increased, the phases n3 and n4 (associated with apoprotein refolding) progressively disappear and, phases n5 and n6 (associated with non-covalent binding of retinal) become evident. The phases n5 and n6 are an order of magnitude smaller than the phases n3 and n4, and thus cannot be resolved when apoprotein refolding is rate-limiting (time of addition of retinal of up to about 5 s). The rate a thus reflects the phase n3 in the region 0-1 s, and phase n5 beyond about 20 s in the refolding of bacterioopsin. Similarly, the rate b reflects the phase n4 in the region 0-1 s, and phase n6 beyond about 50 s in the refolding of bacterioopsin. The rates g, d and e are independent of the time of addition of retinal. The rate g reflects the phase n7 which only becomes apparent at a time of addition of retinal of 5 s and beyond. The phase n7 cannot be resolved for the times of addition of retinal of up to 1 s, because the apoprotein refolding is rate-limiting. When the refolding time is increased to 5 s, this rate-limiting barrier is at least partially removed and thus the phase n7 becomes apparent. The rates d and e reflect the phases n8 and n9, respectively. Since the phases n7, n8 and n9 have been assigned to the formation of Schiff base (see Chapter 5), these would not be expected to be dependent upon the time of addition of retinal. Time-resolved absorbance data S[t,l] at single wavelengths of 380, 430 and 560 nm shown over the two time scales (Fig 6.1b) were simultaneously fitted to a sum of exponentials in a global analysis using the function

S[t,l] = å {ai[l] exp(- ni t)} + d[l] [6.2]

where ai[l] and ni are respectively the wavelength-dependent amplitude and the observed rate constant of an exponential phase i, and d[l] is the wavelength-dependent baseline offset. Data up to a time of addition of retinal of 1 s effectively required a sum of three exponentials, except that an additional exponential was necessary for data recorded at a time of addition of retinal of 0.1 s. This additional exponential accounts for the phase n2, which is prominent over the first 0.5 s in the mixing of bO/SDS with DMPC/CHAPS. Data at all other times of addition of retinal between 5-1000 s required a sum of four exponentials. The dependence of rate constants of exponentials on the time of addition of retinal for absorbance measurements is shown in Fig 6.1d. The rates a and b observed in

Kinetics of Retinal Binding at Various Times in the Refolding of Bacterioopsin 181 fluorescence experiments cannot be distinguished in absorbance measurements. These are collectively denoted by rate ab. Because the phases n3 and n4, and the phases n5 and n6 cannot be distinguished in absorbance measurements (see Chapter 5), the rate ab here reflects the phase n3,4 in the region 0-1 s, and the phase n5,6 in the region beyond about 50 s in the refolding of bacterioopsin. Otherwise, the data shown in Fig 6.1d complement those shown in Fig 6.1c.

182 Chapter 6

6.3 Discussion

Fluorescence and absorbance spectroscopy are used here to probe retinal binding at various times in the refolding of the apoprotein bacterioopsin. The kinetics of retinal binding are virtually identical up to a time of addition of retinal of about 1 s (Figs 6.1c and 6.1d). This is interpreted as an evidence that retinal has no effect on the phases n1 and n2, which have been previously assigned to the formation of the putative intermediates I1 and I2 in the refolding of bacterioopsin (see Chapter 4). The fact that this is so lends support to the view that retinal does not bind to these apoprotein intermediates. The only assumption made here is that retinal binding is accompanied by quenching of intrinsic protein fluorescence. This is indeed likely to be the case as the retinal binding pocket contains four Trp residues (Henderson et al 1990, Grigorieff et al 1996). It is therefore difficult to imagine how retinal binding could occur without fluorescence quenching. Thus as speculated in previous chapters, the intermediates

I1 and I2 are unlikely to represent productive folding. It may be incorrect to refer to these species as intermediates at all. The fluorescence and absorbance changes which have been assigned to the formation of these species may simply result from a change in the micelle environment upon mixing SDS with DMPC/CHAPS. It is equally conceivable that these species have a transient existence during the initial mixing and, in this respect, should play no active role in bacterioopsin folding. Of the several kinetic phases associated with retinal binding, only two rate constants depend on the time of addition of retinal from 1 s up to about 100 s (Fig 6.1c). If retinal bound to a single apoprotein species, then only the formation of this species will be rate-limiting for the subsequent binding of retinal. That is, there will only be a single rate-limiting step in apoprotein refolding for the subsequent binding of retinal. The fact that two rate constants are dependent on the time of addition of retinal implies that there are two independent rate-limiting steps in apoprotein refolding. That this is so demonstrates strongly that retinal binds to at least two kinetically distinct apoprotein species. These two rate-limiting steps coincide with the phases n3 and n4, implying that both the apoprotein intermediates I3 and I4 very probably bind retinal independently. The fact that exponentials are used to approximate the bimolecular reaction between retinal and the apoprotein casts some doubt over whether the two rate-limiting steps observed here indeed represent two independent retinal binding steps. It is

Kinetics of Retinal Binding at Various Times in the Refolding of Bacterioopsin 183 conceivable that a single retinal binding step approximates to two exponentials here. It is however unlikely, although not impossible, that the rate constants of both of these exponentials would depend upon the time of addition of retinal. Thus although the possibility that there is only one retinal binding step cannot be unequivocally excluded, the most likely interpretation of this finding is that retinal binds in a parallel fashion to the apoprotein intermediates I3 and I4. A number of reasons can be put forward to further support this hypothesis. Firstly, the fact that cis-trans proline isomerization may contribute to parallel pathways in bacterioopsin folding is a strong possibility. There are three transmembrane proline residues in bacteriorhodopsin (Henderson et al 1990, Grigorieff et al 1996). Thus, unfolded polypeptide chains containing all native proline isomers will refold faster than those with some or all non-native proline isomers. Proline isomerization has been observed to be the cause of parallel folding pathways in water- soluble proteins RNAse A (Mayr et al 1996, Houry & Scheraga 1996), RNAse T1 (Odefey et al 1995), FKBP (Veeraraghavan et al 1996) and chymotrypsin inhibitor-2 (Jackson & Fersht 1991). The parallel nature of apoprotein refolding will support the view that there may be more than one retinal binding step. Secondly, the denatured state of bacterioopsin in SDS is likely to be highly heterogenous and the various conformations may follow distinct kinetic pathways during apoprotein refolding. It is further conceivable that not all polypeptide chains are likely to experience similar constraints imposed upon them by the lipid bilayer during refolding. The refolding of some molecules, for instance, may be hampered more than others by the lipid bilayer. Thus a fraction of protein may refold faster than the rest, giving rise to parallel pathways. These plausible speculations would support the presence of more than one bimolecular reaction involving retinal and bacterioopsin. Thirdly, the new view of protein folding considers folding as a distribution of structures from the unfolded to the native state rather than being confined to a well-defined pathway (Baldwin 1994, Baldwin 1995, Dill & Chan 1997). Unfolded molecules are envisaged to reach the native state via parallel alternative routes rather than a specific pathway. Any folding event is not considered as being specific but rather representing an ensemble of structures or reactions. It thus should not be surprising in view of such statistical mechanical ideas that more than one retinal-binding step is

184 Chapter 6 involved in bacteriorhodopsin refolding. Such considerations combined with the above results shift the balance in favour of the idea that the binding of retinal occurs in a parallel fashion to the apoprotein intermediates I3 and I4. In this regard, the formation of the apoprotein state with a native-like secondary structure in the micelles (bO¢) would be essentially a two-state process. That is, the formation of the state bO¢ from bacterioopsin partially denatured in SDS (bO) does not require intermediates. Thus in line with this argument the model of bacteriorhodopsin folding suggested in the previous chapter, in which the apoprotein intermedaites I3 and I4 were postulated to lie on a sequential pathway, gains little credibility. The results presented here however do provide further support for the model

k 8 k I 5 I 3 1 k k 3 1 k bO 7 R I430 bR

k 2 k

4

I I k 2 4 6 k 9

in which the intrinsic rate constants k1-k9 equate to the corresponding observed rate constants n1-n9 (see Chapter 5). In this model, the retinal chromophore is envisaged to bind independently to the apoprotein intermediates I3 and I4. The product of these bimolecular reactions is the 430-nm intermediate. This intermediate then subsequently decays into bacteriorhodopsin. Alternatively the bimolecular reactions involving the apoprotein intermediates I3 and I4, and retinal may result in the formation of bacteriorhodopsin without the accumulation of the 430-nm intermediate. Thus, the 430-nm intermediate is not proposed to be critical for bacteriorhodopsin folding.

To reiterate, the role and the nature of the species I1 and I2 is not clear. These species very probably represent off-pathway intermediates or may only have a transient

Kinetics of Retinal Binding at Various Times in the Refolding of Bacterioopsin 185 existence. What is clear however is that these species are very unlikely to be involved in the productive folding of bacteriorhodopsin. The foregoing argument is also in support of the view that retinal is unlikely to have an obligatory role in initiating or directing apoprotein folding (London & Khorana 1982, Booth et al 1995). The results presented above are also consistent with the model

I1 I3 k k 3 k k8 1 5 k bO R 7 I430 bR k 2 k 4 I2 I4 k 6 k9

in which the 430-nm is considered to be absolutely necessary for bacteriorhodopsin folding. That is, bacteriorhodopsin cannot be formed directly from the reaction between retinal and the apoprotein intermediates I3 and I4. In previous studies on bacteriorhodopsin regeneration from apomembrane (purple membrane minus retinal) and retinal at lower temperatures, two species absorbing around 400 nm and 430/460 nm were envisaged to lie on a sequential pathway in the refolding of bacteriorhodopsin (Oesterhelt et al 1973, Oesterhelt & Schuhmann 1974, Schreckenbach et al 1977, Schreckenbach et al 1978a). Retinal was proposed to be in a planarized conformation and non-covalently bound to the protein within the 400-nm and 430/460-nm species. The Schiff base formation would occur upon the decay of the 430/460-nm species into bacteriorhodopsin. In light of such observations, it is conceivable that the 430- nm intermediate observed here is a composite of more than one absorbing species. On the basis of these considerations, the 400-nm and the 430/460-nm species observed in the previous study could thus lie on parallel pathways. This is the first report that suggests that there is likely to be more than one bimolecular reaction involving retinal and bacterioopsin. Previous studies on bacteriorhodopsin folding have pointed to a single retinal binding step (Booth et al

186 Chapter 6

1995, Booth et al 1996, Booth & Farooq 1997). Although kinetics of retinal binding at various times in the refolding of bacterioopsin have been studied previously (Booth et al 1996), no attempt was made to rationalize the data at a quantitative level. A single retinal binding step is also proposed in the refolding of bacteriorhodopsin from apomembrane (purple membrane minus retinal) and retinal at lower temperatures (Schreckenbach et al 1977), but multiple retinal binding events still seem possible. The coordination of haem to its axial ligands in cytochrome c (Elove et al 1994) is also thought to occur via parallel alternative routes. Non-native histidine ligands can become trapped during refolding of cytochrome c. The parallel pathways result from the competition between the native and non-native ligands. In the unfolded protein, haem may coordinate to non-native ligands which then become exchanged with native ligands during refolding. Alternatively, some fraction of the unfolded protein may be directly converted to the native conformation in which the haem only binds native ligands. Formation of intradisulphide linkages during BPTI folding is also thought to proceed via alternative routes (Weissman & Kim 1995). That is, there is no specific sequence in which the various disulphide linkages are formed. A number of additional conclusions can be drawn from the data presented here. The kinetics of retinal binding beyond about 100 s in the refolding of bacterioopsin remain the same (Figs 6.1c and 6.1d). This implies that bO equilibrates with the micelles within about two minutes after initial mixing of bO/SDS with DMPC/CHAPS. This finding is consistent with the assignment of the slow downward drift in intrinsic protein fluorescence observed after about 250 s in the refolding of bacterioopsin to the photobleaching of aromatic residues rather than due to a further conformational change within the apoprotein (see Chapter 4). The state of bacterioopsin in equilibrium with the micelles is envisaged to contain a native-like secondary structure (Huang et al 1981, London & Khorana 1982). It thus looks increasingly likely, in view of the results presented here, that this partially structured state (bO¢) represents an intermediate in the folding reaction of bacteriorhodopsin. Folding intermediates are generally thought of as being transitory. The state bO¢ could thus represent another stable state in addition to the ground states (the unfolded and the native conformations). In an analogy with water-soluble proteins, a third stable state has also been observed under mild denaturing conditions (Thomas et

Kinetics of Retinal Binding at Various Times in the Refolding of Bacterioopsin 187 al 1983, Baldwin 1990, Christensen & Pain 1991, Redfield et al 1994). This so-called stable molten globule is thought to contain substantial secondary structure with fluctuating tertiary interactions, and structurally resembles a transient intermediate (kinetic molten globule) observed in the folding of some proteins (Ikeguchi et al 1986, Ptitsyn et al 1990, Chaffotte et al 1992). The state bO¢ could thus represent the membrane protein-equivalent of a molten globule as suggested previously (Chapter 4). In summary, the results of this chapter are in support of the view that retinal binds solely to the putative apoprotein intermediates I3 and I4, and does so in a parallel fashion. The formation of these intermediates is rate-limiting for the subsequent binding of retinal. There is no evidence to suggest that the species I1 and I2 play an active role in bacteriorhodopsin folding. The effect of varying retinal concentration on the kinetics of bacteriorhodopsin refolding should reveal further clues to the mechanism of retinal binding (the subject of next chapter).

Chapter 7

DEPENDENCE OF REFOLDING KINETICS OF BACTERIORHODOPSIN ON RETINAL CONCENTRATION

7.1 Overview 190

7.2 Results 191 7.2.1 Bacteriorhodopsin Refolding Under Limiting Amounts of Retinal 191 7.2.2 Bacteriorhodopsin Refolding in Excess Retinal 196

7.3 Discussion 207

190 Chapter 7

7.1 Overview

The results presented in previous chapters lend support to the idea that the apoprotein refolding is rate- limiting for the subsequent binding of retinal. Other lines of evidence for this notion are also available (London & Khorana 1982, Booth et al 1995). To provide further experimental evidence in support of this view, the refolding of bacteriorhodopsin is carried out at a range of retinal concentrations using stop-flow fluorescence and photodiode array spectroscopy. Further, these studies should also provide a deeper insight into the role of the putative apoprotein intermediates I1 and I2 in the folding of bacteriorhodopsin. It has been demonstrated in the previous chapter that retinal most likely binds in a parallel fashion to the apoprotein intermediates I3 and I4. This finding is in contrast to the previous studies in which only a single retinal binding step was thought to be involved (Schreckenbach et al 1977, London & Khorana 1982, Booth et al 1995). Consider the retinal-binding step

k+R bO¢ + R I430 k-R where bO¢ is a state of the apoprotein bacterioopsin with a native-like seconday structure in the micelles,

R is retinal and I430 is the 430-nm intermediate in the folding reaction of bacteriorhodopsin. The elementary association and dissociation rate constants k+R and k-R are related to the observed rate constant nR of the retinal binding step and the retinal concentration [R], in the region where it is in excess over bacterioopsin, by the relationship

nR = k+R [R] + k-R

This is essentially an equation of a straight line. The experimentally observed rate constant should thus follow a linear increase with increasing retinal concentration. Thus if there is indeed more than one retinal binding step, one would expect at least two observed rate constants associated with bacteriorhodopsin refolding to depend on retinal concentration. That is, there should be at least two independent bimolecular reactions involving retinal and bacterioopsin. How thermodynamic information can be obtained on the 430-nm intermediate will also be demonstrated.

Dependence of Refolding Kinetics of Bacteriorhodopsin on Retinal Concentration 191

7.2 Results

7.2.1 Bacteriorhodopsin Refolding Under Limiting Amounts of Retinal

Addition of retinal in DMPC/CHAPS to partially denatured bO in SDS (bO/SDS) and bO pre- equilibrated with the micelles (bO/DMPC/CHAPS), at a final protein to retinal molar ratio of 1:1, has been studied in Chapter 5 using both the intrinsic protein fluorescence and retinal absorbance. The same experiments are repeated here under conditions where retinal concentration is limiting. Refolding is monitored by intrinsic protein fluorescence only. Absorbance measurements are not presented due to the poor signal-to-noise ratio obtained with the low concentrations of retinal used here. Fig 7.1a shows changes in intrinsic protein fluorescence upon the addition of retinal/DMPC/CHAPS to bO/SDS at a range of protein to retinal molar ratios. Data were collected over various time scales from 0.1 to 1000 s. Only two such time scales are shown. As discussed before (see Chapters 4 and 5), the overall rise in fluorescence observed upon bacterioopsin refolding in the absence of retinal is due to the transfer of aromatic residues from a less hydrophobic environment in SDS to a more hydrphobic environment in DMPC/CHAPS (protein to retinal molar ratio of 1:0). Addition of retinal to bacterioopsin in equimolar amounts results in an overall fluorescence decay, as a result of quenching of protein fluorescence by retinal (protein to retinal molar ratio of 1:1). This quenching of protein fluorescence is due to energy transfer from Trp residues to the retinal moiety (Polland et al 1986a). As can be seen, the amplitude of fluorescence decay in the presence of retinal in equimolar amounts is about four-fold greater than the fluorescence rise in its absence. Addition of retinal in excess to bacterioopsin has no further effect on the rate or the amplitude of the fluorescence decay (data not shown). However between these two extremes, a competition between fluorescence rise and decay is evident. At a protein to retinal molar ratio of 16:1, the overall amplitude of fluorescence rise appears slightly diminished. This results from the binding of retinal to a small fraction of apoprotein molecules. Upon increasing retinal concentration to a protein to retinal molar ratio of 8:1, the overall fluorescence rise is interrupted by a small fluorescence decrease (inset). Thus the greater amplitude of fluorescence quenching relative to fluorescence rise means that the binding of retinal to the

192 Chapter 7

Fig 7.1 (pp 193-194). Bacteriorhodopsin refolding under limiting amounts of retinal. Retinal/DMPC/CHAPS was mixed, in the stop-flow cuvette, with an equal volume of bO/SDS or bO/DMPC/CHAPS. bO/DMPC/CHAPS was prepared by manually mixing bO/SDS with an equal volume of DMPC/CHAPS and allowed to equilibrate for 30 min. Final protein concentration was constant at 2 mM. Retinal binding was monitored by intrinsic protein fluorescence.

(a) Addition of retinal/DMPC/CHAPS to bO/SDS. Data at protein to retinal molar ratios of 1:0, 16:1, 8:1, 4:1, 2:1 and 1:1 are shown. Data are shown over two time periods of 0-0.25 s and 0.25-1000 s, and represent an average of eight and two transients respectively. Data over these two time periods were collected over full scales of 0.5 and 1000 s, respectively. For each scale, 4000 data points were recorded. Note that the insets shown for data at protein to retinal molar ratios of 8:1 and 4:1 are over the time periods of 0.25-100 s and 0.25-500 s, respectively.

(b) Addition of retinal/DMPC/CHAPS to bO/DMPC/CHAPS. Data at protein to retinal molar ratios of 1:0, 16:1, 8:1, 4:1, 2:1 and 1:1 are shown. Data are shown over two time periods of 0-1 s and 1-1000 s, and represent an average of eight and two transients respectively. Data over these two time periods were collected over full scales of 1 s and 1000 s, respectively. For each scale, 4000 data points were recorded.

Dependence of Refolding Kinetics of Bacteriorhodopsin on Retinal Concentration 193

a

3

2

1 1:0 0 0.00 0.25 500 1000

3

2

1 16:1 0 0.00 0.25 500 1000

3

2 2.73

1 2.66 8:1 50 100 0 0.00 0.25 500 1000

3

2 2.6

1 2.4 4:1 0 250 500 0.00 0.25 500 1000

3

2

1 2:1 0 0.00 0.25 500 1000

3

2

1 1:1 0 Relative fluorescence / V 0.00 0.25 500 1000 Time / s

194 Chapter 7

b

3

2

1 1:0 0 0.0 0.5 1.0 500 1000

3

2

1 16:1 0 0.0 0.5 1.0 500 1000

3

2

1 8:1 0 0.0 0.5 1.0 500 1000

3

2

1 4:1 0 0.0 0.5 1.0 500 1000

3

2

1 2:1 0 0.0 0.5 1.0 500 1000

3

2

1 1:1 0

Relative fluorescence / V 0.0 0.5 1.0 500 1000 Time / s

Dependence of Refolding Kinetics of Bacteriorhodopsin on Retinal Concentration 195 apoprotein is initially seen as a decrease in fluorescence. The fact that retinal can only bind to a small fraction of the apoprotein molecules means that this initial decrease in fluorescence is subsequently superceded by a fluorescence rise. A similar explanation can be offered regarding fluorescence changes observed at a protein to retinal molar ratio of 4:1(inset), except that an overall fluorescence decay is interrupted by a small fluorescence rise. This can be explained by the fact that retinal binding to bacterioopsin occurs in fast and slow stages. Thus although apoprotein refolding is almost complete 100 s after initiation, retinal binding (fluorescence quenching) occurs over stages that can take from hundreds of milliseconds to several minutes (see Chapter 5). When retinal concentration is increased further to a protein to retinal molar ratio of 2:1, an overall decrease in fluorescence is evident because of the larger amplitude of fluorescence quenching relative to fluorescence rise. Another important observation that can be derived from Fig 7.1a is that varying retinal concentration has no effect on fluorescence changes associated with phases n1 and n2 (initial fluorescence rise followed by a slight decrease observed over the first 0.25 s). Fig 7.1b shows changes in intrinsic protein fluorescence upon the addition of retinal/DMPC/CHAPS to bO/DMPC/CHAPS. Data were collected over various time scales from 0.1 to 1000 s. Only two such time scales are shown. As can be seen, addition of retinal at all concentrations results in an overall fluorescence quenching. This is because retinal is added after the rate-limiting step(s) in apoprotein refolding has been overcome. Increasing retinal leads to a corresponding increase in the amplitude of fluorescence quenching, implying that all retinal that is added binds. The data presented in Fig 7.1 are essentially those published previously (Booth et al 1996).

196 Chapter 7

7.2.2 Bacteriorhodopsin Refolding in Excess Retinal

The purpose of this experiment is to identify the phases in bacteriorhodopsin refolding that are dependent on retinal concentration. Two requirements however need to be satisfied in order for this experiment to succeed. One is the removal of the rate-limiting barrier in apoprotein refolding, and one is that the retinal concentration should not be rate-limiting. The first requirement can be fulfilled by allowing bO to pre-equilibrate with the micelles (bO/DMPC/CHAPS) before the addition of retinal. The second requirement can be satisfied by adding retinal in excess over bacterioopsin. Retinal concentrations used are at least four-fold in excess over bacterioopsin concentration. Fig 7.2a shows changes in intrinsic protein fluorescence obtained upon the addition of retinal/DMPC/CHAPS to bO/DMPC/CHAPS at a range of protein to retinal molar ratios. An increase in the initial rate of fluorescence quenching with increasing retinal concentration, from protein to retinal molar ratio of 1:4 to 1:16, is evident. In fact, under these conditions the binding of retinal is so fast that some fluorescence quenching occurs within the deadtime of the measurement. This can be judged by a drop in the initial fluorescence level recorded. Fig 7.2b shows time-resolved changes in the retinal absorption band upon the addition of retinal/DMPC/CHAPS to bO/DMPC/CHAPS at a range of protein to retinal molar ratios. The time- resolved spectra between 350-750 nm were measured using a photodiode array. In order to clearly observe the changes in the retinal absorption band as a function of retinal concentration, plots of absorbance changes at single wavelengths of 380, 430 and 560 nm are also shown at corresponding protein to retinal molar ratios (Fig 7.2c). As can be seen, the appearance of a native-like chromophore absorption band at 560 nm occurs at the expense of the retinal absorption band at 380 nm at all protein to retinal molar ratios. Increasing retinal from protein to retinal molar ratio of 1:4 to 1:12 leads to an increase in the initial rate of decay of absorbance at 380 nm, which is mirrored by a similar increase in absorbance at 430 nm. This reflects the decay of retinal directly into the 430-nm intermediate (see Chapter 5). The initial rate of change of absorbance at 560 nm as a function of retinal concentration remains the same. These observations are consistent with the decay of retinal into the native chromophore via a spectroscopically distinct intermediate, as noted previously (see Chapters 5 and 6).

Dependence of Refolding Kinetics of Bacteriorhodopsin on Retinal Concentration 197

Time-resolved fluorescence data S[t] shown over the two time scales (Fig 7.2a) at various protein to retinal molar ratios were simultaneously fit to a sum of four exponentials using the function

S[t] = å {ai exp(- ni t)} + d [7.1]

where ai and ni are respectively the amplitude and the observed rate constant of an exponential phase i, and d is the baseline offset. The various exponentials reflect the five phases n5, n6, n7, n8 and n9 associated with the binding of retinal to bacterioopsin (see Chapter 5). The dependence of the rate constants of these five phases on retinal concentration is shown in

Fig 7.2d. As can be seen, the rate constants of only two phases (n5 and n6) appear to be dependent on retinal concentration. Increasing retinal concentration beyond protein to retinal molar ratio of 1:16 is not desirable, as substantial fluorescence quenching within the deadtime of the measurement makes the determination of rate constants associated with the fast phases n5 and n6 less reliable. A preliminary account of the data presented in Fig 7.2d has been given previously (Booth et al 1996). In this report, the binding of retinal to bO pre-equilibrated with the micelles could only be resolved into four kinetic phases instead of five observed here. Further, only one phase appeared to be dependent on retinal concentration instead of two reported here. There are two main reasons for this discrepancy: Firstly, the density of data in the region (0-1 s) where retinal binding predominates was at least ten-fold lower in the previous report. Secondly, data at various protein to retinal molar ratios were analyzed separately over different time scales instead of simultaneous analysis of data performed here. In fact, the time scales used to obtain the rate constants of the two fast phases in the previous report were not the same for various protein to retinal molar ratios. Thus, for example, a time scale of about 1 s was used for data taken at protein to retinal molar ratio of around 1:4 and a time scale of about 0.1 s for the data taken at protein to retinal molar ratio of around 1:16. The residuals obtained were also very poor. The collection of a higher density of data in the region 0-1 s and the availability of Look (which allows

198 Chapter 7

Fig 7.2 (pp 199-203). Bacteriorhodopsin refolding in excess retinal. Retinal in DMPC/CHAPS was mixed, in the stop-flow cuvette, with an equal volume of bO/DMPC/CHAPS. bO/DMPC/CHAPS was prepared by manually mixing bO/SDS with an equal volume of DMPC/CHAPS and allowed to equilibrate for 30 min. Final protein concentration was constant at 2 mM. Refolding was monitored by intrinsic protein fluorescence and retinal absorbance.

(a) Refolding monitored by intrinsic protein fluorescence. Data at protein to retinal molar ratios of 1:4, 1:6, 1:8, 1:10, 1:12 and 1:16 are shown. Data shown were collected over two time periods of 0-1 s and 1-1000 s using a split time scale, and represent an average of two transients. For each interval, 2000 data points were recorded. Data at all protein to retinal molar ratios shown were fit to a sum of five exponentials, corresponding to the five kinetic phases n5, n6, n7, n8 and n9 associated with retinal binding to bacterioopsin. Data shown over the two time scales were analyzed simultaneously using the kinetic analysis software Look. The fit is indicated by a solid line. Residuals are also shown.

(b) Refolding monitored by retinal absorbance using a photodiode array. Time-resolved absorption spectra are shown for protein to retinal molar ratios of 1:4, 1:6, 1:8, 1:10 and 1:12. Data are shown over two time periods of 0-4 s and 4-1000 s. Data over these two time periods were collected over full scales of 4 s and 1000 s, respectively. For each full scale, 400 spectra were recorded between 350-750 nm. Data at each protein to retinal molar ratio were fit to a sum of four exponentials, corresponding to the four kinetic phases n5,6, n7, n8 and n9 associated with retinal binding to bacterioopsin, in a global analysis using Look. Data over the two time periods shown were analyzed simultaneously.

(c) Absorbance changes at single wavelengths of 380, 430 and 560 nm are shown for the data collected in (b) at protein to retinal molar ratios of 1:4, 1:6, 1:8, 1:10 and 1:12. The fit of a four-exponential is indicated by a solid line. The residuals for the absorbance changes at 430 nm are also shown.

(d) The dependence of rate constants of the five kinetic phases n5, n6, n7, n8 and n9 on retinal concentration for fluorescence measurements shown in (a). The solid lines through rate constants of the phases n5 and n6 represent linear fits to the data. Errors were calculated from measurements on three different bO preparations. Error bars are given to one standard deviation.

(e) The dependence of rate constants of the four kinetic phases n5,6, n7, n8 and n9 on retinal concentration for absorbance measurements shown in (b). The solid line through rate constant of the phase n5,6 represents a linear fit to the data. Errors were calculated from measurements on three different bO preparations. Error bars are given to one standard deviation.

Dependence of Refolding Kinetics of Bacteriorhodopsin on Retinal Concentration 199

a

3 1:4

2

1

0 0.02 0.00 -0.02 0.0 0.5 1.0 500 1000

3 1:6

2

1

0 0.02 0.00 -0.02 0.0 0.5 1.0 500 1000

3 1:8

2

1

0 0.02 0.00 -0.02 0.0 0.5 1.0 500 1000

3 1:10

2

1

0 0.02 0.00 -0.02 0.0 0.5 1.0 500 1000

3 1:12

2

1

0 0.02 0.00 -0.02 0.0 0.5 1.0 500 1000

3 1:16

2

1

0 0.02 0.00 -0.02

0.0 0.5 1.0 500 1000 Residuals Relative fluorescence / V

Time / s

200 Chapter 7

b 1.0 1:4 0.5

0.0 400 500 600 700 0 2 4 500 1000 1.0 1:6 0.5

0.0 400 500 600 700 0 2 4 500 1000 1.0 1:8 0.5

0.0 400 500 600 700 0 2 4 500 1000 1.0 1:10 0.5

0.0 400 500 600 700 0 2 4 500 1000 1.0 1:12 0.5

0.0 Absorbance 400 500 600 700 0 2 4 500 1000 Wavelength / nm Time / s

Dependence of Refolding Kinetics of Bacteriorhodopsin on Retinal Concentration 201

c 1.0 1:4

0.5

0.0 0.002 0.000 -0.002 0 2 4 500 1000 1.0 1:6

0.5

0.0 0.002 0.000 -0.002 0 2 4 500 1000 1.0 1:8

0.5

0.0 0.002 0.000 -0.002 0 2 4 500 1000 1.0 1:10

0.5

0.0 0.002 0.000 -0.002 0 2 4 500 1000 1.0 1:12 380 nm

0.5 430 nm Absorbance 560 nm 0.0 0.002 0.000 -0.002 0 2 4 500 1000 Residuals

Time / s

202 Chapter 7 d

20 n5

16

12

8 0 5 10 15 20 25 30 35

12

n6 8

4

0 -1 0 5 10 15 20 25 30 35

n 0.4 7

0.2

0.0 0 5 10 15 20 25 30 35

0.02 n8 Experi men tally observed rate constant / s

0.01

0.00 0 5 10 15 20 25 30 35

0.003

n9 0.002

0.001

0.000 0 5 10 15 20 25 30 35

Retinal concentration / mM

Dependence of Refolding Kinetics of Bacteriorhodopsin on Retinal Concentration 203 e

10 n5,6

5

0 0 5 10 15 20 25 30

0.2 n7

-1 0.1

0.0 0 5 10 15 20 25 30

0.02 n8

0.01 Experimentally observed rate constant / s

0.00 0 5 10 15 20 25 30

0.003 n9 0.002

0.001

0.000 0 5 10 15 20 25 30

Retinal concentration / mM

204 Chapter 7 simultaneous analysis of data of any complexity over different time scales) have led to a refinement in the kinetic resolution of retinal binding to bacterioopsin here. Time-resolved spectra S[t,l] shown over the two time scales (Fig 7.2b) at various protein to retinal molar ratios were simultaneously fit to a sum of five exponentials in a global analysis using the function

S[t,l] = å {ai[l] exp(- ni t)} + d[l] [7.2]

where ai[l] and ni are respectively the wavelength-dependent amplitude and the observed rate constant of an exponential phase i, and d[l] is the wavelength-dependent baseline offset. The various exponentials reflect the phases n5,6, n7, n8 and n9 associated with retinal binding to bacterioopsin (see Chapter 5). The dependence of the rate constants of these four phases on retinal concentration is indicated in Fig 7.2e. These results complement fluorescence data presented above in that only the rate constant of the phase n5,6 is dependent on retinal concentration. Note that, the phase n5,6 denotes the fluorescence phases n5 and n6, which cannot be distinguished in absorbance experiments (see Chapter 5). The slight decrease in rate constants of the phases n7, n8 and n9 with increasing retinal concentration is most likely due to retinal photobleaching, which becomes more pronounced at higher protein to retinal molar ratios. Increasing the intensity of excitation light appears to exaggerate this decrease (data not shown). Conversely, a reduction in the intensity of the excitation light appears to minimize this decrease. Increasing retinal concentration beyond protein to retinal molar ratio of 1:12 is not desirable, as higher concentrations of retinal saturate the photodiode array signal, with the effect that the subsequent changes in absorbance at 380 nm cannot be accurately measured.

The observed rate constants for the phases n5 and n6 and n5,6 obey pseudo-first-order kinetics under conditions where retinal is in excess over bacterioopsin according to the equations (see Chapter 2)

n5 = k+5 [R] + k-5

n6 = k+6 [R] + k-6

n5,6 = k+5,6 [R] + k-5,6

Dependence of Refolding Kinetics of Bacteriorhodopsin on Retinal Concentration 205 where k+5, k+6 and k+5,6 are the intrinsic association rate constants and, k-5, k-6 and k-5,6 the corresponding dissociation rate constants for the reaction

bO¢ + R I430

involving retinal (R) and the state of bacterioopsin with a native-like secondary structure in the micelles

(bO¢) to form the 430-nm intermediate (I430).

A plot of observed rate constants for the phases n5, n6 and n5,6 as a function of retinal concentration is thus expected to be a straight line (Figs 7.2d and 7.2e). The values for the intrinsic association and dissociation rate constants can be obtained from the slope and the y-intercept of these plots. These are listed in Table 7.1. Also shown in Table 7.1 are the corresponding equilibrium constants and the free energy changes calculated for the reaction of retinal with bacterioopsin. The equilibrium constant (K) is a ratio of the intrinsic association rate constant to the intrinsic dissociation rate constant, and the free energy change (DG) is calculated from the relationship

DG = - RT lnK where R is the universal molar gas constant and T is the absolute temperature.

206 Chapter 7

Table 7.1. Intrinsic rate constants (k), equilibrium constants (K) and free energy changes (DG) for the phases n5, n6 and n5,6 associated with reactions involving retinal and bacterioospin at 295 K and pH 6.

Intrinsic rate constant Phase association / s-1mM-1 dissociation / s-1 K / mM-1 - DG / kJmol-1

n5 0.32 ± 0.03 (k+5) 8.41 ± 1.77 (k-5) 0.04 ± 0.01 26 ± 1

n6 0.27 ± 0.02 (k+6) 0.51 ± 0.45 (k-6) 0.53 ± 0.47 32 ± 4

n5,6 0.25 ± 0.03 (k+5,6) 0.71 ± 0.43 (k-5,6) 0.35 ± 0.22 31 ± 2

Footnotes to Table 7.1

Intrinsic association and dissociation rate constants (k) were obtained from the slopes and y-intercepts of the plots of the phases n5, n6 and n5,6 versus retinal concentration respectively (Figs 7.2d and 7.2e).

Equilibrium constant (K) is the ratio of the association intrinsic rate constant to the dissociation intrinsic rate constant.

Free energy change is calculated from the relationship DG = - RTlnK.

Errors were calculated from measurements on three different bO preparations. All errors are given to one standard deviation.

Dependence of Refolding Kinetics of Bacteriorhodopsin on Retinal Concentration 207

7.3 Discussion

This study demonstrates refolding of bacteriorhodopsin at a range of retinal concentrations, in an attempt to gain a deeper insight into the mechanism of retinal binding to the apoprotein bacterioopsin. The data presented here for bacteriorhodopsin refolding under limiting amounts of retinal are consistent with the view that the apoprotein refolding is rate-limiting for the subsequent binding of retinal (Fig 7.1). This is in contrast to the observation that chlorophylls and xanthophylls are necessary for initiating the folding of the plant light-harvesting complex II (Plumley & Schmidt 1987, Paulsen et al 1990, Booth & Paulsen 1996). Thus, the view that ligand molecules may be necessary for facilitating the folding of proteins within biological membranes no longer holds generality. Rather, some membrane proteins may have an absolute requirement for ligand molecules for their assembly whereas in others a ligand molecule may only have a subsidiary role. The initiation of the folding of the water-soluble protein cytochrome c is also dependent upon the presence of its ligand haem (Hamada et al 1993, Elove et al 1994). However apomyoglobin can refold to a molten globule-like state in the absence of haem (Jennings & Wright 1993). Although refolding kinetics of bacteriorhodopsin under limiting amounts of retinal can be explained by a single retinal binding event, these data by no means exclude the possibility that more than one bimolecular reaction may be operating. Studies on Ca2+-binding proteins parvalbumin (Kuwajima et al 1988) and a-lactalbumin (Kuwajima et al 1989) indicate that the binding of Ca2+ occurs after a rate- limiting step in the folding. Parvalbumin retains a native-like secondary structure but little rigid tertiary structure upon the removal of Ca2+ (Sudhakar et al 1995). This state of parvalbumin is thought to resemble a molten globule (Ikeguchi et al 1986, Elove et al 1992). Thus, in analogy with parvalbumin, retinal probably binds to a state of bacterioopsin resembling a molten globule. This molten globule state of bacterioopsin however could be a much more compact structure than that observed in the folding of water-soluble proteins, as a result of the restrictions imposed upon the protein by membrane lipids. Molten globules are thought to be general intermediates in the folding of water-soluble proteins (Ptitsyn et al 1990). Evidence is accumulating that points towards an obligatory role of molten globules in the insertion and folding of proteins into

208 Chapter 7 membranes. Thus, studies on pore-forming toxins colicins (van der Goot et al 1991, Parker & Pattus 1993, Evans et al 1996) and diphtheria (Paliwal & London 1996) indicate that their insertion into model membranes occurs via molten globules. Folding of bacterial porins OmpA and Omp F is also thought to proceed through molten globule intermediates (Surrey & Jahnig 1995, Surrey et al 1996, Kleinschmidt & Tamm 1996). The fact that varying retinal concentration has virtually no observable effect on the fluorescence changes associated with the phases n1 and n2 implies that retinal does not bind to the apoprotein intermediates I1 or I2, as envisaged in previous chapters. This finding further supports the notion that these intermediates may represent off-pathway products in the folding of bacteriorhodopsin. Preliminary experiments indicate that addition of retinal in a large excess over bacterioopsin effectively removes the fluorescence changes associated with the phases n1 and n2, observed under limiting and stoichiometric concentrations of retinal. Instead these phases are accompanied by a decrease in fluorescence and their rate constants remarkably fall close to the values obtained from absorbance measurements, which solely monitor micelle mixing (see Chapter 4). This observation thus lends support to the view that the phases n1 and n2 may solely be a consequence of micelle mixing. Addition of retinal to bO pre-equilibrated with the micelles has allowed the steps associated with retinal binding to be more closely examined (Fig 7.2). The fact that only the phases n5 and n6 have been observed to be dependent on retinal concentration demonstrates unequivocally that there are two bimolecular reactions involving retinal and bacterioopsin. In the presence of excess retinal, the second- order reaction between retinal and bacterioopsin can be described by first-order kinetics, that is, an exponential. Hence the fact that two exponentials are dependent upon retinal concentration can only mean that there are two independent retinal binding steps. Thus as suspected in previous chapters, retinal does indeed bind in a parallel fashion to the apoprotein resulting in the formation of the 430-nm intermediate. A preliminary account of this work has been given previously (Booth et al 1996). In this report, only one second-order reaction involving retinal and bacterioopsin was detected. The discrepancy between the results presented here and those reported earlier is explained by the fact that kinetic analysis is not only performed upon a higher density of data collected here (at least ten-fold higher) but also because analysis is

Dependence of Refolding Kinetics of Bacteriorhodopsin on Retinal Concentration 209 independent of the time scales used between various retinal concentrations. In contrast, different time scales were used to obtain the rate constant of the fastest phase for various retinal concentrations in the previous report. The procedure used in the previous study to measure the various rate constants was thus less accurate.

The slower phases n7, n8 and n9, assigned to the decay of the 430-nm intermediate into bacteriorhodopsin (see Chapter 5), have been shown to be independent of retinal concentration. This indicates that the formation of bacteriorhodopsin must proceed through the 430-nm intermediate. If these phases involved the decay of retinal and bacterioopsin directly into bacteriorhodopsin, then they would be expected to be dependent upon retinal concentration. Thus although the lack of a single isobestic point, in the time-resolved absorption spectra (see Chapter 5), could not rule out the possibility that some fraction of bacterioorhodopsin may also form directly form bacterioopsin and retinal the above results unequivocally demonstrate that the role of the 430-nm intermediate in the folding of bacteriorhodopsin is absolutely necessary. Taken together, the above results are only consistent with the model

I1 I3 k k 3 k k8 1 5 k bO R 7 I430 bR k 2 k 4 I2 I4 k 6 k9

invoked in the previous chapters in which the requirement of the 430-nm intermediate in bacteriorhodopsin refolding is absolute. The apoprotein intermediates I3 and I4 are structurally similar but kinetically distinct species. The intermediates I1 and I2 most likely denote misfolded products, which very probably have only a transient existence. Both the 430-nm intermediate and the native state bacteriorhodopsin are likely to represent an ensemble of structures. Free energy changes associated with the formation of the 430-nm intermediate have also been determined from the kinetic measurements reported here (Table 7.1).

210 Chapter 7

The only assumption made in these calculations is that the retinal-binding steps (phases n5 and n6) are in quasi-equilibrium. This is indeed likely to be the case as the state of bacterioopsin with a native-like secondary structure (bO¢), to which retinal apparently binds, reaches equilibrium within about two minutes after mixing bO/SDS with DMPC/CHAPS (see Chapter 6). Retinal here is added 30 minutes after this mixing. Furthermore the depopulation of the 430-nm intermediate is likely to be negligible - the formation of the 430-nm intermediate is at least an order of magnitude greater than its subsequent decay into bacteriorhodopsin. Both the fluorescence and absorbance measurements presented here agree on a value of around - 30 kJmol-1 for the non-covalent binding of retinal to bacterioopsin. This value is similar in magnitude to the hydrolysis of ATP, which acts as a universal energy currency in the body, that drives many biological processes which are otherwise thermodynamically unfavourable. The binding of a Ca2+ to the small globular protein parvalbumin (Sudhakar et al 1995), the unfolding of the integral outer membrane protein FepA (Klug et al 1995) and the insertion of M23 procoat protein (Soekarjo et al 1996) into lipid bilayers also result in a free energy change of similar magnitude. This value is also similar in magnitude to the free energy of stabilization of several water-soluble proteins relative to their unfolded states (Pace & Vanderberg 1979, Ahmad & Bigelow 1982, Serrano et al 1990, Yao & Bolen 1995). An enthalpy change of about - 400 kJmol-1 has been reported for the binding of retinal to the apomembrane (purple membrane minus retinal) (Kahn et al 1992). Remarkably, the measured enthalpy change for the binding of retinal to bacterioopsin in mixed phospholipid/detergent micelles is similar (Brouillette et al 1987, Brouillette et al 1989). Micellar bacteriorhodopsin however is less stable than the purple membrane (Brouillette et al 1989). The decreased stability of monomeric micellar bacteriorhodopsin is thought to result from the loss of protein-protein interactions, which predominate within the crystalline lattice of purple membrane. The entropic contribution of the crystalline lattice to the free energy of stabilization of bacteriorhodopsin in the purple membrane is thought to be about 20 kJmol-1K-1 (Haltia & Freire 1995). That is, the free energy of stabilization of bacteriorhodopsin is mainly of enthalpic origin. Thus assuming a negligible or favourable entropy change, the enthalpy change of - 400 kJmol-1 represents a free

Dependence of Refolding Kinetics of Bacteriorhodopsin on Retinal Concentration 211 energy change for the refolding of bacteriorhodopsin from the apomembrane and retinal. It therefore seems likely that retinal confers further stabilization on bacteriorhodopsin following its non-covalent binding to the apoprotein. Thus although retinal may not be involved in initiating bacterioopsin refolding, it clearly drives the refolding reaction of bacteriorhodopsin to completion. The interhelical interactions are thought to be the dominant driving force in the packing of transmembrane helices (Popot & Engelman 1990, Popot 1993, Lemmon & Engelman 1994). These may include electrostatic or van der Waals forces. Hydrogen bonding between polar residues or formation of salt bridges between charged residues on adjacent helices could, for example, drive helix association within the membrane (Engelman 1982, Honig & Hubbel 1984). Indeed, several polar residues are present in the transmembrane helices of bacteriorhodopsin (Henderson et al 1990, Grigorieff et al 1996) and the photosynthetic reaction centres (Diesenhofer et al 1985, Allen et al 1987, Komiya et al 1988), and a salt bridge between Glu and Arg on adjacent helices is observed in the atomic model of plant light- harvesting antenna complex II (Kuhlbrandt et al 1994). Salt bridges are also thought to be involved in the assembly of T-cell receptor complex between the T-cell receptor a and CD3d (Manolios et al 1990, Cosson et al 1991). Mutagenesis studies indicate that specific amino acid residues are critical in the assembly of the major histocompatibility complex II (Cosson & Bonifacino 1992), in the dimerization of the single-transmembrane protein glycophorin (Lemmon et al 1992) and the M13 procoat protein (Deber et al 1993). The reasons that make interhelical interactions a dominant force in the close packing of transmembrane helices relative to helices in water are largely thermodynamic ones. Thus, whereas hydrogen bonding and salt bridges in water are only marginally stable the same can not be argued for them in the hydrophobic milieu of the membrane. This is because charged or polar groups have the option of hydrogen bonding with water molecules which is clearly not available in the membrane. Thus the formation of interhelical salt bridges and hydrogen bonds could in theory drive the association of helices in the membrane. A priori approaches indicate that the free energy penalty of leaving a residue-sized cavity within the membrane is about 10 kJmol-1 as a result of unsatisfied van der Walls interactions and vacuum in the cavity (Popot & Engelman 1990). The close

212 Chapter 7 packing of helices so as to minimize any cavities could contribute to free energy of stabilization of tens of kilojoules per mole of each helix. The amino acid residues within the center of the transmembrane regions of the photosynthetic reaction center are as tightly packed as in the centers of water-soluble proteins (Yeates et al 1987). Side-by-side interactions between helices appear to be a hallmark in the stabilization of membrane proteins. Other factors such as ligand binding, folding of interhelical loops and bilayer rearrangement are thought to contribute little to the close packing of transmembrane helices. Several lines of evidence suggest that the cleavage of a polypeptide chain into two or more fragments corresponding to transmembrane regions has no observable effect on the tertiary organization of many polytopic membrane proteins within the lipid bilayer. Bacteriorhodopsin can be refolded to a native-like chromophore from two chymotryptic fragments corresponding to helices A-B and C-G in mixed phospholipid/detergent micelles (Liao et al 1983, Popot et al 1987). The product is active in light-driven proton pumping and its moderately high resolution X-ray structure is indistinguishable from that of intact bacteriorhodopsin and the purple membrane (Popot et al 1986). Refolding of bacteriorhodopsin from fragments corresponding to helices A-E and F-G as well as from helices A-E and C-G has been achieved (Liao et al 1984, Kataoka et al 1992). Further, bacteriorhodopsin can be refolded from the chymotryptic five-helix fragment (helices C-G) and chemically synthseized peptides corresponding to helices A and B (Kahn & Engelman 1992). Although bacteriorhodopsin cannot be refolded from fragments corresponding to individual helices, peptides corresponding to five (A, B, C, D and E) of the seven helices have been shown to be capable of independently forming transmembrane helices in phospholipid bilayers (Hunt et al 1993). Why peptides corresponding to helices F and G cannot form transmembrane helices in isolation is not clear. Several other membrane proteins including lactose permease (Wrubel et al 1990, Bibi & Kaback 1990, Zen et al 1994, Sahin-Toth et al 1996), the b-adrenergic receptor (Kobilka et al 1988), a voltage- dependent Na+ channel (Stuhmer et al 1989), the yeast a-factor transporter STE6 (Berkower & Michaelis 1991) and adenylate cyclase (Tang et al 1991) can also be refolded from their fragments corresponding to transmembrane

Dependence of Refolding Kinetics of Bacteriorhodopsin on Retinal Concentration 213 regions. Mammalian visual rhodopsin can be proteolytically cleaved in its loops to yield three fragments that remain associated (Litman 1979). Thus, in view of the above considerations, it is conceivable that the retinal chromophore may not be directly involved in the close packing of transmembrane helices. The intriguing possibility that retinal may cause further condensation of the apoprotein indirectly however cannot be excluded. Thus retinal binding could for example induce the folding of the extramembraneous loops. Indeed, theoretical calculations indicate that unfolding of the interhelical loops of bacteriorhodopsin could result in an enthalpy change of a magnitude close to that observed for the binding of retinal to apomembrane (Haltia & Freire 1995). The notion that the enthalpy change associated with retinal binding results largely from the folding of interhelical loops gains further credibility in light of the observation that denaturation of bacteriorhodopsin in DMPC/CHAPS micelles gives rise to an enthalpy change of a similar magnitude (Brouillette et al 1987). The denaturation of membrane proteins results in the unfolding of extramembraneous parts with the transmembrane portions remaining largely unperturbed. In an analogy with water-soluble proteins, the interaction of a ligand molecule with the apoprotein induces its folding to a native conformation. Thus, Ca2+ is required for the folding of prothrombin (Nelsestuen & Lim 1977), parvalbumin (Kuwajima et al 1988) and a-lactalbumin (Kuwajima et al 1989). Similarly, haem is required for the folding of cytochrome c (Stellwagen et al 1972, Fisher et al 1973) and myoglobin (Hargrave et al 1994, Chiba et al 1994). The extramembraneous loops in many ways resemble the characteristics of water-soluble proteins. In summary, the results of this chapter are in support of the view that the binding of retinal occurs after a rate-limiting step in the refolding of the apoprotein bacterioopsin. Two bimolecular reactions involving retinal and bacterioopsin are observed, confirming the proposal that retinal binds in a parallel fasion to the apoprotein intermediates I3 and I4. The product of these two parallel reactions is the 430-nm intermediate, for which a free energy change of about - 30 kJmol-1 is determined. The 430- nm intermediate appears to be obligatory in the reaction of retinal with bacterioopsin to form bacteriorhodopsin.

Chapter 8

CONCLUSIONS

8.1 How Does Bacteriorhodopsin Fold? 216

8.2 Implications of This Study on Protein Folding 226

8.3 Closing Remarks 230

216 Chapter 8

8.1 How Does Bacteriorhodopsin Fold?

Kinetics is the study of the rates at which the various species evolve and decay in a process. The study of kinetics is absolutely pivotal to understanding the mechanism of action of a protein or any other molecule for that matter. Many investigators are led by the conviction that those who pursue kinetics may be wasting their time in that once the structure of a protein is solved to atomic level its mechanism of action will ensue. This has proved to be largely incorrect for the atomic structures of hundreds of proteins are known and yet little is known about their mechanism of action. A knowledge of the structure of course assists in unravelling the mechanism of action of a protein. Structure per se however does not hold all the clues to the dynamics of a process. The study of kinetics sets out to obtain information not only on the ground states (the initial and the final species, or in protein folding terminology, the unfolded and the native species) but any intermediate states that may also be involved in a reaction. The transient nature of the intermediate states implies that kinetics is the only method available to study them. Kinetics is by no means a perfect discipline per se. It afterall tells you nothing whatsoever about the structural transitions occuring during a reaction. Further in the case of protein folding studies, kinetics is very limited in that it provides you with non-specific information. That is, the kinetically resolved species in a protein folding reaction do not correspond to any realistic species but rather represent an average of all the microscopic states. Thus, when one talks about a kinetic species in a protein folding reaction it is implied that it consists of an ensemble of substates, albeit structurally similar. Taken together, the results presented in this thesis support a model in which intermediates unequivocally play a vital role in the folding of bacteriorhodopsin (Fig 8.1). The whole process of bacteriorhodopsin folding can be dissected into two parts: The first part involves the formation of an apoprotein state with a native-like secondary structure in the micelles (bO¢) from a denatured state in SDS (bO) and, in the second part, retinal binds to this state to regenerate bacteriorhodopsin (bR). The state bO¢ probably contains closely-packed transmembrane helices stabilized by interhelical interactions. If this is so then this state exhibits all the characteristics of a molten globule intermediate that has been widely observed in the folding of water-soluble proteins (Ptitsyn et al 1990). The formation of the state bO¢ is essentially a two-state

Conclusions 217

100 ms 100 s

I3 5 s

5 s bO retinal micelles I430 bR 20 s

500 s 500 ms

I4

apoprotein apoprotein 430-nm native protein (partially intermediates intermediate (retinal denatured (fully formed (retinal covalently in SDS) native secondary non-covalently bound to the structure + bound to the apoprotein + loops largely apoprotein + loops with unfolded) loops largely native b unfolded) conformation)

Fig 8.1. A paradigm for bacteriorhodopsin folding. Only on-pathway intermediates are shown. The off- pathway species I1 and I2 are omitted for clarity. Note that, in this scheme, folding does not initiate from a random coil but rather from a state that apparently retains at least half the helical content of the native protein. This partially denatured state may be an on-pathway intermediate in the folding of bacteriorhodopsin from a fully unfolded polypeptide chain.

218 Chapter 8 process occuring via alternative parallel pathways. That is, intermediates are very probably not necessary for the formation of the state bO¢ that consists of at least two kinetically distinct but structurally similar species (I3 and I4). These species represent on-pathway intermediates in the folding of bacteriorhodopsin. One possible cause of the formation of the state bO¢ via parallel pathways is attributed to the slow cis-trans isomerization about proline imide bonds (Brandts et al 1975). Thus those molecules in the denaured state in SDS with all native proline isomers will refold faster than those which contain some or all non-native isomers. Off-pathway reactions are also envisaged during the formation of the state bO¢ (species I1 and I2). It is believed that the fraction of the protein that follows such behaviour is very small and is in equilibrium with the denatured state bO such that in the presence of retinal it may be persuaded to take an on-pathway route. The fact that no intermediates are thought to be involved in the formation of the state bO¢ does not in any way challenge the two-stage model proposed for membrane protein folding (Popot & Engelman 1990). The two-stage model is based upon the conceptual folding of a random coil in a biological membrane, whereas folding of the state bO¢ here is carried out from a partially denatured state bO that apparently contains secondary structure equivalent to almost four transmembrane helices in SDS (London & Khorana 1982, Riley et al 1997). This partially denatured state may be an intermediate per se between the fully unfolded apoprotein and the state bO¢. That is, the first stage of the two-stage model is not under study here.

The binding of retinal to the state bO¢ (apoprotein intermediates I3 and I4) occurs in a parallel fashion and results in the formation of a spectroscopically distinct intermediate. Retinal is believed to be non-covalently bound to the apoprotein within this temporal intermediate that absorbs maximally around 430 nm. The formation of this 430-nm intermediate is energetically a favourable process with a concomitant release of free energy of tens of kilojoules. The next step in the folding reaction of bacteriorhodopsin involves the decay of the 430-nm intermediate into bacteriorhodopsin via three kinetically distinct pathways, with lifetimes ranging from several seconds to several minutes. It is proposed that the formation of bacteriorhodopsin from the 430-nm intermediate via the fastest route is kinetically controlled whilst the slower pathways are under thermodynamic control. It is therefore conceivable that the kinetic species

Conclusions 219 interconverts to the thermodynamic species during the course of the reaction. This could thus ensure that all of the native protein lies at a global free energy minimum. This would satisfy the thermodynamic requirement of the protein folding process. Alternatively, the fraction of the 430-nm intermediate that decays into bacteriorhodopsin within seconds may contain retinal in a native configuration whereas the remainder contains retinal in some sort of non-native configuration. The decay of the 430-nm intermediate into bacteriorhodopsin over a time scale of minutes may thus be explained by the slow rearrangement of the retinal chromophore to a native configuration. The formation of bacteriorhodopsin from the 430-nm intermediate is characterized by the formation of a Schiff base between the aldehyde group of retinal and the e-amino group of Lys216. The model of bacteriorhodopsin folding presented above is purely based on kinetic grounds and in no way does it have any implications on the thermodynamics of this process. In this model, it is conjectured that the transmembrane topology of the state bO¢ is almost indistinguishable from that of bacteriorhodopsin. This prediction partly results from the painstaking work of many investigators who have widely demonstrated that transmembrane helices are capable of docking per se via interhelical interactions and that a ligand molecule probably plays no active role in their tertiary organization (Popot & Engelman 1990, Popot 1993, Lemmon & Engelman 1994). It is of course possible that the binding of retinal to the state bO¢ results in rearrangement of helices, but it is equally conceivable that retinal binds to a pre-formed pocket within this state without any noticable rearangements. The major structural difference between the state bO¢ and bacteriorhodopsin is thought to lie in the conformation of the extramembraneous loops. These loops are envisaged to be largely unfolded in the state bO¢ prior to the binding of retinal and remain so in the 430-nm intermediate. Only upon the decay of the 430-nm intermediate into bacteriorhodopsin do such regions take up the native conformation. That is, the role that the retinal chromophore plays in bacteriorhodopsin assembly is largely concerned with triggering the folding of interhelical loops. The extramembraneous loops are likely to be involved in the functional activity of a protein residing in the hydrophobic milieu of the membrane. Indeed, such regions have been implicated in the functional activity of the mammalian visual receptor rhodopsin (Palczewski et al 1991, Thurmond et al 1997,

220 Chapter 8

Unger et al 1997). Further, recent high resolution structural studies on bacteriorhodopsin emphasize the importance of residues within the loop regions in its proton pumping activity (Grigorieff et al 1996, Kimura et al 1997). Thus although a ligand molecule may not have much control over the folding of transmembrane regions of a protein, its interaction may be important in ensuring that the extramembraneous regions also fold correctly for the protein as a whole may not become functional without them. Evidence that some cellular proteins may be disordered in the absence of a ligand is accumulating. Thus, for example, the cell cycle regulator p21 has been shown to lack stable secondary or tertiary structure in the free solution state but adopts an ordered stable conformation when bound to its natural ligand cyclin-dependent kinase (Kriwacki et al 1996). In an analogy with these observations, it is suggested that the interhelical loops of the partially refolded apoprotein bacterioospin are highly disordered with no specific secondary or tertiary structure in the absence of retinal but assume a specific ordered conformation upon its binding. In this way, the interhelical loops probably resemble the p21 regulator in experiencing a disorder-order transition in order to be functionally active. A quantitative description of the energetics of bacteriorhodopsin folding is not possible at this stage. However one can speculate about the probable shapes of free energy profile, energy landscape and the folding funnel of bacteriorhodopsin (Fig 8.2). The free energy profile describes the change in the free energy during the course of the folding reaction of bacteriorhodopsin. It is anticipated that the apoprotein intermediates I3 and I4 are more stable than the denatured state bO. The transfer of an hydrophobic stretch of 20-30 amino acids from an aqueous to an apolar environment is expected a priori to be concomitant with free energy release of hundreds of kJ per mole (Engelman & Steitz 1981,

Lemmon & Engelman 1994). Although formation of the state bO¢ (intermediates I3 and I4) is far from this ideal situation, it is likely that this process is energetically favourable. The knowledge that the denatured state bO can refold spontaneously to the state bO¢ with a native-like secondary structure in the absence of retinal further corroborates this belief. It is also envisaged that the 430-nm intermediate is more stable than the apoprotein intermediates I3 and I4 for its formation is observed here to be energetically

Conclusions 221

bO I3/I4 I430 Free energy

bR

Reaction coordinate

bO

I3/I4

I3/I4 I430

Free energy I430

bR

Conformational entropy

bO

I3/I4

I3/I4 I430

I430

bR

Fig 8.2. Energetics of bacteriorhodopsin folding. Hypothetical diagrams showing free energy profile (top), energy landscape (middle) and folding funnel (bottom) of bacteriorhodopsin.

222 Chapter 8 favourable. The final native protein must of course lie at the global free energy minimum in order to fulfil the thermodynamic requirements of the protein folding process. The realistic energy profile would be multidimensional with a large number of minor bumps superimposed on the basic profile (Fersht 1995b, Fersht 1997). The energy landscape depicts the free energy as a function of conformational entropy of bacteriorhodopsin during its folding. As the protein refolds from the denatured state to a fully folded conformation, it is likely to become more compact. That is, its conformational entropy will decrease. The global energy minimum on this diagram thus corresponds to the native state bacteriorhodopsin, while the local energy minima correspond to the folding intermediates. A multidimensional representation of the energy landscape is embodied in the folding funnel illustrated for bacteriorhodopsin (Wolynes et al 1995, Bryngelson et al 1995, Onuchic et al 1995). This is the first report that invokes the folding of bacteriorhodopsin via parallel alternative branched pathways. Parallel pathways are also thought to operate in the folding of water-soluble proteins lysozyme (Radford et al 1992, Kiefhaber 1995, Matagne et al 1997), RNAse A (Udgaonkar & Baldwin 1990), cytochrome c (Elove et al 1994) and DHFR (Jennings et al 1993, Jones & Matthews 1995). Sequential pathways have been suggested for bacteriorhodopsin folding in all the previous studies reported (Oesterhelt et al 1973, Oesterhelt & Schuhmann 1974, Schreckenbach et al 1977, Schreckenbach et al 1978a, London & Khorana 1982, Booth et al 1995, Booth et al 1996), although possibilities for parallel branched pathways were reserved. This report is thus not in conflict with any of the previous studies. Sequential pathways are also invoked for the three other integral membrane proteins, namely the bacterial porins OmpA (Surrey & Jahnig 1995, Kleinschmidt & Tamm 1996) and OmpF (Surrey et al 1996), and the plant-light-harvesting complex II (Booth & Paulsen 1996), for which the folding kinetics have been studied. Table 8.1 compares the nomenclature used for the folding intermediates of bacteriorhodopsin in the previous reports and this thesis. Notably, only one spectroscopically distinct intermediate is observed here as opposed to two suggested by Oesterhelt and coworkers (Schreckenbach et al 1977). On the basis of kinetic fluorescence refolding experiments, an intermediate was also suggested to be involved

Conclusions 223

Table 8.1. Folding intermediates of bacteriorhodopsin: Comparison of the various nomenclatures used in the previous studies and this thesis.

Nomenclature This Booth Booth London & Schreckenbach thesis et al et al Khorana et al 1977 Assignment 1996 1995 1982 (1) Micelle mixing

I * * * * 1 (3) Protein misfolding

(2) Formation of transmembrane helices I I I * * within the core regions 2 1 1

(1) Formation of a state of bacterioopsin with a I native-like secondary 3 structure

I2 Io bO¢ apomembrane (2) Helix rearrangement and I association 4 (3) Helix formation at the bilayer interfacial regions

400-nm (1) Non-covalent chromophore binding of retinal

I430 IR IR * (2) Ring-chain 430/460-nm planarization of retinal chromophore (Schreckenbach et al 1977)

Retinal-protein intermediate preceding * I * * * the formation of 3 bacteriorhodopsin but following the formation of the 430-nm intermediate

* NOT applicable.

224 Chapter 8 in the decay of the 430-nm intermediate into bacteriorhodopsin (Booth et al 1996). This study however demonstrates unequivocally that this is not so. This section would be incomplete without a reference to the situation in vivo. Bacteriorhodopsin in the purple membrane is organized into a two-dimensional hexagonal lattice, with each unit of the lattice comprising a trimer of bacteriorhodopsin (Blaurock 1982, Mukhopadhyay et al 1996). In contrast, bacteriorhodopsin within the mixed phospholipid/detergent micelles exists as a monomer (Brouillette et al 1989). The role of protein-protein interactions in the folding of bacteriorhodopsin therefore cannot be assessed at this stage. Protein-protein interactions are responsible for the pronounced stability of bacteriorhodopsin in the purple membrane relative to that in the micelles (Brouillette et al 1989, Kahn et al 1992). The purple membrane and the micelles however share a common molecular architecture (Mazer et al 1980). Both contain amphipathic molecules exquisitely arranged in a bilayer fashion surrounded by an aqueous environment. This resemblance in molecular architecture, with similar physico-chemical properties, is reflected in the appearance of a native-like chromophore absorption band upon the regeneration of bacteriorhodopsin in the micelles. Further, the regenerated bacteriorhodopsin in the micelles exhibits all the characteristic photoreactions of the proton pumping cycle of native bacteriorhodopsin (Braiman et al 1987). Of course the proton pumping activity of bacteriorhodopsin cannot be observed in the micelles. Unlike the native purple membrane, the micelles are not an enclosed structure. That is, there is no inside or outside. The proton pumping activity of bacteriorhodopsin is however restored upon the removal of the detergent from the mixed phospholipid/detergent micelles (Huang et al 1981, Liao et al 1983, Braiman et al 1987). This procedure results in the formation of vesicles or liposomes in which the orientation of bacteriorhodopsin is reversed relative to that in the native membrane. There is thus little reason to doubt the applicability of principles of bacteriorhodopsin folding established in vitro to folding in vivo. There are of course likely to be some subtle differences in bacteriorhodopsin folding in vivo compared to the situation in vitro. Thus bacteriorhodopsin is synthesized in vivo as a precursor with an N-terminal leader sequence of 13 amino acids (Seehra & Khorana 1984). This so-

Conclusions 225 called signal peptide undoubtedly plays a critical role in targeting the insertion of the protein into the cell membrane of halobacteria. There is also an Asp residue at the C-terminus of bacteriorhodopsin precursor. What role this residue plays in bacteriorhodopsin folding in vivo remains obscure. Both the signal peptide and the Asp residue are removed following the insertion of bacteriorhodopsin into the membrane. Further the possibility that chaperones may assist bacteriorhodopsin folding in vivo cannot be ruled out. The point being made here is that the basic rules governing protein folding are likely to be the same whether it is taking place inside or outside of the cell. It is afterall the primary sequence of a protein that ultimately determines its 3D shape. It is this argument that makes the model of bacteriorhodopsin folding presented above absolutely relevant to a comprehensive understanding of the mechanism of bacteriorhodopsin folding in vivo.

226 Chapter 8

8.2 Implications of This Study on Protein Folding

Membrane protein folding was almost a forgotten ground, from a kinetics point of view, at the time that this research was initiated. The ab initio progress was hampered by the fact that many well-established biophysical techniques could not be applied to bacteriorhodopsin for it belongs to a hydrophobic class of proteins that have a world of their own. Methods available to gain a better understanding of such proteins remain indirect and superficial, especially from a folding point of view. Despite these technical difficulties, the work presented here has profound implications on the mechanisms of protein folding. The results of this study provide overwhelming support for theoretical models that describe the process of protein folding occuring via multiple parallel routes instead of being confined to a well defined pathway (Baldwin 1994, Baldwin 1995, Dill & Chan 1997). The latter view was put forward by Cyrus Levinthal (Levinthal 1968) - one of the well-respected theoreticians of his time. The field of protein folding has come a long way since its launch as a bona fide research pursuit in the 1960s (Anfinson et al 1961). Specific intermediates were thought to be an essential part of the folding process in the early days (Wetlaufer 1973, Dill 1985, Gregoret & Cohen 1991, Karplus & Weaver 1994). More recently however the role of intermediates in productive folding has been questioned and, indeed, many investigators have suggested that such species merely represent off-pathway species and kinetic traps that slow down rather than accelerate folding (Creighton 1994, Sosnik et al 1994, Schindler et al 1995, Weissman 1995). As a rule of thumb, the most important intermediates in protein folding reactions are those that are unlikely to accumulate to a significant extent and thus may go undetected. Studies on bacteriorhodopsin folding indicate that although intermediates may be responsible for protein misfolding and off-pathway reactions, their role in productive folding is a critical one. Thus the results presented here unequivocally support the hypothesis that all molecules must pass through the 430-nm intermediate on their way to the native chromophore. Two of the kinetic criteria for the involvement of an on-pathway intermediate in a folding reaction are the presence of a lag period in the formation of the native protein and the lack of a single isobestic point in the time-resolved

Conclusions 227 spectra. Both of these conditions are met here regarding the 430-nm intermediate. The interesting feature of this study, however, is that intermediates may not be universal to the process of protein folding. The view that intermediates are not involved in protein folding gains no support whatsoever from this study. The complexity of protein folding reactions intuitively suggests that a large diversity of mechanisms is likely to be in operation and that there may be no common theme among different systems. It would be reasonable to question the applicability of the studies on a membrane protein to water-soluble systems. In this regard, the converse would also be true. That is, can one question the applicability of folding studies on water-soluble proteins to their membrane counterparts. With this consideration in mind, one may therefore conclude that although there is a controversy over as to whether there is an intrinsic role of intermediates in the folding of water-soluble proteins, such species may have an absolute requirement in the case of membrane proteins. This argument brings out a classical demonstration of synergy between experiment and theory. The two-stage model of membrane protein folding inherently invokes the role of intermediates (Popot & Engelman 1990), and all the experimental evidence available to date points towards such species being critical in the folding of membrane proteins. Folding kinetics of the only three other integral membrane proteins studied so far provide compelling evidence in support of this model. These include plant light- harvesting complex II (Booth et al 1996), and the bacterial porins OmpA (Surrey & Jahnig 1995, Kleinschmidt & Tamm 1996) and OmpF (Surrey et al 1996). The proposal that intermediates are likely to be necessary in the folding of membrane proteins may be a consequence of the constraints imposed upon the protein by the lipid bilayer (von Heijne & Manoil 1990, Donnelly et al 1993). Another important lesson that can be learnt from this study is that although the role of intermediates in productive folding of bacteriorhodopsin is unequivocal, there appears to be no specific pathway that the unfolded molecules follow in their quest to attain a native conformation. The parallel nature of the pathways observed here reflects the fact that bacteriorhodopsin folding can be essentially considered as a distribution of structures from the unfolded to the native state. The foregoing argument may seem to contradict the role of intermediates in bacteriorhodopsin folding. The point being made

228 Chapter 8 however is that the intermediates are no well-defined species but represent an ensemble of structures. These observations fit in nicely with the new view of protein folding (Baldwin 1994, Baldwin 1995, Dill & Chan 1997). Although intermediates have been implicated in the folding of membrane proteins previously (Surrey & Jahnig 1995, Kleinschmidt & Tamm 1996, Surrey et al 1996, Booth et al 1996), this study demonstrates for the first time the involvement of parallel pathways in membrane protein folding. Interestingly, multiple pathways are also thought to operate in the proton transfer by bacteriorhodopsin across the cell membrane of halobacteria (Unger et al 1997). Multiple pathways, routes or mechanisms must be a rule rather than an exception in biological phenomena. Why should nature restrict itself to a single option when multiple options can be both kinetically and thermodynamically feasible. Nature favours rather than restricts diversity. The elucidation of the mechanisms of protein folding remains one of the gigantic and mammoth challenges in modern biology. The widespread interest in this area does not solely arise from the need to satisfy the curiosity of academics but more so from the recognized impact that this pursuit might have on medicine and biotechnology. Protein folding in this respect provides an ideal example of pure research directly linked to application. There is a wide acceptance of the idea that membrane proteins may beat their water-soluble counterparts to the challenge of providing the folding code. This argument results from the knowledge that constraints imposed upon proteins within biological membranes reduce the conformational space available to them by many orders of magnitude relative to soluble proteins (von Heijne & Manoil 1990, Donnelly et al 1993). In theory, therefore, it should be an easier task to build up a sequence of events that lead to folding of proteins within biological membranes rather than in aqueous environment. Although theoretical studies of protein folding have had some remarkable success in recent years (Karplus & Shakhnovich 1992, Sali et al 1994a, Sali et al 1994b, Karplus & Weaver 1994, Gutin et al 1995), their potential towards folding of proteins in biological membranes remains yet to be realized. This point deserves some extended consideration. One of the problems with Monte Carlo simulations of protein folding is that they often ignore complexities associated with real proteins. This is because the

Conclusions 229 computer time required to carry out a simulation incorporating all the possible interactions would be so lengthy that it would make the whole process unpractical. As a consequence, theoretical studies of protein folding only provide a simplified picture of the mechanisms of the folding of real proteins. For the folding of proteins in biological membranes, however, not all possible interactions may need to be taken into account. Thus computer simulations of membrane protein folding may provide a picture that is close to realism. This should not be surprising in view of the fact that substantial breakthroughs have already been witnessed in the development of theoretical approaches for de novo design (Whitley et al 1994) and three-dimensional structure prediction (Henderson 1995) of integral membrane proteins. The intriguing implication of this realization is that if folding of proteins within biological membranes could be understood, the mechanisms of protein folding in an aqueous environment may follow. Thus although investigations into membrane protein folding did not take off almost three decades after research on folding of water-soluble proteins began, the sequence in which the folding mechanisms of soluble and membrane proteins are understood may be reversed. It is with respect to this latter view that the research on the folding of a membrane protein presented here bears a special importance. In brief, the research presented here supports the credence that folding occurs via funnelling rather than tunnelling (Baldwin 1995, Dill & Chan 1997). That is, there is no specific pathway that links unfolded and native molecules but rather an infinite options exist for unfolded molecules to attain a native conformation, at least initially. These various multiple parallel pathways may only converge at a late stage in folding.

230 Chapter 8

8.3 Closing Remarks

This study demonstrates that the experimental investigation of the folding kinetics of membrane proteins is feasible. More specifically, the research presented here supports a number of views previously held regarding bacteriorhodopsin folding. These include

(1) Formation of secondary structure elements in the apoprotein bacterioopsin precedes tertiary interactions (Popot & Engelman 1990, Popot 1993, Lemmon & Engelman 1994). (2) Retinal does not initiate the folding of the apoprotein bacterioopsin (London & Khorana 1982, Booth et al 1995). (3) The folding of the apoprotein bacterioopsin is rate-limiting for the subsequent binding of retinal (London & Khorana 1982, Booth et al 1995). (4) Retinal binds to a state of the apoprotein bacterioopsin with a native-like secondary structure (London & Khorana 1982, Booth et al 1995). (5) Spectroscopically distinct intermediates may be involved in bacteriorhodopsin folding (Oesterhelt et al 1973, Oesterhelt & Schuhmann 1974, Schreckenbach et al 1977, Schreckenbach et al 1978a).

The research presented in this thesis also invokes several novel ideas regarding bacteriorhodopsin folding. These include

(1) Multiple parallel branched pathways may be prevalent in the folding of the apoprotein bacterioopsin to a state with a native-like secondary structure. (2) The binding of the retinal chromophore to the apoprotein may occur in a parallel fashion. That is, there may be several bimolecular reactions involving retinal and bacterioopsin. (3) The formation of bacteriorhodopsin from retinal and the apoprotein occurs via a spectroscopically distinct intermediate. This intermediate absorbs maximally around 430 nm. (4) The decay of the 430-nm intermediate into bacteriorhodopsin proceeds through parallel alternative branched pathways. (5) Thermodynamic measurements on the folding intermediates of bacteriorhodopsin are feasible.

Conclusions 231

This study should form the basis of what may hopefully turn out to be a one of the better understood areas of protein biochemistry in the years and decades to come. There are several ideas which cannot be tested experimentally at this stage because of technical difficulties associated with bacteriorhodopsin and other membrane proteins. Thus unlike water-soluble proteins, there are virtually no methods available to obtain high resolution structural information on the folding intermediates of membrane proteins. However a number of experimental ideas have been developed over the last few years which are worth putting to the test in the near future. One of the speculations surrounding this study is that cis- trans proline isomerization may be responsible for the slow refolding kinetics of the apoprotein bacterioopsin. This idea could be tested using a number of techniques. These include site-directed mutagensis, addition of peptidyl prolyl isomerase and double-jump assays. Site-directed mutagenesis invloves removal of the suspected prolines, in this case the three transmembrane residues. One can then compare the refolding kinetics of the wild type and the mutant protein. What effect the replacement of a proline residue by another amino acid residue might have on the protein structure will need to be addressed. Peptidyl prolyl isomerase catalyzes proline isomerization and thus addition of this enzyme should accelerate the apoprotein refolding kinetics. Whether this enzyme will retain its activity in the hydrophobic milieu of the micelles is not known. Double-jump assays involve unfolding of a protein rapidly using a denaturant and refolding is initiated before the prolines have had a chance to isomerize to non-native isomers. Helical integral membrane proteins tend to be resistant to classical denaturants such as guanidinium hydrochloride and urea. This is especially true in the case of bacteriorhodopsin. Whether a suitable denaturant can be found for bacteriorhodopsin will be interesting. Ab initio attempts to denature the protein rapidly with SDS have proved futile. It would also be interesting to determine the nature of the native and non-native proline isomers in bacteriorhodopsin. That is, do the proline residues in the native protein have cis or trans or a mixture of both configurations. One of the techniques that has proved overwhelming in this investigation is the 2D NOESY NMR spectroscopy. This method involves detection of cross-peaks, arising from nuclear overhauser effects

232 Chapter 8

(spatial interactions between two protons which must be within 5 Å of each other), between the d- proton of proline and a-proton of the preceding residue. Calculations suggest that the distance between these protons is = 5 Å only if proline bond is trans instead of cis (Wuthrich et al 1984, Adler & Scheraga 1990). To what extent the solid-state methods render this technique amenable to a membrane protein is another hurdle.

Preliminary observations indicate that the species I1 and I2 can be trapped under alkaline conditions (pH 10-12). One could thus study their role in bacteriorhodopsin folding using steady-state techniques. It is believed that these species very probably represent off-pathway products. Whether this speculation holds any legitimacy can therefore be investigated. It is however possible that under alkaline conditions the species I1 and I2 may not be an accurate representation of the species observed around neutral pH. Refolding kinetics of bacteriorhodopsin in Br-DMPC/CHAPS (brominated micelles) would be an attractive strategy to pursue. The quenching of protein fluorescence by bromine should provide a highly sensitive probe of the qualitative changes of the environment of aromatic residues during refolding. Refolding kinetics of bacteriorhodopsin in the micelles could also be monitored using fluorescent dyes conjugated to DMPC. Stop-flow far-UV CD measurements in a micelle system such as DMPC/DHPC should provide a deeper insight into the mechanism of transmembrane helix formation during bacterioopsin refolding. Note that, far-UV CD measurements cannot be performed on DMPC/CHAPS micelles due to the strong absorbing properties of CHAPS in the far-UV region of the electromagnetic spectrum. It has been shown here that thermodynamic information can be obtained on the transient folding intermediates of a membrane protein. These data combined with protein engineering techniques should form the basis of mapping the structures of these intermediates at an atomic level. The work presented here predicts that the partially refolded state bO¢ structurally resembles the native bacteriorhodopsin. The major difference lies in the state of extramembraneous loops which are predicted to be largely unfolded in the state bO¢ but fully folded in bacteriorhodopsin. This prediction needs to be tested. Although 2D crystals of bacteriorhodopsin can be easily prepared, crystallization of the apoprotein state bO¢ for subsequent crystallographic analysis may be notorious. Solid-state NMR methods may not be applicable in the foreseeable future.

Conclusions 233

One can however use site-directed mutagenesis to introduce cysteine residues at various sites in the protein followed by selective modification of the thiol group with an oxylpyrroline spin label. EPR spectroscopy of the spin-labelled proteins could then be used to compare a low resolution structure of the state bO¢ and bacteriorhodopsin. EPR spectroscopy has had much success with work on bacteriorhodoposin in the past (Hubbell & Altenbach 1994). Alternatively, vibrational spectroscopy could also be used to compare how closely the structure of the partially refolded state bO¢ resembles that of bacteriorhodopsin (Siebert 1990). One of the conclusions drawn from this study is that retinal is non-covalently bound to the apoprotein within the 430-nm intermediate. Further evidence in support of this hypothesis could be derived from kinetic refolding measurements on bacteriorhodopsin mutants Lys216Ala. The covalent bond between the aldehyde group of retinal and the e-amino group of Lys216 is believed to be unimportant for the proton pumping activity of bacteriorhodopsin (Schweiger et al 1994). Addition of retinylidene-n-alkylamines to the mutant Lys216Ala regenerates native-like of bacteriorhodopsin. There is thus strong possibility that addition of retinal to the mutant Lys216Ala will result in the formation of a 430-nm chromophore. This species would become kinetically trapped as the Schiff base between retinal and Lys216 is an obligate pre-requisite for the formation of the native chromophore of bacteriorhodopsin. Thus studies on the mutant Lys216Ala may not only confirm the original hypothesis regarding the nature of the interaction between retinal and the apoprotein within the 430-nm but could also lead to structural information on this species. The only snag however is that expression of this mutant in a bacterial host may not yield sufficient quantities of protein for structural studies. Attempts to express membrane proteins in bacterial hosts are generally fraught with difficulty. Whatever direction the future research on bacteriorhodopsin folding takes, the work presented here will provide an infectious inspiration for the present and the next generation of membrane protein folders.

REFERENCES

Abraham J P, Leslie A G W, Lutter R & Walker J E (1994). Structure at 2.8 Å Resolution of F1- ATPase from Bovine Heart Mitochondria. Nature 370, 621-628.

Adler M & Scheraga H A (1990). Non-Native Isomers of Pro93 and Pro114 Predominate in Heat- Unfolded Ribonuclease A. Biochemistry 29, 8211-8216.

Ahmad F & Bigelow C C (1982). Estimation of the Free Energy of Stabilization of Ribonuclease A, Lysozyme, a-Lactalbumin and Myoglobin. J Biol Chem 257, 12935-12938.

Alex S, Thanh H L & Vocelle D (1992). Studies of the Effect of Hydrogen Bonding on the Absorption and Fluorescence Spectra of All-Trans Retinal at Room Temperature. Canad J Chem 70, 880- 887.

Allen L C (1975). A Model for the Hydrogen Bond. Pro Nat Aca Sci 72, 4701-4705.

Allen J P, Feher G, Yeates T O, Komiya H & Rees D C (1987). Structure of the Reaction Center from Rhodobacter Sphaeroides R-26: The Cofactors. Pro Nat Aca Sci 84, 5730-5734.

Ames B N (1966). Assay of Inorganic Phosphate, Total Phosphate and Phosphatases. Meth Enzymol 8, 115-118.

Anfinson C B, Haber E, Sela M & White F H (1961). The Kinetics of Formation of Native Ribonuclease During Oxidation of the Reduced Polypeptide Chain. Pro Nat Aca Sci 47, 1309-1314.

Anfinson C B (1973). Principles That Govern the Folding of Protein Chains. Science 181, 223-230.

Anson M L (1945). Protein Denaturation and the Properties of Protein Groups. Adv Prot Chem 2, 361- 386.

Balbach J, Forge V, van Nuland N A J, Winder S L, Hore P J & Dobson C M (1995). Following Protein Folding in Real Time Using NMR Spectroscopy. Nat Struct Biol 2, 865-870.

Baldwin R L (1989). How Does Protein Folding Get Started? Trend Biochem Sci 14, 291-294.

Baldwin R L (1990). Molten Globules: Specific or Non-Specific Folding Intermediates? Biochem Mol Biol 2, 379-383.

Baldwin R L (1994). Matching Speed and Stability. Nature 369, 183-184.

Baldwin M A, Cohen F E & Prusiner S B (1995). Prion Protein Isoforms: A Convergence of Biological and Structural Investigations. J Biol Chem 270, 19197-19200.

Baldwin R L (1995). The Nature of Protein Folding Pathways: The Classical Versus the New View. J Biomol NMR 5, 103-109.

Ballew R M, Salbelko J & Gruebele M (1996a). Direct Observation of Fast Protein Folding: The Initial Collapse of Apomyoglobin. Pro Nat Aca Sci 93, 5759-5764.

Ballew R M, Salbelko J & Gruebele M (1996b). Observation of Distinct Nanosecond and Microsecond Protein Folding Events. Nat Struct Biol 3, 923-926.

Barnett S M, Dracheva S, Hendler R W & Levin I W (1996). Lipid-Induced Conformational Changes of an Integral Membrane Protein: An Infra-Red Spectroscopic Study of the Effects of Triton X-100 Treatment on the Purple Membrane of Halobacterium Halobium ET1001. Biochemistry 35, 4558-4567.

Bayley H, Huang K-S, Radhakrishnan R, Ross A H, Takagaki Y & Khorana H G (1981). Site of Attachment of Retinal in Bacteriorhodopsin. Pro Nat Aca Sci 78, 2225-2229.

Bayley H, Hojeberg B, Huang K-S, Khorana H G, Liao M-J, Lind C & London E (1982). Delipidation, Renaturation and Reconstitution of Bacteriorhodopsin. Meth Enzymol 88, 74-81.

Becker R S, Inuzuka K & Balke D E (1971). Comprehensive Investigation of the Spectroscopy of : Theoretical and Experimental Considerations of Absorption Spectra. J Am Chem Soc 93, 38-43.

Beppu Y & Kakatani T (1994). Theoretical Study of Color Control Mechanism in Retinal Proteins: Role of the Tryptophan Residue, Tyrosine Residue and Water Molecule. Photochem Photobiol 59, 660-669.

Berkower C & Michaelis S (1991). Mutational Analysis of the Yeast a-Factor Transporter STE6: A Member of the ATP-Binding Cassette (ABC) Protein Superfamily. EMBO J 10, 3777-3785.

Bevington P R & Robinson D K (1992). Data Reduction and Error Analysis for the Physical Sciences (2nd edition). McGraw Hill, New York.

Bibi E & Kaback H R (1990). In Vivo Expression of the Lac Y Gene in Two Segments Leads to Functional Lac Permease. Pro Nat Aca Sci 87, 4325-4329.

Birge R R, Murray L P, Zidovetzki R & Knapp H M (1987). Two Photon, 13C and Two-Dimensional 1H NMR Spectroscopic Studies of Retinyl Schiff Bases, Protonated Schiff Bases and Schiff Base Salts: Evidence for a Protonation Induced pp* Excited State Level Ordering Reversal. J Am Chem Soc 109, 2090-2101.

Birge R R, Einterz C M, Knapp H M & Murray L P (1988). The Nature of the Primary Photochemical Events in Rhodopsin and Isorhodopsin. Biophys J 53, 367-385.

Birge R R (1990). Nature of the Primary Photochemical Events in Rhodopsin and Bacteriorhodopsin. Biochim Biophys Acta 1016, 293-327.

Blatz P E & Mohler J H (1975). Effect of Selected Anions and Solvents on the Electronic Absorption, Nuclear Magnetic Resonance and Infra-Red Spectra of the N-Retinylidene-n-Butylammonium Cation. Biochemistry 14, 2304-2309.

Blaurock A E (1982). Analysis of Bactriorhodopsin Structure by X-Ray Diffraction. Meth Enzymol 88, 124-132.

Blobel G (1980). Intracellular Protein Topogenesis. Pro Nat Aca Sci 77, 1496-1500.

Bogomolni R A & Spudich J L (1982). Identification of a Third Rhodopsin-Like Pigment in Photoactive Halobacterium Halobium. Pro Nat Aca Sci 79, 6250-6254.

Booth P J, Flitsch S L, Stern L J, Greenhalgh D A, Kim P S & Khorana H G (1995). Intermediates in the Folding of the Membrane Protein Bacteriorhodopsin. Nat Struct Biol 2, 139-143.

Booth P J & Paulsen H (1996). Assembly of Light-Harvesting Chlorophyll a/b Complex In Vitro: Time- Resolved Fluorescence Measurements. Biochemistry 35, 5103-5108.

Booth P J, Farooq A & Flitsch S L (1996). Retinal Binding During Folding and Assembly of the Membrane Protein Bacteriorhodopsin. Biochemistry 35, 5902-5909.

Booth P J, Riley M L, Flitsch S L, Templer R H, Farooq A, Curran A R, Chadborn N & Wright P (1997). Evidence that Bilayer-Bending Rigidity Affects Membrane Protein Folding. Biochemistry 36, 197-203.

Booth P J & Farooq A (1997). Intermediates in the Assembly of Bacteriorhodopsin Investigated by Time-Resolved Absorption Spectroscopy. Eur J Biochem 246, 674-680.

Borel A C & Simon S M (1996). Biogenesis of Polytopic Membrane Proteins: Membrane Segments of P-Gycoprotein Sequentially Translocate to Span the ER Membrane. Biochemistry 35, 10587- 10594.

Braig K, Otwinowski Z, Hegde R, Boisvert D C, Joachimiak A, Horwich A L & Sigler P B (1994). The Crystal Structure of the Bacterial Chaperonin GroEL at 2.8 Å. Nature 371, 578-586.

Braiman M & Mathies R (1982). Resonance Raman Spectra of Bacteriorhodopsin's Primary Photoproduct: Evidence for a Distorted 13-Cis Retinal Chromophore. Pro Nat Aca Sci 79, 403- 407.

Braiman M S, Stern L J, Chao B H & Khorana H G (1987). Structure-Function Studies on Bacteriorhodopsin: Purification and Renaturation of Bacterioopsin Polypeptide Expressed in Escherichia Coli. J Biol Chem 262, 9271-9276.

Braiman M S, Mogi T, Marti T, Stern L J, Khorana H G & Rothschild K J (1988). Vibrational Spectroscopy of Bacteriorhodopsin Mutants: Light-Driven Proton Transport Involves Protonation Changes of Aspartic Acid Residues 85, 96, and 212. Biochemistry 27, 8516-8520.

Brandts J F, Halvorson H R & Brennan M (1975). Consideration of the Possibility that the Slow Step in Protein Denaturation Reactions is Due to Cis-Trans Isomerism of Proline Residues. Biochemistry 14, 4953-4963.

Brouillette C G, Muccio D D & Finney T K (1987). pH Dependence of Bacteriorhodopsin Thermal Unfolding. Biochemistry 26, 7431-7438.

Brouillette C G, McMichens R B, Stern L J & Khorana H G (1989). Structure and Thermal Stability of Monomeric Bacteriorhodopsin in Mixed Phospholipid/Detergent Micelles. PROTEINS: Struct Func Gen 5, 38-46.

Brown L S, Sasaki J, Kandori H, Maeda A, Needleman R & Lanyi J K (1995). Glutamic Acid 204 is the Terminal Proton Release Group at the Extracellular Surface of Bacteriorhodopsin. J Biol Chem 270, 27122-27126.

Bryngelson J D, Onuchic J N, Socci N D & Wolynes P G (1995). Funnels, Pathways and the Energy Landscape of Protein Folding: A Synthesis. PROTEINS: Struct Func Gen 21, 167-195.

Bychkova V E, Pain R H & Ptitsyn O B (1988). The Molten Globule State is Involved in the Translocation of Proteins Across Membranes? FEBS Lett 238, 231-234.

Bychkova V E, Dujsekina A E, Klenin S I, Tiktopulo E I, Uversky V N & Ptitsyn O B (1996). Molten Globule-Like State of Cytochrome c Under Conditions Simulating Those Near the Membrane Surface. Biochemistry 35, 6058-6063.

Bycroft M, Matouschek A, Kellis J T, Serrano L & Fersht A R (1990). Detection and Characterization of a Folding Intermediate in Barnase by NMR. Nature 346, 488-490.

Chaffotte A F, Guillou Y & Goldberg M E (1992). Kinetic Resolution of Peptide Bond and Side-Chain Far-UV Circular Dichroism During the Folding of Hen Egg White Lysozyme. Biochemistry 31, 9694-9702.

Chang C-H, El-Kabbani O, Tiede D, Norris J & Schiffer M (1991). Structure of the Membrane-Bound Protein Photosynthetic Reaction Center from Rhodobacter Sphaeroides. Biochemistry 30, 5352- 5360.

Chiba K, Ikai A, Kawamura-Konishi Y & Kihara H (1994). Kinetic Study of Myoglobin Refolding Monitored by 5 Optical Probe Stopped-Flow Methods. PROTEINS: Struct Func Gen 19, 110- 119.

Chitnis P R, Xu Q, Chitnis V P & Nechushtai R (1995). Function and Organization of Photosystem I Polypeptides. Photosynth Res 44, 23-40.

Christensen H & Pain R H (1991). Molten Globule Intermediates and Protein Folding. Eur Biophys J 19, 221-229.

Cladera J, Torres J & Padros E (1996). Analysis of Conformational Changes in Bacteriorhodopsin Upon Retinal Removal. Biophys J 70, 2882-2887.

Clarke A R (1996). Molecular Chaperones in Protein Folding and Translocation. Curr Opin Struct Biol 6, 43-50.

Cosson P, Lankford S P, Bonifacino J S & Klausner D R (1991). Membrane-Protein Association by Potential Intramembrane Charge Pairs. Nature 351, 414-416.

Cosson P & Bonifacino J S (1992). Role of Transmembrane Domain Interactions in the Assembly of Class II MHC Molecules. Science 258, 659-662.

Cowan S W, Schirmer T, Rummel G, Steiert M, Ghosh R, Pauptit R A, Jansonius J N & Rosenbusch J P (1992). Crystal Structures Explain Functional Properties of Two Escherichia Coli Porins. Nature 358, 727-733.

Cowan S W & Rosenbusch J P (1994). Folding Pattern Diversity of Integral Membrane Proteins. Science 264, 914-916.

Cowan S W, Garavito R M, Jansonius J N, Jenkins J A, Karlsson R, Konig N, Pai E F, Pauptit R A, Rizkallah P J, Rosenbusch J P, Rummel G & Schirmer T (1995). The Structure of OmpF Porin in a Tetragonal Crystal Form. Structure 3, 1041-1050.

Creighton T E (1980). A Three-Disulphide Intermediate in Refolding of Reduced Ribonuclease A with a Folded Conformation. FEBS Lett 118, 283-288.

Creighton T E (1981). Accessibilities and Reactivities of Cysteine Thiols During Refolding of Reduced Bovine Pancreatic Trypsin Inhibitor. J Mol Biol 151, 211-213.

Creighton T E & Goldenberg D P (1984). Kinetic Role of a Metastable Native-Like 2-Disulfide Species in the Folding Transition of Bovine Pancreatic Trypsin Inhibitor. J Mol Biol 179, 497-526.

Creighton T E (1990). Protein Folding. Biochem J 270, 1-16.

Creighton T E (1992). Proteins: Structures and Molecular Properties (2nd edition). W H Freeman, New York.

Creighton T E (1994). The Energetic Ups and Downs of Protein folding. Nat Struct Biol 1, 135-138.

Cross T A & Opella S J (1994). Solid-State NMR Structural Studies of peptides and Proteins in Membranes. Curr Opin Struct Biol 4, 574-581.

DasSarma S & Fleischmann E M (1995). Archae: A Laboratory Manual - Halophiles. Cold Spring Harbor Laboratory Press, New York.

Deber C M, Khan A R, Li Z, Joensson C, Glibowicka M & Wang J (1993). Val to Ala Mutations Selectively Alter Helix-Helix Packing in the Transmembrane Segment of Phage M13 Coat Protein. Pro Nat Aca Sci 90, 11648-11652.

Devos R, Plaetinck G, Cheroutre H, Simons G, Degrave W, Tavernier J, Remaut E & Fiers W (1983). Molecular Cloning of Human Interleukin 2 cDNA and its Expression in Escherichia Coli. Nuc Acid Res 11, 4307-4323.

Diesenhofer J, Epp O, Miki K, Huber R & Michel H (1985). Structure of the Protein Subunits in the Photosynthetic Reaction Center of Rhodopseudomonas Viridis at 3 Å Resolution. Nature 318, 618-624.

Dill K A (1985). Theory for the Folding and Stability of Globular Proteins. Biochemistry 24, 1501- 1509.

Dill K A & Chan H S (1997). From Levinthal to Pathways to Funnels. Nat Struct Biol 4, 10-19.

Donnelly D, Overington J P, Ruffle S V, Nugent J H A & Blundell T L (1993). Modeling a-helices Transmembrane Domains: The Calculation and Use of Substitution Tables for Lipid-Facing Residues. Prot Sci 2, 55-70.

Donnelly D & Findlay J B C (1994). Seven-Helix Receptors: Structure and Modelling. Curr Opin Struct Biol 4, 582-589.

Dornmair K, Kiefer H & Jahnig F (1990). Refolding of an Integral Membrane Protein: OmpA of Escherichia Coli. J Biol Chem 265, 18907-18911.

Eaton W A, Thompson P A, Chan C K, Hagen S J & Hofrichter J (1996). Fast Events in Protein Folding. Structure 4, 1133-1139.

Eaton W A, Munoz V, Thompson P A, Chan C-K & Hofrichter J (1997). Submillisecond Kinetics of Protein Folding. Curr Opin Struct Biol 7, 10-14.

Ebrey T G, Becher B, Mao B & Kilbride P (1977). Exciton Interactions and Chromophore Orientation in the Purple Membrane. J Mol Biol 112, 377-397.

Eisele J-L & Rosenbusch J P (1990). In Vitro Folding and Oligomerization of a Membrane Protein. J Biol Chem 265, 10217-10220.

Elove G A, Chaffotte A F, Roder H & Goldberg M E (1992). Early Steps in Cytochrome c Folding Probed by Time-Resolved Circular Dichroism and Fluorescence Spectroscopy. Biochemistry 31, 6876-6883.

Elove G A, Bhuyan A K & Roder H (1994). Kinetic Mechanism of Cytochrome c Folding: Involvement of the Heme and Its Ligands. Biochemistry 33, 6925-6935.

Engelman D M, Henderson R, McLachlan A D & Wallace B A (1980). Path of the Polypeptide in Bacteriorhodopsin. Pro Nat Aca Sci 77, 2023-2027.

Engelman D M & Steitz T A (1981). The Spontaneous Insertion of Proteins into and Across Membranes: The Helical Hairpin Hypothesis. Cell 23, 411-422.

Engelman D M (1982). An Implication of the Structure of Bacteriorhodopsin: Globular Membrane Proteins are Stabilized by Polar Interactions. Biophys J 37, 187-188.

Engelman D M, Goldman A & Steitz T A (1982). The Identification of Helical segments in the Polypeptide Chain of Bacteriorhodopsin. Meth Enzymol 88, 81-88.

Engelman D M (1996). Crossing the Hydrophobic Barrier: Insertion of Membrane Proteins. Science 274, 1850-1851.

Englander S W, Sosnik T R, Englander J J & Mayne L (1996). Mechanisms and Uses of Hydrogen Exchange. Curr Opin Struct Biol 6, 18-23.

Ermler U, Michel H & Schiffer M (1994). Structure and Function of the Photosynthetic Reaction Center from Rhodobacter Sphaeroides. J Bioenerg Biomem 26, 5-15.

Evans L J A, Goble M L, Hales K A & Lakey J H (1996). Different Sensitivities to Acid Denaturation Within a Family of Proteins: Implications for Acid Unfolding and Membrane Translocation. Biochemistry 35, 13180-13185.

Fahmy K, Zvyaga T A, Sakmar T P & Siebert F (1996). Spectroscopic Evidence for Altered Chromophore-Protein Interactions in Low-Temperature Photoproducts of the Visual Pigment Responsible for Congenital Night Blindness. Biochemistry 35, 15065-15073.

Feng Y, Menick D R, Katz B M, Beischel C J, Hazard E S, Misra S, Ebrey T G & Crouch R K (1994). Probing of the Retinal Binding Site of Bacteriorhodopsin by Affinity Labeling. Biochemistry 33, 11624-11630.

Fersht A R, Matouschek A & Serrano L (1992). The Folding of an Enzyme: Theory of Protein Engineering Analysis of Stability and Pathway of Protein Folding. J Mol Biol 224, 771-782.

Fersht A R (1993). Protein Folding and Stability: The Pathway of Folding of Barnase. FEBS Lett 325, 5-16.

Fersht A R, Itzhaki L S, ElMasry N F, Matthews J M & Otzen D E (1994). Single Versus Parallel Pathways of Protein Folding and Fractional Formation of Structure in the Transition State. Pro Nat Aca Sci 91, 10426-10429.

Fersht A R (1995a). Optimization of Rates of Protein Folding: The Nucleation-Condensation Mechanism and its Implications. Pro Nat Aca Sci 92, 10869-10873.

Fersht A R (1995b). Characterising Transition States in Protein Folding: An Essential Step in the Puzzle. Curr Opin Struct Biol 5, 79-84.

Fersht A R (1997). Nucleation Mechanisms in Protein Folding. Curr Opin Struct Biol 7, 3-9.

Fillingame R H (1996). Membrane Sectors of F- and V-Type H+-Transporting ATPases. Curr Opin Struct Biol 6, 491-498.

Fisher W R, Taniuchi H & Anfinson C B (1973). On the Role of Haem in the Formation of the Structure of Cytochrome c. J Biol Chem 248, 3188-.

Fodor S P A, Ames J B, Gebhard R, van den Berg E M M, Stoeckenius W, Lugtenburg J & Mathies R A (1988). Chromophore Structure in Bacteriorhodopsin's N Intermediate: Implications for the Proton-Pumping Mechanism. Biochemistry 27, 7097-7101.

Freedman R B (1995). The Formation of Protein Disulphide Bonds. Curr Opin Struct Biol 5, 85-91.

Fromme P (1996). Structure and Function of Photosystem I. Curr Opin Struct Biol 6, 473-484.

Frydman J, Nimmesgern E, Ohtsuka K & Hartl F U (1994). Folding of Nascent Polypeptide Chains in a High Molecular Mass Assembly with Molecular Chaperones. Nature 370, 111-117.

Fyfe P K & Cogdell R J (1996). Purple Bacterial Antenna Complexes. Curr Opin Struct Biol 6, 467- 472.

Gartner W & Oesterhelt D (1988). Methoxyretinals in Bacteriorhodopsin: Absorption Maxima, Cis- Trans Isomerisation and Retinal Protein Interaction. J Biochem 176, 641-648.

Garvey E P, Swank J & Matthews C R (1989). A Hydrophobic Cluster Forms Early in the Folding of Dihydrofolate Reductase. PROTEINS: Struct Func Gen 6, 259-266.

Gerber G E, Gray C P, Wildenauer D & Khorana H G (1977). Orientation of Bacteriorhodopsin in Halobacterium Halobium as Studied by Selective Proteolysis. Pro Nat Aca Sci 74, 5426-5430.

Gerber G E & Khorana H G (1982). Primary Structure of Bacteriorhodopsin: Sequencing Methods for Membrane Proteins. Meth Enzymol 88, 56-74.

Gerwert K, Souvignier G & Hess B (1990). Simultaneous Monitoring of Light-Induced Changes in Protein Side-Group Protonation, Chromophore Isomerization and Backbone Motion of Bacteriorhodopsin by Time-Resolved Fourier-Transform Infra-Red Spectroscopy. Pro Nat Aca Sci 87, 9774-9778.

Gilbert A & Baggott J (1991). Essentials of Molecular Photochemistry. Blackwell Scientific Publications, London.

Gilles-Gonzalez M A, Engelman M D & Khorana H G (1991). Structure-Function Studies of Bacteriorhodopsin: Effects of Deletions in Loops B-C and E-F on Bacteriorhodopsin Chromophore and Structure. J Biol Chem 266, 8545-8550.

Gilmanshin R I & Ptitsyn O B (1987). An Early Intermediate of Refolding a-Lactalbumin Forms Within 20 ms. FEBS Lett 223, 327-329.

Golding C, Senior S, Wilson M T & O'Shea P (1996). Time Resolution of Binding and Membrane Insertion of a Mitochondrial Signal Peptide: Correlation with Structural Changes and Evidence for Cooperativity. Biochemistry 35, 10931-10937.

Gregoret L M & Cohen F E (1991). Protein Folding: Effect of Packing Density on Chain Conformation. J Mol Biol 219, 109-122.

Grigorieff N, Ceska T A, Downing K H, Baldwin J M & Henderson R (1996). Electron- Crystallographic Refinement of the Structure of Bacteriorhodopsin. J Mol Biol 259, 393-421.

Gutin A M, Abkevich V I & Shakhnovich E I (1995). Is Burst Hydrophobic Collapse Necessary for Protein Folding. Biochemistry 34, 3066-3076.

Haltia T & Freire E (1995). Forces and Factors that Contribute to the Structural Stability of Membrane Proteins. Biochim Biophys Acta 1228, 1-27.

Hamada D, Hoshino M, Kataoka M, Fink A L & Goto Y (1993). Intermediate Conformational States of Apocytochrome c. Biochemistry 32, 10351-10358.

Hamada D, Segawa S-I & Goto Y (1996). Non-Native a-Helical Intermediate in the Refolding of b- Lactoglobulin - A Predominantly b-Sheet Protein. Nat Struct Biol 3, 868-873.

Han M & Smith S O (1995). NMR Constraints on the Location of the Retinal Chromophore in Rhodopsin and Bathorhodopsin. Biochemistry 34, 1425-1432.

Hao W & Fong H K W (1996). Blue and Ultraviolet Light-Absorbing Opsin from the Retinal Pigment Epithelium. Biochemistry 35, 6251-6256.

Hargrave P A, Hamm H E & Hofmann K P (1993). Interaction of Rhodopsin with the G Protein Transducin. BioEssays 15, 43-50.

Hargrave M S, Krzywda S, Wilkinson A J, Dou Y, Ikedasaito M & Olson J S (1994). Stability of Myoglobin - A Model for the Folding of Heme-Proteins. Biochemistry 33, 11767-11775.

Harrison S C & Durbin R (1985). Is There a Single Pathway for the Folding of a Polypeptide Chain? Pro Nat Aca Sci 82, 4028-4030.

Harrison P M, Bamborough P, Dagget V, Prusiner S B & Cohen F E (1997). The Prion Folding Problem. Curr Opin Struct Biol 7, 53-59.

Hartl F U & Martin J (1995). Molecular Chaperones in Cellular Protein Folding. Curr Opin Struct Biol 5, 92-102.

Henderson R & Unwin P N T (1975). Three Dimensional Model of Purple Membrane Obtained by Electron Microscopy. Nature 257, 28-32.

Henderson R, Baldwin J M, Ceska T A, Zemlin F, Beckmann E & Downing K H (1990). Model for the Structure of Bacteriorhodopsin Based on High Resolution Electron Cryo-Microscopy. J Mol Biol 213, 899-929.

Henderson R (1995). The Potential and Limitations of Neutrons, Electrons and X-Rays for Atomic Resolution Microscopy of Unstained Biological Molecules. Quart Rev Biophys 28, 171-193.

Henry E R & Hofrichter J (1992). Singular Value Decomposition: Application to Analysis of Experimental Data. Meth Enzymol 210, 129-192.

Hlodan R, Craig S & Pain R H (1991). Protein Folding and its Implications for the Production of Recombinant Proteins. Biotech Gen Eng Rev 9, 47-51.

Hoeltzli S D & Frieden C (1996). Real-Time Refolding Studies of 6-19F-Tryptophan Labeled Escherichia Coli Dihydrofolate Reductase Using Stopped-Flow NMR Spectroscopy. Biochemistry 35, 16843-16851.

Honig B, Greenberg A D, Dinur U & Ebrey T G (1976). Visual Pigment Spectra: Implications of the Protonation of the Retinal Schiff Base. Biochemistry 15, 4593-4599.

Honig B, Dinur U, Nakanishi K, Balogh-Nair V, Gawinowicz M, Arnaboldi M & Moto M (1979). An External Point Charge Model for Wavelength Regulation in Visual Pigments. J Am Chem Soc 101, 7084-7086.

Honig B, Dinur U, Birge R R & Ebrey T G (1980). The Isomer Dependence of Oscillator Strengths in Retinal and Related Molecules - Spectroscopic Assignments. J Am Chem Soc 102, 488-494.

Honig B & Ebrey T G (1982). Protein-Chromophore Interactions as Spectroscopic and Photochemical Determinants. Meth Enzymol 88, 462-470.

Honig B H & Hubbell W L (1984). Stability of Salt Bridges in Membrane Proteins. Pro Nat Aca Sci 81, 5412-5416.

Honig B & Yang A-S (1995). Free Energy Balance in Protein Folding. Adv Prot Chem 46, 27-58.

Houry W A & Scheraga H A (1996). Nature of the Unfolded State of Ribonuclease A: Effect of Cis- Trans X-Pro Peptide Bond Isomerization. Biochemistry 35, 11719-11733.

Hu J, Griffin R G & Herzfeld J (1994). Synergy in the Spectral Tuning of Retinal Pigments: Complete Accountng of the Opsin Shift in Bacteriorhodopsin. Pro Nat Aca Sci 91, 8880-8884.

Huang K-S, Bayley H & Khorana H G (1980). Delipidation of Bacteriorhodopsin and Reconstitution with Exogenous Phospholipid. Pro Nat Aca Sci 77, 323-327.

Huang K-S, Bayley H, Liao M-J, London E & Khorana H G (1981). Refolding of an Integral Membrane Protein: Denaturation, Renaturation and Reconstitution of Intact Bacteriorhodopsin and Two Proteolytic Fragments. J Biol Chem 256, 3802-3809.

Huang G S & Oas T G (1995a). Submillisecond Folding of Monomeric l Repressor. Pro Nat Aca Sci 92, 6878-6882.

Huang G S & Oas T G (1995b). Structure and Stability of Monomeric l Repressor: NMR Evidence for Two-State Folding. Biochemistry 34, 3884-3892.

Hubbell W L & Altenbach C (1994). Investigation of Structure and Dynamics in Membrane Proteins Using Site-Directed Spin Labeling. Curr Opin Struct Biol 4, 566-573.

Hunt J F, Bousche O, Earnest T N, Kalghatgi K, Reilly K, Horvath C, Rothschild K J & Engelman D M (1993). A Biophysical Dissection of a Prototypical a-Helical Integral Membrane Protein. Unpublished Observations.

Ikeguchi M, Kuwajima K, Mitani M & Sugai S (1986). Evidence for Identity Between the Equilibrium Folding Intermediate and a Transient Folding Intermediate: A Comparative Study of the Folding Reactions of a-Lactalbumin and Lysozyme. Biochemistry 25, 6965-6972.

Itzhaki L S, Otzen D E & Fersht A R (1995a). The Structure of the Transition State for Folding of Chymotrypsin Inhibitor-2 Analyzed by Protein Engineering Methods: Evidence for a Nucleation- Condensation Mechanism for Protein Folding. J Mol Biol 254, 260-288.

Itzhaki L S, Otzen D E & Fersht A R (1995b). Nature and Consequences of GroEL-Protein Interactions. Biochemistry 34, 14581-14587.

Iwata S, Ostermeier C, Ludwig B & Michel H (1995). Structure at 2.8 Å Resolution of Cytochrome c Oxidase from Paracoccus Denitrificans. Nature 376, 660-669.

Jackson S E & Fersht A R (1991). Folding of Chymotrypsin Inhibitor-2: Evidence for a Two-State Transition. Biochemistry 30, 10428-10435.

Jahnig F (1983). Thermodynamics and Kinetics of Protein Incorporation into Membranes. Pro Nat Aca Sci 80, 3691-3695.

Jansson S, Andersen B & Scheller V H (1996). Cross-Linking Studies of PSI: Nearest Neighbour Analysis of Higher Plant Photosystem I Holocomplex. Plant Physiol 112, 409-420.

Jennings P A & Wright P E (1993). Formation of a Molten Globule Intermediate Early in the Kinetic Folding Pathway of Apomyoglobin. Science 262, 892-896.

Jennings P A, Finn B E, Jones B E & Matthews C R (1993). Reexamination of the Folding Mechanism of Dihydrofolate Reductase from Escherichia Coli: Verification and Refinement of a Four Channel Model. Biochemistry 32, 3783-3789.

Jones C M, Henry E R, Hu Y, Chan C-K, Luck S D, Bhuyan A, Roder H, Hofrichter J & Eaton W A (1993). Fast Events in Protein Folding Initiated by Nanosecond LASER Photolysis. Pro Nat Aca 90, 11860-11864.

Jones B E & Matthews C R (1995). Early Intermediates in the Folding of Dihydrofolate Reductase from Escherichia Coli Detected by Hydrogen Exchange and NMR. Prot Sci 4, 167-177.

Junge W & Kakitani T (1986). Vision. Photobiochem Photobiophys 13, 211-418.

Kahn T W, Sturtevant J M & 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.

Kahn T W & Engelman D M (1992). Bacteriorhodopsin can be Refolded from Two Independently Stable Transmembrane Helices and the Complementary Five-Helix Fragment. Biochemistry 31, 6144-6151.

Kakitani H, Kakitani T, Rodman H & Honig B (1985). On the Mechanism of Wavelength Regulation in Visual Pigments. Photochem Photobiol 41, 471-479.

Kalisky O, Feitelson J & Ottolenghi M (1981). Photochemistry and Fluorescence of Bacteriorhodopsin Excited in Its 280-nm Absorption Band. Biochemistry 20, 205-209.

Karplus M & Shakhnovich E (1992). Protein Folding (edited by Creighton T E). W H Freeman, New York.

Karplus M & Weaver D (1994). Protein Folding Dynamics: The Diffusion-collision Model and Experimental Data. Prot Sci 3, 650-668.

Karplus M & Sali A (1995). Theoretical Studies of Protein Folding and Unfolding. Curr Opin Struct Biol 5, 58-73.

Karrasch S, Bullough P A & Ghosh R (1995). The 8.5 Å Projection Map of the Light-Harvesting Complex I from Rhodospirillum Rubrum Reveals a Ring Composed of 16 Subunits. EMBO J 14, 631-638.

Kataoka M, Kahn T W, Tsujiuchi T, Engelman D M & Tokunaga F (1992). Bacteriorhodopsin Reconstituted from Two Individual Helices and the Complementary Five-Helix Fragment is Photoactive. Photochem Photobiol 56, 895-901.

Kates M, Yengoyan L S & Sastry P S (1965). A Diether Analog of Phosphatidyl Glycerophosphate in Halobacterium Cutirubrum. Biochim Biophys Acta 98, 252-268.

Kates M. (1972). Laboratory Techniques in Biochemistry and Molecular Biology (edited by Work T S & Work E). American Elsevier Publishing Co, New York.

Kates M, Kushwaha S C & Sprott G D (1982). Lipids of Purple Membrane from Extreme Halophiles and of Methanogenic Bacteria. Meth Enzymol 88, 98-111.

Kelly J W (1996). Alternative Conformations of Amyloidogenic Proteins Govern Their Behaviour. Curr Opin Struct Biol 6, 11-17.

Khorana H G (1988). Bacteriorhodopsin: A Membrane Protein That Uses Light to Translocate Protons. J Biol Chem 263, 7439-7442.

Khorana H G (1993). Two Light Transducing Membrane Proteins: Bacteriorhodopsin and the Mammalian Rhodopsin. Pro Nat Aca Sci 90, 1166-1171.

Khorasanizadeh S, Peters I D & Roder H (1996). Evidence for a Three-State Model of Protein Folding from Kinetic Analysis of Ubiquitin Variants with Altered Core Residues. Nat Struct Biol 3, 193- 205.

Kiefhaber T (1995). Kinetic Traps in Lysozyme Folding. Pro Nat Aca Sci 92, 9029-9033.

Kim P S & Baldwin R L (1980). Structural Intermediates Trapped During the Folding of Ribonuclease A by Amide Proton Exchange. Biochemistry 19, 6124-.

Kim P S & Baldwin R L (1984). A Helix Stop Signal in the Isolated S-Peptide of Ribonuclease A. Nature 307, 329-334.

Kim P S & Baldwin R L (1990). Intermediates in the Folding Reactions of Small Proteins. Ann Rev Biochem 59, 631-660.

Kimura Y, Vassylyev D G, Miyazawa A, Kidera A, Matsushima M, Mitsuoka K, Murata K, Hirai T & Fujiyoshi Y (1997). Surface of Bacteriorhodopsin Revealed by High Resolution Electron Crystallography. Nature 389, 206-211.

Kippen A D & Fersht A R (1995). Analysis of the Mechanism of Assembly of Cleaved Barnase from Two Peptide Fragments and Its Relevance to the Folding Pathway of Uncleaved Barnase. Biochemistry 34, 1464-1468.

Kleinschmidt J H & Tamm L K (1996). Folding Intermediates of a b-Barrel Membrane Protein: Kinetic Evidence for a Multi-Step Membrane Insertion Mechanism. Biochemistry 35, 12993-13000.

Klug C S, Su W, Liu J, Klebba P E & Feix J B (1995). Denaturant Unfolding of the Ferric Enterobactin Receptor and Ligand-Induced Stabilization Studied by Site-Directed Spin Labeling. Biochemistry 34, 14230-14236.

Kobilka B K, Kobilka T S, Daniel K, Regan J W, Caron M G & Lefkowitz R J (1988). Chimeric a2- and b2-Adrenergic Receptors: Delineation of Domains Involved in Effector Coupling and Ligand Binding Specificity. Science 240, 1310-1316.

Kohorn B D & Auchincloss A H (1991). Integration of a Chlorophyll-Binding Protein into Escherichia Coli Membranes in the Absence of Chlorophyll. J Biol Chem 266, 12048-12052.

Komiya H, Yeates T O, Rees D C, Allen J P & Feher G (1988). Structure of the Reaction Center from Rhodobacter Sphaeroides R-26 and 2.4.1: Symmetry Relations and Sequence Comparisons Between Different Species. Pro Nat Aca Sci 85, 9012-9016.

Konishi T & Packer L (1977). Chemical Modification of Bacteriorhodopsin with N-Bromosuccinimide. FEBS Lett 79, 369-373.

Kragelund B B, Robinson C V, Knudsen J, Dobson C M & Poulsen F M (1995). Folding of a Four- Helix Bundle: Studies of Acyl-Coenzyme A Binding Protein. Biochemistry 34, 7217-7224.

Krauss N, Hinrichs W, Witt I, Fromme P, Pritzkow W, Dauter Z, Betzel C, Wilson K S, Witt H S & Sanger W (1993). Three-Dimensional Structure of System I of at 6 Å Resolution. Nature 361, 326-331.

Krauss N, Schubert W-D, Klukas O, Fromme P, Witt H T & Saenger W (1996). Photosystem I at 4 Å Resolution: A Joint Photosynthetic Reaction Center and Core Antenna System. Nat Struct Biol 3, 965-973.

Krebs M P & Khorana H G (1993). Mechanism of Light-Dependent Proton Translocation by Bacteriorhodopsin. J Bact 175, 1555-1560.

Kreusch A, Neubuser A, Schiltz E, Weckesser J & Schulz G E (1994). The Structure of the Membrane Channel Porin from Rhodopseudomonas Blastica at 2.0 Å Resolution. Prot Sci 3, 58-63.

Kriwacki R W, Hengst L, Tennant L, Reed S I & Wright P E (1996). Structural Studies of p21 in the Free and Cdk2-Bound State: Conformational Disorder Mediates Binding Diversity. Pro Nat Aca Sci 93, 11504-11509.

Kuhlbrandt W (1992). Two-Dimensional Crystallization of Membrane Proteins. Quart Rev Biophys 25, 1-49.

Kuhlbrandt W, Wang D N & Fujiyoshi Y (1994). Atomic Model of Plant Light-Harvesting Complex by Electron Crystallography. Nature 367, 614-621.

Kuwajima K, Yamaya H, Miwa S, Sugai S & Nagamura T (1987). Rapid Formation of Secondary Structure Framework in Protein Folding Studied by Stopped-Flow Circular Dichroism. FEBS Lett 221, 115-118.

Kuwajima K, Sakuraoka A, Fueki S, Yoneyama M & Sugai S (1988). Folding of Carp Parvalbumin Studied by Equilibrium and Kinetic Circular Dichroism Spectra. Biochemistry 27, 7419-7428.

Kuwajima K, Mitani M & Sugai S (1989). Characterization of the Critical State in Protein Folding: Effects of Guanidine Chloride and Specific Ca2+ Binding on the Folding Kinetics of a- Lactalbumin. J Mol Biol 206, 547-561.

Labhardt A M (1986). Folding Intermediates Studied by Circular Dichroism. Meth Enzymol 131, 126- 135.

Labhart H (1957). FE Theory Including an Elastic s Skeleton: Spectra and Bond Lengths in Long Polyenes. J Chem Phys 27, 957-962.

Laemmli U K (1970). Cleavage of Structural Proteins During the Assembly of the Head of Bacteriophage T4. Nature 227, 680-685.

Lansbury P T, Costa P R, Griffiths J M, Simon E J, Auger M, Halverson K J, Kocisko D A, Hendsch Z S, Ashburn T T, Spencer R G S, Tidor B & Griffin R G (1995). Structural Model for the b- Amyloid Fibril Interstand Alignment of an Antiparallel b-Sheet Comprising a C-Terminal Peptide. Nat Struct Biol 2, 990-998.

Lanyi J K (1986). Halorhodopsin: A Light-Driven Chloride Ion Pump. Ann Rev Biophys Biophys Chem 15, 11-28.

Lanyi J K (1993). Proton Translocation Mechanism and Energetics in the Light-Driven Pump Bacteriorhodopsin. Biochim Biophys Acta 1183, 241-261.

Lazar S W & Kolter R (1996). SurA Assists the Folding of Escherichia Coli Outer Membrane Proteins. J Bact 178, 1770-1773.

Lecker S H, Driesen A J M & Wickner W (1990). ProOmpA Contains Secondary and Tertiary Structure Prior to Translocation and is Shielded from Aggregation by Association with SecB Protein. EMBO J 9, 2309-2314.

Lemke H-D & Oesterhelt D (1981). Lys216 is a Binding Site of the Retinyl Moiety in Bacteriorhodopsin. FEBS Lett 128, 255-260.

Lemmon M A, Flanagan J M, Hunt J F, Adair B D, Bormann B-J, Dempsey C E & Engelman D M (1992). Glycophorin Dimerization is Driven by Specific Interactions Between Transmembrane a-Helices. J Biol Chem 267, 7683-7689.

Lemmon M A & Engelman D M (1994). Specificity and Promiscuity in Membrane Helix Interactions. Quart Rev Biophys 27, 157-218.

Lemmon M A, Treutlein H R, Adams P D, Brunger A T & Engelman D M (1994). A Dimerization Motif for Transmembrane a-Helices. Nat Struct Biol 1, 157-163.

Levinthal C (1968). Are There Pathways for protein Folding? J Chim Phys 65, 44-45.

Li S-C & Deber C M (1994). A Measure of Helical Propensity for Amino Acids in Membrane Environments. Nat Struct Biol 1, 368-373.

Liao M-J, London E & Khorana H G (1983). Regeneration of the Native Bacteriorhodopsin Structure from Two Chymotryptic Fragments. J Biol Chem 258, 9949-9955.

Liao M J, Huang K S & Khorana H G (1984). Regeneration of Native Bacteriorhodopsin Structure from Fragments. J Biol Chem 259, 4200-4204.

Litman B J (1979). Rhodopsin: Its Molecular Substructure and Phospholipid Interactions. Photochem Photobiol 29, 671-677.

Liu R S H, Krogh E, Li X-Y, Mead D, Colmenares L U, Theil J R, Ellis J, Wong D & Asato A E (1993). Analysing the Red-Shift Characteristics of Azulenic, Naphthyl, Other Ring-Fused and Retinyl Pigment Analogs of Bacteriorhodopsin. Photochem Photobiol 58, 701-705.

Liu X, Siegel D L, Fan P, Brodsky B & Baum J (1996). Direct NMR Measurement of the Folding Kinetics of a Trimeric Peptide. Biochemistry 35, 4306-4313.

London E & Khorana H G (1982). Denaturation and Renaturation of Bacteriorhodopsin in Detergents and Lipid-Detergent Mixtures. J Biol Chem 257, 7003-7011.

Lozier R H, Bogomolni R A & Stoeckenius W (1975). Bacteriorhodopsin: A Light-Driven Proton Pump in Halobacterium Halobium. Biophys J 15, 955-962.

Maeda A, Asato A E, Liu R S H & Yoshizawa T (1984). Interaction of Aromatic Retinal Analogues with Apopurple Membranes of Halobacterium Halobium. Biochemistry 23, 2507-2513.

Manolios N, Bonifacino J S & Klausner R D (1990). Transmembrane Helical Interactions and the Assembly of the T-Cell Receptor Complex. Science 249, 274-277.

Marinetti T, Subramaniam S, Mogi T, Marti T & Khorana H G (1989). Replacement of Aspartic Residues 85, 96, 115 or 212 Affects the Quantum Yield and Kinetics of Proton Release and Uptake by Bacteriorhodopsin. Pro Nat Aca Sci 86, 529-533.

Marquardt D W (1963). An Algorithm for Least-Squares Estimation of Non-Linear Parameters. J Soc Indus App Math 11, 431-441.

Marston F A O (1986). The Purification of Eukaryotic Polypeptides Synthesized in Escherichia Coli. Biochem J 240, 1-12.

Marti T, Rosselet S J, Otto H, Heyn M P & Khorana H G (1991). The Retinylidene Schiff Base Counterion in Bacteriorhodopsin. J Biol Chem 266, 18674-18683.

Martin J & Hartl F U (1993). Protein Folding in the Cell: Molecular Chaperones Pave the Way. Structure 1, 161-164.

Martin J & Hartl F U (1997). Chaperone-Assisted Protein Folding. Curr Opin Struct Biol 7, 41-52.

Marwan W & Oesterhelt D (1987). Signal Formation in the Halobacterial Photophobic Response Mediated by a Fourth Retinal Protein (P480). J Mol Biol 195, 333-342.

Matagne A, Radford S E & Dobson C M (1997). Fast and Slow Tracks in Lysozyme Folding: Insight into the Role of Domains in the Folding Process. J Mol Biol 267, 1068-1074.

Mathies R A, Lin S W, Ames J B & Polland W T (1991). From Femtoseconds to Biology: Mechanism of Bacteriorhodopsin's Light-Driven Proton Pump. Ann Rev Biophys Biophys Chem 20, 491- 518.

Mathies R A & Kakitani T (1992). Biophysical Chemistry of Retinal proteins. Photochem Photobiol 56, 857-1190.

Matouchek A, Kellis J T, Serrano L & Fersht A R (1989). Mapping the Transition State and Pathway of Protein Folding by Protein Engineering. Nature 340, 122-126.

Matouchek A, Kellis J T, Serrano L, Bycroft M & Fersht A R (1990). Transient Folding Intermediates Characterized by Protein Engineering. Nature 346, 440-445.

Matouchek A, Serrano L & Fersht A R (1992). The Folding of an Enzyme: Structure of an Intermediate in the Refolding of Barnase by a Protein Engineering Procedure. J Mol Biol 224, 819-835.

Matthews C R (1993). Pathways of Protein Folding. Ann Rev Biochem 62, 653-683.

Mayhew M, Da-Silva A R C, Martin J, Erdjument-Bromage H, Tempst P & Hartl F U (1996). Protein Folding in the Central Cavity of the GroEL-GroES Chaperonin Complex. Nature 379, 420-426.

Mayr L M, Odefey C, Schutkowski M & Schmid F X (1996). Kinetic Analysis of the Unfolding and Refolding of Ribonuclease T1 by a Stopped-Flow Double-Mixing Technique. Biochemistry 35, 5550-5561.

Mazer N A, Benedek G B & Carey M C (1980). Quasielastic Light-Scattering Studies of Aqueous Biliary Lipid Systems. Mixed Micelle Formation in Bile Salt-Lecithin Solutions. Biochemistry 19, 601-615.

McDermott G, Prince S M, Freer A A, Hawthornthwaite-Lawless A M, Papiz M Z, Cogdell R J & Isaacs N W (1995). Crystal Structure of an Integral Membrane Light-Harvesting Complex from Photosynthetic Bacteria. Nature 374, 517-521.

Miranker A D & Dobson C M (1996). Collapse and Cooperativity in Protein Folding. Curr Opin Struct Biol 6, 31-42.

Mogi T, Stern L J, Marti T, Chao B H & Khorana H G (1988). Aspartic Acid Substitutions Affect Proton Translocation by Bacteriorhodopsin. Pro Nat Aca Sci 85, 4148-4152.

Mogi T, Stern L J, Chao B H & Khorana H G (1989). Structure-Function Studies on Bacteriorhodopsin: Substitutions of the Membrane-Embedded Prolines 50, 91 and 186 - The Effects are Determined by the Substituting Amino Acids. J Biol Chem 264, 14192-14196.

Moody M F, Jones P T, Carver J A, Boyd J & Campbell I D (1987). 1H Nuclear Magnetic Resonance Studies of an Integral Membrane Protein: Subunit c of the F1Fo ATP Synthase. J Mol Biol 193, 759-774.

Morozova L A, Haynie D T, Arico-Muendel C, Dael H v & Dobson C M (1995). Structural Basis of the Stability of a Lysozyme Molten Globule. Nat Struct Biol 2, 871-875.

Mukhopadhyay A K, Dracheva S, Bose S & Hendler R W (1996). Control of the Integral Membrane Proton Pump Bacteriorhodopsin by Purple Membrane Lipids of Halobacterium Halobium. Biochemistry 35, 9245-9252.

Mullen E, Johnson A H & Akhtar M (1981). The Identification of Lys216 as the Retinal Binding Residue in Bacteriorhodopsin. FEBS Lett 130, 187-193.

Myers A B & Birge R R (1980). The Effect of Solvent Environment on Molecular Electronic Oscillator Strengths. J Chem Phys 73, 5314-5321.

Nakanishi K, Balogh-Nair V, Arnaboldi M, Tsujimoto K & Honig B (1980). An External Point-Charge Model for Bacteriorhodopsin to Account for its Purple Colour. J Am Chem Soc 102, 7945-7947.

Nall B (1994). Mechanisms of Protein Folding (edited by Pain R H). Oxford University Press, New York.

Needleman R, Chang M, Ni B, Varo G, Fornes J, White S H & Lanyi J K (1991). Properties of Asp212Asn Bacteriorhodopsin Suggest that Asp212 and Asp85 Both Participate in a Counterion and Proton Acceptor Complex Near the Schiff Base. J Biol Chem 266, 11478-11484.

Nelsestuen G L & Lim T K (1977). Equilibria Involved in Prothrombin- and Blood-Clotting Factor X- Membrane Binding. Biochemistry 16, 4164-4171.

Nolting B, Golbik R & Fersht A R (1995). Submillisecond Events in Protein Folding. Pro Nat Aca Sci 92, 10668-10672.

Nolting B, Golbik R, Neira J L, Soler-Gonzalez A S, Schreiber G & Fersht A R (1997). The Folding Pathway of a Protein at High Resolution From Microseconds to Seconds. Pro Nat Aca Sci 94, 826-830.

Odefey C, Mayr L M & Schmid F X (1995). Non-Prolyl Cis-Trans Peptide Bond Isomerization as a Rate-Determining Step in Protein Unfolding and Refolding. J Mol Biol 245, 69-78.

Oesterhelt D & Stoeckenius W (1971). Rhodopsin-Like Protein from the Purple Membrane of Halobacterium Halobium. Nat New Biol 233, 149-152.

Oesterhelt D & Hess B (1973). Reversible Photolysis of the Purple Complex in the Purple Membrane of Halobacterium Halobium. Eur J Biochem 37, 316-326.

Oesterhelt D, Meentzen M & Schuhmann L (1973). Reversible Dissociation of the Purple Complex in Bacteriorhodopsin and Identification of 13-Cis and All-Trans Retinal as its Chromophores. Eur J Biochem 40, 453-463.

Oesterhelt D & Stoeckenius W (1974). Isolation of the Cell Membrane of Halobacterium Halobium and Its Fractionation into Red and Purple Membrane. Meth Enzymol 31, 667-678.

Oesterhelt D & Schuhmann L (1974). Reconstitution of Bacteriorhodopsin. FEBS Lett 44, 262-265.

Oesterhelt D & Tittor J (1989). Two Pumps, One Principle: Light-Driven Ion Transport in Halobacteria. Trend Biochem Sci 14, 57-61.

Oesterhelt D, Tittor J & Bamberg E (1992). A Unifying Concept for Ion Translocation by Retinal Proteins. J Bioenerg Biomem 24, 181-191.

Ohgushi M & Wada A (1983). Molten Globule State: A Compact Form of Globular Proteins with Mobile Side-Chains. FEBS Lett 164, 21-24.

Onuchic J N, Wolynes P G, Luthey-Schulten Z & Socci N D (1995). Toward an Outline of the Topography of a Realistic Protein Folding Funnel. Pro Nat Aca Sci 92, 3626-3630.

Ostermeier C, Iwata S & Michel H (1996). Cytochrome c Oxidase. Curr Opin Struct Biol 6, 460-466.

Otzen D E, Itzhaki L S, ElMasry N F, Jackson S E & Fersht A R (1994). Structure of the Transition State for the Folding / Unfolding of the Barley Chymotrypsin Inhibitor-2 and Its Implications for Mechanisms of Protein Folding. Pro Nat Aca Sci 91, 10422-10425.

Pace C N & Vanderburg K E (1979). Determining Globular Protein Stability: Guanidine Hydrochloride Denaturation of Myoglobin. Biochemistry 18, 288-292.

Pace C N & Creighton T E (1986). The Disulphide Folding Pathway of Ribonuclease T1. J Mol Biol 188, 477-486.

Pain R H (1987). Protein Folding for Pleasure and for Profit. Trend Biochem Sci 12, 309-312.

Palczewski K, Buczylko J, Kaplan M W, Polans A S & Crabb J W (1991). Mechanism of Activation. J Biol Chem 266, 12949-12955.

Paliwal R & London E (1996). Comparison of the Conformation, Hydrophobicity and Model Membrane Interactions of Diphtheria Toxin to those of Formaldehyde-Treated Toxin (Diphtheria Toxoid): Formaldehyde Stabilization of the Native Conformation Inhibits Changes that Allow Membrane Insertion. Biochemistry 35, 2374-2379.

Pan K-M, Baldwin M, Nguyen J, Gasset M, Serban A, Groth D, Melhorn I, Huang Z, Fletterick R J, Cohen F E & Prusiner S B (1993). Conversion of a-Helices into b-Sheets Features in the Formation of the Scrapie Prion Protein. Pro Nat Aca Sci 90, 10962-10966.

Parker M W & Pattus F (1993). Rendering a Membrane Protein Soluble in Water: A Common Packing Motif in Bacterial Protein Toxins. Trend Biochem Sci 18, 391-395.

Pascher T, Chesick J P, Winkler J R & Gray H B (1996). Protein Folding Triggered by Electron Transfer. Science 271, 1558-1560.

Paulsen H, Rumler U & Rudiger W (1990). Reconstitution of Pigment-Containing Complexes from Light-Harvesting Chlorophyll a/b-Binding Protein Overexpressed in Escherichia Coli. Planta 181, 204-211.

Pepe I M & Cugnoli C (1992). Retinal Photoisomerase: Role in Invertebrate Visual Cells. J Photochem Photobiol 13, 5-17.

Pepys M B, Hawkins P N, Booth D R, Vigushin D M, Tennent G A, Soutar A K, Totty N, Nguyen O, Blake C C F, Terry C J, Feest T G, Zalin A M & Hsuan J J (1993). Human Lysozyme Gene Mutations Cause Hereditary Systemic Amyloidosis. Nature 362, 553-557.

Perazzona B, Spudich E N & Spudich J L (1996). Deletion Mapping of the Sites on the HtrI Transducer for Sensory Rhodopsin I Interaction. J Bact 178, 6475-6478.

Pfefferle J-M, Maeda A, Sasaki J & Yoshizawa T (1991). Fourier Transform Infra-Red Study of the N Intermediate of Bacteriorhodopsin. Biochemistry 30, 6548-6556.

Phillips C M, Mizutani A & Hochstrasser R M (1995). Ultrafast Thermally Induced Unfolding of RNAse A. Pro Nat Aca Sci 92, 7292-7296.

Picot D & Garavito R M (1994). Prostaglandin H Synthase: Implications for Membrane Structure. FEBS Lett 346, 21-25.

Picot D, Loll P J & Garavito R M (1994). The X-Ray Crystal Structure of the Membrane Protein Prostaglandin H2 Synthase-1. Nature 367, 243-249.

Pitt G A J, Collins F D, Morton R A & Stok P (1955). Studies on Rhodopsin: Retinylidenemethylamine, an Indicator Yellow Analogue. Biochem J 59, 122-128.

Plaxco K W & Dobson C M (1996). Time-Resolved Biophysical Methods in the Study of Protein Folding. Curr Opin Struct Biol 6, 630-636.

Plumley F G & Schmidt G W (1987). Reconstitution of Chlorophyll a/b Light-Harvesting Complexes: Xanthophyll-Dependent Assembly and Energy Transfer. Pro Nat Aca Sci 84, 146-150.

Polland H J, Franz M A, Zinth W, Kaiser W & Oesterhelt D (1986a). Energy Transfer from Retinal to Amino Acids - A Time-Resolved Study of the Ultraviolet Emission of Bacteriorhodopsin. Biochim Biophys Acta 851, 407-415.

Polland H J, Franz M A, Zinth W, Kaiser W, Kolling E & Oesterhelt D (1986b). Early Picosecond Events in the Photocycle of Bacteriorhodopsin. Biophys J 49, 651-662.

Popot J-L, Trewhella J & Engelman D M (1986). Reformation of Crystalline Purple Membrane from Purified Bacteriorhodopsin Fragments. EMBO J 5, 3039-3044.

Popot J-L, Gerchman S-E & Engelman D M (1987). Refolding of Bacteriorhodopsin in Lipid Bilayers: A Thermodynamically Controlled Two-stage Process. J Mol Biol 198, 655-676.

Popot J-L & Engelman D M (1990). Membrane Protein Folding and Oligomerisation: The Two-stage Model. Biochemistry 29, 4031-4037.

Popot J-L (1993). Integral Membrane Protein Structure: Transmembrane a-Helices as Autonomous Folding Domains. Curr Opin Struct Biol 3, 532-540.

Prockop D J & Kivirikko K I (1995). Collagens: Molecular Biology, Diseases and Potentials for Therapy. Ann Rev Biochem 64, 403-434.

Ptitsyn O B, Pain R H, Semisotnov G V, Zerovnik E & Razgulyaev O I (1990). Evidence for a Molten Globule State as a General Intermediate in Protein Folding. FEBS Lett 262, 20-24.

Ptitsyn O B (1995). Structure of Folding Intermediates. Curr Opin Struct Biol 5, 74-78.

Radford S E, Dobson C M & Evans P A (1992). The Folding of Hen Lysozyme Involves Partially Structured Intermediates and Multiple Pathways. Nature 358, 302-307.

Ramachandran G N & Sasisekharan V (1968). Conformation of Polypeptides and Proteins. Adv Prot Chem 23, 283-437.

Rao V R & Oprian D D (1996). Activating Mutations of Rhodopsin and Other G Protein-Coupled Receptors. Ann Rev Biophys Biomol Struct 25, 287-314.

Rassow J & Pfanner N (1996). Protein Biogenesis: Chaperones for Nascent Polypeptides. Curr Biol 6, 115-118.

Redfield C, Smith R A G & Dobson C M (1994). Structural Characterisation of a Highly-Ordered Molten Globule at Low pH. Nat Struct Biol 1, 23-29.

Rehorek M & Heyn M P (1979). Binding of All-Trans Retinal to the Purple Membrane: Evidence for Cooperativity and Determination of the Extinction coefficient. Biochemistry 18, 4977 - 4983.

Renthal R, Cothran M, Dawson N & Harris G J (1987). Fluorescent Labeling of Bacteriorhodopsin: Implications for Helix Connections. Biochim Biophys Acta 897, 384-394.

Riley M L, Wallace B A, Flitsch S L & Booth P J (1997). Slow a Helix Formation During Folding of a Membrane Protein. Biochemistry 36, 192-196.

Robb F T & Place A R (1995). Archae: A Laboratory Manual - Thermophiles. Cold Spring Harbor Laboratory Press, New York.

Robyt J F & White B J (1987). Biochemical Techniques: Theory and Practice. Brooks/Cole Publishing Co, Monterey (California).

Roder H, Elove G A & Englander S W (1988). Structural Characterization of Folding Intermediates in Cytochrome c by H-Exchange Labelling and Proton NMR. Nature 335, 700-704.

Roder H (1989). Structural Characterization of Protein Folding Intermediates by Proton Magnetic Resonance and Hydrogen Exchange. Meth Enzymol 176, 446-473.

Roder H & Colon W (1997). Kinetic Role of Early Intermediates in Protein Folding. Curr Opin Struct Biol 7, 15-28.

Rothschild K J (1992). FTIR Difference Spectroscopy of Bacteriorhodopsin: Toward a Molecular Model. J Bioenerg Biomem 24, 147-167.

Sahin-Toth M, Kaback H R & Friedlander M (1996). Association Between the Amino- and Carboxy- Terminal Halves of Lactose Permease is Specific and Mediated by Multiple Transmembrane Domains. Biochemistry 35, 2016-2021.

Sali A, Shakhnovich E & Karplus M (1994a). How Does a Protein Fold? Nature 369, 248-251.

Sali A, Shakhnovich E & Karplus M (1994b). Kinetics of Protein Folding: A Lattice Model Study of the Requirements for Folding to the Native State. J Mol Biol 235, 1614-1636.

Sanders C R, Czerski L, Vinogradova O, Badola P, Song D & Smith S O (1996). Escherichia Coli Diacylglycerol Kinase is an a-Helical Polytopic Membrane Protein and can Spontaneously Insert into Preformed Lipid Vesicles. Biochemistry 35, 8610-8618.

Savarese T M & Fraser C M (1992). In Vitro Mutagenesis and the Search for Structure-Function Relationships Among G Protein-Coupled Receptors. Biochem J 283, 1-19.

Schaffer A M, Waddell W H & Becker R S (1974). Visual Pigments: Experimental and Theoretical Investigations of the Absorption Spectra of Retinal Schiff Bases and Retinals. J Am Chem Soc 96, 2063-2068.

Schatz G & Dobberstein B (1996). Common Principles of Protein Translocation Across Membranes. Science 271, 1519-1526.

Schertler G F X, Villa C & Henderson R (1993). Projection Structure of Rhodopsin. Nature 362, 770- 772.

Schindler T, Herrler M, Marahiel M A & Schmid F X (1995). Extremely Rapid Protein Folding in the Absence of Intermediates. Nat Struct Biol 2, 663-673.

Schmid F X & Baldwin R L (1979). Detection of an Early Intermediate in the Folding of Ribonuclease A by Protection of Amide Protons Against Exchange. J Mol Biol 135, 199-215.

Schreckenbach T, Walckhoff B & Oesterhelt D (1977). Studies on the Retinal-Protein Interaction in Bacteriorhodopsin. Eur J Biochem 76, 499-511.

Schreckenbach T, Walckhoff B & Oesterhelt D (1978a). Specificity of the Retinal Binding Site of Bacteriorhodopsin: Chemical and Stereochemical Requirements for the Binding of Retinol and Retinal. Biochemistry 17, 5353-5359.

Schreckenbach T, Walckhoff B & Oesterhelt D (1978b). Properties of the Retinal Binding Site in Bacteriorhodopsin: Use of Retinol and Retinyl Moieties as Fluorescent Probes. Photochem Photobiol 28, 205-211.

Schulz G E (1996). Porins: General to Specific, Native to Engineered Passive Pores. Curr Opin Struct Biol 6, 485-490.

Schweiger U, Tittor J & Oesterhelt D (1994). Bacteriorhodopsin can Function Without a Covalent Linkage Between Retinal and Protein. Biochemistry 33, 535-541.

Seehra J S & Khorana H G (1984). Bacteriorhodopsin Precursor: Characterisation and its Integration into the Purple Membrane. J Biol Chem 259, 4187-4193.

Serrano L, Horovitz A, Avron B, Bycroft M & Fersht A R (1990). Estimating the Contribution of Engineered Surface Electrostatic Interactions to Protein Stability by Using Double-Mutant Cycles. Biochemistry 29, 9343-9352.

Serrano L, Matouchek A & Fersht A R (1992). The Folding of an Enzyme: Structure of the Transition State for Unfolding of Barnase Analysed by a Protein Engineering Procedure. J Mol Biol 224, 805-818.

Shakhnovich E I (1997). Theoretical Studies of Protein Folding Thermodynamics and Kinetics. Curr Opin Struct Biol 7, 29-40.

Sharkov A V, Pakulev A V, Chekalin S V & Matveetz Y A (1985). Primary Events in Bacteriorhodopsin Probed by Subpicosecond Spectroscopy. Biochim Biophys Acta 808, 94-102.

Shortle D R (1996). Structural Analysis of Non-Native States of Proteins by NMR Methods. Curr Opin Struct Biol 6, 24-30.

Siebert F (1990). Resonance Raman and Infra-Red Difference Spectroscopy of Retinal Proteins. Meth Enzymol 189, 123-136.

Sifers R N (1995). Defective Protein Folding as a Cause of Disease. Nat Struct Biol 2, 355-357.

Sinclair J F, Ziegler M M & Baldwin T O (1994). Kinetic Partitioning During Protein Folding Yields Multiple Native States. Nat Struct Biol 1, 320-326.

Smith S O, Pardoen J A, Mulder P P J, Curry B, Lugtenburg J & Mathies R (1983). Chromophore Structure in Bacteriorhodopsin's O640 Photointermediate. Biochemistry 22, 6141-6148.

Soekarjo M, Eisenhawer M, Kuhn A & Vogel H (1996). Thermodynamics of the Membrane Insertion Process of the M13 Procoat Protein, a Lipid Bilayer Traversing Protein Containing a Leader Sequence. Biochemistry 35, 1232-1241.

Sosnick T R, Mayne L, Hiller R & Englander S W (1994). The Barriers in Protein Folding. Nat Struct Biol 1, 149-156.

Sosnick T R, Mayne L & Englander S W (1996). Molecular Collapse: The Rate-Limiting Step in Two- State Cytochrome c Folding. PROTEINS: Struct Func Gen 24, 413-426.

Sowers K R & Schreier H J (1995). Archae: A Laboratory Manual - Methanogens. Cold Spring Harbor Laboratory Press, New York.

Stanier R Y, Ingraham J L, Wheelis M L & Painter P R (1987). General Microbiology (5th edition). Macmillan Education Ltd, London.

Stellwagen E, Rysavy R & Babul G (1972). The Conformation of Horse Heart Apocytochrome c. J Biol Chem 247, 8074-.

Stern L J & Khorana H G (1989). Structure-Function Studies on Bacteriorhodopsin: Individual Substitutions of Arginine Residues by Glutamine Affect Chromophore Formation, Photocycle and Proton Translocation. J Biol Chem 264, 14202-14208.

Stoeckenius W & Rowen R (1967). A Morphological Study of Halobacterium Halobium and its Lysis in Media of Low Salt Concentration. J Cell Biol 34, 365-393.

Stoeckenius W & Kunau W H (1968). Further Characterization of Particulate Fractions from Lysed Cell Envelopes of Halobacterium Halobium and Isolation of Gas Vacuole Membranes. J Cell Biol 38, 337-357.

Stoeckenius W, Lozier R H & Bogomolni R A (1979). Bacteriorhodopsin and the Purple Membrane of Halobacteria. Biochim Biophys Acta 505, 215-278.

Stoeckenius W & Bogomolni R A (1982). Bacteriorhodopsin and Related Pigments of Halobacteria. Ann Rev Biochem 52, 587-616.

Stoeckenius W (1985). The Rhodopsin-Like Pigments of Halobacteria: Light-Energy and Signal Transducers in an Archaebacterium. Trend Biochem Sci 10, 483-486.

Stuhmer W, Conti F, Suzuki H, Wang X, Noda M, Yahagi N, Kubo H & Numa S (1989). Structural Path Involved in Activation and Inactivation of the Sodium Channel. Nature 339, 597-603.

Sudhakar K, Phillips C M, Owen C S & Vanderkooi J M (1995). Dynamics of Parvalbumin Studied by Fluorescence Emission and Triplet Absorption Spectroscopy of Tryptophan. Biochemistry 34, 1355-1363.

Sugiyama Y & Mukohata Y (1996). Dual Roles of DMPC and CHAPS in the Refolding of Bacterial In Vitro. J Biochem 119, 1143-1149.

Surrey T & Jahnig F (1992). Refolding and Oriented Insertion of a Membrane Protein into a Lipid Bilayer. Pro Nat Aca Sci 89, 7457-7461.

Surrey T & Jahnig F (1995). Kinetics of Folding and Membrane Insertion of a b-Barrel Membrane Protein. J Biol Chem 270, 28199-28203.

Surrey T, Schmid A & Jahnig F (1996). Folding and Membrane Insertion of the Trimeric b-Barrel Protein OmpF. Biochemistry 35, 2283-2288.

Takahashi S, Yeh S R, Das T K, Chan C K, Gottfried D S & Rousseau D L (1997). Folding of Cytochrome c Initiated by Submillisecond Mixing. Nat Struct Biol 4, 44-50.

Talluri S, Rothwarf D M & Scheraga H A (1994). Structural Charaterization of a Three-Disulfide Intermediate of Ribonuclease A Involved in Both the Folding and Unfolding Pathways. Biochemistry 33, 10437-10449.

Tang W-J, Krupinski J & Gilman A G (1991). Expression and Characterization of Calmodulin-Activated (Type I) Adenylate Cyclase. J Biol Chem 266, 8595-8603.

Thomas R M, Feeney J, Nicholson R B, Pain R H & Roberts G C K (1983). Identification by NMR Spectroscopy of a Stable Intermediate in the Folding of Staphylococcal b-Lactamase. Biochem J 215, 525-529.

Thurmond R L, Creuznet C, Reeves P J & Khorana H G (1997). Structure and Function in Rhodopsin: Peptide Sequences in the Cytoplasmic Loops of Rhodopsin are Intimately Involved in Interaction With Rhodopsin Kinase. Pro Nat Aca Sci 94, 1715-1720.

Tocanne J-F, Cezanne L, Lopez A, Piknova B, Schram V, Tournier J-F & Welby M (1994). Lipid Domains and Lipid/Protein Interactions in Biological Membranes. Chem Phys Lipid 73, 139- 158.

Tsukihara T, Aoyama H, Yamashita E, Tomizaki T, Yamaguchi H, Shinzawa-Itoh K, Nakashima R, Yaono R & Yoshikawa S (1995). Structure of Metal Sites of Oxidized Bovine Heart Cytochrome c Oxidase at 2.8 Å. Science 269, 1069-1074.

Tsukihara T, Aoyama H, Yamashita E, Tomizaki T, Yamaguchi H, Shinzawa-Itoh K, Nakashima R, Yaono R & Yoshikawa S (1996). The Whole Structure of the 13-Subunit Oxidized Cytochrome c Oxidase at 2.8 Å. Science 272, 1136-1144.

Udgaonkar J B & Baldwin R L (1988). NMR Evidence for an Early Framework Intermediate on the Folding Pathway of Ribonuclease A. Nature 335, 694-699.

Udgaonkar J B & Baldwin R L (1990). Early Folding Intermediate of Ribonuclease A. Pro Nat Aca Sci 87, 8197-8201.

Udgaonkar J B & Baldwin R L (1995). Nature of the Early Folding Intermediate of Ribonuclease A. Biochemistry 34, 4088-4096.

Unger V M, Hargrave P A, Baldwin J M & Schertler G F X (1997). Arrangement of Rhodopsin Transmembrane a-Helices. Nature 389, 203-206.

Unwin N (1993). Nicotinic Acetylcholine Receptor at 9 Å Resolution. J Mol Biol 229, 1101-1124.

Unwin N (1995). Acetylcholine Receptor Channel Imaged in the Open State. Nature 373, 37-43. van der Goot F G, Gonzalez-Manas J M, Lakey J H & Pattus F (1991). A Molten-Globule Membrane Insertion Intermediate of the Pore-Forming Domain of Colicin A. Nature 354, 408-410. van der Steen R, Biesheuvel P, Mathies R & Lugtenburg J (1986). Retinal Analogues with Locked 6-7 Conformations Show That Bacteriorhodopsin Requires the 6-s-Trans Conformation of the Chromophore. J Am Chem Soc 108, 6410-6411.

Varley P G, Gronenborn A M, Christensen H, Wingfield P T, Pain R H & Clore G M (1993). Kinetics of Folding of the All-b Sheet protein Interleukin-1b. Science 260, 1110-1113.

Veeraraghavan S, Holzman T F & Nall B T (1996). Autocatalyzed Protein Folding. Biochemistry 35, 10601-10607. von Heijne G & Manoil C (1990). Membrane Proteins: From Sequence to Structure. Prot Eng 4, 109- 112.

Walmsley A R, Martin G E M & Henderson P J F (1994). 8-Anilino-1-Naphthalenesulfonate is a Fluorescent Probe of Conformational Changes in the D-Galactose-H+ Symport Protein of Escherichia Coli. J Biol Chem 269, 17009-17019.

Weiss M S, Abele U, Weckesser J, Welte W, Schiltz E & Schulz G E (1991). Molecular Architecture and Electrostatic Properties of a Bacterial Porin. Science 254, 1627-1630.

Weiss M S & Schulz G E (1992). Structure of Porin Refined at 1.8 Å Resolution. J Mol Biol 227, 493- 509.

Weissman J S (1995). All Roads Lead to Rome? The Multiple Pathways of Protein Folding. Chem Biol 2, 255-260.

Weissman J S & Kim P S (1995). A Kinetic Explanation for the Rearrangement Pathway of BPTI Folding. Nat Struct Biol 2, 1123-1130.

Wetlaufer D B (1973). Nucleation, Rapid Folding and Globular Intrachain Regions in Proteins. Pro Nat Aca Sci 70, 697-701.

Wetlaufer D B (1990). Nucleation in Protein Folding - Confusion of Structure and Process. Trend Biochem Sci 15, 414-415.

Whitley P, Nilsson I & von-Heijne G (1994). De Novo Design of Integral Membrane Proteins. Nat Struct Biol 1, 858-862.

Wickner W (1979). The Assembly of Proteins into Biological Membranes: The Membrane Trigger hypothesis. Ann Rev Biochem 48, 23-45.

Wickner W (1988). Mechanisms of Membrane Assembly: General Lessons from the Study of M13 Coat Protein and Escherichia Coli Leader Peptidase. Biochemistry 27, 1081-1086.

Wickner W, Driessen A J M & Hartl F-U (1991). The Enzymology of Protein Translocation Across Escherichia Coli Membrane. Ann Rev Biochem 60, 101-124.

Williams D C, Frank R M V, Muth W L & Burnett J P (1982). Cytoplasmic Inclusion Bodies in Escherichia Coli Producing Biosynthetic Human Insulin Proteins. Science 215, 687-693.

Williams K, Causgrove T P, Gilmanshin R, Fang K S, Callender R H, Woodruff W H & Dyer R B (1996). Fast Events in Protein Folding: Helix Melting and Formation in a Small Peptide. Biochemistry 35, 691-697.

Woese C R & Fox G E (1977). Phylogenetic Structure of the Prokaryotic Domain: The Primary Kingdoms. Pro Nat Aca Sci 74, 5088-5090.

Wolff E K, Bogomolni R A, Scherrer P, Hess B & Stoeckenius W (1986). Color Discrimination in Halobacteria: Spectroscopic Characterization of a Second Sensory Receptor Covering the Blue- Green Region of the Spectrum. Pro Nat Aca Sci 83, 7272-7276.

Wolynes P G, Onuchic J N & Thirumalai D (1995). Navigating the Folding Routes. Science 267, 1619- 1620.

Wrubel W, Stochaj U, Sonnewald U, Theres C & Ehring R (1990). Reconstitution of an Active Lactose Carrier In Vivo by Simultaneous Synthesis of Two Complementary Protein Fragments. J Bact 172, 5374-5381.

Wuthrich K, Billeter M & Braun W (1984). Polypeptide Secondary Structure Determination by Nuclear Magnetic Resonance Observation of Short Proton-Proton Distances. J Mol Biol 180, 715-740.

Xia D, Yu C A, Kim H, Xia J-Z, Kachurin A M, Zhang L, Yu L & Diesenhofer J (1997). Crystal Structure of the Cytochrome bc1 Complex from Bovine Heart Mitochondria. Science 277, 60- 66.

Yao M & Bolen D W (1995). How Valid are Denaturant-Induced Free Energy Measurements - Level of Conformance to Common Assumptions Over an Extended Range of Ribonuclease A Stability. Biochemistry 34, 3771-3781.

Yeates T, Komiya H, Rees D, Allen J & Feher G (1987). Structure of the Reaction Center from Rhodobacter Sphaeroides R-26: Membrane-Protein Interactions. Pro Nat Aca Sci 84, 6438- 6442.

Yeh S R, Takahashi S, Fan B & Rousseau D L (1997). Ligand Exchange During Cytochrome c Folding. Nat Struct Biol 4, 51-56.

Yoshizawa T & Kuwata O (1991). Iodopsin: A Red-Sensitive Cone Visual Pigment in the Chicken Retina. Photochem Photobiol 54, 1061-1070.

Zen K H, McKenna E, Bibi E, Hardy D & Kaback H R (1994). Expression of Lactose Permease in Contiguous Fragments as a Probe for Membrane-Spanning Domains. Biochemistry 33, 8198- 8206.

Zimanyi L, Varo G, Chang M, Ni B, Needleman R & Lanyi J K (1992). Pathways of Proton Release in the Bacteriorhodopsin Photocycle. Biochemistry 31, 8535-8543.

Zimanyi L, Cao Y, Needleman R, Ottolenghi M & Lanyi J K (1993). Pathway of Proton Uptake in the Bacteriorhodopsin Photocycle. Biochemistry 32, 7669-7678.

Zingsheim H P, Neugebauer D-C & Henderson R (1978). Properties of the Two Sides of the Purple Membrane Correlated. J Mol Biol 123, 275-278.

Zwanzig R, Szabo A & Bagchi B (1992). Levinthal's Paradox. Pro Nat Aca Sci 89, 20-22.