General Principles

BIOMEMBRANES A M u l t i - V o l u m e Treatise Volume 1

9 1995

GENERAL PRINCIPLES

This Page Intentionally Left Blank

BIOMEMBRANES A Multi-Volume Treatise GENERAL PRINCIPLES

Editor: A. G. LEE Department of Biochemistry University of Southampton Southampton, England VOLUME

1 * 1995

@ Greenwich, Connecticut

JAI PRESS INC.

London, England

Copyright 91995 by JAI PRESSINC 55 Old Post Road, No. 2 Greenwich, Connecticut 06836 JAI PRESSLTD. The Courtyard 28 High Street Hampton Hill, Middlesex TW12 1PD England All rights reserved. No part of this publication may be reproduced, stored on a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, filming, recording, or otherwise, without prior permission in writing from the publisher. ISBN: 1-55938-658-4 Manufactured in the United States of America

CONTENTS

LIST OF CONTRIBUTORS

xi

PREFACE A. G. Lee

ix

THE FUNCTIONAL ROLES OF LIPIDS IN BIOLOGICAL MEMBRANES

David B. Fenske, Myrna A. Monck, Michael J. Hope, and Pieter R. Cullis PRINCIPLES OF MEMBRANE PROTEIN STRUCTURE

M. S. P. Sansom and lan D. Kerr

29

FATTY ACID- AND ISOPRENOID-LINKED MEMBRANE PROTEINS

Marco Parenti and Anthony I. Magee

79

THE BIOSYNTHESIS OF MEMBRANE PROTEINS

David Stephens, Sunita Kulkarni, and Brian Austen

107

SPECIFICITY OF LIPID-PROTEIN INTERACTIONS

Derek Marsh EFFECTS OF LIPID-PROTEIN INTERACTIONS ON MEMBRANE FUNCTION A. G. Lee

137

187

GENERAL PRINCIPLES OF MEMBRANE TRANSPORT

lan C. West

225

MEMBRANE SIGNALING SYSTEMS

C. U. M. Smith

245

This Page Intentionally Left Blank

LIST OF CONTRIBUTORS

Brian Austen

Department of Surgery St. George's Hospital Medical School London, England

Pieter R. Cullis

Liposome Research Unit Department of Biochemistry University of British Columbia Vancouver, Canada

David B. Fenske

Liposome Research Unit Department of Biochemistry University of British Columbia Vancouver, Canada

Michael J. Hope

Skin Barrier Research Laboratory Department of Medicine University of British Columbia Vancouver, Canada

lan D. Kerr

Laboratory of Molecular Biophysics University of Oxford Oxford, England

Sunita Kulkarni

Department of Surgery St. George's Hospital Medical School London, England

A. G. Lee

Department of Biochemistry University of Southampton Southampton, England

Anthony I. Magee

Laboratory of Eukaryotic Molecular Genetics National Institute for Medical Research London, England

vii

viii

LIST OF CONTRIBUTORS

Derek Marsh

Max-Planck-lnstitut ffir Biophysikalische Chemie Gi~ttingen, Germany

Myrna A. Monck

Liposome Research Unit Department of Biochemistry University of British Columbia Vancouver, Canada Departement de Biologie Service de Biophysique Gifsur-Yvette, France

Marco Parenti

Dipartimento di Farmacologia Chemioterapia e Tossicologia Medica Universit~ di Milano Milano, Italy

M. S. P. Sansom

Laboratory of Molecular Biophysics University of Oxford Oxford, England

C. U. M. Smith

Faculty of Life and Health Sciences Aston University Birmingham, England

David Stephens

Department of Surgery St. George's Hospital Medical School London, England

lan C. West

Department of Biochemistry and Genetics University of Newcastle upon Tyne Newcastle upon Tyne, England

PREFACE

Progress in understanding the nature of the biological membrane has been very rapid over a broad front, but still pockets of ignorance remain. Application of the techniques of molecular biology has provided the sequences of a very large number of membrane proteins, and has led to the discovery of superfamilies of membrane proteins of related structure. In turn, the identification of these superfamilies has led to new ways of thinking about membrane processes. Many of these processes can now be discussed in molecular terms, and unexpected relationships between apparently unrelated phenomena are bringing a new unity to the study of biological membranes. The quantity of information available about membrane proteins is now too large for any one person to be familiar with anything but a very small part of the primary literature. A series of volumes concentrating on molecular aspects of biological membranes therefore seems timely. The hope is that, when complete, these volumes will provide a convenient introduction to the study of a wide range of membrane functions. This first volume is devoted to general surveys of the structure and synthesis of membrane proteins and lipids, of the interactions between lipids and proteins, and of the functions of biological membranes. As editor, I wish to thank all the contributors for their efforts and for the staff of JAI Press for their professionalism in seeing everything through to final publication. Anthony G. Lee Southampton, U.K. March 1, 1995

This Page Intentionally Left Blank

THE FUNCTIONAL ROLES OF LIPIDS IN BIOLOGICAL MEMBRANES

David B. Fenske, Myrna A. Monck, Michael J. Hope, and Pieter R. Cullis

I. II. III. IV. V. VI. VII. VIII.

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Lipid Diversity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Lipid Physical Properties and Phase Transitions . . . . . . . . . . . . . . . . . . Lipid Polymorphism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Lipids and M e m b r a n e Fusion . . . . . . . . . . . . . . . . . . . . . . . . . . Orientational Order in Lipid Bilayers . . . . . . . . . . . . . . . . . . . . . . Permeability and the Osmotic Properties of Membranes . . . . . . . . . . . . The Role of Lipids in Protein Function . . . . . . . . . . . . . . . . . . . . . Abbreviations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Biomembranes Volume 1, pages 1-28. Copyright 9 1995 by JAI Press Inc. All rights of reproduction in any form reserved. ISBN: 1-55938-658-4

. . . . . . . . . .

2 2 4 6

11 13

18 . . . . . . . .

20 21 22 23

2

D.B. FENSKE, M. A. MONCK, M. J. HOPE, and P. R. CULLIS I.

INTRODUCTION

Biological membranes surround cells and organelles, divide the interior of eukaryotic cells into distinct compartments, and provide surfaces for the localization of metabolic enzymes, transport proteins, receptors, and various substrates. In addition, membranes are semipermeable barriers which regulate the transport of water, ions, and other metabolites, thereby providing a means of controlling the internal environment. Our basic understanding of membrane structure has changed little since Singer and Nicholson (1972) first proposed the fluid-mosaic model over 20 years ago. Biological membranes are fluid (liquid-crystalline) lipid bilayers, into which proteins can insert or associate at the surface. Until recently, membrane research has focused primarily on the protein components, with the lipid portion viewed as a convenient barrier and environment for enzymes. However, biological membranes contain a wide diversity of lipids, far more than are needed to perform structural functions, and these lipids require elaborate metabolic pathways for their synthesis and transport. This suggests specific roles for the individual lipid components of membranes. Much is known concerning how different lipids affect the physical properties of membranes, and how individual lipids may function in such fundamental processes as fusion. On a more general level, understanding how individual lipids contribute to the overall electrostatic and hydrophobic properties of the membrane is basic to understanding the factors which regulate protein association and insertion, and which result in a sufficiently fluid matrix for the functioning of membrane enzymes. Many comprehensive reviews have been written dealing with the physical properties and functional roles oflipids in membranes (Cullis and De Kruijff, 1979; Cullis et al., 1985; Gruner et al., 1985; Cullis et al., 1986a, b; Lindblom and Rilfors, 1989; Cullis et al., 1990; Seddon, 1990; Cullis and Hope, 1991; Bloom et al., 1991). This review will serve to summarize and update earlier treatments, and thus some familiarity with the different structures and classes of lipids, the common types of model membrane systems and the techniques by which they are formed, and the basic physical properties of bilayers is assumed. This review will focus on lipid polymorphism and its relation to membrane fusion, and the role of membrane order in growth and protein function.

II.

LIPID DIVERSITY

The main classes of lipids found in eukaryotic biological membranes include the glycerophospholipids, the sphingolipids, and cholesterol (Chol) (Cullis and Hope, 1991). Of the former group, phosphatidylcholine (PC) is the major lipid, but phosphatidylethanolamine (PE), phosphatidylserine (PS), phosphatidylinositol (PI), and cardiolipin (CL) are also major lipid species in biological membranes. A representative chemical structure of a common phospholipid, 1-palmitoyl-2-

Roles of Lipids in Biological Membranes

3

O

-O~

II P ~O~CH2~CH2~ I I

§

N(CH3)3 O

O

O

O

II I I CH~CH \

-O~P~O~CH2~CH2~ OH HC~

CH

+ N(CH3) 3

O

NH

I

CH

A B Figure 1. Chemical structures of representative phospholipids POPC (A) and SPM (B).

oleoyl-sn-glycero-3-phosphocholine (POPC), is shown in Figure 1A. The sphingolipids include sphingomyelin (SPM), ceramide (CER), and the glycosphingolipids (GSLs). The structure of SPM is shown in Figure 1B, where its similarity with PC is apparent. CER is a SPM molecule in which the phosphocholine headgroup has been replaced with a hydroxyl group; the GSLs contain carbohydrate headgroups where the number of sugar residues can range from one, in the glucosyland galactosylceramides, to five in the more complex lipids such as ganglioside GM1 (GM1). In procaryotic membranes, the major phospholipids are PE, phosphatidylglycerol (PG), and CL; PC is not usually present (Gurr and Harwood, 1991). Also present in some microorganisms are the glyceroglycolipids, with monoglucosyl- and diglucosyl-diacylglycerol comprising the most abundant species. The galactosyl forms are also found in plant membranes, where they usually form the majority of lipids in organelles such as the chloroplast (Quinn and Williams, 1983). This lipid diversity is increased by variation in fatty acyl-chain content (in chain length and degree of unsaturation), and the presence of minor lipid components. Lipid compositions vary according to cell types, and the composition of the same

4

D.B. FENSKE, M. A. MONCK, M. J. HOPE, and P. R. CULLIS

membrane (e.g., that of the erythrocyte) can vary significantly between species (Lehninger, 1975). Furthermore, the inner and outer monolayers often contain asymmetric lipid distributions (Op den Kamp, 1979). This has been characterized in numerous membranes, where the same trends are observed (for a summary, see Cullis and Hope, 1991). PC and SPM are found primarily on the outer cell surface, and PE and PS are found primarily on the cytosolic side. The glycolipids are exclusively localized on the outer surface of cells. These distributions appear to be maintained by several proteins, such as the aminotranslocase which is responsible for the transport of PE and PS (Devaux, 1991). The location of glycolipids on the outer cell surface is understood in terms of their receptor functions, but the role of the other lipids is not as clear. However, the presence of PS on the outer surface may be a signal of cell senescence (Tanaka and Schroit, 1983), and its localization on the inner surface with PE may be required to maintain a surface capable of fusion in processes such as endocytosis and organdie fusion (Devaux, 1991).

i11. LIPID PHYSICAL PROPERTIES AND PHASE TRANSITIONS Most biological membranes are 'fluid' at physiological temperatures, a requirement for proper function. The fluid-membrane phase usually refers to the liquid-crystalline bilayer phase, although membranes that contain large quantities of Chol can adopt a different fluid-phase known as liquid-ordered (Bloom et al., 1991; see below). How does the organism regulate membrane fluidity? Answers to this question have come from studies of the physical properties of model membranes. Model membranes composed of well-defined lipid species can exist in a number of different phases. The gel-state exists at low temperatures, where hydrophobic interactions between the acyl-chains are maximized. This results in the chains adopting the all-trans conformation where molecular motion is severely restricted; the chains are highly ordered (see Section VI for a discussion of order parameters). Lateral diffusion is extremely slow in the gel-state, with coefficients in the range of 10-9 to 10-11 cm2/s (Wu et al., 1977; Vaz et al., 1982; Schneider et al., 1983; Kapitza et al., 1984). As the temperature is raised above T,n, the gel-to-liquid-crystalline phase transition temperature, the lipid chains melt, forming the fluid liquidcrystalline phase. In this state, the chains are less ordered due to rapid molecular motion (such as rotation about the long molecular axis and trans-gauche isomerization). Lateral diffusion is much more rapid, with coefficients in the range of 10-6 to 10-8 cm2/s (Mackay et al., 1978; Vaz et al., 1982, 1985). For some lipids, a further increase in temperature results in the formation of nonbilayer phases, such as the hexagonal Hn (Cullis and De Kruijff, 1979; Seddon, 1990) and cubic (Lindblom and Rilfors, 1989) phases. This will be discussed more fully in Section IV. These transitions can be monitored by a number of techniques, including nuclear magnetic resonance (NMR), electron spin resonance (ESR), fluorescence, and

Roles of Lipids in Biological Membranes

5

~ 1 kcal*mol-l*K -1

16.0% 19.6%

~

25.0% ~ ~ - ' ~ ' ~ " - - ' - ' - - - " - - " 36.0~ . _ - - . , . - - - - - - - - - " - - - - - - - ' - - - ~ ~ I

10

15

I

I

I

I

i

20 25 30 35 40 Temperature (*C)

I

45

50

Figure 2. Effect of Chol on the gel-to-liquid-crystalline phase transition of SEPC as measured by DSC. The Chol concentration is given in mol %. (Reproduced from Linseisen et al., 1993, with permission.)

differential scanning calorimetry (DSC). All except the latter are spectroscopic techniques. Of the former, NMR is the most useful, and for the study of lipid-phase transitions the most powerful techniques are 2H and 31P NMR (Davis, 1979, 1983; Cullis and De Kruijff, 1979). Both of these are sensitive to the molecular motions present in the different phases, and thus give rise to characteristic spectra for each phase. A more complete discussion of 2H NMR and its application to membranes and polymorphism will be given in Section VI. DSC measures enthalpy changes as a sample undergoes an endothermic or exothermic transition (Biltonen and Lichtenberg, 1993). Some representative DSC scans for 1-stearoyl-2-elaidoyl-snglycero-3-phosphocholine (SEPC) and SEPC:Chol mixtures are shown in Figure 2. Three parameters can be derived from such traces: the transition temperature Tm, characterized by the temperature showing the highest enthalpy; the transition enthalpy, which is proportional to the area under the curve; and the 'cooperativity' of the transition, which can be estimated from the width of the transition. Values of these parameters have been compiled for most common lipids (Biltonen and Freire, 1978; Mabrey and Sturtevant, 1978; McElhaney, 1982; van Osdol et al., 1989).

6

D.B. FENSKE, M. A. MONCK, M. J. HOPE, and P. R. CULLIS

NMR and DSC studies of a wide variety of model systems shed light on the physical state of biological membranes. For instance, the high degree of unsaturation in most lipids is one mechanism whereby the temperature of transition to gel-state is reduced far below the physiological range (see Cullis et al., 1985). PCs with saturated hydrocarbon chains have TmS that increase with chain length, reaching values greater than 40 ~ for acyl-chains greater than 14 carbons in length (Biltonen and Lichtenberg, 1993). For lipids containing monounsaturated chains, the Tm values are reduced by 20--40 ~ Complete abolishment of the main transition in a saturated-lipid system can be accomplished by the presence of high quantities of Chol (Demel and De Kruijff, 1976). Similar behavior has also been observed in unsaturated systems. Figure 2 shows DSC endotherms of mixtures of Chol with the unusual lipid SEPC, which contains a single trans double bond at carbon 9 of the elaidoyl chain (Linseisen et al., 1993). Pure SEPC exhibits a single endothermic peak at 33.2 ~ with an enthalpy of 8.5 kcal/mol. As the concentration of Chol is increased to about 16 mol %, a decrease is observed in both the main transition temperature and the transition enthalpy. Above this concentration the enthalpy continues to decrease, leaving only a broad endothermic feature of unknown origin. When the Chol content is greater than 25 mol %, the system generally undergoes a transition to the liquid-ordered phase, characterized by the existence of highly ordered fluid chains over a wide temperature range (Ipsen et al., 1987; Vist and Davis, 1990). Cholesterol therefore functions to keep the membrane in a fluid environment. An excellent discussion of the effect of Chol from a 2H NMR viewpoint is given in the recent review by Bloom et al. (1991).

IV. LIPID POLYMORPHISM Most biological membranes contain appreciable quantities of lipids which are capable in isolation of forming liquid-crystalline nonbilayer structures. This is referred to as lipid polymorphism. Although numerous nonbilayer phases exist, the predominant one formed by membrane lipids is the hexagonal Hu phase, which consists of lipid cylinders packed in a hexagonal array (Cullis and De Kruijff, 1979; Seddon, 1990). In the Hn phase, the lipids are arranged with the phospholipid headgroups pointing towards the center of the cylinder, forming an aqueous channel with a diameter of approximately 2 nm. Other nonbilayer phases include micellar structures and cubic phases (Lindblom and Rilfors, 1989). Lipid polymorphism has been extensively studied using X-ray diffraction (Luzzati and Husson, 1962; Luzzati et al., 1966; 1968a,b,c; Caffrey, 1985; Gruner et al., 1988; Tate et al., 1992), 2H and 31p NMR (Cullis and De Kruijff, 1979; Sternin et al., 1988; Lafleur et al., 1990a,c; Fenske et al., 1990; 1992; Gawrisch et al., 1992), and freeze-fracture electron microscopy (EM) (Verkleij, 1984; Hope et al., 1989). X-ray diffraction, employed in the early pioneering studies, is perhaps the most powerful technique, allowing elucidation of the precise details of the phase struc-

Roles of Lipids in Biological Membranes

,

~,~

~ PPH

_~'~

,

7

,

~

~

_~'~

,

PPM

Figure 3. 31p NMR as a tool for the characterization of lipid polymorphism. Liquidcrystalline lipid systems can exist as bilayer (A), hexagonal H, (B), or isotropic phase (C) assemblies, or as mixtures of these phases (D-F). (A) multilamellar dispersions of DOPC at 30 ~ (B) Hexagonal phase DOPE:Chol (1:1) at 20 ~ (C) LUVs of 20 mol % PA in DOPC at 20 ~ (D) DOPE:ChoI:DOPC (1:1:2) at 20 ~ The two upfield bilayer peaks originate from PC and PE, and are resolved due to the smaller shielding anisotropy of PE; (E) DOPE:ChoI:DOPC (1:1:1) at 40 ~ where bilayer, Hll, and isotropic phases are resolved; (F) same as (E), but at 60 ~

ture. However, NMR techniques have proven to be convenient and rapid for determining polymorphic phase tendencies of lipid mixtures. The most commonly utilizexl techniques have been 31p NMR, and to a lesser extent 2H NMR, both of which are sensitive to the different motional characteristics of the various lipid phases. Finally, freeze-fracture EM allows visualization of lipid phases, and can reveal irregular variation in local structure which may be unavailable from X-ray or NMR techniques.

8

D.B. FENSKE, M. A. MONCK, M. J. HOPE, and P. R. CULLIS

The 31p NMR lineshapes corresponding to bilayer, HI, and "isotropic" phases (e.g., micelles, vesicles, or cubic phases) are well characterized (Cullis and De Kruijff, 1976, 1979; Seelig, 1978), and representative examples are shown in Figure 3. Large bilayer systems (on the order of microns) exhibit broad, asymmetric 31p NMR spectra with a low-field shoulder and high-field peak, separated by about 40-50 ppm (Figure 3A). HI systems exhibit a reverse asymmetry, and the width is reduced by a factor of two. This is due to additional motional averaging which results from diffusion of phospholipid molecules about the cylinder axis (Figure 3B). The term "isotropic" phase is applied to systems where the motions are sufficiently rapid to average the chemical-shift anisotropy tensor, giving rise to narrow-line, symmetrical NMR spectra. Examples of such systems include micelles, unilamellar vesicles with diameters <100 nm (Figure 3C), and cubic phases (Lindblom and Rilfors, 1989). 31p NMR is also useful in characterizing complex lipid mixtures in which two or more phases coexist, such as occurs during phase-transitions. Examples are given in Figure 3D-F for mixtures of 1,2-dioleoyl-snglycero-3-phosphocholine (DOPC): 1,2-dioleoyl-sn-glycero-3-phosphoethanolamine (DOPE):Chol (see figure legend for details) which exhibit coexistence of bilayer and isotropic phases (Figure 3D); bilayer, HI, and isotropic phases (Figure 3E); and HI and isotropic (Figure 3F) phases. Polymorphic phases can also be identified by freeze-fracture EM (Verkleij, 1984; Hope et al., 1989). Bilayers give rise to fiat, featureless fracture-planes, whereas a regular corrugated pattern is observed with Ha phase. Some "isotropic phase" samples give evidence of small "lipidic particles," which may correspond to interbilayer attachment sites (this is discussed below in Section V). A large variety of pure and mixed lipid systems have been investigated with regard to their polymorphic preferences. A comprehensive tabulation of this data can be found in Cullis et al. (1990). The results from eukaryotic cell phospholipids reveal that a large proportion of membrane lipids can adopt the HH phase under appropriate conditions. Although PC and SPM form only bilayers in isolation, other lipids (PE, PS, PA, and CL) can form both lamellar and Hrt phases depending on the conditions (pH, temperature, and acyl-chain unsaturation). Furthermore, Chol and long-chain unsaturated fatty acids can induce the Ha phase in some lipid mixtures. Of the nonbilayer phase lipids, PE has been the most extensively studied. Depending on the level of unsaturation, many PEs form stable bilayers, undergoing lamellar-to-Hn transitions as the temperature is raised above some critical value. The tendency to form Htt phase is increased with increasing acyl-chain unsaturation. Other lipids, such as PS and phosphatidic acid (PA), will only form Hn phase at sufficiently low pH, where the negative charge on the headgroup is neutralized. Similarly, PA and CL will adopt the Ha phase in the presence of Ca 2+, which neutralizes the negative surface charge. The addition of Ca 2+ to a lipid mixture isolated from human erythrocytes has been shown to trigger the formation of HI phase (Hope and Cullis, 1979). The ability to regulate bilayer-nonbilayer transitions by controlling parameters such as pH and divalent cation concentration is

Roles of Lipids in Biological Membranes

9

clearly important in an isothermal environment, and thus these observations may have biological relevance. Some of the glycolipids found in the membranes of plants and microorganisms also can be divided into bilayer and nonbilayer categories. Diglucosyl diacylglycerol (DGDG) is a bilayer-forming lipid, while monoglucosyl diacylglycerol (MGDG) favors the Ha phase. Observations on pure lipid species, while informative in terms of the factors which modulate polymorphism, are not particularly relevant biologically. The nonbilayer lipids are always present in complex lipid mixtures, and therefore studies on mixed model systems are of interest. In general, increasing the proportion of bilayer-forming lipids results in a progressive stabilization of the bilayer phase, with complete stabilization usually achieved with 20-50% of the bilayer lipid (Cullis and De Kruijff, 1979; Hui et al., 1981; Tilcock et al., 1982; Boni and Hui, 1983; Cullis et al., 1986). The composition of these lipid mixtures remains the same during lamellar-to-nonlamellar transitions, suggesting that the bilayer and nonbilayer lipids are homogeneously dispersed in the two phases. An illustrative example involves lamellar and Ha phases in some PC:PE mixtures. 2H NMR studies on selectively 2H-labeled lipids reveals that the PC and PE are both evenly distributed in the two phases, when one might have predicted that the Ha phase would be enriched in PE and the lamellar phase enriched in PC (Tilcock et al., 1982). This has also been shown using 31p NMR of oriented multibilayers of PC:PE and PC:PE:Chol mixtures, where the resonances originating from the two phospholipids are resolved in the bilayer phase (Fenske and Cullis, 1992). The proportion of PC:PE in the lamellar phase was that of the bulk ratio, and this ratio remained constant during the transition to Ha phase. Increasing the membrane complexity by introducing Chol as a third component is potentially important, due to the ubiquitous distribution of this lipid, and results in some interesting behavior. Cholesterol can induce the formation of HII phase in PE-containing bilayer systems that have been stabilized by PC (Tilcock et al., 1982). Much effort has been directed at understanding the physical basis of lipid polymorphism, i.e., at defining the factors which determine the phase preference of a given lipid species. A qualitative yet successful approach, initially developed by Israelachvili and coworkers (1980), involves consideration of the molecular shapes of individual lipid molecules. This can be simply expressed by means of a shape parameter S: S = V/AoLc

(1)

where V is the hydrocarbon chain volume, Ao is the hydrocarbon-water interfacial area, and Lc is the hydrocarbon chain length. AoLc is the cylinder volume corresponding to the interfacial surface area and hydrocarbon chain length; if V is greater than this value, i.e., S > 1, then the chains must have a cone geometry, which would favor the Hn phase. If V is less than AoLc, then S < 1, and the acyl-chains have an inverted cone geometry. When S = 1, the acyl-chains occupy a cylinder,

10

D.B. FENSKE, M. A. MONCK, M. J. HOPE, and P. R. CULLIS

which favors the bilayer phase. This can be easily understood by expressing the equation above as a function of the cross-sectional areas of the interface (Ao), and of the hydrophobic end of the lipid (Ah) (Cullis et al., 1990): S = 1/311 + (Ah/Ao) 1/2 + Ah/Ao]

(2)

When Ah - Ao, the lipids are cylindrical in shape, and S = 1. For Ah > Ao, the lipids are clearly cone shaped, and S > 1. For Ah < Ao, an inverted cone geometry gives S < 1. Bilayer phase lipids such as PC tend to have a cylindrical geometry, where the cross-sectional area of the headgroup is equivalent to that of the fatty acyl-chains. Lipids which prefer the Htt phase have a cone shape, where the lipid chains occupy a greater cross-sectional area than the headgroup, thereby forcing a curvature to the assembly which favors the hexagonal geometry. Detergent-like lipids, which form micellar structures, have an inverted cone geometry, where the headgroup crosssectional area is greater than that of the chains. The 'shape' of the molecule is determined by all factors which will influence the sizes of the polar and nonpolar regions, and will include such terms as the hydration and charge of the headgroup, possible hydrogen-bonding interactions, the effect of counterions, and the extent of molecular motion. An example of this comes from a comparison of DOPC, which forms a bilayer phase at room temperature, and DOPE, which adopts the hexagonal Ha phase. The latter lipid, with it's smaller, less-hydrated headgroup, adopts a cone shape and favors the Hrt phase. However, as the temperature is reduced to approximately 10 ~ the extent of trans-gauche isomerization is reduced, and the lipid acyl-chains become more ordered. This results in a reduced cross-sectional area being swept out by the acyl-chains, and the shape of the molecule becomes more cylindrical, allowing the aggregate to undergo a hexagonal-to-lamellar transition. Although generally expressed in qualitative terms, at least one recent study provides quantitative support for the shape hypothesis. Lee et al. (1993) examined changes in the midpoint of the La-to-Hu transition of 1-palmitoyl-2-oleoyl-sn-glycero-3' phosphoethanolamine (POPE) in response to the presence of 5 mol % of other lipids with varying headgroup size. This would result in changes in Ao but not Ah. The result was a linear relationship between the headgroup volume (calculated from covalent radii) and the midpoint-temperature of the transition. This adds further support to the idea that molecular shape is important in determining the structural properties of lipid assemblies. A more quantitative approach to understanding the factors involved in the formation of nonbilayer phases has come from the "curvature" concept introduced by Gruner and coworkers (Kirk et al., 1984; Gruner, 1985). In the presence of nonbilayer lipids, a monolayer will tend to curl into cylinders with a spontaneous radius of curvature Ro, essentially executing a lamellar-to-hexagonal transition. However, certain constraints can prevent this transition. The stretching of acylchains is required to prevent the formation of intercylinder spaces in the hydrocarbon matrix of the Ha phase, but this process is energetically unfavorable. The

Roles of Lipids in Biological Membranes

11

addition of hydrocarbon agents such as alkanes allows these spaces to be filled, thereby removing the packing constraints preventing the transition to Ha phase, and allowing the system to adopt its spontaneous radius of curvature, which can be measured by X-ray diffraction. A small radius of curvature indicates a strong tendency to form Hix phase, whereas a larger radius of curvature, such as may occur in a PE/PC mixture, indicates considerably less tendency. The intrinsic radius of curvature thus provides a measure of the polymorphic tendencies of a lipid system. A fascinating example of this involves DOPE and its methylated derivatives (DOPE-Me and DOPC). As expected, the Ro values for DOPE and DOPC were small and large, respectively. However, monomethylated DOPE had an intermediate value, which translated into a propensity to form cubic phases (Gruner et al., 1988). Further advances in modeling lamellar-to-nonlamellar transitions will require quantitative kinetic and thermodynamic data. The transition times for the La to HI transition of several PEs have been measured using time-resolved X-ray diffraction (Caffrey, 1985; Tate et al., 1992). For large temperature jumps, the transitions were reversible, with time constants ranging from 0.1-3 s, in agreement with theoretical values calculated by Siegel (1986a). However, for small temperature jumps, the transition kinetics were much slower, in some cases occurring over several days (Tate et al., 1992). The exchange of lipid between coexisting lamellar and nonlamellar (HI and cubic) phases has been examined using macroscopically oriented samples in conjunction with one- and two-dimensional 31p NMR techniques (Fenske and Cullis, 1992). Although a wide variety of systems exhibited reversible La-to-Hn or La-to-cubic transitions, no exchange of lipid could be detected (on the seconds timescale) between coexisting phases at near-equilibrium conditions. Despite a wealth of information on the polymorphic phase behavior of numerous lipid mixtures in vitro, there is no convincing evidence for the existence of nonbilayer phases in vivo, except perhaps in a transient manner or in certain pathological states (Buchheim et al., 1979). Since the nonbilayer lipids are not required to form a fluid bilayer matrix, and are not expressed as nonbilayer phases, they must be involved in other membrane processes. There are two areas where these lipids may play important roles. One possibility is that some of the intermediate structures of bilayer-to-nonbilayer transitions may function in membrane fusion. Another is that nonbilayer lipids can increase membrane order, which may be important in some situations. Both of these possibilities are discussed below.

V. LIPIDS A N D MEMBRANE FUSION The ability of membranes to fuse is a requirement for such processes as fertilization, cell division, exocytosis, endocytosis, viral infection, and intracellular membrane transport. Much evidence, both experimental and theoretical, suggests that the lipid components of membranes are involved in fusion processes. For example, vesicles

12

D.B. FENSKE, M. A. MONCK, M. J. HOPE, and P. R. CULLIS

can be induced to fuse in the absence of protein with mixing of both the lipid and aqueous contents. Well-characterized systems include small unilamellar vesicles (SUVs) made of saturated PCs near the gel-to-liquid--crystalline phase transition (Lichtenberg et al., 1981), large unilamellar vesicles (LUVs) or SUVs containing PA or PS in the presence of Ca 2§ (Prestegard and O'Brien, 1987), or LUVs containing a mixture of charged and nonbilayer lipids. Thus, LUVs made of certain PE:PS mixtures will fuse upon addition of Ca 2+ to form larger lamellar structures (which exhibit lipidic particle structure via freeze-fracture EM), followed by a transition to Hu phase (Hope et al., 1983). Fusion can also occur in systems in which the PS is replaced by another anionic lipid such as PA. Rapid fusion of LUVs composed of DOPC/DOPE/Pl/1,2-dioleoyl-sn-glycero-3-phosphate (DOPA) occurs in the presence of Ca 2+ (Eastman et al., 1992). These same systems undergo a lamellar-to-hexagonal transition when studied as multilamellar vesicles (MLVs). Certain lipid-soluble fusogens, such as monoolein, are capable of inducing Hn or cubic phase structure in model and biological membranes (Hope and Cullis, 1981; Lindblom and Rilfors, 1989). Topological considerations alone demonstrate that fusion cannot occur without local, transient departures from bilayer morphology at the fusion interface. These observations have led to the hypothesis that fusion may involve membrane lipids with polymorphic capabilities, proceeding via nonbilayer intermediates. Although several models ofbilayer-to-nonbilayer transitions have been proposed (Caffrey, 1985; Gruner, 1985; Siegel, 1984, 1986a,b,c, 1993), the most detailed theories have been developed by Siegel (1984, 1986a,b,c, 1993), who provides a unified description of both lamellar-to-hexagonal and lamellar-to-cubic transitions. Until recently, Siegel proposed that the first fusion intermediates are inverted micellar intermediates (IMI) (Figure 4A), which form between apposed bilayers at temperatures near TH. The short-lived IMI rapidly assemble into either Hrr phase precursors, or into intedamellar attachments (ILA), which are cubic phase precursors. The particular fate of a given lipid mixture depends on the ratio of the area per lipid headgroup in the lamellar and hexagonal phases, or on the spontaneous radius of curvature (Ellens et al., 1989). The transition to Ha phase does not lead to fusion, as leakage of the vesicle contents occurs. The ILA are thought to be fusion intermediates, and have recently been visualized using cryo-transmission EM (Siegel et al., 1989). It is now thought that ILAs give rise to the "lipidic particle" morphology observed with certain lipid systems using freeze-fracture EM (Ellens et al., 1989). Furthermore, the highly curved ILA structure explains the narrow "isotropic" resonances often observed in lipid mixtures by 31p NMR. Isotropic resonances occur in lipid-dispersions containing nonbilayer lipids, and indicate the presence of a structure in which the lipid molecules are able to rapidly sample all possible orientations, such that averaging of the shielding anisotropy occurs (Lindblom and Rilfors, 1989). The only structures which fulfill these requirements are micelles, small vesicles, and cubic phases. Recently, it has been shown that liposome fusion occurs over the same narrow temperature range where isotropic

Roles of Lipids in Biological Membranes r

--

13

::'::~lii~l~l~I?fl/?

1~-'::'::'::~-.'," .... "'-"::':",:'.i

I:".:.'?:'.":"":"."'.',:'.:".'":"'-':.'":'3

!

Stalk

i I

R

I

IMI T'MC

s ion

Pore

IL.A

A

B

F i g u r e 4. Two possible mechanisms of membrane fusion, involving IMI (A) and stalk (B) intermediates, as proposed by Siegel (1993). The fusion pore and ILA are equivalent. See Siegel (1993) for details. (Reproduced from Siegel, 1993, with permission.)

31p NMR resonances are observed (Ellens et al., 1989). Low concentrations of diacylglycerols (2 mol %) are able to lower TH, the temperature at which isotropic resonances appear, and the temperature for fast membrane fusion by 15-20 ~ (Siegel et al., 1989). Siegel has recently proposed a modified theory of membrane fusion which involves the formation of stalk structures (Figure 4B) between apposed bilayers rather than the IMI mentioned above, and claims that this model is energetically more reasonable (Siegel, 1993). The putative stalk structures are transformed into ILA (or fusion pores) under certain conditions, and thus many elements of the two models overlap.

VI. ORIENTATIONAL ORDER IN LIPID BILAYERS Many biological membranes contain significant quantities of bilayer and nonbilayer lipids. One possible role for the nonbilayer lipids, as discussed above, is in membrane fusion. Another role involves the regulation of membrane lipid order,

14

D.B. FENSKE, M. A. MONCK, M. J. HOPE, and P. R. CULLIS

which may be of importance in modulating the activity of certain membrane-bound enzymes. The measurement of hydrocarbon orientational order provides a measure of the relative degree of flexibility of the hydrocarbon chain. As shown later on, this can be related empirically to membrane hydrophobic thickness or to the viability of a living organism. Orientational order can be measured using a variety of techniques, but the following discussion will focus on 2H NMR spectroscopy, which has been widely used for this purpose (Seelig, 1977; Davis, 1983; Bloom et al., 1991). The 2H NMR spectrum of 1-[2H31]palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC-d31) multilamellar dispersions is shown in Figure 5A. The spectrum consists of 15 overlapping "Pake doublets," each of which corresponds to either a methylene group or the terminal methyl group of the fatty acyl-chain. The Pake doublet lineshape arises from random orientations of the C 2 H bond axis with respect to the bulk magnetic field direction. The two maxima in each doublet correspond to lipids located in regions of the bilayer where the angle between the bilayer normal and the external magnetic field is 90 ~ The separation between the maxima, known as the quadrupolar splitting (AVQ), is related to the C-2H bond order parameter SCDby the relation AVQ = (3/4)(e2qQ/h)ScD. SCD = (1/2)(3 1), where denotes the time average of the angular fluctuations of the C-2H bond with respect to the director axis, defined as the normal to the bilayer surface. The order parameter is thus a measure of the angular excursions of the acyl-chains about the surface normal. Over the past two decades, several techniques have been developed or enhanced which facilitate the measurement of SCD in model and biological systems. The problems inherent in the acquisition of undistorted broad-line spectra were overcome by the development of quadrupolar echo pulse methods (Davis et al., 1976). Many experiments using specifically deuteriated systems were employed to determine order parameters at individual carbon atoms (Seelig, 1977); however, the time and effort required to produce a series of specifically acyl-chain-labeled lipids was enormous. The introduction of DePaking methods (Stemin et al., 1983) greatly simplified measurements of order parameters, particularly in systems labeled in many positions. Furthermore, the use of integration methods enabled one to derive a complete order profile from a dePaked spectrum of membrane lipids perdeuteriated in one of the lipid chains (Sternin et al., 1988; Lafleur et al., 1989). These techniques are illustrated in Figure 5, where the powder spectrum of POPC-d31 (Figure 5A, top) has been dePaked to give the 0~ spectrum (Figure 5A, bottom). Each half of the spectrum is integrated and divided into 14 equal areas corresponding to carbons 2-15 (the methyl group splitting is measured directly from the spectrum). The quadrupolar splittings are calculated from each unit area and are used to construct the smoothed order parameter profile (OPP), shown in Figure 5B. The shape of the OPP, which provides a signature of the lipid bilayer phase, consists of a plateau region (C2--C10) in which the order is relatively constant, followed by a rapid decrease in order towards the center of the bilayer. -

I

I

I

,,

--50

--25

0

25

I

0.3

I

I

50

(kHz)

Frequency I

I

I

I

B 9 Q

0.2_

9

m

O Q

I=I

O..I.--

0.0 0

Carbon

Number

(n)

Figure 5. (A) 2H NMR spectrum of POPC-d31, in the liquid-crystalline state (top). Enhanced resolution of the individual Pake doublets is obtained by dePaking the top spectrum, giving the spectrum below, which corresponds to the calculated oriented spectrum for an angle of 0 o between the bilayer normal and the external magnetic field. The DePaked spectrum is integrated to give the smoothed order parameter profile shown in (B) (ll). Also shown in (B) is the smoothed OPP for hexagonal phase POPE (% obtained from the dePaked spectrum of POPE-d31 at 70 ~ The two profiles are normalized, but it should be remembered that the absolute order parameters of the Hil phase are approximately half those of the bilayer phase.

15

16

D.B. FENSKE, M. A. MONCK, M. J. HOPE, and P. R. CULLIS

It has been noted that the shape of the order profile is conserved for a bilayer system. For example, Davis et al. (1980) observed that a more ordered bilayer exhibits a longer plateau region. In addition, temperature variation, a change in headgroup composition, or addition of Chol all affect the shape of the order profile in a predetermined manner. Lafleur et al. (1990b) were able to predict the shape of the profile from the arithmetic average of the order parameters <ScD> determined from all positions. One finds that POPE-d31 bilayers at 60 ~ POPC-d31 bilayers at 20 ~ and POPC-d31:Chol (8:2) bilayers at 40 ~ all exhibit the same order profile. The conservation of the shape of the order profile was convincingly demonstrated by Morrow and Lu (1991) for a series of deuteriated PC bilayers when one adjusted for acyl-chain length at the same reduced temperature. The variation of SCD is quite different in the inverted-hexagonal phase (Stemin et al., 1988; Lafleur et al., 1990c). First, the magnitude of the order parameters is decreased by at least a factor of two due to increased motional averaging resulting from rotation about the long cylinder axis. Second, the plateau region is either very short or nonexistant, giving a relatively linear decrease in order down the length of the chain. To illustrate, the order profile of hexagonal phase POPE-d31 at 70 ~ is shown in Figure 5B, normalized to the order profile of the lamellar phase. The lack of a plateau region occurs because of increased motional freedom (looser packing) of carbon atoms 2 through 8 relative to the bilayer phase. It is of interest that Hn-forming lipids are more ordered when in the lamellar phase than are bilayer-forming lipids (Lafleur et al., 1990c; Monck et al., 1992). This observation may be explained by the increased lateral pressure on the lipid acyl-chains required to maintain a bilayer structure and is further supported by the observations that Hn-forming lipids confer an increase in <ScD> in membrane lipid bilayers when mixed with lamellar-phase lipids (Cullis et al., 1986; Fenske et al., 1990; Lafleur et al., 1990b). In the mycoplasma A. laidlawii, this is a mechanism for modulating <ScD> in the absence of other regulatory factors (Wieslander et al., 1980; Monck et al., 1992). The application of 2H NMR DePaking and integration methods to living systems is illustrated by studies on the microorganismAcholeplasma laidlawii strain B. This organism can be made fatty acid auxotrophic, which means that with suppression of de novo fatty acid biosynthesis, exogenously supplied fatty acids will be incorporated into the membrane lipids. Furthermore, binary mixtures of fatty acids will be incorporated in roughly a 50:50 molar ratio, with the more saturated chain occupying the sn-1 position of the glycerol backbone. Figure 6A shows a representative 2H NMR spectrum for A. laidlawii grown on a fatty acid mixture containing deuteriated-palmitic acid. The spectrum displays the characteristic bilayer order profile. By varying the exogenous fatty acid composition of perdeuteriated-palmitic acid and oleic acid in the range between 80:20 and 20:80 (mol %), outside of which the organism grows poorly or not at all, it was possible to determine the range of order compatible with growth of the microorganism (Monck et al., 1992). This was found to be 0.14 < SCD < 0.18. It was suggested that the

Roles of Lipids in Biological Membranes

17

I

I

I

A

I

--50

I

--25

0.3

I

I

0

Frequency I

I

I

50

25

(kHz) I

I

I

]9 0.2_

I::I

0.1-

0.00 Carbon

Number

(n)

Figure6. (A) 2H NMR spectrum ofA. laidlawiigrownon fatty acid mixture containing perdeuteriated palmitic acid. (B) Smoothed OPP obtained by dePaking and integration of (A).

predominant lipid species, MGDG and DGDG, which in isolation prefer the HH and bilayer phase, respectively, may play an important role in establishing the observed order profile. One may postulate that optimum membrane protein function can only occur within this range of order, perhaps due to a protein requirement for a certain membrane fluidity, or for a certain membrane thickness. A direct relationship between average membrane order <ScD> and membrane thickness has been suggested (Seelig and Seelig, 1974; Ipsen et al., 1990), which leads to the possibility of a correlation between optimal protein function and bilayer thickness.

18

D.B. FENSKE, M. A. MONCK, M. J. HOPE, and P. R. CULLIS

The studies with A. laidlawii demonstrate that membrane order must be regulated within a certain range if growth is to occur. Other studies also demonstrate the necessity for regulating membrane order. Steady-state and time-resolved fluorescence anisotropy spectroscopy is often used to monitor changes in membrane order. A recent study by Behan-Martin et al. (1993) used both techniques to examine changes in order of a brain synaptic membrane fraction from a number of fish, mammalian, and avian species. Using two different fluorescent probes, they observed a striking relationship between membrane order and body temperature, with the membranes of cold-adapted species showing greater disorder than those of warm-adapted species. Similar values of membrane order were observed for all species when the comparison was made at their respective body temperatures, suggesting an optimum range of order compatible with proper brain function. These observations demonstrate conservation of membrane order over a wide range of thermal environments.

VII.

PERMEABILITY AND THE OSMOTIC PROPERTIES OF MEMBRANES

Many of the physico-chemical properties of membranes are a function of the ensemble, and are not attributable to specific lipids or classes of lipids. Put another way, the sum total of interactions between lipids and between lipids and their aqueous environment determine such properties as membrane permeability and the effect of osmotic gradients on membrane structure. This is the most important aspect of membrane function as the ability to act as highly selective semipermeable barriers is essential for cell function and maintenance. The topic of membrane permeability has been reviewed elsewhere (Cullis and Hope, 1991; de Gier, 1993), and will not be discussed in detail here. Instead, we will focus on some recent work regarding the osmotic properties of vesicles, an area of research which may have significant biological relevance. Liquid-crystalline membranes are highly permeable to water, with values of the permeability coefficient (P) in the range of 10-2 to 10-4 cm/s (Deamer and Bramhall, 1986). The permeability coefficients of other nonelectrolytes vary greatly, from 10-1 cm/s for ammonia to 10-6 crn/s for urea and glycerol to 10-11 crn/s for glucose. Lipid bilayers are highly impermeable to most ions, with P values ranging from 10-11 to 10-13 crn/s for ions such as C1- and Na +, respectively (see de Gier, 1993). The exception is H +, for which P is about 10-5 cm/s. These variations in permeability allow the establishment of osmotic gradients and electrochemical potentials across a lipid bilayer, as impermeable agents can be trapped within a vesicle, while semipermeable agents such as water or protons can still cross in response to osmotic or electrochemical potentials. A number of different factors affect the permeability of water, nonelectrolytes, and electrolytes. For example, the surface potential at the membrane surface will

Roles of Lipids in Biological Membranes

19

affect the ability of charged ions to cross. One factor which appears to affect all three classes of permeants mentioned above, is the order of the membrane. Generally, the more ordered the membrane, the less permeable it is. The increase in order can be established either by increasing the saturation of the fatty acyl chains, or by introducing Chol to the system. Gel-state lipids are particularly impermeable, but membranes in which gel and liquid-crystalline lipid coexist can be even more permeable than fluid membranes. This has been attributed to the presence of defects at the gel-liquid boundary. Closed membrane systems containing an impermeable solute are thus susceptible to osmotic forces. For example, a unilamellar liposome encapsulating Na § can be made to swell or shrink by placing it in hypertonic (higher solute concentration) or hypotonic (lower solute concentration) solution, respectively. A few studies have examined the osmotic sensitivity of unilamellar vesicles of different sizes. Lichtenberg et al. (1981 ) examined vesicles prepared using a French pressure cell, with diameters ranging from 20-90 nm. By examining the fractional degree of selfquenching of trapped 6-carboxyfluorescein under conditions of osmotic stress, they found that vesicles with diameters greater than 40 nm were nonspherical under isoosmotic conditions, and could sustain an increase in the volume of the aqueous interior of almost twofold without leakage of the internal contents. Similar results have recently been obtained for much larger LUVs. Mui et al. (1993) characterized the osmotic properties of LUVs prepared by extrusion using cryo-transmission EM. LUVs prepared under isoosmotic conditions were found to be nonspherical, but could be made to "round-up" by placing them in an hypoosmotic medium. When LUVs were exposed to a sufficiently large osmotic differential, lysis was observed by the release of 6-carboxyfluorescein, with a residual osmotic differential remaining after lysis. A membrane tension of 40 dyn/cm was estimated (at 23 ~ from the maximum residual osmotic differentials obtained for LUVs varying in diameter fro m 90 to 340 nm. Several studies have reported that SUVs are osmotically insensitive (Johnson and Buttress, 1973; Milon et al., 1986). Support for the opposite view comes from recent work by Lerebours et al. (1993), who utilized freeze-fracture EM, ESR, and fluorescence spectroscopy to demonstrate that SUVs of diameter 20 nm are osmotically sensitive, swelling or shrinking in response to applied salt gradients of opposite direction. Unlike the results reported for LUVs (Mui et al., 1993), SUV swelling did not result in breakage of the membrane with release of entrapped material. Whether the results of these studies are applicable to the behavior of biological membranes is not known. What they do indicate is that membrane shape is susceptible to osmotic forces, and that membrane curvature may influence the magnitude of the osmotic gradient that can be sustained while maintaining bilayer integrity. The possibility of osmotic forces playing a role in the establishment or maintenance of highly curved membranes of cellular organelles and intracellular membranes is intriguing.

20

D.B. FENSKE, M. A. MONCK, M. J. HOPE, and P. R. CULLIS

VIii.

THE ROLE OF LIPIDS IN PROTEIN FUNCTION

One of the more important roles of membrane lipids is the regulation of membrane protein activity. In some cases, it appears that specific lipids are involved in maintaining the activity of specific proteins; in other cases, protein activity appears to be sensitive to the overall membrane environment, which is modulated by all lipids present. The precise mechanisms involved are not all known, but could involve binding of lipids to protein regulatory sites, or the matching of membrane hydrophobic thickness with protein hydrophobic thickness, as delineated in the "mattress model" of protein-lipid interactions (Mouritsen and Bloom, 1984; Bloom et al., 1991; Nezil and Bloom, 1992; Mouritsen and Bloom, 1993). In this section, we will examine several recent studies in which protein activity is related to variation in membrane properties. The examples cited do not represent an exhaustive review of this field, but merely serve to illustrate an important role of lipids in biological membrane function. Some proteins require specific lipids for maximum activity. Examples include protein kinase C (PKC) and many of the mitochondrial proteins. PKC, which is involved in transmembrane signaling, requires the presence of PS for its activity. Most of the electron transport complexes and several of the mitochondrial translocases require CL, a structurally unique phospholipid found in the inner mitochondrial membrane of eukaryotes, for maximum activity (see Robinson, 1993 and references therein). Although this suggests a direct CL-protein interaction, evidence for the presence of tightly bound CL exists for only two proteins, the ADP/ATP carrier, and the cytochrome c oxidase electron-transport complex. In the case of cytochrome c oxidase, much evidence suggests that two mol of CL binds to high-affinity sites of the bovine heart complex, thereby increasing the rate of electron-transport. The evidence for this view has recently been reviewed by Robinson (1993). Numerous proteins appear sensitive to the overall physical state of the membrane. The activity of the sarcoplasmic reticulum Ca2+-ATPase is modulated by Chol, apparently via changes in membrane fluidity or order (Madden et al., 1979; 1981). PKC is also regulated in this manner. A large number of hydrophobic and amphipathic compounds can alter its activity. Numerous studies have shown that PKC enhancers and inhibitors tend to promote or inhibit the formation of hexagonal phase lipid, respectively, in appropriate model systems (Epand et al., 1991). Thus, PKC is sensitive to the physical state of the membrane, particularly with respect to its propensity to form nonbilayer phases. This suggests that PKC activity may be regulated by membrane order, which is increased by the presence of nonbilayer lipids (see Section VI). Another protein whose activity seems to be influenced by membrane polymorphic characteristics is Gramicidin, a hydrophobic peptide which forms a membranespanning homodimer (linked at the N-terminals) that functions as a cation channel (Wallace, 1990). Depending on the solvent present when Gramicidin is added to

Roles of Lipids in Biological Membranes

21

the membrane, either the channel conformation, or a nonchannel conformation in which the dimer is formed head-to-tail is obtained (Killian et al., 1988a,b; Bano et al., 1989, 1991). Conversion of the nonchannel form to the channel form can be accomplished by heating. Cox et al. (1992) found that the presence of PE in PC bilayers enhanced this conversion rate. In addition, a small amount of Ha phase lipid was detected by 31p NMR in preparations containing the channel form. This led to the proposal that Ha-promoting lipids would induce the nonchannel to channel transition. Other parameters which may influence the kinetics and thermodynamics of Gramicidin dimerization include hydrophobic mismatch between the hydrophobic length of the dimer and the hydrophobic portion of the lipid bilayer (Huang, 1986; Helfrich and Jacobson, 1990; Ring, 1992). As alluded to above, this may be an important factor in optimizing the function of many membrane proteins. It is interesting that some proteins, in the absence of specific (normally required) lipids, have shown good activity levels depending on the thickness of the membrane milieu (Gurr and Harwood, 1991). A specific example of this is the passive glucose transporter of the erythrocyte membrane, for which maximal activity is observed in the presence of PS, but which, in its absence, depends on the length of the lipid acyl-chains (Carruthers and Melchior, 1988). Many other examples can be found in the recent literature of lipid modulation of membrane protein activities. A few examples include the photochemical protein rhodopsin (Gibson and Brown, 1993) and a Mg2+-ATPase from human erythrocytes (Zimmerman and Daleke, 1993). Phospholipase A2 is also regulated by lipid physical properties; of particular interest is the suggestion that lipid lateral phase separation may play a role in modulating the lipase activity (Op den Kamp et al., 1974; Apitz-Castro et al., 1982" Jain et al., 1989; Burack et al., 1993).

ABBREVIATIONS CER, Chol, CL, AVQ, DGDG, DOPA, DOPC, DOPE, DSC, EM, (e2qQ/h), ESR, GM1,

ceramide. cholesterol. cardiolipin. quadrupolar splitting. diglucosyl diacylglycerol. 1,2-dioleoyl-sn-glycero-3-phosphate. 1,2-dioleoyl-sn-glycero- 3-phosphocholine. 1,2-dioleoyl-sn-glycero- 3-phosphoethanolamine. differential scanning calorimetry. electron microscopy. quadrupolar coupling constant. electron spin resonance spectroscopy. ganglioside GM1.

22

D.B. FENSKE, M. A. MONCK, M. J. HOPE, and P. R. CULLIS

glycosphingolipid. hexagonal Hu phase. interlamellar attachments. inverted micellar intermediate. La, lamellar liquid-crystalline phase. LUV, large unilamellar vesicle. MGDG, monoglucosyl diacylglycerol. MLV, multilamellar vesicle. NMR, nuclear magnetic resonance. oPP, order parameter profile. P, permeability coefficient. PA, phosphatidic acid PC, phosphatidylcholine. PE, phosphatidylethanolamine. PG, phosphatidylglycerol. PI, phosphatidylinositol PKC, protein kinase C. PS, phosphatidylserine. POPC, 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine. POPC-d31, 1-[2H31]palmitoyl-2-oleoyl-sn- glycero-3-phosphocholine. POPE, 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoethanolamine. POPE-d31, 1-[2H31]palmitoyl-2-oleoyl- sn-glycero- 3-phosphoethanolamine. Ro, spontaneous radius of curvature. SCD, carbon-deuterium bond order parameter. SEPC, 1-stearoyl-2-el aidoy 1-sn-glycero- 3-phosphocholine. SPM, sphingomyelin. SUV, small unilamellar vesicle. TH, lamellar to hexagonal phase transition temperature. Tm, gel to liquid-crystalline phase transition temperature. TMC, transmonolayer contact. GSL, HII, ILA, IMI,

ACKNOWLEDGMENTS

The authors wish to thank Dr. John Holland and Dr. Thomas D. Madden for permission to publish some of their data prior to publication (the 31p NMR spectra in Figure 3). We also acknowledge the assistance of Dr. Jenifer L. Thewalt, who helped with the preparation of some of the figures.

Roles of Lipids in Biological Membranes

23

REFERENCES Apitz-Castro, R., Jain, M. K., & de Haas, G. H. (1982). Origin of the latency phase during the action of phospholipase A2 on unmodified phosphatidylcholine vesicles. Biochim. Biophys. Acta 688, 349-356. Bano, M. C., Braco, L., & Abad, C. (1989). HPLC study on the 'history' dependence of gramicidin A conformation in phospholipid model membranes. FEBS Lett. 250, 67-71. Bano, M. C., Braco, L., & Abad, C. (1991). Conformational transitions of gramicidin A in phospholipid model membranes. A high performance liquid chromatography assessment. Biochemistry 30, 886-894. Behan-Martin, M. K., Jones, G. R., Bowler, K., & Cossins, A. R. (1993). A near perfect temperature adaptation of bilayer order in vertebrate brain membranes. Biochim. Biophys. Acta 1151, 216-222. Biltonen, R. L., & Freire, E. (1978). Thermodynamic characterization of conformational states of biological macromolecules using differential scanning calorimetry. CRC Crit. Rev. Biochem. 5, 85-124. Biltonen, R. L., & Lichtenberg, D. (1993). The use of differential scanning calorimetry as a tool to characterize liposome preparations. Chem. Phys. Lipids 64, 128-142. Bloom, M., Evans, E., & Mouritsen, O. G. (1991). Physical properties of the fluid lipid-bilayer component of cell membranes: a perspective. Quart. Rev. Biophys. 24, 293-397. Boni, L. T., & Hui, S. W. (1983). Polymorphic phase behaviour of dilinoleoylphosphatidylethanolamine and palmitoyloleoylphosphatidylcholine mixtures. Structural changes between hexagonal, cubic, and bilayer phases. Biochim. Biophys. Acta 73 l, 177-185. Buchheim, W., Drenckhahn, D., & Lullmann-Rauch, R. (1979). Freeze-fracture studies of cytoplasmic inclusions occuring in experimental lipidosis as induced by amphiphilic cationic drugs. Biochim. Biophys. Acta 575, 71-80. Burack, W. R., Yuan, Q., & Biltonen, R. L. (1993). Role of lateral phase separation in the modulation of phospholipase A2 activity. Biochemistry 32, 583-589. Caffrey, M. (1985). Kinetics and mechanism of the lamellar gel/liquid crystal and lamellar/inverted hexagonal phase transition in phosphatidylethanolamine: a real time X-ray diffraction study using synchrotron radiation. Biochemistry 24, 4826--4844. Carruthers, A., & Melchior, D. L. (1988). Effects of lipid environment on membrane transport: the human erythrocyte sugar transport protein/lipid bilayer system. Ann. Rev. Physiology 50, 257-281. Cox, K. J., Ho, C., Lombardi, J. V., & Stubbs, C. D. (1992). Gramicidin conformational studies with mixed-chain unsaturated phospholipid bilayer systems. Biochemistry 3 l, 1112-1118. Cullis, P. R., & De Kruijff, B. (1976). 31p NMR studies of unsonicated aqueous dispersions of neutral and acidic phospholipids. Effects of phase transitions, p2H and divalent cations on the motion in the phosphate region of the polar headgroup. Biochim. Biophys. Acta 436, 523-540. Cullis, P. R., & De Kruijff, B. (1979). Lipid polymorphism and the functional roles oflipids in biological membranes. Biochim. Biophys. Acta 559, 399-420. Cullis, P. R., & Hope, M. J. (1991). Physical properties and functional roles of lipids in membranes. In: Biochemistry of Lipids, Lipoproteins, and Membranes (Vance, D. E., & Vance, J., eds.) pp. 1-41. Elsevier Science Publishers, New York. Cullis, P. R., Hope, M. J., De Kruijff, B., Verkleij, A. J., & Tilcock, C. P. S. (1985). Structural properties and functional roles of phospholipids in membranes. In: Phospholipids and Cellular Regulation (Kuo, J. E, ed.). pp. 1-59. CRC Press, Boca Raton, Florida. Cullis, E R., De Kruijff, B., Verkleij, A. J., & Hope, M. J. (1986a). Lipid polymorphism and membrane fusion. Biochem. Soc. Trans. 14, 242-245. Cullis, P. R., Hope, M. J., & Tilcock, C. P. S. (1986b). Lipid polymorphism and the roles of lipids in membranes. Chem. Phys. Lipids 40, 127-144.

24

D.B. FENSKE, M. A. MONCK, M. J. HOPE, and P. R. CULLIS

Cullis, P. R., Tilcock, C P., & Hope, M. J. (1990). Lipid polymorphism. In: Membrane Fusion (Wilschut, J., & Hoekstra, D., eds.), pp. 35-64. Marcel Dekker, N.Y. Davis, J. H. (1979). Deuterium magnetic resonance study of the gel and liquid crystalline phases of dipalmitoyl phosphatidylcholine. Biophys J. 27, 339-358. Davis, J. H. (1983i. The description of membrane lipid conformation, order and dynamics by 2H NMR. Biochim. Biophys. Acta 737, 117-171. Davis, J. H., Jeffrey, K. R., Bloom, M., Valic, M. I., & Higgs, T. P. (1976): Quadrupolar echo deuteron magnetic resonance spectroscopy in ordered hydrocarbon chains. Chem. Phys. Lett. 42,390-394. Deamer, D. W., & Bramhall, J. (1986). Permeability of lipid bilayers to water and ionic solutes. Chem. Phys. Lipids 40, 167-188. de Gier, J. (1993). Osmotic behaviour and permeability properties of liposomes. Chem. Phys. Lipids 64, 187-196. Demel, R. A., & De Kruijff, B. (1976). The function of sterols in membranes. Biochim. Biophys. Acta 457, 109-132. Devaux, P. E (1991). Static and dynamic lipid asymmetry in cell membranes. Biochemistry 30, 1163-1173. Eastman, S. J., Hope, M. J., Wong, K. E, & Cullis, P. R. (1992). Influence of phospholipid asymmetry on fusion between large unilamellar vesicles. Biochemistry 31, 4262-4268. Ellens, H., Siegel, D. E, Alford, D., Yeagle, E L., Boni, L., Lis, L. J., Quinn, E J., & Bentz, J. (1989). Membrane fusion and inverted phases. Biochemistry 28, 3692-3703. Epand, R. M., Epand, R. E, Leon, B. T.-C., Menger, E M., & Kuo, J. E (1991). Evidence for the regulation of the activity of protein kinase C through changes in membrane properties. Bioscience Reports 1l, 59-64. Fenske, D. B., & Cullis, E R. (1992). Chemical exchange between lamellar and non-lamellar lipid phases. A one- and two-dimensional 31p-NMR study. Biochim. Biophys. Acta 1108, 201-209. Fenske, D. B., Jarrell, H. C., Guo, Y., & Hui, S. W. (1990). Effect of unsaturated phosphatidylethanolamine on the chain order profile of bilayers at the onset of the hexagonal phase transition. A 2H NMR study. Biochemistry 29, 11222-11229. Gawrisch, K., Parsegian, V. A., Hajduk, D. A., Tate, M. W., Gruner, S. M., Fuller, N. L., & Rand, R. E (1992). Energetics of a hexagonal-lamellar-hexagonal-phase transition sequence in dioleoylphosphatidylethanolamine membranes. Biochemistry 31, 2856-2864. Gibson, N. J., & Brown, M. E (1993). Lipid headgroup and acyl chain composition modulate the MI-MII equilibrium of rhodopsin in recombinant membranes. Biochemistry 32, 2438-2454. Gruner, S. M. (1985). Intrinsic curvature hypothesis for biomembrane lipid composition: a role for nonbilayer lipids. Proc. Natl. Acad. Sci. USA 82, 3665-3669. Gruner, S. M., Cullis, P. R., Hope, M. J., & Tilcock, C. P. S. (1985). Lipid polymorphism: the molecular basis of non-bilayer phases. Ann. Rev. Biophys. Biophys. Chem. 14, 211-238. Gruner, S. M., Tate, M. W., Kirk, G. L., So, E T. C., Tumer, D. C., & Keane, D. T. (1988). X-ray diffraction study of the polymorphic behaviour of N-methylated dioleoylphosphatidylethanolamine. Biochemistry 27, 2853-2866. Gurr, M. I., & Harwood, J. L. (1991). Lipid Biochemistry. An Introduction (4th edition), pp. 244-294. Chapman and Hall, London. Helfrich, P., & Jakobsson, E. (1990). Calculation of deformation energies and conformations in lipid membranes containing gramicidin channels. Biophys J. 57, 1075-1084. Hope, M. J., & Cullis, P. R. (1979). The bilayer stability of inner monolayer lipids from the human erythrocyte. FEBS Letters 107, 323-326. Hope, M. J., & Cullis, P. R. (1981). The role of nonbilayer lipid structures in the fusion of human erythrocytes induced by lipid fusogens. Biochim. Biophys. Acta 640, 82-90. Hope, M. J., Walker, D. C., & Cullis, E R. (1983). Ca2+ and pH induced fusion of small unilamellar vesicles consisting of phosphatidylethanolamine and negatively charged phospholipids: a freeze fracture study. Biochem. Biophys. Res. Commun. 110, 15-22.

Roles of Lipids in Biological Membranes

25

Hope, M. J., Wong, K. E, & Cullis, P. R. (1989). Freeze-fracture of lipids and model membrane systems. J. Electron Microscopy Techniques 13,277-287. Huang, H. W. (1986). Deformation free energy of bilayer membrane and its effect on gramicidin channel lifetime. Biophys J. 50, 1061-1070. Hui, S. W., Stewart, T. P., Yeagle, P. L., & Albert, A. D. (1981). Bilayer to non-bilayer transition in mixtures of phosphatidylethanolamine and phosphatidylcholine: implications for membrane properties. Arch. Biochem. Biophys. 207, 227-240. Ipsen, J. H., Karlstrom, G., Mouritsen, O. G., Wennerstrom, H. W., & Zuckermann, M. J. (1987). Phase equilibria in the phosphatidylcholine-cholesterol system. Biochim. Biophys. Acta 905, 162-172. Ipsen, J. H., Mouritsen, O. G., & Bloom, M. (1990). Relationships between lipid membrane area, hydrophobic thickness and acyl-chain orientational order. The effects of cholesterol. Biophys J. 57,405-412. Israelachvili, J. N., Marcelja, S., & Horn, R. G. (1980). Physical principles of membrane organization. Quart. Rev. Biophys. 13, 121-200. Jain, M. K., Yu, B., & Kozubek, A. (1989). Binding of phospholipase A2 to zwitterionic bilayers is promoted by lateral segregation of anionic amphiphiles. Biochim. Biophys. Acta 980, 23-32. Johnson, S. M., & Buttress, N. (1973). The osmotic insensitivity of sonicated liposomes and the density of phospholipid-cholesterol mixtures. Biochim. Biophys. Acta 307, 20-26. Kapitza, H. G., Ruppel, D. A., Galla, H. J., & Sackmann, E. (1984). Lateral diffusion of lipids and glycophorin in solid phosphatidylcholine bilayers. Biophys. J. 45, 577-587. Killian, J. A., Nicholson, L. K., & Cross, T. A. (1988a). Solid-state 15N-NMR evidence that gramicidin A can adopt two different backbone conformations in dimyristoylphosphatidylcholine model membrane preparations. Biochim. Biophys. Acta 943,535-540. Killian, J. A., Prasad, K. U., Hains, D., & Urry, D. W. (1988b). The membrane as an environment of minimal interconversion. A circular dichroism study on the solvent dependence of the conformational behavior of gramicidin in diacylphosphatidylcholine model membranes. Biochemistry 27, 4848-4855. Kirk, G. L., Gruner, S. M., & Stein, D. L. (1984). A thermodynamic model of the lamellar to inverse hexagonal phase transition of lipid membrane-water systems. Biochemistry 23, 1093-1102. Lafleur, M., Fine, B., Sternin, E., Cullis, P. R., & Bloom, M. (1989). Smoothed orientational order profile by 2H NMR. Biophys J. 56, 1037-1041. Lafleur, M., Bloom, M., & Cullis, P. R. (1990a). Lipid polymorphism and hydrocarbon order. Biochemistry and Cell Biol. 68, 1-8. Lafleur, M., Cullis, P. R., & Bloom, M. (1990b). Modulation of the orientational order profile of the lipid acyl chain in the La phase. Eur. Biophys. J. 19, 55-62. Lafleur, M., Cullis, P. R., Fine, B., & Bloom, M. (1990c). Comparison of the orientational order of lipid chains in the La and the Hn phases. Biochemistry 29, 8325-8333. Lee, Y. C., Taraschi, T. E, & Janes, N. (1993). Support for the shape concept of lipid structure based on a headgroup volume approach. Biophys J. 65, 1429-1432. Lehninger, A. L. (1975). Biochemistry (2nd edition), p. 303. Worth Publishers Inc., New York. Lerebours, B., Wehrli, E., & Hauser, H. (1993). Thermodynamic stability and osmotic sensitivity of small unilamellar phosphatidylcholine vesicles. Biochim. Biophys. Acta 1152, 49-60. Lichtenberg, D., Freire, E., Schmidt, C. E, Barenholz, Y., Feigner, P. L., & Thompson, T. E. (1981). Effect of surface curvature on stability, thermodynamic behaviour, and osmotic activity of dipalmitoylphosphatidylcholine single lamellar vesicles. Biochemistry 20, 3462-3467. Lindblom, G., & Rilfors, L. (1989). Cubic phases and isotropic structures formed by membrane lipids--possible biological relevance. Biochim. Biophys. Acta 988, 221-256. Linseisen, F. M., Thewalt, J. L., Bloom, M., & Bayerl, T. M. (1993). 2H-NMR and DSC study of SEPC-cholesterol mixtures. Chem. Phys. Lipids 65, 141-149. Luzzati, V., & Husson, F. (1962). The structure of the liquid-crystalline phases of lipid-water systems. J. Cell Biol. 12, 207-219.

26

D.B. FENSKE, M. A. MONCK, M. J. HOPE, and P. R. CULLIS

Luzzati, V., Reiss-Husson, E, Rivas, E., & Gulik-Kryzwicki, T. (1966). Structure and polymorphism in lipid-water systems, and their possible biological implications. Ann. N. Y. Acad. Sci. 137, 409-413. Luzzati, V., Tardieu, A., & Gulik-Kryzwicki, T. (1968a). Polymorphism of lipids. Nature (London) 217, 1028-1030. Luzzati, V., Gulik-Kryzwicki, T., & Tardieu, A. (1968b). Polymorphism of lecithins. Nature (London) 218, 1031-1034. Luzzati, V., Tardieu, A., Gulik-Kryzwicki, T., Rivas, E., & Reiss-Husson, E (1968c). Structure of the cubic phases of lipid-water systems. Nature (London) 220, 485-488. Mabrey, S., & Sturtevant, J. M. (1978). High-sensitivity differential scanning calorimetry in the study of biomembranes and related model systems. Methods Membr. Biol. 9, 237-274. Mackay, A. L., Burnell, E. E., Nichol, C. P., Weeks, G., Bloom, M., & Valic, M. I. (1978). Effect of viscosity on the width of the methylene proton magnetic resonance line in sonicated phospholipid bilayer vesicles. FEBS Lett. 88, 97-100. Madden, T. D., Chapman, D., & Quinn, P. J. (1979). Cholesterol modulates activity of calcium-dependent ATPase of the sarcoplasmic reticulum. Nature 279, 538-541. Madden, T. D., King, M. D., & Quinn, P. J. (1981). The modulation of Ca2+-ATPase activity of sarcoplasmic reticulum by membrane cholesterol. The effect of enzyme coupling. Biochim. Biophys. Acta 641,265-269. McElhaney, R. N. (1982). The use of differential scanning calorimetry and differential thermal analysis in studies of model and biological membranes. Chem. Phys. Lipids 30, 229-259. Milon, A., Lazrak, T., Albrecht, A.-M., Wolff, G., Weill, G., Ourisson, G., & Nakatani, Y. (1986). Osmotic swelling of unilamellar vesicles by the stopped-flow light scattering method. Influence of vesicle size, solute, temperature, cholesterol, and three cx,or-dihydroxycarotenoids. Biochim. Biophys. Acta 859, 1-9. Monck, M. A., Bloom, M., Lafleur, M., Lewis, R. N. A. H., McElhaney, R. N., & Cullis, P. R. (1992). Influence of lipid composition on the orientational order in Acholeplasma laidlawii strain B membranes: a deuterium NMR study. Biochemistry 31, 10037-10043. Morrow, M. R., & Lu, D. (1991). Universal behaviour of lipid acyl chain order: chain length scaling. Chem. Phys. Lett. 182, 435-439. Mouritsen, O. G., & Bloom, M. (1984). Mattress model of lipid-protein interactions in membranes. Biophys J. 46, 141-153. Mouritsen, O. G., & Bloom, M. (1993). Models of lipid-protein interactions in membranes. Annu. Rev. Biophys. Biomol. Struct. 22, 145-171. Mui, B. L.-S., Cullis, P. R., Evans, E. A., & Madden, T. D. (1993). Osmotic properties of large unilamellar vesicles prepared by extrusion. Biophys J. 64, 443-453. Nezil, E A., & Bloom, M. (1992). Combined influence of cholesterol and synthetic amphiphilic peptides upon bilayer thickness in model membranes. Biophys J. 61, 1176-1183. Op den Kamp, J. A. F. (1979). Lipid asymmetry in membranes. Annu. Rev. Biochem. 48, 47-71. Op den Kamp, J. A. E, de Gier, J., & van Deenen, L. L. M. (1974). Hydrolysis of phosphatidylcholine liposomes by pancreatic phospholipase A2 at the transition temperature. Biochim. Biophys. Acta 345, 253-256. Prestegard, J. H., & O'Brien, M. P. (1987). Membrane and vesicle fusion. Annu. Rev. Phys. Chem. 38, 383-411. Quinn, P. J., & Williams, W. P. (1983). The structural role of lipids in photosynthetic membranes. Biochim. Biophys. Acta 737, 223-266. Ring, A. (1992). Influence of ion occupancy and membrane deformation on gramicidin A channel stability in lipid membranes. Biophys. J. 61, 1306-1315. Robinson, N. C. (1993). Functional binding of cardiolipin to cytochrome c oxidase. J. Bioenergetics Biomembranes 25, 153-163.

Roles of Lipids in Biological Membranes

27

Schneider, M. B., Chan, W. K., & Webb, W. W. (1983). Fast diffusion along defects and corrugations in phospholipid PIs' liquid crystals. Biophys J. 43, 157-165. Seddon, J. M. (1990). Structure of the inverted hexagonal (HI0 phase, and non-lamellar phase transitions of lipids. Biochim. Biophys. Acta 1031, 1-69. Seelig, A., & Seelig, J. (1974). The dynamic structure of fatty acyl chains in a phospholipid bilayer measured by deuterium magnetic resonance. Biochemistry 13, 4839-4845. Seelig, J. (1977). Deuterium magnetic resonance: theory and applications to lipid membranes. Quart. Rev. Biophys. 10, 353-418. Seelig, J. (1978). 31p nuclear magnetic resonance and the head group structure of phospholipids in membranes. Biochim. Biophys. Acta 515, 105-140. Siegel, D. P. (1984). Inverted micellar structures in bilayer membranes. Formation rates and half-lives. Biophys. J. 45, 399-420. Siegel, D. P. (1986a). Inverted micellar intermediates and the transitions between lamellar, cubic, and inverted hexagonal lipid phases. I. Mechanism of the La-HII phase transition. Biophys. J. 49, 1155-1170. Siegel, D. P. (1986b). Inverted micellar intermediates and the transitions between lamellar, cubic, and inverted hexagonal lipid phases. II. Implications for membrane-membrane interactions and membrane fusion. Biophys. J. 49, 1171-1183. Siegel, D. P. (1986c). Inverted micellar intermediates and the transitions between lamellar, cubic, and inverted hexagonal amphiphile phases. III. Isotropic and inverted cubic state formation via intermediates in transitions between Lc~ and Hii phases. Chem. Phys. Lipids 42, 279-301. Siegel, D. P. (1993). Energetics of intermediates in membrane fusion: comparison of stalk and inverted micellar intermediate mechanisms. Biophys J. 65, 2124-2140. Siegel, D. P., Bums, J. L., Chestnut, M. H., & Talmon, Y. (1989). Intermediates in membrane fusion and bilayer/nonbilayer phase transitions imaged by time-resolved cryo-transmission electron microscopy. Biophys. J. 56, 161-169. Singer, S. J., & Nicholson, G. L. (1972). The fluid mosaic model of the structure of cell membranes. Science 175, 720-731. Sternin, E., Bloom, M., & Mackay, A. L. (1983). De-Pake-ing of NMR spectra. J. Magn. Res. 55, 274-282. Stemin, E., Fine, B., Bloom, M., Tilcock, C. P. S., Wong, K. E, & Cullis, P. R. (1988). Acyl chain orientational order in the hexagonal HII phase of phospholipid-water dispersions. Biophys. J. 54, 689-694. Tanaka, Y., & Schroit, A. J. (1983). Insertion of fluorescent phosphatidylserine into the plasma membrane of red blood cells. Recognition by autologous macrophages. J. Biol. Chem. 258, 11335-11343. Tate, M. W., Shyamsunder, E., Gruner, S. M., & D'Amico, K. L. (1992). Kinetics of the lamellar-inverse hexagonal phase transition determined by time-resolved X-ray diffraction. Biochemistry 31, 1081-1092. Tilcock, C. P. S., Bally, M. B., Farren, S. B., & Cullis, P. R. (1982). Influence of cholesterol on the structural preferences of dioleoylphosphatidylethanolamine-dioleoylphosphatidylcholine systems: a phosphorus-31 and deuterium nuclear magnetic resonance study. Biochemistry 21, 4596-4601. van Osdol, W. W., Biltonen, R. L., & Johnson, M. L. (1989). Measuring the kinetics of membrane phase transitions. J. Biochem. Biophys. Methods 20, 1-46. Vaz, W. L.C., Derzko, Z. I., & Jacobson, K. A. (1982). Photobleaching measurements of the lateral diffusion of lipids and proteins in artificial phospholipid bilayer membranes. Cell Surf. Rev. 8, 83-135. Vaz, W. L. C., Clegg, R. M., & Hallmann, D. (1985). Translational diffusion of lipids in liquid crystalline phase phosphatidylcholine multibilayers. A comparison of experiment with theory. Biochemistry 24, 781-786.

28

D. B. FENSKE, M. A. MONCK, M. J. HOPE, and P. R. CULLIS

Verkleij, A. J. (1984). Lipidic intramembranous particles. Biochim. Biophys. Acta 779, 43-92. Vist, M. R., & Davis, J. H. (1990). Phase equilibria of cholesterol/dipalmitoylphosphatidylcholine mixtures: 2H nuclear magnetic resonance and differential scanning calorimetry. Biochemistry 29, 451-464. Wallace, B. A. (1990). Gramicidin channels and pores. Annu. Rev. Biophys. Biophys. Chem. 19, 127-157. Wieslander, A., Christiansson, A., Rilfors, L., & Lindblom, G. (1980). Lipid bilayer stability in membranes. Regulation of lipid composition inAcholeplasma laidlawii as governed by molecular shape. Biochemistry 19, 3650-3655. Wu, E.-S., Jacobson, K., & Papahadjopoulos, D. (1977). Lateral diffusion in phospholipid multibilayers measured by fluorescence recovery after photobleaching. Biochemistry 16, 3936-3941. Zimmerman, M. L., & Daleke, D. L. (1993). Regulation of a candidate aminophospholipid-transporting ATPase by lipid. Biochemistry 32, 12257-12263.

PR! NC! PLES OF MEMB RAN E PROTEIN STRUCTURE

M. S. P. Sansom and lan D. Kerr

I. II.

IlI.

IV.

VI.

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Classification of M e m b r a n e Proteins . . . . . . . . . . . . . . . . . . . . . . . A. Integral and Peripheral M e m b r a n e Proteins . . . . . . . . . . . . . . . . B. Single T M Helices . . . . . . . . . . . . . . . . . . . . . . . . . . . . C. A l l - ~ Topologies . . . . . . . . . . . . . . . . . . . . . . . . . . . . D. All-~ Topologies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . E. " M i x e d " Topologies . . . . . . . . . . . . . . . . . . . . . . . . . . . All-c~ IMPs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A. Experimentally D e t e r m i n e d Structures . . . . . . . . . . . . . . . . . . . B. Topology Prediction . . . . . . . . . . . . . . . . . . . . . . . . . . . C. Structure-Prediction . . . . . . . . . . . . . . . . . . . . . . . . . . . All- I] IMPs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A. Experimentally D e t e r m i n e d Structures . . . . . . . . . . . . . . . . . . . B. Related Structures . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. .

30 31 31 32 33 34 34 34 34 41 48 60 60 63

~13 IMPs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A. Nicotinic Acetylcholine Receptor . . . . . . . . . . . . . . . . . . . . . .

65 65

Biomembranes Volume 1, pages 29-78. Copyright 9 1995 by JAI Press Inc. All rights of reproduction in any form reserved. ISBN: 1-55938-658-4 29

. .

. . . . .

. . . . .

. . . . . .

30

M.S.P. SANSOM and IAN D. KERR

B. Voltage-GatedIon Channels . . . . . . . . . . . . . . . . . . . . . . . . . VI. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

I.

69 71 72 72

INTRODUCTION

Why attempt to elucidate principles of membrane protein structure? First, such principles help to clarify the relationship between the structures of membrane proteins and their diverse functions. Second, an understanding of such principles may enable accurate prediction of membrane protein structures. Sequencing membrane proteins has become relatively straightforward, whereas determination of their three-dimensional (3D) structures remains slow and difficult. Consequently, this predictive aspect is of considerable importance. The following discussion focuses on the structure of integral membrane proteins (IMPs), and in particular, on their transmembrane (TM) domains. Many membrane proteins also possess extensive extramembranous domains. Available evidence suggests the folding of such domains follows the same principles as for soluble proteins. There already exists an extensive literature concerning the latter, to which the interested reader is referred (Schulz and Schirmer, 1979; Chothia, 1984; Branden and Tooze, 1991). High-resolution three-dimensional structures are known for only a handful of integral membrane proteins. It, therefore, may seem premature to discuss principles of membrane protein structure. However, as a result of detailed examination of known 3D structures, and of computational analysis of numerous membrane protein sequences, general principles have started to emerge. The database of high-resolution structures of IMPs is summarized in Table 1. Both X-ray diffraction from three-dimensional crystals (Michel, 1991) and electron diffraction from two-dimensional (2D) crystals (Kuhlbrandt, 1992) have been used to determine these 3D structures. In addition to true IMPs, crystallographic and nuclear magnetic resonance (NMR) structures are known for several amphipathic channel-forming peptides (e.g., melittin, Terwilliger and Eisenberg, 1982; alamethicin, Fox and Richards, 1982; 8-toxin, Tappin et al., 1988) which may be considered as models of helical TM domains (Sansom, 1991, 1993a,b). High-resolution structures have also been determined for a number of membrane-active toxins which provide insights into membrane protein structures (Li, 1992; Parker and Pattus, 1993). Together, these three classes of structure provide sufficient data to enable definition of some general principles. Despite the limited size of the experimental database, prediction of membrane protein structure is more feasible than structure-prediction of soluble proteins. The 2D nature of a bilayer places considerable constraints on possible 3D structures, allowing experimental and theoretical analysis of transbilayer topologies of

Principles of Membrane Protein Structure

Table 1. Protein

Membrane Protein Structures

Method

bacteriorhodopsin halorhodopsin rhodopsin photosynthetic reaction center

31

Resolution ,~a ~, (2D) ~, (2D) ]k

Reference

EM EM EM X-ray

3.5 6.0 9.0 2.3

Henderson et al. (1990) Havelka et al. (1993) Schertler et al. (1993b) Deisenhofer et al. (1985)

X-ray

3.1 ]k

Allen et al. (1987)

EM X-ray X-ray X-ray X-ray EM

3.4 ,~ 6.0/k 1.8/~ 2.0 ,~, 3.0/k 9.0/k

Ktihlbrandt et al. (1994) Krauss et al. (1993) Weiss et al. (1991b) Kreusch et al. (1994) Cowan et al. (1992) Unwin (1993)

(Rhodopseudomonas viridis) photosynthetic reaction center

(Rhodobacter sphaeroides) light harvesting complex photosystem I (Synechococcus) porin (Rhodobacter capsulatus) porin (Rhodopseudomonas blastica) porins OmpF and PhoE (E. coli) nicotinic acetylcholine receptor

Note: a3.5~, resolution in ~,; 10 ,~ resolution in z.

polypeptide chains. Thus, it is possible to generate credible models of IMP structures by combining analysis of amino acid sequences with experimentally derived structural constraints. In the following discussion, the relationship between experimental structures, general principles and structure-prediction will be examined and evaluated.

il. CLASSIFICATION OF MEMBRANE PROTEINS A. Integral and Peripheral Membrane Proteins There are two major types of membrane proteinsmintegral and peripheral. Integral proteins interact strongly with membranes as a result of their polypeptide chains spanning the bilayer one or more times, whereas peripheral proteins interact more weakly with the membrane surface. Peripheral proteins may form electrostatic interactions with phospholipid headgroups or may bind to integral proteins. Peripheral proteins are thought to obey the same structural principles as soluble proteins, and so will not be discussed further. The key structural feature of IMPs is that their polypeptide chains span lipid bilayers, thus excluding those regions of the chains from contact with water molecules. This has two structural consequences: (a) membrane-spanning regions of polypeptide chains adopt a defined secondary structure, often (~-helical; and (b) membrane-spanning sequences are predominantly hydrophobic. If a membranespanning polypeptide was to adopt a random coil conformation, exclusion of water molecules from the bilayer would result in unfulfilled H-bonds, at an energetic cost

32

M.S.P. SANSOM and IAN D. KERR

A

R /)

B

Figure 1. Schematic diagrams of possible transbilayer topologies of IMPs. Cylinders represent (x-helices and arrows represent I~-strands. A represents those membrane proteins which possess a single TM helix. B represents all-c~ IMPs, with multiple a-helices crossing the bilayer. C shows an all-13 topology, and D a "mixed" ~13 topology, with both c~-helical and 13-sheetTM domains.

of ca. +6 kcal/mol/H-bond. Thus, simple thermodynamic arguments favor defined secondary structures for membrane-spanning chains. The first level of characterization of the structure of an IMP is thus to determine the location in the sequence of the membrane-spanning regions, to establish their secondary structure, and to define how the membrane-spanning elements are threaded back and forth across the bilayer. This is referred to as the transbilayer topology of the polypeptide chain. The requirement for hydrophobic side-chains to interact with the bilayer core aids in prediction of such topologies from amino acid sequences.

B. Single TM Helices The simplest structure of an IMP is a single TM helix (Figure 1A). A 20 residue o~-helix (1.5 A/residue) will span a lipid bilayer of thickness 30 A. Longer or shorter

33

Principles of Membrane Protein Structure

helices may form TM domains if local distortion of the bilayer occurs (Mouritsen and Bloom, 1984, 1993). Side-chains of a single TM helix will be predominantly hydrophobic. Such IMPs may posses extensive globular domains on either one or both faces of the bilayer. The TM domain may simply act as an "anchor" to localize a globular domain in a membrane, or may have a more active role. For example, it has been suggested that intramembrane helix-helix interactions may play a role in cellular signaling (Bormann and Engelmann, 1992). The TM helices of glycophorin, an intensively studied single TM helix protein from erythrocytes, are capable of sequence-specific aggregation within the lipid bilayer phase (Bormann et al., 1989; Lemmon et al., 1992). Thus, even though a single TM helix exhibits a simple topology, possible complexities resulting from oligomerization must be considered. For example, Popot and de Vitry (1990) have suggested that inner mitochondrial and thylakoid membranes contain large numbers of single TM helix IMPs which microassemble within the bilayer to form functional oligomers. Such assemblies will be considered alongside more complex all-or IMPs.

C. All-ix Topologies Many IMPs exhibit more complex topologies, their polypeptide chain crossing the bilayer several times. In such molecules, a considerable fraction of the mass of the protein may be located within the bilayer. Arguably, the most common class of such IMPs are those with an all-tx topology (Figure 1B; Table 2), in which several TM helices span the bilayer. This class of IMPs may be extended to include microassemblies of single TM helix proteins (see above), channel-forming peptides (CFPs), and membrane-active toxins which form transbilayer pores by intramembranous association of TM helices to generate parallel bundles. In all-or IMPs, the TM helices are again generally hydrophobic. However, particularly in ion channel and transport proteins, amphipathic helices may occur which associate to form

Table 2. Classification of Integral Membrane Proteins Class

Description

all-o~

bundle of TM u-helices

all-13 ~13

antiparallel l-barrel o~-helicesand 13-strandsin the sameTM domain

Members

bacteriorhodopsin halorhodopsin rhodopsin photosyntheticreaction centers light harvestingcomplex photosystemI porins nicotinicacetylcholinereceptor K+ channels?

34

M.S.P. SANSOM and IAN D. KERR

bundles in which the hydrophilic surfaces of the helices line a central pore or binding site.

D. AII-~ Topologies Initially, it was suggested that all IMPs might contain predominantly tx-helical TM domains (see e.g., Capaldi, 1982). However, since the determination of high-resolution structures of several porins, it has become evident that a [5-sheet may also yield a stable TM fold (Figure 1C; Table 2). In order to avoid unfulfilled H-bonds at its edges, the sheet is folded back on itself to form a transbilayer [5-barrel. It remains to be seen whether this fold will be found in IMPs other than porins and related proteins.

E. "Mixed" Topologies More recently, two IMP structures have been suggested to contain a "mixed" ct/13 topology. Significantly, both of these structures are for ion channels, possessing a central transbilayer pore. For the nicotinic acetylcholine receptor, it is suggested that a central bundle of TM helices is surrounded by a hydrophobic 13-barrel. For voltage-gated channels, it is suggested that the central pore is formed by a [5-barrel which is, in turn, surrounded by an outer bundle of TM helices. In both cases, the functional protein is an oligomer, with o~-helices and [5-strands donated by each of the subunits. Overall, IMP topologies may be divided into three classes: ct, 13, and ct/13, with the first of these apparently the most widespread. In all cases, membrane-spanning segments are either hydrophobic or amphipathic. Furthermore, intramembranous association of TM domains contributed by different subunits may occur.

i11. ALL-ix IMPs In this section we first discuss all-or IMPs whose structures have been determined at atomic or near-atomic resolution. In particular, we focus on general rules which may be derived from these structures. We then describe how such structural principles may be employed in prediction of transbilayer topologies of all-or IMPs. Finally, we assess attempts to predict the intramembranous packing of TM helices.

A. Experimentally Determined Structures Two high-resolution structures are availablemthose of bacteriorhodopsin (bR) and the photosynthetic reaction center (PS/RC). Lower resolution structures are available for halorhodopsin (hR), bovine rhodopsin, a photosynthetic light harvesting complex (LHC), and photosystem I (PS-I). The number of TM helices observed

Principles of Membrane Protein Structure

35

in these proteins are: bR, hR and rhodopsin, 7; PS/PC, 11; LHC, 3; and PS-I, 21. Thus, the two high-resolution structures contain TM helix bundles of an intermediate level of complexity.

Bacteriorhodopsin The seven-helix bundle structure (Figure 2) of bR was first determined at 7 resolution by Henderson and Unwin (1975). This represented a significant breakthrough in the study of IMP structure. The more recent determination of the structure of bR at near-atomic resolution (Henderson et al., 1990) provides much valuable information concerning the packing of TM helices within IMPs. The basic structure is a compact bundle of seven helices (A to G; Figure 2), which lie approximately perpendicular to the plane of the bilayer and surround an elongated central pocket containing the chromophore retinal (covalently linked to Lys-216 of helix G), and a proton-conducting "channel" to the extracellular surface. Viewed from the extracellular face of the membrane (corresponding to the N-terminus of the polypeptide chain) helices A to G are in a counterclockwise order. The anisotropy of the resolution (3.5 A in xy, 10 A in z, where xy is the plane of the bilayer) results in poorer definition of interhelix loops than of TM helices. The development of 3D crystals of bR which are suitable for X-ray diffraction studies (Schertler et al., 1993a) is expected to overcome the latter problem. The lengths of the TM helices and interhelix loops are given in Table 3. A mean helix length of 23 residues (ca. 35 ,~) is sufficient to completely span the hydrophobic region of a lipid bilayer. The interhelix loops (mean length = 10 residues) are all quite short. Thus, most of the polypeptide chain lies within or very close to the bilayer. Helices B, C and F contain proline residues, resulting in kinked helices. The kink angles are: B, 6.6~ C, 17.3~ and F, 30.2 ~ As discussed below, intrahelical prolines occur with a relatively high frequency in certain classes of IMP, and may play important structural and/or functional roles. Although the TM helices are largely hydrophobic, they contain some polar and charged residues, most of which point toward the interior of the bundle. Tryptophan residues are present at the N-termini of helices A, C, and E, i.e., close to the presumed position of the bilayer/water interface (Schiffer et al., 1992). The packing of helices within bR is summarized in Table 4. All adjacent helix pairs are antiparallel, except for the G-A pair which "closes" the bundle. The mean helix crossing angle for the antiparallel pairs is f~ = - 168 ~ which is consistent with class 3-4 "ridges-in-grooves" helix packing as defined by Chothia et al. (1981). Overall, the shape of the bundle is that of a distorted left-handed supercoil. Interhelix separations are also typical of those found for globular proteins (Reddy and Blundell, 1993). Thus bR appears to obey similar rules of helix-helix interactions to those which apply for globular proteins.

A

B

C

D

B

Figure 2. Bacteriorhodopsin. Two views of the seven-helix bundle are shown. A is down a perpendicular to the plane of the bilayer, showing the s-carbon traces of the seven TM helices. The same helices are shown in "ribbon" format in B, in a view approximately perpendicular to that in A. The individual helices and the N- and C-termini of the polypeptide chain are labeled. Interhelix loops are omitted from both diagrams. (Coordinates taken from entry 1BRD of the Brookhaven PDB; Bernstein et al., 1977). 36

Table 3. AII-a IMPS: TM Helices and Loops Protein bR

Helix

Length

A B C D E F G

23 25 21 20 21 25 23

mean + SD PS/RC

Proline.ga

L1 L2 L3 L4 L5 M1 M2 M3 M4 M5 H1 mean + SD

P 13 P12

P20

Loop A-B B-C C-D D-E E-F F-G

Length 5 17 7 9 9 11

23 + 2

mean + SD

10 + 4

21 29 26 30 27 27 30 26 29 27 26

P4, P 10 P2

L l-L2 L2-L3 L3-L4 L4--L5

28 2 29 25

P22 P2

M l-M2 M2-M3 M3-M4 M4-M5

31 2 29 33

mean + SD

22 + 12

27 + 2

Note: aNumber indicates position of the proline within the TM helix.

Table 4. Packingof TM Helices Protein bR

Helix Pair A-B B-C C-D D-E E-F F-G G-A mean o + SD

PS/RC

LA-LB LB-LC LC-MD MD-LE LE-LD MA-MB MB-MC MC-LD LD-ME ME-MD mean + SD

f~ (o)a

R (it)

-155 -167 +173 -160 -170 -170 +8

8.5 9.7 9.7 9.3 11.5 10.8 10.7

-1680 + 10

10.0 + 0 . 9

-154 -149 -124 -127 -157 - 160 -157 -128 -128 -163

7.9 10.3 11.5 10.4 9.2 8.1 10.4 10.9 9.8 8.8

-145 + 15

9.7 __+1.1

Notes: a~ is the helix crossing-angle and R is the interaxial separation at the closest contact of the helices. For an exactly parallel helix pair, f~ = 0~ for an exactly antiparallel pair, f2 = + 180~ ~ the parallel helix pair (i.e., G-A).

37

38

M.S.P. SANSOM and IAN D. KERR

Halorhodopsin The structure of halorhodopsin, a light driven C1-pump with 32% sequence identity with bR, has been determined at 6/k resolution in projection perpendicular to the plane of the membrane (Havelka et al., 1993). The projection-structure ofhR is almost identical to that ofbR at the same resolution, hR contains seven TM helices which have similar tilt angles, relative to the bilayer normal, to those of bR. Thus the strong sequence similarity between the two proteins results in a common transbilayer architecture.

Rhodopsin A 9 ,~ resolution projection-structure has been determined for bovine rhodopsin (348 residues; Schertler et al., 1993b). As predicted (see below), this reveals some structural homologies with bR in that there are seven TM helices in a closed bundle. However, whereas in bR three of the helices are approximately perpendicular to the bilayer (B, C and Dmsee Figure 2) and four are tilted, in rhodopsin the opposite is true, with four helices perpendicular to the bilayer while three are tilted. Furthermore, the rhodopsin bundle (in cross-section) is less elongated than that of bR. This demonstrates that even when two proteins share a common TM topology, significant differences may occur in the packing of their helices.

Photosynthetic Reaction Center Photosynthetic reaction center (PS/RC) structures have been determined for two closely related species (Deisenhofer et al., 1985; Deisenhofer and Michel, 1989; Allen et al., 1987). The TM helix bundles are similar in both structures and so the following discussion will be concerned with the R. viridis PS/RC. The R. viridis PS/RC is made up of four subunits: L, M, H, and a cytochrome (Figure 3). The latter is a peripheral-membrane protein and so will not be considered further. Indeed, the principal difference between the R. viridis and R. sphaeroides complexes is the absence of a cytochrome subunit from the latter. The extramembranous domain of the H subunit is on the cytoplasmic face of the membrane. Thus the IMP may be considered as composed of three subunits L, M, and H. Within the IMP, there are also four bacteriochlorophylls, two bacteriopheophytins, a quinone, and an Fe 2+ ion. Subunits L (273 residues) and M (323 residues) have very similar folds, each containing five TM helices. These are packed together in the order ABCED (Figure 3b). The H subunit (259 residues) is mainly cytoplasmic, and has a single TM helix. Thus, there are 11 TM helices in all. These form a flattened, open-ended bundle, distinct from the closed bundle of bR, with an approximate twofold axis relating the L and M subunits. The helices are tilted somewhat relative to the plane of the bilayer.

Principles of Membrane Protein Structure

39

A

Cyt

~i!ii~!ii~iil

i i i i~,ii i i i!i i il ~ii~i~i!i!!!~ii!iii

Figure 3. Rhodopseudomonas viridis photosynthetic reaction center. A shows the (x-carbon trace of the complete PS/RC molecule, viewed within the plane of the bilayer. The four subunits (H, L, M, and Cyt) and the approximate location of the lipid bilayer (shaded) are indicated. The eleven TM helices are shown in B (same view as in A) and C (viewed perpendicular to the plane of the membrane). TM helices of the L and M subunits are shown as light and dark grey ribbons respectively, with the single TM helix of the H subunit at an intermediate level of shading. (Coordinate entry 1PRC of the P D B ) . (continued)

The TM helices of PS/RC are on average somewhat longer than those of bR (Table 3). With the exception of the short LB-LC and MB-MC loops, the interhelix loops are much longer than those of bR, reflecting the greater proportion of the polypeptide chain outside the bilayer region, intrahelical prolines are present, resulting in kinks in helices LC and MC. Helices LE and ME also have pronounced kinks, although they do not contain prolines.

40

M.S.P. SANSOM and IAN D. KERR

B

C

A

Figure 3. (continued)

Packing of the TM helices (Table 4) differs slightly from that of bR in that the crossing angles deviate somewhat more from-180 ~ (mean D = -145~ Helices LD, LE, MD, and ME form a classic four-helix bundle motif (Weber and Salemme, 1980) at the center of the molecule. The crossing angles between LD and ME (and MD and LE) are close to those for class 1-4 "ridges-in-grooves" packing. Helixhelix separations are similar to those for bR.

Principles of Membrane Protein Structure

41

Overall, the PS/RC provides an example of a more extended, flatter bundle of TM helices. It also demonstrates how TM helices from different subunits may contribute to such bundles. Interactions between helices from different subunits do not differ in essence from those between helices donated by the same subunit. This has implications with respect to methods of predicting packing of TM helices.

Light HarvestingComplex The light harvesting complex (LHC) associated with photosystem II (PSII) from chloroplast membranes has a molecular weight of 25 kDa and contains 15 chlorophyll molecules. The structure of LHC (Ktihlbrandt and Wang, 1991; Ktihlbrandt et al., 1994) reveals three TM helices. Helices A and B are ca. 30 residues long and form a central pair on an approximate twofold axis. They are tilted at angles of ca. 25 to 30 ~ to the bilayer normal. Helix C sits to one side of the A-B pair, is ca. 21 residues long, and is tilted at ca. 10 ~ to the bilayer normal. The order of the helices in the sequence is BCA. Thus, A and B are approximately parallel. They pack with an interhelical separation of R ca. 9 A and a helix-crossing angle of f~ ca. +60 ~ This unusual helix-crossing angle is a result of helix-chromophore interactions. Two carotenoid molecules sit in the grooves of the supercoil defined by helices A and B and thus play a structural role.

Photosystem-i The 6 ,~ resolution X-ray structure for PS-I from Synechococcus (Krauss et al., 1993) provides an example of a large and complex all-c~ IMP. PS-I is made up of two large subunits, A and B (each ca. 83 kDa) and five small subunits (3-15 kDa), in addition to 45 chlorophyll a molecules, and three 4 Fe-4S clusters. There are 28 o~-helices, of which seven lie approximately parallel to the bilayer, and 21 are TM helices. The latter are tilted at angles ranging from 3 ~ to 30 ~ relative to the bilayer normal. The overall TM helix bundle is elongated, with an approximate twofold axis relating helices a to h of subunit A to helices a' to h' of subunit B. At 6 / k resolution, it is not possible to determine the polarity of the helices or to trace the interconnecting loops. However, it is clear that when extended to higher resolution, the structure of PS-I will greatly increase our knowledge of TM helix packing.

B. Topology Prediction In addition to the experimental structures discussed above, there have been several attempts to develop methods for prediction of TM topologies of all-or IMPs. In the following section we review the theoretical background and implementation of such methods, and illustrate their application to bR and related proteins. We also briefly review some methods for experimental evaluation of predicted topologies.

42

M.S.P. SANSOM and IAN D. KERR

Hydrophobicity Scales The underlying theory of topology predictions for all-(x IMPs is that TM segments consist of hydrophobic (x-helices 20-30 residues long, i.e., sufficiently long to span at least the hydrophobic core of a lipid bilayer. Engelman et al. (1986) analyzed the thermodynamics of insertion of an Ala20 helix into a lipid bilayer. U(aq) ~

~+40

0

H(aq)

~-30

(l)

-70 U(lip) ~ n(lip) where U is the unfolded polypeptide chain and H is the (x-helical conformation, aq and lip are the aqueous and lipid bilayer environments, and the figures are AG values in kcal/mol. Thus, transfer of a preformed Ala20 helix from water to a bilayer is favored by ca.-30 kcal/mol. Ala20 is a relatively hydrophobic helix. To determine the locations of TM helices within an amino acid sequence one needs to estimate the corresponding transfer free energies for helices with different sequences. There have been numerous proposed sets of values of single amino acid hydrophobicities, i.e., hydrophobicity scales (Cornette et al., 1987). The derivation of these has ranged from physico-chemical measurements to statistical analyses of known (globular) protein structures. Four such scales are shown in Table 5. The KD (Kyte and Doolittle, 1982) scale is based on a combination of statistical data on distribution of amino acids between surface-exposed and buried locations in globular protein structures, and of water-vapor partition coefficients for amino acid analogues. The EIS scale (Eisenberg et al., 1984) is a consensus scale, derived from five other scales and normalized to a mean of zero and a standard deviation of one. The GES scale (Engelman et al., 1986) was obtained via calculation of transfer free energies (see above) using hydrophobic terms derived from accessible surface area calculations, and hydrophilic terms derived from calculations of Born and H-bonding energies. Finally, the VH (von Heijne, 1992) scale is derived from results of topology predictions and analysis of 135 TM helices from 24 bacterial IMPs. These five hydrophobicity scales show broad similarities. Positive values correspond to hydrophobic residues, negative values to hydrophilic residues. There are, however, some detailed differences in the ranking of the amino acids, particularly with respect to the values for Trp, Tyr and Pro. It is significant that the former two residues are somewhat amphipathic, and that proline plays an idiosyncratic role in TM helices (see below).

Hydrophobicity Profiles In order to locate possible TM helices within amino acid sequences, one searches for clusters of ca. 20 to 25 predominantly hydrophobic residues. Each distinct

43

Principles of Membrane Protein Structure Table5. Hydrophobicity Scales Scales Residue

EISa

Arg (R) Lys (K) Asp (D) Gin (Q) Asn (N) Glu (E) His (H) Ser (S) Thr (T) Pro (P) Tyr (Y) Cys (C) Gly (G) Ala (A) Met (M) Trp (W) Leu (L) Val (V) Phe (F) Ile (I)

-2.53 -1.50 --0.90 -0.85 -0.78 -0.74 -0.40 -0.18 -0.05 +0.12 +0.26 +0.29 +0.48 +0.62 +0.64 +0.81 +1.06 + 1.08 +1.19 +1.38

Notes:

KDb

-4.5 -3.9 -3.5 -3.5 -3.5 -3.5 -3.2 -0.8 -0.7 -1.6 -1.3 +2.5 -0.4 + 1.8 + 1.9 -0.9 +3.8 +4.2 +2.8 +4.5

GES c

-12.3 -8.8 -9.2 -4.1 -4.8 -8.2 -3.0 +0.6 + 1.2 -0.2 -0.7 +2.0 +1.0 + 1.6 +3.4 +1.9 +2.8 +2.6 +3.7 +3.1

VHa

-2.75 -3.00 -2.30 -1.81 -1.99 -2.44 -2.19 -0.12 -0.08 -0.45 -0.39 +1.81 +0.16 +0.27 +0.14 -0.88 +0.62 +0.72 +0.43 +0.97

aEisenberget al. (1984). bKyte and Doolittle (1982). CEngelman et al. (1986). avon Heijne (1992).

hydrophobic cluster is presumed to correspond to a separate T M helix. Simply plotting the hydrophobicity (hi) of each residue as a function of its position in a sequence (i) results in an extremely noisy graph which is difficult to interpret. This is overcome by smoothing the trace in order to obtain a h y d r o p h o b i c i t y p r o f i l e (Kyte and Doolittle, 1982; Engelman et al., 1986; von Heijne, 1992). Pronounced peaks in the hydrophobicity profile correspond to putative TM helices. Smoothing is achieved by averaging the hydrophobicities of e.g., 10 residues on either side of a central residue. This is achieved by multiplication by a suitable window function, i.e.: j=+n

Hi = s

wjhi+j

(2)

j--~n

where Hi is the smoothed hydrophobicity centered about residue i, h is the hydrophobicity of an individual residue and w is the window function. The window

44

M.S.P. SANSOM and IAN D. KERR

length is 2n + 1. Thus n = 10 yields a 21 residue window suitable for detection of TM helices. Differently shaped window functions may be employed. The simplest is a rectangular window, i.e., w = 1 for all 2n + 1 residues within the window. More complex windows, e.g., a trapezoidal function (von Heijne, 1992), place greater weight on central residues of the window and less on outer residues. This is intuitively attractive in that it permits occasional polar residues to occur at the ends of TM helices, i.e., in bilayer/water interfacial locations.

Application to bR and Related Proteins Hydrophobicity profiles for bacteriorhodopsin and two related proteins, halorhodopsin and bovine rhodopsin, are shown in Figure 4. These profiles were calculated using the EIS hydrophobicity scale, with a trapezoidal 21-residue window function as described by von Heijne (1992). As discussed above, hR shares 32% sequence identity with bR, and exhibits the same TM topology (Oesterhelt and Tittor, 1989; Havelka et al., 1993). The profiles for bR and hR are very similar (Figure 4a). Both show five distinct peaks for helices A to E, and a more complex double peak corresponding to helices F and G. It is interesting to note that helices F and G of bR are amphipathic (see below). By selection of a suitable cutoff level, it is possible to automatically assign at least helices A to E to positions in the bR and hR sequences. Much discussion in the literature has concerned suitable protocols for automatic assignment of TM helices on the basis ofhydrophobicity profiles. One approach (von Heijne, 1992; also see Creighton, 1993) has been to analyze distributions of hydrophobicity profile peak heights, , for TM and non-TM helices. The resultant distribution is bimodal and so allows objective assignment of a cutoff value for . Bovine rhodopsin shows little sequence homology with bR, but both hydrophobicity profile analysis and the 9/~ projection structure (see above) support a seven TM helix bundle. The hydrophobicity profile for rhodopsin (Figure 4b) exhibits seven distinct peaks, corresponding to seven TM helices. Peaks A, B, D, E, and F are quite clear, the other two more complex. This analysis permits assignment of the positions of the TM helices in rhodopsin and also in related G-protein coupled proteins (see below).

Additional Methods Two types of extension to the hydrophobicity profile method have been developed: (a) elaboration of sequence rules allowing more accurate location of hydrophobic TM helices, and (b) development of methods for identification of amphipathic TM helices. A simple extension of the hydrophobicity profile method is the so-called "positive inside" rule. This is derived from the observation that positively charged residues (Arg and Lys) are more abundant in cytoplasmic than extracellular loops

A

B

C

D

E

F

G

2

II ;\

"

#

't

!l

9\

'-t' I\

II\

/,~.1 \ f'i v',"

~ ~1 il ;I

~

~

..t'~-,,-\A"

I \

\ ,

t-

v

\J

.

.

.

.

l

,

,

,

\/ ,

50

i

,

9

9

100

.

*

.

.

.

V .

.

|

150

.

.

200

.

250

r e s i d u e

A

B

C

F

E

D

G

_>,

o nO I-

-4

' 0

,

.

.

.

.

100

, 200

,

,

,

,

i

300

residue

Figure 4. Hydrophobicity profiles calculated using the von Heijne (1992) trapezoidal window algorithm and the EIS hydrophobicity scale (Eisenberg et al., 1984). The upper diagram shows profiles for bacteriorhodopsin (solid line) and halorhodopsin (broken line) superimposed. The lower diagram shows the profile for bovine rhodopsin. Positive hydrophobicity values correspond to hydrophobic regions, negative values correspond to hydrophilic regions. Experimentally determined positions of the seven TM helices (A to G) of bacteriorhodopsin are indicated above its profile. The predicted TM helices are indicated for rhodopsin.

45

46

M.S.P. SANSOM and IAN D. KERR

of all-tz IMPs. von Heijne (1992) has incorporated this rule into a predictive procedure, and demonstrated a high success rate for prediction of topologies of all-tz IMPs from bacterial membranes. A similar rule has been shown to apply to eukaryotic all-tz IMPs (Sipos and von Heijne, 1993). A second extension to the hydrophobicity profile method allows the termini of TM helices to be identified more accurately. As discussed for bR, tryptophan residues are frequently found at or near the termini of TM helices. This observation has been extended to the PS/RC by Schiffer et al. (1992). Sipos and von Heijne (1993) demonstrated that in general, Trp and Tyr residues occur with higher frequency at the termini of TM helices. This rule locates amphipathic aromatic residues in the interfacial region of the bilayer, close to polar lipid headgroups and to "penetrating" water molecules (Wiener and White, 1992). The second major extension to the hydrophobicity profile method is the development of procedures for detection of possible amphipathic TM helices. These are of particular importance in the context of channels and related transport proteins, which are thought to contain central bundles of amphipathic TM helices (Oiki et al., 1990; Sansom, 1991). Methods for detection of amphipathic helices range from simple graphical procedures (e.g., the helical wheel of Schiffer and Edmundson, 1967), to more complex sequence analysis procedures, e.g., estimation of hydrophobic moments (Eisenberg et al., 1982, 1984; Eisenberg, 1984), or Fourier analysis of sequences (Finer-Moore and Stroud, 1984). All such methods search for periodicities in sequences of residue hydrophobicities in order to detect the hydrophilic and hydrophobic faces of amphipathic helices. Perhaps the most widespread of the latter methods is the hydrophobic moment, defined by:

2

2 /1/2

[.t=(I~ihisin(Si)l+[~ihicos(Si)] ,

(3)

where the summation is over e.g., the 21 residues of a TM helix and where 8 - 100 ~ is the angle at which successive side-chains point away from the helix axis. The hydrophobic moment is plotted as a function of the central residue and peak values of kt correspond to the centers of putative amphipathic helices.

Evaluation of Predictions It is essential to evaluate the accuracy of TM topology predictions. There may be cases for which the prediction strategy outlined above fails. Wallace et al. (1986) provide a useful critique of the uninformed use of secondary structure prediction methods designed for globular proteins when analyzing IMPs. The simplest test of a predicted topology is to compare the fraction of the polypeptide chain predicted to occur as TM helices with experimental estimates of the secondary structure composition. Two main methods are available for estimation of the secondary structure content of IMPs" circular dichroism (CD; Johnson,

Principles of Membrane Protein Structure

47

1988) and Fourier transform infrared (FTIR; Braiman and Rothschild, 1988; Surewicz et al., 1993) spectroscopy. The results of such studies should, if nothing else, permit assignment of an IMP to the all-~ or all-13 class, and thus allow choice of an appropriate topology prediction methodology. CD has been used for some time to analyze protein secondary structures. However, when applied to membrane proteins, potential artifacts may occur, which if uncorrected can lead to erroneous estimates of secondary structure. By paying careful attention to environmental influences, such as the presence of detergents and the size of membrane vesicles, it is possible to correct for artifacts and so obtain accurate estimates of secondary structure composition (Swords and Wallace, 1993). Of course, not all of the c~-helices present in an IMP may be TM helices. However, recent developments in deconvolution of CD spectra offer the possibility of distinguishing between peripheral and TM helices in IMPs (Park et al., 1992). FTIR spectroscopy also provides estimates of the o~-helical content of IMPs, allowing distinction between all-c~ and all-I] topologies (Hafts and Chapman, 1992). It is applicable to proteins in a variety of media (Haris and Chapman, 1988), and so is well suited to investigations of membrane proteins. Overall, CD and FTIR enable the presence of a significant content of TM helices within an IMP to be detected. Furthermore, recent developments in both CD (Vogel, 1987; Huang and Wu, 1991) and IR (Frey and Tamm, 1991) spectroscopy applied to oriented planar bilayers allow the orientation of c~-helices relative to the bilayer (i.e., parallel or perpendicular) to be determined, thus providing further tests for proposed TM topologies. Spectroscopic methods estimate the overall content of TM helices within an IMP. There are also a number of techniques which permit experimental investigation of the topology of TM crossing by a polypeptide. Among the more established techniques for achieving this are limited proteolysis, chemical labeling, and the use of antibodies against synthetic peptide epitopes (reviewed by Gennis, 1989). A more recent approach, particularly applicable to analysis of bacterial IMPs, uses gene-fusion techniques (Hennessey and Broome-Smith, 1993). A reporter molecule, which exhibits location (i.e., intracellular vs. extracellular) specific phenotypes, is fused to C-terminally truncated versions of an IMP. Varying the length of the truncated IMP and determining the resultant location of the reporter molecule allows one to map the TM topology. Suitable reporter molecules include alkaline phosphatase, [~-lactamase and ~-galactosidase. For example, all three reporter molecules, when applied to the ArsB protein (the IMP subunit of a bacterial anion-transporting ATPase), yielded internally consistent results concerning the topology of this 11-TM helix IMP (Wu et al., 1992). At present, this approach has been largely confined to E. coli IMPs, but initial investigations of eukaryotic systems have been carried out. In summary, there now exist several experimental techniques for confirmation or rejection of predicted TM topologies. Such topologies, in themselves, provide a valuable conceptual framework for further analysis of membrane proteins. Hydro-

48

M.S.P. SANSOM and IAN D. KERR

phobicity profiles are routinely employed to predict TM topologies of a wide range of all-o~ IMPs. The resultant topological models may then be used to direct e.g., structure/function studies via site-directed mutagenesis.

C. Structure-Prediction

TM Helix Packing The current model for folding of all-~ IMPs is the two-stage model proposed by Popot and Engelman (1990; also see Popot, 1993). In the first stage, independently stable TM helices insert across the bilayer. In the second stage, these helices pack together, without transbilayer topological rearrangements, in order to form the tertiary structure of the intramembranous domain of the protein. Thus, TM o~-helices act as autonomous folding units. This theory has important consequences for structure-predictions, implying that simulation of packing of "preformed" TM helices may mimic the in vivo folding process. Before considering structure-prediction studies, it is valuable to review the evidence upon which this model is based. Two systems have been studied in some depth, bacteriorhodopsin and the E. coli lac permease. A number of studies have indicated that isolated TM helices adopt similar conformations to those in intact proteins, and that in vitro assembly of isolated TM helices into functional proteins may occur within a bilayer environment. The most convincing demonstration of TM helices behaving as autonomous folding units comes from NMR studies of isolated TM fragments of bR. TM helix A was prepared as a proteolytic fragment corresponding to residues 1-36 of bR, and its NMR structure determined in both methanol/chloroform and in SDS micelles (Pervushin and Arseniev, 1992). Residues 8-32 adopted an o~-helical conformation in both environments, whereas the N-terminal region was flexible. Note that in intact bR, helix A runs from residue 10 to 32. Thus, the same sequence that forms helix A in the intact structure adopts a helical conformation in an isolated peptide when the latter is present in a membrane-mimetic environment. Helix B peptide (residues 36-65, prepared synthetically with norleucine substituted for methionine) when present in SDS micelles adopted a helical conformation from residues 41 to 62 (Lomize et al., 1992). Helix B runs from residue 38 to 62 in intact bR. The peptide was kinked at Pro-50, with a kink angle of 27 ~ compared to a corresponding angle of 6.6 ~ in intact bR. Thus, the secondary structure in the intact protein was preserved in the isolated peptide, but the proline-induced kink was somewhat larger. This is understandable given the flexibility of proline-induced kinks. Finally, a proteolytic fragment corresponding to residues 163-231, and therefore containing helices F (167-191) and G (203-225), has been studied in chloroform/methanol (Barsukov et al., 1992). Two helical regions were detected (168-191 and 198-227) which correspond well to TM helices in intact bR. Helix F exhibited a proline-induced kink of 25 ~ comparable to a kink angle of 30 ~ in

Principles of Membrane Protein Structure

49

intact bR. However, in this isotropic solvent, helices F and G did not pack together. This provides evidence that a bilayer environment is required to restrict the topology of packing of TM helices. Studies on bR and on lac permease have demonstrated that fragments containing one or more TM helices may insert into bilayers and pack with their partner fragments so as to generate functional proteins. For example, chymotryptic cleavage of bR between helices B and C yields two fragments which will refold in a bilayer to form functionally active bR, which in turn will form 2D crystals indistinguishable from those of the native protein (reviewed by Popot and Engelman, 1990). Furthermore, the five-helix (C to G) proteolytic fragment will refold with two synthetic peptides corresponding to helices A and B to form functional bR (Kahn and Engelman, 1992). Such results are not limited to bR. For example, a series of studies using pairs of polypeptide fragments of lac permease (from independently cloned fragments of the lac Y gene) reveal that comparable functional association within a bilayer can occur for this 12-helix IMP (reviewed by Kaback, 1992). For example, pairs of fragments corresponding to helices 1-6 and 7-12, to helices 1-2 and 3-12, orto helices 1 and 2-12 will associate within bilayers to form functional permease. Overall, these two sets of investigations provide compelling experimental evidence for the second stage of the two-stage folding model. Overview of Prediction

The results of topological prediction and analysis provide definition of TM helices and how they span the bilayer. Furthermore, if an interhelix loop is relatively short, then two TM helices adjacent in the sequence are likely to be adjacent in the 3D structure. The aim of structure-prediction studies is to take such information and by combining it with additional, experimentally-derived restraints, to generate models of TM helix packing in intact IMPs. What types of restraints may one place upon packing of TM helices within a bundle? If one considers a bundle of seven TM helices, as in bR, a reasonable restraint on possible packing models concerns the orientation of the helices within the bundle, i.e., which surfaces of the helices point towards the interior of the bundle, and which surfaces make contact with surrounding lipid molecules. Some progress has been made in deriving such restraints from sequence information. What information can one expect such predictive studies to generate? Certain aspects of TM helix packing may be answered by simulation studies. For example, do helices pack as a left-handed (as in bR; see Figure 2) or a right-handed bundle? How do helices pack relative to one another, i.e., what are interhelix contacts and helix-crossing angles? To what extent are TM helices tilted relative to the bilayer normal? Even if predictive studies cannot accurately place the side-chains of TM helices, one hopes that simulations of TM helix packing may answer such "lowresolution" structural questions.

50

M.S.P. SANSOM and IAN D. KERR

Overall, the input for simulations of TM helix packing should consist of idealized bundles, with the predicted interior side-chains pointing towards the center. Discrimination between alternative models resulting from simulations of packing relies on geometric and energetic analysis of models, and their comparison with experimental data. In the ensuing discussion, we explore how such methods have been applied to bR and to other all-t~ IMPs.

Helix Orientation In order to simulate packing of TM helices, it is necessary to predict an approximate orientation of the constituent helices within a bundle. That is, for each helix one must predict which residues are buffed within the interior of the bundle and which residues are exposed to the surrounding lipid environment. Two approaches to this problem are described, one based upon comparison of TM sequences from homologous proteins, and one based on analysis of TM helix amphipathicity.

Sequence Variability. Yeates et al. (1987) compared the sequences of TM helices of PS/RC proteins from three closely related species (Rb. sphaeroides, R. viridis, R. capsulata). The overall degree of sequence conservation was 50% for the L and M subunits, but this fell to only 16% for those residues deduced to be exposed to the surrounding lipid on the basis of the R. sphaeroides crystal structure. This resulted in the proposal that lipid-exposed residues of TM helices are poorly conserved, whereas those buried within helix bundles are more highly conserved. This is presumed to reflect the ability of a fluid bilayer to pack around a hydrophobic protein surface regardless of the exact shape of the surface. This initial observation has been extended by employing Fourier transform analysis to identify a surface of maximum variability (and hence a lipid-exposed surface) of TM helices. Komiya et al. (1988) used such analysis to detect periodicity of conserved/nonconserved residues for six species of PS/RC protein, providing a rigorous demonstration of the greater variability of lipid-exposed residues. Rees et al. (1989) extended this analysis to 82 sequences taken from six different families of all-0t IMPs. The relationship between the most variable surface and the most hydrophilic surface (defined by hydrophobic moment analysis) of 35 putative TM helices was examined. The mean angle between the variable and hydrophilic faces of the helices was 129 ~ and the mode of the corresponding angular distribution was ca. 180~ This suggests that interior residues of TM helix bundles are more hydrophilic than exposed residues. However, the mean hydrophobicity of interior residues of TM helices was equivalent to that of the interior residues of helices in hemoglobin. Thus, the interior of a TM helix bundle is of equal hydrophobicity te the interior of a globular protein, whereas the surface of a TM helix bundle is more hydrophobic than its interior, and exhibits considerable sequence variability. Extensions and elaborations of this basic approach have been described. Foi example, Donnelly et al. (1993) used Fourier analysis of sequence variability

Principles of Membrane Protein Structure

51

between 33 PS/RC sequences in order to calculate a substitution table for lipid-facing residues of TM helices. This was employed with some success to predict the orientation of helices within the bR helix bundle. Cronet et al. (1993) analyzed the sequences of bR and three related proteins by dividing each TM helix into eight angular zones, ranging from completely exposed to completely internal. This was used to generate a preference parameter (i.e., preferred angular zone within a TM helix in a bundle) for each of the 20 amino acid residues. Expansion of the sequence and structure databases for all-o~ IMPs will allow further refinement of these types of methods.

Helix Amphipathicity. As discussed above, e.g., Rees et al. (1989) have shown that, in general, the interior of a TM helix bundle is more hydrophilic (less hydrophobic) than the lipid-exposed outer surface. This is expected to be particularly so for IMPs, such as ion channels and transport proteins, in which the center of the bundle provides a relatively hydrophilic pathway whereby an ion may cross a membrane. It, therefore, is possible to predict TM helix orientation by defining the most hydrophilic surface. At the level of an amino acid sequence, this may be achieved by calculation of hydrophobic moments. When applied to bR (see e.g., Cronet et al., 1993) such analysis correctly predicts the orientation of helices A, C, and (to a lesser extent) G, fails completely for helices B, E, and F, and places the hydrophilic face of D at an interfacial location. An alternative approach to defining the hydrophilic surface of a TM helix has been developed by Kerr and Sansom (1993a) in the context of ion channels formed by bundles of amphipathic helical peptides. This method requires an atomic resolution model of a TM helix. This model is used to estimate the potential energy of interaction of a water molecule with the protein while placing the water molecule at successive points on a cylindrical grid centered on the helix. The helix-water interaction energies thus defined are displayed as a hydrophilic surface map (HSM). Calculation of HSMs for the (experimentally determined) TM helices of bR correctly predicted the orientation of helices C, E, and G, and placed the hydrophilic faces of A and D at helix interfaces (Kerr and Sansom, 1993b). The method failed for helices B and E However, helix B is not very amphipathic. When applied to helices of unknown three-dimensional structure, simulated annealing (see below) is used to generate an ensemble of models of a helix. The members of the ensemble differ in their side-chain conformations. The HSM may be calculated for each member of an ensemble, and so the degree of variation in the position of the center of the hydrophilic face may be estimated. An example of this method as applied to 8-toxin (a channel-forming peptide from Staphylococcus aureus) is shown in Figure 5. The sequence of 5-toxin is: fM-A-Q-D-I-I-S-T-I-G-D-L-V-K-W-I-I-D-T-V-N-K-F-T-K-K The alternation of hydrophobic (Roman) and hydrophilic (italic) residues within this sequence results in a helical structure (Tappin et al., 1988) of marked amphi-

Figure 5. Hydrophilic surface map analysis for Staphylococcus aureus &toxin (Kerr and Sansom, 1993a; Kerr, 1994). A shows five models of a &toxin helix, selected from an ensemble of 25 structures generated by SNMD. Sidechains D4, D l 1 and D l 8 are labeled. B is a hydrophilic surface map (HSM) for one of the &toxin models. The HSM is depicted as isopotential lines at -1 0 (dotted), -1 5 (solid) and -20 (bold) kcal/mol. The positions of selected C a atoms are indicated. C shows the results of averaging the HSM across all z values, so as to obtain the average water-helix interaction energy, <E>, as a function of the angle of rotation about the helix axis, $. The solid curve is mean (kSD, dotted) of <E> for the five models shown in A.

Principles of Membrane Protein Structure

53

pathicity. The HSM (Figure 5B) reveals a clearly defined hydrophilic surface generated by aspartate side-chains at positions 4, 11 and 18 of the helix. However, variation of side-chain conformations between members of the ensemble (Figure 5A) results in a 20 ~ range in the position of the center of the hydrophilic face. By suitable averaging of the HSMs (Figure 5C), a hydrophilic face may be defined which has the Co~ atom of residue D 11 at its center. This has been used to develop a preliminary model of ion channel formation by a bundle of six parallel 8-toxin helices which may subsequently be refined by simulated annealing (see below).

Simulation Studies In this section, we describe a number of attempts to simulate packing of TM helices within bundles. After reviewing early studies, and investigations of TM helix dimers, we focus on two areas which have been investigated in more detail: (a) parallel helix bundles of ion channels, and (b) bR and related seven-helix bundle proteins.

Preliminary Investigations. Pioneering studies of packing within TM helix bundles were carried out by Pullman and colleagues (Furois-Corbin and Pullman, 1986a,b; 1987a, b). Energy minimization methods were used to pack together Alan helices, treating the latter as rigid bodies. Initial studies examined the stability of Alan dimers and the contributions of different energy terms to dimer stabilization (Furois-Corbin and Pullman, 1986a,b). The electrostatic stabilization increased with increasing n until it reached a plateau at n = 14, whereas the van der Waals stabilization continued to increase with increasing helix length. The van der Waals interactions resulted in stabilization of both parallel and anti-parallel dimers (-17 a n d - 1 4 kcal mo1-1 respectively, for n = 14), despite the unfavorable electrostatic interactions in the former. For both dimers, the interaxial separation of the helices was ca. 8 *. A range of Ala14 helix bundles made up of N alternating antiparallel helices (PN for N = 3 to 7) were examined as possible models of transbilayer pores. The P3 bundle had no central pore, the P4 bundle had a narrow central pore, while the P5 bundle had a pronounced pore running throughout the length of the bundle. The P6 and P7 bundles, initially possessing 6x and 7x symmetry, were considerably distorted during energy minimization, the helix bundles "collapsing" inwards so as to prevent formation of a central pore. At a superficial level the "collapsed" P7 bundle resembled the seven-helix bundle of bR and related proteins. TM Helix Dimers. More recent investigations have employed more powerful simulation methods. In particular, simulated annealing by restrained molecular dynamics (SA/MD) has been used to explore possible packing of TM helix dimers. The SA/MD method is closely related to that used in the determination of globular protein structures on the basis of nuclear Overhauser effect (NOE) restraints derived from NMR experiments (Brtinger and Karplus, 1991; Nilges and Brtinger, 1991).

54

M.S.P. SANSOM and IAN D. KERR

It generates an ensemble of structures, all of which satisfy distance restraints chosen in order to embody biochemical data. Variation between members of an ensemble enables one to ascertain the extent to which a set of distance restraints specify a final structure. A major advantage of SA/MD is that it explores a wider range of conformational space than e.g., energy minimization, so the final structure is less biased towards the initial model. Kerr et al. (1994a,b) have used SA/MD to model parallel Leu20 helix dimers, chosen as simple models of hydrophobic TM helices. The initial model was a Cot atom template corresponding to an exactly parallel (f~ = 0 ~ pair of idealized t~-helices. All atoms other than the Coc atoms were added automatically during the SA/MD run. The resultant ensemble of structures (Figure 6A) had a mean crossing angle of f~ = +18 ~ corresponding to classical "ridges-in-grooves" packing of helices. Similar results were obtained for Ala20 helix dimers. Mean interaxial separations were R = 9.3 and 8.1 ~ for Leu20 and Ala20 helices, respectively. This study demonstrates that SA/MD will automatically pack hydrophobic helices into a favorable geometry with no explicit input other than an approximate Cct atom model and with only weak distance restraints in order to maintain the helices in an approximately parallel orientation. A similar SA/MD method was used by Treutlein et al. (1992) to model dimers formed by the TM helix of glycophorin. In vitro mutagenesis studies (Lemmon et al., 1992) revealed that dimerization of glycophorin TM helices in a detergent environment is sequence specific, and that subtle changes in sequence could alter helix-helix association. Both mutagenesis and SA/MD studies suggested that polar residues did not play an important role at the glycophorin helix-helix interface. Interestingly, the most stable packing of glycophorin TM helix dimers corresponded to f~ = -30 ~ This differs significantly from the classical "ridges-ingrooves" packing of Leu20 helix dimers (above), suggesting that specific side-chain-side-chain interactions must be taken into account when modeling

Figure 6. Models of bundles of TM helices generated by SA/MD. A shows 10 structures selected from an ensemble of 50 models for a parallel dimer of Leu20 helices. The oc-carbon backbones of the helices are shown superimposed. B illustrates a single structure from the ensemble in helical ribbon format. The helix axes, indicated by broken lines, define a helix crossing angle of D. = + 18 ~ C and D are ribbon diagrams of a parallel bundle of five Leu20 helices. In C the N-termini of the helices are uppermost and in D they are towards the viewer. The helix axes in this latter diagram indicate a positive crossing-angle between adjacent helices, resulting in a left-handed supercoil. E and F show a hexameric bundle of parallel helices used to model ion channels formed by Staphylococcus aureus G-toxin. In F, the side-chains of residues D4, D l l and D18 which line the central pore are shown in "ball-and-stick" format. (A to Dmi. D. Kerr et al., 1994a,b; E and Fml. D. Kerr, R. Sankararamkrishnan, and M. S. P. Sansom, in preparation).

:

i !

A

B

$ D C

E

F 55

56

M.S.P. SANSOM and IAN D. KERR

packing of TM helices. These studies provide further evidence that modeling via SMMD can provide new information about TM helix packing beyond that included in the initial model. Parallel Helix Bundles and ion Channels. A parallel bundle of TM helices is a structural motif thought to be present in several ion channel proteins (Oiki et al., 1990). This motif is of particular importance with respect to the nicotinic acetylcholine receptor (nAChR) superfamily. As discussed below, the central pore of the nAChR is formed by a bundle of five parallel amphipathic helices (Unwin, 1989, 1993; Stroud et al., 1990; Sansom, 1992a, 1993c). A similar motif may occur in other ion channel proteins. For example, the influenza M2 channel can be modeled as a bundle of four parallel t~-helices (Sansom and Kerr, 1993). Helix bundles are also found in channels formed by a number of channel-forming peptides (CFPs) e.g., alamethicin (Sansom, 1993a, b) and S. aureus &toxin (Mellor et al., 1988; Sansom et al., 1991). Furthermore, synthetic peptides designed to form amphipathic ct-helices generate cation-selective channels when incorporated into lipid bilayers (Lear et al., 1988). It is, therefore, of some interest to simulate packing within bundles of parallel TM helices. SA/MD has been used to model parallel bundles of Ala20 and Leu20 helices (Kerr et al., 1994a). Interhelix distance restraints were applied to maintain approximate symmetry of the bundles, and thus to permit formation of a central pore. An example of a pentameric bundle of Leu20 helices is shown in Figure 6C. This is from an ensemble of 20 such structures for which the mean crossing angle was f~ - + 13 ~ A positive crossing-angle generates a bundle corresponding to a left-handed supercoil, as commonly found in coiled-coil structures (Parry and Cohen, 1990). Corresponding left-handed supercoils were formed by Ala20 helices, and by tetrameric and hexameric bundles. This suggests a left-handed supercoil is a favorable structure for parallel, symmetric bundles of hydrophobic TM helices. However, packing changed slightly once polar residues were incorporated into the TM helices. Introduction of serine and threonine residues resulted in smaller crossing angles (f~ ca. +5~ apparently in order to accommodate interhelix H-bonds stabilizing the bundles. This parallels the observation on glycophorin dimers that TM helix packing is sequence dependent. The same SA/MD methodology has been used to model ion channels formed by parallel bundles of CFP helices. For example, hexameric bundles of S. aureus &toxin (see above) helices have been modeled by Kerr et al. (I.D. Kerr, R. Sankararamakrishnan, and M.S.P. Sansom, in preparation; see Figure 6E, F). The initial Co~ template was constructed such that the Cot of residue D 11 was directed towards the center of the bundle. In light of the results with Ala20 and Leu20 bundles, the 8-toxin helices were tilted so as to generate an initial crossing angle of f~ - +20 ~ The mean crossing-angle for the resultant ensemble of eight structures was f~ = 19 (+ 2) o. The side-chains of D4, D11 and D18 line the central pore, creating a hydrophilic lining to the central pore. This provides a pathway whereby ions may

Principles of Membrane Protein Structure

57

cross a planar lipid bilayer. Polar side-chains also contribute to stabilization of the helix bundle by formation of interhelix H-bonds and electrostatic interactions. Overall, these studies demonstrate the potential of SA/MD for modeling ion channels formed by bundles of parallel TM helices. It is likely that the method will be further refined and applied to a wider range of channels and transport proteins.

Bacteriorhodopsin. There have been several attempts to simulate packing of TM helices within bR. For example, Jfihnig and Edholm (1992) used molecular dynamics (MD) simulations to evaluate the extent to which tertiary structure-predictions were dependent on prior assumptions concerning the arrangement of the TM helices. These simulations were of particular interest in that a simple hydrophobic potential was used to mimic the presence of a lipid bilayer. This hydrophobic potential had been developed in earlier MD simulations of interactions between the glycophorin TM helix and a lipid bilayer (Edholm and J~nig, 1988). Starting with the helices of bR arranged on the circumference of a circle, it was found that although 25 ps MD simulations resulted in more compact helix bundles, the characteristic seven helix bundle ofbR was not generated. If the helices were placed at their approximate positions in the bR bundle, but with all helices exactly parallel to the bilayer normal, MD simulations then resulted in the TM helices adopting the same tilt angles relative to the bilayer as found in bR. Thus, the helix crossing-angles in the model must have been close to those in bR. Chou et al. (1992) adopted a different approach to modeling bR. The initial model was based on fitting seven TM helices to the experimental bR structure and then refining the resultant model using a combination of simulated annealing and energy minimization. The calculations were performed twicemonce with all-trans-retinal (as in the experimental bR structure) and once with 13-cis-retinal (as in the light-activated state). An energetic analysis of the resultant models was performed. It was concluded that while electrostatic interactions favored the antiparallel packing of adjacent helices, van der Waals interactions drove most adjacent helices close to the ideal packing angle of f~ = -160 for "ridges-in-grooves" packing of antiparallel helices. Indeed, examination of published helix-crossing angles (Table 6) reveals this is the case for these models of bR and for a model of a seven-helix G-protein-coupled receptor (see below). Together these and related studies encourage the use of simulations of TM helix packing to model all-(z IMPs. We will now briefly examine some modeling studies of seven-helix IMPs other than bR. G-Protein-Coupled Receptors. These proteins constitute a superfamily of IMPs whose amino acid sequences contain seven hydrophobic segments. They include rhodopsin, the recent 2D structure determination of which has provided experimental evidence for the presence of seven TM helices. On the basis of analysis of sequence variability within the superfamily, and of constraints imposed by connectivity between adjacent helices, Baldwin (1993) allocated the seven

58

M.S.P. SANSOM and IAN D. KERR

Table 6. Helix Crossing-Angles in Model 7-Helix Bundles fl(~ Helix Pair

bR E x ~

bR TRANs

bR cIs

~12AdR

A-B

-155

-150

-154

B-C

-167

-170

-169

--170 -163

C-D

+173

+168

-170

-170

D-E

-160

-151

-167

-174

E-F

-150

-169

-170

-178

F-G G-A

-150 +8

-165 +10

-162 +16

-171 +6

Notes: abREXrr--experimentally determined bR structure (Henderson et al., 1990); bRTRANS---modelof bR helix bundle containing all-trans retinal, and bR~ of bR helix bundle containing 13-cis retinal (Chou et al., 1992); l~2AdRumodel of 132-adrenergicreceptor (MaloneyHuss and Lybrand, 1992).

hydrophobic segments in the sequence to the seven peaks in the 2D projection map of rhodopsin. As in bR, the seven helices are packed in a left-handed fashion, with helices A and G parallel to one another and "closing" the seven-helix bundle. There have been several attempts to model G-protein-coupled receptors and their interactions with ligands. Carried out in the absence of a rhodopsin structure, most of these attempts have taken bR as their starting model. While such studies are of considerable interest as simulations of TM helix packing, results concerning receptor-ligand interactions should not be overinterpreted. Most of the model structures are quite close in their overall architecture to the initial bR structure. This is perhaps not surprising given that energy minimization was the principal methodology employed in these studies. For example, MahoneyHuss and Lybrand (1992) modeled a l~2-adrenergic receptor by a combination of interactive modeling and energy minimization. Helix packing within the resultant model is an approximate mirror image of that in bR. Homology modeling was adopted by Livingstone et al. (1992) to model dopamine receptors. Amore ambitious approach was adopted by Sylte et al. (1993), modeling a 5-HTla (serotonin) receptor. Sequence alignment and hydrophobicity profile analysis were used to define the seven TM helices. Seven isolated helices were generated and energy minimized so as to regularize side-chain conformations. These helices were interactively packed into a left-handed bundle with their most polar surfaces directed towards the center of the bundle. Interhelix loops were added, and the resultant model energy minimized, followed by a 25 ps MD simulation. Serotonin was added at the putative binding site and a further 100 ps MD simulation performed. The resultant structures clearly differed from that of bR. Unfortunately, from the published information it is not possible to assess to what extent it may match the projection structure of rhodopsin.

Proline in TM Helices. An important aspect of TM helix conformation which is particularly amenable to analysis via simulation is that of the structural role of

Principles of Membrane Protein Structure

59

intrahelical proline residues. Intrahelical prolines occur with a relatively high frequency in IMPs. For example, Brandl and Deber (1986) conducted a statistical survey of putative TM helices of IMPs and concluded that most transport proteins contain intramembranous prolines, whereas this residue is absent from TM helices of membrane proteins which lack a transport or channel function. They proposed two possible roles for intrahelical prolines: (a) facilitation of conformational changes via cis-trans isomerization of Xaa-Pro peptide bonds, and/or (b) provision of cation-liganding sites by the carbonyl oxygen of the Xaa-Pro peptide bond. In their review of ion channel structure, Eisenman and Dani (1987) highlighted a further possible role, stating that an intrahelical proline "frees a non-H-bonded backbone carbonyl oxygen for liganding to a cation". This was further explored in modeling studies of possible cation binding sites introduced into helix bundles by proline residues (Sansom, 1992b). The proline-kinked helices of bR and of PS/RC were analyzed by von Heijne (1991) who concluded that the convex face of a proline-kinked helix is often directed towards the center of a TM helix bundle. Woolfson et al. (1991) analyzed proline residues in amphipathic TM helices from a range of channel and transport proteins, revealing that proline is preferentially located on the hydrophilic face of such helices. On this basis they suggested that proline-kinked helices may form funnel-shaped pores, similar to those proposed as the basis of ion channel formation by alamethicin (Fox and Richards, 1982; Sansom, 1993a,b)and related CFPs (Sansom, 1991; Balaram et al., 1992). Although little is known of their structural role in IMPs other than bR and PS/RC, possible structures may be inferred by analysis of intrahelical prolines in globular proteins. B arlow and Thornton (1988) and Sankararamakrishnan and Vishveshwara (1992) have analyzed kink angles ofproline-containing helices in globular proteins. Both studies arrived at a mean kink angle of 25 ~ with kink angles varying from 9 ~ to 49 ~ Such crystallographic data suggest that proline-induced kinks may produce flexible "hinge" regions between two rigid segments within a TM helix. A number of theoretical studies have addressed this possibility. Polinsky et al. (1992) used a molecular mechanics approach to search for stable conformations of an A8LPFA8 helix. Three low energy trans-proline conformations were described, with kink angles of 29 ~ 78 ~ and 83 ~ It was suggested that interconversions between these might permit hinge-bending movements. A similar conclusion was drawn by Yun et al. (1992) using MD simulations to estimate the free energy change on reducing the kink of a Ac-AsPAs-NHMe helix from 40 ~ to 15 ~ A value of < + 0.5 kcal/mol was obtained, suggesting that hinge-bending is thermodynamically feasible. Hingebending of Ac-A7WA2YPA2WAs-NHMe (a simplified model of helix F of bR) was investigated in MD simulations by Sankararamakrishnan and Vishveshwara (1993). The kink angle varied between 0 ~ and 50 ~ on a 10 ps timescale. A subsequent comparative study of Ac-AI2TPA10-NHMe, Ac-A13PATA8-NHMe, and Ac-A13PA3TA6-NHMe, (Vishveshwara and Vishveshwara, 1993) showed that varying the position of threonine relative to proline modulated the extent and

60

M.S.P. SANSOM and IAN D. KERR

dynamics of variations in kink angle. Simulations of the dynamics of Dns(AUAUA)3PAUAUAW-OMe, an analogue of a synthetic channel-forming peptide, yielded variations in kink angle of between 5 ~ and 45 ~ (Vogel, 1992). Overall, it is highly likely that proline residues within TM helices provide molecular "hinges". Such flexible structural elements may play key roles in the function of transport and receptor IMPs. For example, a strictly conserved proline residue in the second TM helix of connexin seems to play a key role in voltage-gating of gap junction channels (Suchyna et al., 1993).

Overview of Simulation Studies. Molecular modeling studies of all-@ IMPs are at an early stage of development. Further work is required before one may anticipate accurate simulations of TM-helix packing. In particular, developments are needed in: (a) methods for incorporating "biochemical" information in MD simulations (e.g., via the use of distance restraints or additional potential energy terms), (b) use of multiple and/or long MD runs in order to minimize bias towards the starting model, and (c) application to a wider range of all-~ IMPs in order to evaluate the extent to which methods are limited to particular families of all-o~ IMPs. Assuming that these problems can be solved, it is likely that simulation studies will enable accurate prediction of helix packing within all-@ IMPs.

IV. ALL-I~ IMPs This section is concerned with those IMPs which exhibit an all-13 topology. Experimentally determined structures exhibiting this topology will be described, followed by a discussion of other IMPs believed to fall into this class, and by a brief examination of attempts at structure-prediction for all-~ IMPs.

A. Experimentally Determined Structures Por/ns The structures of four prokaryotic porins have been determined at high-resolution (Table 1; reviewed by: Schirmer and Rosenbusch, 1991; Cowan, 1993). All have the same basic fold, first observed in Rb. capsulatus porin, and so the latter structure will be described in detail. Porins are found in outer membrane of Gram-negative bacteria. They form water-filled transbilayer pores, of diameter ca. 10/~, which allow uptake and loss of small hydrophilic compounds through this membrane. As such they act as "molecular sieves" with an exclusion size of ca. 600 Da. The pores allow passive diffusion across the bilayer, are weakly ion selective, and exhibit some degree of voltage-gating. The structure of Rb. capsulatus porin (Weiss et al., 1991) is shown in Figure 7. The functional molecule is a trimer, each monomer of which forms a transbilayer

A

B Figure 7. Rb. capsulatus porin (coordinates from entry 2POR of the PDB). [3-Strands are represented by arrows. A is a porin monomer, viewed parallel to the plane of the membrane with the face of the barrel closest to the three-fold axis towards the viewer. B a porin trimer, viewed perpendicular to the membrane plane (i.e., down the three-fold axis). The three parallel pores are clearly visible. The three dark spheres represent interfacial Ca 2+ ions.

61

62

M.S.P. SANSOM and IAN D. KERR Table 7. Porins: TM 13-Strands and Loops

Poffn Rhodobacter capsulatus Rhodopseudomonas blastica E. coli (OmpF)

~-Strand Length ExternalLoop PeriplasmicLoop (Mean + SD) Length(Mean + SD) Length (Mean + SD)

10.7 + 2.9 10.8 + 2.5 12.4 + 2.6

13.8 + 12.6 12.6 + 12.9 15.4 + 7.5

2.3 + 0.5 2.3 + 1.7 2.4 + 2.2

pore. The monomer is a 16-strand antiparallel 13-barrel. Of the 301 residues, [~-strands account for 57% of the amino acid residues. The [3-strands are tilted relative to the trimer axis by between 30 and 60 ~ and the barrel has a right-handed twist. 13-strand lengths range from 6 to 17, with a mean length of 11 residues (Table 7). The height of the barrel wall correspondingly varies from 20 to 40 ]k, with the lowest region of the wall at the trimer interface. Interactions at the trimer interface are primarily polar, as are the residues which form the lining of the pore. The equatorial surface of the trimer is made up of hydrophobic side-chains. Thus the 13-strands in this region are amphipathic, with alternating hydrophobic (external) and hydrophilic (internal) side-chains. Interconnecting loops are short (two or three residues) at the periplasmic end of the barrel, but are more extensive at the external end. An extended external loop folds back into the barrel to form a so-called "eyelet" which is responsible for the size exclusion limit of the pore. The R. blastica (Kreusch et al., 1994) and E. coli (Cowan et al., 1992) porins exhibit a similar architecture to that just described (Table 7). I]-strand and loop lengths for the R. blastica and Rb. capsulatus porins are very similar. The E. coli porins both have 16-strand 13-barrels, but exhibit subtle differences from the other two porins. In both OmpF and PhoE the topology is pseudocyclic in that the 16th strand is made up of both the N- and C-terminal strands, with an intrastrand salt-bridge between the polypeptide chain termini. On average, the 13-strand length is somewhat higher for the E. coli porins, as is the external loop length (Table 7). An interesting feature of all of the porin structures is the distribution of aromatic residues on the outside of the [~-barrel. These form two rings around the trimet surface, one each side of the central hydrophobic band. This is reminiscent of the distribution of tryptophan residues in the PS/RC structure (see above and Schiffer et al., 1992). It has been suggested that in porins, such aromatic residues may play a role in "shielding" the protein molecule against adverse membrane fluctuations (Kreusch et al., 1994). An alternative explanation is that aromatic side-chains, particularly tyrosine and tryptophan, interact favorably with polar lipid headgroups in the bilayer/water interfacial region. Full evaluation of the quantitative importance of such interactions will require detailed structural information on thei~ molecular nature. The porin structures provide some clues as to the latter. Fo~ example, in the R. blastica crystal structure, electron density is present at the trime~ interface corresponding to the n-alkyl chains of bound detergent molecules.

Principles of Membrane Protein Structure

63

A second unusual feature of porins is the distribution of charged side-chains. There is a marked segregation of charge within the pore, with a cluster of positively charged side-chains close to the trimer axis and of cluster of negatively charged side-chains on the opposite face of the pore lining. This creates a strong transverse electrostatic field across the pore, which has been suggested to play a role in governing the selectivity of porin channels (Weiss et al., 1991a).

Aerolysin It is perhaps premature to classify aerolysin as an "experimentally determined structure". However, a wealth of experimental data points towards a predominantly I3-sheet structure for this protein. Aerolysin is a channel-forming toxin produced by Aeromonas hydrophila, a Gram-negative bacterium. It is secreted as proaerolysin, a 54-kDa soluble protein, which is converted to aerolysin by proteolytic removal of a C-terminal peptide of ca. 40 residues. Sequence analysis of proaerolysin fails to reveal any TM helices. It is, therefore, of interest to determine how a hydrophilic globular protein is converted to a channel-forming IME Proaerolysin has been shown spectroscopically to contain predominantly I3-sheet (Buckley, 1992). The crystal structure has been solved at 2.8/k resolution, and has been shown to consist of 40% I3-sheet (Tucker et al., 1990; Parker et al., 1994). The transbilayer pore formed by aerolysin has been imaged at 25/~ resolution by EM of 2D crystals (Wilmsen et al., 1992). The pore is formed by a heptamer of aerolysin molecules, packed in "barrel-stave" fashion about a central pore, diameter 17 A, the pore axis being coincident with the sevenfold symmetry axis. By combining the EM and X-ray derived structures, a model was generated in which the pore is lined by 21 I3-strands (Parker et al., 1994). A series of elegant experiments have provided valuable clues concerning the pathway from proaerolysin to aerolysin pores (van der Goot et al., 1993). Proaerolysin binds to cell membranes and undergoes proteolysis and subsequent oligomerization before inserting into the bilayer to form channels. Changes in binding of the hydrophobic dye, 8-anilino-l-naphthalene sulfonate (ANS), suggest that oligomerization exposes hydrophobic surface patches which drive membrane insertion. Therefore, there is a conformational transition from globular protein to IME However, in view of the convergence of the EM and X-ray results described above, it is clear that the latter retains a predominantly 13structure.

B. Related Structures

Porin-Like Proteins How widespread are all-l~ IMPs? It seems that some other proteins of bacterial outer membranes may possess porin-like folds. For example, high-affinity, ligandspecific transport proteins may be composed of a "gated-porin" along with a

64

M.S.P. SANSOM and IAN D. KERR

"gate-keeper" protein which facilitates entry of ligands into the porin-like channel (Rutz et al., 1993). Gated-porin proteins are predicted to have an antiparallel l-barrel structure, with the external loops responsible for binding of specific ligands. In eukaryotes, porins are present in outer mitochondrial membranes, where they are also referred to as VDACs (voltage-dependent anion channels; reviewed by Mannella, 1992). A low resolution projection-structure from EM studies of Neurospora crassa mitochondrial porin reveals cylindrical pores of ca. 30 ,~ diameter. These have been modeled as consisting ofbetween 12 and 19 amphipathic [3-strands in a barrel-like structure. It will be interesting to determine the extent to which the structures of mitochondrial porins resemble those of prokaryotic porins. Overall, there is little evidence suggesting that all-13 topologies are as widespread among IMPs as are all-~ topologies. However, it may be that further examples await discovery. For example, Fischbarg et al. (1993) have described a somewhat speculative porin-like model for mammalian facilitative glucose transporters.

TopologyPrediction There are two elements to structure-prediction as applied to putative porin-like topologies (Jahnig, 1989). First, it is supposed that the pore-lining ]3-strands will be amphipathic, with hydrophobic external side-chains and hydrophilic internal side-chains. Such a secondary structure is characterized by a very simple sequence patternmalternating hydrophobic and hydrophilic amino acids. Generally amphipathic 13-strands of 9 or 10 residues in length are searched for, in agreement with the length of such strands in experimentally determined porin structures. The second element is to predict the positions ofinterstrand l-turns. This may be carried out using a standard secondary structure-prediction method, such as that of Chou and Fasman (1978). An early attempt at topology prediction for porin, maltoporin, and OmpA protein from E. coli was made by Vogel and Jahnig (1986), in parallel with Raman spectroscopic measurements to determine the secondary structure composition of these proteins. The two porins were predicted to contain 18 amphipathic [3-strands per monomer, forming a 13-barrel around a central pore. The OmpA 1-177 fragment was predicted to form an eight-strand ]3-barrel. It has since been demonstrated that OmpA forms pores in lipid bilayers, with a low (relative to porins) pore diameter (Saint et al., 1993). A more recent predictive study on members of the porin superfamily was carried out by Jeanteur et al. (1991). Alignment of 14 porin sequences was combined with secondary structure-prediction of amphipathic 13-strands and [3-turns. The result was a consensus prediction of 16 13-strands, thus correctly predicting the number of TM ]3-strands in E. coli OmpF and PhoE porins, the structures of which were not published at the time of this work.

Principles of Membrane Protein Structure

65

Overview Topology predictions appear to work quite well for IMPs which exhibit a porin-like fold. It remains to be seen how widespread this class of fold is within other families of membrane proteins. It is possible that it is a rather specialized structure adopted when a large diameter, weakly selective transbilayer pore is required.

VI. ~/p IMPs In this section we discuss two families of ion channel proteins believed to have a "mixed" TM topology, containing both (z-helices and [3-strands. In neither case is a high resolution structure available, but in both cases evidence in favor of an (z/13 topology is reasonably convincing. In addition to reviewing these two families of IMPs, the possible general relevance of such topologies will be discussed.

A. Nicotinic Acetylcholine Receptor Structure The nicotinic acetylcholine receptor (nAChR) is the most intensively studied member of a superfamily of ligand-gated ion channels. The nAChR is a cation-selective channel found in post-synaptic membranes, which may be obtained in large amounts from Torpedo(an electric fish). It has been investigated using a wide range of biophysical techniques, from cryo-electron microscopy to patch-clamp recording, and also subjected to functional dissection by chemical labeling and sitedirected mutagenesis. The structure determination at 9 A resolution of the nAChR from Torpedo (Unwin, 1993) provides a three-dimensional model of the receptor-channel. Previous structural studies at 17 ,~ resolution (Toyoshima and Unwin, 1988) defined the overall shape of the protein, and revealed that the five homologous subunits ((zz]3yS) are packed pseudosymmetrically about a central axis which runs perpendicular to the plane of the bilayer. The ion channel appeared to lie along the fivefold axis. Hydrophobicity profile analysis of 50 or so nAChR sequences suggested that there were four TM helices (M1, M2, M3, and M4) per subunit. Circular dichroism experiments (Mielke and Wallace, 1988) introduced an element of doubt, as they demonstrated that the total (z-helical content of the nAChR was only 23%. If one assumed there to be four TM helices per subunit, this excluded the presence of (z-helices from elsewhere in the molecule, including the large extramembranous domain. However, the apparent conservation of the four TM helices in the sequences of other ligand-gated ion channels was persuasive evidence in favor of this model (Unwin, 1989; Betz, 1990). Mutagenesis and chemical

66

M.S.P. SANSOM and IAN D. KERR

labeling studies (Changeux et al., 1992) focused attention on the M2 helix, which was shown to interact with permeant ions and with channel-blocking drugs, and thus was proposed to line the ion channel. Determination of the structure of the nAChR used tubular crystals which presented the nAChR in a number of different projections, allowing direct reconstruction of the 3D structure of the protein from 2D images. At 9 A resolution, the most prominent elements of the secondary structure, the t~-helices, were evident. Overall, the nAChR is a 120 ,~ long, 80 ,~ diameter cylinder, with a pseudo-5x symmetry axis running down the center of the molecule perpendicular to the plane of the membrane. The ion channel lies on the fivefold axis. There is a 60/k long and 20 ,~ wide entrance to the channel on the synaptic face, followed by a narrow pore across the bilayer (about 30 ,~ long) which widens to form the cytoplasmic entrance, 20 A wide and 30 ,~, long. The 9 /k resolution map also revealed phospholipid headgroups and so allowed the protein to be positioned accurately relative to the bilayer. The synaptic (extramembranous) domain of the nAChR accounts for 55% of the mass of the protein. It includes, on the two t~-subunits, binding sites for the neurotransmitter acetylcholine. The synaptic domain of each subunit contains a left-handed coil of three t~-helices. In the t~-subunits, these three helices form a cavity near the center of the subunit, which is reached from the surface of the molecule by a deep cleft. This cavity is presumed to correspond to the acetylcholine binding site. The intramembranous domain of the nAChR exhibits a high degree of rotational symmetry. It contains only o n e TM helix per subunit. The central pore is lined by five such helices, but these are surrounded by a continuous rim of density which is presumed to correspond to a 13-barrel (see Figure 8A). The transmembrane helices visible in the nAChR are interpreted to be the M2 helices. The M2 helices are markedly kinked (Figure 8B), and are oriented such that at their closest approach they occlude the central pore, presenting a barrier to ion permeation. Thus, the structure is believed to correspond to the closed conformation of the channel. The kink angle, as measured from the data shown in Figure 8B, is ca. 44 ~ This is of comparable magnitude to the kinks in proline-containing helices, and yet proline is absent from the M2 sequence. The N-terminal segments of the helices form a left-handed supercoil (f~ ca. +17~ as observed in simple models of parallel TM helix bundles (see above). Tentative alignment of the M2 sequence to the helices suggests that the constriction in the channel, corresponding to the apex of the kink, is made up of a ring of conserved leucine residues. This has resulted in the proposal that these leucines form a "gate" which closes the channel to passage of ions. Revah et al. (1992) have shown that mutation of the corresponding Leu in a neuronal nAChR to a smaller residue results in changes in gating and conductance properties of the channel. On this basis, one may begin to speculate about the nature of channel opening. An attractive model of channel opening is one in the which helix orientation changes, and

oc-helix I~-sheet

A

B Figure 8. Schematic diagrams of the TM structure of the nicotinic acetylcholine receptor. A is a section through the intramembranous region of the molecule, viewed perpendicular to the bilayer plane. Five M2 helices, one donated by each subunit, line the central pore, and are surrounded by a band of (putative) [3-sheet. B illustrates the approximate structure of the pentameric bundle of M2 helices, viewed from the synaptic end (i.e., C-termini) of the helices. (Drawn from data in Unwin, 1993).

67

68

M.S.P. SANSOM and IAN D. KERR

thus opening the channel (Unwin, 1995). The surrounding 13-sheet may provide a scaffolding supporting such helix movements.

Verotoxin A possible model for the transbilayer domain of the nAChR is provided by the B subunits of two bacterial A-B toxins (heat-labile enterotoxin, Sixma et al., 1991; and verotoxin-1, Stein et al., 1992) whose crystal structures reveal them to be made up of a central bundle of five parallel {x-helices surrounded by a framework of [3-sheet (see Figure 9). Interactions between 13-strands from adjacent monomers hold the pentameric complex together, and there is a central pore passing through the molecule. Interestingly, it has been suggested that the B subunit of verotoxin-1 may undergo a conformational change and bring about translocation of the A subunit across the membrane of the target cell (Read and Stein, 1993). This further strengthens the proposal that such toxins may be used as models of nAChR-like od~ IMPs.

Figure 9. The structure of the B subunit pentamer of verotoxin (PDB entry 1 BOV), viewed down the approximate five-fold axis. The central pore is lined by five (z-helices, one from each subunit, surrounded by an outer rim of l-sheet.

Principles of Membrane Protein Structure

69

General Significance What is the overall importance of the nAChR structure? It clearly is of relevance to the other members of the ligand-gated channel superfamily, which are presumed to share a similar TM architecture. In the context of IMPs in general, the novel transbilayer architecture warns against uncritical use of hydrophobicity plots to generate models of transmembrane topology. Clearly one should pay close attention to spectroscopic data, particularly for those IMPs believed to contain extensive extramembranous domains. Although hydrophobic helices may be detected by sequence analysis, this does not demonstrate that they occupy TM locations. The need for experimental topology information (see above) is clearly emphasized.

B. Voltage-Gated Ion Channels This superfamily of IMPs includes voltage-gated K +, Na + and Ca 2+ channels (Hille, 1992). Of these, the molecular structure of voltage-gated K + channels is perhaps the best understood. However, in the absence of high resolution data, our understanding of K + channel structure is derived from sequence comparisons and analysis, and from site-directed mutagenesis studies. The latter have resulted in an interesting proposal concerning the topology of the transmembrane domain, in particular the pore region, and it is this which we will describe in more detail. Voltage-gated K + channels, perhaps the best characterized of which is the Shaker channel of Drosophila, are thought to form tetramers in which the four subunits are packed around a central pore (MacKinnon, 1991). Low resolution EM images of purified Shaker channel protein (Li et al., 1994), reveal a square-shaped complex, of dimensions ca. 80 x 80/k, with a large central hole. Sequence analysis of voltage-gated K § channels reveals six putative TM helices (Figure 10A). This pattern is conserved in the corresponding membrane domains of voltage-gated Na + and Ca 2+channels. Na + and Ca 2+channels contain four repeats of a domain corresponding to a single K + channel subunit within one polypeptide chain. Initial proposals for voltage-gated channel structure had a central pore lined by a bundle of amphipathic TM helices. However, recent mutagenesis data are incompatible with such a model and suggest a more complex topology (reviewed by Miller, 1991; Pongs, 1993). Briefly, mutagenesis results indicate that the loop (H5) between TM helices $5 and $6 controls the ion selectivity of K + channels, and also contains those residues responsible for binding toxins (e.g., charybdotoxin, CTX) and simple organic molecules (e.g., tetraethylammonium, TEA) which have been demonstrated to block the channel when it is open. Thus, H5 is concluded to form at least part of the lining of the channel. Furthermore, mutations at either end of the H5 sequence alter interactions with externally applied CTX or TEA, whereas mutations in the center of H5 alter interactions with internally applied TEA. The simplest interpretation of these results is that H5 forms a ]I-loop which traverses the bilayer and forms the lining of the pore of Shaker. In conjunction with evidence

70

M. S. P. SANSOM and IAN D. KERR Shaker $1 $ 2 $ 3 $ 4 $ 5 H5

IRK1 $ 5 H5 $ 6

$6

pore

A

B

,.S5

12 Figure 10. Model structures for K+ channel proteins. A and B show proposed topologies of single subunits of Shakerand IRK1 channels. The pore lining region (I-t5) is modeled as a 13-hairpin flanked by s-helices. Both channel proteins are tetramers, with the central pore formed by an eight stranded antiparalle113-barrel, as indicated in 12, a schematic view of such a channel perpendicular to the plane of the bilayer.

on the tetrameric structure of Shaker this has resulted in the proposal illustrated in Figure 10C, in which the central pore is formed by an eight-stranded antiparallel ~-barrel. Two strands are donated by each subunit, and the 13-barrel is in turn surrounded by a 6 • 4 = 24 TM helix bundle. There have been several attempts to model Shaker channels. Durrell and Guy (1992) used an interactive approach, combined with energy minimization, to model the entire Shakerprotein. Their model was used as the basis of informed speculation concerning possible mechanisms of voltage-dependent activation and inactivation of Shaker. Bogusz et al. (1992) focused on the central 13-barrel, which was modeled using energy minimization on the basis of a theoretical model of an optimal eight stranded 13-barrel structure. Their model was used for calculations of ion permeation energies. Such models demonstrate the feasibility of a "mixed" ~13 structure for K § channels. However, it should be remembered that there is no structural evidence for a [~-barrel. Indeed, synthetic H5 peptide in isolation forms tx-helices in a

Principles of Membrane Protein Structure

71

membrane environment (Haris et al., 1994), although this does not preclude adoption of a I]-strand conformation in the intact protein. What is the general relevance of the proposed structure for voltage-gated channels? A related fold has been suggested for a second family of K § channels, exemplified by an inward rectifier K + channel (IRK1; Kubo et al., 1993). Sequence analysis of IRK1 suggests a smaller TM domain, containing only two putative TM helices per subunit. The loop between the two helices shows sequence homologies with H5 from Shaker. The TM topology proposed for IRK1 is shown in Figure 10B, with two TM helices flanking a central H5 [3-loop. The general significance of Shaker-like topologies, outside ion channel proteins, is as yet unknown. However, it seems possible that further mutagenesis and structural studies may reveal 13-loops within transport proteins previously suggested to be all-t~ IMPs.

VI.

CONCLUSIONS

The past decade has witnessed considerable progress in unravelling the principles of membrane protein structure. The all-o~ family of IMPs is represented by several experimental structures, and more structures will be solved at high resolution in the near future. Furthermore, there is now a good understanding of how TM helices assemble within lipid bilayers. This, in turn, is enabling development of computational procedures to simulate such self-assembly, thus allowing prediction of all-or IMP structures. Rather less is known concerning other classes of IMP. The porins remain the major representative of all-[~ IMPs. It remains to be seen how generally applicable structural principles governing the porins are to other membrane proteins. As far as more complex transbilayer topologies are concerned, the nicotinic receptor provides a tantalizing glimpse of possible ot/13 folds within bilayers. Again, the general relevance of such structures remains uncertain. What might the future hold? Further IMP structures will be determined at high resolution, both by EM and by X-ray diffraction. Both methods are now mainly limited by problems of how to express sufficient material in order to enable growth of suitably diffracting 2D or 3D crystals (Schertler, 1992). In addition to such direct methods, a "hybrid" approach to membrane protein structure may be possible. Such an approach might combine: (a) definition of TM topology, (b) determination of key residues by site-directed mutagenesis, (c) use of solid-state NMR techniques to determine selected inter-residue distances (Smith and Peersen, 1992; Smith, 1993), and (d) computer simulation to integrate all of the experimental information into possible molecular models. Of course, the success of such an approach will depend on increased understanding of structural principles as derived from further experimental structures. An increase in the number of experimental structures will result in a refinement of ideas concerning general principles of membrane protein

72

M.S.P. SANSOM and IAN D. KERR

structure. This, in turn, will e n a b l e g r e a t e r u s e to b e m a d e o f the e x p a n d i n g m e m b r a n e p r o t e i n s e q u e n c e database.

ACKNOWLEDGMENTS This work was supported by the Wellcome Trust. Our thanks to our colleague Dr. R. Sankararamakrishnan for help with calculations on TM helix geometry.

REFERENCES Allen, J. P., Feher, G. Yeates, T. O., Komiay, H., & Rees, D. C. (1987). Structure of the reaction center from Rhodobacter sphaeroides R-26: The protein subunits. Proc. Natl. Acad. Sci. USA 84, 6162-6166. Balaram, E, Sukumar, M., Krishna, K., Mellor, I. R., & Sansom, M. S. P. (1992). The properties of ion channels formed by zervamicins. Eur. Biophys. J. 21, 117-128. Baldwin, J. M. (1993). The probable arrangement of the helices in G protein-coupled receptors. EMBO J. 12, 1693-1703. Barlow, D. J., & Thornton, J. M. (1988). Helix geometry in proteins. J. Mol. Biol. 201,601--619. Barsukov, I. L, Nolde, D. E., Lomize, A. L., & Arseniev, A. S. (1992). Three-dimensional structure of proteolytic fragment 163-231 of bacterioopsin determined by nuclear magnetic resonance data in solution. Eur. J. Biochem. 206, 665-672. Bernstein, E, Koetzle, T. Williams, G., Meyer, E., Brice, M., Rodgers, J., Kennard, O., Shimanouchi, T., & Tasumi, M. (1977). The Protein Data Bank: A computer-based archival file for macromolecular structures. J. Mol. Biol. ll2, 535-542. Betz, H. (1990). Homology and analogy in transmembrane channel design: lessons from synaptic membrane proteins. Biochem. 29, 3591-3599. Bogusz, S., Boxer, A., & Busath, D. D. (1992). An SS 1-SS2 I~-barrel structure for the voltage-activated potassium channel. Prot. Engng. 5, 275-293. Bormann, B. J., & Engelmann, D. M. (1992). Intramembrane helix-helix association in oligomerization and transmembrane signaling. Annu. Rev. Biophys. Biophys. Struct. 21,223-242. Bormann, B. J., Knowles, W. J., & Marchesi, V. T. (1989). Synthetic peptides mimic the assembly of transmembrane glycoproteins. J. Biol. Chem. 264, 4033-4037. Braiman, M. S., & Rothschild, K. J. (1988). Fourier-transform infrared techniques for probing membrane-protein structure. Annu. Rev. Biophys. Biophys. Chem. 17, 541-570. Branden, C., & Tooze, J. (1991). Introduction to Protein Structure. Garland, New York. Brandl, C. J., & Deber, C. M. (1986). Hypothesis about the function of membrane-buried proline residues in transport proteins. Proc. Natl. Acad. Sci. USA 83, 917-921. Briinger, A. T., & Karplus, M. (1991). Molecular dynamics simulations with experimental restraints. Acc. Chem. Res. 24, 54-61. Buckley, J. T. (1992). Crossing three membranes: channel formation by aerolysin. FEBS Lett. 307, 30-33. Capaldi, R. A. (1982). Structure of intrinsic membrane proteins. Trends Biochem. Sci. 7, 292-295. Changeux, J. P., Galzi, J. I., Devillers-Thi6ry, A., & Bertrand, D. (1992). The functional architecture of the acetylcholine nicotinic receptor explored by affinity labelling and site-directed mutagenesis. Quart. Rev. Biophys. 25, 395-432. Chothia, C. (1984). Principles that determine the structure of proteins. Annu. Rev. Biochem. 53, 537-572.

Principles of Membrane Protein Structure

73

Chothia, C., Levitt, M., & Richardson, D. (1981). Helix to helix packing in proteins. J. Mol. Biol. 145, 215-250. Chou, K. C., Carlacci, L., Maggiora, G. M., Parodi, L. A., & Schulz, M. W. (1992). An energy-based approach to packing the 7-helix bundle of bacteriorhodopsin. Prot. Sci. 1, 810--827. Chou, P. Y., & Fasman, G. D. (1978). Empirical predictions of protein conformation. Annu. Rev. Biochem. 47, 251-276. Cornette, J. L., Cease, K. B., Margalit, H, Spouge, J. L., Berzovsky, J. A., & DeLisi, C. (1987). Hydrophobicity scales and computational techniques for detecting amphipathic structures in proteins. J. Mol. Biol. 195, 659-685. Cowan, S. W. (1993). Bacterial porins--lessons from 3 high-resolution structures. Curr. Opin. Struct. Biol. 3, 501-507. Cowan, S. W., Schirmer, T., Rummel, G., Steiert, M., Ghosh, R., Apuptit, R. A., Jansonius, J. N., & Rosenbusch, J. P. (1992). Crystal structures explain functional properties of two E. coli porins. Nature 358, 727-733. Creighton, T. E. (1993). Proteins: Structures and Molecular Properties (2nd ed.), pp. 66--67, Freeman, New York. Cronet, P., Sander, C., & Vriend, G. (1993). Modeling of transmembrane seven helix bundles. Prot. Engng. 6, 59-64. Deisenhofer, J., & Michel, H. (1989). The photosynthetic reaction center from the purple bacterium Rhodopseudomonas viridis. Science 245, 1463-1473. Deisenhofer, J., Epp, O., Miki, K., Huber, R., & Michel, H. (1985). Structure of the protein subunits in the photosynthetic reaction centre of Rhodopseudomonas viridis at 3 A resolution. Nature 318, 618-624. Donnelly, D., Overington, J. P., Ruffle, S. V., Nugent, J. H. A., & Blundell, T. L. (1993). Modeling o~-helical transmembrane domains: the calculation and use of substitution tables for lipid-facing residues. Prot. Sci. 2, 55-70. Durrell, S. R., & Guy, H. R. (1992). Atomic scale structure and functional models of voltage-gated potassium channels. Biophys. J. 62, 238-250. Edholm, O., & Jiihnig, E (1988). The structure of a membrane-spanning polypeptide studied by molecular dynamics. Biophys. Chem. 30, 279-292. Eisenberg, D. (1984). Three-dimensional structure of membrane and surface proteins. Annu. Rev. Biochem. 53, 595-623. Eisenberg, D., Weiss, R. M., & Terwilliger, T. C. (1982). The helical hydrophobic moment: a measure of the amphiphilicity of a helix. Nature 299, 371-374. Eisenberg, D., Schwarz, E., Komaromy, M., & Wall, R. (1984). Analysis of membrane and surface protein sequences with the hydrophobic moment plot. J. Mol. Biol. 179, 125-142. Eisenman, G., & Dani, J. A. (1987). An introduction to molecular architecture and permeability of ion channels. Annu. Rev. Biophys. Biophys. Chem. 16, 205-226. Engelman, D. M., Steitz, T. A., & Goldman, A. (1986). Identifying nonpolar transbilayer helices in amino acid sequences of membrane proteins. Annu. Rev. Biophys. Biophys. Chem. 15, 321-353. Finer-Moore, J., & Stroud, R. M. (1984). Amphipathic analysis and possible formation of the ion channel in an acetylcholine receptor. Proc. Natl. Acad. Sci. USA 81,155-159. Fischbarg, J., Cheung, M., Czegledy, E, Li, J., Iserovich, P., Kuang, K., Hubbard, J., Garner, M., Rosen, O. M., Golde, D. W., & Vera, J. C. (1993). Evidence that facilitative glucose transporters may fold as ]3-barrels. Proc. Natl. Acad. Sci. USA 90, 11658-11662. Fox, R. O., & Richards, E M. (1982). A voltage-gated ion channel model inferred from the crystal structure of alamethicin at 1.5/~, resolution. Nature 300, 325-330. Frey, S., & Tamm, L. K. (1991). Orientation of melittin in phospholipid bilayers: a polarized attenuated total reflection infrared study. Biophys. J. 60, 922-930.

74

M.S.P. SANSOM and IAN D. KERR

Furois-Corbin, S., & Pullman, A. (1986a). Theoretical study of the packing of tx-helices by energy minimization: effect of the length of the helices on the packing energy and on the optimal configuration of a pair. Chem. Phys. Lett. 123, 305-310. Furois-Corbin, S., & Pullman, A. (1986b). Theoretical study of the packing of or-helices of poly(L-alanine) into transmembrane bundles. Possible significance for ion transfer. Biochim. Biophys. Acta 860, 165-177. Furois-Corbin, S., & Pullman, A. (1987a). Theoretical study of potential ion channels formed by a bundle of o~-helices: effect of the presence of polar residues along the inner channel wall. J. Biomol. Str. Dyn. 4, 589-597. Furois-Corbin, S., & Pullman, A. (1987b). Theoretical study of the packing of t~-helices into possible transmembrane bundles: sequences including alanines, leucines and serines. Biochim. Biophys. Acta 902, 31--45. Gennis, R. B. (1989). Biomembranes: Molecular Structure and Function. Springer-Verlag, New York. Haris, P. I., & Chapman, D. (1988). Fourier transform infrared spectra of the polypeptide alamethicin and a possible structural similarity with bacteriorhodopsin. Biochim. Biophys. Acta 943, 375380. Haris, P. I., & Chapman, D. (1992). Does Fourier-transform infrared spectroscopy provide useful information on protein structures? Trends Biochem. Sci. 17, 328-333. Haris, P. I., Ramesh, B., Sansom, M. S. P., Kerr, I. D., Srai, K. S., & Chapman, D. (1994). Studies of the pore-forming domain of a voltage-gated potassium channel. Prot. Engng. 7, 255-262. Havelka, W., Henderson, R., Heymann, J. A. W., & Oesterhelt, D. (1993). Projection structure of halorhodopsin from Halobacterium halobium at 6/~ resolution obtained by electron cryo-microscopy. J. Mol. Biol. 234, 837-846. Henderson, R., & Unwin, P. N. T. (1975). Three-dimensional model of a purple membrane obtained by electron microscopy. Nature 257, 28-32. Henderson, R., Baldwin, J. M., Ceska, T. A., Zemlin, E, 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. Hennessey, E. S., & Broome-Smith, J. K. (1993). Gene-fusion techniques for determining membrane protein topology. Curr. Opin. Struct. Biol. 3, 524-531. Hille, B. (1992). Ionic Channels of Excitable Membranes (2nd ed.), Sinauer, Sunderland MA. Huang, H. W., & Wu, Y. (1991). Lipid-alamethicin interactions influence alamethicin orientation. Biophys. J. 60, 1079-1087. J~nig, F. (1989). Structure prediction for membrane proteins. In: Prediction of Protein Structure and the Principles of Protein Conformation (Fasman, G. D., ed.), pp. 707-717, Plenum, New York. J~nig, E, & Edholm, O. (1992). Modeling of the structure of bacteriorhodopsinma molecular dynamics study. J. Mol. Biol. 226, 837-850. Jeanteur, D., Lakey, J. H., & Pattus, E (1991). The bacterial porin superfamily: sequence alignment and structural prediction. Mol. Microbiol. 5, 2153-2164. Johnson, W. C. (1988). Secondary structure of proteins through circular-dichroism spectroscopy. Annu. Rev. Biophys. Biophys. Chem. 17, 145-166. Kaback, H. R. (1992). The lactose permease of Escherichia coli: a paradigm for membrane transport proteins. Biochim. Biophys. Acta 1101,210-213. Kahn, T. W., & Engelman, D. M. (1992). Bacteriorhodopsin can be refolded from two independently stable transmembrane helices and the complementary five-helix fragment. Biochem. 31, 61446151. Kerr, I. D. (1994). The Biophysics of Peptide Models of Ion Channels. D. Phil. thesis, University of Oxford. Kerr, I. D., & Sansom, M. S. P. (1993a). Hydrophilic surface maps of channel-forming peptides: analysis of amphipathic helices. Eur. Biophys. J. 22, 269-277.

Principles of Membrane Protein Structure

75

Kerr, I. D., & Sansom, M. S. P. (1993b). Hydrophilic and hydrophobic surface map analysis of bacteriorhodopsin. Biochem. Soc. Trans. 2 l, 78S. Kerr, I. D., Sankararamakrishnan, R., Smart, O. S., & Sansom, M. S. E (1994a). Simulated annealing and molecular dynamics modelling of parallel helix bundles. Biophys. J. 67, 1501-1515. Kerr, I. D., Sankararamakrishnan, R., & Sansom, M. S. P. (1994b). Simplified models of the pore domain of the nicotinic acetylcholine rtceptor. Biochem. Soc. Trans. 22, 158S. Komiya, H., Yeates, T. O., Rees, D. C., Allen, J. P., & Feher, G. (1987). Structure of the reaction center from Rhodobacter sphaeroides R-26 and 2.4.1: symmetry relations and sequence comparisons. Proc. Nail. Acad. Sci. USA 85, 9012-9016. Krauss, N., Hinrichs, W., Witt, I., Fromme, E, Pritzkow, W., Dauter, Z., Betzel, C., Wilson, K. S., Witt, H. T., & Saenger, W. (1993). Three-dimensional structure of system I of photosynthesis at 6 A, resolution. Nature 36 l, 326-330. Kreusch, A., Neubtiser, A., Schiltz, W, Wekesser, J., & Schulz, G. E. (1994). The structure of the membrane channel porin from Rhodopseudomonas blastica at 2.0 ~, resolution. Prot. Sci. 3, 58-63. Kubo, Y., Baldwin, J. J., Jan, Y. N., & Jan, L. Y. (1993). Primary structure and functional expression of a mouse inward rectifier potassium channel. Nature 362, 127-133. Kiihlbrandt, W. (1992). Two-dimensional crystallization of membrane proteins. Quart. Rev. Biophys. 25, 1-49. Kiihlbrandt, W., & Wang, D. N. (1991). Three-dimensional structure of plant light-harvesting complex determined by electron crystallography. Nature 350, 130-134. Kiihlbrandt, W., Wang, D. N., & Fujiyoshi, Y. (1994). Atomic model of plant light-harvesting complex by electron crystallography. Nature 367, 614-621. Kyte, J., & Doolittle, R. E (1982). A simple method for displaying the hydrophobic character of a protein. J. Mol. Biol. 157, 105-132. Lear, J. D., Wasserman, Z. R., & DeGrado, W. E (1988). Synthetic amphiphilic peptide models for protein ion channels. Science 240, 1177-1181. Lemmon, M. A., Flanagan, J. M., Treutlein, H. R., Zhang, J., & Engelmann, D. M. (1992). Sequence specificity in the dimerization of transmembrane t~-helices. Biochem. 31, 12719-12725. Li, J. (1992). Bacterial toxins. Curr. Opin. Struct. Biol. 2, 545-556. Li, M., Unwin, N., Stauffer, K. A., Jan, Y. N., & Jan, L. Y. (1994). Images of purified Shaker potassium channels. Curr. Biol. 4, 110-115. Livingstone, C. D., Strange, P. G., & Naylor, L. H. (1992). Molecular modeling of D2-1ike dopamine receptors. Biochem J. 287, 277-282. Lomize, A. L., Pervushin, K. V., & Arseniev, A. S. (1992). Spatial structure of (34-65) bacterioopsin polypeptide in SDS micelles determined from nuclear magnetic resonance. J. Biomol. NMR 2, 361-372. MacKinnon, R. (1991). Determination of the subunit stoichiometry of a voltage-activated potassium channel. Nature 350, 232-235. MaloneyHuss, K., & Lybrand, T. P. (1992). Three-dimensional structure for the 132adrenergic receptor protein based on computer modeling studies. J. Mol. Biol. 225, 859-871. Mannella, C. A. (1992). The 'ins' and 'outs' of mitochondrial membrane channels. Trends Biochem. Sci. 17, 315-320. Mellor, I. R., Thomas, D. H., & Sansom, M. S. P. (1988). Properties of ion channels formed by Staphylococcus aureus ~i-toxin. Biochim. Biophys. Acta 942, 280-294. Michel, H. (1991). Crystallization of Membrane Proteins. CRC Press, Florida. Mielke, D. L., & Wallace, B. A. (1988). Secondary structural analyses of the nicotinic acetylcholine receptor as a test of molecular models. J. Biol. Chem. 263, 3177-3182. Miller, C. (1991). 1990: Annus mirabilis of potassium channels. Science 252, 1092-1096. Mouritsen, O. G., & Bloom, M. (1984). Mattress model of lipid-protein interactions in membranes. Biophys. J. 46, 141-156.

76

M.S.P. SANSOM and IAN D. KERR

Mouritsen, O. G., & Bloom, M. (1993). Models of lipid-protein interactions in membranes. Annu. Rev. Biophys. Biomol. Struct. 22, 145-171. Nilges, M., & Brtinger, A. T. (1991). Automatic modeling of coiled coils: application to the GCN4 dimerization region. Prot. Engng. 4, 649-659. Oesterhelt, D., & Tittor, J. (1989). Two Pumps, one principle--light-driven ion-transport in Halobacteria. Trends Biochem. Sci. 14, 57-61. Oiki, S., Madison, V., & Montal, M. (1990). Bundles of amphipathic transmembrane oc-helices as a structural motif for ion-conducting channel proteins: studies on sodium channels and acetylcholine receptors. Proteins: Struct. Funct. Genetics 8, 226-236. Park, K., Perczel, A., & Fasman, G. D. (1992). Differentiation between transmembrane helices and peripheral helices by the deconvolution of circular dichorism spectra of membrane proteins. Prot. Sci. 1, 1032-1047. Parker, M. W., & Pattus, E (1993). Rendering a membrane protein soluble in water: a common packing motif in bacterial protein toxins. Trends in Biochem. Sci. 18, 391-395. Parker, M. W., Buckley, J. T., Postma, J. P. M., Tucker, A. D., Leonard, K., Pattus, F., & Tsernoglou, D. (1994). Structure of the Aeromonas toxin proaerolysin in its water-soluble and membrane-channel states. Nature 367, 292-295. Parry, C., & Cohen, D. A. D. (1990). a-Helical coiled coils and bundles: how to design an or-helical protein. Proteins: Struct. Funct. Genetics 7, 1-15. Pervushin, K. V., & Arseniev, A. S. (1992). Three-dimensional structure of (1-36) bacterioopsin in methanol-chloroform mixture and SDS micelles determined by 2D 1H-NMR spectroscopy. FEBS Lett. 308, 190-196. Polinsky, A., Goodman, M., Williams, K. A., & Deber, C. M. (1992). Minimum energy conformations of proline-containing helices. Biopolymers 32, 399-406. Pongs, O. (1993). Structure-function studies on the pore of potassium channels. J. Membrane Biol. 136, 1-8. Popot, J. L. (1993). Integral membrane-protein structure: transmembrane o~-helices as autonomous folding domains. Curr. Opin. Struct. Biol. 3, 532-540. Popot, J. L., & de Vitry, C. (1990). On the microassembly of integral membrane proteins. Annu. Rev. Biophys. Biophys. Chem. 19, 369-403. Popot, J. L., & Engelman, D. M. (1990). Membrane protein folding and oligomerization: the two-stage model. Biochem. 29, 4031-4037. Read, R. J., & Stein, P. E. (1993). Toxins. Curr. Opin. Struct. Biol. 3, 853-860. Reddy, B. V. B., & Blundell, T. L. (1993). Packing of secondary structural elements in proteins: analysis and prediction of inter-helix distances. J. Mol. Biol. 233, 464--479. Rees, D. C., DeAntonio, L., & Eisenberg, D. (1989). Hydrophobic organization of membrane proteins. Science 245, 510-513. Revah, F., Bertrand, D., Galzi, J. L., Devillers-Thiery, A., Mulle, C., Hussy, N., Bertrand, S., Ballivet, M.,& Changeux, J. P. (1992). Mutations in the channel domain alter desensitization of a neuronal nicotinic receptor. Nature 353, 846-849. Rutz, J. M., Liu, J., Lyons, J. A., Goranson, J., Armstrong, S. K., Mclntosh, M. A., Feix, J. B., & Klebba, P. E. (1993). Formation of a gated channel by a ligand-specific transport protein in the bacterial outer membrane. Science 258, 471-475. Saint, N., De, E., Julien, S., Orange, N., & Molle, G. (1993). Ionophore properties of OmpA of Escherichia coli. Biochim. Biophys. Acta 1145, 119-123. Sankararamakrishnan, R., & Vishveshwara, S. (1992). Geometry of proline-containing alpha-helices in proteins. Int. J. Peptide Protein Res. 39, 356-363. Sankararamakrishnan, R., & Vishveshwara, S. (1993). Characterization of proline-containing or-helix (helix F model of bacteriorhodopsin) by molecular dynamics studies. Proteins: Struct. Funct. Genetics 15, 26-41.

Principles of Membrane Protein Structure

77

Sansom, M. S. P. (1991). The biophysics of peptide models of ion channels. Prog. Biophys. Mol. Biol. 55, 139-236. Sansom, M. S. P. (1992a). The roles of serine and threonine sidechains in ion channels: a modelling study. Eur. Biophys. J. 21,281-298. Sansom, M. S. P. (1992b). Proline residues in transmembrane helices of channel and transport protines: a molecular modeling study. Prot. Engng. 5, 53-60. Sansom, M. S. P. (1993a). Alamethicin and related peptaibols--model ion channels. Eur. Biophys. J. 22, 105-124. Sansom, M. S. P. (1993b). Structure and function of channel-forming peptaibols. Quart. Rev. Biophys. 26, 365-421. Sansom, M. S. P. (1993c). Acetylcholine receptor: peering down a pore. Curr. Biol. 3, 239-241. Sansom, M. S. P., & Kerr, I. D. (1993). Influenza virus M2 protein: a molecular modelling study of the ion channel. Prot. Engng. 6, 65-74. Schertler, G. E X. (1992). Overproduction of membrane proteins. Curr. Opin. Struct. Biol. 2, 534-544. Schertler, G. E X., Bartunik, H. D., Michel, H., & Oesterhelt, D. (1993a). Orthorhombic crystal form of bacteriorhodopsin nucleated on benzamidine diffracting to 3.6 ,~ resolution. J. Mol. Biol. 234, 156-164.

Schertler, G. E X., Villa, C., & Henderson, R. (1993b). Projection structure of rhodopsin. Nature 362, 771-772. Schiffer, M., & Edmundson, A. B. (1967). Use of helical wheels to represent the structures of protein and to identify segments with helical potential. Biophys. J. 7, 121-135. Schiffer, M., Chang, C. H., & Stevens, F. J. (1992). The functions of tryptophan residues in membrane proteins. Prot. Engng. 5, 213-214. Schirmer, T., & Rosenbusch, J. P. (1991). Prokaryotic and eukaryotic porins. Curr. Opin. Struct. Biol. 1,539-545. Schulz, G. E., & Schirmer, R. H. (1979). Principles of Protein Structure. Springer-Verlag, New York. Sipos, L., &von Heijne, G. (1993). Predicting the topology of eukaryotic membrane proteins. Eur. J. Biochem. 213, 1333-1340. Sixma, T. K., Pronk, S. E., Kalk, K. H., Warton, E. S., van Zanten, B. A. M., Witholt, B., & Hol, W. G. J. (1991). Crystal structure of a cholera toxin related heat-labile enterotoxin from E. coli. Nature 351,371-377. Smith, S. O. (1993). Magic angle spinning NMR methods for internuclear distance measurements. Curr. Opin. Struct. Biol. 3, 755-759. Smith, S. O., & Peersen, O. B. (1992). Solid-state NMR approaches for studying membrane protein structure. Annu. Rev. Biophys. Biomol. Struct. 21, 25-47. Stein, P. E., Boodhoo, A., Tyrrell, G. J., Brunton, J. L., & Read, R. J. (1992). Crystal structure of the cell-binding B oligomer of verotoxin-1 from E. coli. Nature 355, 748-750. Stroud, R. M., McCarthy, M. P., & Shuster, M. (1990). Nicotinic acetylcholine receptor superfamily of ligand gated ion channels. Biochem. 30, 11007-11023. Suchyna, T. M., Xu, L. X., Gao, E, Fourtner, C. R., & Nicholson, B. J. (1993). Identification of a proline residue as a transduction element in voltage-gating of gap junctions. Nature 365, 847-849. Surewicz, W. K., Mantsch, H. H., & Chapman, D. (1993). Determination of protein secondary structure by Fourier transform infrared spectroscopy: a critical appraisal. Biochem. 32, 389-394. Swords, N. A., & Wallace, B. A. (1993). Circular-dichroism analyses of membrane proteins: examination of environmental effects on bacteriorhodopsin spectra. Biochem. J. 289, 215-219. Sylte, I., Edvarsen, O., & Dahl, S. G. (1993). Molecular dynamics of the 5-HTla receptor and ligands. Prot. Engng. 6, 691-700. Tappin, M. J., Pastore, A., Norton, R. S., Freer, J. H., & Campbell, I. D. (1988). High resolution 1H NMR study of the solution structure of 8-hemolysin. Biochem. 27, 1643-1647. Terwilliger, T. C., & Eisenberg, D. (1982). The structure of melittin: II~interpretation of the structure. J. Biol. Chem. 257, 6016-1022.

78

M.S.P. SANSOM and IAN D. KERR

Toyoshima, C., & Unwin, N. (1988). Ion channel of acetylcholine receptor reconstructed from images of postsynaptic membranes. Nature 336, 247-250. Treutlein, H. R., Lemmon, M. A., Engelman, D. M., & Briinger, A.T. (1992). The glycophorin A transmembrane domain dimer: sequence-specific propensity for a fight-handed supercoil of helices. Biochem. 31, 12726-12733. Tucker, A. D., Parker, M. W., Tsernoglou, D., & Buckley, J. T. (1990). Crystallization of a proform of aerolysin, a hole-forming toxin from Aeromonas hydrophila. J. Mol. Biol. 212, 561-562. Unwin, N. (1989). The structure of ion channels in membranes of excitable cells. Neuron 3, 665-676. Unwin, N. (1993). Nicotinic acetylcholine receptor at 9 ,~ 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., Pattus, F., Wong, K. R., & Buckley, J. T. (1993). Oligomerization of the channel-forming toxin aerolysin precedes insertion into lipid bilayers. Biochem. 32, 2636-2642. Vishveshwara, S. S., & Vishveshwara, S. (1993). Effect of constraints by threonine on proline containing ot-helixma molecular dynamics approach. Biophys. Chem. 46, 77-89. Vogel, H. (1987). Comparison of the conformation and orientation of alamethicin and melittin in lipid membranes. Biochem. 26, 4562-4572. Vogel, H. (1992). Structure and dynamics of polypeptides and proteins in lipid membranes. Quart. Rev. Biophys. 25, 433-458. Vogel, H., & J~ihnig, E (1986). Models for the structure of outer-membrane proteins of Escherichia coli derived from Raman spectroscopy and prediction methods. J. Mol. Biol. 190, 191-199. von Heijne, G. (1991). Proline kinks in transmembrane oc-helices. J. Mol. Biol. 218, 499-503. von Heijne, G. (1992). Membrane protein structuremhydrophobicity analysis and the positive-inside rule. J. Mol. Biol. 225, 487-494. Wallace, B. A., Cascio, M., & Mielke, D. L. (1986). Evaluation of methods for the prediction of membrane protein secondary structures. Proc. Natl. Acad. Sci. USA 83, 9423-9427. Weber, P. C., & Salemme, F. R. (1980). Structural and functional diversity in 4-o~-helical proteins. Nature 287, 82-84. Weiss, M. S., Abele, U., Weckesser, J., Welte, W., Schiltz, E., & Schulz, G. E. (1991a). Molecular architecture and electrostatic properties of a bacterial porin. Science 254, 1627-1630. Weiss, M. S., Kreusch, A., Schiltz, E., Nestel, U, Welte, W., Weckesser, J., & Schulz, G. E. (1991b). The structure of porin from Rhodobacter capsulatus at 1.8 ,~ resolution. FEBS Lett. 280, 379-382. Wiener, M. C., & White, S. H. (1992). Structure of a fluid dioleoylphosphatidylcholine bilayer determined by joint refinement of X-ray and neutron diffraction data. III. Complete structure. Biophys. J. 61,434-447. Wilmsen, H. U., Leonard, K. R., Tichelaar, W., Buckley, J. T., & Pattus, E (1992). The aerolysin membrane channel is formed by heptamerization of the monomer. EMBO J. 11, 2457-2463. Woolfson, D. N., Mortishire-Smith, R. J., & Williams, D. H. (1991). Conserved positioning of proline residues in membrane-spanning helices of ion-channel proteins. Biochem. Biophys. Res. Comm. 175, 733-737. Wu, J., Tisa, L. S., & Rose, B. P. (1992). Membrane topology of the ArsB protein, the membrane subunit of an anion-translocating ATPase. J. Biol. Chem. 267, 12570-12576. Yeates, T. O., Komiya, H., Rees, D. C., Allen, J. P., & Feher, G. (1987). Structure of the reaction centre from Rhodobacter sphaeroides R-26: membrane-protein interactions. Proc. Natl. Acad. Sci. USA 84, 6438-6442. Yun, R. H., Anderson, A., & Hermans, J. (1992). Proline in o~-helix: stability and conformation studied by dynamics simulation. Proteins: Struct. Funct. Genetics 10, 219-228.

FATTY ACI D- AN D ISOPRENOi D-LI NKED MEMBRANE PROTEINS

Marco Parenti and Anthony i. Magee

I.

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

80

II.

Protein A c y l a t i o n . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

80

A.

Myristoylation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

81

B. Palmitoylation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . C. Functions o f A c y l a t i o n . . . . . . . . . . . . . . . . . . . . . . . . . . . Protein Prenylation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

83 84 86

A.

I s o p r e n o i d Biosynthesis . . . . . . . . . . . . . . . . . . . . . . . . . . .

87

B.

Protein Prenylation

. . . . . . . . . . . . . . . . . . . . . . . . . . . . .

87

C. D. E.

A s s o c i a t e d Modifications . . . . . . . . . . . . . . . . . . . . . . . . . . E n z y m o l o g y of Prenylation . . . . . . . . . . . . . . . . . . . . . . . . . Inhibition of C-terminal Processing . . . . . . . . . . . . . . . . . . . . .

89 91 92

F.

Function o f Protein Prenylation . . . . . . . . . . . . . . . . . . . . . . .

III.

IV.

94

Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

96

Acknowledgments

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

96

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

96

References

Biomembranes Volume 1, pages 79-105. Copyright 9 1995 by JAI Press Inc. All rights of reproduction in any form reserved. ISBN: 1-55938-658-4 79

80

MARCO PARENT! and ANTHONY I. MAGEE

I. I N T R O D U C T I O N In the last twenty years or so, evidence has been mounting for the covalent attachment of various kinds of lipid moieties to proteins. The 1980s showed an explosion of research in this area, resulting in the elucidation of the structure of four major lipid modifications: palmitoylation, myristoylation, isoprenylation and glycosyl-phosphatidylinositol linkage. The last of these, involving C-terminal addition of a complex glycosylated phosphatidylinositol moiety to a variety of cell surface proteins, has been reviewed recently and will not be dealt with here (McConville and Ferguson, 1993). These four modifications have been found only in eukaryotes, although prokaryotes do perform some different lipid modifications which will not be discussed here. Taken as a whole, these lipid modifications are found on a large number of cellular proteins and new examples are being identified at a high rate. Thus, several percent of proteins in any given cell are modified, making these modifications highly important for cell physiology, and suggesting a comparison with other better known modifications such as glycosylation and phosphorylation. Intuitively, one might imagine that addition of a very hydrophobic lipid moiety to a protein would give it an affinity for membranes. While this is true for many examples, it is not always so. Indeed, many lipid-modified proteins spend all or part of their lifetime in the cytosol. Also, it is beginning to appear that these modifications may often mediate protein-protein interactions rather than proteinmembrane interactions. The purpose of this chapter is to briefly discuss the major structural, enzymological, and functional features of acylation with the fatty acids myristate and palmitate, and of isoprenylation and its associated modifications. We do not intend to provide an exhaustive review of these fields, but rather to give an overview which illustrates the key points. Detailed reviews have appeared recently (Giannakouros and Magee, 1993; Newman and Magee, 1993).

II.

PROTEIN ACYLATION

Protein acylation refers to either co-translational or post-translational modifications that result in the covalent attachment of a fatty acid prosthetic group onto an acceptor protein. Two types of long chain saturated fatty acids serve as major substrates for these modifications, i.e., the 14-carbon myristic and the 16-carbon palmitic acids. Less commonly, several other fatty acids (such as stearic or oleic acids) have been found linked to proteins, and there are also interconversion reactions that occur after the fatty acid has been attached to the protein (Towler et al., 1988). In addition, two photoreceptor proteins, transducin and recoverin, are heterogeneously acylated at their amino-termini with one of four types of fatty acyl residues, including lauroyl (C12:0), myristoyl (C14:0), (cis-deltaS)-tetradecaenoyl

LipidModifications of Proteins

81

(C14:1) and (cis,cis-deltaS,delta8)-tetradecadienoyl (C14:2) moieties (Dizhoor et al., 1992; Neubert et al., 1992). Protein acylation was originally detected in viral membrane glycoproteins (Schmidt et al., 1979; Hruby and Franke, 1993), then extended to include an increasing number of cellular polypeptides with various cellular functions ranging from membrane receptors [e.g., the retinal pigment rhodopsin (Papac et al., 1992) and the ~2-adrenergic receptor (O'Dowd et al., 1989)], proteins involved in transmembrane signaling [e.g., the ct-subunits of many heterotrimeric guanylyl nucleotide-binding regulatory proteins (G-proteins) (Buss et al., 1987; Neubert et al., 1992; Parenti et al., 1993)], proteins involved in the tumorigenic transformation of cells [e.g., the oncoproteins p21 ras and pp60 src (Sefton et al., 1982; Schultz et al., 1985)], and to different enzymes [e.g., the catalytic subunit of cyclic AMP-dependent protein kinase (Carr et al., 1982) and nitric oxide synthase (Busconi and Michel, 1993)]. Palmitoylation of proteins is a post-translational modification. The fatty acid is attached via a thioester bond to internal cysteine residues (one or multiple sites within the same protein) that lie topologically close to the cytoplasmic face of cellular membranes (reviewed in Towler et al., 1988 and Schmidt, 1989). On the other hand, myristoylation is usually a co-translational modification that occurs at the amino terminus of a nascent protein by the time the growing peptide chain has reached 100 amino acid residues in length (Deichaite et al., 1988). The fatty acid is covalently attached to an amino-terminal glycine adjacent to the initiator methionine, via a hydroxylamine-resistant amide bond. However, it has recently been reported that in blood platelets, where protein synthesis is negligible when compared to that in nucleated cells, myristoylation of proteins may occur posttranslationally and via a thioester bond, i.e., with the same characteristics originally thought to be exclusive to the palmitoylation reaction (Muszbek and Laposata, 1993).

A. Myristoylation Myristoylation is catalyzed by the soluble enzyme myristoyl-CoA:protein Nmyristoyl transferase (NMT; EC 2.3.1.97). The enzyme was originally purified from Saccharomyces cerevisiae and its complete sequence is known (Towler et al., 1987a)---it is a 455-residue monomeric protein located in the cytoplasm and not detectably associated with membranes of any cellular organelles (Knoll et al., 1992). Recently, the human enzyme has been cloned and sequenced (Duronio et al., 1992). The gene sequence codes for a protein containing 416 amino acid residues with an estimated molecular mass of 48 kDa, which is 44% identical to S. cerevisiae NMT, and can complement the lethal phenotype of the nmtl-181 mutation of S. cerevisiae, which causes temperature-sensitive myristic acid auxotrophy. Human and yeast NMT have overlapping yet distinct protein substrate specificities as judged by a co-expression system that reconstitutes protein

82

MARCO PARENTI and ANTHONY I. MAGEE

N-myristoylation in Escherichia coli (which has no endogenous NMT activity) by sequentially inducing NMT and several homologous mammalian and yeast Ga proteins as myristate acceptors during growth in medium containing [3H]myristate (Duronio et al., 1991). Rat Goa and S. cerevisiae Gpal (the Ga involved in mating pheromone signal transduction) were equally myristoylated when co-expressed with human and yeast NMT, but the human Gza is N-myristoylated by the human enzyme only. The N-terminal 6-8 residues of NMT substrates appear to provide the principal recognition determinants for many N-myristoylated proteins (Gordon et al., 1991). There are three amino acid differences in this region of Goa and Gza (GCTLSAEE vs. GCRQSSEE, respectively). However, for Goa, recent studies indicate that amino acids C-terminal to these 6-8 residues may also play an important role in modulating enzyme-substrate interactions (Duronio et al., 1991). Kinetic studies have shown that S. cerevisiae NMT has a very high degree of specificity for 14-carbon fatty acids despite the fact that myristoyl-CoA appears to be a minor acyl-CoA species in both yeast and mammalian cells where it is 5-20-fold less abundant than palmitoyl-CoA (Johnson et al., 1993). The Km of palmitoyl-CoA for purified S. cerevisiae NMT is equivalent to that of myristoyl-CoA, but still the efficiency of transfer of C16:0 to a variety of peptide substrates is less than 5% that of C14:0 and palmitoyl-CoA acts as a competitive inhibitor of NMT (Smith and Powell, 1986). This raises the question as to how NMT is able to avoid inhibition by palmitoyl-CoA in vivo. It is possible that NMT is functionally segregated from palmitoyl-CoA due, at least in part, to the different solubilities of myristoyl-CoA and palmitoyl-CoA (critical micellar concentrations = 210 and 42 laM, respectively; Towler et al., 1987b) and differences in their ability to affiliate with membranes and/or intracellular binding proteins. Cooperative interactions between the acyl-CoA and peptidebinding sites of NMT contribute to its specificity (Heuckeroth et al., 1988). The elements required for recognition of acyl-CoA and peptide substrates by NMT reside in both halves of the enzyme (Rudnick et al., 1990). Peptide substrates must have a sterically unencumbered primary amino group associated with an N-terminal glycine residue. A broad spectrum of amino acids is permitted at positions 3 and 4, while strict amino acid requirements are exhibited at position 5, indicating that this residue plays a critical role in the binding of substrates to the enzyme (Duronio et al., 1991). Replacement of Ala in position 5 with Asp ablates the peptide's myristoyl-accepting activity. A serine at this position results in a decrease in the apparent Km by a factor of about 500. Basic residues are preferred at positions 7 and 8 over neutral residues which, in turn, are preferred over acidic residues (Duronio et al., 1991). Thus, the general consensus sequence for myristoylation is H2N(M)GXXXS/T/A/N/C. NMT has an ordered bi-bi reaction mechanism (Rudnick et al., 1991); a high affinity myristoyl-CoA:enzyme complex forms prior to binding of peptide. After catalytic transfer of myristate from CoA to peptide, CoA is released followed by myristoyl peptide. Both the high affinity myristoylCoA:NMT binary complex as well as the myristoyl-CoA:NMT:peptide ternary

Lipid Modifications of Proteins

83

complex have been characterized (Rudnick et al., 1990). The reaction mechanism of NMT implies that all events must be carefully orchestrated in both space and time within the cell. Purified S. cerevisiae NMT has no intrinsic methionylaminopeptidase (MAP) activity (Rudnick et al., 1991), indicating that the initiator methionine residue must be removed from the protein substrates by cellular MAP to expose their Gly 2 residues prior to addition of myristate. It is not known whether efficient processing requires a direct interaction between MAP and NMT and the translational apparatus, e.g., ribosomes. The minimal catalytic domain of NMT resides in sequences between Ile 59 ---) Phe 96 and Gly 451 ---) Leu 455. The amino terminal 59 residues of NMT may play an important noncatalytic role, functioning as a targeting signal so that this cytosolic enzyme can access the cellular myristoyl-CoA pool generated from activation of exogenous C14:0 by acyl-CoA synthetase(s) (Rudnick et al., 1992).

B. Palmitoylation Much less is known about the biochemical mechanisms of post-translational acylation of proteins with fatty acids through thioester linkages. Thus far, only partial purification of palmitoyl transferase has been reported (Kasinathan et al., 1990), not allowing determination of its complete structure. Utilizing a cell-free system for palmitoylation of viral acceptor proteins, an acylating enzyme activity has been characterized as an integral membrane protein, while no activity has been detected in the cytosol (Berger and Schmidt, 1985; Mack et al., 1987). A candidate site for the palmitoylation reaction is a pre-Golgi compartment (Rizzolo et al., 1985; Rizzolo and Kornfeld, 1988), a smooth peri-Golgi cisterna contiguous with the rough endoplasmic reticulum (ER) that may correspond to the transitional ER (Palade, 1975). Palmitoylation of transmembrane proteins, therefore, takes place after the translation of the respective polypeptide and before the trimming of peripheral mannose residues due to mannosidase I, a putative cis-Golgi function, has been completed (Bonatti et al., 1989). Palmitoylation sites for integral membrane proteins are usually close to the cytoplasmic surface of the transmembrane regions, suggesting that the active site of the enzyme is cytoplasmically exposed. Thus, it is necessary only to postulate the existence of a single enzyme for palmitoylation of this class of proteins as well as nonintegral proteins such as ras. However more than one enzyme may exist. The CoA esters of long chain fatty acids are the acyl-donors for the palmitoyl transferase, but the enzyme is much less specific for chain length and degree of unsaturation than is the N-myristoyl transferase. Thus, the fatty acids which are transferred to cysteine residues reflect to some extent the cellular abundance of acyl CoAs; palmitate (Cl6:0) usually being predominant but stearate (Cls:0), oleate (C18:1) and others also being found. There is no clear sequence requirement in the protein acceptor. However, the acylated cysteine residues usually occur in regions

84

MARCO PARENTI and ANTHONY I. MAGEE

close to the cytoplasmic surface of cellular membranes, being located there either by virtue of an associated transmembrane peptide sequence or another modification such as myristoylation or prenylation. Palmitoylation of cellular polypeptides decreases considerably after blocking protein synthesis, but significant incorporation of fatty acid always persists (Magee and Courtneidge, 1985; Mcllhinney et al., 1985; Olson and Spizz, 1986). Turnover of palmitic acid moieties during the lifetime of a protein has been observed in many instances, such as for the transferrin receptor (Omary and Trowbridge, 1981), ankyrin (Staufenbiel, 1987), GAP-43 (Kinsella et al., 1991) and p21 ras, where it shows a remarkably high velocity (tl/2 --20 min) (Magee et al., 1987). Studies of the peripheral membrane protein, ankyrin, have shown that it incorporates fatty acid even in mature erythrocytes that do not contain an extensive endoplasmic reticulum or Golgi apparatus (Staufenbiel, 1987). In these cells, protein synthesis has ceased, and ankyrin is stably integrated into the plasma membrane-associated cytoskeleton. From these arguments, it is concluded that both acylation and deacylation of ankyrin occur at the cytoplasmic side of the plasma membrane (Staufenbiel, 1988). As observed for acylation, the deacylation reaction is also enzyme-driven, and a membrane-associated protein:fatty acylesterase activity has been detected in diverse tissues (Berger and Schmidt, 1986; Basu et al., 1988; Bizzozero et al., 1992). The rapid reversibility of palmitoylation suggests that this modification is of regulatory significance for protein function(s) and the turnover rate of the reaction may be affected by extracellular and/or intracellular signals. Indeed cellular activators [e.g., the platelet-aggregating agents thrombin and phorbol 12-myristate 13-acetate (PMA) (Huang, 1989); hormones, e.g., insulin (Jochen et al., 1991); and mitogens, e.g., serum and growth factors (James and Olson, 1989)] alter the incorporation of palmitic acid into proteins. Furthermore, the palmitoylation of 132-adrenergic receptor is modulated by agonist stimulation (Mouillac et al., 1992).

C. Functions of Acylation Although the precise role(s) of protein acylation is still a matter of debate depending upon the unique characteristics of each acceptor protein, the most common belief is that acylation enhances membrane binding avidity as the addition of the fatty acid side-chains increases protein hydrophobicity and may alter protein folding (Bizzozero and Lees, 1986). Fatty acylated proteins are mostly membraneassociated, but many soluble myristoylated proteins exist, e.g., the catalytic subunit of cyclic AMP-dependent protein kinase, proving that myristoylation is not sufficient for membrane attachment (Towler et al., 1988). Furthermore, in other palmitoylated proteins, the attachment of fatty acids is definitely not required for membrane binding as specific transmembrane segments provide for sufficient anchorage (e.g., rhodopsin, 132-adrenergic receptor, and the sodium channel). These issues and more recent evidence have strengthened the idea of a more complex role of acylation in targeting proteins to specific membranes and/or in facilitating

Lipid Modifications of Proteins

85

protein-protein interactions or even protein functions. The interactions of the capsid proteins of polio virus support the involvement of myristoylation in proteinprotein interactions. In the Mahoney strain of type 1 poliovirus, capsid protein VP4 is myristoylated, and electron density maps of the virion indicate that myristic acid is an integral part of the protein shell. The hydrocarbon chain of the fatty acid molecule is in direct contact with the amino-termini of both VP4 and VP3, apparently stabilizing the interactions between these two proteins (Chow et al., 1987). Cellular transformation by Rous sarcoma virus pp60 v-src tyrosine kinase is dependent on its association with the plasma membrane and is prevented by mutation of the N-terminal myristoylated glycine (Kamps et al., 1986). Recently, pp60 v-src was found to bind to a cytoplasmic-face plasma membrane putative "receptor" in a specific and saturable manner that was dependent on the aminoterminal myristoyl moiety (Resh, 1989). A 32-kDa membrane protein has been identified as a putative component of the receptor for pp60 vsrc based on chemical cross-linking of a myristoylated amino-terminal pp60 vsrc peptide (Resh and Ling, 1990). The pp60 v-srcreceptor does not appear to be a general myristic acid receptor, since other myristoylated peptides do not compete for binding to the receptor in membranes, nor do they inhibit cross-linking. Therefore, in the case of pp60 vsrc, myristoylation provides a recognition signal to direct the protein to specific sites in the plasma membrane where it may come into contact with specific substrates whose phosphorylation activates the pathways leading to oncogenic transformation of cells. All known members of the src family of nonreceptor tyrosine kinases are N-terminally myristoylated and with the exception ofsrc itself and blk, also contain one or more nearby cysteine residues which are potential sites for palmitoylation (see below). Indeed, lck (Schultz et al., 1987), fyn and yes (Koegl et al., 1994) have been shown to incorporate palmitate, which could affect their subcellular localization and interactions with other proteins. Thus, the N-terminal regions of these proteins could become multiply acylated with unique spacings of the acyl groups. Similar concepts apply to other proteins involved in cellular signaling, e.g., heterotrimeric G-proteins. The a-subunits of several members of this family, e.g. ail, ai2, ai3, ao, az and at (transducin), are myristoylated (Buss et al., 1987; Jones et al., 1990; Mumby et al., 1990; Neubert et al., 1992), Myristoylation of the a-subunit of Go, an abundant G-protein in brain, is essential for its membrane attachment (Jones et al., 1990; Linder et al., 1991), and by increasing its affinity for the 13~t-subunit complex, facilitates the formation of the ternary complex. Presumably this finding extends to the other myristoylated G-protein a-subunits. Furthermore, the addition of myristic acid is absolutely necessary for ai2-mediated signal transduction and regulation of effector enzymes, e.g., adenylyl cyclase, in the cell (Gallego et al., 1992). More recently, palmitoylation of some G-protein a-subunits, e.g., ail, ai2, ai3, ao, az, aq and as, has been demonstrated (Linder et al., 1993; Degtyarev et al., 1993; Parenti et al., 1993). The fatty acid is attached onto a cysteine residue located in the N-terminal region of the proteins (Cys 3 in c~

86

MARCO PARENTI and ANTHONY I. MAGEE

and t~s). Therefore, the lipid modifications of G-protein o~-subunits offer a rather complicated and assorted picture, some of them being dually modified with both myristic and palmitic acids (the (Xi subfamily, Oto and O~z), others being only palmitoylated (Otqand Ors), and the retinal transducin being heterogeneously modified by saturated and unsaturated fatty acyl groups. The different set(s) of lipids carried by o~-subunits might contribute to the targeting of the proteins to the "right" coupling with receptors and/or cellular effectors, i.e., enzymes or ion channels. Thus far, the functional significance ofpalmitoylation oft~-subunits is still obscure, except that its prevention by site-directed mutagenesis renders Oto less stably attached to membranes (Parenti et al., 1993), but the same mutation does not affect membrane localization of Ors(Degtyarev et al., 1993). However, if this modification is reversible, as it appears to be in most cases, it may provide an additional mechanism for regulation of signal transduction. Interestingly, membrane receptors of the seven-transmembrane spanning domain family, e.g., 132-adrenergic receptor and rhodopsin, that couple to G-proteins, are also palmitoylated in their C-terminal cytoplasmic tail (Ovchinnikov et al., 1988; O'Dowd et al., 1989), a region that contributes to determine the specificity of their interaction with G-proteins (Conklin and Bourne, 1993). Indeed, mutation of the palmitoylated cysteine of the ]]2-adrenergic receptor into a nonpalmitoylated glycine leads to a form of receptor that is largely uncoupled from Gs and exhibits a drastically reduced ability to mediate agonist stimulation of adenylyl cyclase (O'Dowd et al., 1989). Moreover, the nonpalmitoylated receptor is highly phosphorylated (Moffett et al., 1993), as if it was in a desensitized state, i.e., uncoupled from adenylyl cyclase, induced by agonist exposure. Through modulation of the turnover of the palmitoyl moiety on the 13-receptor (Mouillac et al., 1992), agonists may very well affect the efficiency of receptor-effector coupling. Therefore, for its increasing importance in such cellular events as signal transduction and oncogenic transformation, the time has come for fatty acylation to deserve more respectful consideration as is attributed to other regulatory post-translational modifications of proteins, such as phosphorylation and glycosylation.

II!.

PROTEIN PRENYLATION

This section will deal with protein prenylation, reviewing the basic principles of this field and directing the reader to key references and reviews, as well as introducing the latest information available. Several reviews of this field have appeared recently (Clarke, 1992; Sinensky and Lutz, 1992; Giannakouros and Magee, 1993).

LipM Modifications of Proteins

87

A. Isoprenoid Biosynthesis Isoprenoids are derived from the five-carbon precursor mevalonic acid which is produced from hydroxymethylglutaryl (HMG) CoA by the reductase enzyme. Mevalonic acid is converted to the branched unsaturated compound, isopentenyl diphosphate, which is then polymerized to produce compounds containing multiples of five carbons. The key intermediates for protein prenylation are the 15-carbon compounds farnesyl diphosphate (FPP) and the 20-carbon compound geranylgeranyl diphosphate (GGPP). The GGPP used in protein prenylation is produced by a cytosolic enzyme which uses FPP and isopentenyl diphosphate (Ericsson et al., 1992; Runquist et al., 1992). Condensation of two molecules of FPP produces squalene, the precursor to the steroid family. FPP is also the precursor of dolichols, heme A, ubiquinone, vitamin A (retinoic acid), and a number of other important metabolites (see Goldstein and Brown, 1990; Giannakouros and Magee, 1993). Because of its importance and its role in cholesterol biosynthesis, this pathway has been extensively studied. Specific inhibitors (e.g., compactin, mevinolin, lovastatin, and pravastatin) of HMGCoA reductase, the enzyme catalyzing the first step in cholesterol biosynthesis, have been identified and widely used clinically to control hypercholesterolemia (Endo et al., 1976; Alberts et al., 1980). These compounds have proven useful in the study of protein prenylation since they deplete the cell of FPP and GGPP, thus preventing processing of prenylated proteins, and allowing efficient labeling of isoprenoids by added exogenous radioactive mevalonic acid (Schmidt et al., 1984; Maltese and Sheridan, 1987; Hancock et al., 1989).

B. Protein Prenylation Linkage of derivatives of mevalonic acid to proteins and the importance of these modifications for cellular physiology had been noted throughout the 1980s (Schmidt et al., 1984; Maltese and Sheridan, 1987; Beck et al., 1988). Protein prenylation has been observed in yeasts (Giannakouros et al., 1993a), mammalian cells, and recently in plants (Swiezewska et al., 1993). However, it was not until the late 1980s that the chemical and biological properties of protein prenylation became clear. Studies on ras proteins and nuclear lamins (Wolda and Glomset, 1988; Casey et al., 1989; Hancock et al., 1989), as well as fungal mating factors (Kamiya et al., 1979; Sakagami et al., 1979; Brake et al., 1985), led to the identification of the C-terminal CXXX motif (C = cysteine, X = any amino acid) as a signal for thioether-linked prenylation. Two types of CXIX2X3 motif have now been defined (see Table 1). When X3 is a small, hydrophilic amino acid (e.g., S, M, C, Q), a farnesyl group is transferred from FPP to the sulfhydryl of the cysteine of the CXXX motif. When X3 is a large hydrophobic residue (e.g., L, F, I, N), the attached isoprenoid is geranylgeranyl derived from GGPP. The residue in the X1 and X2 positions can also influence prenylation (see Giannakouros and Magee, 1993; Newman and Magee, 1993). The CCKVL motif of rhoB is unusual in that

88

MARCO PARENTI and ANTHONY I. MAGEE Table 1. Signals for C-terminal Prenyl Modifications of Proteins

Signal

Prenoid Modification

Enzyme

Other Modifications

Farnesyl (X3 = S, M, A, C, Q)

Protein Famesyl Transferase (PFT)

XXX Proteolysis carboxylmethylation (Palmitoylation)

Geranylgeranyl (X3 = L, F, I, N)

Protein Geranylgeranyl Transferase I (PGGTI)

XXX Proteolysis carboxylmethylation (Palmitoylation)

-CXC

Double geranylgeranylation

Protein Geranylgeranyl Carboxyl-methylation Transferase II (PGGTII)

-XCC

Geranylgeranylation

Protein Geranylgeranyl (Palmitoylation) Transferase II (PGGTII)

-CX1X2X3

both farnesyl and geranylgeranyl groups appear to be transferred to it (Adamson et al., 1992). The data suggest that the upstream cysteine may be a site for geranylgeranylation while the downstream cysteine may be farnesylated. Also, the CXXX motifs of some rab family members, e.g., rab5 and rabl 1, are modified differently, demonstrating that the context as well as the sequence of the CXXX motif is crucial (Cox et al., 1993). It should be emphasized that not all CXXX motifs can signal prenylation. The CGLF motif of several G-protein t~-subunits is not significantly modified by prenylation, probably due to the inhibitory effects of the glycine residue in position Xl (Jones and Spiegel, 1990; C.M.H. Newman, M. Parenti and A.I. Magee, unpublished observations) and also to its context (Cox et al., 1993). A large number of cellular polypeptides are now known to be prenylated at CXXX motifs. These include nuclear lamins, many ras superfamily members, G-protein T-subunits, phosphodiesterases, and kinases (see Glomset et al., 1990; Maltese, 1990; Lorenz et al., 1991; Anant et al., 1992; Giannakouros and Magee, 1993; Newman and Magee, 1993). Two other cysteine-containing C-terminal motifs have been clearly shown to be signals for prenylation. These are XCC and CXC and thus far have been found primarily in the rab proteins of the ras superfamily. Only geranylgeranyl has been found attached to these motifs. In the case of the CXC motif, both chemical analysis of purified protein (Farnsworth et al., 1991) and site-directed mutagenesis (Giannakouros et al., 1993b) show that both cysteines are prenylated. The situation with the XCC motif is less clear. No direct evidence for prenylation of both cysteines has been published, although mutational analysis of S. cerevisiae SEC4 and YPT1 proteins suggests that either cysteine can support function, whereas deletion of both cysteines is lethal (Molenaar et al., 1988; Walworth et al., 1989). The S. pombe XCC protein YPT3 has been found to incorporate palmitate as well as geranylger-

Lipid Modifications of Proteins

89

anyl whereas the S. pombe YPTI protein does not appear to be palmitoylated despite also having an XCC motif (Newman et al., 1992). Thus, sequences upstream of the XCC motif may influence the lipid modifications which are appended to the protein. This is supported by the observations that: (a) chimeric rab proteins containing XCC or CXC motifs fail to be correctly processed unless an extensive region of upstream sequence (usually 3-40 residues) is included in the chimera (Chavrier et al., 1991) and (b) the effector domain of rab 1B can affect its prenylation (Wilson and Maltese, 1993). This is in direct contrast to the situation with the CXXX motif, where transfer of this tetrapeptide alone can signal correct C-terminal modification (Hancock et al., 1989). In addition to the well characterized prenylation with farnesol or geranylgeraniol, two other derivatives of the isoprenoid biosynthetic pathway have been reported to be covalently bound to proteins. These are steroid hormones (Takahashi and Breitman, 1992) and retinoids (Takahashi and Breitman, 1989; Takahashi et al., 1991). The nature of the linkages to protein is unclear, evidence having been presented for thioester linkage of both steroids and retinoids (Takahashi and Breitman, 1989, 1992) and against thioester linkage of retinoids (Takahashi et al., 1991). Also, few data are available on the nature of the actual moieties attached to the protein. The significance of these types of modification, therefore, cannot be assessed at this time.

C. Associated Modifications In addition to prenylation, many proteins containing CXXX, CXC, or XCC motifs can be modified by palmitoylation, proteolytic processing and/or carboxylmethylation. Prenylation appears to be the first of these modifications, and is required before the others can occur. After prenylation of CXXX motifs, the XXX residues are proteolytically removed by a membrane-bound peptidase. Several activities capable of cleaving CXXX peptides have been detected (Akopyan et al., 1992; Ashby et al., 1992; Hrycyna and Clarke, 1992; Ma and Rando, 1992). However, most of these can be eliminated due to incorrect subcellular localization (e.g., cytosolic or mitochondrial) or inappropriate substrate specificity (i.e., cleavage of unprenylated as well as prenylated substrate). The best candidates for an authentic prenyl-protein cleaving enzyme are membrane-associated and preferentially cleave prenylated peptides. One of these is a thiol-dependent serine type carboxypeptidase (Akopyan et al., 1992). A better characterized activity is a protease which preferentially removes XXX as a tripeptide, but which can act as a carboxypeptidase and remove a single residue (Ashby et al., 1992; Ma and Rando, 1992). This enzyme will hydrolyze synthetic peptide substrates containing geranyl, farnesyl, or geranylgeranyl groups (Ma et al., 1992). It is much more active on peptides with L-amino acids in the X positions although D-amino acids can be weakly tolerated at X3. The protease does not appear to require upstream sequence for activity, but does require a blocked or-amino group on the substrate. It is likely

90

MARCO PARENTI and ANTHONY I. MAGEE

that a single protease, relatively nonspecific for the XXX amino acids, could be responsible for proteolytic processing of all prenylated CXXX sequences. Subsequent to proteolytic processing, the resulting C-terminal tx-carboxyl group of the prenylated cysteine residue is carboxyl-methylated by a membrane-associated methyltransferase using S-adenosyl methionine as methyl donor (Clarke et al., 1988; Hancock et al., 1991 a). The methyltransferase has not yet been purified, but in yeast, it is the product of the STE14 gene (Hrycyna and Clarke, 1990; Hrycyna et al., 1991). This enzyme is required for processing of the S. cerevisiae mating factor a, hence the sterile phenotype observed when the gene is inactivated. The STE14 gene product appears sufficient to account for all C-terminal carboxylmethylation in S. cerevisiae. XCC and CXC motifs do not appear to be proteolytically processed after prenylation, but the latter motif is carboxyl-methylated. Studies using artificial substrates and inhibitors have revealed a very simple substrate requirement for the methyltransferase enzyme. All that is required is a C-terminal prenylated cysteine residue with a free tx-carboxyl group, and a prenyl group of more than 10 carbons for efficient methylation; farnesylated or geranylgeranylated substrates are equally as good (Tan et al., 1991; Volker et al., 1991; Prrez-Sala et al., 1992). Thus, the simplest compound which will serve as a substrate is S-prenyl propionic acid. An apparently endoplasmic reticulum-associated prenyl substrate: carboxyl methyltransferase, has recently been identified in rat liver (Stephenson and Clarke, 1992). The enzyme is sensitive to N-ethyl maleimide, suggesting that an active site cysteine is involved. The extreme detergent-sensitivity of the enzyme, however, has precluded its purification. Many proteins which are prenylated at CXXX motifs subsequently become palmitoylated at upstream cysteine residues. In the case of N- and H-ras proteins, there are one or two cysteines within six residues of the prenylated cysteine which can act as palmitoylation sites (Hancock et al., 1989, 199 lb). As mentioned earlier, there does not appear to be a clear signal sequence for palmitoylation. Rather, the distance between the prenylation site and the palmitoylation sites seems to be crucial. Work in this laboratory (C.M.H. Newman and A.I. Magee, unpublished observations) suggests that cysteines moved as little as 2-5 residues upstream of the normal palmitoylation sites have vastly reduced abilities to act as palmitate acceptors. This may correlate with the accessibility of the active site of a membranebound acyl transferase to the cysteines (Gutierrez and Magee, 1991). It is clear that prenylation alone is generally insufficient for efficient membrane binding and subcellular targeting. Upstream sequences contribute crucially to both of these phenomena. In many instances, the palmitoylation sites are the relevant cooperative foci. In other cases, such as the K(B)-ras protein and a number of other ras-related proteins, upstream polybasic regions cooperate with prenylated C-termini to effect correct targeting (Hancock et al., 1990). This targeting appears essential for biological function, so that its reversibility could contribute to regulation of that function. Prenylation and proteolytic processing are chemically and

Lipid Modifications of Proteins

91

biologically irreversible modifications. However, both palmitoylation (Magee et al., 1987) and carboxyl-methylation (Chelsky et al., 1987; P6rez-Sala et al., 1991) are dynamically reversible in vivo and have the potential to be regulatory modifications. These modifications exhibit parallels with phosphorylation as a regulatory modification and the elucidation of their relevance will depend on the characterization of the enzyme activities involved, which is as yet at an early stage.

D. Enzymology of Prenylation At least two enzymes have been identified in both mammals and yeast which can prenylate CXXX motifs (see Table 1). Protein:farnesyl transferase (PFT) recognizes CX1X2X3 motifs (where X3 = S, M, A, C, Q) and is a Zn 2+ metalloenzyme (Chen et al., 1993) composed of two subunits designated o~(49 kDa) and 13(46 kDa) in mammals. The [3-subunit is homologous to the product of the yeast RAM 1 gene and is responsible for peptide substrate binding and also has a binding site for FPP (Omer et al., 1993). The o~-subunit is equivalent to the product of the yeast RAM2 gene and also plays a role in catalysis (Andres et al., 1993a). Both subunits are required for enzyme activity and stability. Protein:geranylgeranyl transferase I (PGGTI) recognizes proteins with X3 = L, F, I, N and also has o~- and 13-subunits (Yokoyama and Gelb, 1993). The 13-subunit (43 kDa in the cow) is specific to PGGTI and is homologous to the S. cerevisiae CDC43 gene product (Mayer et al., 1992) and the S. pombe cwg 2+ gene (Diaz et al., 1993). The cz-subunit is identical to that ofPFT (Seabra et al., 1991; Moomaw and Casey, 1992). This enzyme is also a metalloprotein, requiring both Mg 2+ and Zn 2+ for optimal activity (Moomaw and Casey, 1992). Although PFT and PGGTI show preferential use of FPP and GGPP respectively, and preferential activity on the CXXX motifs cited above, they do show some cross-specificity both for diphosphate substrate and for CXXX motif (Caplin et al., 1993; Evans-Trueblood et al., 1993) resulting in some CXXX motifs becoming modified with a mixture of isoprenoid groups. One enzyme (PGGTII) seems to be responsible for geranylgeranylation of both cysteines in CXC motifs and of XCC motifs as well as some CXXX motifs found only in rab proteins (e.g., CCSN of rab5; Kinsella and Maltese, 1992; Cox et al., 1993; T.R. Brtva, R. Khosravi-Far, M. Sinensky, and C.J. Der, personal communication). Unlike PFT and PGGTI, PGGTII is inactive on small synthetic substrates, requiring considerable upstream sequence. It also differs from the other two enzymes in actually being inhibited by Zn 2+ (Seabra et al., 1992b). This enzyme consists of two components, A and B, which are separable in high salt. Component B is a heterodimer of cz (60 kDa) and 13(38 kDa) subunits homologous to those of the other protein prenyl transferases. The S. cerevisiae MAD2 and BET2 genes seem to be the equivalents of the mammalian cz and 13 subunits (Armstrong et al., 1993). Component A is a 95-kDa protein required for enzyme activity and function as it presents unmodified rab protein to component B and remains bound to it until

92

MARCO PARENTI and ANTHONY !. MAGEE

the geranylgeranylated product is removedneither by detergent in vitro, or in vivo more likely by a prenylated rab binding protein such as a guanine-nucleotide dissociation inhibitor (GDI), or by binding to cellular lipid bilayers (Andres et al., 1993b). These authors have proposed the name Rab Escort Protein (REP) for component A. REP shows homology to rab3A GDI, a prenylated rab-binding protein which forms a soluble complex with rab3A thus facilitating its recycling during its active cycle. REP is identical to the product of the human choroideremia gene, which is defective in the eponymous retinal degenerative disease (Seabra et al., 1992a). A related, but distinct gene, has been identified in humans which localizes in the same chromosomal region (1 q31-ter) as the Usher syndrome type II locus (Cremers et al., 1992) and the clinical similarities between these two diseases testify to analogous molecular etiologies.

E. Inhibition of C-terminal Processing Considerable effort has been expended in trying to develop inhibitors for C-terminal prenylation and processing steps. The motivation is to find inhibitors which would interfere with the action of transforming ras proteins, which appear to play a role in as many as 30% of all human cancers (Barbacid, 1987). Such inhibitors could also be useful as anti-inflammatory agents, by blocking the processing of ras-related proteins required for the oxidative burst in phagocytes (Bokoch and Prossnitz, 1992). Several problems are obvious. First, ras proteins are farnesylated, so inhibitors must selectively act on ras farnesylation over the farnesylation of other important cell proteins such as nuclear lamins. Second, geranylgeranylated proteins must be unaffected and inhibitors must be cell permeant. All processing steps have potential for inhibition. The reversible steps, palmitoylation and carboxyl-methylation, should be more rapidly affected than the irreversible steps, prenylation and proteolysis. They also have the possible advantages that both addition and removal of the modification could be targeted and that only a subset of cellular proteins are reversibly palmitoylated or carboxyl-methylated. However, little is known about the enzymes carrying out either modification. Simple inhibitors of carboxyl-methylation have been devised, such as N-acetyl farnesyl cysteine (Volker et al., 1991; P6rez-Sala et al., 1992). Inhibition of carboxyl-methylation has been found to block function of signal transducing pathways involving ras-related proteins and G-proteins (Huzoor-Akbar et al., 1993; Scheer and Gierschik, 1993). However, the apparent lack of specificity of the enzyme makes it unlikely that inhibitors will be sufficiently selective to be clinically useful (Sobotka-Briner and Chelsky, 1992). Similarly, a single C-terminal protease may be responsible for proteolysis of all prenylated sequences, reducing its attractiveness as a target for chemotherapy. Progress has, however, been made in designing potent, specific inhibitors of this enzyme (Ma et al., 1993). The data suggest that the enzyme is likely to be of the

Lipid Modifications of Proteins

93

serine or cysteine type, although it is not inhibited by group-specific inhibitors of either class of enzyme. It is inhibited competitively by noncleavable tetrapeptidebased analogs and by N-Boc-S-all trans-farnesyl-L-cysteine aldehyde, a transition state analog. By far, the greatest success has been achieved with inhibitors of prenylation. Inhibitors of the early steps of isoprenoid biosynthesis, such as HMGCoA reductase, might be expected to have too many side effects due to their general effects on biosynthesis of all isoprenoid end products. Indeed, the doses required to block protein prenylation are considerably higher than those which block steroid synthesis (see Goldstein and Brown, 1990). Nevertheless, in tissue culture model systems, specific effects of this type of inhibitor on ras-dependent pathways, e.g., responses to EGF, insulin and IGF- 1, have been reported, while other responses, e.g., to NGF, PDGF and bFGF, are unaffected (see Giannakouros and Magee, 1993). Pathways involving ras-related proteins, such as rho and rac which regulate the actin cytoskeleton and NADPH oxidase in phagocytes, can also be targeted (Bokoch and Prossnitz, 1992; Fenton et al., 1992; Bifulco et al., 1993). Another potential problem is the possible effects on prenylation of essential growth-related proteins, notably nuclear lamins. However, in most cases tested, little or no affect on nuclear lamin processing is seen at concentrations of inhibitor which interfere with ras activity. This may be due to differences in the affinity of the CXXX motifs of lamins and ras proteins for the PFT thus allowing a window within which ras processing can be selectively blocked. The antitumour effects of the steroid, dehydroepiandrosterone, may also be due to its inhibition of isoprenoid biosynthesis (Schulz and Nyce, 1991). The plant terpene, d-limonene, has recently been found to inhibit prenylation of proteins in vitro and to have anti-cancer activity in vivo (Crowell et al., 1991). This compound is metabolized in vivo into the more active derivatives, perillic acid and perillyl alcohol, and acts upstream from mevalonate in the isoprenoid pathway. It has been suggested that d-limonene may act directly on protein:prenyl transferases, but it is equally likely to block a stage of isoprenoid biosynthesis. Microbial screens have also been used to identify compounds which block protein prenylation. The mycotoxin, patulin, inhibits protein prenylation in cultured cells and inhibits PFT in vivo as well (Miura et al., 1993); however its mechanism of action is unclear. Streptomyces species have been found to produce a group of antibiotics related to manumycin which inhibit PFT in vitro, suppress a prenylation-dependent defect in a genetically engineered yeast strain, and inhibit growth of a K-ras-transformed fibrosarcoma in vivo (Hara et al., 1993). These compounds bear a structural resemblance to the isoprenoid side-chain, and competitively inhibit PFT with respect to FPP, but noncompetitively inhibit PFT with respect to protein acceptor. Due to the hydrophobicity of these compounds, they may be more easily taken up by target cells. However, their specificity and side-effects need to be assessed. An extensive screen of microbial products has revealed three other types of PFT inhibitors which compete with FPP and show selectivity for PFT over PGGT (Gibbs

94

MARCO PARENTI and ANTHONY I. MAGEE

et al., 1993). These are (ct-hydroxyfarnesyl) phosphoric acid, chaetometillic acids, and zaragozic acids, but their usefulness in vivo may be restricted by their high negative charge. More attention has focused on the design of peptide analogs based on the CXXX tetrapeptide sequence (Moores et al., 1991; Reiss et al., 1991). Extensive studies to define the CX1X2X3 sequence requirements of the PFT and PGGT enzymes, combined with molecular modeling, have led to the generation of excellent lead compounds as specific inhibitors of PFT. Two approaches have been taken. James et al. (1993) substituted the X1 and X2 positions with a benzodiazepine peptidomimetic which increased the potency of the inhibitor and its penetration into cells. Effects on prenylation of many proteins, including ras and nuclear lamins, were seen in cell culture, and ras-transformed cells were reverted to a normal phenotype. Kohl et al. (1993) modified a known tetrapeptide inhibitor by introducing a C-terminal homoserine lactone, and substituting two of the peptide bond carbonyls with noncleavable methylenes. These changes had the joint effects of: (a) increasing the stability and permeability of the compound, which was shown selectively to inhibit PFT in vitro, (b) blocking the processing of ras in cell culture, and (c) selectively blocking the growth of ras-transformed tumor cells in culture. These compounds have exciting possibilities as anti-tumor drugs but in vivo tests have not yet been reported.

F. Function of Protein Prenylation It is clear that prenylation of proteins increases their hydrophobic character. Farnesol is similar in hydrophobicity to palmitate, while geranylgeraniol is considerably more hydrophobic (Black, 1992). In most, if not all cases, prenylated proteins are found at least partly associated with cellular membranes and there are many examples showing that prenylation is required for membrane binding and function (e.g., Horiuchi et al., 1992; Kuroda et al., 1992; Itoh et al., 1993). However, a simple intercalation of the isoprenoid moiety into the phospholipid bilayer seems unlikely due to the relative rigidity of the unsaturated backbone, and the presence of bulky methyl side-chains. It is possible that isoprenoid groups could interact directly with membrane subdomains, possibly by lateral colligative interactions between protein-bound prenoids or by interactions with other isoprenoid compounds such as dolichols. To our knowledge, no clear evidence exists for direct insertion of any fatty acid or isoprenoid moiety on a protein into a lipid bilayer. It is, therefore, quite likely that prenyl groups interact with proteins at the membrane surface to mediate binding. This is analogous to the lipid-protein interactions discussed earlier for myristate in poliovirus and pp60 src. One report of such a receptor has appeared (Thissen and Casey, 1993). This protein binds both farnesylated and geranylgeranylated peptides, but not myristoylated peptides, with high affinity. The receptor appears to require an intact CXXX box, suggesting it is involved in capturing newly prenylated peptides and delivering them to the protease and methyltransferase

LipidModifications of Proteins

95

enzymes. Nevertheless, it sets a precedent which may have wider significance for the binding of prenylated proteins to membranes. In few cases is prenylation alone apparently enough to signal membrane binding and correct intracellular targeting. Farnesylated C-termini must cooperate with nearby sequences, which either have a polybasic nature or contain one or more sites for palmitoylation (Hancock et al., 1990, 1991 b) in order to be membrane-bound and targeted. In some cases, a single geranylgeranyl moiety is sufficient to produce membrane binding (Hancock et al., 1991b), while in others two geranylgeranyl groups or additional modifications appear to be necessary (Butrynski et al., 1992; Giannakouros et al., 1993a; Li and Stahl, 1993). Thus, simple hydrophobicity is not the only criterion, but the exact nature of the membrane binding and targeting mechanisms are obscure at the moment. It seems likely, however, that cooperative interactions between prenyl groups and nearby positively charged or palmitoylated sequences involve interactions with both the lipid bilayer and protein "receptors" that may bind to the protein itself or to the lipid modifications. The local topography of the modified sites, their spacing with respect to each other, and their frequency (i.e., the number of modifications), in addition to the structure of the modifying groups (isoprenoid cf. fatty acid) are probably all important. The targeting phenomenon could be due to: (a) the colocalization of protein "receptors" specific for the targeted protein, (b) the lipid bilayer composition in different compartments or sub-domains, or (c) a combination of both. One case where prenylated C-terminal structures clearly interact directly with proteins is in the ras superfamily of small GTP-binding proteins. Guanine nucleotide exchange on these proteins is regulated by families of molecules called guanine nucleotide dissociation stimulators (GDS) or inhibitors (GDI), collectively known as guanine nucleotide exchange proteins or GEPs (reviewed in Takai et al., 1992). In addition to their effects on the nucleotide state of the ras-related protein (rrp), the GEPs are often also involved in regulating membrane binding (Kawamura et al., 1993). The GEPs, acting as a chaperone, can bind to the hydrophobic tail of the rrp and remove it from cellular membranes resulting in the formation of a soluble 1:1 complex which can mediate the physical recycling of the rrp that accompanies its active cycle (Bourmeyster et al., 1992; Regazzi et al., 1992). GEPs can be quite specific for individual rrps, or can act on a spectrum of them. Strangely, a single GEP can bind rrps with somewhat different C-terminal modifications, again highlighting the complex recognition which must be occurring. The interaction between GEPs and rrps can also be modulated by phosphorylation of the rrp, frequently close to the C-terminus (Bokoch et al., 1991; Hata et al., 1991; Miura, Y. et al., 1992; van der Sluijs et al., 1992), or by phosphorylation of the GEP itself (SteeleMortimer et al., 1993). Thus, GEPs can regulate the activity of rrps and their downstream effector pathways both by changing the nucleotide state and by altering subcellular localization and access to activators and substrates (Ando et al., 1992; Bokoch, 1993; Hancock and Hall, 1993).

96

MARCO PARENTi and A N T H O N Y I. MAGEE

IV. CONCLUSIONS Fatty acid and isoprenoid modifications are important for the membrane association and protein-protein interactions of many cytoplasmically disposed proteins. Aclear theme in the use of these modifications is that they can and usually do work, in a complex and poorly understood way, in concert with each other and with associated modifications (proteolysis, carboxyl-methylation), or with primary sequences within the protein carrier (see Magee and Newman, 1992). In some cases, e.g., palmitoylation, direct interaction with membrane bilayers may be important whereas in others (myristoylation, prenylation), it seems likely that binding to proteins plays a role. The in vivo reversibility o f s o m e of the modifications (palmitoylation, carboxyl-methylation) is clearly linked to regulation of protein function. Thus, the isolation and characterization of the enzymes involved in the addition and removal of these modifications, and in their regulation, will be crucial in understanding the role of lipid modification in protein function. In the cases of irreversible modification (myristoylation, prenylation), conformation changes in the protein carrier or binding to carrier proteins (e.g., GEPs), possibly regulated by phosphorylation, may regulate the exposure of the lipid moiety (Inglese et al., 1992; see Pfeffer, 1992). The wide range of important proteins involved in signal transduction, cell transformation and viral replication, which become modified by these lipids make them an excellent target for chemotherapeutic intervention.

ACKNOWLEDGMENTS We thank our many colleagues for the provision of pre-publication data and for helpful discussions and apologize to those whose work has not been cited due to considerations of space. We are also grateful to members of this laboratory, especially Chris Newman and Thomas Giannakouros. We thank Marilyn Brennan for outstanding secretarial help.

REFERENCES Adamson, P., Marshall, C. J., Hall, A., & Tilbrook, P. A. (1992). Post-translational modifications of p21rho proteins. J. Biol. Chem. 267, 20033-20038. Akopyan, T. N., Couedel, Y., Beaumont, A., Fournie-Zaluski, M.-C., & Roques, B. E (1992). Cleavage of farnesylated COOH-terminal heptapeptide of mouse N-ras by brain microsomal membranes. Biochem. Biophys. Res. Comm. 187, 1336-1342. Alberts, A. W., Chen, J., Kuron, G., Hunt, V., Huff, J., Hoffman, C., Rothrock, J., Lopez, M., Joshua, H., Harris, E., Patchett, A., Monaghan, R., Currie, S., Stapley, E., Albers-Schonberg, G., Hirschfield, J., Hoogsten, K., Liesch, J., & Springer, J. (1980). Mevinolin: a highly potent competitive inhibitor of hydroxymethylglutaryl-coenzyme.A reductase and a cholesterol-lowering agent. Proc. Natl. Acad. Sci. USA 77, 3957-3961. Anant, J. S., Ong, O. C., Xie, H., Clarke, S., O'Brien, E, & Fung, B. K. K. (1992). In vivo differential prenylation of retinal cyclic GMP phosphodiesterase catalytic subunits. J. Biol. Chem. 267, 687-690.

LipM Modifications of Proteins

97

Ando, S., Kaibuchi, K., Sasaki, T., Hiraoka, K., Nishyama, T., Mizuno, T., Asada, M., Nunoi, H., Matsuda, I., Matsuura, Y., Polakis, P., McCormick, F., & Takai, Y. (1992). Post-translational processing of racp21 is important both for their interaction with the GDP/GTP exchange proteins and for their activation of NADPH oxidase. J. Biol. Chem. 267, 25709-25713. Andres, D. A., Goldstein, J. L., Ho, Y. K., & Brown, M. S. (1993a). Mutational analysis of (x-subunit of protein famesyltransferase. J. Biol. Chem. 268, 1383-1390. Andres, D. A., Seabra, M. C., Brown, M. S., Armstrong, S. A., Smeland, T. E., Cremers, E P. M., & Goldstein, J. L. (1993b). cDNA cloning of component A or Rab geranylgeranyl transferase and demonstration of its role as arab escort protein. Cell 73, 1091-1099. Armstrong, S. A., Seabra, M. C., Siidhof, T. C., Goldstein, J. L., & Brown, M. S. (1993). cDNA cloning and expression of the (x and 13subunits of rat Rab geranylgeranyl transferase. J. Biol. Chem. 268, 12221-12229. Ashby, M. N., King, D. S., & Rine, J. (1992). Endoproteolytic processing of a farnesylated peptide in vitro. Proc. Natl. Acad. Sci. USA 89, 4613-4617. Barbacid, M. (1987). Ras genes. Annu. Rev. Biochem. 56, 779-828. Basu, A., Glew, R. H., Evans, R. W., & Bandik, G. (1988). Isolation and characterization of a fatty acyl esterase from rat lung. Arch. Biochem. Biophys. 261,384-393. Beck, L. A., Hosick, T. J., & Sinensky, M. (1988). Incorporation of a product of mevalonic acid metabolism into protein in Chinese hamster ovary cell nuclei. J. Cell Biol. 107, 1307-1316. Berger, M., & Schmidt, M. F. G. (1985). Protein fatty acyltransferase is located in the rough endoplasmic reticulum. FEBS Letts. 187, 289-294. Berger, M., & Schmidt, M. E G. (1986). Characterization of a protein fatty acylesterase present in microsomal membranes at diverse origin. J. Biol. Chem. 261, 14912-14918. Bifulco, M., Laezza, C., Aloj, S. M., & Garbi, C. (1993). Mevalonate controls cytoskeleton organization and cell morphology in thyroid epithelial cells. J. Cell. Physiol. 155, 340-348. Bizzozero, O. A., & Lees, M. B. (1986). Spectroscopic analysis of acylated and deacylated myelin proteolipid protein. Biochemistry 25, 6762-6768. Bizzozero, O. A., Leyba, J., & Nunez, D. J. (1992). Characterization of proteolipid protein fatty acylesterase from rat brain myelin. J. Biol. Chem. 267, 7886-7894. Black, S. D. (1992). Development of hydrophobicity parameters for prenylated proteins. Biochem. Biophys. Res. Comm. 186, 1437-1442. Bokoch, G. M. (1993). Biology of the Rap proteins, members of the ras superfamily of GTP-binding proteins. Biochem. J. 289, 17-24. Bokoch, G. M., & Prossnitz, V. (1992). Isoprenoid metabolism is required for stimulation of the respiratory burst oxidase of HL-60 cells. J. Clin. Invest. 89, 402-408. Bokoch, G. M., Quilliam, L. A., Bohl, B. P., Jesaitis, A. J., & Quinn, M. T. (1991). Inhibition of RaplA binding to cytochrome b558 of NADPH oxidase by phosphorylation of RaplA. Science 254, 1794-1796. Bonatti, S., Migliaccio, G., & Simons, K. (1989). Palmitoylation of viral membrane glycoproteins takes place after exit from the endoplasmic reticulum. J. Biol. Chem. 264, 12590-12595. Bourmeyster, N., Satsia, M.-J., Garin, J., Gagnon, J., Boquet, P., & Vignais, P. V. (1992). Copurification of Rho protein and the Rho-GDP dissociation inhibitor from bovine neutrophil cytosol. Effect of phosphoinositides on Rho ADP-ribosylation by the C3 exoenzyme of Clostridium botulinum. Biochemistry 31, 12863-12869. Brake, A. J., Brenner, C., Najarian, R., Laybourn, P., & Merryweather, J. (1985). Structure of genes encoding precursors of the yeast peptide mating pheromone or-factor. In: Protein Transport and Secretion. (Gething, M. J., ed.), p. 103, Cold Spring Harbor, New York. Busconi, L., & Michel, T. (1993). Endothelial nitric oxide synthase. J. Biol. Chem. 268, 841 0-8413. Buss, J. E., Mumby, S. M., Casey, P. J., Gilman, A. G., & Sefton, B. M. (1987). Myristoylated ct subunit of guanine nucleotide-binding regulatory proteins. Proc. Natl. Acad. Sci. USA 84, 7493-7497.

98

MARCO PARENTI and A N T H O N Y I. MAGEE

Butrynski, J. E., Jones, T. L. Z., Backlund, J., & Spiegel, A. M. (1992). Differential isoprenylation of carboxy-terminal mutants of an inhibitory G-protein ct-subunit: neither famesylation nor geranylgeranylation is sufficient for membrane attachment. Biochemistry 31, 8030-8035. Caplin, B. E., Hettich, L. A., & Marshall, M. S. (1994). Substrate characterization of the Saccharomyces cerevisiae protein farnesyltransferase and type-I protein geranylgeranyltransferase. Biochim. Biophys. Acta 1205, 39-48. Carr, S. A., Biemann, K., Shoji, S., Parmelee, D. C., & Titani, K. J. (1982). n-Tetradecanoyl is the NH2-terminal blocking group of the catalytic subunit of cyclic AMP-dependent protein kinase from bovine cardiac muscle. Proc. Natl. Acad. Sci. USA 79, 6128-6131. Casey, P. J., Solski, P. A., Der, C. J., & Buss, J. E. (1989). p21ras is modified by a famesyl isoprenoid. Proc. Natl. Acad. Sci. USA 86, 8323-8327. Chavrier, P., Gorvel, J.-P., Stelzer, E., Simons, K., Gruenberg, J., & Zerial, M. (1991). Hypervariable C-terminal domain of rab proteins acts as a targeting signal. Nature 353, 769-772. Chelsky, D., Olson, J. E, & Koshland, J., DE (1987). Cell cycle-dependent methyl esterification of lamin B. J. Biol. Chem. 62, 4303-4309. Chen, W.-J., Moomaw, J. F., Overton, L., Kost, T. A., & Casey, P. J. (1993). High level expression ot mammalian protein farnesyltransferase in a baculovirus system. J. Biol. Chem. 268, 9675-9680. Chow, M., Newman, J. F. E., Filman, D., Hogle, J. M., Rowlands, D. J., & Brown, E (1987). Myristoylation of picornavirus capsid protein VP4 and its structural significance. Nature 327, 482-486. Clarke, S. (1992). Protein isoprenylation and methylation at carboxyl-terminal cysteine residues. Annu. Rev. Biochem. 61,355-386. Clarke, S., Vogel, J. P., Deschenes, R. J., & Stock, J. (1988). Posttranslational modification of the H-ras oncogene protein: evidence for a third class of protein carboxyl methyltransferases. Proc. Natl. Acad. Sci. USA 85, 4643-4647. Conklin, B. R., & Bourne, H. R. (1993). Structural elements of Goc subunits that interact with G ~ receptors and effectors. Cell 73, 631-641. Cox, A. D., Graham, S. M., Solski, P. A., Buss, J. E., & Der, C. J. (1993). The carboxyl-terminal CXXX sequence of Gict, but not Rab5 or Rabl 1, supports Ras processing and transforming activity. J Biol. Chem. 268, 11548-11552. Cremers, F. P. M., Molloy, C. M., van de Pol, J. R., van denHurk, J. A. J. M., Bach, I., Geurts van Kessel A., HM, & Ropers, H.-H. (1992). An autosomal homologue of the choroideremia gene colocalize~, with the usher syndrome type II locus on the distal part of chromosome 1q. Hum. Mol. Genet. 1 71-75. Crowell, P. L., Chang, R. R., Ren, Z., Elson, C. E., & Gould, M. N. (1991). Selective inhibition ol isoprenylation of 21-26-kDa proteins by the anticarcinogen d-limonene and its metabolites. J Biol. Chem. 266, 17679-17685. Degtyarev, M. Y., Spiegel, A. M., & Jones, T. L. Z. (1993). The G protein OCssubunit incorporate! [3H]palmitic acid and mutation of cysteine-3 prevents this modification. Biochemistry 32 8057-8061. Deichaite, I., Casson, L. P., King, H.-P., & Resh, M. D. (1988). In vitro synthesis of pp60 vsrc myristoylation in a cell-free system. Mol. Cell. Biol. 8, 4295-4301. Diaz, M., Sanchez, Y., Bennet, T., Sun, C. R., Godoy, C., Tamanoi, E, Duran, A., & Perez, P. (1993) The Schizosaccharomyces pombe cwg 2§ gene codes for the 13 subunit of a geranylgerany transferase type I required for 13-glucan synthesis. EMBO J. 12, 5245-5254. Dizhoor, A. M., Ericsson, L. H., Johnson, R. S., Kumar, S., Olshevskaya, E., Zozulya, S., Neubert, ~I A., Stryer, L., Hurley, J. B., & Walsh, K. A. (1992). The NHz terminus of retinal recoverin i~ acylated by a small family of fatty acids. J. Biol. Chem. 267, 16033-16036. Duronio, R. J., Rudnick, D. A., Adams, S. P., Towler, D. A., & Gordon, J. I. (1991). Analysing thq substrate specificity of Saccharomyces cerevisiae myristoyl-CoA: protein N-myristoyl trans

Lipid Modifications of Proteins

99

ferase by co-expressing it with mammalian G protein subunits in Escherichia coli. J. Biol. Chem. 266, 10498-10504. Duronio, R. J., Reed, S. I., & Gordon, J. I. (1992). N-myristoyltransferase cause temperature-sensitive myristic acid auxotrophy in Saccharomyces cerevisiae. Proc. Natl. Acad. Sci. USA 89, 41294133. Endo, A., Kuroda, M., & Tanzawa, K. (1976). Competitive inhibition of 3-hydroxy-3-methylglutaryl coenzyme A reductase by ML-236A and ML-236B fungal metabolites having hypocholesterolemic activity. FEBS Leas. 72, 323-326. Ericsson, J., Runquist, M., Thelin, A., Andersson, M., Chojnacki, T., & Dallner, G. (1992). Distribution of prenyltransferases in rat tissues. J. Biol. Chem. 268, 832-838. Evans-Trueblood, C. E., Ohya, Y., & Rine, J. (1993). Genetic evidence for in vivo cross-specificity of the CaaX-box protein prenyltransferases farnesyltransferase and geranylgeranyltransferase-I in Saccharomyces cerevisiae. Mol. Cell. Biol. 13, 4260-4275. Famsworth, C. C., Kawata, M., Yoshida, Y., Takai, Y., Gelb, M. H., & Glomset, J. A. (1991). C-terminus of the small GTP-binding protein smgp25A contains two geranylgeranylated cysteine residues and a methyl ester. Proc. Natl. Acad. Sci. USA 88, 6196-6200. Fenton, R. G., Kung, H.-E, Longo, D. L., & Smith, M. R. (1992). Regulation of intracellular actin polymerization by prenylated cellular proteins. J. Cell Biol. 117, 347-356. Gallego, C., Gupta, S. K., Winitz, S., Eisfelder, B. J., & Johnson, G. L. (1992). Myristoylation of G~i2 polypeptide, a G protein ~ subunit, is required for its signalling and transformational functions. Proc. Natl. Acad. Sci. USA 89, 9695-9699. Giannakouros, T., & Magee, A. I. (1993). Protein prenylation and associated modifications. In: Lipid Modifications of Proteins. (Schlesinger, M. J., ed.), pp. 135-162, CRC Press Inc. Boca Raton, FL. Giannakouros, T., Newman, C. M. H., Armstrong, J., & Magee, A. I. (1993a). Processing of the small GTP-binding protein spYPTlp in Schizosaccharomyces pombe and in mammalian COS cells. Biochem. Biophys. Res. Comm. 192, 983-990. Giannakouros, T., Newman, C. M. H., Craighead, M. W., Armstrong, J., & Magee, A. I. (1993b). Post-translational processing of S. pombe YPT5 protein: in vitro and in vivo analysis of processing mutants. J. Biol. Chem. 268, 24467-24474. Gibbs, J. B., Pompliano, D. L., Mosser, S. D., Rands, E., Lingham, R. B., Singh, S. B., Scolnick, E. M., Kohl, N. E., & Oliff, A. (1993). Selective inhibition of farnesyl-protein transferase blocks ras processing in vivo. J. Biol. Chem. 268, 7617-7620. Glomset, J. A., Gelb, M. H., & Famsworth, C. C. (1990). Prenyl proteins in eukaryotic cells: a new type of membrane anchor. TIBS 15, 139-142. Goldstein, J. L., & Brown, M. S. (1990). Regulation of the mevalonate pathway. Nature 343,425-430. Gordon, J. I., Duronio, R. J., Rudnick, D. A., Adams, S. P., & Gokel, G. W. (1991). Protein N-myristoylation. J. Biol. Chem. 266, 8647-8650. Gutierrez, L., & Magee, A. I. (1991). Characterization of an acyl transferase acting on p21Nras in a cell-free system. Biochim. Biophys. Acta. 1078, 147-154. Hancock, J., & Hall, A. (1993). A novel role for RhoGDI as an inhibitor of GAP proteins. EMBO J. 12, 1915-1921. Hancock, J. E, Magee, A. I., Childs, J. E., & Marshall, C. J. (1989). All ras proteins are polyisoprenylated but only some are palmitoylated. Cell 57, 1167-1177. Hancock, J. E, Paterson, H., & Marshall, C. J. (1990). A polybasic domain or palmitoylation is required in addition to the CAAX motif to localize p21 ras to the plasma membrane. Cell 63, 133-139. Hancock, J. E, Cadwallader, K., & Marshall, C. J. (1991a). Methylation and proteolysis are essential for efficient membrane binding of prenylated p21K-ras(B).EMBO J. 10, 641-646. Hancock, J. E, Cadwallader, K., Paterson, H., & Marshall, C. J. (1991b). A CAAX or a CAAL motif and a second signal are sufficient for plasma membrane targeting of ras proteins. EMBO J. 10, 4033-4039.

100

MARCO PARENT! and ANTHONY I. MAGEE

Hara, M., Akasaka, K., Akinaga, S., Okabe, M., Nakano, H., Gomez, R., Wood, D., Uh, M., & Tamanoi, F. (1993). Identification of ras famesyltransferase inhibitors by microbial screening. Proc. Natl. Acad. Sci. USA 90, 2281-2285. Hata, Y., Kaibuchi, K., Kawamura, S., Hiroyoshi, M., Shirataki, H., & Takai, Y. (1991). Enhancement of the actions of smgp21 GDP/GTP exchange protein by the protein kinase A-catalyzed phosphorylation of smgp21. J. Biol. Chem. 266, 6571-6577. Heuckeroth, R. O., Glaser, L., & Gordon, J. I. (1988). Heteroatom-substituted fatty acid analogs as substrates for N-myristoyltransferase: an approach for studying both the enzymology and function of protein acylation. Proc. Natl. Acad. Sci. USA 85, 8795-8799. Horiuchi, H., Kaibuchi, K., Kawamura, M., Matsuura, Y., Suzuki, N., Kuroda, Y., Kataoka, T., & Takai, Y. (1992). The posttranslational processing of rasp21 is critical for its stimulation of yeast adenylate cyclase. Mol. Cell. Biol. 12, 4515--4520. Hruby, D. E., & Franke, C. A. (1993). Viral acylproteins: greasing the wheels of assembly. Trends in Microbiology 1, 20-24. Hrycyna, C. A., & Clarke, S. (1990). Famesyl cysteine C-terminal methyltransferase activity is dependent upon the STE14 gene product in Saccharomycespombe. Mol. Cell. Biol. 10, 50715076. Hrycyna, C. A., & Clarke, S. (1992). Maturation of isoprenylated proteins in Saccharomycescerevisiae. J. Biol. Chem. 267, 10457-10464. Hrycyna, C. A., Sapperstein, S. K., Clarke, S., & Michaelis, S. (1991). The Saccharomyces cerevisiae STE14 gene encodes a methyltransferase that mediates C-terminal methylation of a-factor and RAS proteins. EMBO J. 10, 1699-1709. Huang, E. M. (1989). Agonist-enhanced palmitoylation of platelet proteins. Biochim. Biophys. Acta 1011,134-139. Huzoor-Akbar, Wang, W., Kornhauser, R., Volker, C., & Stock, J. B. (1993). Protein prenylcysteine analog inhibits agonist-receptor-mediated signal transduction in human platelets. Proc. Natl. Acad. Sci. USA 90, 868-872. Inglese, J., Koch, W. J., Caron, M. G., & Lefkowitz, R. J. (1992). Isoprenylation in regulation of signal transduction by G-protein-coupled receptor kinases. Nature 359, 147-150. Itoh, T., Kaibuchi, K., Tadayuki, M., Yamamoto, T., Matsuura, Y., Maeda, A., Shimizu, K., & Takai, Y. (1993). The post-translational processing of rasp21 is critical for its stimulation of mitogen-activated protein kinase. J. Biol. Chem. 268, 3025-3028. James, G., & Olson, E. N. (1989). Identification of a novel fatty acylated protein that partitions between the plasma membrane and cytosol and is deacylated in response to serum and growth factor stimulation. J. Biol. Chem. 264, 20988-21006. James, G. L., Goldstein, J. L., Brown, M. S., Rawson, T. E., Somers, T. C., McDowell, R. S., Crowley, C. W., Lucas, B. K., Levinson, A. D., & Marsters, J. C., Jr. (1993). Benzodiazepine peptidomimetics: potent inhibitors of ras farnesylation in animal cells. Science, 1938-1941. Jochen, A., Hays, J., Lianos, E., & Hager, S. (1991). Insulin stimulates fatty acid acylation of adipocyte proteins. Biochem. Biophys. Res. Comm. 177, 797-801. Johnson, D. R., Duronio, R. J., Langner, C. A., Rudnick, D. A., & Gordon, J. I. (1993). Genetic and biochemical studies of a mutant Saccharomycescerevisiaemyristoyl-CoA:protein N-myristoyltransferase, nmt72pLeu9%Pro, that produces temperature-sensitive myristic acid auxotrophy. J. Biol. Chem. 268, 483-494. Jones, T. L. Z., & Spiegel, A. M. (1990). Isoprenylation of an inhibitory G protein ct subunit occurs only upon mutagenesis of the carboxyl terminus. J. Biol. Chem. 265, 19389-19392. Jones, T. L. Z., Simonds, W. E, Merendino, J., Jr., Brann, M. R., & Spiegel, A. M. (1990). Myristoylation of an inhibitory GTP-binding protein o~subunits is essential for its membrane attachment. Proc. Natl. Acad. Sci. USA 87, 568-572.

Lipid Modifications of Proteins

101

Kamiya, Y., Sakurai, A., Tamura, S., & Takahashi, N. (1979). Structure of rhodotorucine A, a peptidyl factor, inducing mating tube formation in Rhodosporidium toruloides. Agric. Biol. Chem. 43, 363-369. Kamps, M. P., Buss, J. E., & Sefton, B. M. (1986). Rous sarcoma virus transforming protein lacking myristic acid phosphorylates known polypeptide substrates without inducing transformation. Cell 45, 105-112. Kasinathan, C., Grzelinska, E., Okazaki, K., Slomiany, B. L., & Slomiany, A. (1990). Purification of protein fatty acyltransferase and determination of its distribution and topology. J. Biol. Chem. 265, 5139-5144. Kawamura, M., Kaibuchi, K., Kishi, K., & Takai, Y. (1993). Translocation of Ki-rasp21 between membrane and cytoplasm by smg GDS. Biochem. Biophys. Res. Comm. 190, 832-841. Kinsella, B. T., & Maltese, W. A. (1992). rab GTP-binding proteins with three different carboxyl-terminal cysteine motifs are modified in vivo by 20-carbon isoprenoids. J. Biol. Chem. 267, 3940-3945. Kinsella, B. T., Erdman, R. A., & Maltese, W. A. (1991 ). Carboxyl-terminal isoprenylation of ras-related GTP-binding proteins encoded by rac 1, rac2 and ralA. J. Biol. Chem. 266, 9786. Knoll, L. J., Aach-Levy, M., Stahl, P. D., & Gordon, I. J. (1992). Analysis of the compartmentalization of myristoyl-CoA: protein N-myristoyltransferase in SaccluTromyces cerevisiae. J. Biol. Chem. 267, 5366-5373. Koegl, M., Zlatkine, P., Ley, S. C., Courtneidge, S. A., and Magee, A. I. (1994). Palmitoylation of multiple src-family kinases at a homologous N-terminal motif. Biochem. J. 303, 749-753. Kohl, N. E., Mosser, S. D., deSolms, S. J., Giuliani, E. A., Pompliano, D. L., Graham, S. L., Smith, R. L., Scolnick, E. M., Oliff, A., & Gibbs, J. B. (1993). Selective inhibition of ras-dependent transformation by a farnesyltransferase inhibitor. Science 260, 1934-1937. Kuroda, Y., Suzuki, N., & Kataoka, T. (1992). The effect of posttranslational modifications on the interaction of Ras2 with adenylyl cyclase. Science 259, 683-686. Li, G., & Stahl, P. D. (1993). Post-translational processing and membrane association of the two early endosome-associated rab GTP-binding proteins (rab4 and rab5). Arch. of Biochem. Biophys. 304, 471-478. Linder, M. E., Pang, I.-H., Duronio, R. J., Gordon, J. I., Sternweis, P. C., & Gilman, A. G. (1991). Lipid modifications of G protein subunits. J. Biol. Chem. 266, 4654-4659. Linder, M. E., Middleton, P., Hepler, J. R., Taussig, R., Gilman, A. G., & Mumby, S. M. (1993). Lipid modifications of G proteins: tx subunits are palmitoylated. Proc. Natl. Acad. Sci. USA 90, 3675-3679. Lorenz, W., Inglese, J., Palczewski, K., Ohorato, J. J., Caron, M. G., & Lefkowitz, R. J. (1991). The receptor kinase family: primary structure of rhodopsin kinase reveals similarities to the ~-adrenergic receptor kinase. Proc. Natl. Acad. Sci. USA 88, 8715-8719. Ma, Y.-T., & Rando, R. R. (1992). A microsomal endoprotease that specifically cleaves isoprenylated peptides. Proc. Natl. Acad. Sci. USA 89, 6275-6279. Ma, Y.-T., Chaudhuri, A., & Rando, R. R. (1992). Substrate specificity of the isoprenylated protein endoprotease. Biochemistry 31, 11772-11777. Ma, Y.-T., Gilbert, B. A., & Rando, R. R. (1993). Inhibitors of the isoprenylated protein endoprotease. Biochemistry 32, 2386-2393. Mack, D., Berger, M., Schmidt, M. E G., & Kurppa, J. (1987). Cell-free fatty acylation of microsomal integrated and detergent-solubilized glycoprotein of vesicular stomatitis virus. J. Biol. Chem. 262, 4297-4302. Magee, A. I., & Courtneidge, S. A. (1985). Two classes of fatty acid acylated proteins exist in eukaryotic cells. EMBO J. 4, 1137-1144. Magee, A. I., & Newman, C. (1992). The role of lipid anchors for small G-binding proteins in membrane trafficking. Trends in Cell Biology 11, 318-323.

102

MARCO PARENTI and ANTHONY I. MAGEE

Magee, A. I., Gutierrez, L., McKay, I. A., Marshall, C. J., & Hall, A. (1987). Dynamic fatty acylation of p21N-ras. EMBO J. 6, 3353-3357. Maltese, W. A. (1990). Posttranslational modification of proteins by isoprenoids in mammalian cells. FASEB J. 4, 3319-3328. Maltese, W. A., & Sheridan, K. M. (1987). Isoprenylated proteins in cultured cells! subcellular distribution and changes related to altered morphology and growth arrest induced by mevalonate deprivation. J. Cell. Physiol. 133, 471--481. Mayer, M. L., Caplin, B. E., & Marshall, M. S. (1992). CDC43 and RAM2 encode the polypeptide subunits of a yeast type I protein geranylgeranyltransferase. J. Biol. Chem. 267, 20589-20593. McConville, M. J., & Ferguson, M. A. J. (1993). The structure, biosynthesis and function of glycosylated phosphatidylinositols in the parasitic protozoa and higher eukaryotes. Biochem. J. 294, 305-324. Mcllhinney, R. A. J., Pelly, S. J., Chadwick, J. K., & Cowley, G. P. (1985). Studies on the attachment of myristic and palmitic acid to cell proteins in human squamous carcinoma cell lines: evidence for two pathways. EMBO J. 4, 1145-1152. Miura, Y., Kaibuchi, K., Itoh, T., Corbin, J. D., Francis, S. H., & Takai, Y. (1992). Phosphorylation of smg p21B/raplBp21 by cyclic GMP-dependent protein kinase. FEBS Leas. 297, 171-174. Miura, S., Hasumi, K., & Endo, A. (1993). Inhibition of protein prenylation by patulin. FEBS Letts. 318, 88-90. Moffett, S., Mouillac, B., Bonin, H., & Bouvier, M. (1993). Altered phosphorylation and desensitization patterns of a human 132-adrenergic receptor lacking the palmitoylated Cys 341. EMBO J. 12, 349-356. Molenaar, C. M. T., Prange, R., & Gallwitz, D. (1988). A carboxyl-terminal cysteine residue is required for palmitic acid binding and biological activity of the ras-related yeast YPT1 protein. EMBO J. 7, 971-976. Moomaw, J. E, & Casey, P. J. (1992). Mammalian protein geranylgeranyltransferase. J. Biol. Chem. 267, 17438-17443. Moores, S. L., Schaber, D., Mosser, S. D., Rands, E., O'Hara, M. B., Garsky, V. M., Marshall, M. S., Pompliano, D. L., & Gibbs, J. B. (1991). Sequence dependence of protein isoprenylation. J. Biol. Chem. 266, 14603-14610. Mouillac, B., Caron, M., Bonin, H., Dennis, M., & Bouvier, M. (1992). Agonist-modulated palmitoylation of 132-adrenergic receptor in Sf9 cells. J. Biol. Chem. 267, 21733-21737. Mumby, S. M., Heukeroth, R. O., Gordon, J. I., & Gilman, A. G. (1990). G-protein oc-subunit expression, myristoylation, and membrane association in COS cells. Proc. Natl. Acad. Sci. USA 87, 728-732. Muszbek, L., & Laposata, M. (1993). Myristoylation of proteins in platelets occurs predominantly through thioester linkages. J. Biol. Chem. 268, 8251-8255. Neubert, T. A., Johnson, R. S., Hurley, J. B., & Walsh, K. A. (1992). The rod transducin oc subunit amino terminus is heterogeneously fatty acylated. J. Biol. Chem. 267, 18247-18277. Newman, C. M. H., & Magee, A. I. (1993). Post-translational processing of the ras superfamily of small GTP-binding proteins. BBA Reviews on Cancer 1155, 79-96. Newman, C. M. H., Giannakouros, T., Hancock, J. E, Fawell, E. H., Armstrong, J., & Magee, A. I. (1992). Post-translational processing of Schizosaccharomyces pombe YPT proteins. J. Biol. Chem. 267, 11329-11336. O'Dowd, B. E, Hnatowich, M., Caron, M. G., Lefkowitz, R. J., & Bouvier, M. (1989). Palmitoylation of the human ~2-adrenergic receptor. J. Biol. Chem. 264, 7564-7569. Olson, E. N., & Spizz, G. (1986). Fatty acylation of cellular proteins. Temporal and subcellular differences between palmitate and myristate acylation. J. Biol. Chem. 261, 2458-2466. Omary, M. B., & Trowbridge, I. S. (1981). Covalent binding of fatty acid to the transfen'in receptor in cultured human cells. J. Biol. Chem. 256, 4715-4718. Omer, C. A., Kral, A. M., Diehl, R. E., Prendergast, G. C., Poers, S., Allen, C. M., Gibbs, J. B., & Kohl, N. E. (1993). Characterization of recombinant human famesyl-protein transferase: cloning,

Lipid Modifications of Proteins

103

expression, farnesyl diphosphate binding, and functional homology with yeast prenyl-protein transferases. Biochemistry 32, 5167-5176. Ovchinnikov, Y. A., Abdulaev, N. G., & Bogachuk, A. S. (1988). Two adjacent cysteine residues in the C-terminal cytoplasmic fragment of bovine rhodopsin are palmitoylated. FEBS Letts. 230, 1-5. Palade, G. (1975). Intracellular aspects of the process of protein synthesis. Science 189, 347-358. Papac, D. I., Thornburg, K. R., Bullesbach, E. E., Crouch, K. R., & Knapp, D. R. (1992). Palmitoylation of a G-protein coupled receptor. J. Biol. Chem. 267, 16889-16894. Parenti, M., Vigan6, M. A., Newman, C. M. H., Milligan, G., & Magee, A. I. (1993). A novel N-terminal motif for palmitoylation of G-protein a subunits. Biochem. J. 291,349-353. Prrez-Sala, D., Tan, E. W., Cafiada, E J., & Rando, R. R. (1991). Methylation and demethylation reactions of guanine nucleotide-binding proteins of retinal rod outer segments. Proc. Natl. Acad. Sci. USA 88, 3043-3046. Prrez-Sala, D., Gilbert, B. A., Tan, E. W., & Rando, R. R. (1992). Prenylated protein methyltransferases do not distinguish between farnesylated and geranylgeranylated substrated. Biochem. J. 284, 835-840. Pfeffer, S. R. (1992). GTP-binding proteins in intracellular transport. Trends in Cell Biology 2, 41-46. Regazzi, R., Kikuchi, A., Takai, Y., & Wollheim, C. B. (1992). The small GTP-binding proteins in the cytosol of insulin secreting cells are complexed to GDP dissociation inhibitor proteins. J. Biol. Chem. 267, 17512-17519. Reiss, Y., Seabra, M. C., Armstrong, S. A., Slaughter, C. A., Goldstein, J. L., & Brown, M. S. (1991). Nonidentical subunits of p21H-ras farnesyltransferase. Peptide binding and famesyl pyrophosphate carder functions. J. Biol. Chem. 266, 10672-10677. Resh, M. D. (1989). Specific and saturable binding of pp60 vsrc to plasma membrane: evidence for a myristyl-src receptor. Cell 58, 281-286. Resh, M. D., & Ling, H.-P. (1990). Identification of a 32K plasma membrane protein that binds to the myristylated amino-terminal sequence of p6Or-src. Nature 346, 84-86. Rizzolo, L. J., & Komfeld, R. (1988). Post-translational protein modification in the endoplasmic reticulum. J. Biol. Chem. 263, 9520-9525. Rizzolo, L. J., Finidori, J., Gonzalez, A., Arpin, M., Ivanov, I. E., Adesnik, M., & Sabatini, D. D. (1985). Biosynthesis and intracellular sorting of growth hormone-viral envelope glycoprotein hybrids. J. Cell Biol. 101, 1351-1362. Rudnick, D. A., McWherter, C. A., Adams, S. A., Ropson, I. J., Duronio, R. J., & Gordon, J. I. (1990). Structural and functional studies of S. cerevisiae myristoyl-CoA: protein N-myristoyltransferase produced in E. coli: evidence for an acyl-enzyme intermediate. J. Biol. Chem. 265, 13370-13378. Rudnick, D. A., McWherter, C. A., Rocque, W. J., Lennon, P. J., Getman, D. P., & Gordon, J. I. (1991). Kinetic and structural evidence for a sequential ordered bi bi mechanism of catalysis by S. cerevisiae myristoyl-CoA: protein N-myristoyltransferase. J. Biol. Chem. 266, 9732-9739. Rudnick, D. A., Johnson, R. L., & Gordon, J. I. (1992). Studies of the catalytic activities and substrate specificities of Saccharomyces cerevisiae myristoyl-coenzyme A: protein N-myristoyltransferase deletion mutants and human/yeast Nmt chimeras in Escherichia coli and S. cerevisiae. J. Biol. Chem. 267, 23852-23861. Runquist, M., Ericsson, J., Thelin, A., Chojnacki, T., & Dallner, G. (1992). Biosynthesis of trans,trans,trans-geranylgeranyl diphosphate by the cytosolic fraction from rat tissues. Biochem. Biophys. Res. Comm. 186, 157-165. Sakagami, Y., Isogai, A., Suzuki, A., Tamura, S., Kitada, C., & Fujino, M. (1979). Structure of tremerogen A-10, a peptidal hormone inducing conjugation tube formation in Tremella mesenterica. Agric. Biol. Chem. 43, 2643-2645. Scheer, A., & Gierschik, P. (1993). Farnesylcysteine analogues inhibit chemotactic peptide receptormediated G-protein activation in human HL-60 granulocyte membranes. FEBS Letts. 319, 110-114. Schmidt, M. E G. (1989). Fatty acylation of proteins. Biochim. Biophys. Acta. 988, 411-426.

104

MARCO PARENTI and ANTHONY !. MAGEE

Schmidt, M. E G., Bracha, M., & Schlesinger, M. J. (1979). Evidence for covalent attachment of fatty acids to Sindbis virus glycoproteins. Proc. Natl. Acad. Sci. USA 76, 1687-1691. Schmidt, R. A., Schneider, C. J., & Glomset, J. A. (1984). Evidence for post-translational incorporation of a product of mevalonic acid into Swiss 3T3 cell proteins. J. Biol. Chem. 259, 10175-10180. Schultz, A. M., Henderson, L. E., Oroszlan, S., Garber, E. A., & Hanafusa, H. (1985). Amino terminal myristoylation of the protein kinase p60 src, a retroviral transforming protein. Science 227, 427-429. Schultz, A. M., Tsai, S.-C., Kung, H.-E, Oroszlan, S., Moss, J., & Vaughan, M. (1987). Hydroxylaminestable covalent linkage of myristic acid in Goa, a guanine nucleotide-binding protein of bovine brain. Biochem. Biophys. Res. Comm. 146, 1234-1239. Schulz, S., & Nyce, J. W. ( 1991). Inhibition of protein isoprenylation and p21 ras membrane association by dehydroepiandrosterone in human colonic adenocarcinoma cells. Cancer Res. 51, 6563-6567. Seabra, M. C., Reiss, Y., Casey, P. J., Brown, M. S., & Goldstein, J. L. (1991). Protein farnesyltransferase and geranylgeranyltransferase share a common tx subunit. Cell 65, 429-434. Seabra, M. C., Brown, M. S., & Goldstein, J. L. (1992a). Retinal degeneration in choroideremia: deficiency of Rab geranylgeranyl transferase. Science 259, 377-381. Seabra, M. C., Goldstein, J. L., Stidhof, T. C., & Brown, M. S. (1992b). Rab geranylgeranyl transferase. J. Biol. Chem. 267, 14497-14503. Sefton, B. M., Trowbridge, I. S., Cooper, J. A., & Scolnick, E. M. (1982). The transforming proteins of Rous sarcoma virus, Harvey sarcoma virus and Abelson virus contain tightly bound lipids. Cell 31,465-474. Sinensky, M., & Lutz, R. J. (1992). The prenylation of proteins. BioEssays 14, 25-31. Smith, R. H., & Powell, G. L. (1986). The critical micelle concentration of some physiologically important fatty acyl-coenzyme As as a function of chain length. Arch. Biochem. Biophys. 244, 357-360. Sobotka-Briner, C., & Chelsky, D. (1992). COOH-terminal methylation of lamin B and inhibition of methylation by farnesylated peptides corresponding to lamin B and other CAAX motif proteins. J. Biol. Chem. 267, 12116--12122. Staufenbiel, M. (1987). Ankyrin-bound fatty acid turns over rapidly at the erythrocyte plasma membrane. Mol. Cell. Biol. 7,2981-2984. Staufenbiel, M. (1988). Fatty acids covalently bound to erythrocyte proteins undergo a differential turnover in vitro. J. Biol. Chem. 263, 13615-13622. Steele-Mortimer, O., Gruenberg, J., & Glague, M. J. (1993). Phosphorylation of GDI and membrane cycling of rab proteins. FEBS Letts. 329, 313-318. S tephenson, R. C., & Clarke, S. (1992). Characterization of a rat liver protein carboxyl methyltransferase involved in the maturation of proteins with the -CXXX C-terminal sequence motif. J. Biol. Chem. 267, 13314-13319. Swiezewska, E., Thelin, A., Dallner, G., Andersson, B., & Ernster, L. (1993). Occurrence of prenylated proteins in plant cells. Biochem. Biophys. Res. Comm. 192, 161-166. Takahashi, N., & Breitman, T. R. (1989). Retinoic acid acylation (retinoylation) of a nuclear protein in the human acute myeloid leukemia cell line HL60. J. Biol. Chem. 264, 5159-5163. Takahashi, N., & Breitman, T. R. (1992). Covalent modification of proteins by ligands of steroid hormone receptors. Proc. Natl. Acad. Sci. USA 89, 10807-10811. Takahashi, N., Jetten, A. M., & Breitman, T. R. (1991). Retinoylation of cytokeratins in normal human epidermal keratinocytes. Biochem. Biophys. Res. Comm. 180, 393--400. Takai, Y., Kaibuchi, K., Kikuchi, A., & Kawata, M. (1992). Small GTP-binding proteins. Int. Rev. Cytol. 133, 187-230. Tan, E. W., Ptrez-Sala, D., Cafiada, F. J., & Rando, R. R. (1991). Identifying the recognition unit for G protein methylation. J. Biol. Chem. 266, 10719-10722. Thissen, J. A., & Casey, P. J. (1993). Microsomal membranes contain a high affinity binding site for prenylated peptides. J. Biol. Chem. 268, 13780-13783.

Lipid Modifications of Proteins

105

Towler, D. A., Adams, S. P., Eubanks, S. R., Towery, D. S., Jackson-Machelski, E., Glaser, L., & Gordon, I. J. (1987a). Purification and characterization of yeast myristoyl CoA: N-myristoyltransferase. Proc. Natl. Acad. Sci. USA 84, 2708-2712. Towler, D. A,, Eubanks, S. R., Towery, D. S., Adams, S. P., & Glaser, L. (1987b). Amino-terminal processing of proteins by N-myristoylation. J. Biol. Chem. 262, 1030-1036. Towler, D. A., Gordon, J. I., Adams, S. P., & Glaser, L. (1988). The biology and enzymology of eukaryotic protein acylation. Annu. Rev. Biochem. 57, 69-99. van der Sluijs, P., Hull, M., Huber, L. A., Mille, P., Goud, B., & Mellman, I. (1992). Reversible phosphorylation-dephosphorylation determines the localization of rab4 during the cell cycle. EMBO J. 11, 4379-4389. Volker, C., Lane, P., Kwee, C., Johnson, M., & Stock, J. (1991). A single activity carboxyl methylates both farnesyl and geranylgeranyl cysteine residues. FEBS Letts. 295, 189-195. Walworth, N. C., Goud, B., Kabcenell, A. K., & Novick, P. J. (1989). Mutational analysis of SEC4 suggests a cyclical mechanism for the regulation of vesicular traffic. EMBO J. 8, 1685-1693. Wilson, A. L., & Maltese, W. A. (1993). Isoprenylation ofRabl B is impaired by mutations in its effector domain. J. Biol. Chem. 268, 14561-14564. Wolda, S. L., & Glomset, J. A. (1988). Evidence for modification of lamin B by a product of mevalonic acid. J. Biol. Chem. 263, 5977-6000. Yokoyama, K., & Gelb, M. H. (1993). Purification of a mammalian protein geranylgeranyltransferase. J. Biol. Chem. 268, 4055-4060.

This Page Intentionally Left Blank

THE BIOSYNTHESIS OF MEMBRANE PROTEINS

David Stephens, Sunita Kulkarni, and Brian Austen

I. II.

III.

IV.

V.

Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A. Mechanisms of Protein Synthesis . . . . . . . . . . . . . . . . . . . . . . B. Insertion into the Phospholipid B ilayer . . . . . . . . . . . . . . . . . . . C. Targeting of Newly-Synthesized Membrane Proteins . . . . . . . . . . . D. Use of in Vitro Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . Targeting to the ER Membrane . . . . . . . . . . . . . . . . . . . . . . . . . . A. Signal Sequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B. Signal Recognition Particle . . . . . . . . . . . . . . . . . . . . . . . . . C. Heat-Shock Chaperones . . . . . . . . . . . . . . . . . . . . . . . . . . . D. The Translocation Complex . . . . . . . . . . . . . . . . . . . . . . . . . Establishing Topography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A. Assembly of Single-Pass Membrane Proteins . . . . . . . . . . . . . . . B. Assembly of Multi-Spanning Proteins . . . . . . . . . . . . . . . . . . . C. Membrane Proteins with C-terminal Anchors . . . . . . . . . . . . . . . Processing in the ER L u m e n . . . . . . . . . . . . . . . . . . . . . . . . . . . A. Folding and Disulfide Pairing . . . . . . . . . . . . . . . . . . . . . . . . B. Formation of Oligomers . . . . . . . . . . . . . . . . . . . . . . . . . . .

Biomembranes Volume 1, pages 107-135. Copyright 9 1995 by JAI Press Inc. All rights of reproduction in any form reserved. ISBN: 1-55938-658-4

107

108 108 108 109 109 111 113 113 113 114 114 115 115 117 118 118 118 119

10~

DAVID STEPHENS,SUNITA KULKARNI, and BRIAN AUSTEN

VI. VII.

119 120 120 121 122 122 122 124 124 124 125 125 126 126 127 127 128 128 130 130 130 132 132

Retention of Membrane Proteins in the ER . . . . . . . . . . . . . . . . . . . Transport of Newly Synthesized Membrane Proteins in Vesicles . . . . . . . A. Budding Reactions in the Golgi Complex . . . . . . . . . . . . . . . . . B. Fusion of Vesicles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . C. Vesicle Movement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . VIII. Glycosylation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . IX. Retention in the Golgi Complex . . . . . . . . . . . . . . . . . . . . . . . . . X. Constitutive and Regulated Movement to the Cell Surface . . . . . . . . . . . A. Receptor-Mediated Sorting . . . . . . . . . . . . . . . . . . . . . . . . . B. Passive-Sorting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . C. Recycling of Membrane Components . . . . . . . . . . . . . . . . . . . XI. Movement to the Cell Surface via the Endosome . . . . . . . . . . . . . . . . XII. Lysosomal Membrane Proteins . . . . . . . . . . . . . . . . . . . . . . . . . XIII. Lipid Modification of Membrane Proteins . . . . . . . . . . . . . . . . . . . XIV. Biosynthesis of Mitochondrial Membrane Proteins . . . . . . . . . . . . . . . A. Import into Mitochondria . . . . . . . . . . . . . . . . . . . . . . . . . . B. Transit Sequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . C. Components Involved in Mitochondrial Import . . . . . . . . . . . . . . D. Insertion Directly into the Inner and Outer Membranes . . . . . . . . . . XIV. Biosynthesis of Bacterial Membrane Proteins . . . . . . . . . . . . . . . . . A. Outer Membrane Proteins . . . . . . . . . . . . . . . . . . . . . . . . . B. Membrane Proteins in the Cytoplasmic Membrane . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

i. ABSTRACT Membrane proteins are synthesized on the ribosomal machinery of cells and then inserted into membranes. In eukaryotic cells, proteins are either first inserted co-translationally into the membrane of the endoplasmic reticulum, or post-translationally into membranes of mitochondria, the nucleus, or peroxisomes. From the endoplasmic reticulum, membrane proteins may travel in vesicles to the Golgi complex, the lysosome, or the cell surface, and on the way may be subjected to post-translational modification. In bacteria, proteins assemble either into the cytoplasmic or outer membrane. The final location and topography depends on targeting sequences, hydrophobic domains, or covalently-linked lipid.

!1. I N T R O D U C T I O N A. Mechanisms of Protein Synthesis A l t h o u g h a small n u m b e r o f proteins are synthesized in mitochondria, m o s t m e m b r a n e proteins are synthesized by the same cytoplasmic ribosomal m a c h i n e r y as other proteins, using m R N A originating in the nucleus. Briefly, proteins are synthesized from the N- to the C-terminus on ribosomes which m o v e along the

The Biosynthesis of Membrane Proteins

109

mRNA from the 5'- to the 3'-end. In eukaryotes, mRNA translation starts some way from the 5'-end cap structure at a triplet codon sequence (AUG) which sets the reading frame, and continues until a stop codon (e.g., UAA) is reached. Initiation requires that Met-tRNA binds to the 40S subunit of the ribosome together with initiation factors and GTP. This initiation complex then binds the mRNA at the cap structure, and scans along until the first AUG is reached. The 60S subunit then binds, and elongation of the polypeptide occurs in an ordered manner according to the triplet codon sequence in the mRNA, which dictates (by base-pairing to the anticodon) the amino acyl-tRNA that will bind in each position.

B. Insertion into the Phospholipid Bilayer Many proteins are integrated into membranes via interactions between their transmembrane sequences, composed primarily of about twenty consecutive hydrophobic amino acids, and the phospholipid bilayer. Integrated membrane proteins cross the bilayer once, twice, or cross over and back many times, leaving substantial parts of their structure extending into the aqueous environment on either or both sides. In the course of biosynthesis, membrane proteins fold into a characteristic three-dimensional structure. A major question in biosynthesis concerns how these proteins move from the ribosomal machinery in the aqueous environment of the cytoplasm into and across the hydrophobic barrier of the phospholipid bilayer of intracellular membranes, adopting as they move their folded and sometimes oligomeric structures. In eukaryotic cells, membrane proteins that assemble into the plasma membrane are first assembled into the endoplasmic reticulum (ER) membrane. The initial stages of the membrane assembly process is closely aligned to that of protein secretion. For example, mammalian immunoglobulin M heavy chain exists in a cell-bound and a secreted form, both of which are made with the cleavable signal sequence that brings about the initial insertion into the membrane, but only the former contains a C-terminal hydrophobic domain that ends up integrated across the membrane. Proteins that are targeted to the ER membrane (Route 1 in Figure 1) are differentiated early on from those that go to mitochondrial, nuclear, or peroxisomal membranes (Roman numerals by arrows in Figure 1).

C. Targeting of Newly-Synthesized Membrane Proteins Membrane proteins that assemble in the ER may go to a variety of destinations. Some proteins may stay in the ER, while others move through the ER to the Golgi complex (route 2 in Figure 1). Some membrane proteins are carded from the trans-Golgi network (TGN) to become integrated into the cell surface (routes 5 and 6 in Figure 1), while others are diverted to lysosomes (routes 3 and 4 in Figure 1). In polarized cells, membrane proteins are targeted to specific parts of the cell

110

DAVID STEPHENS, SUNITA KULKARNi, and BRIAN AUSTEN Coated pit

Apical plasmamembrane ..~

4-

j

Regulatedpathway

/

Constitutivepathway

''

Earlyendo

~~ 5ec:retorygranule Lysosome ~

b

~

~

Lateendosome ~

4-~ !1

Peroxisome ~

J~"~,

o

I r"~" ~l,

Roughendoplasmic"-----41 ~ ~ reticulum / ~ ~ ~ ~

trans-Golgi network

~ 3 o

i

cis Golgi

~-Mitochondrion

.

Basolateralplasma membrane

Figure 1. Protein targeting pathways. The p~thways that newly-synthesized membrane proteins take from membrane-bound ribosomes at the rough ER to the TGN (routes 1 and 2) and from the TGN to the apical cell surface via regulated (route 5) or bulk flow (route 6) secretory granules, or via the endosome (route 3)'to the lysosome (route 4) or the cell surface, or to the basolateral surface (route 9) are shown. Membrane protein import from free ribosomes in the cytoplasm is into mitochondria (route !), the nucleus (11),or the peroxisome (111).Membranes coated with clathrin prior to budding are indicated (+); membrane proteins are also internalized from the cell surface by endocytosis (route 7), from where they may be moved to other parts of the cell surface by transcytosis (route 8). The endomembrane system is stippled.

surface, e.g., the basolateral surface (route 9 in Figure 1). Proteins that are found at the basolateral surface are not found in the apical membrane, and vice versa. In bacteria, proteins are targeted to and inserted into either the cytoplasmic or the outer membrane. This sorting is achieved by specific targeting sequences, short sequences found in the newly synthesized protein that bring about localization into the membrane of a particular subcellular compartment. Examples of these are listed in Table 1, and described more fully under the appropriate subheading. Some proteins are linked to membranes via fatty-acids, lipids or isoprenoids, which are important for targeting. A consideration of assembly must, therefore, also include a question of how these anchors become attached, and how they are inserted into phospholipid bilayers.

The Biosynthesis of Membrane Proteins Table 1.

111

Examplesof TargetingSequences

Target Destination

Amino Acid Sequence +a

O

+oo

o

O

O+

Translocation into ER lumen

MKWVTFLLLLFISGSAFS^RGVF

ER stop-transfer sequence

-KS SIASFFFIIGLIIGLFLVLRVGIH-

ER retention sequence (membrane protein)

-RRSFIDEKKMP

++

+

+

- -++

++

+++

Golgi retention sequence (membrane protein) -KKWVLIVAFLLFVLIFLSFKKK+ +O 0+O MKANAKTIIAGMIALAISHTAMA

Translocation into bacterial periplasm +

o

+

++o

ooo

---- + ^DDIK-

+

Mitochondrial targeting

MLSALARPVGAALRRSFSTSAQNN^AKVAVL-

IMS targeting

-KLTQKLVTAGVAAAGITASTLLYAD-

Endocytosis

-NPVY-

Lysosomaltargeting (membrane)

-GY-

+

0

+

O

O

OO

a+ = positively charged o = neutral - = negatively charged

D. Use of in Vitro Systems Targeting, assembly and processing of newly-synthesized proteins into membranes derived from the ER (microsomes), mitochondria, peroxisomes, etc., have been studied in in vitro translation systems, mRNA is generated most conveniently by inserting the gene coding for the protein into a vector downstream of a promoter for SP6 or T7 polymerase. Transcription in vitro yields the mRNA, which after purification, is added to a lysate of reticulocytes, wheat germ, or yeast together with ATP, GTP, and creatine phosphate to supply energy, a radioactive amino acid, plus membranes derived from the rough ER or mitochondria of mammalian or yeast cells. The radioactive protein generated is studied by SDS-polyacrylamide gel electrophoresis (SDS-PAGE). If the protein is integrated into the membrane and contains a cleavable targeting sequence, there will be an increase in mobility on SDS-PAGE as the cleavage site reaches the processing peptidase, the active site of which is confined to the specific membrane compartment. If there is extensive core glycosylation, an addition which occurs in vitro with microsomal membranes, an increase in size and decrease in mobility will be seen. An exogenous protease such as trypsin may be added after translation and assembly. The parts of the membrane protein structure that have

Figure 2. Translation and processing of TSH-receptor in vitro. Shown is SDS-PAGE (10 o~/o acrylamide) separation of the S35-labeled receptor (TSHR) translated from in vitro transcribed mRNA by a reticulocyte lysate with or without pancreatic microsomes (DPRM). In the presence of microsomes, the 86 kD initially translated form is partially converted into a 100 kD glycosylated form.

112

The Biosynthesis of Membrane Proteins

113

reached the lumen of the membrane vesicles where they are protected from proteolysis can be determined. In vitro translation and assembly of the thyrotropin receptor, a seven-transmembrane receptor with a large extracellular domain, from mRNA generated by transcription of an SP6 vector, is shown in Figure 2. In the presence of pancreatic microsomes, an increase in Mr of 15 kDa is seen, in keeping with the addition of six core oligosaccharides to six potential glycosylation sites within the protein. There is also loss of a cleavable signal sequence.

Iil.

TARGETING TO THE ER MEMBRANE A. Signal Sequences

The initially translated forms of proteins that are integrated into the membrane of the ER are targeted to the membrane by a signal sequence in their primary structure (Blobel and Dobberstein, 1975). Signal sequences are generally located at the N-terminus of the protein and are 12 to 30 residues in length, containing a span of about ten hydrophobic uncharged residues, with charged residues in the sequences on each side of the hydrophobic core. The C-terminal side of the sequence often terminates at a small uncharged residue, such as Ala or Ser, where signal peptidase cleaves. Signal sequences of membrane proteins are sometimes cleaved off upon their arrival into the lumen of the ER, but also may remain uncleaved, becoming the portion of the protein that becomes integrated into the membrane (signal-anchor sequence).

B. Signal Recognition Particle When a nascent chain of a membrane protein emerges from the ribosome during biosynthesis, the signal sequence binds tightly to signal recognition particle (SRP), a ribonucleoprotein consisting of a 7S RNA species and six protein subunits of 72, 68, 54, 19, 14, and 9 kDa, which acts to slow or halt elongation (Walter and Blobel, 1981) until contact is made with a heterodimeric SRP-receptor protein associated with the ER membrane (Figure 3) (Meyer et al., 1982). Interaction with SRPreceptor reverses elongation stalling so that translation continues, and insertion of the nascent chain across the membrane begins. Sequence analysis and cross-linking have shown that the C-terminal domain of the 54-kDa subunit contains a methionine-rich domain which binds to the signal sequence (Zopf et al., 1990), while the N-terminal domains of both the 54-kDa subunit of SRP and the ~-subunit of SRP-receptor contain consensus GTP-binding sequences (release of SRP from the ribosome-nascent chain complex is known to require GTP). Components homologous to the SRP 54- and 19-kDa subunits, and the 7S RNA species occur in yeast, where gene knock-out experiments have shown that only a subset of

114

DAVID STEPHENS, SUNITA KULKARNI, and BRIAN AUSTEN

~

~

CAP

Cytoplasm

,~r

"/v Gr , / ~

~9,~'" ... ~o~e9'~ ..oo~.,9'~ ~ ~,~o ~ ~s'~~o,~o,-

~" - ~"

z S I g n a ~ ase

~J~"'J~-~.~

~tPA ~ v

L u m e n

~/

Figure 3. Mechanisms of targeting to the ER membrane. Shown are the components involved in the translocation of secretory proteins and in the assembly of membrane proteins. The abbreviated symbols are described in the text.

membrane and periplasmic proteins requires SRP for translocation across the ER (Hann et al., 1992).

C. Heat-Shock Chaperones Some proteins do not require SRP for targeting to the ER membrane. An alternative pathway involves heat-shock protein hsp70 (Figure 3) (Deshaies et al., 1988). This chaperone protein, the expression of which is induced in cells subjected to stress, is thought to bind precursor proteins in the cytoplasm, helping them to remain loosely folded until they are released with concomitant ATP-hydrolysis near the ER membrane prior to their insertion.

D. The Translocation Complex A number of membrane proteins resident in the ER are thought to comprise a translocase complex, which is likely to include a protein channel through which nascent membrane proteins can pass from the cytoplasm to the lumen (Figure 3). The evidence so far indicates that some, but not all components, are also involved in the translocation of secretory proteins. Generation of temperature-sensitive yeast

The Biosynthesis of Membrane Proteins

115

mutants defective in translocation of secretory proteins has identified three genes, Sec61, Sec62, and Sec63 (Stifling et al., 1992) which code for membrane proteins of the ER. In addition to a requirement for Sec61 and Sec63, efficient insertion of some polytopic membrane proteins also requires three other genes, Sec71, Sec72, and Sec73 in yeast (Green et al., 1992). Conductivity measurements across microsomal membranes show the existence of a channel which is opened by puromycin, an antibiotic which clears nascent chains from the channel (Simon and Blobel, 1991). The channel is kept open by the binding of the ribosome to the membrane, indicating that its point of attachment to the membrane is close to the channel through which the nascent chain passes. Photoreaction of truncated nascent chains containing photoreactive amino acids of both type 1 and type 2 membrane proteins (see below) paused in the act of translocation across pancreatic microsomal membranes cross-link to a mammalian homologue of Sec61 (High et al., 1991). The sequence of Sec61 predicts that this protein has multiple transmembrane segments, making it a likely candidate for a component of a channel into which insertion of newly synthesized membrane proteins takes place. Truncated transferrin receptor, which contains a type 2 transmembrane region (see below) additionally photo cross-links to a 42-kDa protein. A similar integral membrane protein, with a 15-kDa cytoplasmic domain, is reported to cross-link to a photoreactive synthetic signal peptide ligand (Robinson et al., 1987), and may act as a membrane receptor for signal sequences. The receptor is specifically located in the membranes of the rough ER and has a 50-kDa homologue in yeast microsomal membranes.

IV. ESTABLISHING T O P O G R A P H Y A. Assemblyof Single-PassMembrane Proteins The topography of single-pass membrane proteins is determined by the structure and relative position of signal sequences and transmembrane domains. These domains consist of an extensive (over 20 residues) region of adjacent hydrophobic amino acids. The charges on the residues near these hydrophobic domains, and the way they interact with the translocation channel in the membrane (Figure 4), determine the final topography of the protein. There are three types of single-pass membrane proteins. The first type contains a cleavable signal sequence at the N-terminus, and a second 'stop-transfer' sequence within the mature sequence of the proteins, which becomes the transmembrane domain. These proteins, typified by the G-protein of the vesicular stomatitis virus, and class 1 and 2 histocompatibility antigens, adopt a 'Type I' topography with the N-terminus of the mature protein on the lumenal side of the ER, and eventually in the extracellular environment when the protein moves to the cell surface. The C-terminus remains in the cytoplasm.

Cleaved

signal

,

Type1

tTC

~,,.,~___~..~

~

TC

N/~

il~,,i,'i i i~i, i~i ~i i~',~i i i, , ,ii~iii;,~i,l,i,i~i i/ii,i,i!,i~',i~i:ii li~i~i;~211,~:iii!l ,~,~,~,~i~,!! ,i~,,,,~ii,

5

,

s , ,, ,., ,, , ,, r,,

~ ,.

3'

c

c

J

!iiii'~iiii'!i!~ 5'

3'

NN

N

R.

Signal anchor Type1

J

N

.

~

.~:

~ii.:".:.~!~i~L::.:.~.i'}}i~i~i.:.~i~!~.:~ ' .';.~~ii~~il~:.~ i i~i!i~i~i-~

Figure 4. Mechanisms of assembly of single-pass membrane proteins. The three different modes of assemblyfor proteins involve interaction with SRP,SRP-receptor (DP) and the translocationcomplex (TC).Type 1 topographyleavesthe N-terminus in the ER lumen, and the C-terminus in the cytoplasm. The opposite orientation is type 2. The orientation of proteins with uncleavedsignals is influenced by the presenceof charged residues(+ or-).

116

The Biosynthesis of Membrane Proteins

117

It is thought that the signal sequence first loops across the membrane, initiating translocation, and is then cleaved on the lumenal side by the exopeptidase, signal peptidase. Signal peptidase consists of a complex of six proteins and glycoproteins (Shelness et al., 1993), with its active site confined to the ER lumen. Translocation across the membrane continues until the stop-transfer sequence in the mature part of the protein enters the translocation channel. This entry stops the translocation process, allowing the protein to move laterally from the translocation channel into the phospholipid bilayer (Figure 4). In the second type of single-pass membrane protein, the signal sequence is uncleaved, and after interacting with SRP and inserting into the translocation complex, the signal sequence becomes the transmembrane segment, being then termed a signal-anchor sequence (Figure 4). Whether or not cleavage by signal peptidase occurs, depends on the nature of the hydrophilic residues on either side of the hydrophobic domain (Haeuptle, 1989), and the length of the hydrophobic domain (Sakaguchi et al., 1992). These proteins may adopt either type 1 or type 2 topography. Type 2 topography refers to a cytoplasmic N-terminus, and an extracytoplasmic C-terminus (Figure 4) (e.g., sialyl transferase, transferrin receptor), and is adopted if there are several positively charged residues on the N-terminal side of the signal-anchor (Parks and Lamb, 1991). Proteins with type 1 orientation (N-terminus lumenal; e.g., cytochrome P450, epoxide hydratase) tend to have acidic or neutral residues on the N-terminal side of the signal-anchor sequence.

B. Assemblyof Multi-Spanning Proteins Multiple spanning proteins pose the more difficult problem of how several hydrophobic segments that span the membrane with alternating orientations are assembled. One possibility is that these proteins contain successive helical hairpin loops consisting of an N-terminal signal-anchor sequence followed by a stop-transfer sequence (see Figure 5), so that insertion of the N-terminal loop would occur first, followed by successive loops towards the C-terminus. Studies with fusion proteins containing a number of signal-anchor sequences show that only the N-terminal one is required for SRP-mediated targeting (Wessels and Speiss, 1988). One signal-anchor sequence placed on the C-terminal side of another signalanchor sequence can act as a stop-transfer sequence, but when three hydrophobic signal anchors are present, not all insert correctly (Lipp et al., 1989). The cytoplasmic loops of seven-transmembrane receptor proteins are more highly positively charged than extracytoplasmic loops; the second cytoplasmic loop is the most highly concentrated in positively-charged residues, suggesting that the topography may be dictated by the charge asymmetry on each side of the third and fourth transmembrane domains, and that other loops may be of lesser importance.

118

DAVID STEPHENS, SUNITA KULKARNI, and BRIAN AUSTEN ~

,TC

/'~TC

r

AT(:

/"~.

D. ("}.

Figure 5. Mechanisms of assembly of a seven-transmembrane receptor protein. Mechanism of membrane assembly is thought to involve successive interactions of signal-anchor sequences with first SRP and then the translocation complex (TC), followed by insertion of successive loops into the membrane.

C. Membrane Proteins with C-terminal Anchors A class of membrane proteins do not have signal sequences, do not require SRP for membrane insertion, and have membrane anchors at their C-terminus. An example is cytochrome bs, which is cytoplasmically orientated (Fleming and Strittmatter, 1978). These proteins are found in several cellular compartments, such as the plasma membrane (e.g., middle T-antigen), the outer mitochondrial membrane (e.g., OM cytochrome bs), and as components of transport vesicles (e.g., Sly2) (see Kutay et al., 1993). Details of their mechanism of assembly are presently unknown, but they are thought to insert into the bilayer after completion of synthesis on ribosomes.

V. PROCESSING IN THE ER LUMEN A. Folding and Disulfide Pairing The extracytoplasmic parts of proteins fold and oligomerize in the ER lumen. Misfolded, aggregated, and incorrectly assembled proteins are degraded. Folding and oligomerization are helped by proteins resident in the lumen of the ER, including the heavy chain binding protein (BiP) and GRP94 (Gething and Sambrook, 1992), prolyl cis-trans isomerase (Stamnes et al., 1991), and protein disulfide isomerase, a redox enzyme required for correct pairing of disulfides (Freedman, 1984). To allow disulfide formation, the redox conditions are more oxidizing in the ER lumen than in the cytosol. The folding of newly synthesized G-protein from vesicular stomatitis virus involves rapid formation of a noncova-

The Biosynthesis of Membrane Proteins

119

lent, multimeric complex containing BiP, from which trimeric oligomers containing intrachain disulfides gradually emerge (de Silva et al., 1993).

B. Formationof Oligomers The forces that drive oligomerization are different for different proteins. In the case of the T-cell receptor, an octamer with six different subunits, the energy comes from ion pairing across transmembrane domains within the phospholipid bilayer. The acetylcholine receptor, a pentamer of four different subunits, oligomerizes via interactions in the extracellular domains of the subunits (Hall, 1992).

VI.

RETENTION OF MEMBRANE PROTEINS IN THE ER

Resident membrane proteins are involved in the synthesis and post-translational modification of other proteins, steps that occur in the ER and Golgi complex. Many of these proteins would ordinarily be swept through the secretory pathway by the process of "bulk flow". This must be avoided so the cell does not have to continually synthesize the proteins; retaining them in their required location is a more efficient process. The anterograde vectorial movement that takes membrane proteins to the cell surface occurs by default and retention of ER resident proteins is an active process. Some ER resident membrane proteins have C-terminal tetrapeptides in their cytoplasmic tails to specify ER retention. The sequences-KKXX or-KXKXX are used (X = any amino acid) and are found at the C-terminus of a number of type I membrane proteins. Mutagenesis studies have shown that the lysine at position-3 (numbering is relative to the carboxyl terminus) is essential for retention, while those at residues -4 and-5 merely have to be positive (Shin et al., 1991). These studies also indicated that the transmembrane domain is important. The "double lysine" motif specifies localization to the rough ER (e.g., the 13-subunit of the signal sequence receptor; Wada et al., 1991) and the smooth ER (e.g., HMGCoA reductase; Jackson et al., 1990). The distribution of proteins containing the double lysine sequence would suggest a receptor-mediated retrieval mechanism similar to that required for soluble proteins (Pelham, 1992). As yet, no such receptor has been identified. Most ER localized transmembrane proteins do not contain this motif and there is evidence that they are retained because of their transmembrane domains. An example of this is the rotavirus protein VP7 (Stirzaker et al., 1990). ER retention by transmembrane domains is an analogous situation to that observed for retention of a number of Golgi complex resident proteins (see below). It is thought that the proteins aggregate in the ER and the resulting clusters are excluded from budding vesicles and are thus retained.

120

VII.

DAVID STEPHENS, SUNITA KULKARNi, and BRIAN AUSTEN

TRANSPORT OF NEWLY SYNTHESIZED MEMBRANE PROTEINS IN VESICLES A. BuddingReactions in the Golgi Complex

Budding and fusion of transport vesicles is central to the intracellular movement of membrane proteins between the ER and the Golgi complex, through the various cisternae of the Golgi complex, and from the TGN to the cell surface or other subcellular organelles (Figure 1). The majority of the information about mechanisms has been obtained from yeast genetics, and by reconstitution of transport steps in cell free systems (Melancon et al., 1991). Vesicles that bud from the TGN during regulated secretion and lysosomal sorting are coated in the same way as those involved in endocytosis, but are distinct from the coated vesicles involved in ER-Golgi complex traffic which have non-clathrin coats. Coated pits are specialized membrane domains that cause clustering of molecules. The coat is formed from clathrin molecules which form a cage around the invaginating membrane and resulting vesicle. Individual clathrin molecules have a triskelion (three-legged) structure which is a complex of three 180 kDa and three smaller (33-35 kDa) molecules. Purified clathrin is well known for its properties of self-assembly into the structures that coat endocytic vesicles (for a review see Pearse and Crowther, 1987). Vesicles that carry proteins between the Golgi cisternae are coated with proteins that pre-exist together as a 600 kDa complex in the cytosol termed a "coatomer"--a complex of a number of coat proteins (COPs) (x- (160 kDa), [3- (110 kDa), y- (105 kDa), and 5-COP (61 kDa), and other small polypeptides (Rothman and Orci, 1992). 13-COP is also believed to be a key component of the structural scaffold that maintains the shape of the Golgi complex and may be a target site for brefeldin A, a drug which blocks movement of proteins to the cell surface by causing a reversible redistribution of the Golgi complex back into the ER. The coats of the Golgi-derived vesicles have also been shown to contain a small, 21 kDa, GTP-binding protein called ADP-ribosylation factor (ARF) (Figure 6). It is so named as it is a cofactor in the cholera toxin mediated ADP-ribosylation of G-protein c~-subunits. The formation of a coat of this kind on Golgi membranes induces budding (Orci et al., 1993). Binding of coat proteins is also thought to involve the large, hetero trimeric G-proteins as well as the small monomeric ones. The effects of GTP-y-S, a non-hydrolyzable analog of GTP, (causing accumulation of coated vesicles) can be mimicked with AIF4-, which is known to activate the ot-subunits of large G-proteins, while not affecting small G-proteins (Kahn, 1991). The target site of these large G-proteins and their mode of action on Golgi membranes remain unknown.

The Biosynthesis of Membrane Proteins

121

recycling of coat components

///

~

SNAP GDP + [ ]

SNARE

":

E> Coaled Vesicle

UNCOATING

FUSION

BUDDING Donor Membrane

Target Membrane

I = ARF

I []

= COATOMER

Figure 6. Transport in vesicles. Membrane proteins are carried in vesicles which bud off from donor compartments with the aid of coatomers and a small GTP-binding protein, ARF. After fusion with the acceptor compartment, the components are recycled for another round of budding and fusion. Abbreviations are described in the text.

B. Fusion of Vesicles The final step of any vesicle transport is fusion of transport vesicles with their target membranes (Figure 6). Membrane fusion is energetically unfavorable due to the polar nature of the two surfaces of the lipid bilayers. NSF (NEM-sensitive fusion protein) is a soluble protein with two ATP binding domains, and provides the energy required for fusion. Three soluble NSF attachment proteins (SNAPs) have been purified, characterized, cloned, and sequenced (Rothman and Orci, 1992). SNAP receptors (SNAREs) have been isolated by affinity chromatography on NSF and SNAP columns (S611ner, 1993). Three proteins, VAMP-2, syntaxins A and B, and SNAP-25 (for synaptosomal associated protein), were eluted by ATP. Due to their relative intracellular distributions (syntaxin on the presynaptic membrane, VAMP-2 on the vesicle surface) a simple fusion mechanism is envisaged where NSF/SNAPs link the two molecules and induce bilayer fusion. Specificity of fusion would presumably be mediated by membrane-specific SNAREs. A family of proteins related to VAMP-2 and syntaxin have been identified and are thought to perform the same role in other transport steps. SNAP-25 is unusual in that it is an integral membrane protein that contains no hydrophobic stretches. It is thought that it may serve as a fatty acyl-CoA acceptor since acyl-CoA is required for NSF-dependent fusion. This would not only serve to anchor SNAP-25 to the membrane, but if acylated on enough sites it would create a hydrophobic interface to serve as a trigger for bilayer fusion. These key findings reveal an important role

122

DAVID STEPHENS,SUNITA KULKARNI, and BRIAN AUSTEN

for NSF and SNAPs in constitutive and regulated secretion as well as in synaptic transmission.

C. Vesicle Movement Vesicles move along microtubules. Microtubules extend from the centrosome in a unipolar manner. The minus-end of the microtubule is always located at the nuclear-end and extension is mediated by addition of monomer units to the plus-end. Movement is generated by interactions between the cytoskeleton and motor proteins which hydrolyze an energy source to generate directional movement. There are two principal classes of motor, those that generate plus-end directed movement (e.g., kinesin) and those that transport their cargo to the minus-end of the microtubule, i.e., to the cell periphery (e.g., cytoplasmic dynein). Both are ATPases. The vesicle itself is believed to bind to the light chains at the tail-end of the motor. The head-ends of motor proteins are responsible for generating movement by ATP/GTP hydrolysis and are the regions that bind microtubules. The light chains of the kinesin superfamily (which includes bimC, ncd, and KAR3 are thought to be functionally interchangeable, forming an array of adaptors for various cargos (Goldstein, 1991). Thus, the light chains must bind to specific receptor molecules on the vesicle surface that specify the direction of transport. As yet, these receptor molecules remain unidentified. VIii.

GLYCOSYLATION

Membrane proteins are subject to many post-translational modifications during their transport to the cell surface. Many proteins are glycosylated. N-glycosylation is initiated in the ER lumen as a core oligosaccharide, composed of mannose (Man), N-acetyl-glucosamine (GlcNAc), and glucose (Glc), is transferred from a dolichol intermediate (Figure 7) to a protein sequon, Asn-X-Thr(Ser), where X is any amino acid residue in the acceptor protein. The enzymes involved in the transferase step comprise a complex of two 66-68 kDa glycoproteins and a 48-kDa resident ER membrane protein (Kelleher et al., 1992). Additional modifications that may occur to oligosaccharides as proteins are transported through the Golgi cisternae are shown in Figure 7.

IX.

RETENTION IN THE G O L G i COMPLEX

Mechanisms exist for retaining selected membrane-bound enzymes involved in oligosaccharide maturation within the cisternae of the Golgi complex. For example, N-acetylglucosaminyltransferase, ]3-galactoside ~-l,6-sialyltransferase (ST) are located in the medial- and trans-cisternae respectively. It has been shown that ST

~anrGolgi

Mediat-Golgi

c~s-o~Igi

0

Figure 7. Glycosylation of membrane proteins in the ER and Golgi. Successive steps of processing of oligosaccharide side-chains on newly-synthesized membrane proteins as they move through the ER and cisternae of the Golgi complex are shown. The disulfide-pairing enzyme, protein disulfide isomerase (PDI), and chaperone heavy chain binding protein (BiP) are involved in early stages of folding. The open squares depict N-acetyl-glucosamine; the open circles, mannose; the open triangles, glucose; the black circles, galactose and sialic acid. 123

124

DAVID STEPHENS,SUNITAKULKARNi,and BRIANAUSTEN

is retained in the trans-Golgi by its 17-residue transmembrane sequence (Wong et al., 1992). o~-1,4-Galactosyltransferase, another trans-Golgi enzyme, is also retained by its transmembrane sequence (Nilsson et al., 1991). Hydropathy plots show that the Golgi complex enzymes have slightly shorter transmembrane spans than plasma membrane proteins, and it has been suggested that the sequences would be unstable in the presence of the greater cholesterol concentration in the plasma membranes, which thus prevents their movement to the cell surface (Bretscher and Munro, 1993).

X. CONSTITUTIVE A N D REGULATED MOVEMENT TO THE CELL SURFACE Proteins are transported to the plasma membrane, either directly from the TGN (constitutive pathway; Figure 1), or indirectly, after a period of storage in the membrane of secretory granules (regulated pathway). The latter of these, regulated exocytosis, occurs in response to an external stimulus.

A. Receptor-Mediated Sorting Some membrane proteins pass through the Golgi complex to the TGN where they are incorporated into constitutive secretory vesicles by bulk flow (Figure 1). The molecular features of this are the same as for budding and fusion of transport vesicles within the Golgi apparatus, and non-clathrin coats are required. For proteins that are segregated into secretory granules for the regulated pathway, or into vesicles to be transported to endosomes, targeting signals are required, and budding involves clathrin. At the TGN, these proteins must be sorted into the appropriate carriers for secretion. Targeting signals are obviously required to sort proteins to their appropriate destination.

B. Passive-Sorting An alternative hypothesis to receptor-mediated sorting is "passive-sorting". This hypothesis suggests there is a selective aggregation of proteins in the TGN that results in subsequent transport to the secretory granules. Homotypic interactions are thought to occur in cells that sort proteins into distinct granule populations, while heterotypic associations account for cells such as exocrine acinar cells where a number of polypeptides are found in the same secretory granule population. Aggregation may, at least in some cases, be caused by the decrease in pH on passage from the ER to the TGN. The increased acidity of the TGN has been shown to cause aggregation of GP-2, a major component of zymogen granules (Fukuoka et al., 1992).

The Biosynthesis of Membrane Proteins

125

As well as aggregation, sorting of other proteins may be due to cytoplasmic tail sequences. P-selectin (CD62) is a granule membrane protein found in platelets and endothelial cells. Deletion of the C-terminal 23 amino acids results in a plasma membrane localization for the protein in transfected cells. Fusion of the same 23 amino acid sequence to a plasma membrane protein results in its retention in secretory granules (Disdier et al., 1992). As with other vesicle mediated processes, GTP binding proteins are proposed to be involved in regulated exocytosis. Padfield and Jamieson (1991) reported localization of at least seven small GTP-binding proteins on the cytoplasmic face of pancreatic zymogen granule membranes. Large, hetero trimeric G-proteins are als0 implicated in the process. In fact, the same subunit (Gai3) is thought to regulate secretory granule and constitutive secretory vesicle formation (Barr et al., 1991).

C. Recyclingof Membrane Components Secretory granules can be formed by an endocytic pathway as well as by the biosynthetic pathway direct from the TGN. This fits well with the passive-sorting mechanism, in that clustering of proteins in the plasma membrane would allow effective recycling of granule components and re-use in conditions of continued secretagogue stimulation. This has been demonstrated for synaptic vesicle components such as synaptophysin (Cameron et al., 1992). Indeed, there is some evidence that the endocytic route is the only pathway for the biogenesis of synaptic vesicles.

XI. MOVEMENT TO THE CELL SURFACE VIA THE ENDOSOME Some membrane proteins can move from the TGN to the cell surface via the endosome (Route 5 in Figure 1). One example is the mannose-6-phosphate receptor which transports soluble proteins destined for lysosomes from both the TGN and the cell surface via endosomes. The class 2 histocompatibility complex of macrophages consists of two subunits, the ~- and I]-chains. Intracellularly, they are also associated with invariant chain, a transmembrane protein exposing 30 amino acids on the cytoplasmic side of the membrane (Claeson et al., 1983). This complex moves via the TGN to the endosome, where the invariant chain is degraded; the class 2 antigen binds to an antigenic peptide, which is then presented on the cell surface. In melanoma cell lines, invariant chain is also localized in endosomes (Pieters et al., 1991). The endosomal targeting sequence may lie in the cytoplasmic tail of the invariant chain, and its degradation is known to occur at different rates in different cell types.

126

DAVID STEPHENS, SUNITA KULKARNI, and BRIAN AUSTEN

XII.

LYSOSOMAL MEMBRANE PROTEINS

Newly synthesized lysosomal membrane proteins, such as lysosomal acid phosphatase (LAP), are targeted to the lysosome from the TGN by signals in their cytoplasmic tail sequences. The information is usually centered in a cluster of amino acids surrounding a Tyr residue. A hexapeptide -PGYRHV- (residues 24-29 of the 163 amino acid cytoplasmic tail) is the important region for transport of LAP to lysosomes. Tyr is required for stabilization of a tight turn structure and for interaction with the cytoplasmic receptor. The interaction results in clustering of the lysosomal proteins in clathrincoated regions in the TGN or the cell surface from which they are transported to late or early endosomes (Pathway 3 in Figure 1), and subsequently on to lysosomes where the transmembrane region is cleaved to form the mature LAP molecule.

XIII.

LIPID MODIFICATION OF MEMBRANE PROTEINS

A group of post-translational modifications of proteins with various lipid modifications (exemplified in Figure 8) have been identified in eukaryotic cells, many of which are required for subcellular localization. The biosynthesis of these lipid groups has been studied by measuring the metabolic incorporation of radioactive fatty acid or mevalonate into the protein of interest. Farnesylation of ras proteins at the C-terminal sequence with the motif-CAAX (C = cysteine, A = aliphatic amino acid, X = any amino acid), together with the nearby presence of a polybasic sequence is sufficient to localize ras proteins at the cytoplasmic membrane. The -CAAX is first farnesylated at the Cys residue, the terminal AAX residues are removed, and the C-terminal Cys is methyl-esterified. In the case of the ras-related Krevl/rapl proteins, which are found in the Golgi complex, the X as leucine causes the protein to receive a geranylgeranyl group. The 14-carbon saturated fatty acid myristate is amide-linked to the amino-terminal glycine residue of a range of proteins, exemplified by the pp60src (proto) oncogene product. Site-directed mutagenesis has revealed this acylation is required both for membrane localization and transforming activity. Myristoylation occurs by donation from myristoyl-CoA co-translationally in the cytoplasm at the consensus sequence Met-Gly-X-X-X-Ser(Thr) or Gly. The enzyme is highly specific for acyl-chain length. A number of cell surface proteins contain palmitic acid residues covalently linked by thioester bonds to cysteines. The acquisition of these residues occurs in the intermediate compartment between the ER and the cis-Golgi complex (Bonatti et al., 1989). Some cell surface proteins are attached by the carboxyterminal residue to ethanolamine in an amide linkage, which is then linked via glycans to phospholipid. The proteins are initially integrated by a stop-transfer sequence into the membrane,

The Biosynthesis of Membrane Proteins

127

/NH

1CH2)2

)

f

Ni~i i~i~i~.~..:i!i-;.i.:..'i!~.~.~,.' ~i ~i i i i~i :,.'~.:.:.:.' i i?!~..~i i ~i:i N.:.:.:.:....:~....:..........:.::.~.:.:.:+.....:.:~:.~.:`:.:.:;:,:.......:.:.:.:...:...:.:.:.:e.:..~e.:e.:`..........:~.....~......:.:.:.::?~:..~!~..e.$~.~.~. i:,.'.::?~ii;.?~i~~i~i ii~ii !i:"~ii~.~;.'!i~iii.~!~iii:ii~i ~ .;::,.':;.i~ii..'.,:ii!~~i~ii~:,."~i~:,:!i!..'.:,.!:i~j !'~l-~Jiii.~ ~ ~:,::,.".;~.~j~i.;.' !~ i~, ~:,.'~.@.::~.~ ~ g~:;:+"~N ....~?~ .~;,:,.:~.:.:.~~.:.~:"-.:..-.:...., 9 . . .,:~,;:,,,~,z:~..',.~,.:' .~.~.:j ~:.:.::!:~:~.:.:.:.:.:..' ~.:ee.:.:::::.~::e.:.:e.:.:.:.:.: ..:.:.:.:.:..' "~"".~z.?' :':':':" ".:.:-:.:.:-.' ..:..' . . . .,:,~x.,,:.:.~.::.:::~:.x.:-' ~i!!ii~i!~~:~!ii~':"."~i~'"~i!iii ........................iiiiiii~iiiiii~i~!b:iiiiiiii.~i~i~1~!iiiiiifii~iii~ili~i~~ilii~!i

':::::*'~ .... ~': .......... "::.~~,~'"::" .................... "

.,',.~i~i~li~.~!~!i..:~!i~i~i~i~i~~,.~"~i. ~",.:-:.i~:i~,~, ........~i i.~.:.~.,~. ~!:.:..:`...~i~:...`:~.i~`.:.~`~i~..:..~...~i~i~~.`~.~:~..~i~i~....~i~i~.:~Z~ ~ ~...................~.:.~....................~:..,:~:....,:...:~.....................~...........................................:,.............,...................... ......................................................... ~i~...i.~i~..i~..i~..!.~.!.-.:-.~i...;..:.~.:....:.~.i.~..:.:~..":.i .i:.,.~ ................. K

tCys~

(

jC%

s~

O--C'_

o

/r

k"C

<S

~

~NH

c~/

Addition:

Palrnitate

Farnesyl

Myristate

Inositol glycolipid

Linkage:

Thioester

Thioether

Amicle

Amide

Examples:

Transferrin receptor

p21 '*'

p60 ~ AMP-depmndentprotein kinase

Th),- 1

YPT1

5'-Nucleofidase

IgE receptor Ankyrin

N-CAM 120

IL-2 receptor

VSV-G protein

cholinesterase

Figure 8. The structures of lipids found covalently linked to membrane proteins.

and then transferred to the p h o s p h o l i p i d anchor. T h e g l y c o p h o s p h o l i p i d anchor plays a role in locating proteins to the apical m e m b r a n e in s o m e polarized cells.

XIV.

BIOSYNTHESIS OF MITOCHONDRIAL PROTEINS A.

MEMBRANE

I m p o r t into M i t o c h o n d r i a

M i t o c h o n d r i a contain 1 0 - 2 0 % of the intracellular proteins, m o s t o f which are b i o s y n t h e s i z e d on free r i b o s o m e s (Pathway I in F i g u r e 1) in the c y t o p l a s m , and then imported. M i t o c h o n d r i a contain two m e m b r a n e s , so the targeting m a c h i n e r y

128

DAVID STEPHENS, SUNITA KULKARNI, and BRIAN AUSTEN

must differentiate proteins of these two membranes. Most mitochondrial membrane proteins are encoded on nuclear genes, translated in the cytoplasm, and then must find their way into their final locations, either in the inner membrane or the outer membrane. Proteins that insert into or translocate the inner membrane are initially made with N-terminal transit sequences; import is post-translational and requires energy from both ATP hydrolysis and a membrane potential across the inner membrane.

B. Transit Sequences Mitochondrial proteins such as cytochrome cl or b that end up in the inner membrane are initially translocated into the mitochondrial matrix by similar mechanisms to those followed by matrix proteins. They are translated with complex N-terminal signals of between 40 and 70 residues in length. The extreme N-termini are rich in basic and hydroxylated amino acid residues, which are cleaved off after import into the matrix (Hurt et al., 1984). Transit sequences have been shown by physical methods to form an amphiphilic helix in the presence of either organic solvents, or acidic phospholipids such as dimyristyl phosphatidyl ethanolamine, which are enriched in mitochondrial membranes (L.K. MacLachlan, P.I. Haris, D.G. Reid, J. White, D. Chapman, J.A. Lucy, and B.M. Austen, unpublished results). The amphiphilic sequence may insert into the phospholipid bilayer of the mitochondrial membrane with its hydrophobic side penetrating the hydrophobic interior, and the hydrophilic surface in contact with the charged aqueous surface. Synthetic transit sequences have been shown to disrupt phospholipid bilayers, and this is likely to be relevant to the way that precursors are translocated. In precursors of proteins that are inserted predominantly in the IMS (inter-membrane space) side of the inner membrane, the amphiphilic sequence is followed by a hydrophobic sequence that acts to re-export the protein to the IMS. Cleavage takes place in two stages. The amphiphilic sequence is removed in the matrix, and the hydrophobic sequence is removed in the IMS. The Rieske Fe-S protein, a component of the cytochrome bcl complex, also enters the IMS via the matrix. Its targeting sequence is cleaved in two steps in the matrix, and the mature form is then translocated back across the membrane after complexing with Fe and S. Some complex presequences are implicated in a variety of secondary functions. For example, the ATPase inner membrane proteolipid subunit of 81 amino acids from N. crassa, a small hydrophobic inner membrane protein, is synthesized initially with a 66 basic presequence, which may be involved in maintaining solubility during transport (Hartl et al., 1989).

C. Components Involved in Mitochondrial Import Most recent ideas about mechanisms of import have been reviewed (Schatz, 1993) (Figure 9). In yeast, newly translated precursors first bind the cytosolic

The Biosynthesis of Membrane Proteins Pre Cytochrome cl

Transit sequence 1~ § + Signallsequence hsp70 /% / \ 1~' " ~ "- . . . . . ~"~ \ \ \ \ \ 3 j ~

129

ADP/ATP carder protein

Figure 9. Mechanisms of insertion of membrane proteins into mitochondria. Precursors of inner membrane proteins are kept in an unfolded state by binding to hsp70, prior to recognition by surface receptors Morn72 or Mom19. These move bound precursors laterally to a proteinaceous import site situated where inner and outer membranes contact. AAC then moves directly into the inner membrane. In the matrix, the proteins are helped to fold by interaction with mitochondrial hsp70, followed by hsp60. Cleavage of the amphiphilic N-terminal sequence exposes a hydrophobic region which actively retranslocates the protein to the periplasm. chaperone hsp70, and the complex then contacts a farnesylated protein YDJ 1 at the outer membrane, which triggers release of the precursor in an ATP-dependent manner close to the membrane. These interactions serve to keep the precursor unfolded. Several receptors (e.g., MOM 19 and MOM72 in Neurospora) are present on the outer membrane where they interact with subsets of mitochondrial proteins and carry them to an import site at contact sites between inner and outer membranes. A yeast protein, MP11, identified by genetic approaches, and Isp45 identified by cross-linking precursors stuck in the translocation site, are potential components of a proteinaceous channel through which the precursors pass (Figure 9). In the matrix, translocated proteins contact the mitochondrial chaperone hsp70, which concomitant with ATP hydrolysis, pass them on to hsp60, which chaperones their folding and oligomerization. Transit sequences are released by the proteolytic action of the Mn 2+ -dependent protease, c~-MPP, and the protease-enhancing protein, ~-MPP.

130

DAVID STEPHENS, SUNITA KULKARNI, and BRIAN AUSTEN

D. Insertion Directly into the Inner and Outer Membranes The ADP/ATP carrier, in contrast, is thought to translocate directly from the cytoplasm into the inner membrane at contact sites between the inner and outer membranes. The carrier is biosynthesized without an N-terminal cleavable extension, but the information for targeting is in the N-terminal 115 residues of the mature protein. The precursor binds to the MOM72 receptor on the cell surface in an ATP-dependent manner, moves to the proteinaceous channel, then under the influence of membrane potential, inserts directly into the inner membrane (Figure 9). There is no special requirement for cleavable transit sequences for proteins that insert directly into the outer membrane. However, the N-terminus of the outer membrane protein OMM70 fused to other proteins actually causes import into the matrix, indicating that OMM70 must contain a stop-transfer sequence in the rest of its structure (Nakai et al., 1989).

XIV.

BIOSYNTHESIS OF BACTERIAL M E M B R A N E PROTEINS A. Outer Membrane Proteins

The biosynthesis of a number of outer membrane proteins, including OmpA, OmpF, LamB and lipoprotein of E. coli have been studied (Baker et al., 1987). These are initially translocated across the cytoplasmic membrane in a similar manner to periplasmic proteins. The pre-protein forms are initially made with N-terminal signal sequences similar in structure to those of eukaryotic cells, i.e., with a hydrophobic core, a positively charged amino-terminus, and a more polar C-terminus terminating in a signal peptidase cleavage site (von Heijne, 1981 ). These proteins are translocated across the cytoplasmic membrane by sec-dependent pathways. Genetic techniques have been used to identify the genes. Mutations in one of these, secY, which is structurally similar to the sec61 of eukaryotic ER membranes, was found to reverse the blockage of protein secretion caused by defects in the signal sequences of exported proteins, suggesting that secY interacts directly with signal sequences. Proton motive force across the membrane is required for translocation (Driessen and Wickner, 1991). In the cytosol, precursors form stoichiometric complexes with chaperones such as secB, GroEL, or trigger factor, which prevent aggregation and contribute to targeting specificity (Randall et al., 1987). Transfer across the membrane is catalyzed by pre-protein translocase, a multi-subunit enzyme of secA, a protein bound to acidic phospholipids on the cytoplasmic surface, SecY, SecE and a smaller polypeptide, all of which are embedded in the membrane (Brundage et al., 1990). A diagram depicting the interactions involved in the translocation of pro-OmpA is

TheBiosynthesisof MembraneProteins

131

~A

3" N

l

Figure 10. Translocation of pro-OmpA through the bacterial cytoplasmic membrane. SecB binds to pro-OmpA, retaining it in an open conformation. Near the membrane, pro-OmpA binds the peripheral ATPase, SecA, which under the stimulation of membrane components including acidic phospholipids, hands pro-OmpA to secY in the translocation complex. After translocation, which requires the electrochemical potential across the membrane, the signal sequence is cleaved by leader peptidase (LP).

shown in Figure 10. SecY is similar in structure to Sec61, proposed as a protein channel in eukaryotic membranes of the ER. Deletion of the ffh gene, which is homologous to the mammalian gene for SRP54K (see above), has been shown to be deleterious for processing of the outer membrane proteins, LamB and OmpF (Phillips and Silhavy, 1993), but processing of other periplasmic proteins is unaffected. Ffh binds signal sequences, and may act as an alternative chaperone in the cytoplasm to keep precursors unfolded. In the periplasm, the signal sequence is cleaved by signal peptidase I, a 36-kDa protein integrated into the membrane. The cleaved protein is then transferred from the cytoplasmic to the outer membrane. The protein may be translocated across in vesicles, or alternatively may move directly across at Bayer's patches, which are areas on the bacterial surface where the inner and outer membranes come together.

132

DAVID STEPHENS, SUNITA KULKARNI, and BRIAN AUSTEN

E. c o l i produces several lipoproteins which are initially synthesized as pre-proteins. They are translocated across the cytoplasmic membrane, and then lipid modified at their amino termini prior to cleavage by the globomycin-sensitive signal peptidase II. The mature lipoproteins then attach to the periplasmic leaflet of either the cytoplasmic or outer membrane via their lipid tails. The final location is determined by an Asp present at the +2 position of the cytoplasmic lipoproteins.

B. Membrane Proteins in the Cytoplasmic Membrane Very few prokaryotic cytoplasmic membrane proteins have cleavable signal sequences. These include the major coat protein encoded by gene 8 of bacteriophage M 13, and the minor coat protein encoded by gene 3 of the fl bacteriophage adopt class 1 topology (N-terminus out). The small (50 residue) M13 coat protein is assembled directly with correct topology into the E. coli cytoplasmic membrane without the aid of the sec gene products, but does require membrane electrochemical potential. Assembly of the fl procoat was proposed to be an integral part of the folding pathway of the protein; the membrane was proposed to trigger a conformational change within the precursor so that its buried hydrophobic surfaces became exposed, and then assembled into the phospholipid bilayer (Wickner et al., 1991). The gene 8 procoat has also been shown to assemble spontaneously after translation on ribosomes that are not attached to the cytoplasmic surface. Signal peptidase itself is anchored to the membrane by two hydrophobic C-terminal domains. The N-terminal domain can insert independently of the Sec machinery, while translocation of the periplasmic C-terminal domain is Sec-dependent (Andersson and von Heijne, 1993). As with eukaryotic membrane proteins, a concentration of positive charges prevent a segment of a membrane protein moving from the cytoplasm to the periplasm. These may either interact with the negatively charged headgroups of phospholipids or with SecA. Analyses of multi-transmembrane proteins has shown that positive charges are more frequent in the loops on the cytoplasmic side of the membrane than in those in the periplasm (von Heijne, 1986). The dipoles associated with the lipid headgroups presumably make the membrane more easily penetrated by anions than cations. Alternatively, it is the charge distribution in the proteins in the translocation complex that are important in preventing positive charges translocating.

REFERENCES Andersson, H., & von Heijne, G. (1993). See dependent and see independent assemblyof E. coli inner membrane proteins: the topologicalrule depends on chain length. EMBOJ. 12, 683-691. Baker, K., Mackman,N., & Holland,I. B. (1987).Geneticsand biochemistryof the assemblyof proteins into the outer membrane of E. coli. Prog. Biophys. Molec. Biol. 49, 89-115.

The Biosynthesis of Membrane Proteins

133

Barr, E A., Leyte, A., Mollner, S., Pfeuffer, T., Tooze, S.A., & Huttner, W. B. (1991). Trimeric G-proteins of the trans-Golgi network are involved in the formation of constitutive secretory vesicles and immature secretory granules. FEBS Lett. 294, 239-243. Blobel, G., & Dobberstein, B. (1975). Transfer of proteins across membranes. 1. Presence of proteolytically processed and unprocessed nascent immunoglobulin light chains on membrane-bound ribosomes of mouse myeloma. J. Cell Biol. 67, 835-851. Bonatti, S., Migliaccio, G., & Simons, K. (1989). Palmitylation of viral membrane glycoproteins takes place after exit from the endoplasmic reticulum. J. Biol. Chem. 264, 12590-12595. Bretscher, M. S., & Munro, S. (1993). Cholesterol and the Golgi apparatus. Science 261, 1280-1281. Brundage, L., Hendrick, J. P., Schiebel, E., Driessen, A. J. M., & Wickner, W. (1990). The purified E. coli integral membrane proteins secY/E is sufficient for reconstitution of secA dependent precursor protein translocation. Cell 61,649-657. Cameron, P. L., Sudhof, T. C., Jahn, R., & De Camilli, P. (1992). Colocalisation of synaptophysin with transferrin receptors; implications for synaptic vesicle biogenesis. J. Cell Biol. 115, 151-164. Claesson, L. C., Larhammer, D., Rask, L., & Peterson, P. A. (1983). cDNA for the human invariant chain of class II histocompatibility antigens and its implications for protein structure. Proc. Natl. Acad. Sci. USA 85, 7395-7399. de Silva, A., Braakman, I., & Helenius, A. (1993). Posttranslational folding of vesicular stomatitis virus G protein in the ER: involvement of noncovalent and covalent complexes. J. Cell Biol. 120, 647-655. Deshaies, R. J., Koch, B. D., Werner, W. M., Craig, E. A., & Scheckman, R. (1988). A subfamily of stress proteins facilitates translocation of secretory and mitochondrial precursor polypeptides. Nature 332, 800-805. Disdier, M., Morrissey, J. H., Fugate, R. D., Baintob, D. E, & McEver, R.P. (1992). Cytoplasmic domain of P-selectin (CD62) contains the signal for sorting into the regulated secretory pathway. Mol. Cell Biol. 3, 309-321. Driessen, A. J. M., & Wickner, W. (1991). Proton transfer is rate-limiting for translocation of precursor proteins by the Escherichia coli translocase. Proc. Natl. Acad. Sci. USA 88, 2471-2475. Fleming, P. J., & Strittmatter, P. (1978). The nonpolar segment of cytochrome bs. J. Biol. Chem. 247, 8198-8202. Freedman, R. B. (1984). Native disulphide bond formation in protein biosynthesis; evidence for the role of PDI. Trends Biochem. Sci. 9, 438--441. Fukuoka, S., Freedman, S. D., Yu, H., Sukhatme, V. P., & Scheele, G. A. (1992). GH-2/THP gene family encodes self-binding glycosylphosphatidylinositol-anchored proteins in apical secretory compartments of pancreas and kidney. Proc. Natl. Acad. Sci. USA 89, 1189-93. Gething, M.-J., & Sambrook, J. (1992). Protein folding in the cell. Nature 355, 33-45. Goldstein, L. S. B. (1991). The kinesin superfamily; tales of functional redundancy. Trends Cell Biol. 1, 93-98. Green, N., Fang, H., & Walter, P. (1992). Mutants in three novel complementation groups inhibit membrane protein insertion into and soluble protein translocation across the ER membrane of S. cerevisiae. J. Cell Biol. 116, 597-604. Haeuptle, M.-T., Flint, N., Giough, N. M., & Dobberstein, B. (1989). A tripartite structure of the signals that determine protein insertion in the endoplasmic reticulum membrane. J. Cell Biol. 108, 1227-1236. Hall, Z. W. (1992). Recognition domains in assembly of oligomeric membrane proteins. Trends Cell Biol. 2, 66-68. Hann, B. C., Stirling, C. J., & Walter, P. (1992). Sec65 gene product is a subunit of the yeast signal recognition particle required for its integrity. Nature 356, 532-533. Hartl, F.-U., Pfanner, N., Nicholson, D. W., & Neupert, W. (1989). Mitochondrial protein import. Biochim. Biophys. Acta. 988, 1-45.

134

DAVID STEPHENS,SUNITA KULKARNI, and BRIAN AUSTEN

High, S., Gorlich, D., Wiedmann, M., Rapoport, T. A., & Dobberstein, B. (1991). The identification of proteins in the proximity of signal anchor sequences during their targeting to and insertion into the membrane of the ER. J. Cell Biol. 113, 35--44. Hurt, E. C., Pesoid-Hurt, B., & Schatz, G. (1984). The cleavable prepiece of an imported mitochondrial protein is sufficient to direct cytosolic dihydrofolate reductase into the mitochondrial matrix. FEBS Lett. 178, 306-310. Jackson, M. R., Nilsson, T., & Peterson, P. A. (1990). Identification of a consensus motif for retention of transmembrane proteins in the ER. EMBO J. 9, 3153-3162. Kahn, R. A. (1991). Fluoride is not an activator of the smaller (20-25 kDa) GTP-binding proteins. J. Biol. Chem. 266, 15595-15597. Kelleher, D. J., Kreibich, G., & Gilmore, R. (1992). Oligosaccharyltransferase activity is associated with a protein complex composed of ribophorins I and II and a 48 kDa protein. Cell 69, 55-65. Kutay, U., Hartmann, E., & Rapoport, T. (1993). A class of membrane proteins with a C-terminal anchor. Trends Cell Biol. 3, 72-75. Lipp, J., Flint, N., Haeuptle, M.-T., & Dobberstein, B. (1989). Structural requirements for membrane assembly of proteins spanning the membrane several times. J. Cell Biol. 109, 2013-2022. MacLachlan, L. K., Hafts, P. I., Reid, D. G., White, J., Chapman, D., Lucy, J. A., & Austen, B. M. (1994). A spectroscopic study of the mitochondrial transit peptide of rat malate dehydrogenase. Biochem. J. submitted for publication. Melancon, P., Franzusoff, A., & Howell, K. E. (1991). Vesicle budding; insights from cell-free assays. Trends Cell Biol. 1, 165-171. Meyer, D. I., Krause, E., & Dobberstein, B. (1982). Secretory protein translocation across membranesm the role of the docking protein. Nature 297, 647-650. Nakai, M., Hase, T., & Matsubara, H. (1989). Precise determination of the mitochondrial import signal contained in a 70 kDa protein of yeast mitochondrial outer membrane. J. Biochem. 105, 513-519. Nilsson, T., Lucocq, J. M., Mackay, D., & Warrren, G. (1991). The membrane spanning domain of beta-l,4-galactosyltransferase specifies trans Golgi localisation. EMBO J. 10, 3567-3575. Orci, L., Palmer, D. J., Ravazzola, M., Perrelet, A., Amherdt, M., & Rothman, J. E. (1993). Budding from Golgi membranes requires the coatomer complex of non-clathrin coat proteins. Nature 362, 648-652. Padfield, P. J., & Jamieson, J. D. (1991). Low molecular weight GTP-binding proteins associated with zymogen granule membranes from rat pancreas. Biochem. Biophys. Res. Comm. 174, 600-605. Parks, G. D., & Lamb, R. A. (1991). Topology of eukaryotic type 2 membrane proteins: importance of N-terminal positively charged residues flanking the hydrophobic domain. Cell 64, 777-787. Pearse, B. M. E, & Crowther, R. A. (1987). Structure and assembly of coated vesicles. Annu. Rev. Biophys. Biophys. Chem. 16, 49-68. Pelham, H. R. B., Roberts, L. M., & Lord, M. (1992). Toxin entry: how reversible is the secretory pathway. Trends Cell Biol. 2, 183-5. Phillips, G. J., & Silhavy, T. J. (1993). The E. coli ffh is necessary for viability and efficient protein export. Nature 359, 744-746. Pieters, J., Horstmann, H., Bakke, O., Griffiths, G., & Liupp, J. (1991). Intracellular transport and localisation of major histocompatibility complex class II molecules and associated invariant chain. J. Cell Biol. 115, 1213-1223. Randall, L. L., Hardy, S. J. S., &Thom, J. R. (1987). Export of protein: a biochemical view. Annu. Rev. Microbiol. 41,507-541. Robinson, A., Kaderbhai, M. A., & Austen, B. M. (1987). Identification of signal sequence binding proteins integrated into the rough endoplasmic reticulum membrane. Biochem. J. 242, 767-777. Rothman, J. E., & Orci, L. (1992). Molecular dissection of the secretory pathway. Nature 355,409--415. Sakaguchi, M., Tomiyoshi, R., Kuroiwa, T., & Mihara, K. (1992). Functions of signal and signal-anchor sequences are determined by the balance between the hydrophobic segment and the N-terminal charge. Proc. Natl. Acad. Sci. USA 89, 16-19.

The Biosynthesis of Membrane Proteins

135

Sato, T., Sakaguchi, M., Mihara, K., & Omura, Y. T. (1990). The amino-terminal structures that determine topological orientation of cytochrome P-450 in microsomal membranes. EMBO J. 9, 2391-2397. Schatz, G. (1993). The protein import machinery of mitochondria. Prot. Sci. 2, 141-146. Shelness, G. S., Lin, L., & Nicchitta, C. V. (1993). Membrane topology and biogenesis of eukaryotic signal peptidase. J. Biol. Chem. 268, 5201-5208. Shin, J., Dunbrack, R. L., Lee, S., & Strominger, J. L. (1991). Signals for retention of transmembrane proteins in the endoplasmic reticulum: studies with CD4 translocation mutants. Proc. Nail. Acad. Sci. USA 88, 1918-1922. Simon, M., & Blobel, G. (1991). A protein-conducting channel in the endoplasmic reticulum. Cell 65, 371-380. S/511ner, T., Whiteheart, S. W., Brunner, M., Erdjument-Bromage, H., Geromanca, S., Tempst, P., & Rothman, J. E. (1993). SNAP receptors implicated in vesicle targeting and fusion. Nature 362, 318-355. Stamnes, M. A., Shieh, B.-H., Chuman, G. L., Harris, G. L., & Zuker, C. S. (1991). The cyclophilin homologue ninaA is a tissue specific integral membrane required for the proper synthesis of a subset of Drosophilia rhodopsins. Cell 65, 219-227. Stifling, C., Rothblatt, J., Hosobuchi, M., Deshaies, R., & Schekman, R. (1992). Protein translocation mutants defective in the insertion of integral membrane proteins in the ER. Mol. Biol. Cell 3, 129-142. Stirzaker, S. C., Poncet, D., & Both, G. W. (1990). Sequences in rotavirus glycoprotein VP7 that mediate delayed translocation and retention of the protein in the ER. J. Cell Biol. 111, 1343-1350. von Heijne, G. (1981). Models for transmembrane translocation of proteins. Biochem. Soc. Symp. 46, 259-273. von Heijne, G. (1986). The distribution of positively charged residues in bacterial inner membrane protein correlates with the trans-membrane topology. EMBO J. 5, 3021-3027. Wada, I., Rindress, D., Cameron, D. A., Oy, W. J., Doherty, J. J., Louvard, D., Bell, A. W., Dignard, D., Thomas, D. Y., & Bergeron, J. J. M. (1991). SSRo~ and associated calnexin are major calcium binding proteins of the endoplasmic reticulum membrane. J. Biol. Chem. 266, 19599-19610. Walter, P., & Blobel, G. (1981). Translocation of proteins across the ER. III. Signal recognition particle (SRP) causes signal-sequence dependant and site-specific arrest of chain elongation that is released by microsomal membranes. J. Cell Biol. 91,557-561. Wessels, H. P., & Spiess, M. (1988). Insertion of a multispanning membrane protein occurs sequentially and requires only one signal sequence. Cell 55, 61-70. Wickner, W., Driessen, A. J. M., & Hartl, E-U. (1991). Enzymology of protein translocation. Annu. Rev. Biochem. 60, 101-124. Wong, S. H., Low, S. H., & Hong, W. J. (1992). The 17-residue transmembrane domain of beta-galactoside alpha2,6-sialyltransferase is sufficient for Golgi retention. Cell. Biol. 117, 245-258. Zopf, D., Bernstein, H. D., Johnson, A. E., & Walter, P. (1990). The methionine-rich domain of the 54 kd protein subunit of SRP contains an RNA binding site and can be cross-linked to a signal sequence. EMBO J. 9, 4511-4517.

This Page Intentionally Left Blank

SPECIFICITY OF LIPID-PROTEIN INTERACTIONS

Derek Marsh

I. II.

III. IV.

V.

VI.

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Specific Lipid-Requiring Membrane Enzymes . . . . . . . . . . . . . . . . . A. Phospholipase A2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B. Protein Kinase C . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . C. 13-Hydroxybutyrate Dehydrogenase . . . . . . . . . . . . . . . . . . . . Thermodynamics and Dynamics of Lipid Selectivity . . . . . . . . . . . . . . Acyl-Chain Dependence of Lipid Selectivity . . . . . . . . . . . . . . . . . . A. Chain Length Dependence of Lipid-Protein Association . . . . . . . . . B. Hydrophobic Matching of Lipids with Integral Proteins . . . . . . . . . . Headgroup Dependence of Lipid Selectivity with Integral Proteins . . . . . . A. Dependence of Selectivity on Lipid Class . . . . . . . . . . . . . . . . . B. Ionic Strength Dependence of Selectivity . . . . . . . . . . . . . . . . . C. pH Dependence of Selectivity . . . . . . . . . . . . . . . . . . . . . . . D. Exchange Rates of Protein-Associated Lipids . . . . . . . . . . . . . . . E. Gel Phase Lipids and Selectivity . . . . . . . . . . . . . . . . . . . . . . Integral Protein Structure and Lipid Selectivity . . . . . . . . . . . . . . . . . A. Myelin Proteolipid Protein . . . . . . . . . . . . . . . . . . . . . . . . . B. M 13 Bacteriophage Coat Protein and IsK Protein . . . . . . . . . . . . .

Biomembranes Volume 1, pages 137-186. Copyright 9 1995 by JAI Press Inc. All rights of reproduction in any form reserved. ISBN: 1-55938-658-4

137

138 139 139 142 144 146 148 149 152 153 154 158 160 162 164 165 165 167

138

DEREK MARSH

C. Mitochondrial ADP/ATPTranslocator . . . . . . . . . . . . . . . . . . . D. Cytochrome c Oxidase . . . . . . . . . . . . . . . . . . . . . . . . . . . VII. Lipid Selectivity of Peripheral Proteins . . . . . . . . . . . . . . . . . . . . A. Influence of Protein Charge . . . . . . . . . . . . . . . . . . . . . . . . B. Lipid Headgroup Selectivity . . . . . . . . . . . . . . . . . . . . . . . . C. Mixed Lipid Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . D. Specific Lipid Binding and Covalently Linked Chains . . . . . . . . . . VIII. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

170 171 172 172

174 177 179 180 180

!. I N T R O D U C T I O N Lipid-protein interactions are ubiquitous in biological membranes" nearly all membrane proteins are exposed to lipid. In the case of integral proteins, interactions take place both with the lipid headgroups and with the fatty acyl-chains. Hydrophobic interactions with the latter are responsible for sealing the protein into the membrane and not only preserve the integrity of the membrane permeability barrier, but also interface the protein with the fluid lipid milieu that gives rise to the dynamic features that, in many cases, are essential for protein function. Polar interactions of integral proteins with the lipid headgroups also contribute to the stability of the protein-lipid membrane assembly, give rise to a selectivity of interaction with specific negatively charged lipids, and in certain instances, modulate or control the function of the protein. In the case of peripheral proteins, electrostatic interactions with lipid headgroups are responsible for the binding of the protein to the membrane which generally gives an absolute requirement for negatively charged lipids and, in addition, other lipids are required for the activation of certain peripheral membrane-associated enzymes. In general, lipid-protein interactions can be divided into two broad types: the highly specific and the obligatory. The highly specific class are those exhibiting an absolute requirement for a particular lipid for activity, either as a substrate in the case of the phospholipases and other lipid-metabolizing enzymes, or as an activator as in the case, for example, of protein kinase C, or ]3-hydroxybutyrate dehydrogenase. The obligatory class are those involving the classical membrane proteins, where the lipid/protein interaction is essential for the structural and functional integrity of intrinsic membrane proteins, or for the binding and ultimately, the function of peripheral proteins. In the first case, the lipid selectivity is obvious and different examples will be dealt with individually. In the second case, the lipid selectivities may be more subtle, although in certain cases also obligatory, and these will be treated generally with distinction being made where necessary between integral and peripheral proteins.

L ipid-Protein Specificity

139

il. SPECIFIC LIPID-REQUIRING MEMBRANE ENZYMES Three examples of membrane enzymes that have an absolute requirement for a specific type of lipid are discussed first. These are all considered to be peripheral membrane proteins. Nevertheless, they illustrate various different features of the specificity of lipid-protein interactions that may be relevant, to a greater or a lesser extent, to lipid interactions with membrane proteins in general. Of particular interest are the X-ray crystal structures that define the types of contact and interactions that different parts of the lipid molecule may have with various amino acid side-chains in the protein.

A. PhospholipaseA2 Phospholipase A2 (PLA2) is an enzyme that catalyzes the hydrolysis of the 2-acyl ester bonds of glycerophospholipids with a wide variety of polar headgroups (X) producing a lysophospholipid and free fatty acid: PLA2/Ca2+ R1COOCH2.CH(OOCR2)-CH2OP(O2)OX + H20 ---> R1COOCHz-CH(OH)-CH2OP(O2)OX + RzCOOH The enzyme uniquely requires a lipid-water interface for complete activation and Ca 2+ is an obligatory cofactor. Although soluble phospholipases A2 have a broad specificity for their phospholipid substrates, in contrast to general membrane proteins, they contain a highly specific binding site for a single phospholipid molecule and have an absolute catalytic stereospecificity for sn-3 phospholipids with an ester bond at the sn-2 position. This class of enzymes, therefore, provides an example of both a rather specific lipid-protein interaction and a particular protein-membrane surface association. The nature of the specific phospholipid interaction and the likely mode of binding of the protein at the membrane interface has been deduced from the crystal structure of the PLA2 enzyme complex with a phosphonate-containing analog of phosphatidylethanolamine, a specific inhibitor designed to be a transition state analog (Scott et al., 1990a; White et al., 1990). In the 2-/k resolution structures of PLA2 from the evolutionarily divergent Class I/II and Class III families, the inhibitor is bound with the sn-1 and sn-2 chains approximately parallel, in a similar configuration to that in lipid membranes (see Figure 1). The sn-2 chain is bent tetrahedrally; in the case of the inhibitor, this bend immediately follows the phosphonate group which emulates the transition-state ester configuration and corresponds to the C-2 atom of the sn-2 chain in the normal bilayer configuration (see Figure 2A). Unlike the situation in phospholipid bilayer aggregates, however, the lipid headgroup is not folded back over the sn-2 ester region, but forms specific polar interactions with the enzyme and its Ca2+-cofactor.

16 A

,~(~,, 124

34 ~

4

~

~

79

3 109 49

Figure I. X-ray crystal structures (oc-carbon backbone) of (A) cobra-venom, and (B) bee-venom phospholipases A2 of the Class I/II and Class III families, respectively, complexed with the phospholipid transition-state analog (heavy lines), 1-O-octyl-2heptyl-phosphonyl-sn-glycero-3-phosphoethanolamine. (Scott et al., 1990b).

A,

B,

Figure 2. (A) Conformation of the inhibitor phospholipid, 1-O-octyl-2-heptyl-phosphonyl-sn-glycero-3-phosphoethanolamine, when complexed to phospholipase A2 (Scott et al., 1990b). (B) Schematic view looking down the hydrophobic channel in Class I/II cobra-venom phospholipase A2 showing the residues forming the channel, and those at the opening of the channel that are presumed to be part of the interracial binding site (squares). The lipid chains (sn-1 and sn-2) are indicated by filled circles. (White et al., 1990). 140

Lipid-Protein Specificity

141

Of the polar groups of the inhibitor that are common to normal phospholipids and are not involved directly in the catalytic mechanism, the amino group (of phosphatidylethanolamine) forms a hydrogen bond with a poorly conserved surface asparagine, Asn 53, in the Class I/II enzyme, and a similar interaction occurs with an unrefined portion of the carbohydrate in the Class III enzyme. A similar hydrogen bonding capacity is also present in phosphatidylserines, but not in phosphatidylcholines, and therefore may generally make some contribution to the headgroup selectivity of lipid-protein interactions. The formal oxy-anion of the sn-3 phosphate group provides an axial ligand for the CaZ+-cofactor which is further hepta-coordinated by the carboxylate of the nearly invariant Asp 49 (Asp 35 in Class III) and the backbone carbonyls of residues 28, 30, and 32 (residues 8, 10, and 12 in Class III). It is likely that this CaZ+-mediated interaction with the lipid phosphate is a specific feature of PLA2, and possibly of other CaZ+-dependent lipid-protein associations. In addition, Tyr 69 forms a hydrogen bond distally with the sn-3 phosphate group in the cobra venom enzyme. In other Class I/II enzymes, residue 69 is alternatively a lysine whose ~-amino group could equally well bind to the phospholipid phosphate group. In the Class III enzyme, this role is fulfilled by a hydrogen bond between the guanidino group of Arg57 and the sn-3 phosphate. These latter types of polar group interaction could conceivably be of more general relevance to associations between phospholipids and membrane proteins. The chains of the lipid inhibitor lie in a hydrophobic channel extending approximately 14 .~ (9/k for the Class III enzyme) from the catalytic site (His 48-N81; His 34 in Class III) to the enzyme surface. The faces of this channel are made up mainly of invariant or highly conserved hydrophobic residues (see Figure 2B). In the Class I/II enzyme, one side is formed by the invariant Ile 9, together with the invariant Phe 5 at the channel floor, and both of these residues are in contact with the sn-2 chain. The other side consists of the nearly invariant Leu 2 (occasionally a Val) in contact with the sn-1 chain and a potentially mobile hydrophobic flap provided by the ring of Tyr 69, or otherwise, the methylene groups of Lys 69. In the inhibitor complex, this flap is anchored by the interaction of the functional group of residue 69 with the sn-3 phosphate group. Finally, the roof of the channel is formed by Trp 19 which contacts both the sn-1 and sn-2 chains; this position is occupied by a hydrophobic residue in 85% of the known sequences. In the Class III enzyme, the proximal methylene groups of Arg 57 form the potentially mobile flap, and the interactions of the lipid chains with the hydrophobic channel appear to be effectively identical to those of the Class I/II enzyme. The hydrophobic channel is shorter than are the chains of normal membrane phospholipids. Therefore, it is thought that the remainder of the lipid chains remain anchored within the lipid bilayer. Diffusion of the phospholipid into the active site is facilitated by the essentially hydrophobic and dehydrated nature of the protein channel which must be in close contact at its mouth with the membrane surface. The interfacial contact site is assumed to be composed of those residues that are close to the channel opening. In the Class I/II enzymes, this includes the exposed

142

DEREK MARSH

residues in the first two turns of the N-terminal helix, Tyr(Trp) 3 and Lys 6, the usually hydrophobic residue at position 19, as well as Arg 31 and possibly the highly conserved Tyr 75 (see Figure 2B). In the Class III enzyme, the third helix (residues 76 to 90) is analogous to the N-terminal helix of the Class I/II enzymes and will contribute residues to the interfacial binding site. In the latter case, the residues contributing to the channel, to the interfacial site, or to both are: Ile 1, Tyr 3, Cys 9, His 11, Thr 56, Arg 57, Leu 59, Val 83, Met 86, Tyr 87 and Ile 91 (Scott et al., 1990a). Thus, although because of its function PLA2 differs from normal membrane proteins in the structure of the complex with the specific phospholipid substrate, the high resolution structures of the complex do illustrate, in considerable detail, some of the side-chain interactions, both polar and hydrophobic, as well as interfacial, that may generally occur between lipids and membrane proteins. B. Protein Kinase C Protein kinase C (PKC) is a Ca 2§ and lipid-dependent regulatory enzyme which transduces signals from second messengers that result from the hydrolysis of phosphatidylinositol bisphosphate (PIP2) or other phospholipids. During activation, the enzyme is recruited from the cytoplasm to the membrane surface. A mixed-micelle assay has been used to determine the specificities and stoichiometries of the lipid cofactors (Hannun et al., 1985). Activation uniquely requires one molecule of diacylglycerol (or the functionally equivalent phorbol ester). In addition, PKC requires four molecules of the phospholipid, phosphatidylserine, for activation, and optimal activity is obtained with ca. ten molecules of phosphatidylserine. The importance of a lipid-water interface for the activation is demonstrated by the abrupt increase in activity at the critical micelle concentration and the lack of dependence on micelle concentration for fixed micelle composition in the mixed-micellar assay (Hannun et al., 1986). The activation by diacylglycerol is specific to the naturally occurring sn-l,2 stereoisomer; neither the sn-2,3 enantiomer nor the sn-1,3 isomer are effective (Boni and Rando, 1985). Both carbonyl groups of the stereospecific oxygen esters and the 3-hydroxyl moiety of diacylglycerol are necessary for maximal activity (Ganong et al., 1986). Structural alignment with the diacylglycerols suggests that the hydroxyl group at C-9 fulfills this third requirement in the phorbol diesters. On the other hand, it appears that the ester hydrocarbon chains are relatively unimportant in the specificity of the diacylglycerols (Boni and Rando, 1985; Hannun et al., 1986). It has been found from in vitro experiments that PKC binds to negatively charged phospholipids in a Ca2+-dependent manner (Bazzi and Nelsestuen, 1987). Although binding takes place to a variety of anionic phospholipids, only phosphatidylserine is capable of fully activating the enzyme (Hannun et al., 1986). The nonactivating binding is of a classical electrostatic origin, and is dependent on the membrane

L ipid-Protein Specificity

143

Table 1. Phospholipid-Mediated Activation of Protein Kinase C a'b Headgroup

10 mol %

15 mol %

-O-P(O2)-O-CH2-CH(COO-)-NH3 +

37.4

49.0

-O-P(O2)-O-CH2-CH(NH3+)-COO-

17.2

31.5

-O-P(O2)-O-CH2-CH2-CH(COO-)-NH3 +

6.0

19.2

-O-P(O2)-O-CH2-CH(COOCH3)-NH3+

3.6

10.6

-O-P(02)--O-CH2-CH(COO-)-NHCOCH3

1.3

0.7

-O-P(02)-O--CH2-CH(CH20H)-NH3+

0

0 4.9

-O-P(O2)-O-CH2-CH2-CH2OH

4.8

-O-P(O2)-O-CH2-CH2-CH2-COO-

4.0

3.2

-O-P(02)-O-CH2-CH2-COO-

7.3

7.3

-O-P(O2)-O--CH2-COO-

0.6

0

-O-P(O2)-O-CH2-CH3

2.4

1.6

-O-P(O2)-O-CH3

3.5

4.0

Notes: aSource: Lee and Bell (1989). bprotein kinase C activity (pmol/min) in the presence of 1,2-dioleoyl-sn-glycero-3phospho-L-serine or various headgroup analogs, at 10 and 15 mol % in mixed micelles, and of 100 nM phorbol 12,13-dibutyrate. A similar activation pattern by these phospholipids was observed with 2 mol % diolein in the mixed micelles, instead of the phorbol ester cofactor.

surface charge and on the ionic strength, but is unaffected by the presence of diacylglycerol (Orr and Newton, 1992a). The binding and activation by phosphatidylserine, on the other hand, is specifically enhanced by diacylglycerol (or phorbol ester), but is relatively insensitive to the membrane surface charge and to ionic strength. Unlike the normal hyperbolic kinetics displayed for activation by diacylglycerol, a high degree of cooperativity is observed in both the binding and activation by phosphatidylserine, with Hill coefficients in the region of 4-6 (Hannun et al., 1986; Orr and Newton, 1992b). Half-maximal binding, (X0.5) in a mixed micellar assay was found at 4 mol % phosphatidylserine compared with 7 mol % phosphatidic acid, with Hill coefficients nH of 6.2 + 0.7 and 2.9 + 0.5, respectively, for the two lipids (Orr and Newton, 1992b). This corresponds to a difference in free energy of association (AG -- nHRTlnX0.5) of approximately -7.4 kcal/mol for the specific binding with phosphatidylserine compared to the nonspecific binding with phosphatidic acid. Several features of the phosphatidylserine molecule are necessary for its specificity in activation (Lee and Bell, 1989). These can be seen from the enzyme activities given in Table 1 for phosphatidylserine or phospholipids with other headgroup variants present in the micellar assay. Both the carboxyl and amino groups are required, and the stereospecificity of the L-serine as well as the distance between the carboxyl and amino groups also are important. In addition, the interfacial conformation of the polar headgroup is critical, because neither lysophosphatidylserine nor the 2-acetyl derivative support activity. However, the

144

DEREK MARSH

stereochemistry of the glycerol backbone is not an important factor; both 1,2-racphosphatidylserine and 1,3-phosphatidylserine fully support activity. The PKC enzyme, therefore, presents an example of a highly specific binding with rather stringent requirements for the lipid structure. Presumably, there is a unique binding site for diacylglycerol, with allosteric modulation of the cooperative interaction with phosphatidylserine. Calcium also plays an essential role in the lipid-protein interaction, as is also the case with PLA2. The lipid interaction also appears necessary for the specific binding of the protein to the membrane on activation. Other examples of membrane association accompanying activation are afforded by the peripheral enzymes, ~-hydroxybutyrate dehydrogenase and pyruvate oxidase. In the latter case, attachment to the membrane is induced by the binding of substrate (Blake et al., 1978).

C. 13-HydroxybutyrateDehydrogenase The lipid-requiting enzyme, D-13-hydroxybutyrate dehydrogenase, is a "peripheral" membrane protein that is bound tightly at the inner face of the mitochondrial inner membrane. It may be purified as a soluble apoprotein devoid of lipid, in which state it is inactive, but will insert spontaneously into phospholipid vesicles. Although the enzyme will bind to a variety of phospholipids, of the normal phospholipids, it has an absolute requirement for phosphatidylcholine for activity (Isaacson et al., 1979). The two substrates of the enzyme, D(-)-3-hydroxybutyrate and NAD § are water soluble, and the binding of phosphatidylcholine confers on the enzyme the ability to bind the coenzyme, NADH (Gazzotti et al., 1974). Activation of the enzyme by phosphatidylcholine is of a cooperative allosteric nature with a Hill coefficient of-~ 2.4, indicating interactions between at least three phosphatidylcholine binding sites (Sandermann et al., 1986). The enzyme is thought to be present in the membrane as a tetramer, thus it is likely that there is one phosphatidylcholine activation site per monomer. The specific structural features of the phospholipid molecule that are necessary for efficient activating capacity have been determined by the use of various synthetic phosphatidylcholines and their analogs. The hydrophobic portion of phosphatidylcholine is not of crucial importance for the specificity of activation of the enzyme, although it is essential for binding. Whereas dioleoyl-phosphatidylcholine supports the highest activity of the symmetrical diacyl-phosphatidylcholines alone, optimal activity is obtained by phospholipids of single saturated mixed-chain species in appropriate mixtures, irrespective of the precise chain composition, provided that the lipid bilayer is in the fluid phase (Churchill et al., 1983). The glycerol backbone of phosphatidylcholine is also not an important feature of the activating species as: 1) D- and L-stereoisomers activate equally well, 2) phosphatidylcholines with ether-linked rather than ester-linked chains activate, although less efficiently, 3) octadecyleicosyl-phosphorylcholine, a branched alkyl

Lipid-Protein Specificity

145

Table 2. Activation of D-13-HydroxybutyrateDehydrogenase ([3DH) with

Phosphatidylcholine (PC) Analogs of Different Headgroup Structuresa'b (PC/~iDH)o.5 Headgroup

phosphoryl: -O-P(02)-O-CH2-CH2-N +(CH3)3 -O-P(O2)-O-CH2-CH2-CH2-N +(CH3)3 -O-P(O2)-O-CH2-CH2-CH2-CH2-N +(CH3)3 -O-P(O2)-O-CH2(CH2)-CH2-N +(CH3)3 -O-P(O2)-O-CH2-CH2-N +(C2HS)(CH3)2 -O-P(02)-O-CH2-CH2-N +(C2H5)3 phosphonyl: -P(02)-O-CH2-CH2-N +(CH3)3 -CH2-P(O2)-O-CH2-CH2-N +(CH3)3 phosphinyl: -P(O2)-CH2-CH2-N +(CH3)3

Act max (%)

(mol/mol)

72 70 45 0 64 0

5.2 4.8 4.0

42 34

3.8 3.7

3.9 m

Notes: aSource:Isaacson et al. (1979). bThemaximum activity (Act max)is expressed relative to mitochondrial phospholipid, and (PC/~DH)0.5 is the mole ratio of the phosphatidylcholine analog to I]DH giving half-maximal activity. The phosphatidylcholine analogs were codispersed with mitochondrial phosphatidylethanolamine and cardiolipin at a mole ratio of 1:0.8:0.4. The phosphoryl analogs are 2-octadecyl-eicosyl derivatives, the phosphonyl analogs are 2,3dimyristoyloxypropyl, and 3,4-dimyristoyloxybutyl derivatives, and the phosphinyl analog is a 2,3distearoyloxypropylderivative.

phosphorylcholine in which the glycerol moiety is absent, activates to the same extent as does mitochondrial phosphatidylcholine (Isaacson et al., 1979). Only certain features of the phosphorylcholine lipid headgroup are essential for efficient activating capacity (see Table 2) as: 1) when the --CH2CH2- group linking the phosphate and trimethyl ammonium groups is increased by either one or two further methylene units, activation is not greatly altered, 2) replacement of this link by an isopropyl group, however, abolishes activation, 3) whereas phosphatidylethanolamine cannot activate, the N-ethyl-N,N-dimethyl analog can, but the N,N,N-triethyl analog cannot, 4) the enzyme is activated by phosphono-analogs of phosphatidylcholine in which the glycerol to phosphorus oxygen is either missing or is replaced isosterically by a methylene group, and 5) the phosphinate-analog of phosphatidylcholine, in which both oxygens of the phosphate diester links are missing, does not support activity, however (Isaacson et al., 1979). Although a three-dimensional structure of the enzyme is currently not available, these results constitute a detailed mapping of the structural requirements of the lipid headgroup binding site. The above examples of particular highly specific lipid-protein interactions give some idea of the ways in which selective associations may take place between lipids and membrane proteins in general. In the following sections, these more global features of the obligatory lipid-protein associations in membrane structures are addressed, starting first with overall thermodynamic considerations.

146

DEREKMARSH !11. T H E R M O D Y N A M I C S A N D D Y N A M I C S OF LIPID SELECTIVITY

The selectivity of interaction of lipids with membrane proteins can be characterized thermodynamically by the relative association constants of the different lipids with the protein. An exchange equilibrium is established between lipids (L and L*) competing for sites (P) on the protein: -1 '~b

PL + L* ,-----" PL* + L

(1)

-1 '~f

The relative association constant for lipid L* with respect to lipid L can then be defined by: Kr = [L*]b.[L]f/[L]b.[L*]f

(2)

where the subscripts b and f correspond to lipids associated with the protein and free in the bilayer, respectively. Square brackets indicate concentrations, and hence Kr is the relative association constant that is normally determined experimentally when activity coefficients are taken as unity. The relative association constant, K~r, in which activity coefficients are taken into account explicitly is given by: K~r= Kr(~*~tf/~tf* )

(3)

where Yi are the activity coefficients of the different species (defined in terms of mole fractions). This latter form is particularly useful in analyzing effects of the composition of the aqueous environment (e.g., ionic strength) on the selectivity of the lipid-protein interaction. Experimental determinations of the selectivity of lipid association with membrane proteins can be analyzed by a multi-site binding model. The only difference for membrane-associated proteins compared with soluble proteins is that all lipid sites on the protein are occupied (see Figure 3). The appropriate equation for equilibrium exchange association is (Brotherus et al., 1981)" Nb =f{nt* + n t / [ f + Kr(1 - f ) ] }

(4)

where f is the fraction of the lipid L* that is associated with the protein, nt* and nt are the total numbers oflipids L* and L, respectively, per protein, and Nb is the total number of lipid association sites on the protein. The relative association constant is then the average value taken over all sites, i.e.,/QrV= Z niKi/Nb. For the case that the lipid L* is present only in probe amounts (for instance in experiments with labeled lipids), considerable simplifications ensue because then nt* = O, yielding the following relation for the fractionfof lipid L* that is associated with the protein: (1 - f ) / f = (nt/Nb - 1)/Kr

(5)

Lipid-Protein Specificity

147

np

Nb Figure 3. Schematic indication of the association of bilayer lipids with integral and peripheral membrane proteins. The integral protein (P0 is surrounded by a first shell of Nb lipids, the chains of which interact with the hydrophobic surface of the protein, and the headgroups of which interact with the protein residues located at the polar-apolar interface of the membrane. The peripheral protein (Pp) is bound at the polar surface of the lipid bilayer and bears np specific sites (+) at which the lipid headgroups can associate. The exchange equilibrium for lipid association with the protein is described by Eq. 4 (where Nb -n p in the case of peripheral proteins) and is characterized by an association constant (Kr) relative to the background lipid. (Marsh, 1990).

This is the form used routinely in analyzing electron spin resonance (ESR) experiments with spin-labeled lipids interacting with integral proteins (Marsh, 1985). The thermodynamics of the lipid-protein selectivity have inevitable consequences on the dynamics of lipid exchange at the association sites on the protein. At equilibrium, the exchange rate of lipids off the protein (xgl) and that of lipids onto the protein (7:?1) are related by mass balance (cf. Eq. 1): fl~b 1 "-(l - f ) ' r , ? 1

(6)

Hence from Eq. 5, the ratio of the off- and on-rates for lipid exchange is given by: "r,-bI/T,-( 1 = ( n t / N b - 1 ) / K r

(7)

The on-rate can be assumed to be the same for all lipids because it is diffusion controlled: it is determined by the overall relative lipid concentration, nt/Nb. Therefore, the intrinsic off-rates for exchange of different lipids are inversely proportional to their relative association constants, i.e., for lipids A and B:

148

DEREK MARSH

,gl (A) /'r,~l(B ) = Kr(B ) /Kr( A)

(8)

As must be the case, lipids with higher specificity have a longer residence time, *b, on the protein. An interesting case of lipid selectivity arises for membrane enzymes which can exist in two (or more) conformations, E1 or E2. The lipid L* may then have different relative association constants, Kr,1 and Kr,2, for the El and E2 conformations, respectively. The conformational equilibrium for the protein in an environment composed of the lipid L* then will be shifted relative to that in the reference lipid, L. The lipid association and conformational equilibria are expressed by the scheme: K . , E1 + L* ~

L*'EI

E2 + L* ~

L*.E2 Kr,2

where K is the conformational equilibrium constant (El/E2) in the reference lipid, and K* is that in the lipid L*. From the cyclic nature of the equilibria, the shift in conformational equilibrium is given by" K* = g'(Kr,1/Kr,2)

(9)

Correspondingly, the relative association constant measured at low concentrations of the probe lipid (L*) is: Kr = (Kr,2 + K'Kr,1)/( I + K)

(1 O)

which expresses the relative association constant measured (i.e., Kr) in terms of the intrinsic relative association constants, Kr,~ and Kr,2 for the two conformations, and the conformational equilibrium constant in the background lipid. An example of such a system (the Ca2+-ATPase) is given later.

IV. ACYL-CHAIN DEPENDENCE OF LIPID SELECTIVITY The association of the lipid chains with proteins is determined to a large extent by hydrophobic interactions. For integral membrane proteins, this will have a more subtle effect on the selectivity of interaction compared to that for soluble proteins because in the former case, the lipid chains are essentially in a hydrophobic environment both at the protein interface and in the lipid bilayer regions of the membrane.

Lipid-Protein Specificity

149

A. Chain Length Dependence of Lipid-Protein Association The case of soluble lipid-binding proteins is considered first. Pyruvate oxidase is a peripheral membrane enzyme from E. coli that is water soluble, but is greatly activated by lipids and other amphiphiles. Association with the membrane or with lipids is induced by a conformational change on binding the substrate that exposes a hydrophobic region for interaction with lipid. The endogenous electron acceptor (ubiquinone) is membrane bound. The dependence of the concentration (X0.5) at which half-maximal activation is achieved on the chain-length of the activating lipid is given in Figure 4. In all cases shown, the lipid is in the monomeric form (i.e., is below the critical micelle concentration) in water. There is a very clear selectivity for the interaction and activation depending on the lipid chain-length, which is given to a good approximation by (see Figure 4)"

11 9

~9

v

7 tr

c5

X [-~ 5

I I

4

u

I

6

J

I

8

I

1

10

i

1

12

i

I

14

u

I

16

t

18

ncH ( C - a t o m s ) Figure 4. Chain length dependence of the mole fraction of lipid monomer in water required for half maximal activation of pyruvate oxidase by fatty acids (o), lysophosphatidylcholines (A), and diacyl-phosphatidylcholines (m). All lipids have straight, saturated chains, and nCH is the number of aliphatic carbon atoms in one chain. Straight lines are linear regressions. (Data from Blake et al., 1978). The unitary free energies of association of diacyl-phosphatidylcholines with [3-hydroxybutyrate dehydrogenase (O) from Cortese et al. (1982) are also given.

1 50

DEREK MARSH

1.0 9 ~ 0.5-

,-~ o.o!

!

I

t

-0.5

I-1.0

-

-

C~ -1.5 -

-2.0

12

, 13

i 14

I 15

n

I 16

, 17

~

I 18

I 19

(C-atoms)

Figure 5. Apparent free energy of association, RTIn(f/1-f), of a fixed spin-labeled phosphatidyicholine with rhodopsin (I) in recombinants with saturated diacylphosphatidylcholines of different chain lengths at a constant lipid/protein ratio, where n is the total number of aliphatic carbon atoms in one chain. All systems are in the fluid lipid phase. The zero level corresponds to Kr = 1 and Nb = 22 in Eq. 5 (Ryba and Marsh, 1992). The free energies of association of 9-trans-monounsaturated phosphatidylcholines with the Ca2+-ATPase (O), where the zero level corresponds to the association with dioleoyl-phosphatidylcholine, are also given (Caffrey and Feigenson, 1981). The lines are linear regressions.

-R TlnX0.5 = AGI.nCH + AGo

(11)

where AGI and AGo are the gradient and intercept, respectively, of the plot in Figure 4, riCH is the number of aliphatic carbon atoms in the chain, and X0.5 is in mole fraction units. The gradient of the free energy of association (AG =-RTInX0.5) is approximately 650 cal/mol/CH2 for the single-chain lipids and 500 cal/mol/CH2 per chain for the double-chain lipids. These values are comparable to the free energy of transfer from an aqueous to a hydrophobic environment (see e.g., Cevc and Marsh, 1987), as indicated for instance, by the chain length dependence of the critical micelle concentrations of the lipids which is very similar to that for activation (Blake et al., 1978). This indicates that activation occurs primarily by binding of the lipid chains to hydrophobic regions that are exposed on the protein

Lipid-Protein Specificity

1 51

accompanying the substrate-induced conformational change. A similar result has also been obtained for the activation of 13-hydroxybutyrate dehydrogenase by short-chain diacyl-phosphatidylcholines (Cortese et al., 1982). The unitary free energies of lipid monomer association with this protein are also given in Figure 4 for comparison. The incremental free energy of association in the case of 13-hydroxybutyrate dehydrogenase is approximately 600 cal/mol/CH2 per chain. The above results suggest that the hydrophobic interactions of lipid chains with apolar regions on the protein are rather similar to those with the lipid chains themselves. Thus for membrane bound proteins, particularly those of the integral type, the effects of lipid chain length on the selectivity of association are likely to come mostly from a limited modulation of the overall interaction. Lipid selectivity based on chain composition is, therefore, likely to be only modest for membrane bound proteins, at least over the range of chain lengths that can reasonably match the hydrophobic span of the protein. Results obtained on the association of spin-labeled phosphatidylcholine with the integral protein rhodopsin when reconstituted in bilayers composed of phosphatidylcholines of different chain lengths are given in Figure 5. In this case, the chain composition of the spin-labeled reporter lipid is held constant. Therefore, any chain length dependent selectivity in the effective relative association constant arises from the different chain lengths of the background host lipid, because this defines the standard state to which the relative association constants are referred. The chain length dependence of the apparent free energy of lipid association with rhodopsin (given in Figure 5) is consistent with Eq. 11, but the gradient with respect to chain length is approximately 80 cal/mol/CH2 per chain, which is much smaller than the total free energy of association for lipids binding to proteins in aqueous solution that was considered above. This demonstrates that, over the range given in Figure 5, the chain length selectivity for association ofphosphatidylcholine with rhodopsin arises from limited differences in the energetics of interaction between the lipid chains and the intramembranous section of the protein and those of interaction of the membrane lipid chains with themselves. Similar to the case of rhodopsin, it has been found that the association constants of phosphatidylcholines with the Ca2+-ATPase do not depend greatly on chain composition when assayed by competition from the fluorescence yield of the protein in mixtures with a quenching lipid (Caffrey and Feigenson, 1981; Froud et al., 1986). The relative free energy of association for those lipids which show the largest effects are included for comparison with rhodopsin in Figure 5. The results with trans-monounsaturated phosphatidylcholines have a gradient with chain length over the range C(14: It) to C(18: It) of approximately 40 cal/mol/CH2 per chain, which is smaller even than that found for saturated phosphatidylcholines with rhodopsin. An alternative approach to studying the selectivity of chain interactions with the protein is to compare lipids with different numbers of hydrocarbon chains. For cytochrome oxidase (Powell et al., 1987) and the Na+,K§ (Esmann et al.,

152

DEREK MARSH

1988a), it is found that there is little selectivity in interaction between spin-labeled cardiolipin and its monolyso-derivative, or between spin-labeled phosphatidylcholine and spin-labeled lyso-phosphatidylcholine in the case of cytochrome oxidase. Additionally, rhodopsin which exhibits little selectivity of interaction between diacyl-phospholipids with different headgroups (cf. below) also evidences little selectivity between spin-labeled lipids with different numbers of chains (Marsh et al., 1982). These results again suggest that the difference in the free energy of interaction of the lipid chains with each other and with the hydrophobic surface of the protein is not large.

B. Hydrophobic Matching of Lipids with Integral Proteins The limited effects of acyl-chain length on the association of phospholipids with integral proteins can be expected only for the range of chain lengths that matches reasonably well with the hydrophobic span of the transmembrane protein. The latter presumably is adapted to the average effective chain length of the natural membrane lipids. For rhodopsin in bilayers of phosphatidylcholines with chains either longer or shorter than those shown in Figure 5, the effects of chain length on the lipid-protein interaction are much more dramatic than for the chain length range C(13:0) to C(17:0). In these more extreme cases, the lipid chain is no longer able to match the apolar intramembranous stretch of the protein, and an aggregation or phase separation of the protein takes place (Ryba and Marsh, 1992). However, the lipid-protein interactions are very different at the two chain length extremes. For saturated C(18:0) chains, the protein separates from the lipid and there is essentially very little interaction between the two; simultaneously the protein bleaches and lipid-lipid interactions are favored over lipid-protein interactions. For C(12:0) chains, on the other hand, the protein has modified photochemical activity, but lipids are trapped in the protein aggregates. Probably, in this case, interactions of the spin-labeled lipid with the protein are favored over those with the lipids of C(12:0) chain length (Ryba and Marsh, 1992). A selectivity based on chain length is also observed in the ability of lipids to support the activity of integral membrane enzymes, as was indicated above for the photochemical activity of rhodopsin. For the Ca2+-ATPase from sarcoplasmic reticulum, an optimum chain length of C(18:1) to C(20:1) has been found for supporting activity of the enzyme reconstituted in di-monounsaturated phosphatidylcholines (Caffrey and Feigenson, 1981; Johannsson et al., 1981). It might be assumed that this chain length dependence reflects the optimum matching of the lipid chains to the hydrophobic stretch of the protein, as mentioned already for lipid-protein interactions with rhodopsin. Further support for the concept of hydrophobic matching comes from data on the lipid chain length-dependent aggregation of rhodopsin from rotational diffusion measurements (Kusumi and Hyde, 1982; Ryba and Marsh, 1992), and also from electron microscopy (Chen and Hubbell, 1973), in the latter case, data has also been obtained for bacteriorhodopsin

Lipid-Protein Specificity

153

(Lewis and Engelman, 1983). However, it is found in certain cases, that the protein can be incorporated in lipids with a relatively broad range of chain lengths about the optimum. Also, the overall lipid-protein interactions do not exhibit such a relatively sharp maximum in the chain length dependence as does the enzyme activity, particularly in the case of the Ca2+-ATPase. This suggests that the enzyme activity reflects more subtle variations in lipid-protein interactions during the enzymatic cycle than those registered by overall measurements in the resting state. The relative association constants with the Ca2+-ATPase in different functional conformations have been derived for phosphatidylcholines of different chain lengths on the basis of the analysis given by Eqs. 9 and 10 above (Froud et al., 1986). The Ca2+-ATPase enzyme cycles between the E1 and E2 conformations, where the former corresponds to high affinity binding of Ca 2§ at outward-facing sites, and the latter to Ca 2+ release from low-affinity, inward-facing sites. The association constant of C(14: lc) phosphatidylcholine relative to that of C(18: lc) phosphatidylcholine for the enzyme in the E1 state was deduced to be Kr,E, = 2.5, and that for the enzyme in the E2 state to be Kr,E2 = 0.2. This reflects the fact that the conformational equilibrium constant (El/E2), obtained from other measurements, is 13 times greater in the C(14:1c) lipid than in the C(18:1c) lipid, even though the association constant for the C(14: lc) lipid relative to that ofthe C(18: lc) is only Kr = 0.83 in the resting state. These results show that the lipid-protein interactions can affect the conformational state of the enzyme and that this probably is determined by different affinities of the lipid for the two conformations. The higher affinity of the C(18:1c) lipid relative to the C(14:1c) lipid for the E2 conformation could indicate that the forward translocating conformational step is accelerated in the C(18:1 c) lipid, hence accounting for the enhanced activity in this lipid. This analysis suggests that in mixed-lipid systems, a selectivity in lipidprotein interaction could be expressed for those lipids whose chain lengths optimally match the hydrophobic span of the protein, at least when the different conformations are taken into account.

V. HEADGROUP DEPENDENCE OF LIPID SELECTIVITY WITH INTEGRAL PROTEINS The selectivity of interaction with integral proteins for lipid species of different polar headgroup types has been studied extensively by ESR spectroscopy with spin-labeled lipids (see e.g., Marsh, 1985, 1987). The results have been analyzed on the basis of Eq. 5 to yield average association constants relative to the background host lipid in titrations of the total lipid/protein ratio, or to yield relative values between different spin-labeled lipids from measurements at fixed lipid/protein ratio. In the latter case, the association constants are usually normalized to that for spin-labeled phosphatidylcholine. It has been demonstrated in several cases with reconstituted systems that there is little or no selectivity of

154

DEREK MARSH

Table 3. Relative Association Constants (Kr) for the Interaction of Various Spin-Labeled Lipids with Different Integral Membrane Proteins a Protein

Ptd2Gro b

Myelin proteolipid d

PtdH

StH

PtdSer PtdGro PtdEtn PtdCho

Rej e

1.5 e

2.9 e

7.0 e

1.4 e

1.1 e

0.5 e

1.0

1

3.0 f

2.4 f

2.9 f

1.4 f

2.0 f

1.7 f

1.0

2

N a+, K+-ATPas e

3.8

1.5

1.7

1.7

0.9

0.9

1.0

3

Cytochrome c oxidase

5.4

1.9

--

1.0

1.0

1.0

1.0

4

3.8

4.3

4.1

2.4

0.8

w

1.0

5

n-AcCho receptor

D

2.7

4.1

0.7

D

1.1

1.0

6

1.1

1.5,2.5 g 2.2

--

w

1.0

7

M13 coat protein

4.2 h

4.2 h

2.3 h

2.1 h

1.6 h

0.9 h

1.0

8

1.6 i

1.2 i

1.2 i

1.1 i

1.0 i

1.0

9

1.0

1.0

1.0

1.0

1.0

1.0

10,11

ADP-ATP carrier

Rhodopsin

1.0

Notes: aAll values are referred to spin-labeled phosphatidylcholine. bptd2Gro, cardiolipin; PtdH, phosphatidic acid; StH, stearic acid; PtdSer, phosphatidylserine; PtdGro, phosphatidylglycerol; PtdEtn, phosphatidylethanolamine; PtdCho, phosphatidylcholine; n-AcCho, nicotinic acetylcholine. CReferences: 1. Brophy et al. (1984); 2. Sankaram et al. (1991); 3. Esmann et al. (1985); 4. Knowles et al. (1981); 5. Horv~ithet al. (1990a); 6. Ellena et al. (1983); 7. Raines and Miller (1993); 8. Datema et al. (1987); 9. Peelen et al. (1992); 10. Watts et al. (1979); 11. Marsh et al. (1982). aNatural mixture of the proteolipid protein and the DM-20 isoform. eln dimyristoyl-phosphatidylcholine. fin dimyristoyl-phosphatidylglycerol. s'Values for protonated and charged forms, respectively, in the absence of salt. A value of 3.4 was obtained for the charged form at high ionic strength. hl]-sheet form of the protein in dimyristoyl-phosphatidylcholine/phosphatidylglycerol(80:20 mol/mol). ia-helical (partly) form of the protein in dimyristoyl-phosphatidylcholine.

spin-labeled phosphatidylcholine relative to unlabeled phosphatidylcholine when this is used as the reconstituting lipid. In all cases, the chain composition of the spin-labeled diacyl-phospholipids is maintained constant. For the spin-labeled lipids with different numbers of chains, there is little evidence for an appreciable selectivity arising from this account (cf. above).

A. Dependence of Selectivity on Lipid Class The relative association constants for a variety of different lipid classes interacting with a range of different integral proteins are summarized in Table 3. A selectivity of interaction for the different lipid species is found with most of the integral proteins, rhodopsin and the Ca2+-ATPase (see Marsh, 1987) being exceptions. For the latter, a somewhat lower selectivity for phosphatidylethanolamine and a considerably lower selectivity for phosphatidylcholines in the gel phase (relative to the fluid phase) or for anionic phospholipids bound by Ca 2§ has been detected by a fluorescence quenching assay (London and Feigenson, 1981; East

Lipid-Protein Specificity

155

and Lee, 1982). In general, phosphatidylcholine exhibits one of the lowest selectivities (as already mentioned), with that for phosphatidylethanolamine sometimes being somewhat lower. Preferential selectivities are normally found for negatively charged lipids, an exception being phosphatidylglycerol. Values of Kr range up to 5-6 (or even higher), relative to phosphatidylcholine, corresponding to mean excess free energies of association of approximately 1 kcal/mol. Either cardiolipin (diphosphatidylglycerol), phosphatidic acid, or free fatty acid are found to have one of the highest selectivities, the detailed selectivity pattern depending on the particular protein, a point that will be returned to later. It must be remembered that the association constants (Kr) given in Table 3 are all referred to the background lipid bilayer (mostly composed ofphosphatidylcholine) as standard state. Different values may be expected with different background lipids, not only because of the selectivities between the various spin-labeled lipids at probe amounts as evidenced by the values of Kr, but also because in the bulk lipid, the properties of the standard state may be modified for other, more global reasons, e.g., by the degree of hydration or by the electrostatic surface potential of the whole bilayer. This can be seen in the comparison of the lipid spin-label selectivity patterns for the myelin proteolipid protein in dimyristoyl-phosphatidylcholine and in dimyristoyl-phosphatidylglycerol bilayers, respectively (first two entries in Table 3). Also of importance is the influence of other membrane proteins on the selectivity of lipid-protein interactions. It may be expected that the presence of other proteins effectively will modify the standard state to which the values of the relative association constant (Kr) of lipids interacting with the protein in question is referred, hence modifying the lipid selectivity by competition. In experiments on the association of the myelin basic protein with dimyristoyl-phosphatidylglycerol bilayers containing the myelin proteolipid protein (Sankaram et al., 1991), it was found that the binding of the peripheral basic protein reduced the association constants relative to phosphatidylcholine for all spin-labeled lipids interacting with the integral proteolipid protein to an extent that correlates with the selectivity of their interaction with the basic protein (cf. see below). The analysis of lipid selectivity given so far has been concerned with relative association constants averaged over all lipid sites at the protein interface, without considering whether these values arise from a small number of highly selective sites or from a smaller, generalized selectivity for all sites. It is not possible to distinguish between these two situations by experiments in which the test lipid is present only at low relative concentrations, such as those normally employed with probe lipid species, particularly spin-labels (Brotherus et al., 1981 ). To do this, it is necessary to increase the relative concentration of the test lipid to check for possible effects of saturation of highly specific sites. This has been done in the case of cardiolipin interacting with cytochrome oxidase, for which it displays a relatively high specificity. It was found that the fraction of spin-labeled cardiolipin (present at probe amounts) interacting with the protein changed only a little with increasing amounts of unlabeled cardiolipin in the background lipid at constant total lipid/protein ratio

156

DEREK MARSH

ooooo aaAaZ~ +++++ xxxxx

800 -

'm

600

DMPC CL DiMeCL MonolysoCL

-

i

:~

_

400

-

r'---1

r t.--.J

~200 O

0

,

0

l

40

~

I

80 -1

^~T--

I 120

7"

--i 160

Vo(s ) Figure 6. Eadie-Hofstee plots of the activity of cytochrome oxidase reconstituted in dimyristoyl-phosphatidylcholine alone (O), and in dimyristoyl-phosphatidylcholine supplemented with 1 7 mol % of dimethylcardiolipin (+), monolysocardiolipin (X), or native cardiolipin (A). The full lines are simulations according to the kinetic model of Brzezinski and Malmstr6m (1986) which involves two conformations of the protein with different affinities for protons and two redox centers ofthe enzyme. The activation by cardiolipin and its analogs is simulated by increases in the rate constants for the conformational change of the unprotonated enzyme and of those for binding cytochrome c. (Abramovitch et al., 1990). (Powell et al., 1985). On the basis of an analysis using a more complete form of Eq. 4, it was concluded that the selectivity of cardiolipin for cytochrome oxidase was expressed at all lipid sites on the protein and not only at a few highly specific sites. The specificity of cytochrome oxidase for cardiolipin may be of functional significance in the inner mitochondrial membrane, since cardiolipin is unique to mitochondria in mammalian systems and is also found to be an efficient activator of the cytochrome oxidase enzyme (see Marsh and Powell, 1988; Abramovitch et al., 1990; Robinson, 1993; and references therein). This activating capacity is illustrated in Figure 6 for preparations of reconstituted cytochrome oxidase free of endogenous cardiolipin and to which cardiolipin, or derivatives of cardiolipin, have been added. Conversely, free fatty acids are found to inhibit the Na§247

Lipid-Protein Specificity

1 57

(Ahmed and Thomas, 1971), which correlates with the specificity for association of spin-labeled fatty acids with this protein. On the other hand, negatively charged phospholipids are required for efficient reactivation of delipidated Na§ in many reconstitution systems (Kimelberg and Papahadjopoulos, 1974; Wheeler et al., 1975; Palatini et al., 1977; de Pont et al., 1978; Mandersloot et al., 1978; Cornelius and Skou, 1984). With the exception of phosphatidylglycerol, this correlates qualitatively with the phospholipid spin-label selectivity pattern given in Table 3. Fatty acids are also among the class of noncompetitive blockers that have a local anesthetic-like effect on the acetylcholine receptor. They are able to block the agonist-induced ion flux, without giving rise to receptor desensitization (Andreasen and McNamee, 1980). This property correlates well with the selectivity of the receptor for spin-labeled fatty acids (see Table 3). In addition, spin-labeled local anesthetic analogs have been shown to substitute, with a greater specificity than for phospholipids, in sites at the lipid-protein interface (Earnest et al., 1986; Horv~ith et al., 1990b), which is also in agreement with the site of action proposed from inhibition studies (Heidmann et al., 1983). Further discussion of the functional effects of different lipids in activation can be found in Marsh (1987). Of the lipids not included in Table 3, interestingly, gangliosides show relatively little selectivity over that for phosphatidylcholine in interaction with the Na§ § ATPase or with the acetylcholine receptor (Esmann et al., 1988b; L.I. Horv~ith, H.R. Aries, G. Schwarzmann, K. Sandhoff, D. Marsh, and EJ. Barrantes, unpublished results). However, gangliosides certainly are not excluded from interaction with the protein, indicating that their unique headgroup structure may play a significant role at the lipid-protein interface. In plant photosynthetic membranes, the lipid composition is very different from mammalian and other systems. For thylakoid membranes, monogalactosyl-diglyceride appears to take the part played by phosphatidylcholine in mammalian membranes, and phosphatidylglycerol displays a marked selectivity relative to monogalactosyl-diglyceride both in whole thylakoid membranes and in subfractions enriched in either photosystem I or photosystem II (Li et al., 1989). Neither destacking of the thylakoids nor phosphorylation or illumination were found to affect the selectivity for phosphatidylglycerol (Li et al., 1990a). Therefore, the redistribution of thylakoid protein complexes in the membrane, under these various conditions, takes place with conservation of the properties of the lipid/protein interface. However, the lipid-protein interactions in thylakoids from chilling-insensitive plants were found to be different from those for chilling-sensitive plants (Li et al., 1990b). In particular, a greater selectivity for phosphatidylglycerol was found in the chilling-insensitive plants and it is conceivable that this might lead to a stabilization of the thylakoid protein complexes (cf. also, Tr6moli6res et al., 1981; R6my et al., 1982). An interesting case of lipid headgroup selectivity arises for a small peptide (K26) that represents the single apolar span of a small 130-residue protein (IsK) which, on expression, induces slowly activating voltage-gated K § channels (Takumi et al., 1988). The 26-residue K26 peptide has an amino acid sequence: KEALY-

1 58

DEREK MARSH

ILMVLGFFGFFTLGIMLSYIR, which contains Lys 1, Glu 2, and Arg 26 at the Nand C-terminals, respectively, as the sole charged residues. Nevertheless, on reconstitution into lipid bilayers, the K26 peptide displays a pronounced selectivity for certain negatively charged phospholipids (Horv~ith, et al., 1995). The relative association constants found for different spin-labeled lipids in a dimyristoyl-phosphatidylcholine host are: Kr = 4.3 for phosphatidic and stearic acids, 2.5 for phosphatidylserine, and -- 1 for phosphatidylglycerol and phosphatidylcholine. Therefore, in spite of the relative simplicity of the system and the relatively small number of basic residues, the structure of the peptide assemblies in lipid bilayers is such as to present a hydrophobic surface that is capable of motionally restricting the lipid chains and to exert a lipid selectivitymat least in part, via the polar termini. These are points that will be returned to in a later section.

B. Ionic Strength Dependence of Selectivity As already noted, the highest selectivities in Table 3 are observed for negatively charged phospholipids, and thus it is likely that electrostatic interactions play a significant role in the lipid specificity of integral membrane proteins. Electrostatics are not invariably the sole contributing factor, however, because lipids with the same formal charge have different selectivities. In particular, negatively charged phosphatidylglycerol in most cases displays no selectivity relative to the zwitterionic phosphatidylcholine. The electrostatic contribution to the lipid selectivity can be defined as that part of the total selectivity which is eliminated by screening in solutions of high ionic strength. In many cases, this does not account for the whole of the lipid selectivity observed (Marsh, 1985). The ionic strength dependence of the relative association constant can be described, at least qualitatively, by the Debye-Htickel theory of electrolytes. The average relative lipid association constant in the presence of ions is given by Eq. 3 which in this case is written as: Kr = ~(%'r

(12)

where/Qr is the association constant in the absence of screening, and ~ , ~,, and %1, are the activity coefficients of the spin-labeled lipid, protein, and lipid-protein complex, respectively. The ionic activity coefficients are given by the DebyeHtickel expression (Robinson and Stokes, 1955)" In ~ = [-Z~e2/(8rteoekT)][~/(1 + ~caj)]

(13)

where Zj is the charge on species j, and aj is the interaction distance of species j with counterions. The inverse Debye screening length in Eq. 13, which governs the range of the electrostatic interaction, is given by" = (2000NAe21/~oekT) 1/2

(14)

Lipid-Protein Specificity

159

where I is the ionic strength of the solution. Equations 13 and 14 have been found to be reasonably successful in interpreting the ionic strength dependence for the interaction of various negatively charged lipids with the Na§247 (Esmann and Marsh, 1985), cytochrome oxidase (Powell et al., 1987), the myelin proteolipid protein (Horv~ith et al., 1988a), and the ADP-ATP carrier protein (Horv~ith et al., 1990a). The precondition for this is that the ionic interaction distances (aj) should be treated as semi-adjustable parameters, in order empirically to extend the range of validity of the Debye-Htickel theory to higher ionic strengths (Robinson and Stokes, 1955). The ionic strength dependence of the relative association constants for different spin-labeled lipids with the myelin proteolipid protein is shown in Figure 7, as an example.

10

r~

o

6 A

~

4

2

o lt o.o

w. . l. . I. . l. . I. . I. . I. . .I . .I . .I . ~

0.5

1.0

t

i1/2

I

t

I

1.5

I

J

J

I

2.0

Figure 7. Ionic strength (I) dependence of the relative association constants (Kr) for spin-labeled phosphatidic acid (o), stearic acid (A), and phosphatidyiserine (m), with the myelin proteolipid protein in recombinants with dimyristoyl-phosphatidylcholine. The values are normalized to the relative association constant ( KPC r ) for spin-labeled phosphatidylcholine (dashed line). Solid lines are fits to the Debye-H0ckel theory (cf. Eq. 13) with ZL = -1, Zp = +1, ap = 1.65 nm, and aL = 0.12-, 0.24-, 0.33-nm for phosphatidic acid, stearic acid, and phosphatidylserine, respectively. (Horv,~th et al., 1988a).

160

DEREK MARSH

In general, the experiments on the ionic strength dependence show that the selectivity for negatively charged lipids cannot be screened out completely by high concentrations of salt. This is the case for phosphatidic acid and phosphatidylserine interacting with the Na+,K+-ATPase, for cardiolipin interacting with cytochrome oxidase, for phosphatidic acid and stearic acid interacting with the myelin proteolipid protein, and for cardiolipin, stearic acid, and phosphatidic acid interacting with the ADP-ATP carrier. These results indicate a nonelectrostatic, in addition to an electrostatic, contribution to the lipid selectivity. On the other hand, the selectivity of stearic acid for the Na§ and of phosphatidylserine for the ADP-ATP carrier is eliminated completely at high ionic strength (I > 1.5). For the myelin proteolipid protein, most of the selectivity displayed by phosphatidylserine can also be screened by moderately high concentrations of salt (see Figure 7). In these latter two cases, the lipid selectivity is presumably of a predominantly electrostatic character, and also directly attributable to positively charged residues on the protein.

C. pH Dependence of Selectivity The lipid selectivity is additionally affected by pH titration, if the pKas of the groups involved in the specificity of interaction are in the accessible range. The pH dependences of the fraction of various spin-labeled lipids that are associated with the myelin proteolipid protein in recombinants with dimyristoyl-phosphatidylcholine of identical lipid/protein ratios, as determined by ESR experiments, are shown in Figure 8, as an example. Besides changing the electrostatics by neutralizing the negative charge of the lipid polar headgroup (as seen from the figure), pH titration may also have other effects such as changing the degree of polar group hydration on protonation. For the myelin proteolipid protein, the fractions of stearic and phosphatidic acids interacting with the protein titrate with an effective pKa of 7.5 for both lipids, and phosphatidylserine also begins to titrate at pHs greater than 8 (Horvfith et al., 1988a). In the protonated stme, stearic acid displays a selectivity which is still greater than that for phosphatidylcholine and not much smaller than that of phosphatidylserine in the singly negatively charged state. The selectivity for phosphatidic acid in the singly negatively charged state, at low pH, is yet greater than that for both the latter lipids. Both phosphatidic acid and stearic acid also display a well-defined titration behavior in their interaction with the Na+,K+-ATPase, with effective pKas of 6.6 and 8.0, respectively (Esmann and Marsh, 1985). In the singly protonated states, neither lipid displays a preferential selectivity relative to phosphatidylcholine, although phosphatidic acid still bears a net negative charge. In this case, protonation of phosphatidic acid clearly has a greater effect on the selectivity than simply that of removing the electrostatic contribution, because a much smaller reduction is found for the doubly charged state at high ionic strength. Similar to these effects of protonation of the lipid headgroup, removal of the charge on cardiolipin by

L ipid-Protein Specificity

161

0.8

I I l a

0.7

_

--

~.~ 0.6 -r-.(

9

9

G) 0.5

0,4

0.3

/&

9

--

6

'

I

7

I

8

'

I

9

10

pH Figure 8. pH dependence of the fraction (t) of spin-labeled phosphatidic acid (e), stearic acid (A), and phosphatidylserine (m), associated with the myelin proteolipid protein in recombinants with dimyristoyl-phosphatidylcholine of lipid/protein ratio 37:1 mol/mol. The corresponding values for spin-labeled phosphatidylcholine are indicated by the horizontal dashed line. The full lines are fits to a standard pH titration dependence. The vertical dashed lines are the corresponding pKa values for the titrations. (Horv~th et al., 1988a).

methylation of the phosphate groups leads to a decreased affinity for cytochrome oxidase (Powell et al., 1987). The decrease is similar to that obtained by screening at high ionic strength, and still represents a significant selectivity relative to phosphatidylcholine. These results, and those of pH titration, are in accordance with the conclusions reached regarding the relative electrostatic and nonelectrostatic contributions to the selectivity determined from the ionic strength dependence. The differential selectivities of the protonated and unprotonated lipids for the protein will have thermodynamic consequences which are manifested in a shift in pKa of the lipid at the lipid-protein interface, relative to that in the background lipid bilayer. The lipid-protein and acid-base equilibria are depicted by the following scheme:

162

DEREK MARSH

P + L* ~

KaI P + L*H ~

L*" P

I KaLP* L*H-P

where pKa = -logKa and pKLP= -logK~aP are the pKs of the lipid, L*, in the lipid bilayer and at the lipid-protein interface, respectively. Kr(L*H) and Kr(L*) are the relative association constants of the protonated and deprotonated forms of the lipid, respectively. From the cyclic nature of the equilibria, the acid dissociation constant at the lipid-protein interface is given by:

K~aP = Ka'[Kr(L*)/Kr(L*H)]

(15)

Hence the shift in pK at the lipid-protein interface, ApK~aP = pKLP- pKa, is given by: ApK~aP = log[Kr(L*H)/Kr(L*)]

(16)

Using values for the relative association constants for myelin proteolipid protein predicts values of ApK~aP = -0.2, -0.5, and -0.5 for phosphatidic acid, stearic acid and phosphatidylserine, respectively, from Eq. 16 (Horvfith et al., 1988a). If the titration of the fraction of motionally restricted lipid is taken as a measure of the pK at the lipid-protein interface (cf. Figure 8), it is found that the effective pKs quoted above are in reasonable agreement with these predictions for phosphatidic acid and for phosphatidylserine, but the pK shift of stearic acid is in the opposite direction to that predicted (Horvfith et al., 1988a). For the Na+,K+-ATPase, the upward shift in the pKa of stearic acid at the lipid-protein interface is even greater (Esmann and Marsh, 1985). Some interaction, possibly a change in polarity, must outweigh the stabilization of the dissociated form of the fatty acid by the protein in these latter cases.

D. Exchange Rates of Protein-Associated Lipids The selectivities of different lipids for a particular membrane protein are reflected also in their exchange rates at the lipid-protein interface. The intrinsic off-rates for lipid exchange, determined either by simulation of the conventional ESR spectra of the spin-labeled lipids (Horv~ith et al., 1988b), or from the power saturation properties of the ESR spectra (Horv~ith et al., 1993a), generally are found to be inversely proportional to the relative association constants, in accordance with Eq. 8 (see e.g., Fig. 9, solid line). This is the case for the myelin proteolipid protein (Horv~ith et al., 1988a, b), the DM-20 myelin protein (Horv~ith et al., 1990c), and

L ipid-Protein Specificity

163

.o -,

0.8 []

/ /

o.6

/ / /

I

[]

/ /

~0.4

/

0/13"13

b 0.2

0~

0.0 -

i

I

0.2

i

I

0.4

i

I

0.6

KrPC/K

'

I

0.8

i

1

Figure 9. Correlation ofthe off-rate for exchange ('tb 1) of spin-labeled stearic acid (A), phosphatidylserine (m),and cardiolipin (D) associated with the mitochondrial ADP/ATP carrier in recombinants with egg phosphatidylcholine, with the corresponding inverse relative association constants (Kr). Both values are normalized to those for spin-labeled phosphatidylcholine, the dependence for which is given by the full line. T = 10 ~ (Horv,~th et al., 1990a).

the M 13 phage coat protein (Wolfs et al., 1989; Peelen et al., 1992), and with the exception of cardiolipin (see Figure 9 and below), also for the ADP-ATP carrier (Horv~ith et al., 1990a). In addition, when the lipid selectivity is reduced either by pH titration or electrostatic screening at high ionic strength, the on-rates remain constant, indicating that they are diffusion-controlled, whereas the off-rates increase, reflecting the reduced specificity (Horv~ith et al., 1988a, 1990a). The intrinsic off-rates for exchange of the nonselective lipid, phosphatidylcholine, at the interface with different proteins are given in Table 4. From this, the off-rates of other lipids with higher specific 9 can be obtained from the values of the association constants relative to phosphatidylcholine by means of Eq. 8 (cf. Figure 9). In general, the off-rates for phosphatidylcholine are in the region of 10v s-1. These values are of the same order, but significantly slower, than the intrinsic lipid-lipid exchange rates (< 108 s-l; Sachse et al., 1987) arising from lateral diffusion in fluid lipid bilayers. The reason for this could, in part, be a steric

164

DEREK MARSH

Table 4. Lipid Exchange Off-Rate Constants ('rg1) and Activation Energies (Ea) for Various Integral Membrane Proteins Reconstituted in Phosphatidylcholine Bilayers a Protein/Lipid b

T (~

1:~1 (s -l)

Ea (kJ/mol)

Ref c

Myelin proteolipid protein/myr2PtdCho

30

1.6 X 107

20

1

Myelin DM-20 protein/myr2PtdCho

30

1.5 X 107

--

2

Rhodopsin/myr2PtdCho M 13 coat protein/myr2PtdCho

30 30

1.6 X 107 2.3 X 107,d

20 __

3 4

30

5.3 X 106,e

m

5

M13 coat protein/ole2PtdCho ADP-ATP carrier/egg PtdCho

24

3.0 X 107,d

~

4

10

1.4 x 107

~

6

Ca 2+-ATPase/ole2PtdCho

37

1-2 • 107

~

7

Notes: aData obtained by simulation of the conventional ESR spectrum of spin-labeled phosphatidylcholine. bptdCho, phosphatidylcholine; myr, myristoyl; ole, oleoyl. CReferences: 1. Horv~ith et al. (1988a,b); 2. Horv~ith et al., (1990c); 3. Ryba et al. (1987); 4. Peelen et al. (1992); 5. Wolfs et al. (1989); 6. Horv~ithet al. (1990a); 7. East et al. (1985). dot-helical form (partly) of the protein. el3-sheet form of the protein.

hindrance of the diffusive motion by the protein, but it could also reflect a somewhat energetically more favorable interaction with the protein than the purely lipid-lipid interactions (cf. the section above on chain dependence of lipid selectivity). In contrast to the above general correlation between exchange rate and the average relative association constant, it has been found that the intrinsic off-rate for exchange of cardiolipin associated with the ADP-ATP carrier (which is a highly basic protein) is much slower than that which would be predicted from its relative association constant by using Eq. 8 (Figure 9, dashed line). This is because the method used to determine the exchange rates does not distinguish readily between labels that are exchanging at the average rate, and those that are exchanging much more slowly (Horv~ith et al., 1990a). Therefore, it was concluded that there was a smaller population (approximately 30% of the total) of more highly specific sites for cardiolipin on the lysine-rich ADP-ATP carrier, at which the exchange rate was undetectably slow (< 106 s-l). This seems to be the exception rather than the rule, at least for those proteins for which the lipid exchange rates have been measured; the possible origin of this effect is discussed in more detail later.

E. Gel Phase Lipids and Selectivity Finally, a brief mention is given of the selectivity of interaction with proteins in gel phase lipids. All the results mentioned so far have referred solely to proteins in a fluid phase lipid environment. The association of spin-labeled lipids with the myelin proteolipid protein in gel phase dimyristoyl-phosphatidylcholine has been studied by saturation transfer ESR spectroscopy (Horv~ith et al., 1993a,b). It is

L ipid-Protein Specificity

165

found that the rate of exchange of the lipids associated with the protein in the gel phase is slow (~106 s-l). However, the number of lipids associated with the protein and their selectivity is similar to that found in the fluid phase. Thus, it appears that the myelin proteolipid protein is solvated rather similarly by lipids in the gel and fluid phases, although this must not generally be the case, and in addition, it can not be excluded that the solvated protein is aggregated in the gel phase. How far these results can be extrapolated to other integral proteins is not yet clear. For instance, it has been found by NMR techniques that the lipid stoichiometries of rhodopsin and cytochrome oxidase in the gel phase differ somewhat from those found by spin-label ESR techniques in the fluid phase (Bloom and Smith, 1985). Additionally, as already mentioned, when in competition with fluid phase lipids, gel phase lipids display a reduced selectivity for the Ca2+-ATPase (London and Feigenson, 1981; East and Lee, 1982). This latter is, however, a different situation from that described in the preceding paragraph for spin-labeled lipids at probe amounts interacting with the myelin proteolipid protein, where all lipids essentially are in the gel phase.

VI. INTEGRAL PROTEIN STRUCTURE A N D LIPID SELECTIVITY The features of the protein sequence and structure that could contribute to the selectivity observed for negatively charged lipids, namely the placing of uncompensated basic residues (particularly those at the ends of the transmembrane sections) close to the lipid polar headgroups at the membrane surface, have been discussed in detail previously (Marsh, 1993). Only a few examples, for which a clear connection between the protein sequence and lipid selectivity have been established experimentally, will be treated extensively here.

A. Myelin Proteolipid Protein The proteolipid protein from myelin (PLP) displays a marked selectivity for certain negatively charged phospholipids, particularly phosphatidic acid for which Kr = 10.4 (Horv~ith et al., 1990c). The DM-20 myelin protein is a 26 kD isoform of the proteolipid protein differing from the parent protein by the deletion of 35 contiguous residues (positions 116-150) in the major charged hydrophilic loop (Figure 10). This protein, while possessing a similar number of sites (Nb) for the association of nonselective lipids such as phosphatidylcholine, has a greatly reduced selectivity of interaction with negatively charged lipids as compared with the parent protein (Table 5). (Note that the values of Kr given for PLP in Table 3 above correspond, in fact, to the natural mixture ofPLP + DM-20.) For phosphatidic acid, the relative association constant is reduced to Kr = 2.3 (Horv~ith et al., 1990c). Thus, the sequence deleted in the DM-20 isoform makes the major contribution to

166

DEREK MARSH

+~, |174174174 ~|

_ +|

Cc~

,-~

r

|

z.,

14

+,

0~

~|174174 "~

~'_

, ~

([tt'_-_.

~__'~ ", +

-

_

OUT

+~

:i4

F

I )~

|

|174174 O - s - s . ' ~ + |169~ ~,~|

+~

~

+ ~-

|

II ~

|

-~'s.

~ |174 +

+,.,e+~ ~.

III ~

IV

+

IN5 +

'

+

Figure 10. Transmembrane topology of the myelin proteolipid protein proposed on the basis of the amino acid sequence, and the acylation and reduction states of the cysteine residues. The region that is deleted in the DM-20 myelin protein isoform (residues 116-150) is indicated by the short horizontal lines. (Weimbs and Stoffel, 1992). the lipid selectivity of the myelin proteolipid protein. This demonstrates a clear connection between primary structure and the selectivity of lipid-protein interactions that can be interpreted in terms of models for the transmembrane disposition of the protein. On the four-helix model proposed for the transmembrane topology of the myelin proteolipid protein by Weimbs and Stoffel (1992) that is given in Figure 10, there are seven acidic residues, six basic residues, and two histidines, at the membrane outer surface. Thus, there is a net extracellular charge of either-1 or + 1, depending on the protonation state of the histidines. Additionally, there are only two histidines (one of which, if charged, probably would be compensated by a glutamic acid) and two extra acidic residues that are definitely placed close to the lipid headgroups. Therefore, it seems unlikely that the extracellular domains of the protein are likely to contribute greatly to the selectivity for negatively charged lipids. At the inside surface of the membrane, there are five acidic residues and thirteen basic residues, plus four histidines, giving a net intracellular charge of +8 or +12, depending on the pKas of the histidines. Of these, 5-6 uncompensated basic residues plus two histidines and two uncompensated acidic residues are definitely placed within five

Lipid-Protein Specificity

167

Table 5. ParametersCharacterizing the Lipid-Protein Interactions and Selectivity of the Myelin Proteolipid Apoproteins, PLP and DM-20, in DimyristoyI-Phosphatidylchol ine Bilayers a Parameter

PLP

Zb

+7 - + 1 3

Nb (mol/mol) Xt~1 (PtdCho)

9-10 1.6 • 107 s-1

Kr (PtdH)C/Kr(PtdCho)

DM-20

+3 - + 5 9-10 1.5 • 107 s-1

10.4

2.3

(StH)

6.5

2.0

(PtdSer)

2.2

1.1

(PtdGro)

1.8

1.0

Notes

aSource: Horv~ithet al. (1990c).

bZ is the total charge on the protein, the lower (higher) value assigns a charge of 0 (+1) to each histidine. cptdCho, phosphatidylcholine; PtdH, phosphatidic acid; StH, stearic acid; PtdSer, phosphatidylserine; PtdGro, phosphatidylglycerol.

residues of the lipid headgroups in the model. Thus, it would appear that these residues, and additionally other basic residues that might be situated close to the membrane surface in the protein folding, are likely to be responsible for at least part of the selectivity of the proteolipid protein for anionic lipids. The section of the proteolipid protein that is deleted in the 26 kD DM-20 isoform contains two acidic residues and six basic residues plus all four of the cytoplasmic histidines. Thus, this region bears at least 50% or, depending on the protonation state of the histidines, up to 65% of the net positive charge on the protein. Of these, however, only two histidines and one uncompensated basic residue are placed within five residues of the lipid headgroups. Therefore, within the framework of the model, it seems likely that other basic residues must also be placed close to the lipid headgroups within the three-dimensional structure of the parent PLP protein to account for the large contribution of this section of the protein to the lipid selectivity.

B. M13 Bacteriophage Coat Protein and lsK Protein The coat protein from the M 13 bacteriophage is a small bipolar protein with a single 19-residue hydrophobic stretch (Figure 11). It is known to exhibit a conformational polymorphism, in that the protein extracted with phenol has a secondary structure consisting of [3-sheets, whereas that extracted with cholate is predominantly or-helical in structure (Nozaki et al., 1976; Spruijt et al., 1989). Depending on the lipid/protein ratio, ionic strength, and lipid type, the t~-helical structure of the cholate-extracted protein is either largely preserved upon incorporation of the protein in phospholipid bilayers, or it converts partially to the polymeric 13-sheet

168

DEREK MARSH Acidic

domain:

+H3N_A~.a_G1 u-Gly-Asp-Asp-Pro-Ala-Lys-Ala-A~a-Phe-Asn-Ser-Leu5 0 15

20

G ln-Ala-Ser-Ala-Thr-Gl uHydrophobic

domain: 25

3o

Tyr-Ile-Gly-Tyr-Ala-Trp-Ala-Met-Val-Val-Val-Ile-Val-Gly35 Ala-Thr-Iie-Gly-IleBasic

domain:

40

45

5o

Lys-Leu-Phe-Lys-Lys-Phe-Thr-Ser-Lys-Ala-Ser-CO0-

Figure 11. Primary sequence ofthe M 13 bacteriophage coat protein (Van Wezenbeek et al., 1980). Hydrophobic residues are in bold face and charged residues are in italics. The predominantly negatively charged N-terminal domain, the positively charged C-terminal domain, and the central hydrophobic stretch are indicated.

structure (Spruijt et al., 1989; Spruijt and Hemminga, 1991). Thus, this reconstituted system allows comparison of the lipid interactions with a protein in either of the two secondary structural states commonly found for the transbilayer section of integral membrane proteins (Table 6). One interesting feature found from spin-label ESR studies of this protein in dimyristoyl-phosphatidylcholine (Peelen et al., 1992)

Table 6. Parameters Characterizing the Lipid-Protein Interactions and Selectivity of the Cholate-Extracted and Phenol-Extracted M13 Phage Coat Protein in DimyristoyI-Phosphatidylchol ine Bilayersa'b Parameter IRmax c

Cholate-Extracted (predominantly ct-helix) 1654 cm -1

Phenol-Extracted (predominantly ~-sheet) 1625 cm -1

Nb (mol/mol) ~1~1 (PtdCho) d

4-5 2-3 x 107 s-1

4 5 x 106 s-1

Kr(PtdH)/Kr(PtdCho) (StH)

1.6 1.2

4.2 2.3

(PtdSer)

1.2

2.1

(PtdGro) (PtdEtn)

1.1 1.0

1.6 0.9

Notes: aSource:Peelen et al. (1992). bin dimyristoyl-phosphatidylcholine bilayers, cholate- and phenol-extracted MI 3 coat protein is predominantly in the or-helical and p-sheet conformations, respectively. qRmax is the position of the peak maximum in the amide I region of the Fourier transform infrared spectrum of the protein.

Lipid-Protein Specificity

169

is that the exchange rates (xgl) of lipids at the hydrophobic surface of the ]3-sheet form of the protein are up to 4-5 times slower than are those for the protein in a predominantly or-helical conformation. Whereas this may be due, in part, to trapping oflipids within the polymeric ]3-sheet structures, the more extended nature of the peptide chain in a 13-sheet conformation, as compared with a more compact transmembrane or-helix (cf. Marsh, 1993), may cause the lipid chains to assume a more extended, and hence less flexible, structure which would also serve to reduce the lipid exchange rates. Regarding the specificity of the lipid-protein interactions, considerable differences were also found in the selectivity for negatively charged lipids between the t~-helical and 13-sheet forms of the M 13 bacteriophage coat protein (Table 6, Peelen et al., 1992). A considerably more pronounced range of lipid selectivities was found for the 13-sheet form of the protein than for the protein in a predominantly t~-helical conformation. This can be attributed largely to a different location of the charged residues in the N- and C-terminal domains of the protein relative to the lipid polar headgroups at the membrane surface. This will be determined in part by the way in which the hydrophobic section of the protein traverses the bilayer, in particular whether the [3-sheet is strongly tilted relative to the membrane normal, or forms a 13-hairpin structure crossing the bilayer twice, and whether the strands are arranged in a parallel or in an antiparallel fashion (cf. Marsh, 1993). Also, the clustering of charged residues close to the membrane surface will depend on the secondary structure of the polar extramembranous portion of the protein, because of the difference in pitch for equivalently oriented residues (2 for a [3-sheet conformation and 3.6 for an t~-helical conformation). This effect may also contribute to the enhanced lipid selectivity of the ]3-sheet form of the M13 coat protein. Interestingly, the K26 peptide, which corresponds to the single putative transmembrane stretch of the IsK protein that is associated with slow voltage-gated K + channels, tends also to display a plasticity in secondary structure. The lipid selectivity of this peptide in dimyristoyl-phosphatidylcholine bilayers has been described in a previous section. The sequence of the K26 peptide, with the notation given in Figure 11, is: +H3N- Ly s-G lu- A l a- Leu- Tyr- lle- Leu- M e t-Val- L e u - G l y- Ph e- Ph e-G l y- Ph e- Ph e-

Thr-Leu-Gly-Ile-Met-Leu-Ser-Tyr-Ile-Arg-CO0which contains a continuous stretch of 23 apolar residues that is of sufficient size to span the membrane in an t~-helical conformation (cf. Marsh, 1993). However, the peptide reconstituted in dimyristoyl-phosphatidylcholine bilayers is present wholly in a ]3-sheet conformation (Horv~ith et al., 1995). This result is significant because molecular models that have been suggested recently for other potassium channels contain pore-lining segments in ]3-sheet conformations, rather than the more familiar transmembrane t~-helices proposed for many membrane proteins (Guy and Conti, 1990; Bogusz et al., 1992). Approximately 2.5 lipids per peptide

1 70

DEREK MARSH

monomer, independent of the lipid species, are motionally restricted by direct interaction with the peptide assembly in the bilayer (Horv~ith et al., 1995). This value is consistent with a 13-barrel structure for the peptide in which the 13-strands either are strongly tilted (by 60 ~ or have a reverse turn at their center, both of which would give a transmembrane span compatible with the hydrophobic thickness of the lipid bilayer (e.g., Marsh, 1993). The pronounced lipid selectivity of the peptide for negatively charged lipids, as described above, suggests that the single basic residues at the N- and C-terminals are positioned close to the lipid headgroups in the peptide assemblies. This is again consistent with the K26 peptide being integrated in the bilayer as a strongly tilted ~-sheet structure, or as an assembly of 13-hairpin conformations, and not protruding appreciably from the bilayer surface. The rather high values of the relative association constants for certain of the anionic lipids suggests a very effective association with the limited number of the basic residues in the peptide. Presumably this is, at least in part, for the reasons advanced above for the higher lipid selectivities with the [3-sheet conformation of the M13 phage coat protein. In particular, the alternation in orientation of adjacent residues relative to the lipid in a ~-sheet structure would tend to mitigate the possible cancellation of the charges of opposite sign (Lys and Glu) at the N-terminal of the peptide.

C. Mitochondrial ADP/ATP Translocator Another integral protein for which particularly interesting features of the lipid selectivity have been found relative to the protein structure, is the ADP/ATP cartier of the inner mitochondrial membrane. This protein is not only hydrophobic, but also extremely basic in character, with an isoelectric point of >10. The primary sequence contains six glutamic acids, 15 aspartic acids, 17 arginines, and 23 lysines, giving a net positive charge of +18. Chemical labeling studies have identified several lysine residues, both on the cytosolic and matrix faces of the protein, which are thought to be located directly in the vicinity of the phospholipid headgroups, and thus could give rise to a specificity of interaction with negatively charged lipids (Bogner et al., 1986). The carrier protein is, indeed, found to display a pronounced selectivity for anionic lipids, in particular for cardiolipin, a lipid peculiar to mitochondria in mammalian systems (Horv~ith et al., 1990a). A unique feature of the selectivity for cardiolipin is that approximately 30% of this spin-labeled lipid associated with the carrier has an exchange rate that is atypically slow (cf. Figure 9) and, therefore, is associated at sites of considerably higher specificity than those of other integral proteins studied to date. This is most certainly a reflection of the unusually high density ofpositively charged residues close to the membrane surface in this protein.

Lipid-Protein Specificity

1 71

D. Cytochrome cOxidase Chemical modification studies have also identified lysine groups as being involved in the specificity of another mitochondrial protein, cytochrome c oxidase, for cardiolipin (Powell et al., 1987). Cytochrome oxidase is a large, multimeric integral protein with a subunit structure of considerable complexity. Affinity labels attached at the phospholipid headgroups have been used to identify those lysine groups on the enzyme that are situated at the polar-apolar interface of the membrane accessible to lipid (McMillen et al., 1986; Kuppe et al., 1987). It was found (Table 7) that the lysine-specific, benzaldehyde lipid probes incorporated selectively into the smaller subunits (V-VIII). Similarly, in a different study, an arylazido analog of cardiolipin was found to label cardiolipin-depleted cytochrome oxidase at the smaller subunits, probably subunit VIIc and possibly, but less likely, subunits VIIb and VIII (Robinson, 1993). The most hydrophobic, mitochondrially coded subunits (I-III) were labeled relatively little, and the lysine-rich subunit IV not at all. Subunits I and III were heavily labeled by nonspecific photoactivatable lipids, however, indicating that they contribute to the protein-lipid interface, but possibly not to the specificity for negatively charged lipids. Of the smaller polypeptides (VII-VIII) that are heavily labeled by the lysine-reacting lipid probe, each has a hydrophobic sequence consisting of approximately 20 amino acids flanked by ionic residues, including lysines, which presumably therefore contribute to the specificity of cytochrome c oxidase for negatively charged lipids. Table 7. Labeling Patterns of Cytochrome c Oxidase with Lysine-Specific

Benzaldehyde Probes and with Photoaffinity Arylazido Lipid Probesa'b

Subunit

Mr

Transmembrane No. of StoichioSegments Lysines metry

Lipid Benzaldehydes

Aryl-azido Water-Soluble Lipid Benzaldehyde

I

56993

several

9

1

+

+++

+

II

26049

2

6

1

+

0

+

HI

29918

several

3

1

+

+++

+

IV

17153

1

18

1

0

+

+++

Va

12436

none

7

1

+ ]

Vb

10670

none

6

1

+

Via

9419

1

3

1

-

"1

VIb

10068

none

6

1

-

~

VIc

8480

1

8

1

VIIc

5441

1

4

1

VIII

4962

1

5

[2]

++

intensity of labeling is ranked according to: 0, +, ++, +++.

++ +++ +

+

+++ J

Note." aSource:Kuppe et al. (1987). ~

+

+ +

+++

++-I-

1 72

DEREK MARSH

VII.

LIPID SELECTIVITY OF PERIPHERAL PROTEINS

The basic feature of lipid selectivity for most peripheral proteins is an absolute requirement for negatively charged lipids. Classically, peripheral proteins are released from the membrane at high ionic strength, indicating the obligatory nature of electrostatic surface interactions for their binding. Although there is various evidence that certain peripheral proteins, e.g., the myelin basic protein (Boggs et al., 1988; Sankaram et al., 1989a), penetrate lipid bilayers, their removal by high salt and the lack of binding to zwitterionic bilayers indicates that the accompanying hydrophobic interaction is energetically insufficient to sustain an effective lipidprotein association.

A. Influence of Protein Charge Model experiments with oligo-lysine peptides (lysn; Kim et al., 1991), have indicated that the intrinsic association constant (KA) with negatively charged lipids increases by approximately a factor of ten for each additional lysine residue, corresponding to an intrinsic free energy of association per residue of AGy = -1.3 kcal/mol at an ionic strength of I = 0.1 (see Figure 12). Similar results were also obtained for the binding of oligo-arginine peptides (argn) to phosphatidylserine bilayers. These values correspond to an experimental situation in which the surface potential (~) of the lipid vesicles is neutralized. In general, one would expect the total free energy for association of peripheral proteins to be given by (cf. Sankaram and Marsh, 1993): - R T InKA = neffAG]' +AG~ + Zeffe~

(17)

where neff is the effective number of positively charged residues binding at the negatively charged lipid surface, and Zeffe is the effective net charge on the protein. The first two terms on the right of Eq. 17 correspond to the intrinsic free energy of association of the protein, and the final term corresponds to the increase in activity (or concentration) of the protein due to electrostatic attraction to the membrane surface. For the lysn peptides (Figure 12), the effective number of interacting residues is neff = n-1 for AG~ = 0. In these experiments, little difference (i.e., selectivity) was found between the binding to phosphatidylglycerol or to phosphatidylserine bilayers for both the lysn and argn peptides. However, admixtures of zwitterionic phosphatidylcholine in a 2:1 ratio reduced the binding strength of the peptides by an order of magnitude. For a variety of peripheral membrane proteins, on the other hand, the free energy of association with negatively charged lipid bilayers lies in the range of-0.2 to-0.6 kcal/mol per net positive charge on the protein, corresponding to values of neff = 2-4 (Sankaram and Marsh, 1993). This reduction in free energy is illustrated more correctly in Figure 12 for the binding of myelin basic protein and its two comple-

L ipid-Protein Specificity

1 73

10 -I

~,10

-2.

D ~

9

10 -~

10

-( _ _

l

i

I

I

2

3

4

5

6

n

Figure 12. Peptide concentration (C0.5) required to neutralize the surface potential of phosphatidylglycerol vesicles by binding lysn (o) peptides, as a function of the number of positively charged residues (n) in the peptide. (I = 0.1, pH 7.0, T = 25 ~ The full line is a linear regression to the data for the lysn peptides, with a gradient corresponding to a free energy of association of-1.3 Kcal/mol per lysine residue. (Data from Kim et al., 1991). The reciprocal association constants (1/KA) of the myelin basic protein and its complementary fragments (residues 1-116 and 11 7-170) for binding to dimyristoyl-phosphatidylglycerol dispersions (E~)as a function of the net charge on the protein and fragments are also shown. For the latter, the values of n have to be multiplied by a factor of four. (I = 0.02, pH 8.0). The dashed line is a linear regression to the data for the myelin basic protein and fragments, with a gradient corresponding to a free energy of association of-O.1 Kcal/mol per net positive charge on the protein. (Data from Sankaram et al., 1989b).

mentary fragments of Mr 12.6 kDa and 5.8 kDa to bilayers of dimyristoyl-phosphatidylglycerol. The incremental free energy of association per net positive charge on the protein deduced from the gradient of this plot is o n l y - 0 . 1 kcal/mol. These values for the peripheral proteins have not been corrected for the residual surface potential, but the latter may be relatively small because the binding constants were determined primarily in the region of the binding curve approaching saturation. Nevertheless, the free energies are much smaller than those predicted from the incremental values obtained with the lysn or argn peptides. This reflects the spatial

1 74

DEREK MARSH

distribution of positive charges over the surface of the protein; apparently the secondary and tertiary folding do not allow many of the basic residues to interact optimally with the negatively charged lipid surface.

B. Lipid Headgroup Selectivity The selectivity of interaction of different lipids with peripheral proteins may be probed in several ways. One is via competition in mixed lipid systems, another is the relative binding strengths to a single lipid type, and a third is the ability to penetrate and increase the surface pressure of monolayers composed of different lipids. Clearly, not all these different methods necessarily reflect the same features of the selectivity of the interaction. In particular, for mixed lipid systems, the standard state to which the interaction is referred may vary with the different lipid mixtures. It has been found, for instance, that the mode of interaction of the myelin basic protein with mixtures of phosphatidylcholine and phosphatidylglycerol is different at the higher mole fractions of the zwitterionic lipid from what it is at the higher mole fractions of the negatively charged component (Sankaram et al., 1989c). The order of selectivity for a range of spin-labeled lipids present at probe concentrations and interacting with a variety of peripheral proteins bound to bilayers composed solely of negatively charged phosphatidylglycerol are given in Table 8. These selectivity series were determined from the degree of perturbation of the ESR spectra from the spin-labeled lipids (Sankaram et al., 1989c,d). On the whole, the degree of selectivity is not large, an exception being phosphatidylserine interacting with the myelin basic protein. The maximum differential free energies of interaction for most lipids are in the region of 0.5 kcal/mol per lipid, or 1.7-2.6 kcal/mol per protein. In general, the more highly charged form of the lipid displays a selectivity relative to the protonated species, which extends also to the interactions with fatty acids (Sankaram et al., 1990). Even the zwitterionic lipids, phosphatidylcholine and phosphatidylethanolamine, display a selectivity in certain cases, although the peripheral proteins bind much less or not at all to bilayers composed solely of these lipids. The latter is probably a result of the fact that, once the protein is bound electrostatically to the negatively charged bilayers, other interactions with the lipid headgroups may lead to a significant degree of association and hence to a limited specificity. Under these conditions, the neutral lipid, diacylglycerol, also displays some degree of selectivity of interaction in certain cases, which might have implications for the activation of protein kinase C by this particular lipid (Sankaram et al., 1989d). The stoichiometries of binding of the myelin basic protein to bilayers of phosphatidylserine, phosphatidylglycerol, phosphatidic acid, and phosphatidylethanolamine under comparable conditions have been determined to be: 1:11, 1:35, 1:40 and 1:70 mol/mol, respectively (Boggs and Moscarello, 1978). These values are in the same order as that obtained for the selectivity of interaction obtained with probe

Lipid-Protein Specificity

175

Table 8. Selectivity of Interaction of Spin-Labeled Phospholipids with Peripheral Proteins Bound to DimyristoyI-Phosphatidylglyceroi Bilayers a Peripheral Protein

Interaction Hierarchy

myelin basic protein

PS- > C L - > PA 2- > P G - > P F > PA- > PE + > PC ~

apocytochrome c

PI- > CL- > PS- > PC ~ > P G - > PE +

cytochrome c

PI- > P G - > C L - > P S - = PC ~ > PE +

lysozyme

C L - > P G - >> PE + > PC ~ > P S - > PI-

poly-lysine

C L - > PS-___ PG- > PI- > PC~ > PE +

Notes: aSource:Sankaram et al. (1989c,d). bps, phosphatidylserine; CL, cardiolipin; PA, phosphatidic acid; PG, phosphatidylglycerol; PI, phosphatidylinositol; PE, phosphatidylethanolamine; PC, phosphatidylcholine.

amounts of the different lipids when the protein is bound to phosphatidylglycerol bilayers (Table 8), and presumably reflect differences in the strength of binding to the different lipids. However, as mentioned above, this must not necessarily be the case; with cytochrome c and apocytochrome c, for instance, the situation is rather different. The strength of binding of apocytochrome c to bilayers composed of 1:1 mol/mol mixtures of phosphatidylcholine with various negatively charged lipids is in the order: cardiolipin > phosphatidylglycerol > phosphatidylinositol >> phosphatidylserine, with the association constant for cardiolipin being more than ten times that for phosphatidylserine, corresponding to an increased free energy of interaction of ca. 1.5 kcal/mol (Table 9). This selectivity series, although possessing certain similarities, is markedly different from that obtained with probe amounts of these lipids in phosphatidylglycerol (cf. Table 8), presumably for the reasons given above. For mixtures of phosphatidylserine with phosphatidylcholine, it was found that the strength of binding of apocytochrome c remained approximately constant (/Co ---20 BM) down to 15 mol % of the negatively charged lipid and, although the extent of binding relative to total lipid decreased with increasing amounts of the zwitterionic phospholipid, the stoichiometry relative to the negatively charged component remained essentially constant at ca. ten phosphatidylserine/apocytochrome c, which corresponds to approximate electroneutrality (Rietveld et al., 1986). Similarly for cytochrome c, the strength of binding to negatively charged lipid bilayers was found to be in the order: cardiolipin > phosphatidylglycerol > phosphatidylserine >> phosphatidylcholine (Rietveld et al., 1983). This again differs from the selectivity series obtained for cytochrome c with probe amounts of these lipids in phosphatidylglycerol (cf. Table 8). Selectivity series for the association of cytochrome c and apocytochrome c with negatively charged lipids have also been established from their binding to lipid monolayers and from their ability to penetrate the monolayers (Demel et al., 1989). For cytochrome c, the efficiency in increasing the monolayer surface pressure is in

1 76

DEREK MARSH

Table 9. Association Constants, Stoichiometries, and Relative Free Energies of

Association For Binding of Apocytochrome c to Unilamellar Vesiclesa'b Lipid

KA (M-l) c

N (mol/mol) d

n (mol/mol)

AG (kcal/mol) e

PtdSer/PtdCho

54 + 12 x 102

16 + 1

8 + 0.5

Ptdlns/PtdCho

20 + 3 x 103

17 + 1

8.5 + 0.5

--0.75

0

PtdGro/PtdCho

33 + 7 x 103

14 + 1

7 + 0.5

-1.1

Ptd2Gro/PtdCho

73 + 24 x 103

13 + 1

6.5 + 0.5

-1.5

Note: aSource: Rietveld et al. (1986). bUnilamellar vesicles were composed of 50 mol % egg phosphatidylcholine mixed with various negatively charged lipids. CAssociation constant is defined by: KA = [PL]/[P][L] where PL is bound protein, P is unbound protein and L is unbound negatively charged lipid. dStoichiometries: N is total lipid, n is the negatively charged lipid component. eFree energy of association AG is defined relative to the PtdSer/PtdCho mixture. fPtdSer, bovine brain phosphatidylserine; Ptdlns, yeast phosphatidylinositol; PtdGro, egg phosphatidylglycerol; Ptd2Gro, cardiolipin; PtdCho, egg phosphatidylcholine.

the order: cardiolipin >> phosphatidylserine--, phosphatidylinositol > phosphatidylcholine ---phosphatidylethanolamine. For apocytochrome c, the series is in the same order, but the effects on cardiolipin are notmuch greater than are those of phosphatidylserine, in contrast to the situation with cytochrome c. Overall, the size of the effects with apocytochrome c is much greater than is that with cytochrome c, indicating that the former has a much greater ability to penetrate lipid monolayers, a feature that is also found in bilayer systems (Grrrissen et al., 1986). These selectivity series established with lipid monolayers differ in certain respects both from the binding affinities to lipid vesicles and from the selectivity sequences determined using a fixed background host lipid containing spin-labeled probe lipids. Presumably hydrophobic components of the interaction, which are relevant to the translocation competence of apocytochrome c, but are less important for cytochrome c, are emphasized in the monolayer-penetration assay. The selectivity of binding of apocytochrome c to different negatively charged lipids may have implications for the specific targeting of this precursor protein to mitochondria, since no receptor protein for apocytochrome c has been found in the outer membrane. The outer mitochondrial membrane contains a higher proportion of negatively charged lipids than, for instance, microsomal membranes (Demel et al., 1989) and among these is the unique mitochondrial lipid, cardiolipin (Hovius et al., 1990), for which apocytochrome c displays the strongest binding (Table 9). This suggests that lipid selectivity may be a major contributing factor in determining the targeting of apocytochrome c from the cytoplasm, as is evidenced by the considerably greater extent of binding to and ability to penetrate monolayers of lipids from the outer mitochondrial membrane than those from microsomal membrane lipids (Demel et al., 1989). In addition, it is found also from lipid monolayer

L ipid-Protein Specificity

1 77

experiments that holocytochrome c has a much greater selectivity for interaction with the lipids of the inner mitochondrial membrane, than with those of the outer membrane (Demel et al., 1989). This selectivity may be attributed to the preferential concentration of cardiolipin, for which cytochrome c has a high affinity, in the inner membrane (Hovius et al., 1990) and will contribute to the localization of cytochrome c in the vicinity of its redox partners in the mitochondrial electron transport chain. From the point of view of activation of peripheral proteins by association with lipids, conformational changes in cytochrome c that involve a shift in redox potential have been demonstrated upon binding of the protein to negatively charged lipid membranes (Heimburg et al., 1991). The conformational equilibrium of the bound protein was found to be sensitive both to lipid composition and to lipid phase state. In particular, quite low concentrations of diacylglycerol were found to have marked conformational effects. It is to be anticipated that the selectivity of different lipids for interaction with cytochrome c in mixtures (cf. Table 8) may play at least a modulatory role in this control of the protein conformation by lipid composition.

C. Mixed Lipid Systems Systematic studies on mixed lipid systems indicate that a considerably greater degree of complexity can be introduced into the lipid selectivity by this additional degree of compositional freedom. The effects of admixture of zwitterionic phosphatidylcholine with negatively charged phosphatidylglycerol on the selectivity of interaction of the myelin basic protein have been studied in competition experiments over a wide range of lipid composition (Sankaram et al., 1989c). The degree of binding of the protein decreases progressively with the phosphatidylcholine content in the mixture, but to an extent that is greater than would be obtained if a fixed binding stoichiometry relative to the negatively charged component were maintained. This varying stoichiometry possibly results from conformational changes of the bound protein in response to changes in the lipid composition (cf. Heimburg et al., 1991), with consequent changes in the extent of binding. The degree of perturbation of the overall lipid mobility, as registered by spin-labeled phospholipid probes, is an indication of the lipid selectivity and was found to be greater for phosphatidylglycerol than for phosphatidylcholine, up to a mole fraction of 0.25 for the zwitterionic component. Beyond this, a uniform degree of perturbation was experienced by both lipids, which points to a different mode of association with reduced lipid selectivity at higher mole fractions of phosphatidylcholine (Figure 13a). The results for the spin-labeled phosphatidylcholine component could be predicted with reasonable success on the basis of Eq. 4, assuming direct competition for a fixed number of association sites on the protein and an association constant (Kr = 0.484) relative to phosphatidylglycerol (Sankaram et al., 1989c). The same model held for the spin-labeled phosphatidylglycerol component, in the range over which this displayed a clear selectivity relative to phosphatidylcholine. The

1 78

DEREK MARSH

6

(_9 v26 x o

E
24 0.5

b.

-

c

0.4

o 0.3

13

o 0.2 0.1 0.0

m

0.0

I

0.5

I

1.0

I

1.5

I

2.0

I

2.5

l O0/n t (mol/mol)

Figure 13. Dependence of: a) the outer hyperfine splitting (Amax) in the ESR spectra of phosphatidylglycerol (O) and phosphatidylcholine (E3)spin-labeled on the 5-C atom of the sn-2 chain, and b) the fraction (f) of phosphatidylglycerol (O) and phosphatidylcholine (E3)spin-labeled on the 12-C atom ofthe sn-2 chain that is motionally restricted by the membrane penetrating part of the protein, on the myelin basic protein/lipid ratio (1/nt) in complexes with dimyristoyl-phosphatidyiglycerol:dimyristoyi-phosphatidylcholine mixtures with mole fractions of dimyristoyl-phosphatidylglycerol in the range 0 to 1. The solid lines represent the dependence depicted from Eq. 4, with Kr(PG:PC) = 2.07 for phosphatidylglycerol relative to phosphatidylcholine, and Nb = 17 for the number of specific (electrostatic) lipid association sites on the protein. (Sankaram et al., 1989c).

breakdown of the model, for the reduction in selectivity of phosphatidylglycerol at mole fractions of phosphatidylcholine greater than 0.25, illustrates the complexity of the lipid selectivity in mixed systems. In addition to its electrostatic surface association, part of the myelin basic protein also penetrates into the lipid bilayer. The penetration of the protein can be detected by the specific motional restriction of the lipid chains spin-labeled closer to the terminal methyl end (GOrrissen et al., 1986), in a manner analogous to that found for true integral membrane proteins (cf. Marsh, 1985). This effect is visible in the ESR spectra of the spin-labeled lipids for protein-bound lipid mixtures in which

Lipid-Protein Specificity

179

the mole fraction of dimyristoyl-phosphatidylcholine is below 0.25, and again exhibits a selectivity between the spin-labeled phosphatidylglycerol and phosphatidylcholine components. The dependence of the fraction of lipid that is directly associated with the penetrant portion of the protein on the extent of protein binding is given in Figure 13b. Again, the effects of lipid composition on the lipid specificity and degree of association with the protein could be predicted on the basis of Eq. 4, using the same parameters that were necessary for the previous fits. Thus a consistent description of the lipid selectivity in the presence of low amounts of the zwitterionic component can be made in terms of simple competition, but at higher mole fractions of the zwitterionic lipid the situation is qualitatively different, exhibiting additional mixing effects. These further degrees of complexity in the lipid-protein interaction almost certainly occur with the heterogeneous lipid composition present in biological membranes. Particularly interesting in this respect, is the finding that sulphatide, the unique anionic lipid of nerve myelin, displays a selectivity relative to phosphatidylglycerol for the basic protein bound to sulphatide, but with phosphatidylglycerol as the background lipid the reverse is true (M.B. Sankaram, P.J. Brophy, G. Schwarzmann, D., Rotschiedt, K. Sandhoff, and D. Marsh, unpublished results). This gives some indication of the way in which the lipid interactions with peripheral membrane proteins may be modulated by the complex lipid composition of biological membranes.

D. Specific Lipid Binding and Covalently Linked Chains Finally, a brief mention is given of the highly specific interaction of certain peripheral proteins with a single unique type of lipid, for which a generalized electrostatic association at a charged membrane surface is not required. An archetypal example of this type of interaction is afforded by the binding of the 13-subunit of cholera toxin to the ganglioside GM1, or more generally by the binding of anticardiolipin antibodies to diphosphatidylglycerol. The association of avidin or streptavidin with N-biotinyl derivatives of phosphatidylethanolamine serves as a model system for such interactions. It has been found that the highly specific binding of avidin with spin-labeled N-biotinyl phosphatidylethanolamine causes a large and selective restriction in the chain motion of the biotin lipid, without appreciably perturbing the mobility of the lipid chains in the bulk of the membrane (Swamy and Marsh, 1993). Such a selective disruption of the lipid packing might contribute, for instance, to the mechanism of import into the cell of the ~-subunit of cholera toxin by the specific surface binding of the 13-subunits (cf., e.g., Tomasi and Montecucco, 1981). Additionally, because of the extraordinarily high affinity of avidin in the essentially irreversible binding to biotin, these results may be extrapolated to the lipid interactions with proteins bearing covalently attached fatty acyl-chains. It is possible, for example, that the specific immobilization of the protein-linked chains could enhance interactions with less mobile membrane lipid

180

DEREK MARSH

domains, such as has been suggested for the coupling of the sorting and targeting of glycosyl phosphatidylinositol-anchored proteins with that of glycolipids in polar epithelial cells (Lisanti and Rodriguez-Boulan, 1990). The more general aspects of covalently linked fatty acid chains in lipid interactions with peripheral proteins have been reviewed recently (Sankaram and Marsh, 1993).

VIii.

CONCLUSION

The effective integration of proteins in membrane structures, although determined primarily by hydrophobic effects for integral proteins and electrostatic effects for peripheral proteins, depends in detail on the interactions with the particular lipids making up the membrane. This modulation, even if subtle, most likely plays an essential part in maintaining membrane protein stability and function, as well as preserving the integrity of the membrane permeability barrier. The many diverse aspects controlling the selectivity of these lipid-protein interactions have been presented here. Additionally, there are many examples of highly specific lipid-protein associations that are required absolutely for function. The precise structural and thermodynamic features involved in the activation by these lipid cofactors are almost certainly also involved, to a greater or lesser extent, in the contribution of lipid selectivity to the stability of the overall membrane architecture.

REFERENCES Abramovitch, D. A., Marsh, D., & Powell, G. L. (1990). Activation of cytochrome c oxidase by cardiolipin and analogues of cardiolipin. Biochim. Biophys. Acta 1020, 34--42. Ahmed, K., & Thomas, B. S. (1971). The effects of long chain fatty acids on sodium plus potassium ion-stimulated adenosine triphosphatase of rat brain. J. Biol. Chem. 246, 103-109. Andreasen, T. J., & McNamee, M. G. (1980). Inhibition of ion permeability control properties of acetylcholine receptor from Torpedo californica by long-chain fatty acids. Biochemistry 19, 4719-4726. Bazzi, M. D., & Nelsestuen, G. L. (1987). Association of protein kinase C with phospholipid vesicles. Biochemistry 26, 115-122. Blake, R., Hager, L. P., & Gennis, R. B. (1978). Activation of pyruvate oxidase by monomeric and micellar amphiphiles. J. Biol. Chem. 253, 1963-1971. Bloom, M., & Smith, I. C. P. (1985). Manifestations of lipid-protein interactions in deuterium NMR. In: Progress in Protein-Lipid Interactions (Watts, A., and De Pont, J.J.H.H.M., eds.), Vol. 1, pp. 61-88, Elsevier, Amsterdam. Boggs, J. M., & Moscarello, M. A. (1978). Effect of basic protein from human central nervous system myelin on lipid bilayer structure. J. Membr. Biol. 39, 75-96. Boggs, J. M., Rangaraj, G., & Koshy, K. M. (1988). Photolabeling of myelin basic protein in lipid vesicles with the hydrophobic reagent 3-(trifluoromethyl)-3-(m-[125I]iodophenyl)diazirine. Biochim. Biophys. Acta 937, 1-9. Bogner, W., Aquila, H., & Klingenberg, M. (1986). The transmembrane arrangement of the ADP/ATP carrier as elucidated by the lysine reagent pyridoxal 5-phosphate. Eur. J. Biochem. 161,611-620.

Lipid-Protein Specificity

181

Bogusz, S., Boxer, A., & Busath, D. D. (1992). An SS1-SS2 ~-barrel structure for the voltage-activated potassium channel. Prot. Engineering 5, 285-293. Boni, L. T., & Rando, R. R. (1985). The nature of protein kinase C activation by physically defined phospholipid vesicles and diacylglycerols. J. Biol. Chem. 260, 10819-10825. Brophy, P. J., Horv~ith, L. I., & Marsh, D. (1984). Stoichiometry and specificity of lipid-protein interaction with myelin proteolipid protein studied by spin-label electron spin resonance. Biochemistry 23, 860-865. Brotherus, J. K., Griffith, O. H., Brotherus, M. O., Jost, P. C., Silvius, J. R., & Hokin, L. E. (1981). Lipid-protein multiple binding equilibria in membranes. Biochemistry 20, 5261-5267. Brzezinski, P., & Malmstrrm, B. G. (1986). Electron-transport-driven proton pumps display nonhyperbolic kinetics: Simulation of the steady state kinetics of cytochrome oxidase. Proc. Natl. Acad. Sci. USA 83, 4282-4286. Caffrey, M., & Feigenson, G. W. (1981). Fluorescence quenching in model membranes. 3. Relationship between calcium adenosinetriphosphatase enzyme activity and the affinity of the protein for phosphatidylcholines with different acyl chain characteristics. Biochemistry 20, 1949-1961. Cevc, G., & Marsh, D. (1987). Phospholipid Bilayers. Physical Principles and Models, J. Wiley, New York. Chen, Y. S., & Hubbell, W. L. (1973). Temperature- and light-dependent structural changes in rhodopsin-lipid membranes. Exp. Eye Res. 17, 517-532. Churchill, P., Mclntyre, J. O., Eibl, H., & Fleischer, S. (1983). Activation of D-]]-hydroxybutyrate apodehydrogenase using molecular species of mixed fatty acyl phospholipids. J. Biol. Chem. 258, 208-214. Cornelius, E, & Skou, J. C. (1984). Reconstitution of (Na+ + K+)-ATPase into phospholipid vesicles with full recovery of its specific activity. Biochim. Biophys. Acta 772, 357-373. Cortese, J. D., Vidal, J. C., Churchill, P., McIntyre, J. O., & Fleischer, S. (1982). Reactivation of D-l]-hydroxybutyrate dehydrogenase with short-chain lecithins: stoichiometry and kinetic mechanism. Biochemistry 21, 3899-3908. Datema, K. P., Wolfs, C. J. A. M., Marsh, D., Watts, A., & Hemminga, M. A. (1987). Spin-label electron spin resonance study of bacteriophage M 13 coat protein incorporation into mixed lipid bilayers. Biochemistry 26, 7571-7574. De Pont, J. J. H. H. M., van Prooijen-van Eeden, A., & Bonting, S. L. (1978). Role of negatively charged phospholipids in highly purified (Na+ + K+)-ATPase from rabbit kidney outer medulla. Biochim. Biophys. Acta 508, 464-477. Demel, R. A., Jordi, W., Lambrechts, H., Van Damme, H., Hovius, R., & De Kruijff, B. (1989). Differential interactions of apo- and holocytochrome c with acidic membrane lipids in model systems and the implications for their import into mitochondria. J. Biol. Chem. 264, 3988-3997. Earnest, J. P., Limbacher Jr., H. P., McNamee, M. G., & Wang, H. H. (1986). Binding of local anesthetics to reconstituted acetylcholine receptors: effect of protein surface potential. Biochemistry 25, 5809-5818. East, J. M., & Lee, A. G. (1982). Lipid selectivity of the calcium and magnesium ion dependent adenosinetriphosphatase, studied with fluorescence quenching by a brominated phospholipid. Biochemistry 21, 4144 4151. East, J. M., Melville, D., & Lee, A. G. (1985). Exchange rates and numbers of annular lipids for the calcium and magnesium ion dependent adenosinetriphosphatase. Biochemistry 24, 2615-2623. Ellena, J. E, Blazing, M. A., & McNamee, M. G. (1983). Lipid-protein interactions in reconstituted membranes containing acetylcholine receptor. Biochemistry 22, 5523-5535. Esmann, M., & Marsh, D. (1985). Spin label studies on the origin of the specificity of lipid-protein interactions in Na+,K+-ATPase membranes from Squalus acanthias. Biochemistry 24, 35723578.

182

DEREK MARSH

Esmann, M., Watts, A., & Marsh, D. (1985). Spin label studies of lipid-protein interactions in (Na+,K+)-ATPase membranes from rectal glands of Squalus acanthias. Biochemistry 24, 13861393. Esmann, M., Powell, G. L., & Marsh, D. (1988a). Spin label studies on the selectivity of lipid-protein interaction of cardiolipin analogues with the Na§247 Biochim. Biophys. Acta 941, 287-292. Esmann, M., Marsh, D., Schwarzmann, G., & Sandhoff, K. (1988b). Ganglioside-protein interactions: spin label electron spin resonance studies with (Na§247 membranes. Biochemistry 27, 2398-2403. Froud, R. J., Earl, C. R. A., East, J. M., & Lee, A. G. (1986). Effects of lipid fatty acyl chain structure on the activity of the (Ca2§247 Biochim. Biophys. Acta 860, 354-360. Ganong, B. R., Loomis, C. R., Hannun, Y. A., & Bell, R. M. (1986). Specificity and mechanism of protein kinase C activation by sn- 1,2-diacylglycerols. Proc. Natl. Acad. Sci. USA 83,1184-1188. Gazzotti, P., Bock, H.-G., & Fleischer, S. (1974). Role of lecithin in D-13-hydroxybutyratedehydrogenase function. Biochem. Biophys. Res. Commun. 58, 309-315. G6rrissen, H., Marsh, D., Rietveld, A., & De Kruijff, B. (1986). Apocytochrome c binding to negatively charged lipid dispersions studied by spin label electron spin resonance spectroscopy. Biochemistry 25, 2904-2910. Guy, H. R., & Conti, E (1990). Pursuing the structure and function of voltage-gated channels. Trends Neurosci. 13,201-206. Hannun, Y. A., Loomis, C. R., & Bell, R. M. (1985). Activation of protein kinase C by triton X-100 mixed micelles containing diacylglycerol and phosphatidylserine. J. Biol. Chem. 260, 1003910043. Hannun, Y. A., Loomis, C. R., & Bell, R. M. (1986). Protein kinase C activation in mixed micelles. Mechanistic implications of phospholipid, diacylglycerol, and calcium interdependencies. J. Biol. Chem. 261, 7184-7190. Heidmann, T., Oswald, R. E., & Changeux, J.-P. (1983). Multiple sites of action for noncompetitive blockers of acetylcholine receptor rich membrane fragments from Torpedo marmorata. Biochemistry 22, 3112-3127. Heimburg, T., Hildebrandt, P., and Marsh, D. (1991). Cytochrome c-lipid interactions studied by resonance Raman and 31p NMR spectroscopy. Correlation between the conformational changes of the protein and the lipid bilayer. Biochemistry 30, 9084-9089. Horv~ith, L. I., Brophy, P. J., & Marsh, D. (1988a). Influence of lipid headgroup on the specificity and exchange dynamics in lipid-protein interactions. A spin label study of myelin proteolipid apoprotein-phospholipid complexes. Biochemistry 27, 5296-5304. Horv~ith, L. I., Brophy, P. J., & Marsh, D. (1988b). Exchange rates at the lipid-protein interface of myelin proteolipid protein studied by spin-label electron spin resonance. Biochemistry 27, 46-52. Horv~ith, L. I., Drees, M., Beyer, K., Klingenberg, M., & Marsh, D. (1990a). Lipid-protein interactions in ADP-ATP carrier/egg phosphatidylcholine recombinants studied by spin label ESR spectroscopy. Biochemistry 29, 10664-10669. Horv~ith, L. I., Arias, H. R., Hankovszky, H. O., Hideg, K., Barrantes, F. J., & Marsh, D. (1990b). Association of spin-labeled local anesthetics at the hydrophobic surface of acetylcholine receptor in native membranes from Torpedo marmorata. Biochemistry 29, 8707-8713. Horv~ith, L. I., Brophy, P. J., & Marsh, D. (1990c). Influence of polar residue deletions on lipid-protein interactions with the myelin proteolipid protein. Spin label ESR studies with DM-20/lipid recombinants. Biochemistry 29, 2635-2638. Horv~ith, L. I., Arias, H. R., Schwarzmann, G., Sandhoff, K., Marsh, D., & Barrantes, E J. (1993). Ganglioside-protein interactions in acetylcholine receptor-rich membranes from Torpedo marmorata. ESR spin label studies., to be published.

L ipid-Protein Specificity

183

Horv~ith, L. I., Brophy, P. J., & Marsh, D. (1993a). Exchange rates at the lipid-protein interface of the myelin proteolipid protein determined by saturation transfer electron spin resonance and continuous wave saturation studies. Biophys. J. 64, 622-631. Horv~ith, L. I., Brophy, P. J., & Marsh, D. (1993b). Spin label saturation transfer ESR determinations of the stoichiometry and selectivity of lipid/protein interactions in the gel phase. Biochim. Biophys. Acta 1147, 277-280. Horv~th, L. I., Heimburg, T., Kovachev, P., Findlay, J. B. C., Hideg, K., & Marsh, D. (1995). Integration of a K+ channel-associated peptide in a lipid bilayer: conformation, lipid-protein interactions and rotational diffusion. Biochemistry, in press. Hovius, R., Lambrechts, H., Nicolay, K., & De Kruijff, B. (1990). Improved methods to isolate and subfractionate rat liver mitochondria. Lipid composition of the inner and outer membrane fractions. Biochim. Biophys. Acta 1021, 217-226. Isaacson, Y. A., Deroo, P. W., Rosenthal, A. E, Bittman, R., Mclntyre, J. O., Bock, H.-G., Gazzotti, P., & Fleischer, S. (1979). The structural specificity of lecithin for activation of purified D-13-hydroxybutyrate apodehydrogenase. J. Biol. Chem. 254, 117-126. Johannsson, A., Keightley, C. A., Smith, G. A., Richards, C. D., Hesketh, T. R., & Metcalfe, J. C. (1981). The effect of bilayer thickness and n-alkanes on the activity of the (Ca2§ + Mg2+)-dependent ATPase of sarcoplasmic reticulum. J. Biol. Chem. 256, 1643-1650. Kim, J., Mosior, M., Chung, L. A., Wu, H., & McLaughlin, S. (1991). Binding of peptides with basic residues to membranes containing acidic phospholipids. Biophys. J. 60, 135-148. Kimelberg, H., & Papahadjopoulos, D. (1974). Effects of phospholipid acyl chain fluidity, phase transitions and cholesterol on (Na§ + K§ adenosine triphosphatase J. Biol. Chem. 240, 1071-1080. Knowles, P. E, Watts, A., & Marsh, D. (1981). Spin-label studies of head-group specificity in the interaction of phospholipids with yeast cytochrome oxidase. Biochemistry 20, 5888-5894. Kuppe, A., Mrsny, R. J., Shimizu, M., Firsan, S. J., Keana, J. E W., & Griffith, O. H. (I 987). Labeling of bovine heart cytochrome c oxidase with analogues of phospholipids. Synthesis and reactivity of a new cardiolipin benzaldehyde probe. Biochemistry 26, 7693-7701. Kusumi, A., & Hyde, J. S. (1982). Spin-label saturation-transfer electron spin resonance detection of transient association of rhodopsin in reconstituted membranes. Biochemistry 21, 5978-5983. Lee, M.-H., & Bell, R. M. (1989). Phospholipid functional groups involved in protein kinase C activation, phorbol ester binding, and binding to mixed micelles. J. Biol. Chem. 264, 1479714805. Lewis, B. A., & Engelman, D. M. (1983). Bacteriorhodopsin remains dispersed in fluid phospholipid bilayers over a wide range of bilayer thicknesses. J. Mol. Biol. 166, 203-210. Li, G., Knowles, P. E, Murphy, D. J., Nishida, I., & Marsh, D. (1989). Spin label ESR studies of lipid-protein interactions in thylakoid membranes. Biochemistry 28, 7446-7452. Li, G., Knowles, P. E, Murphy, D. J., & Marsh, D. (1990a). Lipid-protein interactions in stacked and destacked thylakoid membranes and the influence of phosphorylation and illumination. Spin label ESR studies. Biochim. Biophys. Acta 1024, 278-284. Li, G., Knowles, P. E, Murphy, D. J., & Marsh, D. (1990b). Lipid-protein interactions in thylakoid membranes of chilling-resistant and-sensitive plants studied by spin label electron spin resonance spectroscopy. J. Biol. Chem. 265, 16867-16872. Lisanti, M. P., & Rodriguez-Boulan, E. (1990). Glycophospholipid membrane anchoring provides clues to the mechanism of protein sorting in polarized epithelial cells. Trends Biochem. Sci. 15, 113-118. London, E., & Feigenson, G. W. (1981). Fluorescence quenching in model membranes. 2. Determination of the local lipid environment of the calcium adenosine triphosphatase from sarcoplasmic reticulum. Biochemistry 20, 1939-1948. Mandersloot, J. G., Roelofsen, B., & de Gier, J. (1978). Phosphatidylinositol as the endogenous activator of the (Na§ + K+)-ATPase in microsomes of rabbit kidney. Biochim. Biophys. Acta 508, 478--485.

184

DEREK MARSH

Marsh, D. (1985). ESR spin-label studies of lipid-protein interactions. In: Progress in Protein-Lipid Interactions (Watts, A., and De Pont, J. J. H. H. M., eds.), Vol. 1, pp. 143-172, Elsevier, Amsterdam. Marsh, D. (1987). Selectivity of lipid-protein interactions. J. Bioenerg. Biomemb. 19, 677-689. Marsh, D. (1990). Lipid-protein interactions in membranes. FEBS Lett. 268, 371-375. Marsh, D. (1993). The nature of the lipid-protein interface and the influence of protein structure on protein-lipid interactions. In: New Comprehensive Biochemistry, Vol. 25: Protein-Lipid Interactions (Watts, A., ed.), pp. 41-66, Elsevier, Amsterdam. Marsh, D., & Powell, G. L. (1988). Properties of cardiolipin and functional implications for cytochrome oxidase activity. Bioelectrochem. Bioenerg. 20, 73-82. Marsh, D., Watts, A., Pates, R. D., Uhl, R., Knowles, P. E, & Esmann, M. (1982). ESR spin-label studies of lipid-protein interactions in membranes. Biophys. J. 37, 265-274. McMillen, D. A., Volwerk, J. J., Ohishi, J., Erion, M., Keanna, J. F. W., Jost, P. C., & Griffith, O. H. (1986). Identifying regions of membrane proteins in contact with phospholipid head groups: covalent attachment of a new class of aldehyde lipid labels to cytochrome c oxidase. Biochemistry 25, 182-193. Nozaki, Y., Chamberlain, B., Webster, R., & Tanford, C. (1976). Evidence for a major conformational change of coat protein in assembly of fl bacteriophage. Nature (London) 259, 335-337. Orr, J. W., & Newton, A. C. (1992a). Interaction of protein kinase C with phosphatidylserine. 2. Specificity and regulation. Biochemistry 31, 4667-4673. Orr, J. W., & Newton, A. C. (1992b). Interaction of protein kinase C with phosphatidylserine. 1. Cooperativity in lipid binding. Biochemistry 31,4661-4667. Palatini, P., Dabbeni-Sala, E, Pitotti, A., Bruni, A., & Mandersloot, J. C. (1977). Activation of (Na§ + K+)-dependent ATPase by lipid vesicles of negatively charged phospholipids. Biochim. Biophys. Acta 466, 1-9. Peelen, S. J. C. J., Sanders, J. C., Hemminga, M. A., & Marsh, D. (1992). Stoichiometry, selectivity, and exchange dynamics of lipid-protein interaction with bacteriophage M 13 coat protein by spin label electron spin resonance. Effects of protein secondary structure. Biochemistry 31, 2670-2677. Powell, G. L., Knowles, P. E, & Marsh, D. (1985). Association of spin-labelled cardiolipin with dimyristoylphosphaidylcholine-substituted bovine heart cytochrome c oxidase. A generalized specificity increase rather than highly specific binding sites. Biochim. Biophys. Acta 816, 191-194. Powell, G. L., Knowles, P. E, & Marsh, D. (1987). Spin-label studies on the specificity of interaction of cardiolipin with beef heart cytochrome oxidase. Biochemistry 26, 8138-8145. Raines, D. E., & Miller, K. W. (1993). The role of charge in lipid selectivity for the nicotinic acetylcholine receptor. Biophys. J. 64, 632--641. R6my, R., Tr6moli~res, A., Duval, J. C., Ambard-BretteviUe, E, & Dubacq, J. P. (1982). Study of the supramolecular organization of light-harvesting chlorophyll protein (LHCP). Conversion of the oligomeric form into the monomeric one by phospholipase A2 and reconstitution with liposomes. FEBS Lett. 137, 271-275. Rietveld, A., Sijens, P., Verkleij, A. J., & De Kruijff, B. (1983). Interaction of cytochrome c and its precursor apocytochrome c with various phospholipids. EMBO J. 2, 907-913. Rietveld, A., Jordi, W., & De Kruijff, B. (1986). Studies on the lipid dependency and mechanism of translocation of the mitochondrial precursor protein apocytochrome c across model membranes. J. Biol. Chem. 261, 3846-3856. Robinson, N. C. (1993). Functional binding of cardiolipin to cytochrome c oxidase. J. Bioenerg. Biomembr. 25, 153-163. Robinson, R. A., & Stokes, R. H. (1955). Electrolyte Solutions, Butterworths, London. Ryba, N. J. P., & Marsh, D. (1992). Protein rotational diffusion and lipid/protein interactions in recombinants of bovine rhodopsin with saturated diacylphosphatidylcholines of different chain-

Lipid-Protein Specificity

185

lengths studied by conventional and saturation-transfer electron spin resonance. Biochemistry 31, 7511-7518. Ryba, N. J. P., Horvfith, L. I., Watts, A., & Marsh, D. (1987). Molecular exchange at the lipid-rhodopsin interface: spin label electron spin resonance studies of rhodopsin-dimyristoylphosphatidylcholine recombinants. Biochemistry 26, 3234-3240. Sachse, J.-H., King, M. D., & Marsh, D. (1987). ESR determination of lipid translational diffusion coefficients at low spin label concentrations in biological membranes, using exchange broadening, exchange narrowing, and dipole-dipole interactions. J. Magn. Reson. 71,385-404. Sandermann, H. Jr., McIntyre, J. O., & Fleischer, S. (1986). Site-site interaction in the phospholipid activation of D-~-hydroxybutyrate dehydrogenase. J. Biol. Chem. 261,6201-6208. Sankaram, M. B., & Marsh, D. (1993). Protein-lipid interactions with peripheral membrane proteins. In: New Comprehensive Biochemistry, Vol. 25: Protein-Lipid Interactions (Watts, A., ed.), pp. 127-162, Elsevier, Amsterdam. Sankaram, M. B., Brophy, P. J., & Marsh, D. (1989a). Spin label ESR studies on the interaction of bovine spinal cord myelin basic protein with dimyristoylphosphatidylglycerol dispersions. Biochemistry 28, 9685-9691. Sankaram, M. B., Brophy, P. J., & Marsh, D. (1989b). Interaction of two complementary fragments of the bovine spinal cord myelin basic protein with phospholipid bilayers. An ESR spin-label study. Biochemistry 28, 9692-9698. Sankaram, M. B., Brophy, P. J., & Marsh, D. (1989c). Selectivity of interaction of phospholipids with bovine spinal cord myelin basic protein studied by spin-label electron spin resonance. Biochemistry 28, 9699-9707. Sankaram, M. B., de Kruijff, B., & Marsh, D. (1989d). Selectivity of interaction of spin-labelled lipids with peripheral proteins bound to dimyristoylphosphatidylglycerol bilayers, as determined by ESR spectroscopy. Biochim. Biophys. Acta 986, 315-320. Sankaram, M. B., Brophy, P. J., Jordi, W., & Marsh, D. (1990). Fatty acid pH titration and the selectivity of interaction with intrinsic proteins in dimyristoylphosphatidylglycerol dispersions. Spin label ESR studies. Biochim. Biophys. Acta 1021, 63-69. Sankaram, M. B., Brophy, P. J., & Marsh, D. ( 1991). Lipid-protein and protein-protein interactions in double recombinants of myelin proteolipid apoprotein and myelin basic protein with dimyristoyiphosphatidylglycerol. Biochemistry 30, 5866-5873. Sankaram, M. B., Brophy, P. J., Schwarzmann, G., Rotschiedt, D., Sandhoff, K., & Marsh, D. (1993). Glycosphingolipid specificity and binding of myelin basic protein to sulphatide and phosphatidylglycerol dispersions. ESR spin label studies, to be published. Scott, D. L., Otwinowski, Z., Gelb, M. H., & Sigler, P. B. (1990a). Crystal structure of bee-venom phospholipase A2 in a complex with a transition-state analogue. Science 250, 1563-1566. Scott, D. L., White, S. P., Otwinowski, Z., Yuan, W., Gelb, M. H., & Sigler, P. B. (1990b). Interfacial catalysis: the mechanism of phospholipase A2. Science 250, 1541-1546. Spruijt, R. B., & Hemminga, M. A. (1991). The in situ aggregational and conformational state of the major coat protein of bacteriophage M 13 in phospholipid bilayers mimicking the inner membrane of host E. coll. Biochemistry 30, 11147-11154. Spruijt, R. B., Wolf's, C. J. A. M., & Hemminga, M. A. (1989). Aggregation-related conformational change of the membrane-associated coat protein of bacteriophage M13. Biochemistry 28, 9158-9165. Swamy, M. J., & Marsh, D. (1993). Interaction of avidin with spin-labelled N-biotinyl phosphatidylethanolamine in a lipid membrane. FEBS Lett. 324, 56-58. Takumi, T., Ohkubo, H., & Nakanishi, S. (1988). Cloning of a membrane protein that induces a slow voltage-gated potassium current. Science 242, 1042-1045. Tomasi, M., & Montecucco, C. (1981). Lipid insertion of cholera toxin after binding to GMl-containing liposomes. J. Biol. Chem. 256, 11177-11181.

186

DEREK MARSH

TrEmoli~res, A., Dubacq, J.-P., Ambard-Bretteville, E, & Rtmy, R. (1981). Lipid composition of chlorophyll-protein complexes. Specific enrichment in trans-hexadecenoic acid of an oligomeric form of light-harvesting chlorophyll a/b protein. FEBS Lett. 130, 27-31. Van Wezenbeek, P. M. G. E, Hulsebos, T. J. M., & Schoenmakers, J. G. G. (1980). Nucleotide sequence of the filamentous bacteriophage M13 DNA genome: comparison with phage fd. Gene 11, 129-148. Watts, A., Volotovski, I. D., & Marsh, D. (1979). Rhodopsin-lipid associations in rod outer segment membranes. Identification of immobilized lipid by spin-labels. Biochemistry 18, 5006-5013. Weimbs, T., & Stoffel, W. (1992). Proteolipid protein (PLP) of CNS myelin: positions of free, disulfide-bonded, and fatty acid thioester-linked cysteine residues and implications for the membrane topology of PLP. Biochemistry 31, 12289-12296. Wheeler, K. P., Walker, J. A., & Barker, D. M. (1975). Lipid requirement of the membrane sodium-pluspotassium ion-dependent adenosine triphosphatase system. Biochem. J. 146, 713-722. White, S. P., Scott, D. L., Otwinowski, Z., Gelb, M. H., & Sigler, P. B. (1990). Crystal structure of cobra-venom phospholipase A2 in a complex with a transition-state analogue. Science 250, 1560-1563. Wolfs, C. J. A. M., Horv~ith, L. I., Marsh, D., Watts, A., & Hemminga, M. A. (1989). Spin-label ESR of bacteriophage M 13 coat protein in mixed lipid bilayers. Characterization of molecular selectivity of charged phospholipids for the bacteriophage M 13 coat protein in lipid bilayers. Biochemistry 28, 9995-10001.

EFFECTS OF LIPID-PROTEIN INTERACTIONS ON MEMBRANE FUNCTION

A. G. Lee

Io II. III. IV.

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Lipids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Viscosity-Dependent Effects . . . . . . . . . . . . . . . . . . . . . . . . . . . Effects of Lipids on Membrane Proteins . . . . . . . . . . . . . . . . . . . . . A. The Glucose Transporter . . . . . . . . . . . . . . . . . . . . . . . . . . B. The Nicotinic Acetylcholine Receptor . . . . . . . . . . . . . . . . . . . C. Rhodopsin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . D. Cardiolipin and Proteins of Mitochondria . . . . . . . . . . . . . . . . . z+ 2+ E. The (Ca - M g )-ATPase . . . . . . . . . . . . . . . . . . . . . . . . . . Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

i.

187 188 198 206 210 211 213 216 216 221 221

INTRODUCTION

M e m b r a n e proteins are unique in that they function in an e n v i r o n m e n t c o m p o s e d o f a p h o s p h o l i p i d bilayer. T h e majority o f the p h o s p h o l i p i d s s u r r o u n d i n g a protein

Biomembranes Volume 1, pages 187-224. Copyright 9 1995 by JAI Press Inc. All rights of reproduction in any form reserved. ISBN: 1-55938-658-4

187

188

A.G. LEE

in the membrane will interact with it nonspecifically; the membrane-penetrant part of the protein, consisting of stretches of hydrophobic amino acids arranged as m-helices, will interact with the fatty acyl-chains of the phospholipids, and the charged headgroups of the phospholipids will interact with charged amino acid residues at either boundary of the o~-helices. The phospholipids can thus be pictured as interacting with the membrane-penetrant surface of the protein, rather than as binding to discrete 'binding sites' on the protein. Such a picture is consistent with the data available from the crystal structure of the photosynthetic reaction center (Roth et al., 1989). Detergent molecules (N,N-dimethyldodecylamine N-oxide) surround the set of transmembranous m-helices as an irregular shell or annulus, and it is assumed that, in the native membrane, the phospholipid molecules are arranged in much the same way. As well as these solvent-like phospholipids, specific interactions of a few phospholipid molecules at specific sites on a protein are also possible, although difficult to establish unambiguously. It can be estimated that any biological membrane contains tens to hundreds of chemically distinct species of phospholipid. In part, this complexity reflects the complex fatty acid composition of the diet from which some of the fatty acids found in the membrane are derived. Although the lipid composition of the membrane has been found to vary with diet in very many systems, this variation is limited. It is clear that the lipid composition of the membrane is controlled by the cell; the limits acceptable to the cell have to be defined together with the reason for these limits and the mechanisms used by the cell to maintain the composition within these limits. A second reason for a complex lipid composition lies in the general role played by lipids in the physiology of the cell (e.g., signaling by phosphoinositides, the use of polyunsaturated fatty acids stored in the membranes for the synthesis of prostaglandins, etc.). Finally, the phospholipid composition of the membrane must be compatible with close to optimum functioning of the proteins in the membrane (Houslay and Stanley, 1982). It is this aspect of the problem that will be reviewed here.

!!. LIPIDS The most detailed information about the structures adopted by phospholipids has come from X-ray diffraction studies (Hauser et al., 1981). Crystal structures of dilauroylphosphatidylethanolamine [di(C12:0)PE] and dimyristoylphosphatidylcholine [di(C14:0)PC] are shown in Figure 1. In both, the orientation of the headgroup is approximately parallel to the bilayer surface. The thickness of the polar region of the lipids is ca. 10/k. The conformation of the glycerol diester group is such that the initial part of the sn-2 fatty acyl-chain extends parallel to the bilayer surface with a sharp bend at the second carbon to become parallel with the sn-1 chain. As a consequence, the sn-1 chain extends further into the bilayer than does the sn-2 chain by about three methylene units (3.7 ]k). In naturally occurring

Lipid-Protein Interactions and Membrane Function

(a)

DLPE

189

(b)

DMPC 2

~o

,,,o

,qo

o

3

Figure 1. Molecular conformations of (a) dilauroylphosphatidylethanolamine and (b) dimyristoylphosphatidylcholine, showing the two molecular arrangements found in the crystal. (Reproduced with permission from Hauser et al., 1981 ).

phospholipids this difference is minimized, because longer fatty acyl-chains are normally found at the sn-2 position than at the sn-1 position. Structures adopted by these phospholipids when dispersed as bilayers in water are very similar to those adopted in the crystal, with similar orientations of the headgroups and glycerol diester groups. In the liquid-crystalline phase, the phospholipids have considerable motional freedom including: diffusion in the plane of the membrane, rotation of the whole lipid molecule about its long axis, and rotation about C-C bonds in the fatty acyl-chains and in the headgroup region; for this reason, the lipids are said to be fluid. Nevertheless, the fatty acyl-chains have an average orientation that is still perpendicular to the bilayer surface. There is disagreement as to how the thickness of the fatty acyl-chain region of a phospholipid bilayer changes with chain length. Cornell and Separovic (1983) reported that the major change upon altering the fatty acyl-chain length was in the area occupied by the phospholipid at the phospholipid-water interface rather than in the thickness of the membrane, but the bulk of the data (Lewis and Engelman,

190

A.G. LEE

1983b; Sperotto and Mouritsen, 1988) suggests that, in the liquid-crystalline phase, the thickness, d (/~), of the hydrophobic region is a linear function of the acyl-chain length, nc, given by: d = 1.75 (nc-1)

(1)

This formula gives a thickness for the hydrophobic region of di(C18:1)PC very close to the most recent estimate (32 A) from X-ray diffraction measurements (Wiener and White, 1992). It is encouraging that this matches the estimated hydrophobic thickness of the photosynthetic reaction center (Michel and Deisenhofer, 1990). In the gel phase, the thickness is about 30% greater than in the liquid-crystalline phase (Sperotto and Mouritsen, 1988). Thus, in the liquid-crystalline phase, a change in fatty acyl-chain length from C18 to C14 results in a thinning of the bilayer by 7/~, and for a C14 chain, a transition to the gel phase from the liquid-crystalline phase results in a thickening of about 7 A. In describing motion in the liquid-crystalline phase, a major difficulty is in distinguishing between the rate of a motion and its extent or range. The term 'fluidity' strictly refers only to the rate of motion. Unfortunately, most methods for measuring rates of motion, e.g., nuclear magnetic resonance (NMR), electron spin resonance (ESR), and fluorescence depolarization, are sensitive to both rates and extents of motion, and separating the two is difficult. One might expect that, as a general rule, increasing rate and increasing extent of motion will go together; although this is probably true when, for example, the temperature of a lipid bilayer is changed, it cannot be assumed to be true when the perturbation consists of a change in the chemical composition of the system (a change in lipid structure, the addition of cholesterol, etc.). The extent of motion in the bilayer is normally characterized by a molecular order parameter S m~ describing the order of the long axis of the acyl-chain. The order parameter S~n~ for carbon atom j is defined as: SJn~ 1/2 <3 cos2~j- 1>

(2)

where 13j is the angle between the bilayer normal and the long molecular axis (defined as the normal to the plane spanned by the two C-H vectors of the carbon atom j of the fatty acyl-chain). ~j is also equal to the angle between the bilayer normal and the vector joining carbons j - 1 and j + 1. The NMR order parameter is defined by: SjN ~ = 1/2 <3 cos2~j- 1>

(3)

where ~j is the angle between the bilayer normal and the C-H vectors at carbon atom j. If the CH2 segment undergoes axially symmetric movement about the long molecular axis then the two definitions are related by: sjNMR= --0.5 SJn~

(4)

Lipid-Protein Interactions and Membrane Function

191

0.5 ~,~ 0.4

N

~ O.3 ~ O.2 0 O.I

.,...

I

2

....

I

l

6 LabeLed

l

.,

l

I0

l

1

14

carbon atom

Figure 2. Normalized order profiles for different bilayers. The molecular order parameter (Sm~ is shown as a function of chain position for di(C16:O)PC (o), (C16:0,C18:1)PC (z~),1,2-di(C16:O)PS (E3),and Acholeplasma laidlawi (x). Order parameters are compared at a fixed temperature of 18 ~ above the phase transitiontemperature of the lipid. (Reproduced with permission from Seelig and Browning, 1978).

The order parameter SJn~ = 1 in a perfectly ordered system, and sJn~ = 0 in a totally disordered system. Thus, in free solution where a molecule has freedom of motion around all possible axes of rotation (isotropic motion), the order parameter will be 0. In a crystal, where motion is highly restricted, the order parameter will have its maximum value of 1. The order parameter profiles for a number of phospholipid bilayers in the liquid-crystalline phase are remarkably similar (Figure 2). For all, a plateau region of constant order is followed by a region of rapidly decreasing order towards the center of the bilayer. The similarity of the order parameter profiles for the saturated and unsaturated phospholipids is particularly interesting. It shows that the order parameter is essentially determined by the distance from the lipid-water interface, and not by the geometry of the chain. Seelig and Waespe-Sarcevic (1978) have described the hydrocarbon chain interior of the bilayer as a series of strata running parallel to the bilayer surface; segments of the fatty acyl-chains that are located in the same stratum are, therefore, characterized by the same segmental order parameter.

192

A . G . LEE

It might have been thought that the presence of a single cis double bond towards the middle of a fatty acyl-chain would have led to a significant packing problem because of the sharp bend in the chain at the double bond. This, indeed, is the basis of the large decrease in phase transition-temperature observed on introduction of a single double bond into phosphatidylcholines or phosphatidylethanolamines. However, the data in Figure 2 suggest that a sharp bend is unlikely to be present in bilayers of phospholipids containing monounsaturated fatty acyl-chains in the liquid-crystalline phase. Presumably, the unsaturated fatty acyl-chain can adopt a "jog form" in which the two saturated portions of the chain are roughly parallel, although not colinear, with two bends in opposite senses adjacent to the two ends of the double bond (Applegate and Glomset, 1986). Of the unsaturated fatty acyl-chains, those of the o)-3 family are particularly important, especially 4, 7, 10, 13, 16, 19-docosahexaenoic acid (C22:6) found in phosphatidylethanolamine and phosphatidylserine of neuronal membranes, and in rod outer segment membranes. Since this cannot be synthesized de novo by animal cells, it has to be synthesized from linolenic acid (C18:3) obtained from plants by the addition of four carbon atoms at the carboxyl end of the acyl-chain, and the introduction of three extra cis double bonds. It is often assumed that the role of polyunsaturated fatty acyl-chains is to lower the gel-to-liquid-crystalline phase transition-temperatures and to increase the 'fluidity' of the bilayer. This, however, seems very unlikely. Thus, phase transition-temperatures for a series of phosphatidylcholines containing stearic acid at the sn-1 position, and unsaturated chains from oleic acid (C18:1) to arachidonic acid (C20:4) at the sn-2 position, show a minimum at -16 ~ for linoleic acid (C18:2) with an increase to -12 ~ for the more unsaturated species (Coolbear et al., 1983). Similarly, monolayer studies show that introduction of one double bond in the sn-2 position of phosphatidylcholine greatly increases the monolayer area, but that introduction of further double bonds have little further effect, with, in fact, substitution of a C18:1 chain by a C22:6 chain actually leading to a slight condensation of the monolayer (Demel et al., 1972). It seems likely, therefore, that any uniqueness of the C22:6 chain must lie in its packing properties rather than in its motional characteristics. Molecular modeling has, indeed, suggested unique conformational properties for chains such as linoleic acid or docosahexaenoic acid containing a 1,4-pentadiene structure (Applegate and Glomset, 1986). Figure 3 shows the minimum energy conformation calculated for 1,4-pentadiene. In this conformation, both carbon-carbon single bonds are rotated in the same direction by either +118 ~ or-118 ~ with the planes of the two double bonds being positioned essentially at right angles with respect to one another, the double-bond directions being parallel (Applegate and Glomset, 1986). In an NMR study of bilayers of (C16:0, C18:2A6'9)pc, it was suggested that the isolinoleic acid chain adopted just such a structure in the bilayer with the planes of the two double bonds being essentially at fight angles, the two double bond axes being parallel but tilted away from the bilayer normal (Baenziger et al., 1992; Figure 4). The orientation of the two double bonds is similar to that observed in crystals

Lipid-Protein Interactions and Membrane Function

193

Figure 3. Minimum energy conformation for 1,4-pentadiene, showing the spatial organization of the planes of the double bonds and the double bond directions. (Reproduced with permission from Applegate and Glomset, 1986).

of linoleic acid (C 18"2A9'12) with the methylene carbons on either side of the double bonds adopting a compact all-trans conformation that would pack well into a lipid bilayer (Ernst et al., 1979). The NMR data suggested jumping between two related conformations as shown in Figure 4, mainly arising from rotations about the C7-C8 and C8-C9 single bonds. Baenziger et al. (1992) argued that an acyl-chain containing three or more double bonds would be very unlikely to undergo similar jump motions, as this would tend to project the chain parallel to the bilayer surface. Thus, more unsaturated chains might adopt highly ordered, compact structures with relatively little local motion. Applegate and Glomset (1986) showed two likely low energy conformations for the (C22:6) chain (Figure 5). In the first, a helical conformation, the five-methylene carbon atoms intervening between the double bonds define a straight central axis, with the double bonds lying parallel to this axis. In the second possible conformation, shaped like an angle iron, the five-intervening methylene carbon atoms again define a straight central axis, but now, successive double-bond planes project outwards from the axis in two, rather than in four, different directions; double bonds

COOH

COOH a)

t I l

m m m t

0

CH~

CH3

~b

L~

b) ~k~ ' ~

c7)

~

CHS

COOH

Figure 4. Suggested model for the isolinoleoyi chain in (C16:0, C18:2A6'9)pc. The 1,4-pentadiene segment is proposed to adopt one of two conformations, the whole segment being tilted away from the bilayer normal. The double bond positions are parallel, with the planes of the double bonds being at 90 ~ with respect to each other. The two conformations interconvert at a fast rate (> 105 s-l). (Reproduced with permission from Baenziger et al., 1992).

194

Lipid-Protein Interactions and Membrane Function

195

A B 1

2

3

3

Figure 5. Possible extended conformations for the hexaenoic 19 carbon chain. (A) helical conformation and (B) angle iron-shaped conformation. (I)is a longitudinal view in the plane defined by three of the six double bonds. The molecules are oriented left to right from CI to the ethyl terminus. (2) is an oblique view. (3) shows cross sections (left) from the CI end and (right) from the ethyl terminus, sp2 carbons of the double bonds are indicated by dark-shading, other carbons by light-shading, and hydrogens are unshaded. (Reproduced with permission from Applegate and Glomset, 1986).

1, 3, and 5 form one plane and double bonds 2, 4, and 6 form a second plane perpendicular to the first. Simulations of diacylglycerols containing both saturated and C22:6 chains suggested that the angle iron-shaped conformation gave significantly better packing than the helical conformation (Applegate and Glomset, 1986). A conformation for the diacylglycerol was constructed based on the crystal structure of dilauroylphosphatidylethanolamine, in which the sn-1 acyl-chain and the glycerol group were aligned along a single axis, with the sn-2 chain initially projecting outward from this axis at an angle of about 90 ~, and then bending to become parallel to the sn-1 chain. With the saturated chain at the sn-1 position and the unsaturated chain at the sn-2 position, the tight bend in the sn-2 chain produced no packing problems and, indeed, the lipid could adopt a conformation that was almost as regular as that observed for crystalline dilauroylphosphatidylethanolamine. In contrast, with an arachidonic acid chain (C20:4) in the sn-2 position instead of the C22:6 chain, much less regular structures were predicted, partly due to the presence of the additional methylene group proximal to the A5 double bond in arachidonic acid, which precluded the sharp bend in the chain which was, however, compatible with the C22:6 chain as described (Applegate and Glomset, 1986). It was suggested

196

A.G. LEE

that both a A4 double bond and a terminal 03-3 double bond were required for optimal packing of a diacylglycerol (and presumably a phosphatidylethanolamine) containing an sn-2 polyunsaturated fatty acyl-chain (Applegate and Glomset, 1986). Since the choline headgroup occupies a significantly greater area in the membrane surface than an ethanolamine headgroup, structures adopted by phosphatidylcholines containing polyunsaturated fatty acyl-chains could be significantly different to those described above. It has been suggested that the relatively high degree of order observed in bilayers of lipids containing polyunsaturated fatty acyl-chains could be important (Baenziger et al., 1992). Although such lipids will be in the liquid-crystalline phase at physiological temperatures, the presence of polyunsaturation should maintain a high degree of local order with relatively restricted motions within the fatty acyl-chain region of the bilayer. It is possible that this is important in 'sealing' membrane proteins into the bilayer and reducing the permeability of the bilayer to small ions, but this has yet to be studied in any detail. Since the bilayer is anisotropic, rates of motion cannot be described by one single rate parameter and a single 'viscosity' value cannot accurately describe the resistance to motion (frictional forces) experienced by a molecule in the bilayer. The most common measure of membrane fluidity has been the extent of depolarization of the fluorescence of a probe molecule such as diphenylhexatriene (DPH) incorporated into the membrane, measured in a conventional steady-state fluorimeter. Initially, a 'microviscosity' was assigned to the membrane by finding an aliphatic oil of known viscosity in which DPH had the same fluorescence polarization value as in the membrane. This approach was flawed for two reasons. First, aliphatic oils of different structures, but the same viscosity, were shown to give different fluorescence polarization values; this, presumably, was a result of anisotropic motion of the DPH molecule. Second, measurements of the time-dependence of fluorescence depolarization showed that steady-state measurements were sensitive to both rates and extents of motion, and that, in fact, under many conditions the steady-state measurements were more sensitive to order than to rate. Time-dependent fluorescence depolarization experiments allow separation of rate and order. Such experiments with DPH have been interpreted in terms of 'wobbling in a cone', where the long axis of the DPH molecule is assumed to wobble freely in a cone of half-angle (x) around the normal to the plane of the membrane, with a wobbling diffusion coefficient (De0) (Kinosita et al., 1981; Stubbs et al., 1981). Values obtained for a series of phosphatidylcholines are given in Table 1. From the values of Dco, values of viscosity (1"1)were estimated using the equation: Oco = kT/6rl Vf

(5)

where V and f are the effective volume and a shape factor for DPH, respectively (Kinosita et al., 1981). The viscosity (1"1)represents the dynamic friction against the wobbling motion. Subsequent analysis showed that interpretation of the DPH

Lipid-Protein Interactions and Membrane Function

197

results was rather more complex than this, since DPH molecules were found to orient with their long axes both normal to and parallel to the plane of the bilayer. Straume and Litman (1987) analyzed their data using such a model, expressing the order not in terms of a half-cone angle (1:), but rather in terms of a fractional volume (fv) available for reorientation of the probe. Comparison of the data for di(C16:0)PC and di(C18:I)PC at 50 ~ at which temperature both phospholipids are in the liquid-crystalline phase, shows that introduction of the first cis double bond into the fatty acyl-chains results in a marked increase in the half-cone angle for wobbling; that is, in a decrease in order (Table 1). The analysis of Stubbs et al. (1981) also suggests that after the introduction of one cis double bond into the lipid fatty acyl-chains, introduction of further cis double bonds has relatively little effect on order in the bilayer (Table 1). Effects of cis double bonds on the rates of motion of DPH are rather modest, except for polyunsaturated fatty acyl-chains where an increased rate of motion is seen (Table 1). In contrast, an NMR study of fatty acyl-chain motion in a phosphatidylcholine containing an isolinoleoyl chain (C 18:2 66'9) suggests reduced motion of the unsaturated chain compared to a saturated chain, as might have been expected from the conformational properties of the unsaturated chain, as described above (Baenziger et al., 1992). The viscosity estimated for the phospholipid bilayer from the DPH results is ca. 0.1-0.2 P (Table 1). This viscosity corresponds essentially to the viscosity experienced when wobbling within the cone, about an axis perpendicular to the plane of the bilayer (Best et al., 1987). The viscosity corresponding to rotation of DPH about its long axis is very low, but that for diffusing in the plane of the membrane (lateral diffusion) is ca. 1 P. The viscosity experienced by a protein rotating in the plane of the membrane about its long axis is ca. 1-4 P. Best et al. (1987) have suggested that three viscosities are necessary to characterize motion in a bilayer. The first corresponds to motions involving positional displacements of phospholipid molecules,

Table 1. Motional Parameters for Phosphatidylcholine Bilayers at 37 ~ From Analysis of DPH Fluorescence Depolarization a Lipid Fatty Acyl Chains 2

"c(~

Do) (nsec-1)

1] (P)

C18:1

1

C18:1

77.6

0.215

0.18

C16:0

C18:1

69.8

0.209

0.18

C16:0

C18:2

68.6

0.241

0.16

C16:0 C 16:0 C18:1

C20:4 C 16:0b C18:1 b

74.5 60.9 84.8

0.283 0.290 0.340

0.14 0.13 0.11

Notes: aSeeStubb et al., 1981. bAt 50 ~

198

A.G. LEE

and would include translation and rotation of proteins, and translation of small molecules such as phospholipids (viscosity ca. 1-4 P). The second corresponds to motions coupled to orientational fluctuations of phospholipid molecules, such as wobbling of small molecules (viscosity ca. 0.2 P). Finally, motions such as rotation of small molecules about their long axes are unhindered by phospholipid molecules and, thus, for such motions, the viscosity vanishes. All these estimates of viscosity derive from studies of the motion of foreign molecules incorporated into lipid bilayers. An analysis of NMR spin-lattice relaxation times has suggested that trans-gauche isomerization rates (109- 101~ s-1) in fatty-acyl chains in a lipid bilayer are very similar to those in a free chain, and that the effective viscosity for the bilayer is ca. 0.01 P (Poastor et al., 1988). This argues that the internal dynamics of a fatty acyl-chain in a membrane are very similar to those in a neat alkane, despite the reduction in orientational freedom for the chain. The very large range of estimates for membrane viscosity reflects the dangers inherent in assuming that one single parameter can describe all motional properties of the bilayer. It is likely that strong interactions at the charged surface of the bilayer are a dominant factor in determining the rates of lateral diffusion of lipids and proteins in the membrane. 'Wobbling in a cone' of a lipid fatty acyl-chain is also likely to involve cooperative motion of a number of lipid molecules, making it less favorable than simple trans-gauche isomerization within a single fatty acyl-chain. It is not possible to say which estimate of viscosity is most 'relevant' for the function of a membrane protein. Intuitively, one feels that motions of small groups on the hydrophobic surface of the protein (rotation of-CH3 groups, etc.) might be sensitive to the trans-gauche isomerization rate in the fatty acyl-chains and, thus sense a small frictional force, whereas larger scale motions (conformation changes, etc.) might involve changes at the lipid-water interface and, thus experience larger frictional forces. A possible way to obtain information about phospholipid packing comes from application of free volume theory. This postulates that a process such as diffusion of phospholipids in the plane of the membrane will occur as the result of movement of phospholipids into voids in the membrane created by the random redistribution of the free volume within the membrane (Galla et al., 1979). The model should be applicable to any process requiting large amplitude molecular motions within the membrane. The greater the cohesive forces between molecules in the membrane, the smaller the coefficient describing the rate of lateral diffusion in the membrane, and the greater the phospholipid order parameter.

III.

VISCOSITY-DEPENDENT EFFECTS

Fluidity describes the ease of movement of a liquid and its inverse, viscosity, describes the resistance of a fluid to movement. The bulk viscosity of a liquid can be measured, for example, by measuring the rate at which a liquid falls under

Lipid-Protein Interactions and Membrane Function

199

gravitational pull, as in an Oswald viscometer. This viscosity describes the resistance which an ordinary liquid offers to its own flow or to the movement of a solid object through it. The rate of motion of an object in the liquid can be easily calculated from the viscosity. Thus, for example, the Stokes-Einstein equation gives the dependence of the diffusion coefficient (D) of a sphere on viscosity (1"1): D = kT/6xqa

(6)

where k is Boltzmann's constant, a is the radius of the sphere, and the viscosity is expressed as poise (P), with units of g m-is -1, or as cP, where l c P - 0.01P. Although this equation was formulated to describe motion of noninteracting spheres in a continuous medium (i.e., one in which the solvent molecules are much smaller than the size of the diffusing object), it has been shown to describe the motion of small molecules comparable in size to the molecules of solvent. However, as described below, it will not necessarily describe the effect of viscosity on the rates of rotations about bonds in a molecule, or the rates of conformational changes in a molecule. Rates of conformational changes are described in terms of transition state theory, but conventional transition state theory has to be modified to include the effects of viscosity. The progress of any change is pictured in transition state theory as in Figure 6, which shows how the energy of the system changes as the initial state (A) turns into the final state (B), the diagram illustrates equally well, a conformational change of A to B, or a chemical reaction with A (the reactant) and B (the product). In the course of the change, A distorts, the energy rises to a maximum (the transition state), and then falls as B is produced. The horizontal axis of the diagram represents the course of the change, and is called the reaction coordinate. For a reaction such as the hydrolysis of ATP (ATP ~ - ADP + Pi), in which the reactant is going to split into two parts, the reaction coordinate can be pictured as the distance between the atoms to be separated. For more complex processes, more than one dimension needs to be considered since the potential energy surfaces are multidimensional. It is then conventional to equate the reaction coordinate with the coordinate of motion that is most important in determining the rate of the reaction and to ignore the other coordinates (Frauenfelder and Wolynes, 1985). In transition state theory, a particular molecule going from A to B crosses the barrier only once before being trapped for a long time in state B; at some later time, the molecule might revert from state B back to state A, but this would then be a completely unconnected event. With this assumption, a rate constant can be calculated from the equilibrium concentration of the transition state. The rate of the reaction is a function of the activation energy: kAB "" AAB exp(-AH,~B/RT)

(7)

kBA -- ABA exp(-AH~A/RT)

(8)

and

200

A.G. LEE

kAB

>,,

,/kH*eAB

C7) c LU

kaA

AH,A ~

A,

----

~ -AH'BA

"-//

ZXi~ Reaction Coordinate Figure 6. The energy barrier over which a system has to go from state A to state B, and the effect of a change in the equilibrium properties of a system on the energy barrier. The enthalpy of activation in the forward direction from state A1 to B is AH AB, and from the modified state A2 is AH**AB. The enthalpy of activation in the reverse direction AH*BA is unaltered. The change in energy of state A relative to state B results in a change in the energy difference between A and B from AG~ to AG~ where kAB and kBA are the forward and backward rate constants, respectively, and AAB and ABA, and Z ~ k a and AH~A are pre-exponential factors and enthalpies of activation in the forward and backward direction, respectively. The pre-exponential factors are given by AAB = "r,k T / h exp(AS~.B)

(9)

and ABA = "r

exp(AS~A)

(~o)

where x is a transmission coefficient and AS~B and ASiA are the entropies of activation in the forward and backward directions, respectively. In transition state theory, x is made equal to 1. Possible effects of solvent on a reaction include both static and dynamic effects. Static effects of solvents on rates occur when solvent interacts preferentially with

Lipid-Protein Interactions and Membrane Function

201

one of the two states A or B or with the transition state (Figure 6). Figure 6 illustrates the consequences of changing to a solvent which interacts less well with state A than with state B, thus raising the energy of state A relative to that of B. An increase in Go (corresponding to an increase in the equilibrium constant B/A) will decrease the activation energy for the reaction A to B and hence will increase the rate kAB of the forward reaction, but will have no effect on the activation energy or rate kBA of the back reaction. This case is easily distinguished from other effects of solvent on rates since it results in a change in equilibrium constant. Static effects of solvent can also affect rates without affecting the equilibrium constant, if interactions of solvent with the transition state are different to those with the states A and B. For example, a solvent that lowers the energy of the transition state will increase the rates of both the forward and backward reactions, but will have no effect on the equilibrium constant. Thesestatic effects need to be distinguished from effects of solvent dynamics. Transition state theory takes no account of solvent viscosity (fluidity). This follows from a basic postulate of the theory that a particular molecule going from A to B crosses the barrier only once before being trapped for a long time in state B (Wigner, 1938). Kramers (1940) introduced frictional effects into kinetic theory by having the transmission coefficient = less than 1. Physically this means that even when the system is poised at the transition state and going from A to B, it will not necessarily get there, because the direction of motion may change to go back to A. Thus, even when the molecule has sufficient energy to overcome the barrier, it will not do so in one attempt. The frictional forces exerted by the environment will cause the molecule to undergo Brownian motion to and fro over the barrier. In the limit of high friction, Kramers (1940) showed that the transmission factor (x) will be inversely related to friction, so that: kAB = AAB/~exp(-AH~,B/RT)

(11)

where ~ is the friction coefficient. It is usually assumed that ~ is linearly related to the viscosity (11) of the solvent, so that the rate becomes inversely related to solvent viscosity: kAB = AAB/rl exp(-AH~,B/RT)

(12)

kBA = ABA/q exp(-AH~A/RT)

(13)

and correspondingly,

However, the equivalence between frictional effects on movement over a molecular energy barrier and bulk viscosity is not obvious, since movement over the energy barrier will often correspond to very small molecular changes, whereas viscosity is a measure of resistance to molecular motion through a solvent involving longtime collective motion of solvent molecules. This is clear if we consider the time scale of motion over an energy barrier. For chemical reactions, barriers are typically

202

A.G. LEE

very sharp, since potential energies vary by many kcals over fractions of/~, and the time scale for motion of an individual molecule in the region of the barrier top will be typically 10-12 to 10-13 s (Hynes, 1985) (the overall rate constant for a particular reaction will, of course, be many orders of magnitude slower than this because the reaction is rare, that is, a molecule only rarely has sufficient energy to overcome the activation barrier). Over very short time scales it is the short-time solvent response that will be important, not the long-time overall response given by the viscosity. What matters is what the solvent can do on this short time scale. In fact, for very sharp barriers there is insufficient time for completed collisions with solvent molecules and there is hardly any friction; the friction-independent transition state theory then describes the system very well. For broad energy barriers of the type observed for isomerizations, the time scale for barrier crossing will be longer, giving time for frictional effects to be felt, and rates may then become proportional to bulk viscosity. To account for these differences, it has been suggested that rates should be expressed as a power-law dependence on viscosity" k ~ 11-x

(14)

where x can be a value between 1 and 0 (Bagchi and Oxtoby, 1983). For x = 1, the full effect of solvent viscosity will be seen, whereas for x = 0, the rate will be viscosity-independent. Viscosity is temperature-dependent, and it has been shown that for many liquids this temperature-dependence can be expressed by the Andrado equation:

11 = rio exp(-En/RT)

(15)

where E n is the activation energy for viscous flow and 1"1ois a constant over a wide range of temperatures. Equation 15 can be combined with, for example, equations 12 and 14 to give: kAB = AAB/TI~ exp[-(AH~B + xEo)/RT]

(16)

The equation demonstrates that when the rate constant depends on viscosity, the activation energy for the process will include a term for viscous flow of the medium. It should, however, be emphasized that equation 16 includes viscosity-dependent terms in both the pre-exponential and exponential factors. Changing solvent will change both factors, 11o and E n. Both static and dynamics effects of solvents could be important in understanding effects on protein mobility. Protein motions involve atomic fluctuations, collective motions (both fast and slow, frequent and infrequent), and a variety of larger scale conformational changes (Ringe and Petsko, 1985; Karplus and Petsko, 1990; McCammon and Harvey, 1987; Brooks et al., 1988). Atomic fluctuations involve vibrational motions characterized by spacial displacements of ca. 0.01 to 1 ~, on a time scale of ca. 10-15 to 10-11 s. These small-amplitude fluctuations are essential to all other motions in the protein and Brooks et al. (1988) have described them as

Lipid-Protein Interactions and Membrane Function

203

the 'lubricant' that makes possible larger scale displacements, such as domain movement. Movement of side-chains and protein 'loops' involve the collective motion of groups of linked atoms, from a few, as in the rotation of a-CH3 group, to many hundreds, as in the movement of loops on the surface of the protein. The fastest of these collective motions, such as the oscillation of a small group within a single energy well, will occur on a time scale of 10-12 s. Other motions, such as the 'flipping' of aromatic side-chains will be slow because of a high activation barrier; an individual ring flip is fast, occurring on a time scale of 10-12 s, but the probability that a ring has sufficient energy to overcome the high activation barrier is small, giving a slow macroscopic rate (k -- 1 s-l). Larger scale motions, such as the movement of loops on the protein surface, will be slower, and are likely to be sensitive to environmental effects; such motions can result in the transmission of solvent effects from the external surface into the protein interior. Rigid-body motions, in which protein domains move relative to one another, are also likely to be affected by solvent since they involve movement of relatively large portions of the protein surface through the surrounding medium. Particularly important rigidbody motions are the domain movements observed in the opening and closing of the active-site regions of proteins such as hexokinase and liver alcohol dehydrogenase. Motion in proteins as pictured by molecular dynamics simulations has been summarized by McCammon and Harvey (1987) as follows" Individual atoms have vibrational motions that are rapid and of small amplitude (hundredths of a nm, hundredths of a ps) superimposed on the slower, larger amplitude motions of groups of atoms (motions on the order of 0.1 nm with characteristic times of about 10 ps). The differences in amplitude of the atomic displacements in different parts of a protein are primarily due to the slower, collective motions. Frequent collisions render the atomic motions chaotic, so group motions on the subpicosecond scale are diffusive, like those in a liquid. Over longer time scales, as groups of atoms wander away from their average, crystallographically observed positions, significant restoring forces come into play. The group motion is then oscillatory about the average position, but with significant damping from collisional effects. Since the magnitude and direction of atomic motions are dominated by group displacements about their mean positions on the time scale of 10 ps, atomic motion on this time scale has a solid-like character, particularly in the interior of globular proteins.

Over short time periods (< 10-12 s), small amplitude motions in proteins are similar to motions in a simple liquid. Each group is temporarily trapped, rattling in a cage made up of other neighboring groups in the protein and, at the surface of the protein, of molecules of the surrounding solvent (McCammon and Harvey, 1987). Frequent collisions of the cage atoms with the encaged group rapidly randomize its motion. Such collisions are the microscopic basis of the frictional forces that limit the rate of movement of the group. It is then important to understand the relative importance of the solvent and of neighboring groups on the protein in making up the cage. In a study of the rotation of Trp residues in small proteins, it was concluded that, at low temperatures, small fast motions of Trp residues were dominated by the frictional resistance of the solvent. However, as the temperature

204

A.G. LEE

was increased, the amplitude of motion increased, until the amplitude was such that further increases became limited by the local peptide environment. When it is the local peptide environment that limits motion, the peptide can be considered to be effectively unsolvated. The small fast motions of the Trp residue appeared to be strongly dependent on viscosity only for solvent viscosities of the order of several P. At viscosities on the order of 1 cP, the amplitude of the local motions was largely determined by neighboring groups on the peptide, which thus made up the effective 'solvent cage' (Rholam et al., 1984). The results of the study of Trp motion are consistent with the results of molecular dynamics simulations. Brooks and Karplus (Brooks and Karplus, 1989; Brooks et al., 1988) have concluded that for a solvent to have a significant effect on the motion of a particular group, not only must the group be in direct contact with the solvent (i.e., be exposed on the surface of the protein), but also the rate of motion of the group must be comparable to the rate of motion of the solvent (they must be dynamically coupled). Thus, small amplitude, high-frequency motions are independent of solvent, whether or not the group is on the surface of the protein, because their frequency is too high to couple with solvent motion. Motions within the protein interior are solid-like, the problem for a group being one of finding sufficient room to move. It is likely that small cavities exist inside proteins and, presumably, motion involves the movement of a group into these cavities with the consequent creation of new cavities, thereby allowing movement of further groups. The highly collective nature of motions within the protein explains their high activation energies and relative infrequency. Such motions are also unlikely to be affected by solvent. Protein motions that can become coupled with the solvent are large-scale fluctuations such as movement of loops on the surface of the protein or movement of domains, both of which involve movement of regions of the protein surface through the solvent (Brooks et al., 1988). Particular motions within the protein interior could then be affected by solvent if the motion is a collective one involving parts of the surface of the protein. It is clear that there is not going to be any simple, universal relationship between solvent viscosity and the rate of motion of any particular group in a protein; motion of some groups will be dependent on solvent viscosity, and that of others will be solvent independent. In the spirit of the approach of Kramers, Ansari et al. (1992) have distinguished between the frictional forces of the solvent, retarding the motion of atoms on the surface of the protein, and the internal friction of the protein, slowing the motion of protein atoms relative to each other. Assuming that the effects of the friction of the protein and solvent are additive, Ansari et al. (1992) described the rate constant in the high-friction limit for a protein conformation change as: B

k = t~'pt~+ (1 - ct.~. s)g exp(-Eo/RT)

(17)

where Eo is the average height of the potential energy barrier separating the protein conformations, B is a viscosity- and temperature-independent parameter that de-

Lipid-Protein Interactions and Membrane Function

205

pends on the shape of the potential surface, ~p is the friction-coefficient for motion in the protein, and ~s is the friction-coefficient for motion in the solvent; ~s will, by Stokes' law, be proportional to solvent viscosity. The parameter o~is the fraction of protein atoms involved in the conformation change that are not in contact with the solvent and accounts for the relative contributions of protein and solvent friction. Ansari et al. (1992) studied the effect of solvent viscosity on the rate of the major conformational change in myoglobin after ligand dissociation, probably corresponding to a small global displacement of protein atoms on the proximal side of the heme. They found that at low solvent viscosities, the rate of the conformational change was independent of viscosity, but at high viscosities, it depended on approximately the inverse first power of the viscosity. The data was fitted to a modified form of equation 17: k- C exp(-Eo/RT) o~ +'q

(18)

where C is a constant, 11 is the solvent viscosity, and ~ has the units of viscosity and is the contribution of the protein friction to the total friction. The value of was 4.1 cP. With this value for t~, the effect of solvent viscosity could be divided into three regions. In the first, below a viscosity of ca. 1 cP (the viscosity of water at 20 ~ solvent friction made only a very small contribution to the total friction, so that the rate was independent of viscosity. In the second region, between solvent viscosities of 1 and 15 cP, both protein and solvent friction contributed to decreasing the rate constant. Finally, in the third region, above 15 cP, the solvent friction dominated and the rate fitted to Kramer's equation. The results of Ansari et al. (1992) with myoglobin suggest that the rates of conformational changes involving fairly global displacements ofprotein atoms (not necessarily large displacements) in contact with solvent will be dependent on solvent viscosity when the latter is high. This is in agreement with the theoretical discussions presented above. Viscosity-dependence of the motion of surface-exposed loops on proteins has been demonstrated experimentally (Lee, 1991). As expected, different processes in the same protein can show different dependencies on solvent viscosity. Thus, in a study of the kinetics of oxygen and carbon monoxide rebinding to heme proteins after flash photolysis, Beece et al. (1980) found that solvent viscosity affected the rates of processes occurring on the surface of the protein, but did not affect the rates of processes occurring in the center of the protein. Beece et al. (1980), therefore, proposed a power-law dependence of rate on viscosity similar to that shown in equation 14 so that kAB = AAB/rlx exp(-AH~B/RT)

(19)

kBA = ABA/YIx exp(-zi/-/~A/RT)

(2o)

and

206

A.G. LEE

Recent studies have also shown that changing solvent can have large effects on rates of protein motion not due to changes in solvent viscosity, but rather to changes in the solvent dielectric constant. Changes in dielectric constant will have large effects on the interaction energies between charged groups on a protein and, since such groups tend to be concentrated near the protein surface, this effect will be strongest at the protein-solvent interface. Affleck et al. (1992) modeled the charged groups on a protein as point charges on the ends ofmechanical springs. It was shown that vibrational frequencies of oppositely charged residues increased with increasing solvent dielectric constant because of the reduced electrostatatic interaction between them. These conclusions were consistent with experimental data showing that protein motions were faster in more hydrophilic (higher dielectric constant) solvents. In summary, static effects of solvents need to be sorted out before any dynamic influences of solvents on rates can be clearly revealed. The most likely effect of a change in the solvent around a protein is a change in the relative energies of the different conformational states of the protein, and thus a change in the rates of one or more conformational transitions. These static effects can be recognized by the corresponding changes in equilibrium properties of the system. Changes in the viscosity of the system could also affect rates, but since viscosity is a dynamic property, changes in viscosity cannot affect any equilibrium property of the system. The order parameter is a thermodynamic property of the system which, in principle, could affect equilibrium properties of the system, but it is hard to imagine how such a change could come about. If both a change in rate and in order parameter are observed, then most likely these changes are a reflection of some other more fundamental change in the system, such as a change in cohesive forces in the system, which could lead to both a change in order parameter and, through changes in solvation, to a change in conformational state for a protein.

IV. EFFECTS OF LIPIDS ON MEMBRANE PROTEINS Since the phospholipid bilayer provides part of the environment of a membrane protein, it is possible that the viscosity of the bilayer might affect the rates of some of the processes carried out by the protein. As described (Lee, 1991), a test is to look for changes in equilibrium properties of the protein. If the equilibrium constant for a process changes, then changes in rate constant cannot be due to changes in viscosity, but must follow from differences in relative energies of the various states of the protein. Even if no change in equilibrium constant is observed, observed changes in rate constants may still not be due to changes in viscosity, since differences in solvation of the transition and initial and final states may be important. In general, it seems rather unlikely that the functions of membrane proteins will be affected by the kind of change in membrane viscosity (fluidity) seen in biological

Lipid-Protein Interactions and Membrane Function

207

systems. Although changes in rates have often been correlated with changes in membrane fluidity, the link need not be a causal one. Changes in the phospholipid or sterol composition of the membrane or binding of hydrophobic molecules to the membrane could result in changes in membrane fluidity. These changes in membrane composition could directly affect the conformational state of proteins in the membrane and provide a more likely explanation for any observed changes in function than a change in membrane viscosity (Lee, 1991). If phospholipids of different structure bind with different strengths to any two conformational states, E1 and E2, of a protein, then changing the phospholipid composition of the membrane will change the equilibrium constant El/E2. Since the equilibrium constant is equal to the ratio of the forward to the backward rate constant, a change in equilibrium constant implies that at least one of the two rate constants must have changed. If this rate constant is, or becomes, a slow step in the reaction sequence, then the rate of the overall reaction will be affected. Similarly, addition of any molecule to the membrane that binds differently to E 1 or E2 will affect the E l/E2 equilibrium and so will affect the overall rate. A particularly simple change in the phospholipid composition of a membrane is a change in the length of the fatty acyl-chains. As the average chain length for a phospholipid in a biological membrane is about C18, it seems likely that the thickness of the hydrophobic regions of membrane proteins will match that of a bilayer of phospholipids with C 18 chains. It is unlikely that any mismatch between the thickness of the hydrophobic region of a protein and of a phospholipid bilayer will result in significant exposure of these regions of the protein to water because the Gibbs' free energy of exposure of hydrophobic residues to water is high (Tanford, 1973). Rather, changes in conformation of either the protein or the phospholipid can be expected to minimize the mismatch. If the protein can adopt more than one conformation with different hydrophobic thicknesses, then a change in the conformational equilibrium could result. Since hydrophobic tx-helices in membrane proteins are generally flanked by charged residues, a major change in the length of the helices is unlikely. A tilting of the or-helices to change the effective 'thickness' of the hydrophobic region of the protein is more likely. Distortion of the fatty acyl-chains of the phospholipids on the surface of the protein is also likely. For a chain which is shorter in the liquid-crystalline phase than the hydrophobic thickness of the average membrane protein, the hydrophobic mismatch can be reduced by a reduction in the number of gauche conformations in the chain. However, the effective length of the fatty acyl-chain cannot be longer than that in the gel phase, so that a C 14 chain would be about the shortest that could be made to match a membrane protein of hydrophobic thickness 30 ,~. For a chain which is too long, an increase in the number of gauche conformations would effectively shorten the chain. Both could, however, result in unfavorable solvation of a protein. If different conformational states of the protein had different hydrophobic thicknesses, then the differences in solvation energies between phospholipids with different fatty acyl-chains would lead to changes in the equilibrium between the

208

A.G. LEE

conformational states; the state with the thinnest hydrophobic region would be favored by the phospholipid with the shortest chain. The argument can also be put in terms of surface pressure (Baldwin and Hubbel, 1985a,b). If the volume occupied by a protein in the membrane is different in different conformations, the conformational equilibrium will be affected by the surface tension at the lipid-protein interface. Effects of surface tension were first described for liquid-air interfaces (Adam, 1941), but the same arguments apply in principal for any interface, including the lipid-protein interface. Surface tension arises at an interface because of the tendency of molecules in a liquid to stick together. When a molecule is at a surface, it experiences a force which tends to pull it back into the bulk phase, because it has more neighboring liquid molecules in the bulk phase than it does at the interface. The force is attractive, and is referred to as a tensionmthe surface tension. Increasing the area of the surface involves doing work against the cohesive forces in the liquid, that is, against the surface tension. A tension is a negative pressure, and so an alternative description of surface tension is in terms of a surface pressure. To relate these effects to membrane proteins, we can consider a solid cylinder (the protein) embedded in a liquid (the lipid bilayer). The work done, dW, in forming a surface of area A is proportional to the area of the surface formed so that: dW=?A

(21)

where q( is the surface tension. For a cylinder of radius r and height h, the total surface free energy will be 2r~rh?. If the radius of the cylinder were to decrease by 8r, the surface free energy would decrease by -2rch?Sr. The tendency of the cylinder to lower its surface free energy by shrinking is counterbalanced by an excess pressure inside the cylinder as compared to outside. If the radius of the cylinder decreases by 8r, then the volume will decrease by 2rcrhSr. The increase in free energy (PAV) of the cylinder is AP(2rcrhSr). At equilibrium the two changes in free energy must add to zero, so that: 2nh?~)r = Atr2~rhSr

(22)

AP = ?/r

(23)

and so,

The change in surface free energy (AG) is in units of energy/mole, AG = 2Nox?h~r

(24)

where No is Avogadro's number. A membrane protein can be described as a cylinder of radius r and height h (typically 50 ,~) in a phospholipid bilayer. The interface between the protein and phospholipid fatty acyl-chains can then be characterized in terms of the interfacial

Lipid-Protein Interactions and Membrane Function

209

surface tension ~L (Baldwin and Hubbel, 1985a,b). For a finite interfacial tension, there exists a pressure differential across the interface so that: PP = PL + ~L/r

(25)

where Pp and PL are the pressures within the protein and lipid phases, respectively. The small radius of curvature of a membrane protein means that quite small interfacial tensions produce very large effective pressures on protein molecules. Thus, for a protein ofradius 20.&, a surface tension of 10 dyne/cm, typical of organic liquid interfaces, will produce a pressure of 50 atm. Poor solvation of a protein will lead to a high surface tension (~L) and thus, a high internal protein pressure, shifting the equilibrium in favor of the conformationof the protein with the smallest volume. If the two conformations differ in radius by 8r, then the free energy difference (AG~ between the two conformations becomes AG ~ = AG ~ + 2Non~LhSr

(26)

where AG ~ is the flee energy difference when 7PL is 0. A change in radius of 2 ,~ would correspond to an increase in circumference equivalent to about four lipid molecules in the bilayer around the protein. There is, as yet, no evidence that points to any marked change in the number of annular phospholipids between different conformation states for a protein, so this would seem to be the upper limit for the possible change in radius. A change in radius of 2 ]k with a surface tension (TPL)of 10 dyne/cm corresponds to a change in free energy of 9 kcals/mole. As described below for the (Ca2+-Mg2+)-ATPase, changing phospholipid from di(C18:1)PC to di(C14:1)PC changes the El/E2 equilibrium constant from 0.5 to 5.2, corresponding to a change in AG~ of 1.5 kcal/mole and equivalent to a change in surface tension of 1.7 dyne/cm. Thus, only very modest changes in surface tension would be required to explain the observed change in equilibrium constant. Differences in solvation could also result from changes in phospholipid headgroup or from addition of hydrophobic additives to the membrane. Thus a 'cone-shaped' phospholipid such as phosphatidylethanolamine favors the hexagonal Ha phase rather than the bilayer phase since it has a smaller optimum hydrophilic (headgroup) area than its optimum hydrophobic (fatty acyl-chain) area. Incorporation of phosphatidylethanolamine into a phospholipid bilayer will then increase the surface tension (surface free energy) of the membrane because of increased water contact with the hydrophobic area leading to an increased lateral compression of the bilayer. This could be the underlying reason for the decrease in membrane permeability observed upon addition of phosphatidylethanolamine (Papahadjopoulos and Watkins, 1967). A possible response of a membrane protein to poor solvation by phospholipid could be to increase its aggregation state in the membrane, since increased hydrophobic contact between protein molecules implies less protein-lipid contact. However, in general, it appears that changing fatty acyl-chain length does not change

210

A.G. LEE

the aggregation state of membrane proteins. Thus, both rhodopsin and bacteriorhodopsin remain freely dispersed in a variety of phospholipid bilayers (Lewis and Engelman, 1983a; Baldwin and Hubbell, 1985b).

A. The Glucose Transporter The glucose transporter from red blood cells is a passive transporter, facilitating the movement of glucose across a membrane, down its concentration gradient. Translocation involves the vectorial transfer of glucose from extracellular to intracellular binding sites via integral (internal bilayer) transport domains. There is strong evidence for the simultaneous existence of two substrate binding sites (influx and efflux). The transporter is predicted to contain 12 transmembrane or-helices, and these are presumed to be arranged to form a water-filled (or accessible) channel that spans the membrane. The sugar transporter has been reconstituted into sealed phospholipid vesicles and the rate of sugar transport measured to give both the turnover number of the transporter (Vmax/site) and the affinity for glucose (Can'uthers and Melchior, 1984; Tefft et al., 1986). Figure 7 shows results for a variety of phosphatidylcholines. At 60 ~ when all will be in the liquid-crystalline effect, the turnover number decreases in the sequence di(C18:0)PC > di(C16:0)PC = di(C18:lt)pc > di(C18:1)PC > di(C20:4)PC. This sequence shows no correlation between activity and fluidity, but rather suggests a relationship with bilayer thickness, optimal thickness being observed in di(C18:0)PC. However, the data in Figure 7 also suggests that the physical phase of the phospholipid, i.e., liquid-crystalline or gel, is important. Thus, in di(C14:0)PC or di(C16:0)PC, no activity is observed in the gel phase. In di(C18:0)PC, activity is seen in the gel phase, although it is considerably less than in the liquid-crystalline phase. Following the phase transition in di(C 14:0)PC, there is a decrease in Km for glucose, followed by a gradual increase with increasing temperature (Figure 8). In dimyristoylphosphatidylglycerol (di(C14:0)PG), transport activity is observed in the gel phase, although it is rather low and activity does not increase until a temperature considerably above that of the phase transition. There is, however, a change in Km at the phase transition. In a series of phospholipids containing C 14:0 fatty acyl chains, Vmax/Sitein the liquid-crystalline phase was found to decrease in the order PS > PA > PG > PC. The basis for this headgroup effect has not been established (Tefft et al., 1986). In summary, the transporter shows a clear chain length preference for activity, with the optimum chain being C18:0. Activity is not dependent on fluidity in the liquid-crystalline state, although activity in the phosphatidylcholines (but not in the phosphatidylglycerols) is decreased in the gel phase (Carruthers and Melchior, 1984; Tefft et al., 1986). The observed effects on Km values also demonstrate that effects of phospholipids cannot be described in terms of membrane viscosity, but have to be understood in terms of changes in conformational state for the trans-

Lipid-Protein Interactions and Membrane Function TEMP. |

1

1

!

!

60

-18

C !

20

211

1

I

I

1

60

I

1

o

O

-

DEL

_

DML

-23 -

0

oo

o

o

DPLIchol

22 l

2.9

l

l

1

1

3.4 I/K

1

l

I

1

1

2.9

1

I

~

-

3.6

X 103

Figure 7. Arrhenius plots of Vmax/sitefor D-glucose zero-trans exit from reconstituted

liposomes: DAL, diarachidonylphosphatidylcholine (di(C20:4)PC); DEL, dielaidoylphosphatidylcholine (di(C18:1t)pc); DML, dimyristoylphosphatidylcholine (di(C14:0)PC); DOL, dioleoylphosphatidylcholine (d i(C18 :1)PC); DPOL, dipalmitoleoylphosphatidylcholine (di(C16:1)PC); DPL, dipalmitoylphosphatidylcholine (di(C16:0)PC); DSL, disteaorylphosphatidylcholine (di(C18:0)PC). (Reproduced with permission from Carruthers and Melchior, 1984). porter. Any more detailed analysis of the effects ofphospholipids must await a more detailed kinetic analysis of transporter function.

B. The Nicotinic Acetylcholine Receptor The nicotinic acetylcholine receptor is a pentameric membrane protein made up of ~, 13, 7, and 8 subunits in a molar ratio of 2:1"1"1, respectively. The binding of acetylcholine to the receptor opens a cation-specific channel which is an intrinsic

212

A . G . LEE

=

5

E

3

d

..

100

"13 m :33 C3

80

m

60 40

20 .~..---':

o

z --.I C)

o -13 rm -.I

5 z

50 40 ~1~

E 30 20 10

0'2'0'4'0"60 TEMPERATURE,

"C

Figure 8. Values of Vmax and Km as a function of temperature for the glucose transporter reconstituted into di(C14:O)PC (solid line) or di(C14:0)PG (dashed line). Shown in the center is the integrated heat flow of the bilayer transition for each system, being approximately equal to the percent completion of the transition. (Reproduced with permission from Tefff et al., 1986).

part of the receptor. In the continued presence of an activating ligand, the ionic conductivity of the receptor decreases, a process referred to as receptor desensitization. Desensitization is correlated with a shift of the receptor from a state with low affinity for activating ligands to a state with high affinity. In a number of studies, it has been shown that the opening of the ion channel and the process of desensitization are sensitive to the structures of the phospholipids surrounding the receptor in the membrane (Fong and McNamee, 1986; McNamee and Fong, 1988; Sunshine and McNamee, 1992).

Lipid-Protein Interactions and Membrane Function

213

It has been shown that the presence of both negatively charged lipids and a neutral lipid are necessary for optimal ion flux response to agonist binding. Initial studies showed good fluxes in mixtures of phosphatidylcholine and phosphatidic acid containing cholesterol. Later studies showed that phosphatidic acid could be replaced by phosphatidylglycerol, phosphatidylinositol, or cardiolipin and that cholesterol could be replaced by other neutral lipids such as squalene or tocopherol (Fong and McNamee, 1986; Sunshine and McNamee, 1992). It is believed that the effects of cholesterol follow from direct binding to the receptor (Middlemas and Raftery, 1987; Jones and McNamee, 1988). It has been shown that local anesthetics compete for binding with spin-labeled analogs of cholesterol, suggesting that both might bind to the same sites on the receptor. Local anesthetics themselves are believed to bind to two classes of sites on the receptor. The first class of sites are of low affinity and are thought to be of relatively large number, possibly at the lipid-protein interface (Valenzuela et al., 1992); fluorescence quenching experiments suggest that cholesterol does not bind at the lipid-protein interface on the receptor (Jones and McNamee, 1988). The second class of sites for the local anesthetics are of high affinity and appear to be located on the extracellular side of the M 1 and M2 transmembrane helices of the ot-subunits, and near the intracellular side of the M2 transmembrane helices of all the subunits (Valenzuela et al., 1992). If the neutral lipids were to bind to these same sites, then binding between the m-helices in a relatively hydrophobic environment would be indicated. The nature of the interactions of negatively charged phospholipids with the receptor is unknown, but would appear to be relatively nonspecific. Receptor function would also seem to depend on the structure of the zwitterionic phospholipid in the system. It has been reported that the affinity-state transition occurs for the receptor reconstituted with dielaidoylphosphatidylcholine (di(C18: It)pc), but not di(C 18:1)PC (Fong and McNamee, 1986). It has also been reported that ion fluxes do not occur for the receptor reconstituted in mixtures of sterols, negatively charged phospholipids, and either di(C14:0)PC or di(C18:I)PC, but do occur in mixtures containing (C18:0,C18:1)PC (Criado et al., 1984). This would seem to indicate an unusual degree of sensitivity of the receptor to its lipid environment.

C. Rhodopsin In the retinal rod, absorption of light by rhodopsin leads within a few ms to a quasi-stable equilibrium between two photochemical intermediates, metarhodopsin I (~ax 478 nm) and metarhodopsin II (X,max380 nm). Metarhodopsin II appears to be identical to the form of rhodopsin which catalyzes the exchange of GTP for GDP on the G-protein, transducin. It has been shown that the metarhodopsin I-metarhodopsin II transition is lipid dependent. In the absence of lipid, bleaching of rhodopsin leads directly from metarhodopsin I to opsin and free retinalnnot to metarhodopsin II (Baldwin and Hubbell, 1985a). In the presence of lipid, the ratio of metarhodopsin I/metarhodopsin II is strongly dependent on the fatty acyl-chain

214

A.G. LEE

$ o) ROS membrones b) ROS phospholipids

c) di(22:6)PE/egg PC (1"1)

d) di(22:6)PC 0 .,,,.. 0

;>

e) di(22:6)PC/egg PC (1"1)

p-

=E

(L f) egg PE/egg PC (1"1) ~

~ l

0

g) egg PC

l

l

20

40

1

I

60 flO Time / ms

l

I

,, J

I00

120

140

Figure 9. Influence of phospholipid on the ratio of metarhodopsin !/metarhodopsin Ii. Flash photolysis transients were measured at 480 nm and the increase in photomultiplier (PMT) voltage at 480 nm following the flash (arrow) reflects the loss of metarhodopsin ! leading to the formation of metarhodopsin Ii. (Reproduced with permission from Wiedmann et al., 1988).

length. In di(C14:0)PC or di(C14:1)PC in the liquid-crystalline phase, very little metarhodopsin II is formed (Baldwin and Hubbell, 1985a, b; Wiedmann et al., 1988) and essentially none is formed in the gel phase. As shown in Figure 9, reasonable amounts of metarhodopsin II are formed when rhodopsin is reconstituted with egg phosphatidylcholine, but the amount formed is still less than that formed in the retinal rod. Increasing the chain length and unsaturation to di(C22:6)PC results in a very significant increase in metarhodopsin II. The highest level of metarhodopsin II is, however, seen upon reconstitution of rhodopsin with mixtures of egg phosphatidylcholine and di(C22:6)PE (Figure 9; Wiedmann et al., 1988). It has been shown that the low proportion of metarhodopsin II found in egg phosphatidylcholine can be attributed to a pK shift; the metarhodopsin I-metarhodopsin II transition is known to be pH-sensitive, and the effect of egg phosphatidylcholine is to shift the pK for metarhodopsin II production (Gibson and Brown, 1993). The effects ofphospholipids cannot, therefore, be viscosity-dependent. Figure 10 shows

Lipid-Protein Interactions and Membrane Function

215

that lipid asymmetry in the retinal rod membrane is not important; phospholipids are believed to be arranged asymmetrically between the two halves of the membrane, and yet the function of rhodopsin is identical to that observed in the native membrane after reconstitution with phospholipids extracted from the membranem a process during which the asymmetry will be lost. In di(C 14:0)PC in the liquid-crystalline phase, the rate of formation of metarhodopsin II has been reported to be about ten-fold slower than in the native system (Mitchell et al., 1992). Formation of metarhodopsin II is also slower for rhodopsin reconstituted with egg phosphatidylcholine, but only by a factor of three (Mitchell et al., 1992). In general, effects on the rate of the reverse metarhodopsin II-metarhodopsin I process are much more marked, giving rise to the large observed effects on the metarhodopsin I/metarhodopsin II ratio (Mitchell et al., 1990; Mitchell et al., 1992). The partial molar volume of rhodopsin increases from metarhodopsin I to II, and as a result, the conformational equilibrium is pressure-dependent; as pressure increases, the metarhodopsin II population decreases (Lamola et al., 1974). Working between 30 and 150 atm, Attwood and Gutfreund (1980) found that in the metarhodopsin II state, rhodopsin is expanded by 179 ]k3 relative to the metarhodopsin I state. At 1 atm and a temperature of 1 ~ a pressure of 150 atm was required to reduce Kequil by a factor of two. It is possible, therefore, that effects of phospholipid structure on the metarhodopsin I/metarhodopsin II equilibrium follow from changes in surface tension attributable to different solvation energies (Baldwin and Hubbel, 1985a,b). A similar idea has been proposed by Mitchell et al. (1990) who observed a correlation between the metarhodopsin I/metarhodospin II ratio in egg yolk phosphatidylcholine/cholesterol systems, and a fractional volume parameter (f) describing motion of the fluorescence probe diphenylhexatriene (DPH) in phospholipid bilayers, which held whether the system was modified by addition of cholesterol or by changing temperature (see Lee, 1991); the fractional volume parameter (f) gives a measure of the cohesiveness of the phospholipids in the bilayer. However, when a wider range of phospholipids was studied, the correlation broke down, although for each particular phospholipid, a correlation was observed in that phospholipid betweenfand gequil (Mitchell et al., 1992); other factors are also clearly involved. Possible changes in the state of aggregation of rhodopsin in the membrane with changing phospholipid composition are rather unclear. Measurements of the rate of rotation of rhodopsin in reconstituted systems in the liquid-crystalline state using ESR have suggested the presence of monomeric protein in di(C15:0)PC, but with aggregation increasing with either shorter or longer chains, aggregation being particularly marked in di(C12:0)PC (Ryba and Marsh, 1992). Similar studies on bacteriorhodopsin, but using electron microscopy have suggested that extensive aggregation only occurs in di(C 10:0)PC and di(C24:1)PC, and not in phosphatidylcholines with chain lengths between C12 and C22 (Lewis and Engelman, 1983a).

216

A.G. LEE

D. Cardiolipin and Proteins of Mitochondria The presence of cardiolipin in the inner mitochondrial membrane and the strong binding of cardiolipin to a variety of proteins in this membrane (substrate carriers, NADH dehydrogenase, cytochrome bcl, cytochrome c oxidase, ATP synthase) suggests a specific role for cardiolipin in the function of the membrane, although an effect on the permeability of the membrane to H + may also be important (Hoch, 1992). The effects of cardiolipin and other phospholipids on the functions of these proteins is, however, still relatively unclear. In part, this is because the strong binding of cardiolipin to the proteins makes it difficult to totally remove all the cardiolipin without denaturing the protein. Thus, although it has been reported that a small number of cardiolipin molecules are essential for the activity ofcytochrome c oxidase, it has also been reported that activity is observed if all the phospholipids around the oxidase are replaced with di(C14:I)PC; addition of cardiolipin to the oxidase reconstituted with di(C 14:0)PC resulted in a ca. twofold increase in activity (Abramovitch et al., 1990). These experiments also suggest that the activity of the oxidase is relatively insensitive to the fatty acyl-chain length of the surrounding, bulk phospholipids. It was originally reported that the ATP synthase lost activity on removal of phospholipid and that addition of cardiolipin to the delipidated ATP synthase restored activity, whereas addition of phosphatidylcholine did not. However, this appears to be an artifact of the reconstitution procedure. Reconstitution of the delipidated ATP synthase using octylglucoside led to preparations with maximal activities similar to those observed with cardiolipin, phosphatidylcholine, or phosphatidylethanolamine, although a lower Km for ATP was observed on reconstitution with cardiolipin (Laird et al., 1986). These studies also, therefore, suggest that the catalytic activity of the synthase is relatively insensitive to the headgroup composition of the bulk surrounding phospholipids. An absolute requirement of the ATP/ADP carrier for cardiolipin has, however, been demonstrated by Hoffmann et al. (1994). It has been suggested that this could be associated with the large number of lysine residues on the carrier. The effects of the structure of the cardiolipin molecule itself on the function of cytochrome c oxidase have been reported (Dale and Robinson, 1988). It has been shown that reconstitution of the oxidase with cardiolipins having C18"1 or C14:0 acyl-chains led to the same oxidase activities, but that cardiolipins with C6:0 chains gave lower activities (Dale and Robinson, 1988).

E. The (Ca2*-Mg2+)-ATPase The (CaZ+-Mg2§ of skeletal muscle sarcoplasmic reticulum (SR) is responsible for the accumulation of Ca 2+ within the lumen of the SR. Hydrolysis of one molecule of ATP leads to the accumulation of two Ca 2+ ions. It is thought that the mechanism of the ATPase involves two conformational states, E1 and E2 (Figure 10). In the E 1 conformation, the two Ca 2+binding sites per ATPase molecule

Lipid-Protein Interactions and Membrane Function

E1

l

E2 ~

-~ Ca2E 1'

= Ca2E 1 'MgATP

E2Pi ~

= Ca2E I " M g A T P

E2P ~

217

~ Ca2E 1 'P

l

Ca2E2'P

Figure 1#. Kinetic scheme for the (Ca2+-Mg2+)-ATPase.

are outward facing and of high affinity, whereas in the phosphorylated form of E2 (E2P), they are of low affinity and inward facing. Following binding of MgATP and Ca 2§ the ATPase undergoes a conformation change involving the relative movement of the nucleotide and phosphorylation domains on the ATPase, bringing the y-phosphate of bound MgATP close to the residue (Asp-351) on the phosphorylation domain which is to be phosphorylated. Phosphorylation of the ATPase is followed by loss of ADP and a conformation change to Ca2E2P; this is the transport step. Since the Ca 2+ binding sites on E2P are of low affinity, Ca 2+ is lost from Ca2E2P to the lumen of the SR, followed by dephosphorylation of the ATPase and recycling to E 1 (de Meis and Vianna, 1979; de Meis, 1981; Gould et al., 1986). The (Ca2+-Mg2+)-ATPase can be reconstituted into bilayers of defined phospholipid composition, so that the effects of phospholipid structure on ATPase activity can be studied. As shown in Figures 11 and 12, the activity of the ATPase in reconstituted systems is sensitive to both the fatty acyl-chain length and the phospholipid headgroup. The phospholipid supporting highest activity is di(C18:I)PC, and phosphatidylcholines with longer or shorter fatty acyl-chains support lower activities (Figure 11). Close to optimal ATPase activities are obtained in phosphatidylcholine bilayers with fatty acyl-chain lengths in the range C 1 6 C20, which matches the composition of the native SR membrane. Changing the phospholipid headgroup from phosphatidylcholine also leads to lower ATPase activities (Figure 12). Again, the observed dependence on headgroup is consistent with phosphatidylcholine being the predominant phospholipid in the native SR. These experiments provide no evidence that the composition of the native membrane is in any way special, since the activity of the ATPase in pure di(C18:I)PC is higher than that observed in the native membrane (Figure 12). The physical phase of the phospholipid is also important. Figure 13 shows the activity of the ATPase reconstituted with di(C18:I)PC and di(C14:0)PC as a function of temperature. The temperature of the gel to liquid-crystalline phase transition for di(C18:1)PC is-20 ~ so that over the whole temperature range in Figure 13, di(C18:1)PC will be in the liquid-crystalline phase and a simple gradual decrease in activity is seen with decreasing temperature, as expected for any enzyme. The picture is different for the ATPase reconstituted with di(C14:0)PC. This phospholipid has a phase transition at 24 ~ so that below 24 ~ it will be in

218

A.G. LEE

25 20 E S:)

15

,===m

4k=a , m om.

10

,I,=a

U

I 14

Chains I

I

I

I

16

18

20

22

24

Chain Length Figure 11. ATPase activities (IU/mg protein) for the (Ca2+-Mg2+)-ATPase reconstituted with phosphatidylcholines containing monounsaturated fatty acyl-chains of the given chain length. Activities measured at 37 ~ and MgATP = 2.1 mM, pH 7.2. (Lee, 1983).

the gel phase; as shown in Figure 13, ATPase activity is essentially zero below 24 ~ It can be concluded that phospholipids need to be in the liquid-crystalline phase for activity. There is, however, no evidence that the exact fluidity in the liquid-crystalline phase is important. Thus, as described below, many of the equilibrium properties of the ATPase are observed to change when the phospholipids change (Michelangeli et al., 1990; Michelangeli et al., 1991) showing that differences exist in the way different phospholipids interact with the various conformational states of the ATPase and, thus, that the dependence of ATPase activity on phospholipid structure cannot be due to effects of fluidity alone. Further, no correlation was observed between ATPase activity and lipid order parameter for the ATPase reconstituted into a variety of phospholipid bilayer systems (East et al., 1984). When the ATPase is reconstituted with the short-chain phospholipid, di(C 14:1)PC, all of the measurable steps of the reaction sequence (Figure 10) are modified. First, there is a shift in the El/E2 equilibrium towards El. Small differences between the E1 and E2 conformations are suggested by the value of the equilibrium constant E l/E2 (0.14 at pH 7.0), since this corresponds to a free energy

L i p i d - P r o t e i n Interactions a n d M e m b r a n e Function

219

///

0.8 ~

0

, m

0.6 0.4 0.2 0

/// /// /// /// /// /// /// /// /// /// /// /// /// /// /// /// /// /// /// / / / /

/ / / /

"//~

",///

/ / / /

/ / / / / / / / /

/// ///

PC

/ / / / / / / / / / / / / / /

/ / / / / / / / / / / / / / /

/ / / / / / / / / / / / / / /

PE

// // A/

PS

SR

Figure 12. ATPase activities (IU/mg protein) for the (Ca2+-Mg2+)-ATPase reconstituted with di(C18:1)phosphatidylcholine (PC), di(C18:1)phosphatidylethanolamine (PE), and brain phosphatidylserine (PS), and for the unreconstituted purified ATPase (SR). Activities measured at 37 ~ and MgATP = 2.1 mM, pH 7.2. (Lee, 1983).

difference between E1 and E2 of just 1.4 kcal/mole (AG ~ = -RTlnKequil). A small free energy difference between E1 and E2 means that small differences in the interaction energies of different phospholipids with E1 and E2 can result in relatively large changes in the equilibrium constant. If the hydrophobic portion of the ATPase is thinner in the E1 conformation than in the E2 conformation, substitution of di(C18:1)PC by di(C14:1)PC would be expected to shift the equilibrium toward E 1. The rate of phosphorylation of the ATPase by ATP decreases markedly on reconstitution with di(C14:1)PC, and the equilibrium constant for the phosphorylation of the ATPase by Pi is decreased. The most surprising observation is, however, that on reconstitution with di(C14:I)PC or di(C24:l)PC, the stoichiometry of Ca 2+ binding changes from the usual 2:1 to 1:1 (Michelangeli et al., 1990; Starling et al., 1993). Ca 2+binding to the ATPase is normally cooperative, and has been interpreted in terms of a sequence E1 --~ E1Ca -~ El'Ca --) El'Ca2, where a single Ca 2+binding site exists on E 1 and binding of Ca 2+ to this site results in a conformational change (E 1Ca --~ E l'Ca) and the creation of the second Ca 2+

220

A.G. LEE ,

I

,

1

~

I

~

1

"

i{C18:1)PC

6

,/

E m

4

diCC 14:01PC

,-.In=

o

g

0

0

10

20

30

40

50

Temperature (*C} Figure 13. Effect of temperature on the activity (IU/rng protein) of the (Ca2+-Mg2+)ATPase reconstituted with di(C18:1)PC and di(C14:0)PC. Activities measured at MgATP = 2.1 raM, pH 7.2 (Warren et ai., 1974).

binding site. Binding of a single Ca 2+ ion to the ATPase would be observed either if the E1Ca ~ El'Ca conformational change was prevented, or if a conformational change E1 ---) E l ' occurred in the absence of the first bound Ca 2+ ion, so that the second site was created with the first site empty and occluded. The observed Ca 2§ affinity of the single site on the ATPase reconstituted with di(C14:1)PC would be consistent with the latter model. The (Ca2+-Mg2+)-ATPase from red blood cells is significantly different from that of SR in that the former is sensitive to calmodulin and acidic phospholipids; both calmodulin and acidic phospholipids increase Vmaxfor the ATPase and increase the affinity of the ATPase for Ca 2+. Acidic phospholipids appear to bind to two sites on the ATPase, one specific for acidic phospholipids and located just upstream from the third-transmembrane cz-helix, the other being the calmodulin-binding domain (Brodin et al., 1992). The cytoplasmic loop forming the first of these sites contains a large number of charged residues and is absent from the SR (Ca2+-Mg2+)-ATPase. The SR (Ca2+-Mg2+)-ATPase also lacks the calmodulin-binding domain found at the C-terminus of the red blood cell (Ca2+-Mg2+)-ATPase.

Lipid-Protein Interactions and Membrane Function

221

SUMMARY It appears that membrane proteins are relatively insensitive to the chemical structures of the majority of the phospholipids that surround them in the membrane. As long as these bulk phospholipids lie within an acceptable range of compositions, which generally means a fatty acyl-chain length between about C 16 and C22, then close to optimum activities will be obtained. Some membrane proteins show specific interaction with a small number of phospholipid molecules, usually negatively charged phospholipids, such as cardiolipin or phosphatidylserine. The effects of phospholipids of suboptimal structures on the activities of membrane proteins are best understood in terms of static effects, that is, in terms of effects of phospholipids on the relative energies of the different conformational states of the proteins. The particular effects of a phospholipid of suboptimal structure on any particular protein are likely to be unique to that protein.

REFERENCES Abramovitch, D. A., Marsh, D., & Powell, G. L. (1990). Activation of beef-heart cytochrome c oxidase by cardiolipin and analogues of cardiolipin. Biochim. Biophys. Acta 1020, 34--42. Adam, N. K. (1941). Physics and Chemistry of Surfaces, Oxford University Press, Oxford. Affleck, R., Haynes, C. A., & Clark, D. S. (1992). Solvent dielectric effects on protein dynamics. Proc. Nad. Acad. Sci. USA 89, 5167-5170. Ansari, A., Jones, C. M., Henry, E. R., Hofrichter, J., & Eaton, W. A. (1992). The role of solvent viscosity in the dynamics of protein conformational changes. Science 256, 1796-1798. Applegate, K. R., & Glomset, J. A. (1986). Computer-based modelling of the conformation and packing properties of docosahexaenoic acid. J. Lipid Res. 27, 658-680. Attwood, P. V., & Gutfreund, H. (1980). The application of pressure relaxation to the study of the equilibrium between metarhodopsin I and II from bovine retinas. FEBS Lett. 119, 323-326. Baenziger, J. E., Jarrell, H. C., & Smith, I. C. P. (1992). Molecular motions and dynamics of a diunsaturated acyl chain in a lipid bilayer: implications for the role of polyunsaturation in biological membranes. Biochemistry 31, 3377-3385. Bagchi, B., & Oxtoby, D. W. (1983). The effect of frequency dependent friction on isomerization dynamics in solution. J. Chem. Phys. 78, 2735-2741. Baldwin, P. A., & Hubbell, W. L. (1985a). Effects of lipid environment on the light-induced conformational changes of rhodopsin. 1. Absence of metarhodopsin II production in dimyristoylphosphatidylcholine recombinant membranes. Biochemistry 24, 2624--2632. Baldwin, P. A., & Hubbell, W. L. (1985b). Effects of lipid environment on the light-induced conformational changes of rhodopsin. Biochemistry 24, 2633-2639. Beece, D., Eisenstein, L., Frauenfelder, H., Good, D., Marden, M. C., Reinisch, L., Reynolds, A. H., Sorensen, L. B., & Yue, K. T. (1980). Solvent viscosity and protein dynamics. Biochemistry 19, 5147-5157. Best, L., John, E., & Jahnig, E (1987). Order and fluidity of lipid membranes as determined by fluorescence anisotropy decay. Eur. Biophys. J. 15, 87-102. Brodin, P., Falchetto, R., Vorherr, T., & Carafoli, E. (1992). Identification of two domains which mediate the binding of activating phospholipids to the plasma-membrane Ca 2§ pump. Eur. J. Biochem. 204, 939-946.

222

A.G. LEE

Brooks, C. L., & Karplus, M. (1989). Solvent effects on protein motion and protein effects on solvent motion. J. Mol. Biol. 208, 159-181. Brooks, C. L., Karplus, M., & Pettitt, B. M. (1988). Proteins: A Theoretical Perspective of Dynamics, Structure and Thermodynamics, Wiley, New York. Carruthers, A., & Melchior, D. L. (1984). Human erythrocyte hexose transporter activity is governed by bilayer lipid composition in reconstituted vesicles. Biochemistry 23, 6901--6911. Coolbear, K. P., Berde, C. B., & Keough, K. M. W. (1983). Gel to liquid-crystalline phase transition of aqueous dispersions of polyunsaturated mixed-acid phosphatidylcholines. Biochemistry 22, 1466-1473. Cornell, B. A., & Separovic, E (1983). Membrane thickness and acyl chain length. Biochim. Biophys. Acta 733, 189-193. Criado, M., Eibl, H., & Bah'antes, E J. (1984). Functional properties of the acetylcholine receptor incorporated in model lipid bilayers. J. Biol. Chem. 259, 9188-9198. Dale, M. P., & Robinson, N. C. (1988). Synthesis of cardiolipin derivatives with protection of the free hydroxyl: its application to the study of cardiolipin stimulation of cytochrome c oxidase. Biochemistry 27, 8270-8275. de Meis, L. (1981). The Sarcoplasmic Reticulum, Wiley, New York. de Meis, L., & Vianna, A. L. (1979). Energy interconversion by the Ca2+-dependent ATPase of the sarcoplasmic reticulum. Annu. Rev. Biochem. 48, 275-292. Demel, R. A., Geurts Van Kessel, W. S., & Van Deenen, L. L. M. (1972). The properties of polyunsaturated lecithins in monolayers and liposomes and the interactions of these lecithins with cholesterol. Biochim. Biophys. Acta 266, 26-40. East, J. M., Jones, O. T., Simmonds, A. C., & Lee, A. G. (1984). Membrane fluidity is not an important physiological regulator of the (Ca2+-Mg2+)-dependent ATPase of sarcoplasmic reticulum. J. Biol. Chem. 259, 8070-8071. Ernst, J., Sheldrick, W. S., & Fuhrhop, J. H. (1979). The structures of the essential unsaturated fatty acids, crystal structure of linoleic acid as well as evidence for the crystal structure of alpha-linolenic acid and arachidonic acid. Z. Naturforsch. [B] 84, 701-711. Fong, T. M., & McNamee, M. G. (1986). Correlation between acetylcholine receptor function and structural properties of membranes. Biochemistry 25, 830-840. Frauenfelder, H., & Wolynes, P. G. (1985). Rate theories and puzzles of hemeprotein kinetics. Science 229, 337-345. Galla, H. J., Hartmann, W., Theilen, U., & Sackmann, E. (1979). On two-dimensional passive random walk in lipid bilayers and fluid pathways in biomembranes. J. Membr. Biol. 48, 215-236. Gibson, N. J., & Brown, M. E (1993). Lipid headgroup and acyl chain composition modulate the MI-MII equilibrium of rhodopsin in recombinant membranes. Biochemistry 32, 2438-2454. Gould, G. W., East, J. M., Froud, R. J., McWhirter, J. M., Stefanova, H. I., & Lee, A. G. (1986). A kinetic model for the Ca2§ + Mg2+-activated ATPase of sarcoplasmic reticulum. Biochem. J. 237, 217-227. Hauser, H., Pascher, I., Pearson, R. H., & Sundell, S. (1981). Preferred conformation and molecular packing of phosphatidylethanolamine and phosphatidylcholine. Biochim. Biophys. Acta 650, 21-51. Hoch, F. L. (1992). Cardiolipins and biomembrane function. Biochim. Biophys. Acta 1113, 71-133. Hoffmann, B., Stock, A., Schlame, M., Beyer, K., & Klingenberg, M. (1994). The reconstituted ADP/ATP carrier activity has an absolute requirement for cardiolipin as shown in cysteine mutants. J. Biol. Chem. 269, 1940-1944. Houslay, M. D., & Stanley, K. K. (1982). Dynamics of Biological Membranes, Wiley, Chichester. Hynes, J. T. (I 985). In: Theory of Chemical Reaction Dynamics Vol. IV (Baer, M., ed.), pp. 171-234, CRC Press, Boca Raton. Jones, O. T., & McNamee, M. G. (1988). Annular and nonannular binding sites for cholesterol associated with the nicotinic acetylcholine receptor. Biochemistry 27, 2364-2374.

Lipid-Protein Interactions and Membrane Function

223

Karplus, M., & Petsko, G. A. (1990). Molecular dynamics simulation in biology. Nature 347, 631-639. Kinosita, K., Kataoka, R., Kimura, Y., Gotoh, O., & Ikegami, A. (1981). Dynamic structure of biological membranes as probed by 1,6-diphenyl-1,3,5-hexatriene: a nanosecond fluorescence depolarization study. Biochemistry 20, 4270-4277. Kramers, H. A. (1940). Brownian motion in a field of force and the diffusion model of chemical reactions. Physica. 7, 284-304. Laird, N. M., Parce, J. W., Montgomery, R. L., & Cunningham, C. C. (1986). Effect of phospholipids on the catalytic subunits of the mitochondrial F0.FI-ATPase. J. Biol. Chem. 261, 14851-14856. Lamola, A. A., Yamane, T., & Zipp, A. (1974). Effects of detergents and high pressure upon the metarhodopsin I-II equilibrium. Biochemistry 13, 738-745. Lee, A. G. (1983). In: Membrane Fluidity in Biology Vol. 2 (Aloia, R.C., ed.), pp. 43-88, Academic Press, New York. Lee, A. G. (1991). Lipids and their effects on membrane proteins: evidence against a role for fluidity. Prog. Lipid Res. 30, 323-348. Lewis, B. A., & Engelman, D. M. (1983a). Bacteriorhodopsin remains dispersed in fluid phospholipid bilayers over a wide range of bilayer thicknesses. J. Mol. Biol. 166, 203-210. Lewis, B. A., & Engelman, D. M. (1983b). Lipid bilayer thickness varies linearly with acyl chain length in fluid phosphatidylcholine vesicles. J. Mol. Biol. 166, 211-217. McCammon, J. A., & Harvey, S. C. (1987). Dynamics of Proteins and Nucleic Acids, Cambridge University Press, Cambridge. McNamee, M. G., & Fong, T.M. (1988). In: Advances in Membrane Fluidity, Vol. 2 (Aloia, R.C., Curtain, C.C., & Gordon, L.M., eds.), pp. 43-62, Alan R. Liss, New York. Michel, H., & Deisenhofer, J. (1990). The photosynthetic reaction center from the purple bacterium Rhodopseudomonas viridis: aspects of membrane protein structure. Curr. Top. Memb. Transp. 36, 53-69. Michelangeli, E, Orlowski, S., Champeil, P., Grimes, E. A., East, J. M., & Lee, A. G. (1990). Effects of phospholipids on binding of calcium to the (Ca2+-Mg2+)-ATPase. Biochemistry 29, 8307-8312. Michelangeli, E, Grimes, E. A., East, J. M., & Lee, A. G. (1991). Effects of phospholipids on the function of the (Ca2+-Mg2+)-ATPase. Biochemistry 30, 342-351. Middlemas, D. S., & Raftery, M. A. (1987). Identification of subunits of acetylcholine receptor that interact with a cholesterol photoaffinity probe. Biochemistry 26, 1219-1223. Mitchell, D. C., Straume, M., Miller, J. L., & Litman, B. J. (1990). Modulation of metarhodopsin formation by cholesterol-induced ordering of bilayer lipids. Biochemistry 29, 9143-9149. Mitchell, D. C., Straume, M., & Litman, B. J. (1992). Role of sn-l-saturated, sn-2-polyunsaturated phospholipids in control of membrane receptor conformational equilibrium: effects of cholesterol and acyl chain unsaturation on the metarhodopsin I-metarhodopsin II equilibrium. Biochemistry 31,662-670. Papahadjopoulos, D., & Watkins, J. C. (1967). Phospholipid model membranes. II. Permeability of hydrated liquid crystals. Biochim. Biophys. Acta 135, 639-652. Poastor, R. W., Venable, R. M., Karplus, M., & Szabo, A. (1988). A simulation based model of NMR T1 relaxation in lipid bilayer vesicles. J. Chem. Phys. 89, 1128-1140. Rholam, M., Scarlata, S., & Weber, G. (1984). Frictional resistance to the local rotations of fluorophores in proteins. Biochemistry 23, 6793-6796. Ringe, D., & Petsko, G. A. (1985). Mapping protein dynamics by X-ray diffraction. Prog. Biophys. Mol. Biol. 45, 197-235. Roth, M., Lewit-Bentley, A., Michel, H., Deisenhofer, J., Huber, R., & Oesterhelt, D. (1989). Detergent structure in crystals of a bacterial photosynthetic reaction centre. Nature 340, 659-662. Ryba, N. J. P., & Marsh, D. (1992). Protein rotational diffusion and lipid/protein interactions in recombinants of bovine rhodopsin with saturated diacylphosphatidylcholines of different chain lengths studied by conventional and saturation-transfer electron spin resonance. Biochemistry 31, 7511-7518.

224

A.G. LEE

Seelig, J., & Browning, J. L. (1978). General features of phospholipid conformation in membranes. FEBS Lett. 92, 41--44. Seelig, J., & Waespe-Sarcevic, N. (1978). Molecular order in cis and trans unsaturated phospholipid bilayers. Biochemistry 17, 3310-3315. Sperotto, M. M., & Mouritsen, O. G. (1988). Dependence of lipid membrane phase transition temperature on the mismatch of protein and lipid hydrophobic thickness. Eur. Biophys. J. 16, 1-10. Starling, A. P., East, J. M., & Lee, A. G. (1993). Effects of phosphatidylcholine fatty acyl chain length on calcium binding and other functions of the (Ca2+-Mg2+)-ATPase. Biochemistry 32, 15931600. Straume, M., & Litman, B. J. (1987). Equilibrium and dynamic structure of large, unilamellar, unsaturated acyl chain phosphatidylcholine vesicles. Higher order analysis of 1,6-diphenylhexatriene and trimethyamino-diphenylhexatriene anisotropy decay. Biochemistry 26, 5113-5120. Stubbs, C. D., Kouyama, T., Kinosita, K., & Ikegami, A. (1981 ). Effect of double bonds on the dynamic properties of the hydrocarbon region of lecithin bilayers. Biochemistry 20, 4257-4262. Sunshine, C., & McNamee, M. G. (1992). Lipid modulation of nicotonic acetylcholine receptor function: the role of neutral and negatively charged lipids. Biochim. Biophys. Acta 1108, 240-246. Tanford, C. (1973). The Hydrophobic Effect, Wiley, New York. Tefft, R. E., Carruthers, A., & Melchior, D. L. (1986). Reconstituted human erythrocyte sugar transporter activity is determined by bilayer lipid head groups. Biochemistry 25, 3709-3718. Valenzuela, C. E, Kerr, J. A., & Johnson, D. A. (1992). Quinacrine binds to the lipid-protein interface of the Torpedo acetylcholine receptor: a fluorescence study. J. Biol. Chem. 267, 8238-8244. Warren, G. B., Toon, P. A., Birdsall, N. J. M., Lee, A. G., & Metcalfe, J. C. (1974). Reversible lipid titrations of the activity of pure adenosine triphosphatase-lipid complexes. Biochemistry 13, 5501-5507. Wiedmann, T. S., Pates, R. D., Beach, J. M., Salmon, A., & Brown, M. F. (1988). Lipid-protein interactions mediate the photochemical function of rhodopsin. Biochemistry 27, 6469-6474. Wiener, M. C., & White, S. H. (1992). Structure of a fluid dioleoylphosphatidylcholine bilayer determined by joint refinement of X-ray and neutron diffraction data. Biophys. J. 61,434--447. Wigner, E. P. (1938). The transition state method. Trans. Faraday Soc. 34, 29-41.

GENERAL PRINCIPLES OF MEMBRANE TRANSPORT

lan C. West

II.

Ill.

Basal Transport Properties of Biological Membranes . . . . . . . . . . . . . . A. Molecular Process of Diffusion . . . . . . . . . . . . . . . . . . . . . . . B. Rate o f Basal Diffusion . . . . . . . . . . . . . . . . . . . . . . . . . . . Protein Catalysts o f Diffusion . . . . . . . . . . . . . . . . . . . . . . . . . . A. W h y Catalyze Transport? . . . . . . . . . . . . . . . . . . . . . . . . . . B. Strategies of Transport Catalysts . . . . . . . . . . . . . . . . . . . . . . C. Functional Considerations . . . . . . . . . . . . . . . . . . . . . . . . . . D. Structural Considerations . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Biomembranes Volume 1, pages 225-243. Copyright 9 1995 by JAI Press Inc. All rights of reproduction in any form reserved. ISBN: 1-55938-658-4

225

226 226 227 228 228 230 234 239 242 242

226

IAN C. WEST

BASAL TRANSPORT PROPERTIES OF BIOLOGICAL MEMBRANES A. Molecular Processof Diffusion Molecular Models of Diffusion It is worth spending a little time to consider how a molecule moves across a biological phospholipid bilayer membrane by simple (nonfacilitated) diffusion. Should we visualize this diffusion as resembling, on the one hand, a scaled down version of macroscopic diffusion through a (viscous) fluid, or, on the other hand, diffusion through a structured polymer matrix? In other words, is there any process akin to viscous flow (where some of the matrix/medium is seen as flowing with the diffusing molecule), or must a hole appear in the matrix adjacent to a diffuser? These questions have been thoroughly studied by Lieb and Stein (1986) over many years and are well reviewed in Stein (1986). The answer that emerges from kinetic studies (see below) seems to be that, though the movement of solutes through biological membranes in many respects resembles diffusion through a 4 nm layer of light parafin oil (e.g., hexadecane) of viscosity 2 poise, nevertheless, the macroscopic viscosity model is inadequate for diffusants as small as single molecules of solute, which see the membrane as a structured matrix, necessitating the formation of holes.

Forces Leading to Passive Diffusion Let us next take a brief look at the fundamental forces that cause osmosis, diffusion, and osmotic pressure. Imagine a membrane (Figure 1) composed of ten leaflets separating a 1 mM solution (of solute X in solvent Y) in the left hand phase, from a 10 mM solution in the fight hand phase. If the partition coefficient of X between Y and membrane (M) were unity, we would expect at steady-state, something close to 1 mM X in layer 1 and 10 mM X in layer 10. Once a steady-state was established there would be a net flow of X from fight to left and, incidentally, a net flow of Y in the opposite direction. Layer 5 would lose and gain molecules of X to layer 6 on the one side and layer 4 on the other, respectively. At 293 K all the molecules (solute and solvents) will be jigging around with kinetic energy ofkTjoules. Each molecule of X will have the same average velocity (this will depend on the size of the molecule, and the structure and viscosity of the solvent and membrane matrix). However, for purely numerical reasons, and because we are talking about a steady-state, layer 5 will lose molecules to layer 4 and gain (on average) the same number of molecules from layer 6. (Note that with this concept we have in essence predicted a linear gradient and Fick's law.)

General Principles of Membrane Transport

5

6

227 I I

8,9

10

10 m M X

5 --->

1 mMX

5

i

I I

I i

Figure 1. Diffusion is driven statistically by random-walk.

The energy required to overcome friction comes from the thermal energy of the system, but that has no directional vector and can not do work. If a force is required to cause net solute flow, the force here is this purely statistical or entropic consideration. It has been called an imaginary force. It can, nevertheless, do real work. B. Rate of Basal Diffusion

Two factors clearly control the rate of diffusion of small molecules through membranes; hydrophilicity and size. The two factors can be separated by studying similarly sized molecules of varying hydrophilicity, or series of molecules of similar hydrophilicity, but increasing size.

Hydrophilicity We have seen above that solutes diffusing through biological membranes encounter these as resembling light machine oil, and that the concentration of solute molecules in the membrane depends on the partitioning of solute between the membrane and external solution (essentially water). Numerous studies have shown a strong correlation between rates of diffusion across biological membranes and oil/water partition coefficients (over l O00-fold range). There is, however, some discussion as to whether octanol, olive oil, or hexadecane most closely resembles any one particular membrane (Stein, 1981). These experiments can be understood

228

IAN C. WEST

in terms of a model in which the diffusing solute, in order to cross the membrane, must first sever its links (H-bonds, dipolar attractions) with water (Stein, 1967).

Solute Size Once the solute has partitioned into the membrane, other factors, of course, play a part in determining diffusion rates--factors such as the size of the diffusant and the physical structure of the membrane. Stein and colleagues have shown that after correcting for partitioning effects, there is a steep inverse relationship between permeability and molecular size. Stokesian diffusion holds where the diffusion coefficient (D) is related to radius (r) of a diffusing sphere by: D = kT/6rtrl r

(1)

(where 1"1is the coefficient of viscosity). In biological membranes, as in synthetic polymer membranes, the dependence on r is much steeper than r-l; more like r -3 (Stein, 1986). This is consistent with the notion that improbable gaps must appear between lipid chains to allow the passage of diffusant; bigger gaps for bigger diffusants.

Activation Energy If the process of simple (basal, unfacilitated) diffusion through biological membranes involves, as a rate-limiting step, the opening of a cavity between acyl-chains of bilayer phospholipids, it should be expected to show an appreciable activation energy, and a ten degree temperature coefficient (Ql0) of two or greater, as is characteristic of protein-mediated transport processes and enzymic reactions. This is, indeed, the case (Lieb and Stein, 1986). It is, therefore, not possible to discriminate between basal and carrier mediated routes by means of activation energies or al0.

ii.

PROTEIN CATALYSTS OF D I F F U S I O N

A. Why Catalyze Transport? There are a number of reasons why the basal route for transmembrane traffic is unsuitable. For some compounds, it is too slow; inflow of monosaccharide and disaccharide sugars, for example, would be very slow. For other compounds, basal transport is too fast; it has fairly recently been appreciated that many cells contain proteins engaged in active extrusion of a wide range of hydrophobic compounds which the cell membrane is unable to exclude. A third reason for invoking a protein-mediated route is thermodynamic; the basal route can never maintain a thermodynamic potential at the destination side of the membrane greater than that

General Principles of Membrane Transport

229

at the departure side. For uncharged solutes, thermodynamic potential more or less means log concentration. Such a positive concentration gradient is often advantageous; for example, in trapping nutrients from a dilute milieu when the cytoplasmic concentration is optimally greater (and occasionally much greater) than that outside the cell. Similarly, the active extrusion of certain toxins or waste-products into an environment already containing supra-optimal concentrations can clearly be advantageous. Likewise, volume homeostasis will always require the active pumping of solute, and pH homeostasis often will. For ions, thermodynamic potential is often dominated by electrical terms, particularly for solutes carrying more than one (net) charge. Thus, Ca 2+, excluded alike from animal, plant, and bacterial cells, will have to be pumped against a large thermodynamic potential. Incidentally, if transport catalysts have to show at least some specificity, this is even more the case for these free-energy consuming active pumps. Following our analysis above, rate enhancement will be achieved by raising the concentration of the diffusing species in the membrane, or alternatively (or additionally), by providing a diffusion pathway in which there are fewer, or lower, obstacles. For hydrophilic substrates, this will more or less amount to the provision of proteins, embedded in and traversing the membrane, that contain pores or channels lined with polar amino acid residues m residues that can donate and accept H-bonds with water or with the substrate. There is some evidence from electron microscopy for the existence of hydrophilic (stain-filled) pores or channels through certain types of transport catalysts where the pore is rather large, e.g., gap junctions or the porin pores of gram-negative bacteria (Unwin and Zampighi, 1980; Steven et al., 1977; and see below). Though the primary structures (amino acid sequences) are now known for a great number of transport proteins, there are only one or two cases where enough is known about the tertiary structure (i.e., the folding of the peptide across the membrane), to allow for informed guesses as to the polar residues that form the substrate-conducting pore or channel. Substrate specificity presumably involves contacts between substrate and catalyst, but next to nothing can yet be said as to how many contacts are required, the binding energies involved, the consequences ofligation (whether or not this triggers conformational changes of the protein), or the types of amino acid residues involved (let alone specific residues). Crystallographic data on soluble proteins (such as enzymes, or the periplasmically located arabinose and galactose binding proteins of E. coli (Quiocho, 1990a, b), allows us to identify these contacts. At present, the best we can do for transport catalysts is to guess that similar residues and conformational changes are involved there also. Active transport, the transport of solutes (either inward or outward) from a lower to a higher electrochemical potential, can be achieved, and can only be achieved, by coupling the free-energy consuming transport process to a free-energy yielding process. This may be a chemical reaction for which the free-energy change is negative, e.g., the hydrolysis of ATP by water to ADP and Pi, in which case the

230

IAN C. WEST

combined (coupled) process is described as primary active transport. The freeenergy yielding process may also be another transport reaction, e.g., the flow of Na + (or H +) down its electrochemical potential gradient into the animal (or plant, or bacterial) cell. It is quite helpful to emphasize (to oneself) that the two parts of the combined (coupled) process do not actually exist in themselves as separate entities with the energy mysteriously travelling from one to the other. Such a view raises unanswerable questions as to the mechanism of such energy transduction. The process that takes place is the combined process, for which the free-energy change must be negative. It is easy to see why the secondary active transport processes are so called; the electrochemical potential gradient of the driving solute can not arise spontaneously m it has to be established by another active transport process, ultimately primary, to avoid an infinite regression.

B. Strategies of Transport Catalysts

Large Open Pores~Channels There are a few examples in biology of transport proteins that facilitate transmembrane transport by the simple expedient of maintaining an open water-filled pore across the membrane, though these are not the predominant type of transport protein. Gap junctions of animal cells, and porins of the outer membranes of gram-negative bacteria and eukaryotic mitochondria operate in this way. Gap junctions link cells of essentially identical physiology, so there should be no physiological gradients between the cells, provided both are healthy. Characteristically, the porin pores are large, permitting the entry of molecules with Mr 700-900 D in the case of bacteria, or 2000 D in the case of mammalian mitochondria, and display little selectivity towards molecules below these cut-off sizes. In gram-negative bacteria, where the most extensive information is available, porins are known with detectable preferences for anions over cations, for oligoglucosides over similar sized molecules containing other sugars, and for organic phosphates (See Table 1). It is important to note that these outer membranes appear to serve an entirely different role from the plasma or cell membrane. In bacteria, they are unnecessary for normal physiological and bioenergetic processes, but protect the cell from many environmental toxins and antibiotics. In mitochondria, they define the intramembranal space and retain cytochrome c (and maybe other macromolecules) in that space. Large, open, relatively nonspecific pores of this sort would presumably be lethal in the cell membrane, and indeed they are. Complement and certain bacterial toxins, such as streptolysin O, appear to form such pores, and are lethal to the affected cell.

General Principles of Membrane Transport

2 31

Regulated and Selective Channels (Pores) There are pores, or "channels", in the plasma membranes of many cell types in which a considerable degree of selectivity is observed, and where the pore is by no means always open. However, though these regulated channels may show "open" and "closed" states, they nevertheless do not close and open between each transport event. These selective channels have characteristically been studied by electrophysiologists and as a consequence, relatively little is known about the extent to which these channels allow the passage of small uncharged hydrophilic solutes. The best characterized channels of this type are the two Na+-specific channels of excitable tissues, regulated by acetylcholine in one case, and membrane potential in the other. In both cases, anions are rejected. The acetylcholine regulated channel is rather unspecific; its discrimination is based on ionic size, much as is the ion selectivity of ion-selective glass electrodes. The electrically-gated Na § channel is more specific for Na + and must respond to Na + in some way, for it partially rejects both smaller and larger cations. As already mentioned, these regulated pores or channels display two states; open and closed. In the closed state, they present a low to very low permeability to all substrates. In the open state, they present essentially no barrier to the diffusion of their selected substrates other than the geometrical consequence of their small diameter. The transition between these two states involves a conformational change of the protein that can be triggered by ligand binding or by small to moderate changes (20-50%) in the transmembrane electric field. (If this makes them seem over-sensitive, it is worth noting that because the membrane is very thin, a 60-mV transmembrane potential difference corresponds to a very large field strength of around 2 • 105 V cm-1). Modem techniques have made it possible to record the conductance of a single channel. It turns out that the effector, be it ligand or field, affects not the open conductance, but the fraction of time the channel spends in each of the two states (Figure 2). Typically, a channel may open for three ms and in that time, allow the transport of 103-104 ions.

Table 1. Some Bacterial Porins Pore Size Name

OmpF OmpC PhoE LamB Tsx

Diameter(rim)

1.15 1.02 1.12 0.7

Note: aLong,thin maltodextrins

Substrate Cut-Off Size ( M r )

600 600 600 340-1000a

Specificity

neutral, cations > anions neutral, cations > anions anions > cations, neutral maltodextrinsand cations nucleosides

232

IAN C. WEST

Closed -135

t '

50ms

~

SpA] /

r

'~J

'

i

II

Open

'1

UL

Figure 2. Voltage-gated ion channel. Membrane potential determines 'open probability' or the fraction of tirfie spent in the 'open' conformation. (Modified from Hartshorne et al., 1986).

The n u m b e r of k n o w n ion channels is rising rapidly, including channels m o r e or less specific for various cations and anions, and channels regulated by amino acids, ATP, cAMP, cell volume, and a variety of other signals (see Table 2).

Table 2. Some Ion Channels Class 1. Voltage-gated

2. Extracellularly activated

3. Intracellularly Activated

Operator membrane potential membrane potential membrane potential GABA Glycine Glutamate Acetylcholine cGMP ATP Ca2+

Ionic-Specificity Na+ K+ Ca2+ CI-, HCO3CI-, HCO3Na+, K+, Ca2+ Na+, K+, Ca2+ Na+, K+ K+ C2+

Conductance (pS)

Helices/Subunit

4--20 2-20

6 6

10-20 30-70 120 4-50

4 4 4 4 6 6 6

General Principles of Membrane Transport

233

Gated (Closed~Closed) Pores (Channels) There is another very important type of pore that displays a different kind of gating; here the two states are both closed states, but they differ in that a cavity in the protein is accessible to one or the other of the two aqueous media. In one state, the cavity is open to the extracellular phase, but closed to the cytoplasm; in the other state, vice versa. This class of transport catalyst is very important because it contains the majority of known transport carriers. (I do not think there is much linguistic difference between a channel and a pore, but traditionally, ion-channels (see above) have a "channel" through the protein, while these gated-pores do not; they only have a pore, and of course it does not run right through.)

Flippase There is a class of compounds that is highly lipophilic while at the same time bearing a polar group; these may be called amphipathic, and examples include cholesterol and phospholipids. Such compounds will partition into the membrane (by virtue of their lipophilic functions), but will be anchored at one or the other of the two water/lipid interfaces by their polar group. Thus it is thought by some that cholesterol and steroids in general, pass freely across cell membranes because of their highly hydrophobic planar rings. Others, however, tell us that the half-time for cholesterol flip-flop is measured not in seconds, but in days. A number of considerations and observations lead to the conclusion that there are catalysts of the rate-limiting step, the flip-flop; these have been called flippases (Bishop and Bell, 1985; Devaux, 1988, 1991). The strategy for these transport catalysts should be interestingly different from that of the carriers considered until now. Previously, the substrates were regarded as hydrophilic, and it was assumed that their binding site in the carrier would be in a protected cavity in the center of the protein. Here, the substrates are at least partly hydrophobic; their binding site could be on the surface of the carrier, at its interface with the phospholipid. Previously, the gated pore carrier was seen to have a binding site accessible to one or the other aqueous phase, but not to both. Here the binding site(s) may be accessible to the lipid, and we know nothing about the gating rules. It has been suggested that the m d r l gene product (the multidrug resistance pump or P-glycoprotein which pumps a number of hydrophobic drugs out of cells) may be a flippase (Higgins and Gottesman, 1992).

Multifunctional Carders A certain amount of protein is presumably needed to form a transmembrane pore with a substrate binding site and a gate (see Section II.D). Bacteriorhodopsin (Mr 26 kD), which transports an unhydrated proton, may well indicate the minimum amount of protein required for this. If the carrier is gated by a second substrate, a

234

IAN C. WEST

cytoplasmic regulator, or by phosphorylation, there will presumably have to be additional sites on the carrier, and the protein may well be bigger. Some examples follow.

Carrier and Enzyme: TransportATPase. Primary activetransport carders that catalyze phosphoryl-transfer from ATP require, in addition to the carrier, an ATP binding site, and phosphoryl-transferase enzymology. The ~-subunit of Na+/K+ATPase combines all these functions on a peptide of 100 kDa. Carrier and Channel. A carrier that also functioned as an open (ungated) channel could similarly be a hybrid protein, but it could alternatively be a gated carder with a modified or mutated gate. This seems to be the case in certain mutants of known carriers or with certain alternative substrates.

Carrier, Enzyme and Channel. The multidrug resistance glycoprotein (P-glycoprotein), known to have both ATP-binding sites and ATPase enzymatic functions, and to catalyze active transport, has been claimed to change on command from some signal, to become a volume-regulated C1- channel (Gill et al., 1992; Trezise et al., 1992; Valverde et al., 1992); however, this interpretation is doubted by others (Rasola et al., 1994). With a protein Mr of 140 kD that crosses the membrane twelve times, there is presumably enough protein. C. Functional Considerations From one point of view, the functional properties of a uniporter (e.g., the Glut 1 glucose uniporter), an antiporter (e.g., the coupled exchange of the Band 3 anion exchanger), and the secondary active transport of a symporter (e.g., the LacY lactose-H + carrier) seem highly disparate. Yet from another viewpoint, these proteins of very similar structure are all doing essentially the same thing; it is just a question of their "gating rules". Nor are the E1E2 ATPases and the light-driven bacteriorhodopsin proton pump greatly different, neither in essential structural principals nor in function.

Gating Rules It is instructive to examine some of the gating rules that have evolved, and to explore their consequences.

Gate if Loaded: Antiport. Suppose, for example, that a gated-pore carrier can not change from its outwardly accessible conformation to its inwardly accessible conformation unless it is loaded (i.e., liganded to its specific substrate or a similarly shaped alternative). It is easy to imagine a mechanism for this, though none has as yet been demonstrated; ligand binding energy could stabilize a transition-state.

General Principles of Membrane Transport

235 Uniport

CO

Ci

S~ ~

So

T CSo

CS i

CA i r

CA o

Ai

Ao Ci

Co

Bi

Bo CB i ......... Ci

Ai

Antiport

-',

T T

CA i B i -r

CAB i ~

~ CB o ~

Co

l"

Symport Ao

CAo

Bo ~ CABo

Figure 3. Kinetic schemes for gated-pore carriers. These schemes are not intended to suggest that the carrier (C), or any part of it, moves across the membrane. The horizontal arrows indicate transitions in the time domain between a conformation with an outwardly accessible substrate binding site (Co, etc.) and one with the same site accessible from the inside (Ci, etc.). S, A, and B indicate substrates; vertical arrows show attendant changes in the carrier. Such a carrier would be unable to catalyze simple, one way, net transport (uniport), but would be able to catalyze rapid exchange between labeled and unlabeled molecules of substrate, or between different members of the substrate family (i.e., antiport or counter-transport; see Figure 3). Examples of such carriers are Band 3 (see above) and the adenine nucleotide exchanger of the mitochondrial inner membrane, both of which show ping-pong kinetics.

Gate If Doubly Loaded: Symport. Another gating rule could be that the important conformational change could only take place if the carrier is loaded with two substrates. This is equally easy to visualize, and results in symport (also called

236

IAN C. WEST

co-transport), i.e., in secondary coupled active transport (See Figure 3). Examples abound. In animals, the second substrate is often a Na § ion. In bacteria, plants, and the bioenergetic organelles of eukaryotes, it is typically the H + ion. Symport would also arise if the critical "rule" were one that prevented substrate binding except to carrier that had already bound the co-substrate; and indeed, the binding of substrate in the absence of co-substrate is surely to be avoided, as it must slow up transport by sequestering the cartier in a nonproductive form. There is more to say (see below) about the thermodynamics of symport when the co-substrate is charged and when there is a membrane potential (as there normally is).

Gate If Loaded and Phosphorylated. If one carrier transition (e.g., from inwards to outwards facing) required the protein to be phosphorylated, and the other transition required dephosphorylation, ATP hydrolysis would drive the protein back and forth~but futilely, unless further gating rules required specific substrates, either for the phosphoryl-transfer reactions or the carrier transitions. The best understood ion-translocating ATPases are probably the Ca2+-ATPase and Na+/K +ATPase of animal plasma membranes. Gated by Electric Field. There are many examples of channels that are gated by membrane potential and also examples of ATPases (e.g., in the chloroplast), that are regulated in this way. However, it is important to distinguish this phenomenon from mechanisms that explain how membrane potential drives the accumulation of substrates by electromotive ion-coupled symport (see below). An electric field could force the carrier into a conformation that kinetically allowed or forbade flux, but there would be no thermodynamic consequences unless net charge moved down the field with each transport event.

Thermodynamic Consequencesof Gating Rules It is both a strength and a weakness of thermodynamics that it is a tautology; thermodynamics tells us many truths with great certainty, but ultimately, these are not the truths we want to be told, because we know them already. Let us take symport as an example. If the cyclic process by which lactose is carried into E. coli results in a stoichiometric inflow of H § we can see that equilibrium will occur when the electrochemical potential gradient for lactose is equal but opposite to, that of the proton; lactose will be accumulated to the extent that the proton is expelled from the cell. At zero membrane potential: Log ([lactose]in/[lactose]out) = pHin - pHout

(2)

while every 60 mV of membrane potential (negative inside) would increase the lactose accumulation ratio (at 27 ~ 10-fold. Or we can argue in reverse and conclude that it is not sufficient for lactose accumulation that there be a proton bound in the ternary complex and a pH gradient, or a membrane potential; we can

General Principles of Membrane Transport

237

see that net transport of H + across the membrane with each transport cycle is essential. For another example we could take the Na+/K + ATPase. The property that drives the expulsion of Na + and the uptake of K + is the free energy difference between free ATP plus water and free ADP plus phosphate. However, it is not a simple matter to point to the free energy yielding or consuming step; we must know the free energies of binding of the nucleotides and ions. We might suspect that no one step will have a free energy change very far from zero, for if such were the case and an intermediate either occurred at vanishingly low concentration or built up tremendously, we might expect kinetic sluggishness, or 'slip' as intermediates spilled over, out of the reaction pathway. However, to test that expectation we would have to measure intermediate concentrations at equilibrium or steady-state, or make kinetic measurements. In other words, a complete study of kinetics answers the interesting thermodynamic questions while the reverse is not true.

Kinetic Properties of Carriers Mass Action Effects. Both in primary and secondary active transport, the energetic input can often be seen as raising the concentration of a participating reactant. J. Keith Wright and others have called this "recruiting". If thermodynamics comes down to kinetics, kinetics comes down to mass action; two reactant molecules make product twice as fast as one. Once again, take lactose-proton symport in order to fix the mind. We know that the carrier must bind both lactose and proton (the lactose cannot bind H +) and then undergo a conformational change. Also, we can see that if there are more protons outside the cell than inside, mass action alone will lead to the more rapid formation of the competent (ternary) complex from the outside (leading to influx) than from the inside (leading to effiux). There are still options the carrier can adopt; whether to bind the sugar and wait for the proton or vice versa, or not to bother about the order of binding. This may not matter on the outside; but on the inside there will be inhibition of inflow if there is too-ready binding of sugar into a binary complex that subsequently, for want of a proton, does nothing but sit there. For many substrates, this is clearly disadvantageous; but with other (perhaps toxic or mutagenic) substrates, "product inhibition" of this sort appears to have evolved and is presumably advantageous. The external proton recruits the lactose carrier to the outside-accessible conformation ready for influx. In an analogous manner cytoplasmic ATP converts the E2(K +) form of the Na+/K+ ATPase to the E 1 form ready for Na + binding. Kinetic Consequences of an Electric Field. Many substrates or co-substrates of transport reactions are charged, and it is clear from thermodynamical considerations that transmembrane electric fields must affect the free energy of transport by affecting the kinetics. Understanding the effects of transmembrane electric fields

238

IAN C. WEST

has proven difficult; there are a number of monograms that grapple with the subject (Heinz, 1981; L~iuger, 1991). A simple, basic concept of general application is once again the idea of recruitment; the field can be seen as leading to a raised (or lowered) concentration of the substrate ion (exemplified as Mitchell's "proton well" Mitchell, 1977), or of the carrier itself (Wright's recruiting concept). Mass action will then see to the kinetics (and hence the thermodynamics) of transport. Electrophoretic accumulation of H + down a proton-specific (West, 1980) tube (the "proton well") is relatively easy to visualize. But how does the electric field affect carrier orientation? Any molecular species with a dipole will sense the vectorial (transmembrane) asymmetry of the electric field. Protonation must change the dipole. Wright has pointed out (Wright et al., 1986) that we can not know a priori the point or points in the transport cycle when the dipole moves or reorients in the field; it could be the carrier eversion step, but it could be substrate binding/unbinding; experiments are needed. Any or all of these effects of the electric field may operate to affect the kinetics and thermodynamics of carrier-mediated transport.

How to Interpret Kinetic Experiments. Kinetics, like biophysics and thermodynamics, is unfortunately a subject where enormous refinements can be made simply by taking thought. Scientists whose expertise is in doing experiments often make only limited incursions into this quagmire. Nevertheless, a large part of our understanding of how carriers work comes from a thorough analysis of kinetic experiments. The very concept of a carrier is the product of kinetic analysis. The observation that the transport of glucose through a biological membrane is faster than through a phospholipid bilayer, suggests a catalyst; that transport can be slowed by inhibitors supports this conclusion, especially if the inhibitors are substrate analogs, or very small amounts of chemicals known to react with proteins. Saturation (Michaelis-type) kinetics suggests a saturable substrate-binding site, while stereoisomers and homologous series define some of the steric properties of that binding. Trans-stimulation (the speeding of isotopic flux by substrate or substrate analog on the opposite side of the membrane) distinguishes the so-called carriers from the so-called channels; in fact, trans-stimulation indicates the importance of carrier reversion as a key step in the kinetic cycle (see Figure 3). So, long before we had any experimental information about the shape of the carrier protein in the membrane, kinetics afforded us a mental picture of a protein that binds a substrate in a stereospecific cavity, changes shape, and releases substrate to regenerate free carrierDbut free carrier that is different from the original. On the other hand, kinetics have been frequently misinterpreted. For example, KM values ([S] when v = Vmax/2) have been interpreted as dissociation constants, indicating affinity (Wright et al., 1985). Analyses have been simplified by assuming that certain steps are infinitely fast relative to other steps.

General Principles of Membrane Transport

239

D. Structural Considerations

Crossing the Membrane It is now a familiar idea that, by looking at the primary structure (i.e., amino acid sequence) of a membrane protein, it is possible to determine the membrane spanning a-helices. However, that is an oversimplification. Some of the algorithms that are used for predicting membrane spanning helices are based entirely on hydrophobicity; these inevitably ignore the facts that hydrophobic amino acid residues are not confined to membrane-spanning helices, and that some polar, and even ionizable residues, are found buffed in the membrane. Other algorithms are entirely ad hoc, noting the residues found in regions of peptide known to traverse the membrane, but these are inevitably based on very little data. Yet other algorithms combine both approaches. There are very few membrane proteins where accurate tertiary structural information allows us to test the accuracy of the various algorithms, and these fall into only two families m the bacteriorhodopsin family, and the bacterial photosynthetic reaction center family. No algorithm is better than 77% reliable on both families (Crimi and Degli Esposti, 1991). Another problem is that not all membrane-spanning stretches of peptide are m-helical; some membrane proteins clearly have small portions of 13-sheet (Yellen et al., 1991), while others (e.g., porins) are almost entirely composed of ]]-structure (Nikaido, 1994). The upshot of all this is that predicted membrane-spanning helices must be treated with appropriate care, as an initial hypothesis only, for the framing of more definitive experiments such as the labeling of exposed residues, the identification of salt-bridges and antibody binding sites, etc. (see Jahnig, 1990; Fasman and Gilbert, 1990).

Negatively Charged Residues von Heijne (1986) first pointed out the tendency for (putative) cytoplasmically facing loops to carry a slight predominance of positively charged amino acid residues, while extracellularly facing segments tend to carry an excess of negatively charged residues. It is tempting to think that charged patches at the point where the hydrophobic membrane phase abuts on the hydrophilic phases may play a part in defining the limits of the buried helices, and fixing the protein in its desired orientation. As the membrane potential, whether small (9 mV in the erythrocyte) or large (120 mW in E. coli), is always of the same orientation, with the cytoplasmic phase negative with respect to the extracellular, it is tempting to see this polarity of each membrane-spanning helix as in some way helping the protein to insert correctly into the membrane, whether this insertion be a simple physical event or a complex, protein assisted process (Dalbey, 1990).

240

IAN C. WEST NI-! ,

kLA 177

TYR 236

(;LU ,325 HIS ;322 LYS )319

ARG 302, SER 306

COOH

Figure 4. Suggested model for the tertiary structure of the lactose permease of Escherichia coll. The model depicts 12 cylindrical transmembrane segments, each in o~-helical conformation, connected by hydrophilic segments; they are arranged in such a way that residues known to affect sugar-substrate binding all face a central pore. (After Collins et al., 1989 with permission).

Buried Salt-Bridges An example of a salt-bridge in the lactose permease buried in the hydrophobic phase of the E. coli membrane has been beautifully analyzed by T.H. Wilson and colleagues (King et al., 1991; Lee et al., 1992, 1993). Mutants show that the amino group of a lysyl side-chain is very uncomfortable in the core of the membrane; its high affinity for a proton, and hence water, presumably forces it out into one or the other aqueous phase, unless it can pair closely with a buried aspartyl side-chain, presumably sharing the proton. On the other hand, an unpaired Asp is similarly disruptive. These experiments have proven to be very informative for, besides indicating the ability of a single charge to drag the membrane-spanning helix out to the aqueous medium, they have revealed which helices were juxtaposed in the as yet unknown tertiary structure. It should be possible to engineer additional salt-bridges in the protein to test putative juxtapositions not revealing themselves spontaneously.

General Principles of Membrane Transport

241

Size The small C peptides of the FiF0 ATPases cross the membrane only twice. However, it is thought that 6-9 or even 10 copies of the subunit are required to form the proton-conducting channel. Similarly, each subunit of the acetylcholine-regulated channel crosses the membrane four times, but five subunits combine to make a pore in the o~2,~,y,8 structure. The K § channel peptide appears to contain six membrane-spanning or-helices; if it were functional as the monomer, it would be one of the smallest membrane transport catalysts, but it is presumed (partly by analogy with the fourfold larger Na + channels) to function as a tetramer (Jan and Jan, 1992). Bacteriorhodopsin (26 kDa)is the prototype of a widespread family of or-helical membrane proteins that cross the membrane seven times. However, bacteriorhodopsin translocates only protons, while most members of the family translocate nothing more than information. The large number of membrane transport catalysts possessing 12 membrane-spanning helices is certainly very striking and has led P.C. Maloney to annunciate what he has called the "rule of 12" (Maloney, 1989). Many sugar- and anion-porters look as though they may be members of a single superfamily, constructed as a tandem duplication of a six-helix prototype. Other apparently unrelated families, e.g., the OppD/Mdr family, also contain 12 membrane-spanning helices, again as a tandem duplication. On the other hand, the phosphate-porter family of the mitochondrial inner membrane has six membrane-spanning helices, (a 2 + 2 + 2 triplication). This, and a number of other clues, suggest that six helices packed into a ring (see Figure 4) are just sufficient to enclose a central tube that can act as the transit channel for substrates ranging in size from nitrate to lactose. I therefore prefer the concept of an ancestral proto-carrier having six membrane-spanning helices (a "rule of 6").

Substrate Binding Sites Not much can, as yet, be said about substrate binding sites. The band 3 anion exchange carrier, referred to above, has been said to have only one type of chloride binding site (from NMR evidence). This seems very possible and may prove to be a general feature; one site accessible from one or the other surface. Another approach is the collection of mutants with altered substrate specificity that can be mapped very precisely to specific amino acid side-chains (see Figure 4) Collins et al. (1989). Though suggestive, this approach can not yet be said to have given us a complete picture of a binding site.

Flexibility There are already a few clues as to the types of mobility required of a gated-pore carrier. The relationship between rate of turnover and temperature has been studied for a number of carriers. A linear Arrhenius plot indicates a characteristic activation

242

IAN C. WEST

energy for the rate-limiting step of transport, which seems always to be the conformational change in those cases where there is evidence. Characteristically, there is a discontinuity in the Arrhenius plots between two linear portions, showing a fairly abrupt rise in activation energy as the temperature falls below a critical point. However, what "freezes" at this transition temperature is not so clear m lipid or protein (Overath et al., 1976).

Iii.

CONCLUSIONS

I hope this chapter has shown that there are emerging two or three strategies by which solutes are translocated across membranes, two or three types of carrier structure, and a rapidly expanding picture of the kinetic process that is transmembrane transport.

REFERENCES Bishop, W. R., & Bell, B. M. (1985). Assembly of the endoplasmic reticulum phospholipid bilayer: the phosphatidylcholine transporter. Cell 42, 51-60. Collins, J. C., Permuth, S. E, & Brooker, R. J. (1989). Isolation and characterization of lactose permease mutants with an enhanced recognition of maltose and diminished recognition of cellobiose. J. Biol. Chem. 264, 14698-14703. Crimi, M., & Degli Esposti, M. (1991). Structural predictions for membrane proteins: the dilemma of hydrophobicity scales. Trends Biochem. Sci. 16, 119. Dalbey, R. E. (1990). Positively charged residues are important determinants of membrane protein topology. Trends Biochem. Sci. 15, 253-257. Devaux, P. E (1988). Phospholipid flippases. FEBS Lett. 234, 8-12. Devaux, P. E (1991). Static and dynamic lipid asymmetry in cell membranes. Biochemistry 30, 1163-1173. Fasman, G. D., & Gilbert, W.A. (1990). The prediction of transmembrane protein sequences and their conformation: an evaluation. Trends Biochem. Sci. 15, 89-92. Gill, D. R., Hyde, S. C., Higgins, C. E, Valverde, M. A., Mintenig, G. M., & Sepulveda, E V. (1992). Separation of drug transport and chloride channel functions of the human multidrug resistance P-glycoprotein. Cell 71, 23-32. Hartshorne, R., Tamkun, M., & Montal, M. (1986). In: Ion Channel Reconstitution (Miller, C., ed.) pp. 337-362, Plenum Press, New York. Heinz, E. (1981). Electrical Potentials in Biological Membrane Transport, Springer-Verlag, Berlin. Higgins, C. E, & Gottesman, M. M. (1992). Is the multidrug transporter a l~ppase? Trends Biochem. Sci. 17, 18-21. Jan, L. Y., & Jan, Y. N. (1992). Structural elements involved in specific K+ channel functions. Annu. Rev. Physiol. 54, 537-555. J~hnig, E (1990). Structure predictions of membrane proteins are not that bad. Trends Biochem. Sci. 15, 93-95. King, S. C., Hansen, C. L., & Wilson, T. H. (1991). The interaction between aspartic acid 237 and lysine 358 in the lactose carrier of Escherichia coli. Biochim. Biophys. Acta 1062, 177-186. L~iuger, P. (1991). Electrogenic Ion Pumps, W.H. Freeman and Co., Oxford. Lee, J. I., Hwang, P. P., Hansen, C., & Wilson, T. H. (1992). Possible salt bridges between transmembrane alpha-helices of the lactose cartier of Escherichia coli. J. Biol. Chem. 267, 20758-20764.

General Principles of Membrane Transport

243

Lee, J. I., Pung, P., & Wilson, T. H. (1993). Lysine 319 interacts with both glutamic acid 269 and aspartic acid 240 in the lactose carrier of Escherichia coli. J. Biol. Chem. 268, 20007-20015. Lieb, W. R., & Stein, W. D. (1986). In: Transport and Diffusion Across Cell Membranes (Stein, W.D., ed.), pp. 69-112, Academic Press, San Diego. Maloney, P. C. (1989). Resolution and reconstitution of anion exchange reactions. Phil. Trans. Roy. Soc. B 326, 437-454. Mitchell, P. (1977). Epilogue: from energetic abstraction to biochemical mechanism. Symp. Soc. Gen. Microbiol. 27, 383-423. Nikaido, H. (1994). Porins and specific diffusion channels in bacterial outer membranes. J. Biol. Chem. 269, 3905-3908. Overath, P., Thilo, L., & Tr~iuble, H. (1976). Lipid phase transitions and membrane function. Trends Biochem. Sci. 1, 186-189. Quiocho, F. A. (1990a). In: Microbial Membrane Transport Systems (Kornberg, H.L., & Henderson, P.J.E, eds.), pp. 341-351, The Royal Society, London. Quiocho, E A. (1990b). Atomic structures of periplasmic binding proteins and the high-affinity active transport systems in bacteria. Phil. Trans. Roy. Soc. B 326, 341-351. Rasola, A., Galietta, L. J. V., Gruenert, D. C., & Romeo, G. (1994). Volume-sensitive chloride currents in four epithelial cell lines are not directly correlated to the expression of the mdr- 1 gene. J. Biol. Chem. 269, 1432-1436. Stein, W. D. (1967). The Movement of Molecules Across Cell Membranes, Academic Press, London. Stein, W. D. (1981). In: Membrane Transport (Bonting, S. L., & dePont, J. J. H. H. M. eds.), pp. 1-28, Elsevier, Amsterdam. Stein, W. D. (1986). Transport and Diffusion Across Cell Membranes, Academic Press, San Diego. Steven, A. A., ten Heggeler, B., MUller, R., Kistler, J., & Rosenbusch, J. P. (1977). Ultrastructure of a periodic protein layer in the outer membrane of Escherichia coli. J. Cell Biol. 72, 292-301. Trezise, A. E. O., Romano, P. R., Gill, D. R., Hyde, S. C., Sepulveda, F. V., Buchwald, M., & Higgins, C. F. (1992). The multidrug resistance and cystic-fibrosis genes have complementary patterns of epithelial expression. EMBO J. 11, 4291-4303. Unwin, P. N. T., & Zampighi, G. (1980). Structure of the junction between communicating cells. Nature 283, 545-549. Valverde, M. A., Diaz, M., Sepulveda, E V., Gill, D. R., Hyde, S. C., & Higgins, C. E (1992). Volume-regulated chloride channels associated with the human multidrug-resistance P-glycoprotein. Nature 355, 830-833. von Heijne, G. (1986). The distribution of positively charged residues in bacterial inner membranes proteins correlates with the trans-membrane topology. EMBO J. 5, 3021-3027. West, I. C. (1980). Energy coupling in secondary active transport. Biochim. Biophys. Acta 604, 91-126. Wright, J. K., Dornmair, K., Mitaku, S., M6r6y, T., Neuhaus, J. M., Seckler, R., Vogel, H., Weigel, U., J~nig, E, & Overath, P. (1985). Lactose: H§ cartier of Escherichia coli: kinetic mechanism, purification, and structure. Annals New York Acad. Sci. 456, 326-341. Wright, J. K., Seckler, R., & Overath, P. (1986). Molecular aspects of sugar:ion cotransport. Annu. Rev. Biochem. 55, 225-248. Yellen, G., Jurman, M., Abramson, T., & MacKinnon, R. (1991). Mutations affecting internal TEA blockade identify the probable pore-forming region of a K§ channel. Science 251,939-941.

This Page Intentionally Left Blank

MEMBRANE SIGNALING SYSTEMS

C. U. M. Smith

Io II. III. IV. V. VI. VII. VIII. IX. X. XI.

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . General Orientation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Primary Messengers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Membrane Receptors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . G-Proteins . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Membrane-Bound G-Proteins . . . . . . . . . . . . . . . . . . . . . . . . . . G-Protein Signaling Systems in Biomembranes . . . . . . . . . . . . . . . . . Networks of G-Protein Signaling Systems . . . . . . . . . . . . . . . . . . . . Effectors and Second Messengers . . . . . . . . . . . . . . . . . . . . . . . . Pathologies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

I.

245 246 248 250 252 255 258 259 261 266 266 268

INTRODUCTION

M e m b r a n e signaling systems are ubiquitous and evolutionarily ancient. T h e y have, for instance, been studied in bacterial m e m b r a n e s w h e r e r e c e p t o r - t r a n s d u c e r m e c h a n i s m s play important roles (see S i m o n et al., 1985) and are also w e l l - k n o w n

Biomembranes Volume 1, pages 245-270. Copyright 9 1995 by JAI Press Inc. All rights of reproduction in any form reserved. ISBN: 1-55938-658-4

245

246

C . U . M . SMITH

in yeasts (Engelberg et al., 1989; Kaziro et al., 1991). In this chapter, however, we shall concentrate our attention on mammalian systems. Here, they are of far more than merely theoretical interest. We know already that ailments such as whoopingcough, scourges such as cholera, and perhaps some forms of cancer are due to defects in such systems.

I!. GENERAL ORIENTATION Membranes respond to a great variety of 'primary messengers' in the extracellular compartment (see Table 1 below). These messengers or agonists interact with precisely tailored receptor molecules (Rs) embedded in the membrane. There are two major types of response: direct and indirect. In the first case, the receptor is, in a sense, its own effector; in the second, the receptor and effector are separate structures in the membrane. The best known examples of direct responses in mammalian cell membranes are those where a receptor doubles as an ion channel. When an appropriate messenger (M) arrives, it causes the receptor to undergo a conformational change such that the conductivity of the channel is altered. Consequently, the flows of ions along their electrochemical gradients through the membrane are altered. Such receptors are defined as ionotropic. In the second case, the receptor is linked via a complex membrane biochemistry to a separate effector (E), also embedded in the membrane. When the messenger interacts with the receptor, the membrane biochemistry is activated and the action of the effector protein changed. Such receptors are defined as metabotropic. Whereas the ionotropic response is very rapid and quickly over, the metabotropic response is slower, sustained, and complex (Figure 1). These two major types of membrane responses to an extracellular effector play very different roles in cellular physiology. Ionotropic responses, because they occur and finish rapidly, underly the rapid responses at, for example, the neuromuscular junction. Here nicotinic-acid receptors (nAChRs) on the sarcolemmae of striated muscle fibers respond very rapidly (ms) to acetylcholine released from overlying motor nerve terminals and hence ensure prompt action by the muscle. Ionotropic receptors are thus of central importance in the neuromuscular system and the behavioral movements it generates. These topics, however, fall outside the scope of this chapter. In what follows, we shall concentrate our attention on the signaling systems involving metabotropic responses. These responses are typically several orders of magnitude slower than ionotropic responses (seconds to tens of seconds). The primary messengers metabotropic systems respond to are many and various (see Table 1). They include neurotransmitters and hormones and, in the case of sensory cells, odorant molecules and photons. In fact, about 80% of all known hormones and neurotransmitters act through metabotropic systems (Birnbaumer et al., 1990a). These systems are thus of overwhelming importance in the life of mammalian cells and in the coordination of a multicellular body.

Membrane Signafing Systems

247 1M

(ii) 1M

\

E

(i)

(ii)

2M

Figure 1. Ionotropic and metabotropic responses. (a) Ionotropic, (b) metabotropic. R, receptor molecule; R*, activated receptor; E, effector molecule; E, activated effector; 1M, primary messenger; 2M, second messenger.

Although primary messengers are very heterogeneous, the signaling systems through which they work remain remarkably similar, whether in the rod cell outer segment, cardiac muscle, erythrocyte, olfactory cell, or neuron. An outline of our present understanding forms the substance of this chapter. They depend on G-proteins associated with the cytoplasmic leaflet of the membrane. G-proteins are so-called because a central feature of their biochemistry is the binding and catalysis of guanine nucleotides (GDP and GTP). We shall look at them in detail in sections IV and V, below. The operation of G-protein signaling systems depends on the fluidity of the lipid cores of biomembranes. This depends on a number of factors including the cholesterol content and the length and degree of saturation of fatty-acid chains of the phospholipids. In many cases, the fluidity of lipid bilayers is said to approach that of a light machine oil (but see section VIII). Lateral movement of proteins in the membrane under the influence of thermal forces is thus very easy and has been well-established by, for instance, laser bleaching and other techniques (Jacobson et al., 1987). G-proteins are thus easily able to shuttle between receptor (R) and effector (E). Effectors take several forms. The most important are enzymes, such as adenylyl cyclase, phospholipase C, and phosphodiesterase, and channels such as some K + and Ca 2+ channels. In the first case, a variety of second messengers (2M) are generated: cAME DAG, InsP3, Ca 2+, etc.; in the second, flows of the relevant ions

248

C . U . M . SMITH

'• Input

Output

~M

,~

Extracellular fluid

G -protein coupled

G -protein coupled

receptor

effector

Molecular Diversity

>40

>80

>20

J,

v

Cytosol

2M >6

>6

Partly after Birnbaumeret al., 1990

Figure 2. G-protein membrane signaling systems. Explanation in text.

across the membrane are affected. Figure 2 summarizes the unity in diversity of Gprotein signaling systems. Figure 2 shows that the G-protein signaling system 'transduces' the primary message (1M) arriving at the extracellular surface of the membrane. The end result of the transduction depends, as we noted above, on the effector. The transduction process confers, as we shall see, great flexibility on the membrane signaling systems. Not only may the transduction be either inhibitory or excitatory, but the activity signaled by the primary message may be greatly amplified and sustained over an appreciable period of time. In this introductory chapter we shall proceed as follows. After a brief review of the nature of primary messengers, we shall look at the structure and function of membrane receptors, especially metabotropic receptors. The chapter then proceeds with a discussion of G-proteins and G-protein signaling systems within membranes, and ends with an account of membrane 'effectors' and the intracellular second messengers they generate.

III.

PRIMARY MESSENGERS

The cells of a multicellular organism communicate with each other in a variety of ways. In some tissues, especially epithelial tissues, gap junctions allow direct communication between the cytosols of neighboring cells. In other cases, especially in some parts of vertebrate nervous systems, tight junctions allow direct transmission of electrical signals from one cell to the next. These types of communication do not, however, come under the remit of this chapter and no more will be said of them. The intercellular communications of interest to us here are those made by way of 'informational substances' (ISs) (Schmitt, 1986) or better, 'messenger substances' (MSs), in the extracellular space. These substances take a variety of

Membrane Signaling Systems

249

(a)

(b)

(c)

(d)

II

Ii /

~

II I

---- - - ~

)

Figure 3. Types of intercellular signaling systems. (a) autocrine, (b) paracrine, (c) endocrine, (d) neuroendocrine, (e) neurocrine.

molecular forms and we shall review them briefly below. First, however, it is worth noting the principal pathways by which these substances exert their effects. These pathways are shown in Figure 3. The membranes of all cells, from bacteria, through the Protista, to multicellular forms, are chemo-sensitive. Figure 3 shows that by the elaboration and secretion of MSs cells can communicate with each other either very directly, as in the autocrine, paracrine, and neurocrine cases; or at greater distances, as in the neuroendocrine and endocrine cases. Next, let us turn to the chemical nature of the messenger substances. They are, of course, extremely various. Upwards of 30 have been recognized in the nervous system alone, and there are many others in other anatomical systems. An outline summary is given in Table 1.

250

C.U.M. SMITH Table 1. Major Types of Primary Messengers

Class

Example

Endocrine

amino acid derivatives catecholamines steroids peptides polypeptides/proteins glycoproteins

thyroxine epinephrine androgens; estrogens vasopressin; secretin somatostatin; insulin hematopoietic factors; FSH

Paracrine

amino acid derivatives arachidonic acid derivatives

histamine prostaglandin

Neuroendocrine

peptides

oxytocin;ADH

Neurocrine

inorganic amino acids catecholamines indoleamines purines peptides polypeptides

nitric oxide (NO) glycine; glutamate norepinephrine; dopamine serotonin adenosine substance P; encephalins ~-endorphin

Table 1 shows that MSs range from very simple molecules, such as nitric oxide (NO) and glycine, to large and complex structures, such as polypeptides and glycoproteins, whose molecular weights may be several kDa.

IV. MEMBRANE RECEPTORS Membrane receptors take the form of large globular proteins embedded in the membrane. It is only recently that the techniques of molecular biology have enabled the structures of these molecules to be determined. In essence, what is done is to sequence purified fragments of a receptor and use these to construct a complementary oligonucleotide probe. The probe is used to screen an appropriate cDNA library for the cDNA of interest, and this is then cloned, or, more recently, amplified by the polymerase chain reaction (PCR), to provide sufficient DNA for sequencing. The amino acid sequence of the receptor can then be established from the polynucleotide sequence (for an introductory account see Smith, 1990). Finally, by locating the position of runs of hydrophobic amino acids in the polypeptide chain, the orientation of the protein in the membrane can (at least) be hypothesized (Fasman and Gilbert, 1990; J~ihnig, 1990).

Membrane Signaling Systems

251

We noted in section I that receptors can be divided into two large groups: ionotropic and metabotropic. The first receptor to yield its structure to the molecular biologist was the nicotinic acetylcholine receptor (nAChR) (Numa et al., 1983). This is an ionotropic receptor and, consequently, we shall pay it no more attention. The first metabotropic receptor to have its structure determined was the 132-adrenergic receptor (132-AR) (Dixon et al., 1986). It is still probably the best understood of all the metabotropic receptors and much of what has been learned of its structure-function relationships will (hopefully) be relevant to other G-protein coupled receptors (see Kobilka, 1992). Its structure is shown in Figure 4.

NH2

HUMAN ~2ADRENERGIC RECEPTOR EXTRACELLULARSURFACE

CYTOPLASMIC SURFACE

Figure 4. The [32-adrenergic receptor. The seven-transmembrane helices are labeled I-VII. These helices are grouped to form a hollow cylinder in the membrane (see text). The extracellular and intracellular loops are labeled e-1 to e-3 and i-1 to i-3, respectively. The amino acids are designated by the conventional single letter code, and the glycosylated residues on the N-terminal chain in the extracellular space by 'horns'. Black residues in the intracellular space are potential sites for phosphorylation by PKA (protein kinase A) or 13-ARK ([3-adrenergic receptor kinase) both of which desensitize the receptor. (Reproduced with permission from Kolbika, 1992).

252

C . U . M . SMITH

Figure 4 shows that hydrophobicity analysis suggests that the 152-AR has seven transmembrane helices (often abbreviated as 7-TM). It has subsequently become apparent that a large number (not all, see Okamato et al., 1990) of other G-protein coupled receptors share this 7-TM structure. These include the ~l- and 132-adrenergic receptors (Yarden et al., 1986; Frielle et al., 1987), a family of ct-adrenergic receptors (~IA, (XIB, 0~IC,Ct2A, ~2B, 1~2C)(Cotecchia et al., 1988), the ml, m2, m3, m4, and m5 muscarinic acetylcholine receptors (mAChRs) (Kubo et al., 1986; Bonner et al., 1990), the D1A, D1B, D2, D3, D4, and D5 dopaminergic receptors (Anderson et al., 1988), the Al and A2 adenosine receptors (Libert et al., 1991), the angiotensin AT1 receptor (Jackson et al., 1988), four subtypes of the serotinergic receptor, 5HTla, 5HTlc, 5HT1D and 5HT2 (Fargin et al., 1988; Julius et al., 1988; Pritchett et al., 1988), a family of metabotropic glutamate receptors (Tanabe et al., 1992), substance K receptor (Masu et al., 1987), the H1 and H2 histamine receptors (Gantz et al., 1991; Yamashita et al., 1991), NK1, NK2, and NK3 tachykinin receptors (Shigemento et al., 1990), rod opsin and the three cone opsins (Nathans et al., 1986), bacteriorhodopsin (Henderson and Unwin, 1975), chemo-attractant receptors (Klein et al., 1988), and a large group of closely related olfactory receptors (Buck and Axel, 1991; Reed, 1992). It is clear that we are dealing with an evolutionarily related superfamily of membrane-embedded proteins sharing major structure-function characteristics. This evolutionary dimension is supported by the observation that receptors for the same agonist in different organisms have somewhat different primary structures and, even more interestingly, that receptors for the same agonist in different parts of the same organism also have slightly different structures, and consequently, slightly different response characteristics. Much detailed biochemistry is now being carried out on the 7-TM receptors to determine genetic expression, translational control, and precise structure-function relationships. It is believed that the seven-transmembrane helices form the pillars of a hollow column, orientated rather like the iris diaphragm of a camera, embedded in the lipid bilayer of the membrane. Small agonists, such as muscarinic acetylcholine, indoleamines, catecholamines, and retinal, are believed to occupy binding sites deep within the central cavity of the column. It is easy to see that their presence may well alter the packing of the seven-helical segments. Large agonists, such as the glycoprotein hormones (thyrotropin, luteinizing hormone) and polypeptides such as neuropeptide Y, cannot fit into so small a cavity. Instead, their receptors develop lengthy N-terminal chains, projecting from the extracellular face of the membrane (see Figure 4), and it is believed that these provide the necessary binding sites.

V. G-PROTEINS The G-proteins involved in membrane signaling systems are members of a large and ubiquitous superfamily of guanine-nucleotide binding proteins. The superfa-

Membrane Signaling Systems

253 Inactive state ["off"]

@

Pi

II,~o

GTP

Figure 5. Activation-deactivation cycle of a G-protein. For symbols and further explanation, see text.

mily includes cytoskeletal proteins such as tubulin, the initiation and elongation factors of the ribosomal translation process, vesicular transport proteins, and at least 30 small (20-25 kDa) regulatory proteins similar to the p21 protein (i.e., 21 kDa) product of the ras oncogenes and proto-oncogenes (Bourne et al., 1991). Ras oncogenes were originally identified in two different retroviruses causing rat sarcomas--hence the name. They have since been found in a wide variety of organisms ranging from yeasts to humans. Each G-protein, as Bourne et al. (1991) remarked, 'is a precisely engineered switch that can change its affinity for other macromolecules.' Structurally, the best-known are the ras 21 protein and the bacterial elongation factor, EF-Tu, which have both been subjected to detailed X-ray crystallography (Wittinghofer and Pai, 1991). All the G-proteins are switched on by binding to GTP and switched off by the hydrolysis of GTP to GDP. This hydrolysis is catalyzed by the ATPase activity of the G-protein itself. It is, however, a slow process having a half life from a few seconds to a few tens of seconds. This accounts for the observation, mentioned above, that metabotropic receptors show a much longer, more sustained response than ionotropic receptors.

254

C . U . M . SMITH Pi

9

~o

GTP

Figure 6.

Regulation of the G-protein response cycle by GNRP and GAP. Symbols as in Figure 5; see text for further explanation.

A simplified response cycle of a G-protein is shown in Figure 5. In the 'off', or inactive state, the G-protein is bound to GDP. The binding is quite firm: the rate constant for GDP release (Kdiss.GDP) is < 0.03 min -1. When GDP is released, however, the G-protein enters a very transient, 'empty', or, 'neutral', state. As there is normally a higher concentration of GTP than GDP in the cytosol, the former enters the vacated nucleotide-binding site on the G-protein. The binding of GTP causes the G-protein to undergo a conformational change to the 'on' or active state (usually represented as G*). In this active state, it is able do its biochemical work. However, it is also able to catalyze the hydrolysis of GTP to GDP + Pi and, thus return to its original 'off' state. It is clear from the forgoing that the length of time the switch stays 'on' is determined by the rate at which the activated G-protein can hydrolyze GTP. To complicate the issue, it has been found that both dissociation of GDP from the inactive G-protein and catalysis of GTP to GDP by the active G-protein are under the control of regulatory proteins - - the guanine nucleotide release proteins (GNRPs) and the GTPase activating proteins (GAPs). In the presence of these regulatory factors, the rates of both processes are greatly increased. In Figure 6, these regulatory proteins have been added to the G-protein cycle.

Membrane Signaling Systems

255

VI. MEMBRANE-BOUND G-PROTEINS Having said a little about G-proteins in general, let us turn to the G-proteins of membrane signaling systems. They have been shown to differ from other members of the superfamily in having a heterotrimeric structure consisting of a large ot-subunit (ca. 45 kDa) and smaller 13- and y-subunits (Figure 7). The ot-subunit is homologous to the 21 kDa monomeric GTPases such as the ras 21 protein and possesses the characteristic guanine-nucleotide binding site. Several workers have proposed structures for this subunit based on analogies with the EF-Tu bacterial elongation factor and, more convincingly, with the p21 ras protein (see Coleman et al., 1994). Both pertussis (PTX) and cholera toxins (CTX) are highly specific for most (but not all) ct-subunits and have proved useful tools in their biochemical and pharmacological analysis. We shall discuss the pathologies caused by these toxins in section X. The smaller 13- and y-subunits are bound so firmly together that it is impossible to separate them in nondenaturing conditions. In the inactive state the resulting complex is firmly attached to the u-subunit. The y-subunit has a hydrophobic tail in the form of a 20-C geranylgeranyl chain which serves to link it into the lipid bilayer of the membrane (Spiegel et al., 1991 ). In this, it resembles the t~-subunit which, unlike most of the other G-proteins in the superfamily, is also attached to the cytoplasmic leaflet of biomembranes. It is believed that this attachment is by way of a 14-C myristic acid chain linked to the N-terminal glycine. These linkages to biomembranes greatly enhance the efficiency of the signaling system for it ensures that the G-protein is always close to both receptor and effector, rather than somewhere in the body of the cytosol. In the last few years, molecular biological techniques have revealed that the ot-subunit comes in a large number of different variants. Analysis of the amino acid sequences allows an evolutionary classification to be attempted (Table 2). It has been found, for instance, that Gt~sl and G~s2 (GO~olf) are very closely related,

iiiiiiiii: !iiii iiii iiiii ::::::::::::::::::::::::::::::::::::::::::"Lh~;i(;i~caci~i i~a~n"~' ::"--"~":":""~~-:::~ ~

Figure 7. Trimeric G-protein of membrane signaling systems.

256

C . U . M . SMITH

Table 2. Classification of G-Protein o~-Subunitsa'b ~-Subunit

Expression

Toxin Sensitivity

Effector c

+/_d

Gs class G~sl

all tissues

CTX

AC Ca 2+ channel

+ +

Go~s2(Ctolf)

olfactory neurons

CTX

AC

+

Grail

brain

PTX

AC

-

Gt~i2

all tissues

PTX

PLC

+

Gtxi3

all tissues

PTX

K + channel

+ -

Gi class

Go~o

brain

PTX

Ca 2+ channel

Gt~tl

retina (rods)

CTX, PTX

cGMP-PDE

+

Gtxt2

retina (cones)

CTX, PTX

cGMP-PDE

+

Gcxa

smooth muscle

~

PLC

+

Gocx(Go~z)

brain

--

PLC

-

Ga class

K + channel G12 class Goq2

all tissues

--

?

GCXl3

all tissues

--

?

Notes: aSequence variants on all the above cx-subunits (with the exception of GCXxhave been demonstrated by the techniques of molecular biology. bpartly after Kaziro et al., 1991; Simon et al., 1991. cAC, adenylyl cyclase; PLC, phospholipase C; cGMP-PDE, cGMP-phosphodiesterase. d+, activation;-, inhibition.

differing in only 12% of their residues. Hence, they can be classified together in the class Gs. Cloning and PCR techniques are constantly detecting new subtypes and any classification must, at present, be provisional. Whereas o~-subunits, as Table 2 indicates, are highly variable, the [3- and ysubunits are much more homogeneous. Even so, five distinct I]-subunits and perhaps ten distinct ~t-subunits have been identified. The number of possible permutations of all three subunits is thus very large. It has been shown that when the o~-subunit binds GDP, its affinity for the 13- and T-subunits is enhanced. The presence of GDP thus stabilizes the heterotrimeric structure. Furthermore, it can be shown that the trimeric form of the G-protein has a greater affinity for an activated R than any of the three subunits have on their own (Bourne et al., 1991). Having briefly considered the structure of membrane signaling G-proteins, we can now return to the cycle outlined in Figure 6. It is possible to see the G-protein linked metabotropic receptors (Rs) as ligand-activated GNRPs, and at least some effectors (Es) as GAPs (Figure 8). In other words, when the heterotrimeric G-pro-

Membrane Signaling Systems

257

EFFECT

f

GDP

GTP

Figure 8. G-protein membrane-signaling system. L, ligand; R*, activated receptor molecule; E* activated effector molecule. Other symbols are as in Figure 5; further explanation in text.

tein comes into contact with a ligand-activated receptor, the release of GDP is catalyzed and the o~-subunit accepts GTP in its vacant nucleotide site. Because the affinity of the [3- and ~t-subunits for cz-GTP is less than for ~-GDE they tend to dissociate. The free ~-GTP is then able to shuttle in the membrane to its destined effector. The effector (enzyme or channel) will be regulated and remain activated or inhibited while the o~-subunit retains its GTE As soon, however, as the ~subunit's GTPase activity can exert itself and hydrolyze GTP to GDP, the subunit returns to its inactive state and its influence on and association with the effector is ended. This GTPase activity has been shown (in the cases of PLC and PDE) to be markedly accelerated by the effector (Arshavsky and Bownds, 1992; Berstein et al., 1992). This finding answers one of the puzzles of G-protein control systems. It had long been felt that the KcatGTPof the c~-subunit (several orders of magnitude less than in the ras p21 protein; Kaziro et al., 1991) made the system too sluggish to account for the comparatively rapid responses elicited by hormones and neurotransmitters. The recognition that Es speed up the rate of m-catalyzed GTP-hydrolysis by up to fifty times and, moreover, that different Es and different E isotypes

258

C . U . M . SMITH

effect different accelerations, solves this particular puzzle. Once back in its inactive, GDP-bound state, the cz-subunit binds once again to the 13- and T-subunits, and the cycle is ready to begin again. Until recently, it had been assumed that the 13- and T-subunits played no real part in membrane signaling. This assumption is now being questioned. It is becoming evident that these two smaller subunits, in the form of a ]3T-complex (for, as we have already noted, they are not separable under physiological conditions) sometimes have an independent and significant role (Federman et al., 1992; Kleuss et al., 1992). There are two possibilities: (a) either they are able to inhibit free cz-subunits by a process called subunit exchange (see below), or (b) they may act independently on specific effectors. It is now beginning to seem likely that both processes occur. The G-protein signaling systems in biomembranes become increasingly complex and subtle the more they are probed by biochemical and molecular biological techniques. It is, however, to a short resume of these systems that we turn in the next sections.

Vii. G-PROTEIN SIGNALING SYSTEMS IN BIOMEMBRANES The preceding two sections began to give some indication of the complexity of G-protein signaling systems in biomembranes. Some of the many processes controlled by these systems are shown in Table 3. The ubiquity of these systems is evident. It is clear that once the evolutionary process had hit on the switchcharacteristics embodied in G-protein structure, it became employed in a whole

Table 3. Some ProcessesControlled by G-Protein Signaling Systems Stimulus

Cell

G-Protein

Effector

Effect

epinephrine

hepatocytes

Gs

ACa

antidiuretic hormone peptide growth factors (e.g., EGF) odorants

kidney cells

Gs

AC

glycogen hydrolysis retention of H20

epithelium

Gs

PLC

cell division

olfactory cells

Golf

AC

acetylcholine

cardiac muscle

Gi

K+ channel

photons photons angiotensin

rods cones smooth muscle cells

Gtl Gt2 Ga

cGMP-PDE cGMP-PDE PLC

detection of odorants deceleration of contractile cycle detection of light detection of light enhanced contractile force

Notes: aAC, adenylylcyclase; PLC, phospholipase C; cGMP-PDE,cGMP-phosphodiesterase; EGE epidermal growth factor.

Membrane Signaling Systems

259

gamut of control mechanisms ranging from the response of sensory cells to their appropriate stimuli to the modulation of the heart beat. In some, but far from all, cases the functions of the G-proteins are symbolized by their subscripts. Inconsistencies are due to the historical development of the subject. G-proteins named for an effect in one system sometimes do not show that effect or even an opposite effect in subsequently discovered systems. Moreover, as in other parts of pharmacology, the newer molecular biological techniques are providing a classification based on amino acid sequence which does not always (or even often) fit with older classifications based on physiological function (see Bimbaumer et al., 1990a,b). As a first approximation, however, one can say that Gs and Gi indicate stimulation and inhibition of adenylyl cyclase, respectively (exceptions to both of these are now known); Gq indicates PTX insensitivity; Gk (Goq3) indicates opening of a specific K + channel; Golf indicates a location in olfactory cells, and Gt stands for t r a n s d u c i n - the G-protein located in rods and cones. It is also worth noting that the preceding nomenclature refers to the whole trimeric G-protein. It will not necessarily always agree with the classification of ~-subunits given in Table 2. Doubtless, further progress in molecular and pharmacological analysis will, in time, lead to a complete and consistent nomenclature (see B imbaumer et al., 1990a, b).

VIII.

NETWORKS OF G-PROTEIN SIGNALING SYSTEMS

We have to bear in mind that the illustrations of the systems presented so far in this chapter have been greatly simplified. In any given area of membrane, there may be several different 7-TM receptors, several different G-proteins, and several different effectors. Indeed, experiments with cloned cells have demonstrated the existence of mRNAs for seven different ~-subunits and three or four different [3- and y-subunits in the same cell. The number of possible different heterotrimers in the same cell membrane could thus approach a hundred (Simon et al., 1991 ). Biological membranes are fully as complex as, and far more dynamic than, the miniaturized circuit boards of late twentieth century electronics. If there are, indeed, a multitude of G-protein signaling systems in a small area of membrane, it is necessary that there should be a minimum of cross-talk, or if there is cross-talk, that it should be 'intended'. It is important that the appropriate G-protein be connected to the appropriate receptor and effector. We are just beginning to understand how this may be done. It has been shown that the G-protein binding domains of the 7-TM receptors are short regions of the second loop (i.e., the loop that connects TM2 to TM3), a section of the large third loop connecting TM5 and TM6, and probably the cytoplasmic tail leaving TM7. There seems to be a distinction between the 7-TM receptors coupled to G-proteins signaling the activation of adenylyl cyclase and those coupled to G-proteins which signal its inhibition. Whereas the former have a short third loop and a long C-terminal tail,

260

C . U . M . SMITH

the latter have a lengthy third loop and a short C-terminal ending. There is little doubt that further work will show yet other structural adaptations so that the multitude of 7-TM receptors (section III) have different affinities for the different G-protein subtypes. There will also, no doubt, be stereochemical differences in the membrane-bound effectors. Although a great deal is now known about the molecular biology and biophysics of ion channels (Agnew et al., 1988), the other effectors are not as well understood. Already, however, it is apparent that adenylyl cyclase (AC), like the 7-TM receptors, is a protein with numerous (probably 12) transmembrane segments and numerous isoforms (Krupinski et al., 1989; Gao and Gilman, 1992). We have already noted that different isoforms probably induce different degrees of activation in their matching G-txs. The cGMP-phosphodiesterase (the effector in the vertebrate phototransduction process) is also comparatively well-understood. It consists of four subunits (txl3Y2) embedded in the 'cytoplasmic' leaflet of outer-segment disc membranes. Activation by Gct*t leads to the displacement of the two y-subunits (Humphries et al., 1992). The region of the transducin ct-subunit (Gctt) which activates this enzyme has been determined (Rarick et al., 1992). It turns out to be a sequence of residues near the C-terminal end of the molecule which, interestingly, is homologous to a similar C-terminal sequence in Gets responsible for activating AC. These stereochemical adaptations ensure that specific G-proteins recognize specific receptors and effectors and thus minimize the danger of unwanted 'crosstalk'. This is not to say, however, that several G-proteins may not act on a single effector. For example, at least three G-proteins seem to affect certain K + channels. Vice versa, a single G-protein may have more than one function. A single species of Gcts, for instance, is known to stimulate both AC and a Ca 2+channel (Birnbaumer et al., 1990b). Finally, as noted above, it is believed that the [3y-complex also plays a role in signaling systems. It may do this directly (as noted) or indirectly by subunit exchange. We are now in a position to understand the nature of the latter process. If both Gs and Gi G-proteins are present in a patch of membrane, it is possible for the I]y-complex of one to bind the tx-subunit of the other. Suppose that Gi is activated so that its 13y-complex dissociates from its tx-subunit. Then, if Gs is activated, its dissociation will be into a membrane space already populated with 13y-complexes. Thermodynamic considerations indicate that this diminishes the likelihood of the dissociation. Hence, the stimulatory effect of Gs is diminished. It is clear, then, that if we look at a 'plan' view of a membrane rather than the conventional 'elevation' we see a very intricate signaling network. Some possibilities are shown in Figure 9. It appears (Birnbaumer et al., 1990a,b) that the simplest case, 9a, is the only one for which it has been difficult to find examples. It must be kept in mind that we are only following the ct-subunits in Figure 9; the 13y-complexes may travel different routes. It must also be kept in mind that biomembrane fluidity, and hence G-protein diffusion constants, is likely (indeed known) to vary from place to place. The situation is thus (as usual) far from being straightforward;

Membrane Signaling Systems a)

1M1

b)

1M1

261 R1

R2 ~

1M2 ,,,

e)

1M1

G1 "

G1

E1 = R1

~ G1

E2 E3

d)

1M 1

1M2 "'

G1

~ E1

G2

= E2

R1

=

R1 .......

= G2 ~ / E 1

After Birnbaumeret al, 1990

Figure 9. Networks of signals in a biomembrane. Explanation and symbols in text.

there are many ways in which the diffusion of large molecules, such as proteins, may be constrained and the membrane may, indeed, be a mosaic of compartments within which, but not between which, G-proteins can shuttle (see McCloskey and Poo, 1984).

IX. EFFECTORS A N D SECOND MESSENGERS We have already noted, in passing, several of the effectors with which the o~subunits of G-proteins interact. The summary in Table 3 shows three enzymes and an ion channel, and several other effectors are becoming known. The earliest effector to be discovered was AC (Sutherland and Rall, 1957) and much progress has been made in working out its structure. It is a large glycoprotein (110 kDa) with 12 transmembrane segments (Figure 10). Its function is to catalyze the conversion of ATP into cAMP. cAMP diffuses into the cell's cytosol where it exerts its major effect on protein kinases. We have already seen that AC is under the control of both stimulatory (Gs) and inhibitory (Gi) G-protein systems. Finally, it is beginning to seem likely that it also acts as a K § channel, perhaps one of the K § channels which controls the resting potential (Vm) (Schultz at al., 1992). A second important and equally well-known effector is cGMP-phosphodiesterase (PDE; Figure 11). This enzyme is found in the disc membranes of the outer segments of vertebrate photoreceptor cells. When retinal in rhodopsin or iodopsin

262

C . U . M . SMITH AC

lili iili ATP

cAMP + 2P i

Various biochemical effects

Figure 10. Second messenger formation at adenylyl cyclase (AC). AC is shown as a 12-TM protein with a large cytosolic domain.

is activated by a photon, it changes from the bent 11-cis to the straight all-trans form. This activates the opsin molecule which, in the ways discussed above, leads to activation of PDE by a G-protein system (Gt or transducin). The activation consists of the dissociation of the two ),-subunits which releases the catalytic activity of the czl3-complex. This activity catalyzes the transformation of cGMP into 5'-GMP. This is important as cGMP is involved in maintaining the inward flow of Na § through minute channels in the boundary membranes of the outer segments. This flow constitutes the so-called 'dark current'. When the cGMP concentration is reduced, the Na § channels close. The 'dark-current' is interrupted, and the rod or cone cell hyperpolarizes. This voltage change across the membrane is detected by the neuron networks in the neural retina and ultimately signaled to the visual areas of the brain via the optic nerve and tract. As usual Gcz switches itself off by hydrolyzing GTP to GDP (assisted, as mentioned above, by the activating influence of the PDE ctl3-complex). The inward Na § current (the dark-current) is restored by fresh supplies of cGMP produced by the catalytic action of guanylyl cyclase on 5'-GMP.

Membrane Signaling Systems

263 PDE Jk,

cGMP

~._-.._-.._-.._-.._~._~ . . . . . . . . . . . . . . . .. .. .. .. .. .. .. . .

. .

. .

. .

. .

. .

5'GMP

. .

;:;5::5::3:<~::3:.-3: ................ ...~.~.~.~.~.~ ~-~-~-~-~-.~-.~ ~.~.~.~.~.~.~ . . . . . . . . . . . . . . ~..~..~..~..~..~..~. V. \-. \-. \-. \'.

Na

\'.

~?

,

Figure 11. Second messenger formation at phosphodiesterase (PDE). The figure shows a flow diagram for the inactivation of Na + channels in vertebrate photoreceptor outer-segments. Further explanation in text.

A third well-known effector is the enzyme, phospholipase C (PLC or PhLC; Figure 12). This enzyme (120 kDa) is located in the cytoplasmic leaflet ofbiomembranes and is known to exist in a number of isoforms. When activated by an appropriate Got-subunit it hydrolyzes phosphatidylinositol biphosphate (PIP2). PIP2 is itself derived by ATP-driven phosphorylation of phosphoinositol, a phospholipid present primarily in the inner leaflet of the membrane. The result of PLC hydrolysis is the production of two significant second messengers: inositol triphosphate (IP3 or InsP3) and diacylglycerol (DAG). Both these second messengers diffuse into the cytosol to exert biochemical effects. InsP3 causes the endoplasmic reticulum to release its stored Ca 2§ ions; hence the increased contractility of smooth muscle cells noted in Table 3. DAG acts on an important protein kinase, protein

Ca2+

various biochemical

J

11=3~,~ ER membranes

changes

~

Ca2+

F i g u r e 12. Second messenger formation at phospholipase C (PLC). When PLC is activated by an appropriate Go~, it hydrolyzes PIP2 into DAG and IP3 (InsP3) which have different effects (see text). Activation of protein kinase C (PKC) also requires the assistance of phosphatidyl serine (PS) which is normally present in the inner leaflet of the membrane. Further explanation in text.

Ca 2+

. ~:~:~;.~:~:~:~:~:~.:~:~:~;.~.~:;~:~:~:~.~:.::~:~:~?~i~:~?Z~:~Z~ / ~ ~

F i g u r e 13.

G-protein action on an ion channel. 264

. . .:.;~.~.~-::i

Membrane Signaling Systems

265 K|

J ATP

cA

Figure 14. Effectors and second messengers. Goc activates adenylyl cyclase (AC) so that the second messenger, cAMP, is produced, cAMP then influences the state of a K+ channel.

kinase C. PLC is regulated by both Gis and Gqs; both have a stimulatory effect but the PTX-insensitive Gqs appear to be more powerful. Once again, PLC has been shown to have GAP activity on its relevant Gt~-subunit. Less important and less well-known than phospholipase C is another membranebound phospholipase: phospholipase A2 (PLA2). It is G-protein coupled and when activated, arachidonic acid is liberated into the cytosol. It has become increasingly apparent in recent years that G-proteins interact not only with membrane-bound enzymes, but that they also have an important role in controlling ion channel conductivity. We have already mentioned that G~sl can affect the conductivity of the voltage-gated dihydropyridine Ca 2§ channel and that Gcti3 has a similar affect on some K § channels. These affects appear to be via the direct action of the t~-subunit on the channel-protein (Figure 13). Finally it has to be kept in mind that many of the second messengers mentioned above, especially cAMP and Ca 2§ are able to work back on the membrane to affect its biophysical characteristics, especially ion channel conductivity (Figure 14). In

266

C.U.M. SMITH

this way, complex feed-back and feed-forward modulations are achieved, which (fortunately) are far outside the scope of this chapter.

X.

PATHOLOGIES

As would be expected in any system as complex as the signaling systems we have been reviewing, things can go wrong. We have already noted (Section V) that many ct-subunits are affected by two important toxins: cholera and pertussis. Pertussis toxin (PTX) catalyzes ribosylation of a cysteine residue near the C-terminal end of tx-subunits. In so doing, it blocks the ability of the subunit to exchange GDP for GTP, and thus prevents its transformation from the inactive to the active state. As the t~-subunit most affected is G~i, which normally inhibits the activity of AC, the consequence is that the enzyme is left in a permanently 'on' state (Figure 15a). It catalyzes a continuous supply of its second messenger, cAMP, and this, among other things, stimulates excessive pumping of electrolytes. The electrolytes drag significant quantities of water with them. PTX is released by Bordatella pertussis, the bacterium responsible for whooping cough (pertussis), and the effect is most marked in the epithelial cells of the respiratory tract. An understanding of G-protein signaling systems thus enables us to understand the etiology of one of the most troubling childhood diseases. Related to PTX is cholera toxin (CTX). This also affects t~-subunits, especially Gt~. It inhibits the ATPase activity of the subunit and hence leaves the switch operating AC in the 'on' position (Figure 15b). Once again, the excess cAMP causes a continuous secretion of large quantities of electrolytes which carry large amounts of water with them. CTX exerts its major effects on the epithelial cells of the gastrointestinal tract m hence the severe diarrhea and dehydration which characterize patients suffering from a cholera infection.

XI.

CONCLUSION

This review of signaling systems serves to emphasize just how far the concept of biomembranes has evolved since the time of Davson and Danielli in the late 1930s. Biological membranes are no longer seen as the inert (somewhat leaky) packaging surrounding and holding in place the dynamic and fluid cytoplasms of cells, but as highly active boundary zones. It should, of course, not be forgotten that great systems of membranes (e.g., the endoplasmic reticulum and the Golgi apparatus) and membranous organelles (e.g., mitochondria) are found within cells and that these too, support complex signaling systems (see Ercolani et al., 1990). At the time of writing, the techniques of molecular biology and molecular pharmacology are providing a continuous flow of new information about G-proteins and membrane receptors. When this phase of discovery, description, and qualitative analysis has passed it will be succeeded by an attempt to develop a more quantitative and

Gio~

a)

PTX ~Gi~*

AO ATP

cAMP + PPi

L

b)

.• CTX--'L_"

stimulatespumpingof

electrolytes through epithelial cell membranes, especially in respiratory tract.

Gs~

1

AC ATP

cAMP + 2P i L

stimulates pumping of electrolytes through epithelial cell membranes, especially of the GI tract.

Figure 15. The effect of PTXand CTX on G-protein signaling systems.Explanation in text.

267

268

C . U . M . SMITH

predictive theory. A start has already be made in this enterprize by making use of quantitative data accumulated for the well-understood vertebrate phototransduction process (Lamb and Pugh, 1992). When a similar quantitative understanding of other G-protein signaling systems has been achieved, we can look forward to a comprehensive membrane dynamics and informatics. We shall then be a long step further forward in our understanding of the life of the cell in health and disease.

REFERENCES Agnew, W. S., Claudio, T., & Sigworth, E J. (1988). Molecular biology of ionic channels. Current Topics in Membrane Transport 33, 1--445. Anderson, P. H., Gingrich, J. A., Bates, M. D., Dearry, A., Falardean, P., Senogles, S. E., & Caron, M. G. (1990). Dopamine receptor subtypes: beyond the D1/D2 classification. Trends Pharmacolog. Sci. 11, 231-236. Arshavsky, V. Y., & Bownds, M. D. (1992). Regulation and deactivation of photoreceptor G-protein by its target enzyme and cGMP. Nature 357, 416-417. Berstein, G., Bank, J. L., Jhon, D.-Y., Exton, J. H., Rhee, S. G., & Ross, E. M. (1992). Phospholipase C-ctl is a GTPase-activating protein for Gq/11, its physiological regulator. Cell 70, 411-418. Birnbaumer, L., Abramowitz, J., & Brown, A. M. (1990a). Receptor--effector coupling by G-proteins. Biochim. Biophys. Acta 1031, 163-224. Birnbaumer, L., Mattera, R., Yatani, A., VanDongen, A., Graf, R., Sanford, J., Codina, J., & Brown, A. M. (1990b). Roles of G-proteins in coupling receptors to ion channels In: Transmembrane Signalling: Intracellular Messengers and Implications for Drug Development (Nahorski, S.R., ed.), pp. 43-71, Chichester, Wiley. Bonner, T. I., Buckley, N. J., Young, A. C., & Brann, M. R. (1987). Identification of a family of muscarinic genes. Science 237, 527-532. Bourne, H. R., Sanders, D. A., & McCormick, E (1991). The GTPase superfamily: conserved structure and molecular mechanism. Nature 349, 117-127. Buck, L., & Axel, R. (1991). A novel multi gene family may encode oderant receptors: a molecular basis for oderant recognition. Cell 65, 175-187. Coleman, D. E., Bergluis, A. M., Lee, E., Linder, M. E., Gilman, A. G., & Sprang, S. R. (1994). Structures of active conformations of Gitxl and the mechanism of GTP hydrolysis. Science 265, 1405-1412. Cotecchia, S., Schwinn, D. A., Randall, R. R., Lefkowitz, R. J., Caron, M. G., & Kobilka, B. K. (1988). Molecular cloning and expression of cDNA for the hamster txl-adrenergic receptor. Proc. Nat. Acad. Sci. USA 85, 7159-7163. Dixon, R. A. E, Kobilka, B. K., Strader, D. J., Benovic, J. L., Dohlman, H. G., Frielle, T., Bolanowski, M. A., Bennett, C. D., Rands, E., Diehl, R. E., Mumford, R. A., Slater, E. E., Sigal, I. S., Caron, M. G., Lefkowitz, R. J., & Strader, C. D. (1986). Cloning of the gene and cDNA for mammalian beta-adrenergic receptor and homology with rhodopsin. Nature 321, 75-79. Engelberg, D., Perlman, R., & Levitski, A. (1989). Transmembrane signalling in Saccarmomyces cerevisiae. Cell. Signaling, 1, 1-7. Ercolani, L., Stow, J. L., Boyle, J. E, Holtzman, E. J., Lin, H., Grove, J. R., & Ausiello, D. (1990). Membrane localisation of the pertussis toxin-sensitive G-protein subunits cti-2 and ~i-3 and expression of metallothionein---~i.2 fusion gene in LLC- Pkl cells. Proc. Natl. Acad. Sci. USA 87, 4635-4639. Fargin A., Raymond, J. R., Lohse, M. J., Kobilka, B. K., Caron, M. G., & Lefkowitz, R. J. (1988). The genomic clone G-21 which resembles an t~-adrenergic receptor sequence encodes the 5HTla receptor. Nature 335, 358-360.

Membrane Signaling Systems

269

Fasman, G. D., & Gilbert, W. A. (1990). The prediction of transmembrane sequences and their conformation: an evaluation. Trends Biochem. Sci. 15, 89-92. Federman, A. D., Conklin, B. R., Schrader, K. A., Reed, R. R., & Bourne, H. R. (1992). Hormonal stimulation of adenylyl cyclase through Gi protein [3ysubunit. Nature 356, 159-161. Frielle, T., Collins, S., Daniel, K. W., Caron, M. G., Lefkowitz T. J., & Kobilka, B. K. (1987). Cloning of the cDNA for the human O~l-adrenergicreceptor. Proc. Natl. Acad. Sci. USA 84, 7920-7924. Gantz, I., Schaeffer, M., Devalle, J., Logsdon, C., Campbell, V., Uhler, M., & Yamada, T. (1991). Molecular cloning of a gene encoding the histamine H2 receptor. Proc. Natl. Acad. Sci. USA 88, 429-433. Gao, B., & Gilamn, A. G. (1992). Cloning and expression of a widely distributed (type IV) adenylyl cyclase. Proc. Natl. Acad. Sci. USA 88, 10178-10182. Humphries, P., Kenna, P., & Farrar, G. (1992). On the molecular genetics ofretinitis pigmentosa. Science 256, 804-808. Henderson, R., & Unwin, P. N. T. (1975). Three-dimensional model of purple membrane obtained by electron microscopy. Nature 257, 28-32. Jackson, T. R., Blair, A. C., Marshall, J., Goedert, M., & Hanley, M. R. (1988). The mas oncogene encodes an angiotensin receptor. Nature 335, 437-440. Jacobson, K., Ishihara, A., & Inman, R. (1987). Lateral diffusion of proteins in membranes. Annu. Rev. Physiol. 49, 163-175. J/ihnig, E (1990). Structure predictions of membrane proteins are not that bad. Trends Biochem. Sci. 15, 93-95. Julius, D., MacDermott, A. B., Axel, R., & Jessell, T. M. (1988). Molecular characterization of a functional cDNA encoding the serotonin 1c receptor. Science 241,558-564. Kaziro, H., Itoh, H., Kazasa, T., Nakafuku, M., & Satoh, T. (1991). Structure and function of signal transducing GTP-binding proteins. Annu. Rev. Biochem. 60, 349-400. Klein, P. S., Sun, T. J., Saxe, C. L., Kimmel, A. R., Johnson, R. L., & Devreotes, P. N. (1988). A chemoattractant receptor controls development in Dictyostelium discoideum. Science 241, 14671472. Kleuss, C., Scherbtihl, H., Hescheler, J., Schultz, G., & Wittig, B. (1992). Different [3-subunits determine G-protein interaction with transmembrane receptors. Nature 358, 424-426. Kobilka, B. (1992). Adrenergic receptors as models for G-protein-coupled receptors. Annu. Rev. Neurosci. 15, 87-114. Kubo, T., Maeda, A., Sugimoto, K., Akiba, I., Mikami, A., Takahashi, H., Haga, T, Haga, K., Ichiyama, A., Kangawa, K., Matsuo, H., Hirose, T., & Numa, S. (1986). Primary structure of porcine cardiac muscarinic acetylcholine receptor deduced from cDNA sequence. FEBS Lett. 209, 367-372. Krupinski, J., Coussen, E, Bakalyar, H. A., Tang, W.-J., Feinstein, P. G., Orth, K., Slaughter, C., Reed, R. R., & Gilman, A. G. (1989). Adenylyl cyclase amino-acid sequence: possible channel or transporter-like structure. Science 244, 1558-1564. Lamb, T. D., & Pugh, E. N. (1992). G-protein cascades: gain and kinetics. Trends in Neurosci. 15, 291-298. Libert, E, Sciffman, S. N., Lefort, A., Parmentier, M., Gerard, C., Dumont, J. E., Vanderhaenger, J. J., & Vassart, G. (1991). The orhan receptor cDNA RDC7 encodes an A 1 adenosine receptor. EMBO J. 10, 1677-1682. Masu, Y., Nakayama, K., Tamaki, H., Harada, Y., Kuno, M., & Nakanishi, S. (1987). cDNA cloning of bovine substance-K receptor through oocyte expression system. Nature 329, 836-838. McCloskey, M., & Poo, M.-M. (1984). Protein diffusion in cell membranes: some biological implications. Internat. Rev. Cytol. 87, 19-81. Nathans, J., Thomas, D., & Hogness, D. S. (1986). Molecular genetics of human color vision: the genes encoding blue, green and red pigments. Science 232, 193-202.

270

C . U . M . SMITH

Numa, S., Noda, M., Takahashi, H., Tanabe, T., Toyosata, M., Furutani, Y., & Kikyotani, S. (1983). Molecular structure of the nicotinic acetylcholine receptor. Cold Spring Harbor Symp. Quant. Biol. 48, 57--69. Okamato, T., Katada, T., Murayama, Y., Ui, M., Ogata, E., & Nishimoto, I. (1990). A simple structure encodes G-protein-activating function of the IGF-11/mannose-6-phosphate receptor. Cell 62, 709-717. Pritchett, D. B., Bach, A. W. J., Wozny, M., Taleb, O., Toso, R. D., Shih, J. C., & Seeburg, P. H. (1988). Structure and functional expression of cloned rat serotonin 5HT2-receptor. EMBO J. 7, 41354140. Rarick, H. M., Artemeyev, N. O., & Hamm, H. E. (1992). A site on rod G-protein 13-subunit that mediates effector action. Science 256, 1031-1033. Reed, R. R. (1992). Signaling pathways in oderant detection. Neuron 8, 205-209. Schultz, J. E., Klumpp, S., Benz, R., Schirhoff-Goeters, W. J. C., & Schmid, A. (1992). Regulation of adenylyl cyclase from Paramecium by intrinsic potassium conductance. Science 255, 600-603. Schmitt, E O. (1986). In: Fast and Slow Chemical Signaling in the Nervous System (Iverson, L.L. and Goodman, E., eds.), pp. 239-243, Oxford University Press, New York. Shigemento, R., Yokoto, Y., Tsuchida, K., & Nakanishi, S. (1990). Cloning and expression of a rat neuromedin K receptor cDNA. J. Biol. Chem. 265, 623-628. Simon, M. I., Krikos, A., Mutoh, N., & Boyd, A. (1985). Sensory transduction in bacteria. Current Topics in Membranes and Transport 23, 3-15. Simon, M. I., Strathmann, M. P., & Gautam, N. (1991). Diversity of signal transduction. Science 252, 802-808. Smith, C. U. M. (1990). Elements of Molecular Neurobiology Wiley, Chichester. Spiegel, A. M., Backland, P. S., Butrynski, J. E., Jones, T. L. Z., & Simmonds, W. E (1991). The G-protein connection: molecular basis of membrane association. Trends Biochem. Sci. 16, 338-341. Sutherland, E. W., & Rail, T. W. (1957). The properties of an adenine ribonucleotide produced with cellular particles, ATP, Mg § and epinephrine or glucagon. J. Am. Chem. Soc. 79, 3608. Tanabe, Y., Masu, M., Ishii, T., Shigemoto, R., & Nakashini, S. (1992). A family of metabotropic glutamate receptors. Neuron 8, 169-179. Wittinghofer, A., & Pai, E. E (1991). The structure of the Ras protein: a model for a universal molecular switch. Trends Biochem. Sci. 16, 382-387. Yamashita, M., Fukui, H., Sugama, K., Horio, Y., Ito, S., Miziguchi, H., & Wada, H. (1991). Expression cloning of a cDNA encoding the bovine histamine Hi receptor. Proc. Natl. Acad. Sci. USA 88, 11515-11519. Yarden Y., Rodrquez, H., Wong, S. E K., Brandt, D. R., May, D. C., Burnier, J., Harkins, R. N., Chen, E. Y., Ramachandran, J., Ullrich, A., & Ross, E. M. (1986). The avian ~-adrenergic receptor: primary structure and membrane topology. Proc. Natl. Acad. Sci. USA 83, 6795-6799.

INDEX AC (adenylyl cyclase), 247, 258, 260262 n-AcCho receptor, 154 Acetylcholine receptor, 157 Acetylcholine-regulated channel, 241 N-Acetyl farnesyl cysteine, 92 Acholeplasma laidlawii strain B, 16 Activation energy, basal diffusion rate and, 228 Active transport, 229 Activity coefficients, 146, 158 Acylation, protein, 80-81 functions, 84-86 myristoylation and, 81-83 palmitoylation and, 83-84 S-Adenosyl methionine, 90 Adenylyl cyclase (AC), 247, 258, 260, 262 ADP-ATP carrier (ADP-ATP translocator), 154, 160, 163, 164, 170 Aerolysin, 63 Affinity labels, 171 Aggregation state, 209, 215 Alamethicin, 30, 56 All-t~ integral membrane proteins, 33-34 experimentally determined structures, 34-41 structure-prediction, 48-60 helix orientation, 50-53 overview, 49-50

271

simulation studies, 53-60 TM helix packing, 48-49 topology prediction, 41-48 evaluations of, 46-48 extension of hydrophobicity profile method, 44, 46 hydrophobicity profiles and, 4244 hydrophobicity scales and, 42, 43 t~-hydroxyfamesyl phosphoric acid, 94 Annealing, simulated, 51 Antiporter, 234, 235 Apocytochrome c, 175 Arrhenius plots, 241,242 Association constants, 146, 176 ATP/ADP carrier, 216 ATP synthase, 216 Avidin, 179 Bacterial membrane proteins, biosynthesis, 130-132 Bacteriorhodopsin, 31, 33, 34, 37, 215, 233, 241 hydrophobicity profile, 44, 45 structure, 35-37 Band 3 anion, 234 fl-barrels, 34, 62, 64 Bayer's patches, 131 fl:-adrenergic receptor, 58, 251 fitS-complex, 258 fl-hydroxybutyrate dehydrogenase, 144, 151

272

/3-sheets, 63, 167, 169 Bilayer-nonbilayer transitions, 8 Binding model, multi-site, 146 Biotin, 179 N-Biotinyl phosphatidylethanolamine, 179 N-Boc-S-all trans-f arnesyl-L-cysteine aldehyde, 93 Ca >, 8, 247 Ca2-ATPase, 20, 151,152, 164, 165 Ca2+-Mg2+-ATPase, 216-220 Carboxylmethylation, 89 Cardiolipin, 2, 8, 20, 155, 156, 160, 171,176, 216 Ceramide, 3 CFPs (channel-forming peptides), 30, 33 Chaetometillic acids, 94 Channel-forming peptides (CFPs), 30, 33 Chaperones, 114 Chloride ion channels, volumeregulated, 234 Cholera toxin (CTX), 179, 266 Cholesterol, 2, 6, 8, 9, 19, 20, 213 Choroideremia, 92 Circular dichroism, 46 Cryo-transmission electron microscopy, 12, 19, 65 CTX (cholera toxin), 179, 266 Cubic phases, 6, 8, 11, 12 Curvature concept, 10 CXC motif, 88 CXXX motif, 87 Cyclic AMP, 247 Cyclic GMP-phosphodiesterase, 260 Cytochrome bc~, 216 Cytochrome c, 175 Cytochrome c oxidase, 20, 154, 171, 216 Cytochrome oxidase, 151, 156, 159, 161,165

INDEX

DAG (diacylglycerol), 142, 174, 247, 263 Deacylation, 84 Debye-Hiickel theory, 158 DePaking methods, 14 DGDG (diglucosyl diacylglycerol), 9,17 Diacylglycerol (DAG), 142, 174, 247, 263 Differential scanning calorimetry, 5 Diffusion, 8 basal, rate of, 227-228 coefficient, 199 molecular models, 226 passive, forces leading to, 227, 226 protein catalysts, 228-242 Diglucosyl diacylglycerol (DGDG), 9,17 Dilauroylphosphatidylethanolamine, 188-189 Dimyristoylphophatidylcholine, 188189 Diphenylhexatriene (DPH), 196, 215 DM-20 myelin protein, 162, 165 DOPC, 7, 8, 10, 11 DOPE, 7, 8, 10, 11 DSC, 5, 6 Effectors, 261-266 EIS hydrophobicity scale, 42, 44 Electron diffraction, 30 Electron spin resonance, 4 Endoplasmic reticulum, 109 membrane protein retention, 119 protein targeting, 113-115 Endosome, 125 Enterotoxin, 68 Enzymes lipid-requiring, 139-145 in protein prenylation, 91-92 Equilibrium constant, 207 Erythrocyte membrane, 21

Index

ESR spectroscopy, 153, 174 Exchange lipid, 147 Farnesyl, 87 Farnesylation, 126 Farnesyl diphosphate, 87 Fatty acids, 80 Fatty acyl-chains, 189, 207, 217 Flippase, 233 Fluidity, membrane, 4, 17, 20, 210 ATPase activity and, 218 defined, 190, 198 measure, 196 polyunsaturated fatty acyl-chains and, 192 transition state theory and, 201 Fluid-mosaic model, 2 Fluorescence depolarization, 190 Fluorescence quenching, 154 Fourier transform infrared, 47 Free fatty acids, 156 Free volume theory, 198 Freeze-fracture electron microscopy, 6,7,8,12 Frictional effects, 201 Fusion, membrane, 2, 4 lipids and, 11-13 vesicle, 121-122 8-toxin, 30, 51, 56 Ganglioside GM1, 179 Gangliosides, 157 Gap junctions, 230 GAPs (GTPase activating proteins), 254, 256 Gated-pores, 233 Gating rules, 234, 235, 236 Gel phase, 4, 19, 210, 218 Gel phase lipids, 164-165 Geranylgeranyl, 87 Geranylgeranyl diphosphate, 87 GES hydrophobicity scale, 42, 45 Glucose transporter, 210

273

Glucose uniporter, 234 Glyceroglycolipids, 3 Glycerophospholipids, 2 Glycolipids, 4, 9 Glycophorin, 54 P-Glycoprotein, 233, 234 Glycospingolipids, 3 Glycosylation, 122, 123 Glycosyl phosphatidylinositolanchored proteins, 180 Glycosyl-phosphatidylinositol linkage, 80 GMP-phosphodiesterase, 261 GNRPs (guanine nucleotide release proteins), 254, 256 Golgi complex budding reactions, 120 retention in, 122, 124 G-protein(s), 247, 252-254 classification, 256 Heterotrimeric, 85 membrane-bound, 255-258 signaling systems, 258-259 networks of, 259-261 a-subunits, 85 G-protein-coupled receptors, 58 Gramicidin, 20 GTPase activating proteins (GAPs), 254, 256 Guanine nucleotide dissociation inhibitors, 95 Guanine nucleotide dissociation stimulators, 95 Guanine nucleotide exchange proteins, 95 Guanine nucleotide release proteins (GNRPs), 254, 256 Halorhodopsin hydrophobicity profile, 44, 45 structure, 31, 33, 34, 38 Hexagonal Hn phase, 6, 8, 10, 12, 16, 20, 21

274

HMGCoA reductase, 87 Hydrophobicity, 33, 42 basal diffusion rate and, 227-228 profiles, 43 bacteriorhodopsin, 44, 45 extension of method, 44, 46 halorhodopsin, 44, 45 rhodopsin, 44, 45 topology prediction and, 42-44 scales, topology prediction and, 42, 43 Hydrophobic matching, 152 Hydrophobic mismatch, 21 Hydrophobic moment, 46 Hydrophobic thickness, 190, 207 ILA (interlamellar attachments), 12 IMI (inverted micellar intermediates), 12 InsP3, 247, 263 Integral membrane proteins (IMPs), 31, 71-72, 165 all-a. See All-a integral membrane proteins all-/3, 33, 34 experimentally determined structures, 60-63 related structures, 63-65 a~ fl, 65-71 nicotinic acetylcholine receptor, 65-69 voltage-gated ion channels, 6971 classification, 33 headgroup dependence of lipid selectivity and, 153-165 mixed topologies, 33, 34 single TM helix, 32-33 structure, lipid selectivity and, 165-171 Interbilayer attachment sites, 8 Intercellular signaling systems, 249 Interhelix loops, 35

INDEX

Interlamellar attachments (ILA), 12 Inverted micellar intermediates (IMI), 12 in vitro translation systems, 111 Inward rectifier, 71 Ion channels, 33, 65, 232 I~: protein, 157 Lk protein, 167 Isoprenoids, 87 Isoprenylation, 80 Isotropic phases, 8 K+ channels, 69, 157, 169 KD hydrophobicity scale, 42 Lac permease, 49 Lactose, 236 Lactose-H + carrier, 234, 237 Lactose permease, 240 Lac Y, 234 Large unilammelar vesicles, 12 Lateral diffusion, 4 Light harvesting complex (LHC), 31, 33, 34, 41 d-Limonene, 93 Linoleic acid, 193 Linolenic acid, 192 Lipid(s), 2, 188-198 bilayers, 2, 14, 189, 210 orientation order in, 13-18 structural characteristics, 4, 8 10, 16 class, lipid selectivity and, 154-158 competition, 177, 179 diversity, 2-4 effects on membrane proteins, 206-220 exchange rates, 162 gel phase, lipid selectivity and, 164-165 headgroup selectivity, 174 interaction with proteins, 138, 188 membrane fusion and, 11-13

Index

mixed systems, 177 modifications, 80 glycosyl-phosphatidylinositol linkage, 80 isoprenylation, 80 of membrane proteins, 126-127 myristoylation, 81-83 palmitoylation, 83-84 monolayers, 175, 192 order, 13 phase transitions, 4-6 physical properties, 4-6 polymorphism, 6-11 protein-associated, exchange rates of, 162-164 protein function and, 20-21 residues, 171 selectivity acyl-chain dependence of, 148153 gel phase lipids and, 164-165 headgroup dependence of, 153165 integral protein structure and, 165-171 ionic strength dependence of, 158-160 of peripheral proteins, 172-180 pH dependence of, 160-162 themodynamics/dynamics of, 146-148 sorting, 180 viscosity-dependent effects, 198206 Liquid-crystalline bilayer phase, 4, 210,214 Liquid-ordered fluid-phase, 4, 6 Local anesthetic, 157 LUVs (large unilamellar vesicles), 12, 19 Lysine, 171 Lysine peptides, 172 Lysine residues, 170

275

Lysosomal acid phosphatase, 126 Lysosomal membrane proteins, 126 Lysozyme, 175 Manumycin, 93 Mass action effects, 237 "Mattress model," 20 M 13 bacteriophage coat protein, 154, 163, 164, 167 Melittin, 30 Membrane(s), 2 basal transport properties, 226-228 curvature, 19 fluidity. See Fluidity, membrane fusion. See Fusion, membrane order, 18, 20 osmotic properties, 18-19 permeability, 18-19 proteins. See Protein(s) receptors, 250-252 shape, 19 signaling systems. See Signaling systems thickness, 17 transport, 225 Membrane-spanning helices, 239, 241 Methyltransferase, 90 Mevalonic acid, 87 Mga-ATPase, 21 MGDG (monoglucosyl diacylglycerol), 9, 17 M2 helix, 66 Micelles, 8 Microviscosity, 196 Mitochondria, 216 Mitochondrial membrane, 128, 176 Modeling, 4, 54 Monogalactosyl-diglyceride, 157 Monoglucosyl diacylglycerol (MGDG), 9, 17 Monolayers, lipid, 175, 192 Motional averaging, 8

276

Motional freedom, 189 Multidrug resistance, 233, 234 Multilamellar vesicles, 12 Myelin basic protein, 172, 174, 175, 177 Myelin D M-20 protein, 164 Myelin proteolipid protein, 154, 159, 160, 162, 164, 165 Myristate, 80 Myristolylation, 81-83, 126 Myristoylation, 80 Myristoyl-CoA, 82 Myristoyl-CoA:protein N-myristoyl transferase, 81 Na+, K§ 151 Na+ channel, electrically-gated, 231 nAChRs. See Nicotinic acetylcholine receptors NADH dehydrogenase, 216 Na+,K+-ATPase, 154, 156, 159, 160 Nicotinic acetylcholine receptors, 211,246, 251 classification, 33 importance, 69 structure, 31, 34, 65-68 verotoxin, 68 Nonbilayer intermediates, 12 Nonbilayer lipids, 9, 11, 13 Nonbilayer phases, 6, 10, 11, 20 Oil/water partition coefficients, 227 Oligo-arginine peptides, 172 Oligomers, 119 Order parameter, 190, 218 Order parameter profiles, 14, 16 Orientation order, in lipid bilayers, 13-18 Pake doublets, 14 Palmitate, 80 Palmitoylation, 80, 83-84 Palmitoyl-CoA, 82

INDEX

Palmitoyl transferase, 83 PA (phosphatidic acid), 8, 12 Partial molar volume, 215 Patulin, 93 PC (phosphatidylcholine), 2, 4, 8, 10, 21,144, 192, 217 PCR (polymerase chain reaction), 250 1,4-Pentadiene, 192 PE (phosphatidylethanolamine), 2, 4, 8, 9, 21,209 Peripheral proteins classification, 31 lipid selectivity, 172-180 Permeability coefficient (P), 18 Pertussis, 266 Pertussis toxin (PTX), 266 Phase transition, 192, 210, 217 Phosphatidic acid (PA), 8, 12 Phosphatidylcholine (PC), 2, 4, 8, 10, 21,144, 192, 217 Phosphatidylethanolamine (PE), 2, 4, 8, 9, 21,209 Phosphatidylinositol, 2 Phosphatidylserine (PS), 2, 4, 8, 12, 20, 21,142 Phosphodiesterase, 247 Phospholipase, 263 Phospholipase A2, 21, 139, 265 Phospholipase C, 247 Photoactivatable lipids, 171 Photosynthetic reaction center (PS/ RC), 31, 33, 34, 37-41,188 Photosystem I, 31, 33, 34, 41 Ping-pong kinetics, 235 pKa shift, 161 Poly-lysine, 175 Polymerase chain reaction (PCR), 250 Polyunsaturated fatty acyl-chains, 197 POPC, 3, 15 POPE, 10

Index

Porin-like proteins, 63-64 Porins structure, 31, 33, 34, 60-63 transmembrane transport, 230, 231 Positive inside rule, 44, 46 Potassium ion channels, 69, 157, 169 p21 protein, 253 Prenylation, protein, 86-95 associated modifications, 89-91 enzymology, 91-92 function, 94-95 inhibition of C-terminal processing and, 92-94 isoprenoid biosynthesis and, 87 Primary active transport, 230 Primary messengers, 248-250, 250 Product inhibition, 237 Proline, 35, 58 Protein disulfide isomerase, 118 Protein:farnesyl transferase, 91 Protein:geranylgeranyl transferase I, 91 Protein kinase C, 20, 142 Protein-lipid interactions, 20 Proteins, membrane, 17, 21. See also specific membrane proteins activity, 20 acylation, 80-86 functions of, 84-86 myristolylation and, 81-83 palmitoylation and, 83-84 bacterial, biosynthesis of, 130-132 classification, 31-34 C-terminal anchor, 118 cytoplasmic, 132 disulfide pairing, 118 folding, 118 function, lipids and, 20-21 insertion into phospholipid bilayer, 109 integral. See Integral membrane proteins (IMPs)

277

interaction with lipids, 138, 188 lipid effects on, 206-220 lipid modifications, 126-127 lysosomal, 126 mitochondrial, biosynthesis of, 127-130 multi-spanning, assembly of, 117118 newly formed, transport in vesicles, 120-122 outer, 130-132 peripheral, 31-32 prenylation. See Prenylation, protein single-pass, assembly of, 115-117 structures, 30, 31 synthesis mechanisms of, 108-109 processing in ER lumen, 118119 targeting, 109-111 in vitro translation systems, 111113 Proton well, 238 PS (phosphatidylserine), 2, 4, 8, 12, 20, 21,142 PS / RC (photosynthetic reaction center), 31, 33, 34, 37-41, 188 Pyruvate oxidase, 149 Quadrupolar echo pulse method, 14 Quadrupolar splitting, 14 Rab proteins, 88, 92 Ras oncogenes, 253 Ras proteins, 87 Rate-limiting step of transport, 242 Recruitment, 237, 238 Rhodopsin, 151,154, 165, 213 hydrophobicity profile, 44, 45 hydrophobic matching with lipids, 152

278

lipid exchange off-rate constant, 164 structure, 21, 34, 38 Ridges-in-grooves helix, 35, 40 Salt-bridge, 240 Sarcoplasmic reticulum, 216 Secondary active transport, 230 Secondary structure prediction method, 46 Second messengers, 261-266 Semipermeable barriers, 2, 18 Shape hypothesis, 10 Signaling systems, 245-246, 268 effectors, 261-266 general orientation, 246-248 G-proteins and, 252-254 membrane-bound G-proteins and, 255-258 membrane receptors and, 250-252 pathologies, 266, 267 primary messengers and, 248-250 second messengers and, 261-266 Signal recognition particle, 113 Signal sequences, 113 Small unilamellar vesicles (SUVs), 12, 19 SNAPs, 121 SNAREs, 121 Sodium ion channel, electricallygated, 231 Solute size, basal diffusion rate and, 228 Sphingolipids, 2 Sphingomyelin (SPM), 3, 4, 8 Spin-labeled lipids, 153, 159, 174 SPM (sphingomyelin), 3, 4, 8 Spontaneous radius of curvature, 10, 12 Steptolysin O, 230 Stokesian diffusion, 228 SUVs (small unilamellar vesicles), 12, 19 Symporter, 234, 235, 236

INDEX

Targeting apocytochrome, 176 endosomal, 125 glycosyl phosphatidylinositolanchored proteins, 180 newly-synthesized membrane proteins, 109-110 Temperature, phase-transition, 192, 242 Thioester, 81 Thylakoid, 157 TM helices, 33 all-a-topologies, 33-34 experimentally determined structures, 34-41 structure-prediction, 48-60 topology prediction, 41-48 all-fl-topologies, 34 amphipathicity, 51-53 bundles, 53, 56-57 dimers, 53-56 mixed topologies, 34 orientation, 50 packing, 48 simulation studies, 53-60 proline in, 58-60 sequence variability, 50-51 single, 32-33 termini, 46 7-TM structure, 252, 259 Topology prediction, 42 all-/3 IMPs, 64 all-a integral membrane proteins. See under All-t~ integral membrane proteins Toxins cholera, 179, 266 enterotoxin, 68 d-toxin, 30, 51, 56 membrane-active, 33 pertussis, 266 verotoxin, 65-68 Transducin, 260, 262

Index

trans-Golgi network, 109 Transition state theory, 199 Transit sequences, 128 Translocase complex, 114 Transmembrane (TM) domains, 30 helices. See TM helices potential difference, 231 topology, 69, 166 Transmission coefficient, 200, 201 Transport proteins, 33 Trans-stimulation, 238 Trimeric G-protein, 255 Uniporter, 234, 235 Unsaturated fatty acyl-chains, 192

279

VDACs (voltage-dependent anion channels), 64 Verotoxin, 65-68 Vesicles, 8, 121-122 Viscosity, 196-198, 201-206 Voltage-dependent anion channels (VDACs), 64 Voltage-gated ion channels, 34, 6971 Whooping cough, 266 XCC motif, 88 X-ray diffraction, 6, 11, 30, 188 Zaragozic acids, 94

Advances in Molecular and Cell Biology

.1 A l P R E S S

Edited by E. Edward Bittar, Department of Physiology, University of Wisconsin-Madison

Volume 7, Biology of the Cancer Cell 1993, 268 pp. ISBN 1-55938-624-X

$97.50

Edited by Gloria H. Heppner, Director, Breast Cancer Biology Program, Michigan Cancer Center

CONTENTS: Introduction. Cellular Genetic Alterations; Models of Breast and Colon Cancer, S.R. Wolman, D. Visscher, Wayne State University School of Medicine. Altered Expression of Transforming Growth Factor ~ and Transforming Growth Factor-~ Autocrine Loops in Cancer Cells, M.G. Brattain, Medical College of Ohio, K.M. Mu/der, Pennsylvania State University, School of Medicine, S.P. Wu, G. Howell, L. Sun, Bay/or College of Medicine, J.K.V. Wilson, Case Western Reserve University, B.L. Ziober, Bay/or College of Medicine. Altered Signal Transduction in Carcinogenesis, Catherine A. O'Brian, Nancy E. Ward, Constantin G. Ioannides, Universty of Texas, M.D. Anderson Cancer Center, Houston. The Significance of the Extracellular Matrix in Mammary Epithelial Carcinogenesis, Calvin D. Roskelley, Ole W. Petersen, Mina J. Bissell, Lawrence Berkeley Laboratory. EpitheliaI-Stromal Cell Interactions and Breast Cancer, Sandra Z. Has/am, Laura J. Counterman, Katherine A. Nummy, Michigan State University. The Tissue Matrix and the Regulation of Gene Expression in Cancer Cells, Kenneth J. Pienta, Wayne State University School of Medicine, Brian C. Murphy, Robert H. Getzenberg, Donald S. Coffey, Johns Hopkins School of Medicine. Tumor Cell Interactions in Cancer Growth and Expression of the Malignant Phenotype, Fred R. Miller, Bonnie E. Miller, Michigan Cancer Foundation. Effect of Class I MHC Gene Products on the Immunobiological Properties of BL6 Melanoma, Misoon Kim, Elieser Gorelik, Pittsburgh Cancer Institute. The Role of Angiogenesis in Tumor Progression and Metastasis, Janusz W. Rak, Erick J. Hegmann, Robert S. Kerbel, Reichmann Research Institute. Subject Index.

Volume 8, Organelles In Vivo 1994, 191 pp. ISBN 1-55938-636-3

$97.50

Edited by Lan Bo Chen, Dana-Father Cancer Institute, Harvard Medical School

CONTENTS: Confocal Redox Imaging of Cells, Barry R. Masters, Walter Reed Army Medical Center. Calcium Channels and Vasodilation, Alison M. Gumey and Lucie H. C/app, UMDS, St. Thomas Hospital, London. Changes in DNA Supercoiling Status of Cells Treated with Antineoplastic Drugs, William D. Wright and J.L. Roti Roti, Washington University School of Medicine. Func-

tional Morphology of the Golgi Region: A Lectino-Electromicroscopic Exploration, Margit Pavelka and Adoff Ellinger, Universitdt Innsbruk. Receptor Mediated Endocytosis of Plasminogen Activators, Gouujun Bu, Phillipa A. Morton, and Alan L. Schwartz, Washington University School of Medicine. Lipid Trafficking in Hepatocytes: Relevance to Biliary Lipid Secretion, Kristien J.M. Zaal, Jan Willem Kok, Folkert Kuipers, and Dick Hoekstra, University of Gro;ningen, Faculty of Medicine. Identification and Characterization of Functional Secretory Cells: Advantages of Multiparameter Flow Cytometry Kinetics, Elizabeth R. Simons and Theresa A. Davies, Boston University. An Outline of Neurosecretion, Jane Somsel Rodman, Tufts University School of Medicine.

Volume 9. Homing Mechanisms and Cellular Targeting 1994, 303 pp. $97.50

.1 A l

ISBN 1-55938-686-X Edited by Bruce R. Zetter, Harvard Medical School and Childrens Hospital

CONTENTS: Introduction, Bruce R. Zetter. Directed Cell Migration in Embryonic Blood Vessel Assembly, Thomas J. Poole, SUNY Health Science Center at Syracuse. Leukocyte Interaction with Endothelium and Extracellular Matrix: The Role of Selectins and CD44, Ivan Stamenkovic, Massachusetts General Hospital and Harvard Medical School Oligosaccharide-dependent Mechanisms of Leukocyte Adhesion, John B. Lowe, Univesity of Michigan Medical School. Cell Adhesion Molecules: Novel Therapeutic Targets for Chronic Inflammatory Diseases of the Central Nervous System, Gregory N. Dietsch, Gary M. Peterman and W. Michael Gallatin. Molecular Mechanism of Targeting of Hemopoietic Stem Cells to the Bone Marrow After Intravenous Transplantation, Mehdi Tavassoli, University of Mississippi School of Medicine. Tumor Cell Adhesion and Growth in Organ Preference of Tumor Metastasis, Gareth L. Nicolson, The University of Texas M.D. Anderson Cancer Center. Cancer Cell Chemotaxis: Mechanisms and Influence on Site-Specific Tumor Metastasis, F. William Orr, McMaster University. Experimental Orthotopic Models of Organ-specific Metastasis by Human Neoplasms, Isaiah J. Fidler, The University of Texas, M.D. Anderson Cancer Center. Inter and Intracellular Targeting of Drugs, Smadar Cohen, Ben Gurion University of the Negev and Robert Langer, Massachusetts Institute of Technology. Chimeric Molecules Constructed with Endogenous Substances, Gregory T. Lautenslager and Lance L. Simpson, Jefferson Medical College. Organ-Specific Targeting of Synthetic and Natural Drug Carriers, S. Moein Moghimi, Lisbeth Ilium and Stanley S. Davis, University of Nottingham.

Volume 10, Molecular Processes of Photosynthesis 1994, 437 pp. $97.50 ISBN 1-55938-710-6

P R E S S

Edited by J. Barber, Imperial College of Science, Technology and Medicine, London

J A l P R E

S S

CONTENTS: Organisation and Dynamics of Thylakoid Membranes, B. Andersson, Stockholm University and J. Barber Wolfson Laboratories, England. Antenna Pigment-Protein Complexes of Higher Plants and Purple Bacteria, J.P. Thomber, University of California, R. Cogdell, The University of Glasgow, Scotland, P. Chitnis, D.T. Morishige, G.F. Peter, S. Gomez, S. Anandan, S. Preiss, B. Welty, A. Lee, T. Takeuchi, C. Kerfield, Kansas State University. Adaptive Variations in Physobilisome Structure, A.N. Glazer, University of California. Photoprotection and Photoinhibitory Damage, W.S. Chow, CSlRO Division of Plant Indutry, Australia. Molecular Genetic Manipulation and Characterization of

Mutant Photosynthetic Reaction Centres from Purple Non-Sulfur Bacteria, E. Takahashi, C.A. Wraight, University of Illinois. PROTON-TRANSLATING NAD(P)-H Transhydrogenase and NADH Dehydrogenase in Photosynthetic Membranes, J.B. Jackson, University of Birmingham and A. McEwan, University of East Anglia, England. Structural Elements Involved in the Assembly and Mechanims of Action of Rubisco, S. Gutteridge, Dupont Company, T. Lundqvist, Swedish University of Agriculture. The Ferredoxin/Thioredoxin System: Update on its Role in the Regulation of Oxygenic Photosynthesis, R. Buchanan, University of California. Identification, Cellular Localization and Participation of Chaperonins in Protein Folding, A.A. Gatenby, P. Viitanen, E.L duPont de Nemours & Co., V. Speth, lnstitut fur Biologie, R. Grimm, Hewlett-Packard GmbH. Translocation of Proteins Across Chloroplast Membranes, B. Bruce, K. Keegstra, University of Wisconsin, Madison. Volume 11, In preparation, Summer 1995

ISBN 1-55938-844-7

Approx. $97.50

List of Contributors. Preface, Kevin M. Brindle. Metabolic Channeling in Organized Enzyme Systems: Experiments and Models, Pedro Mendes, Douglas B. Kell, and G. Rickey Welch. Metabolic Control Analysis in Theory and Practice, Athel Cornish-Bowden. Experimental Approaches to Studying Enzymes in Vivo: The Application of Nuclear Magnetic Resonance Methods to Genetically Manipulated Organisms, SimonPeter Williams, Alexandra M. Fulton, and Kevin M. Brindle. Glycolysis in Vivo: Fluorescence Microscopy as a Tool for Studying Enzyme Organization in Living Cells, Len Pagliaro. The cooperative Behavior of Krebs Tricarboxylic Acid Cycle Enzymes, Paul A. Srere, Craig R. Malloy, A. Dean Sherry, and Balazs Sumegi. NMR Studies of Erythrocyte Metabolism, Hilary A. Berthon and Philip W. KucheL Studies of Physiological Control of ATP Synthesis, K. F. LaNoue and C. Doumen. Hepatic High Energy Phosphate Metabolism in Transgenic Livers Expressing Creatine Kinase as Reveled by 31p NMR, Alan P. Koretsky, Kenneth R. Miller and Jessica M. Halow. Index. CONTENTS:

Also Available: Volumes 1-6 (1987-1993)

$97.50 each

Advances In Cell and Molecular Biology of Membranes and Organelles (Previosly published as Advances In Cell and Molecular Biology of Membranes) Edited by Alan M. Tartakoff, Institute of Pathology, Case Western Reserve University Volumes of Advances in Biochemistry and the Biology of Cells are intended to present interrelated reviews concerned with structure and function of membranes of prokaryotes and eukaryotes. Topics may include membrane protein and lipid biosynthesis, membrane-cytosketal relations, membrane permeability, signal transduction, etc. Individual volumes will have a common theme.

Volume 1, Endosomes and Lysosomes: A Dynamic Relationship 1993, 434 pp. ISBN 1-55938-362-3

$97.50

Edited by Brian Storrie, Department of Biochemistry and Nutrition, Virginia Polytechnic Institute and State University and Robert F. Murphy, Department of Biological Sciences, Carnegie Mellon University

CONTENTS: Preface, Brian Storrie, Virginia Polytechnic Institute and State University and Robert F. Murphy, Carnegie Mellon University. Models of Endosome and Lysosome Traffic, Robert F. Murphy, Carnegie Mellon University. Endocytic Receptors, Michael G. Roth, University of Texas Southwestern. Functions of the Mannose 6-Phosphate Receptors, Bernard Hoflack, European Molecular Biology Laboratory, Germany and Peter Lobel, University of Medicine and Dentistry of New Jersey. Chemistry of Lysosomal Cysteine Proteinases, Robert W. Mason, Virginia Polytechnic Institute and State University and Donna Wilcon, Washington University School of Medicine. Mechanism and Regulation of Autophagic Degradation of Cellular Proteins, William A. Dunn, Jr., University of Florida. Cell-Free Systems for Endocytosis, William A. Braell, Harvard Medical School Genetic Analysis of Membrane Traffic in Mammalian Cells, Penelope A. Colbaugh and Rockford K. Draper, The University of Texas at Dallas. Plasma Membrane Lipid Transport in Cultured Cells: Studies Using Lipid Analogs and Model Systems, Michael Koval, Washington University School of Medicine. Endosomes, Lysosomes, and Trans-Golgi-Related Systems in Conventional Neurons and the Grof Retina: Shards and Suppositions, Eric Holtzman, Eliene Augenbraun, Robert St. Jules, and Maria Santa-Hernandez, Columbia University. The Role of Endocy-

.1 A [ P R E S S

J A [

tosis in Epidermal Growth Factor Signaling, Bryan K. McCune, Johns Hopkins University School of Medicine, William R. Huckle, and H. Shelton Earp, University of North Carolina at Chapel Hill. Membrane Traffic Through the Late Stages of the Yeast Secretory, Eric A. Whitters, Henry B. Skinner, and Vytas A. Bankaitis, University of Alabama at Birmingham. Regulation of Lysosomal Trafficking and Function During Growth and Development of Dictyostelium Discoideum, James A. Cardelli, Louisiana State University Medical Center. Towards an Understanding of the Inheritance of Mammalian Lysosomes and Yeast Vacuoles, Brian Storrie, Virginia Polytechnic Institute and State University. Volume 2, M e m b r a n e Transport in Protozoa

1993,483 pp. 2 Part Set Set ISBN 1-55938-628-2

$195.00

Edited by Helmut Plattner, Fakult~t fearBiologie, Universit&t Konstanz

P R E S S

P A R T A - C O N T E N T S : Preface. Involvement of the TransGolgi Network, Coated Vesicles, Vesicle Fusion and Secretory Product Condensation in the Biogenesis of Pseudomicrothorax Trichocysts, Robert K. Peck, Barbara Swiderski and Anne-Marie Tourme/, Geneva, Switzerland. Early Steps of the Secretory Pathway in Paramecium: Ultra-structural, lmmunocyto-Chemical and Genetic Analysis of Trichocyst Biogenesis, Nicole Garreau de Loubresse, Gif-sur-Yvette, France. Calcium and Trichocyst Exocytosis in Paramecium: Genetic and Physiological Studies, Jean Cohen and Daniel Kerboeuf, Gifsur-Yvette, France. Exocytotic Events During Cell Invasion by Apicomplexa, Jean Francois Dubremeta and Roff Entzeroth, Villeneuve dAsca, France/Bonn, FRG. Pathways of Lysosomal Enzyme Secretion in Tetrahymena, Arno Tiedtke, Thomas Kiy, Christian Vossk(Jhler and Left Rasmussen, M(Jnster, FRG/Odense, Denmark. Synchronization of Different Steps of the Secretory Cycle in Paramecium Tetraurelia: Trichocyst Exocytosis, Exocytosis-Coupled Endocytosis and Intracellular Transport, Helmut Plattner, Gerd Knoll, and Regina Pape, Konstanz, FRG. The Ciliary Membrane and its Engagement In Conjugation, Jason Wolfe, Middletown, CT. Ciliary and Plasma Membrane Proteins in Paramecium: Description, Localization and Intracellular Transit, Yvonne Capdeville, Ren#e Charret, Claude Anthony, Julienne Delorme, Pierre Nahon and Andre. Adoutte, Orsay, Francea. P A R T B -- C O N T E N T S : Endocytosis and Intracellular Transport of Variant Surface Glycoproteins in Trypanosomes, Michael Duszenko and Andreas Seyfang, T(Jbingen, FRG. A Comparative Survey on Phagosome Formation in Protozoa, Klaus Hausmann and Renate Radek, Berlin, FRG. Endosomal Membrance Traffic of Ciliates, Richard D. Allen and Agnes K. Fok, Honolulu, HI. Membrane Flow in the Digestive Cycle in Paramecium, Agnes K. Fok and Richard D. Allen, Honolulu, HI. Signal Coupling During Endocytosis in Amoeba Proteus,

Robert D. Prusch, Spokane, WA. Membrane Recycling and Turnover in Large, Free-Living Amoebae, Kwang W. Jeon, Knoxville, TN. Food Uptake and Digestion in Amoebae, Wilhelm Stockem and Melpo Christofidou-Solomidou, Bonn, FRG. The Lysosomal System in Malaria Parasites, Christian Slomianny, Villeneuve d'Ascq, France. Membrane and Microtubule Dynamics in Heliozoa, Toshinobu Suzaki and Yoshinobu Shigenaka, Hiroshima, Japan. The Host-SymbiontInterface in Ciliate-Algae Associations: Inhibition of Membrane Fusion, Wemer Reisser, G6ttingen, FRG. Lipid Composition of Membranes Involved in Membrane Traffic in Tetrahymena, Shigenobu Umeki and Yoshinori Nozawa, Okayama, Japan.

Volume 3 Signal Transduction Through Growth Factor Receptors 1994, 223 pp. ISBN 1-55938-344-5

.1 A l

Approx: $97.50

Edited by: Yasuo Kitagawa, BioSciences Center, Laboratory of Organogenesis, Nagoya University and Ryuzo Sasaki, Faculty of Agriculture, Departrnent of Food Science and Technology, Kyoto University,

CONTENTS: The Hepatocyte Growth Factor/c-MET Signaling Pathway, D. P. Bottario, A. M.-L. Chan, J. S. Rubin, E. Gak, E. Fortney, J. , Schindler, M. Chedid, and S., A. Aaronson. Insulin Receptor, Y. Ebina, H. Hayashi, F. Kanai, S. Karnohara, and Y. Nishioka. Interleukin-3 Receptor: Structure and Signal Transduction, 7". Kitarnura, and A. Miyajirna. Interleukin-5 Receptor, K. Takatsu. Interleukin-6 Receptor and Signal Transduction, 7. Matsuda, 7. Nakajirna, 7. Kaisho, K. Nakajirna, and 7-. Hirano. Receptor for Granulocyte ColonyStimulating Factor, S. Nagata, & R. Fukunaga. Receptor for Granulocyte/Macrophage Colony-stimulating Factor, K. Kurata, T. Yokota, A. Miyajirna, & K. AraL Perspectives On The Structure And Mechanisms Of Signal Transduction By The Erythropoietin Receptor, S. S. Jones. Interleukin-1 Signal Transduction, J. E. Sims, T. A. Bird, J. G. Giri, and K. S. Dower Volume 4, Protein Export and Membrane Biogenesis In preparation, Spring 1995 ISBN 1-55938-924-9 Approx. $97.50 Edited by Ross E. Dalbey, Department of Chemistry, The Ohio State University

CONTENTS: Introduction to the Series, Alan M. Tartakoff, Case Western Reserve University. Preface, Ross E. Dalbey, The Ohio State University. Membrane Protein Assembly, Paul Whitley and Gunnar von Heijne, Karolinska Institute Center of Structural Biochemistry. Membrane Insertion of Small Proteins: Evolutionary and Functional Aspects, Dorothee Kiefer

P R E S S

J A l P R E

S S

and Andreas Kuhn, Universitat Karlsruhe (TH). Protein Translocation Genetics, Koreaki Ito, Kyoto University. Biochemical Analyses of Components Comprising the Protein Translocation Machinery of Escherichia Coli, Shin-ichi Matsuyama, University of Tokyo and Shoji Mizushima, Tokyo College of Pharmacy. Pigment Protein Complex Assembly in Rhodobacter Sphaeroides and Rhodobacter Capsulatus, Amy R. Vargas, Abbott Laboratories, Illinois and Samuel Kaplan, University of Texas Health Science Center at Houston. Identification and Reconstitution of Anion Exchange Mechanisms in Bacteria, Atul Varadhachary and Peter C. Maloney, Johns Hopkins University School of Medicine. Helix Packing in the C-Terminal Half of Lactose Permease, H. Ronald Kaback, Kirsten Jung, Heinrich Jung, Jianhua Wu, Gilbert C. Prive, and Kevin Zen, University of Califomia Los Angeles. Export and Assembly of Outer Membrane Proteins in E. Coli, Jan Tommassen and Hans de Cock, Utrecht University. StructureFunction Relationships in the Membrane Channel Porin, Georg E. Schulz, Albert-Ludwigs-Universitat. Role of Phospholipids in Escherichia Coli Cell Function, William Dowhan, The University of Texas-Houston. Mechanism of Transmembrane Signaling in Osmoregulation, Alfaan A. Rampersaud, The Ohio State University. Index.

FACULTY/PROFESSIONAL discounts are available in the U.S. and Canada at a rate of 40% off the list price when prepaid by personal check or credit card and ordered directly from the publisher.

JAI PRESS INC. 55 Old Post Road # 2 - P.O. Box 1678 Greenwich, Connecticut 06836-1678 Tel: (203) 661- 7602 Fax: (203) 661-0792