CHAPTER .0

The Role of Electrostatic andNonpolar Interactions in theAssociation of Peripheral Proteins

with Membranes

Diana Murray,*'l,2 Anna Arbuzova,t Barry Honig,*

and Stuart McLaughlinf

*Department of Biochemistry and Molecular Biophysics, Columbia University, New York, New York

10032; tDepartment of Physiology and Biophysics, Health Sciences Center, SUNY, Stony Brook,

New York 11794

I. IntroductionII. Electrostatic and Hydrophobic Interactions Mediate the Membrane Association

of Important Biological ProteinsIII. The Membrane Interaction of Simple Basic Peptides: Theory and ExperimentIV. The Combination of Experimental and Theoretical Approaches Provides a

Detailed Comprehensive Picture of the Membrane Association of Myristoylated

ProteinsA. Src: Myristate plus Basic ClusterB. MARCKS

V. Future DirectionsA. Membrane TargetingB. Computational StudiesReferences

I Address correspondence to Diana Murray (Dim2007@med.comell.edu) or Stuart McLaughlin

(smcl@epo.som.sunysb.edu ).2Current address: Diana Murray, Dept. of Microbiology and Immunology, Weill Medical College

of Comell University, 1300 York Avenue, Box 62, New York, NY 10021.

277Current Topics in Membranes, Volume 52Copyright 2002, Eisevier Science (USA). All rights reserved.1063-5823/02 $35.00

278 Murray et al.

The membrane association of acylated and prenylated peripheral proteins,such as Src, the myristoylated alanine-rich C kinase substrate (MARCKS), andK-ras4B, plays an important role in cellular signal transduction. This chapter re-views experimental and computational studies of the membrane partitioning ofpeptides corresponding to the membrane-interacting regions of these proteins. Thecomputational model partitions the membrane interaction free energies into threecomponents: electrostatic attraction between basic groups on the protein and acidicphospholipids in the membrane, desolvation of the protein and membrane as theyassociate, and nonpolar burial of aromatic side chains into the membrane inter-face. The electrostatic components of the binding free energy are calculated bysolving the Poisson-Boltzmann equation for protein/membrane systems repre-sented in atomic detail (finite-difference Poisson-Boltzmann [FDPB] method).The nonpolar component is calculated as the product of an interfacial hydropho-bicity coefficient and the change in solvent-accessible surface area of aromatic sidechains as they penetrate the interface. The model predicts how membrane associ-ation changes as a function of the electrostatic properties of the system and howdifferent combinations of electrostatic and nonpolar forces dictate a wide range ofmembrane-binding properties. The success of the FDPB methodology in describ-ing experimentally characterized biophysical systems establishes the applicabilityof classical electrostatics and the continuum approach to protein/ membrane sys-tems and justifies its extension to predicting the structural origins of the interfacialassociation of proteins of known structure. The biological implications of recentexperimental measurements of the partitioning ofMARCKS onto membranes con-taining phosphatidylinositol 4,5-bisphosphate are discussed.

INTRODUCTION

The binding of peripheral proteins to membranes is crucial to many biologicalprocesses, such as signal transduction, vesicle trafficking, and retroviral assembly,and is often accomplished through protein motifs that interact with membranelipids. Many peripheral proteins use domains of well-defined structure to targetspecific lipids (Hurley and Misra, 2000). For example, the pleckstrin homology(PH) domain of phospholipase C-8 {PLC-8) anchors the protein to phosphatidyli-nositoI4,5-bisphosphate (PIP2) in the plasma membrane, the CI domain of pro-tein kinase C (PKC) binds to diacylglycerol in nuclear and plasma membranes,and the FYVE domain of the early endosome associated protein (EEA I) bindsspecifically to phosphatidylinositol 3-phosphate, a lipid often found localized toendosomal membranes. Other proteins use unstructured motifs that bind mem-branes through nonspecific electrostatic (Murray et at., 1997) and nonpolar inter-actions (Bhatnagar and Gordon, 1997; Johnson and Comell, 1999; Resh, 1999). Forexample, a substrate of PKC, MARCKS, binds to the plasma membrane throughhydrophobic insertion of an N-terminal myristate into the membrane hydrocarbon

27910. Electrostatic Association of Peripheral Proteins

and electrostatic interaction of a cluster of basic amino acids with acidic phospho-lipid head groups; the nonreceptor tyrosine kinase Src binds in a similar manner(McLaughlin and Aderem, .1995). In this chapter, we focus on the nonspecificelectrostatic and nonpolar interactions responsible for the membrane localizationof proteins like MARCKS and Src. We stress the interplay of experiment andcomputation, which provides a powerful approach to describing the underlyingmolecular and physical basis of these interactions.

II. ELECTROSTATIC AND HYDROPHOBIC INTERACTIONS

MEDIATE THE MEMBRANE ASSOCIATION OF IMPORTANT

BIOLOGICAL PROTEINS

In viva studies have shown that a number of proteins, for example, K-ras4B(Hancock et al., 1990), SIc (Sigal et al., 1994), MARCKS (Seykora et al., 1996;Swierczynski and Blackshear, 1996), and human immunodeficiency virus type IGag polyprotein (HIV-I Gag) (Rein et al., 1986; Zhou and Resh, 1996), requireboth a lipophilic attachment (either an acyl chain or a prenyl group) and a clus-ter of basic residues to bind to the plasma membrane and that membrane asso-ciation is crucial for their biological functions. For example, the small GTPaseK-ras4B has a CAAX motif at its C-terminus, which is sequentially farnesylated,AAX-proteolyzed, and, finally, methylated (Magee and Marshall, 1999). The far-nesyl chain (a 15-carbon isoprenoid) partitions hydrophobically into the mem-brane hydrocarbon, but does not provide sufficient energy to anchor the proteinto membranes (Ghomashchi et al., 1995; Silvius and I'Heureux, 1994). K-ras4Bhas a second membrane-binding motif, a C-terminal cluster of basic residues (seeTable I). Hancock et al. (1990) showed that the plasma membrane targeting ofK-ras4B requires both the CAAX motif and the adjacent basic cluster. Biophys-ical studies with peptides corresponding to the C-terminus of K-ras4B show that

TABLE I

Clusters of Basic Residues That Interact with Acidic Phospholipids'

Src myristate-GSSKSKPKDPSQRRR

MARCKS myristate- ...KKKKKRFSFKKSFKLSGFSFKKNKK

HIV-l Gag myristate-GARASVLSGGELDRWEKIRLRPGGKKKYKL

K-ras4B GKKKKKKSKTSC-famesyl

Rapl-A PVEKKKPKKKSC-geranylgeranyl

Rac-l CPPPVKKRKRKC-geranylgeranyl

Rho-A LQARRGKKKSGC-geranylgeranyl

BASP-l myristate-GGKLSKKKKGY

aBa.,ic residues are boldface. acid residues are italic.

280 Murray et at.

partitioning to membranes containing 20 mol% acidic phospholipid is 300-foldstronger than the partitioning to pure zwitterionic phospholipid membranes, sug-gesting the basic sequence may contribute favorable electrostatic interactions to themembrane association of the intact protein ( Ghomashchi et al. , 1995; Leventis andSilvius, 1998). Other small GTPases, for example, Rapl-A, Rac-l, and Rho-A, alsocontain an isoprenyl group and an adjacent cluster of basic residues (see Table I).

The role of hydrophobic and electrostatic interactions in the membrane asso-ciation of myristylated proteins (Src, MARCKS, HIV-l Gag; see Table I) is alsowell established. Like the famesyl group of K-ras4B, myristate (a 14-carbon fattyacid) partitions hydrophobically into the membrane hydrocarbon ( Bhatnagar andGordon, 1997; Ghomashchi et al., 1995; Peitzsch and McLaughlin, 1993; Resh,1999; Silvius and l'Heureux, 1994). Src and MARCKS are discussed in moredetail in the sections below. HIV-1 Gag plays a central role in the assembly ofnew virions in HIV-infected cells (Gamier et al., 1998; Wills and Craven, 1991).Studies from many labs have shown that plasma membrane association of Gag iscrucial for virion formation and that myristoylation is required for this association(Bryant and Ratner, 1990; Rein et al., 1986; Zhou et al., 1994). As for fame-syl, myristate alone is not sufficient to anchor proteins at the plasma membrane,and like K-ras4B, HIV-1 Gag has a cluster of basic residues adjacent to the siteof lipid modification (Table I) that contributes to membrane association throughelectrostatic interactions with acidic lipids (Zhou et al., 1994). The structure of theN-terminal matrix domain of mv-1 Gag revealed that these basic residues forma flat, positively charged basic surface adjacent to the site of myristoylation (Hillet al., 1996).

What is the purpose of nonspecific association to membrane surfaces? Thetranslocation of a protein from the cytoplasm to a membrane facilitates its inter-actions with other membrane-bound or embedded molecules by a "reduction-of-dimensionality" mechanism. Kinetic analyses suggest that the most important fea-ture of this phenomenon is a simple enhancement of the equilibrium concentrationsof the molecules at the membrane surface (Kholodenko et al., 2000; McCloskeyand Poo, 1986). Membrane association effectively increases the concentration ofa peripheral protein in a thin (d ~ 1 nm) surface layer adjacent to the membrane.Assuming the cell is a sphere of radius R equal to a few micrometers, one findsthat the volume of the surface phase (V = 47CR2d) is about 1/1000 the volume ofthe cell (V = 47C R3 /3). Thus, anchoring a protein to a cell's plasma membrane

increases its concentration 1000-fold, greatly enhancing its ability (or the abilityof proteins it may recruit to the plasma membrane) to interact with effectors andsubstrates. For example, Ras-GTP mediates the plasma membrane associationof the serine/threonine kinase Raf1; this recruitment facilitates a number of post-translational modifications to Raf1 which activate the kinase (Morrison andCutler, 1997). Similarly, plasma membrane association of mv-1 Gag polypro-teins facilitates the homo-oligomerization required for virion assembly and bud-ding (Gamier et al., 1998), and translocation of PKC from the cytoplasm to the

28110. Electrostatic Association of Peripheral Proteins

membrane (Oancea and Meyer, 1998) facilitates its ability to phosphorylate itsmembrane-bound substrates, such as MARCKS.

A number of recent reviews describe the different "switch" mechanisms bywhich the membrane association of lipid-modified proteins may be regulated(Johnson and Cornell, 1999; McLaughlin and Aderem, 1995; Resh, 1999). Theavailability of myristate for membrane partitioning is regulated in a number ofproteins. The myristate moieties of the inactive forms of both the small GTPaseArf and the visual transduction protein recoverin are sequestered in hydrophobiccavities in the proteins, keeping the proteins cytosolic. Activation of Arf and re-coverin by GTP and calcium, respectively, causes protein conformational changesthat result in the extrusion of myristate and concomitant membrane targeting.Recent work indicates that the membrane association of the BIV-l matrix do-main is regulated by proteolysis: In the context of a newly formed virion, the Gagpolyprotein is cleaved by the viral protease, exposing an as-yet-unknown "signal"on the matrix domain that sequesters the myristate chain (Spearman et at., 1997;Zhou and Resh, 1996). Since the hydrophobic energy due to myristate partitioningalone is not enough to keep a protein adsorbed to the plasma membrane (Peitzschand McLaughlin, 1993), mechanisms that interfere with other membrane-bindingmotifs may result in significant translocation of the protein to the cytosol. As dis-cussed in more detail below, the membrane association of MARCKS is regulatedby a myristoyl electrostatic switch (McLaughlin and Aderem, 1995). Phosphor-ylation of serines within the basic cluster by PKC weakens the electrostaticattraction of the basic effector region to acidic lipids and results in the desorp-tion of MARCJ(S from plasma membrane to cytosol in many cell types. The useof multiple membrane-interacting motifs by peripheral proteins allaws for a widevariety of signal-specific membrane recruitment and translocation mechanisms.

Proteins in the same family may use different motifs to bind to membranes(Bhatnagar and Gordon, 1997; Resh, 1999). Src appears to be the only memberof the Src family of nonreceptor tyrosine kinases that uses an acyl chain plus ba-sic cluster. Other family members are also N-myristylated, but have an adjacentpalmitate (16-carbon fatty acid) moiety instead of a basic cluster. The Ras isoformsB-Ras and N-Ras lack the C-terminal basic cluster of K-ras4B, but are palmity-lated on C-terminal cysteines. The different types of lipid modifications seem to becorrelated with localization. For example, the dually acylated Src family membersare localized in detergent-resistant, cholesterol-enriched fractions of membranesknown as rafts, as are other molecules with multiple saturated acyl chains, such asGAP43 (D. A. Brown and London, 2000; Simons and Ikonen, 1997).

Other peripheral proteins, such as the type II .B-phosphatidylinositol 3-kinase,AKAP79, mylein basic protein, and a number of proteins containing C2 domains,use basic groups to bind to membrane surfaces, but have no lipophilic modifica-tions. Although the focus here is on proteins that use both a basic cluster and alipid group to bind to membranes, the principles that are described should be ofgeneral applicability to a wide range of peripheral proteins.

282 Murray et at.

III. THE MEMBRANE INTERACTION OF SIMPLE BASIC PEPTIDES:

THEORY AND EXPERIMENT

Traditionally, the electrostatic properties of membranes and peptide/membranesystems have been described by "smeared-charge" models based on Gouy-Chapman (GC) theory (McLaughlin, 1989). This theory assumes the charges dueto the acidic lipids are smeared uniformly over a planar membrane surface and canbe described by a uniform surface charge density. Studies from a number of labshave established that the electrostatic properties of phospholipid membranes canbe described adequately by smeared-charge theory, at least for membranes con-taining physiological concentrations ( > 10% ) of monovalent acidic lipids (Peitzschet at., 1995). As illustrated in Fig. I A, the electrostatic equipotential profiles abovea bilayer containing 33 mol% acidic phospholipid in 100 mM KCl are essentiallyflat and uniform. When the membrane contains a low concentration of acidic lipid,it can be shown that the essentially hemispherical equipotential profiles aroundeach lipid are given approximately by twice the Debye-Hiickel expression; thefactor of two arises because of the "image charge" effect (McLaughlin, 1989).Figure IR illustrates the -25 mY electrostatic equipotential profile ofa 33 mol%acidic lipid membrane that contains a single PIP2 molecule; the multivalent acidiclipid significantly enhances the negative potential of the membrane in its vicinity.

GC theory has also been applied to describe the membrane binding of multi-valent peptides (Ren-Tal et at., 1996; Heimburg and Marsh, 1996; Mosior andMcLaughlin, 1991, 1992). The charges on the bound peptide are treated as pointcharges that do not perturb the electrostatic potential of the membrane. Theirpositive charges, like the charges on acidic lipids, are smeared uniformly on theplanar membrane surface and contribute to a net surface charge density. In orderto reproduce quantitatively the experimentally determined membrane partition-ing, GC theory typically has to invoke an effective charge for the peptide thatis significantly less than its net charge (Heimburg and Marsh, 1996; Mosior andMcLaughlin, 1991, 1992). In addition, smeared-charge models cannot account forcomplex patterns of electrostatic potential, which depend upon a protein 's shape aswell as the specific location of its charged and polar groups (Honig and Nicholls,1995). Recent computational studies show that in order to describe the observedelectrostatic properties of membranes with adsorbed basic peptides, it is crucial toaccount for the molecular detail of the peptides (Murray et at., 1999). As shown inFig. 1 C, an adsorbed basic peptide perturbs the electrostatic potential in a highlylocalized manner and produces a strong positive potential in its vicinity.

Work from the Honig lab has applied the Poisson-Roltzmann (PR) equationto describe the electrostatic properties of proteins and nucleic acids (Honig andNicholls, 1995). The PR equation is solved numerically in the finite-differenceapproximation (FDPR) for static, atomic-level models of the biologicalmacromolecules. The surrounding aqueous phase is represented implicitly as a

28310. Electrostatic Association of Peripheral Proteins

FIGURE 1 The negative electrostatic potential of a 2:1 PC/PS membrane is significantly per-turbed by multivalent molecules. The red (blue) lines represent the two-dimensional -25 ( +25) m V

equipotential contours obtained from FDPB calculations. In each panel, the membrane compositionis 2:1 PC/PS and the ionic strength, [KCI], is lOO rnM. (A) The -25 mV electrostatic equipotentialcontour is essentially flat and located ~ 10 A above the membrane surface. (B) A trivalent PIPzlipid

(head group colored yellow) greatly increases the negative potential of the membrane in its vicinity.

(C) A membrane-adsorbed basic peptide (heptalysine; blue) introduces a significant positive potential

at the membrane surface. (See color plate.)

homogeneous medium of dielectric constant 80. Monovalent salt ions are treatedin the mean-field approximation, so that ion-ion interactions and the desolvationof ions near molecular surfaces are ignored. In the FDPB implementation, an ionexclusion layer of 2 A thickness is applied to the molecular surfaces of allmolecules (see Fig. 3 of Ben Tal et at., 1996). As recently reviewed (Vlachy,1999), PB theory has been tested extensively against computer simulations thattreat ions explicitly, and for many situations, its description is adequate. Indeed, ithas been highly successful in predicting salt effects observed for macromolecular

systems (Honig and Nicholls, 1995) (see below).

284 Murray et al.

Recently the FDPB methodology has been successfully applied to describe theelectrostatic properties of membranes and protein/membrane systems (Ben- Talet al., 1996; Murray et al., 1997; Peitzsch et al., 1995). The assumption that wa-ter molecules adjacent to a membrane surface can be treated theoretically as adielectric continuum is supported by surface force, X-ray diffraction, and otherexperiments, as reviewed elsewhere (McLaughlin, 1989). The FDPB approachsacrifices accuracy in the representation of molecular details and motions, but re-tains the ability to accurately describe long-range electrostatic interactions, whichare important for protein/membrane interactions. Molecular dynamics (MD) sim-ulations, on the other hand, sacrifice accuracy in the physical model (e.g., throughtruncation or approximation oflong-range electrostatic interactions) to achieve ac-curacy in molecular representation. Hence, MD simulations have been extremelyvaluable in depicting the short-time (less than microseconds) detailed motions oflipids in a membrane environment (Merz and Roux, 1996), but have been lim-ited in their capacity to describe the electrostatic properties of membranes andprotein/membrane systems (Jakobsson, 1997). As illustrated by the examples be-low, the FDPB approach provides a computational, structure-based framework for(1) testing detailed physical models, (2) predicting the electrostatic properties ofprotein/membrane systems, and (3) dissecting the various energetic componentsof protein/membrane interactions and assessing quantitatively their importance.

The FDPB methodology is described in detail elsewhere (Ben- Tal et al., 1996;Gilson and Honig, 1988; Misra et al., 1994). Here an overview is given that focuseson issues relevant to membrane and protein/ membrane systems. The application ofthe methodology is illustrated by describing the interaction of a heptalysine peptide(Z = + 7) with a membrane composed of zwitterionic (phosphatidylcholine, PC)and acidic (phosphatidylserine, PS) lipids (Fig. lC). This type of system is agood model for how the N-terminal basic cluster on Src and the C-terminal basiccluster on K-ras4B interact with membranes (Murray et al., 1997). The membranemodels are built as described in Peitzsch et al. (1995): Atomic-level models ofPC and PS lipids are arranged on a hexagonal lattice in which each lipid occupiesapproximately 68 A 2 in the plane of the membrane (Mclntosh and Simon, 1986).

The peptide in Fig. I C is docked in the aqueous solution above the membranein its minimum electrostatic free energy orientation. The electrostatic interactionfree energy can be calculated as a function of the orientation of the peptide at themembrane surface, the mole percent acidic lipid in the membrane, and the ionicstrength of the solution.

For the numerical solution of the PB equation, each atom of the peptide andmembrane is assigned a radius and a partial charge, which is located at its nucleus.The peptide/membrane model is then mapped onto a three-dimensional latticeof points, each of which represents a small region of the peptide, membrane, orsolvent. Smooth molecular surfaces for the peptide and membrane are generatedby rolling a spherical probe with the radius of a water molecule (1.4 A) over the

28510. Electrostatic Association of Peripheral Proteins

surfaces defined by the van der Waals radii of the constituent atoms. Lattice pointsthat lie within the molecular surfaces of the peptides and bilayer are assigned adielectric constant of 2, and lattice points outside the molecular surfaces, corre-sponding to the aqueous phase, are assigned a dielectric constant of 80. Salt ionsare excluded from a region that extends 2 A (the radius of a Na+ ion) beyond thevan der Waals surfaces of the peptide and membrane, as illustrated in Fig. 3 ofBen Tal et al. (1996). The electrostatic potential and the mean distribution of themonovalent salt ions at each lattice point are calculated by solving the nonlinearPoisson-Boltzmann equation,

V[E(r)V4>(r)] -ErK(r)2 sinh[4>(r)] +

where E(r) is the dielectric constant, 4>(r) is the electrostatic potential, and pkr)is the charge density of the fixed charges. The PB equation is mapped onto thecubic lattice and solved for 4>(r) using the finite-difference approximation and thequasi-Newton method (HoIst, 1993) combined with three levels of multigriddings(HoIst and Saied, 1993).

A sequence of focusing runs of increasing resolution is employed to accu-rately calculate the electrostatic potentials. In a typical initial calculation, thepeptide/membrane model fills a small percentage of the lattice (""'10%) and thepotentials at the boundary points of the lattice are approximately zero. This pro-cedure ensures that the system is electroneutral. Subsequent calculations employsuccessively finer resolutions and extract the potentials at boundary points from apreceding, coarser grain calculation. The solutions 4>(r) to the Poisson-Boltzmannequation are used to calculate the electrostatic free energy of the peptide/membranesystem Gel(P .M), using an equation given by Sharp and Honig (1990). The elec-trostatic free energy of interaction ~Gel is calculated as the difference betweenthe electrostatic free energy of the peptide and membrane when they are closetogether, Gel(P .M), and that when they are far apart, Gel(P) and Gel(M):

dGel = Gel(P .M) -{Gel(P) + Gel(M)}

The curve given by the triangles in Fig. 2 illustrates the calculated electrostaticfree energy of interaction as a function of the distance R between the van der Waalssurfaces of the peptide and the membrane for an orientation in which the peptideis parallel to the membrane surface as in Fig. 1 C. When the peptide is far from themembrane surface, it experiences an electrostatic attraction that drives it towardthe negatively charged membrane. This "Coulombic" attraction increases as thepeptide approaches the membrane. Close to the membrane surface, charged andpolar groups on both the peptide and membrane are desolvated, or stripped of watermolecules, which is energetically unfavorable and results in a repulsion at smalldistances. The minimum electrostatic free energy of interaction for heptalysine is

286 Murray et at.

Lys7 on 2:1 PC:PS, 100 mM KCI

12

8

4

0

-4

GI -8GI~ -12

-16 ..., ., 0 2 4 6 8 10 12 14

Distance, Angstroms

FIGURE 2 FDPB calculations of the peripheral membrane association of heptalysine. The elec-trostatic interaction free energies are plotted as a function of distance between the van der Waalssurfaces of a heptalysine peptide and a 2:1 PC/PS bilayer in 100 mM KCI. The circles illustrate thelong-range electrostatic attraction between basic groups on the peptide and the acidic phospholipidhead groups. The squares illustrate the unfavorable desolvation energy. The balance between the long-range attraction and short-range desolvation repulsion results in an electrostatic free energy minimum(triangles) when the peptide and membrane are separated by a layer of water. See Murray etal. (1999)

for details of the calculations.

predicted to occur where the van der Waals surfaces of the peptide and membraneare separated by a distance of about the thickness of a layer of water. At thisdistance the Coulombic attractive force is balanced by the repulsive desolvationforce.

The components of the electrostatic free energy of interaction, namely theCoulombic attraction and the desolvation repulsion, can be calculated explicitly.The desolvation penalty tlGdesolv is determined at each R by calculating to whatextent the presence of the peptide and the membrane shield the others fromfavorable interactions with the solvent (Misraand Honig, 1995; Misraetal., 1994).This is done by discharging the peptide and membrane in turn and calculating theelectrostatic free energy of the system:

~Gdesolv = ~Gel(P(Q = 0) .M) + ~Gel(P .M(Q = 0)).

The Coulombic component ~GCoul is simply the difference between the totalelectrostatic free energy of interaction and the desolvation penalty:

dGCoul = dGel -dGdesolvo

Both the Coulombic and desolvation components are plotted in Fig. 2. The cir-cles illustrate the long-range electrostatic attraction between basic groups on thepeptide and the acidic lipid head groups; this attraction increases as the peptide

287lO. Electrostatic Association of Peripheral Proteins

approaches the membrane surface. The squares illustrate the shorter range desol-vation repulsion; as the peptide approaches the membrane surface, both moleculesare stripped of water molecules, resulting in the transfer of charged and polargroups from a region of high dielectric constant in the aqueous phase to a regionof low dielectric constant. As the peptide comes very close to the membrane, therepulsive desolvation force becomes greater than the attractive Coulombic force,resulting in an electrostatic minimum free energy orientation in which the peptideadsorbs outside the envelope of the polar head group region of the membrane. Thisprediction is consistent with nuclear magnetic resonance (NMR; Roux et at., 1988),high-pressure fluorescence (Montich et at., 1993), and monolayer surface pressuremeasurements (Ben- Tal et at., 1996), which indicate that simple basic peptides donot penetrate the membrane interface. As shown below, favorable nonpolar in-teractions, mediated by hydrophobic groups, may contribute significantly to themembrane association and tip the force balance toward interfacial penetration.

In the FDPB calculations of the electrostatic component of peptide/membraneinteractions to date, it has been assumed that the lipids are frozen, both verticallyand laterally. Specifically, the calculations assume the lipids do not demix laterallyat the membrane surface as the peptide adsorbs, that the orientation of the lipidhead groups is constant, and that there are no membrane undulations or lipid bob-bing motions. Although these approximations are extreme (see, e.g., Mclntosh andSimon, 1994), the FDPB calculations apparently capture the essence of the equi-librium binding of peripheral peptides. Experiments showing that pentalysinepartitions similarly onto dimyristoylphosphatidylcholine/dimyristoylphospha-tidylglycerol (DMPC/DMPG) membranes in the liquid crystalline and gel phasessuggest that, at least for this type of peripheral membrane association, lipid motionsand demixing can be ignored (Kim et at., 1991). In addition, electron paramagneticresonance (EPR) experiments showed that spin-Iabeled acidic lipids do not accu-mulate around pentalysine when it adsorbs to the membrane surface (Kleinschmidtand Marsh, 1997). A recent theoretical paper addresses the issue of lipid demixingproduced by the adsorption of an idealized basic protein and predicts that underphysiological conditions the demixing effect should not be large for monovalentacidic lipids (May et at., 2000). Below, experimental results are discussed that sug-gest demixing effects can be very large for the multivalent acidic lipid PIP2. Asidefrom the greater computational expense, there is no reason why lateral diffusionof lipids and lipid demixing cannot be incorporated into FDPB calculations.

The concentration of peptide at each distance from the membrane surface isproportional to the exponent of its interaction free energy. As suggested by theelectrostatic free energy curve (Fig. 2, triangles), the bound peptides, which as-sociate with the membrane through long-range electrostatic interactions, can belocated at an appreciable distance from the membrane surface. For example, thepeptide concentration at R = 6 A is predicted to be only 10-fold lower than thatat R = 3 A and 5000-fold greater than the concentration in the bulk. This is the

288 Murray et at.

hallmark of a "nonspecific" electrostatic interaction and is more accurately de-scribed by a partition coefficient, rather than a binding constant, which assumesthe formation of a 1 : 1 complex between a peptide and a lipid (White et at., 1998).The "excess" peptide concentration at a distance R from the membrane is definedas the difference between the peptide concentration at R and the bulk peptide con-centration. Integrating the excess peptide concentration over R gives the Gibbssurface excess (see Fig. 2 in Murray et at., 1997), that is, moles of peptide boundper unit area of membrane surface. The Gibbs surface excess is simply related tothe molar partition coefficient that is measured experimentally.

In practice, the Gibbs surface excess is calculated by considering many ran-domly chosen orientations of the peptide with respect to the membrane at eachR (Ren-Tal et at., 1996; Murray et at., 1998). Although the peptide model isrigid throughout the calculations, the incorporation of many different orientationsapproximates a complete sampling over different peptide structures. For simplebasic peptides like heptalysine, the molar partition coefficient calculated in thisway consistently gives a binding energy that underestimates the experimentallydetermined value by 1.5 kcal/mol. This suggests that the model ignores and/ormisrepresents some interactions. Nevertheless, an important result obtained fromthe comparison of computational predictions and experimental results for simplemodel systems is that the theoretical model correctly predicts how the membranebinding depends on factors that affect the electrostatic properties of the system.[Indeed, for predicting relative binding effects, it has been found that considerationof a single orientation of the peptide with respect to the membrane is sufficient(Ren- Tal et at., 1996,1997; Murray etat., 1998)]. For example, the model correctlypredicts that the binding of charybdotoxin (Z = +4) decreases by five orders ofmagnitude when the monovalent ion concentration in solution is increased from10 to 150 mM (Ren-Tal et at., 1997). The model also adequately describes theobservation that the binding free energy of LyS3, LyS5, and LyS7 onto 2 : 1 PC/PSin 100 mM KCl is -3, -5 and -7 kcal/mol, respectively, that is, the bindingfree energy under physiological conditions is about -1 kcal/mol per basic residue(Ren- Tal et at., 1996). Computational studies on experimentally well characterizedsystems such as these show that the FDPR methodology can account for how themembrane binding increases as the ionic strength of the solution decreases, as themole percent acidic lipid increases, and as the net charge of the peptide increases.

Experimental studies show that the binding of model peptides to membranesdepends only weakly on the chemical nature of either the basic residues (Arg vs.Lys) or the monovalent acidic lipid (PS vs. PG, phosphatidylglycerol) (Kim et at.,1991; Mosior and McLaughlin, 1992). Taken together, these observations supporta model in which long-range, nonspecific electrostatic interactions drive the mem-brane association and that simple basic sequences of the form (LyS)N and (Arg)N(N ~ 2) can contribute significant membrane binding energy under physiologicalconditions (Ren-Tal et at., 1996; Kim et at., 1991; Mosior and McLaughlin, 1991,1992). For example, the molar partition coefficient of pentalysine onto 2 : 1 PC/PS

28910. Electrostatic Association of Peripheral Proteins

membranes in 100 mM KCl is 103 M-: 1. Although not strong enough by itselfto anchor a protein at the plasma membrane surface, a basic cluster may act inconjunction with other mernbrane-binding motifs to effect stable plasma mem-brane association, for example, a myristate (as for Src) or farnesyl group (as for

K-ras4B).The agreement of computational predictions based on solutions to the Poisson-

Boltzmann equation with results from experimental studies on simple model sys-tems supports the idea that nonspecific electrostatic interactions drive the asso-ciation of basic clusters with membranes containing acidic lipids. The FDPBmethodology provides a framework for dissecting protein/membrane interactionsand for examining in detail the balance between the attractive and repulsive forcesinvolved, and it suggests ways to examine the effect of additional forces, for exam-pIe, the hydrophobic partitioning of lipophilic groups into the membrane interioror of aromatic groups into the polar head group region. The next section describesapplications to more biological systems as well as the incorporation of nonpolar

interactions.

IV. THE COMBINATION OF EXPERIMENTAL AND THEORETICALAPPROACHES PROVIDES A DETAILED COMPREHENSIVEPICTURE OF THE MEMBRANE ASSOCIATION OFMYRISTOYLATED PROTEINS

A. Src: Myristate plus Basic Cluster

Src belongs to a family of nonreceptor protein tyrosine kinases that are foundassociated principally with cellular membranes (M. T. Brown and Cooper, 1996;Cooper, 1990; Parsons and Parsons, 1997). Membrane binding is required for itsfunction, and myristate plays a key role in its membrane localization (Resh, 1999).Nonmyristoylated v-Src mutants are found in the cytoplasm and do not transformcells, even though the kinase activity is the same as that of wild-type protein(Buss et at., 1986; Kamps et at., 1985). Although myristate is required for Srcmembrane binding, it alone is not sufficient (Buser et at., 1994; Sigal et at., 1994).Src contains a second membrane-binding motif, an N-terminal cluster of basicresidues (Kaplan et at., 1990). Mutating away these basic residues also producesnontransforming phenotypes (Sigal et at., 1994). This suggests that Src uses acombination of motifs-a myristate plus a cluster of basic residues-to effect

membrane localization.Biophysical studies of the isolated protein show that Src binds to zwitterionic

(PC) membranes with a molar partition coefficient of 103 ~ 1 (Sigal etat., 1994).

Simple myristoylated peptides bind to PC membranes with a molar partitioncoefficient of 104 ~ 1 (Peitzsch and McLaughlin, 1993). Measurements of the

membrane partitioning of acylated peptides show that each CH2 group contributes

290 Murray et al.

about 0.8 kcal/mol to the membrane binding energy, in agreement with Tanford'sobservations on the hydrophobic partitioning of fatty acids into oil from water.Silvius (this volume, Chapter 13) considers the membrane partitioning of acyl-ated and prenylated peptides in more detail. Myristoylated peptides correspond-ing to the N-terminus of Src bind to PC membranes with a partition coefficientsimilar to that of other myristoylated peptides (K = 104 ~ I). The partitioningof the intact Src protein onto PC membranes is about 10-fold lower for reasonsthat are not well understood. Incorporating acidic lipids into the membrane in-creases the binding of both the myristoylated peptide and the intact Src proteinby exactly the same factor. Specifically, the partitioning of both the isolated Srcprotein and myristoylated peptides corresponding to its N-terminus is 1000-foldhigher when the membrane contains 33 mol% acidic lipid (Buser et al., 1994).There is good evidence that the N-terminal cluster of basic residues is respon-sible for the electrostatic interaction between the intact protein and membranes,because mutating away the basic residues decreases the partitioning of the pro-tein onto PC/PS vesicles (Sigal et al., 1994). These observations provide strongevidence that the N-terminal basic residues do indeed contribute significantly tothe membrane association of Src by interacting electrostatically with acidic lipids.Computational predictions obtained by applying the FDPB methodology to modelsof the Src N-terminal peptide (Fig. 3) agree with these experimental observationsand provide further support for the role of nonspecific electrostatic interactionsin the membrane association of Src. Of course, these interactions do not rule outprotein-protein interactions (e.g., utilizing the SH2 and SH3 domains of Src) thatcould help direct Src to different membranes or affect the lateral organization ofSrc within a given membrane.

For peptides corresponding to the N-terminus of Src, the individual hydropho-bic and electrostatic contributions to the binding energy are additive (Buser et al.,1994). If KA is the "hydrophobic" molar partition coefficient for the myristolyatedpeptide onto PC vesicles and KB is the "electrostatic" molar partition coefficientfor the nonmyristoylated peptide into PC/PS vesicles, the overall binding of themyristoylated peptide to PC/PS vesicles is K = aKAKB, where a is dependenton the "distance" between the two binding sites, as can be seen from modelsthat consider the myristate and basic cluster as points connected by a flexiblestring of length L. Partitioning of the myristate into the membrane confines thebasic cluster to a hemisphere of radius L above the membrane surface and fa-cilitates its adsorption to the membrane; the shorter the string, the stronger the"synergism" between the two membrane-binding motifs. For the short "string"of Src, a ~ I. This "ball-and-string" model applies to the membrane partition-ing of peptides corresponding to the C-terminus of K-ras4B as well (Table I)(Ghomashchi et al., 1995). Although the ball-and-string model accounts for theadditivity of electrostatic and hydrophobic interactions in the membrane partition-ing of Src, it is descriptive rather than predictive. Figure 3 shows the predicted

29110. Electrostatic Association of Peripheral Proteins

Myr-Src(2-19)

FIGURE 3 Molecular model of the N-tenninal region of Src interacting with the upper leafletof a 2 : 1 PC/PS bilayer. The front portion of the bilayer has been removed to expose the myris-tate (green). In agreement with ESR and CD measurements, the 18-residue peptide (myristate-

GSSKSKPKDPSQRRRSLE-NH2) is shown in an extended conformation. The six basic residues areblue and the two acidic residues are red. In the bilayer, PS is identified by its exposed nitrogen (blue);oxygen is red, phosphorus is yellow, and carbon and hydrogen are gray. Reprinted from Murray et at.(1997), copyright 1997, with permission from Elsevier Science. (See color plate.)

minimum free energy orientation of the myristoylated Src(2-19) peptide abovea 2: 1 PC!PS membrane in 100 mM KCl obtained using the FDPB methodol-ogy (Murray et al., 1998). The conformation is consistent with monolayer, circu-lar dichroism, and electron paramagnetic resonance (EPR) measurements (Buseret al., 1994; Victor and Cafiso, 1998). The Gibbs surface excess and molar parti-tion coefficient were determined by calculating and averaging over the electrostaticfree energies of interaction for many different sampled orientations of the peptideanchored at its N-terminus to the membrane surface (Murray et al., 1998). The pre-diction underestimates the absolute binding energy by 1 kcal!mol, but, as shownin Fig. 4, the relative binding, as functions of the mole percent acidic lipid in themembrane (Fig. 4, left) and the ionic strength of the solution (Fig. 4, right), is

accurately predicted.The three-pronged approach to the problem of Src membrane association-

(I) in viva transformation studies of wild-type and mutant v-Src, (2) in vitrabiophysical studies of the intact proteins and peptide models of its membrane-interacting regions, and (3) computational studies with molecular models-isa powerful one and provides a detailed molecular and mechanistic picture ofhow Src binds to membranes. The approach also can be applied to obtain in-sight into the membrane-binding mechanisms of other peripheral proteins, forexample, MARCKS, K-ras4B, and HIV-1 Gag.

cnDo

--t)Do

.-Nt:0m.-I(\I

'0..cnu~~Eoo~.5$'I(\I

u..In

(j

0:-O0:

.-

..10

E:;

.;: c.

..EC.o

~U.0

"-0".-a.me-

.-"~

c.a.

e

~o

wu

.0I"?

M

~

Il) "'

I I

I I

810w/le:»t

'A6J8U

8 88J::1

"!" ~ f

";1

aIOw

/le:»t 'A

6Jaua aaJ.:1

0 .~

~

Z

;- ~

Il) ~

"":=°

,.. ..~

~

.=

"' -~

~

o~.,-

~.'85

~

15 .c

.c--.,0.,Q

.Q..'

~

£~~

-;,E

~

.,-

~~

.s~~

-.c 0

., ~

.(U

.c 'O

O~

~~

-"' '0

.C

-C

I .~

0

~

., '0

°c

U-"O

'O."'

O

, .c

.~

~,..Q

) cn.'

~"'~

..

e~

o-""'

0 .=

cn 0

"'U~

U

~~

'3>':o

.-I

O

-.c Q

.C

N

~:i3'3e

..Q

~.,~

~8

.s.s :0'O

.~

>, ""'

~

§

~.c~

u"a0

'3 .~

=

=

'.,0

0/) U

~

.~

e,~

~

.~

~

U

~.

.=

'3 0

1S

.c

§ 6. .§

8 &

Q.

-u 0<

""0~~

.,~

§

., .g

~...>

, .s

.;: ..,

8-""'(U"'

"' 3

0 Q

. .§.,~

.s.,~:g8~

.S

§Q

. .c .,

~

u

&.,~

"".dQ

. "'

"' u

'0 0<

u

., .~

oB.'

.2 :0 i

° ~

'Oo"a

,.. >

, .,

.~

>

~o~

.,.,OO

~

.2 .s

'iU

Ct\

.c c

'0 e

~>

, B

~

.~

~

"Ce

., ~

0<

.

.-.;0: 0

-Q

. ~

~~

...':!.-0

"'~~

--~

---"'

U

""'~~

~>

,.-0

.~

'0 .,

~"C

U

.~

.-0/)

tC

Q.

~

...U

~

~

"'

:010

]8~]:;::

-.~

u'O

e e

~

.c ~

.~

>

, 0

U

.'0~"'~

~

~

:~

'E

5 ~

Q.~

[@~

8.!

8~

s=o

_.'_0~

.,

0 Q

. ~

.=

!5e-a§~~

.'e~

£.,

!5 .,

~

.S

'.:!3 ~

.;::>e

;;I .§

0 ]

§10

~

£ §

.~

.~IS

e

.=

E

~B

~

B

e~

"

., ""'

.~

...'0 0

10. Electrostatic Association of Peripheral Proteins 293

B. MARCKS

MARCKS is a widely distributed major PKC substrate implicated in phago-cytosis, secretion, and membrane recycling (Allen and Aderem, 1996). Althoughseparated by 150 residues (Figure 5, left), both an N-terminal myristate and acluster of basic residues [the effector region, MARCKS(151-175); Table I] arerequired for the plasma membrane association of MARCKS.

Unlike those of K-ras4B and Src, MARCKS's basic cluster contains aromaticresidues that penetrate the membrane interface. EPR measurements of spin-Iabeledpeptides corresponding to the effector region of MARCKS showed that the peptidelies at the membrane interface in a nonhelical conformation with its basic residuesin the aqueous phase and its five Phe penetrating the polar head group region (Fig. 5,right) (Qin and Cafiso, 1996). This conformation agrees with independent circulardichroism (CD) and monolayer penetration experiments (Kim et al., 1994a). Asfor Src, there is good evidence that the membrane-binding properties measuredin vitro with intact MARCKS and peptides corresponding to its effector regionreflect the membrane association ofMARCKS in cells. Figure 5 (left) illustrates themechanisms by which MARCKS binds to the plasma membrane: The N-terminalmyristate partitions into the membrane hydrocarbon core, the basic residues in theeffector region interact electrostatically with acidic lipids, the aromatic residuesin the effector region partition into the membrane interface, and the rest of theprotein, which has regions that are highly acidic, should be repelled from theplasma membrane and remain in the cytoplasm.

The MARCKS basic effector region contains 13 basic residues and 5 aromaticPhe residues (Table I). Experimental studies show that peptides corresponding tothis region bind strongly to membranes containing acidic lipids (K '"" 106 M-Ifor 5: I PC/PS in 100 mM KCI) (Arbuzova et al., 2000; Kim et al., 1994b).Unlike heptalysine and Src(2-19), the MARCKS peptide penetrates the membranewhen it binds. Experimental studies show that the area of a monolayer kept atconstant pressure increases upon addition of MARCKS peptide to the subphase.In addition, the MARCKS peptide binds weakly, but detectably, to neutral PCvesicles, indicating that forces other than electrostatic ones are involved in themembrane association (Arbuzova et al., 2000).

Because of its high net charge (Z = + 13), the MARCKS peptide experiencesvery strong electrostatic interactions with the membrane surface. FDPB calcu-lations with MARCKS(151-175) provide insight into the physical factors thatdrive membrane association (Arbuzova et al., 2000). The circles and squares inFig. 6 (left) show, respectively, the Coulombic attraction to and desolvation re-pulsion from a 5: I PC/PS membrane in 100 mM KCI. Note that the magnitudesof these interactions are much larger than those for heptalysine (Fig. 2), eventhough the mole percent acidic lipid is actually lower in this case. Most pro-teins that use electrostatic interactions to bind peripherally to membranes also

~0cr:

~~

-LO1'00~

I~LO

Ccn~()(I:~~

§ >< ></

'>i-<

'><

'

>-<

E c~

")le~-;'T+:1=

+++

"

In('I.-

.5 ~

.5

3 §

" =

"'

u ..

~'0,,(U

""e

.~

=

't=

""~:2B

'2cn

"'S

~

u "

.$

~

.~

.." ...

U:2u-5e-

~

~

.~e~

~

"" '0

"" .

"""e-;c~

(U(U

"O\

Uu

'00\(U

.~

"

0 -

~~

-5e~.~

.0

'0 ~

o

-5"0§~~

"O§.s--;;U

" .;j

" .=

'0

0(U>

="

.;j C

) .;j

0 (\1

U

0 U

~

(U-"(\1,,

.."""";=:,

B-5~

<,-,

.5 ...;;

>.

,,~00".:.0~

-"' 0

" C

I) "'

.;

'.'=bI)

ii"~

" .~

'0

(\10

>.

"' .~

-

-"' .;

~

""

~'2=

Ue-~

o ~

.§

~

"" o

e-=,,'2U

§ 8

'2 .5

] ~

0 .~

.."

(U

~

~~

]]k8<

.E

~

e

~

5..>

. bI)

" (3

~

e .0;

e o

cn,

., "'

" "'

..~

~=

-"'~"

..c -,.,

" .>

cn~

""bI)"e

"~

."

" "'

U

" =

.;j

" ri1

U~

§:E~

~e

<..>

.,,"'

1~

"S.o"~

e~

"~8.e""

oe~",~

§"

~

U

~

~

.~0

"' .;j

>

-~.~

""

(U

0 ~

.~

.S

5 5

5 ~

E

u...t:~

,.," 0

bI) U

"

""

~]"'

0-5"

'0 ~

-.~

" .t:

B

,. is

~

~]5]~

oo

.o~"',,-O

\

e~

,,~O

\"'"""',,,-

8S:~

~~

.d"-,,"lbI)

-"' ~

c

,.,"'(U,,"'>

.~

§'8

~;::,g.

"" p

" cn

u~

6b ~

.a

~

.

.g]~§~

~~

B§~

~;::'

.~

.~

.9 -:

-"' .t:

bI) u

~

~~

~.B

°- ~

~""".'

&:

="O

(U

0 U

.~

>

R

.t: .0;

~

0

""' (U

bI)

(U

.~

N

S-55.0~

.E

.~

.s ,s

~

<

29510. Electrostatic Association of Peripheral Proteins

contain aromatic residues that partition into the membrane interface. Work fromSteve White's lab has shown that aromatic residues (Trp, Tyr, and Phe) partitionvery favorably into the interface of palmitoyloleoylphosphatidylcholine (POPC)bilayers (White and Wimley, 1999; Wimley and White, 1996). Their experimen-tally derived interfacial hydrophobicity scale was used to incorporate the aromaticinteractions into the FDPB calculations. The aromatic component ~Garom is given

by

LlGarom = YifLlA,

where yif is the interfacial aromatic surface tension coefficient (Yif =

0.13 kcal/mol/A1 derived from the partitioning data of Wimley and White (1996)and M is the change in solvent-accessible surface area of the aromatic side chainsas the peptide associates with the membrane interface. In addition to the long-rangeelectrostatic attraction and short-range desolvation penalty, there is a favorableshort-range aromatic contribution (diamonds, Fig. 6, left) when the peptide isclose enough to the membrane surface so that the Phe side chains can partitioninto the membrane interface. The favorable Coulombic and aromatic contribu-tions combine to outcompete the desolvation penalty and allow the peptide topartially penetrtlte the membrane interface. Indeed, the calculations predict thata peptide corresponding to the MARCKS effector region has a shorter averagedistance ( <R> = 1.5 A, less than a layer of water) from the membrane surfacethan either heptalysine or a MARCKS peptide in which all five Phe are mu-tated to Ala ( <R> = 2.5 A). Figure 6 (right) illustrates the total "electrostaticplus aromatic" interaction free energy of the MARCKS peptide with PC (tri-angles) and 5: 1 PC/PS (upside-down triangles) membranes. In agreement withexperiment (Arbuzova et al., 2000), the calculations predict that the peptide bindsweakly to PC membranes and that the partitioning of the peptide increases 104-foldwhen 17 mol% acidic lipid is incorporated into the membrane. The computationalmethodology provides a physical model for how different magnitudes of theseforces (Coulomb attraction, desolvation repulsion, aromatic attraction), which arereflected in the residue character and structure of proteins, combine to producedifferent membrane-binding behaviors.

The extrapolation from basic peptide to protein is not as straightforward forMARCKS as for Src. In the case of Src, the membrane binding energies of themyristate and basic cluster are additive for both the myristoylated peptide andthe intact protein (Buser et al., 1994; Resh, 1999; Sigal et al., 1994). As dis-cussed above, incorporating'-..' 30% monovalent acidic lipid into a PC vesicleincreases the binding of both the myristoylated Src peptide and the intact Src pro-tein about 103-fold. In the case of MARCKS, incorporating ,-..,20% monovalentacidic lipid into a PC vesicle increases the binding of the effector domain peptide104-fold, as predicted theoretically (Arbuzova et al., 2000), but increases the bind-ing of the intact MARCKS protein only 102-fold (Kim et al., 1994b). This can berationalized by considering the physical character of the entire MARCKS protein,

(3~~Eoo.-<na.-ua..-Li;~0I;).-I

.-II).-(7)'~U11:~~ (3~~Eoo.-cna-

ua.c0G'

1"-.-I.-Lt).-W~Ua:«~

m (/)(/)0.0.

(3(30.0.0-,:;.;;

~~

~

.,(/) .,

., =

.,"V

;.'~

-"'-"'0.'.,-.:=

-""

"" .-.:=

-

~

U.,-"'

b/).-8

~

~

5 .a

8 §

.:J ~

"'

E:

.:J .E

~

..~

~

., U

~

N

...'r)-c.c.e.-.'~

::=~

-"'d}O-,"

"' ""

"' ..

cn.S"".,"'.-S

E

.~§]-g~

:.--""~

0

~

., .0) °§"i3

~

§..."

"" --

-"' .~

.,

c. 0

.,cn

0Q.!5:a~

§':gC

) ~

., .'

OC

.~

C

"" c.

..='""

=

.,=

(/).'b"'. c.

~

oj~§

.8~(/)

Gf

.u u ~

"' ~

"E

. ~u

~~

e§'~

uC

D

C

~

~

~

=

.~

~

ca 0

"'Q!j.-S

<

"W

Ja.,.,~."":2::

~i5

.£.S.S

:i,§g.,~

"""".c~

.s!5

0 §

.-e .,

C..,.'

o=

.,

.-" ~

N..0

"" °

"' -=

.,

ct .c

.:J .d

., ~

=

.'~~

~.sB

.8...~

C.°0.,~

--,-.c.,.d-.=~

0 ~

.,

.s c. .5

b"' .c-~

0

~

~

N

Sb/).~

"'...'I

I !.

"0 .5 :9

~

~

.sI

~

"' ~

~

:9 .C

IJ "0~

.,>

"'=low

/le:»I '1.6Jaua

aaJ.:1 V

i' ~

°

.s ~

.~

~

't: "'

., :t;o U

u ~

.g

-g ~

.~

~

~

oj u=

<~

0.,.,0",ou>

-"">(/)"'

:2:: 5

b/) .:J

t;: ...,

-=

;>I)C

."'1JOU

0 ...=

&

..,

.=

~

...;.--C

.£., ""=

U

.0 ~

.,

., .:J

., I)

~

~§:0.sC

..d",,]

.0) >

e

C

&

U

C

o.,0°.,Coj

~.S

>"'.S

~>

8oj

.s §'

.c .,

~5c

~"0.,.S

.-.,

=

ccn

U

~

., c

E

5 .-

E

~

"' .s

.~

.0 <

.,

~

,. 0

., .c

., ~

E

§~

~

..~

., -§

.c "'

., oj

-:=

.-U

~

C

N

.c ::=

v. .,

~

, .c

C)-"'~

~~

O~

U(/)

I C

""

"'- :.-

b/)~A

( ;

"0:a::=~

.'.~

~t::

"'.,=

"'&

:-"'

,.c"" ""

Q)

=0...

C."O

..0 -~

C

C

0

~

.--=

0 0

U0

"'r

C

~

.9 C

' .:J

.:J .,

---"'u=

.c-...=

U

.,

~

.c ---0

b/)"W

U

C

.=

~

.E

§.:J

§.-~

.s

b"' ~

U

.E

c uoj...;u8t!,§

~",U

.:JU

...c~

~

"'

"' U

-'

oj ~

O""~

~.:J'3,,,o

to;Q]

Doj..oj""

;; :;E

""" ,

.9 U

oojQ.g

'=

c.8

.£ ~

~

;;

.~~

., .,

N.9c.

~

,!,"=c.=

'0 ~

~=

b/)-~

~

~

O

CD

N

N

C

D

O

~

~

"' .-§

~

.~

I I...

...'-' .~

I)

...~

>-

..

I I

Se!J>

-bo*Ee!J'r)

."gcoo.""'1

1 '1.6Jaua

aaJ.:1 5

:E

.9 ~

.t=

5

§

~

r

~

\ }

~~~

lO. Electrostatic Association of Peripheral Proteins 297

depicted in Fig. 5 (left). Several acidic residues (red minus signs) on either side ofthe basic effector domain are confined within a few Debye lengths ( II K = I nmfor 0.1 M salt) of the negatively charged membrane surface and thus are expectedto contribute a significant electrostatic repulsion to the membrane interaction ofthe protein. In addition, the length of the "string" ( 150 amino acids) that connectsthe myristate and basic cluster should decrease the coupling between the twomembrane-binding motifs (myristate and basic cluster), and hence the total bindingenergy should be significantly less that the sum of the individual binding energies.

In many other respects, though, the basic effector domain peptide, MARCKS(151-175), faithfully mimics how the intact protein interacts with membranes, asdiscussed in detail elsewhere (Arbuzova et al., 1998). In brief, the peptide en-capsulates three important aspects of the MARCKSlmembrane interaction. First,binding of calcium-calmodulin (Ca-CaM) to the peptide and the protein occurswith the same high affinity (1-10 nM Kd). Increasing the level of Ca-CaM pro-duces translocation of both the peptide and purified protein from phospholipidvesicles to the surrounding aqueous solution; in many cell types it also producestranslocation of the intact protein from membrane to cytoplasm. The mechanismby which Ca-CaM rapidly dissociates the bound effector domain peptide fromthe membrane-solution interface is described in detail elsewhere (Arbuzova et al.,1997). Second, phosphorylation of the effector region by PKC introduces threephosphates (six negative charges) into the cluster of basic residues. This weakensthe binding of the effector domain peptide to negatively charged membranes andalso produces translocation of the intact protein from phospholipid vesicles (Kimet al., 1994a,b ). The experimental evidence suggests that phosphorylation exerts itseffects in both cases through a simple electrostatic mechanism; FDPB calculationssupport this model (Murray, unpublished calculation). In many cell types, PKCphosphorylation of MARCKS produces translocation from membrane to cyto-plasm (Allen and Aderem, 1995; Guadagno et al., 1992; Rosen et al., 1990;Swierczynski and Blackshear, 1995). Third, recent results show that MARCKS(151-175) can bindPIP2 with high affinity (Arbuzovaetal., 2000). For example, in-troducing 1% PIP2 into a PC vesicle increases the binding ofMARCKS(151-175)by a factor of 104. The binding has been measured with a centrifugation techniqueusing a radioactive peptide and with a fluorescent technique using acrylodan-labeled peptide. Several lines of evidence suggest that one MARCKS(151-175)binds 3-4 PIP2 to form an electroneutral complex (Wang et al., 2001). For ex-ample, the fluorescence from PCIPIP2 vesicles containing NBD-Iabeled PIP2 isquenched by addition ofMARCKS(151-175), which suggests the peptide inducesa strong demixing of PIP2 in the plane of the membrane (Wang et al., 2001).

The interaction ofPIP2 with MARCKS(151-175) may be compared to its inter-action with neomycin and the PH domain of PLC-o 1, two other well-characterizedmolecules that bind PIP2 with high affinity. Addition of"' 10- 5, 10- 6, and 10- 8 M

PIP2 in the form ofPCIPIP2 vesicles binds 50% of neomycin (Gabev et al., 1989),

298 Murray et at.

the PR domain (Garcia et at., 1995), and MARCKS effector domain (Arbuzovaet at., 2000), respectively, that is, MARCKS(151-175) partitions most stronglyonto the PC/PIP2 membranes. Both MARCKS(151-175) and neomycin interactas strongly with PI(3,4)P2 as with PI(4,5)P2, whereas the PR domain of PLC-8 isspecific for PI(4,5)P2. Neomycin and the PR domain form 1 : 1 complexes withPIP2 (Feiguson et at., 1996; Gabev et at., 1989; Garcia et at., 1995; Lemmonet at., 1995), whereas MARCKS(151-175) interacts with several PIP2 (Wanget at., 2001). Electrostatic interactions are important, because increasing the saltconcentration from 100 to 500 mM decreases the binding ofMARCKS(151-175)to 99.9: 0.1 PC/PIP2 vesicles one hundred-fold. The role of other, more specificlipid-protein interactions remains to be determined.

Studies of the inhibition of PLC-mediated hydrolysis of PIP2 suggest that thestrong interaction between PIP2 and MARCKS(151-175) occurs with the intactMARCKS protein as well. Both the protein and the peptide (see Fig. 7) decreasethe PLC-induced hydrolysis ofPIP2 in phospholipid vesicles (Glaser et at., 1996).The simplest interpretation of these results is that both the effector domain in theintact protein and the effector domain peptide bind PIP2 with high affinity andcompete successfully with the catalytic domain of PLC for this lipid. Experimentswith monolayers (Wang et at., 2001) show that these observations are not artifacts,due to peptide-induced aggregation of vesicles, a problem which unfortunatelycomplicates many experiments with basic peptides and lipid vesicles (Murrayet at., 1999).

Can the strong interaction between the effector domain of MARCKS and PIP2observed with in vitra model systems be extrapolated to living cells? This remainsto be determined, but the hypothesis that MARCKS can bind a significant fractionof the PIP2 in atypical cell suggests three testable corollaries. First, MARCKSand PIP2 should have comparable cellular conceptrations; they do in many celltypes. In nerve cells, for example, both MARCKS and PIP2 are present at about10 ,uM (Albert et at., 1987). Second, overexpression of MARCKS should induceenhanced production of PIP2 if MARCKS buffers a significant fraction of thePIP2. It does, at least in one cell type (Laux et at., 2000). Third, MARCKS andPIP2 should be colocalized in cells. MARCKS is not uniformly distributed in tlieplasma membrane of some cell types. In fibroblasts, for example, it is concentratedin ruffles (Myat et at., 1997). If it binds a significant fraction of PIP2, then PIP2should be colocalized with MARCKS in ruffles, and it is (Ronda et at., 1999; Tallet at., 2000). Of course, there are many other, unrelated mechanisms by whichPIP2 and PIP3 (Czech, 2000; Martin, 1998) could be sequestered in ruffles, forexample, through localized synthesis.

Although the functional role of the sequestration of PIP2 by MARCKS canonly be determined in cells or organisms, experiments on model systems andtheoretical calculations can help guide cell biologists in designing and interpret-ing in viva studies. Phosphoinositides are important for the production of second

FIGURE 1 The negative electrostatic potential of a 2:1 PC/PS membrane is significantly per-turbed by multivalent molecules. The red (blue) lines represent the two-dimensional-25 (+25) mY

equipotential contours obtained from FDPB calculations. In each panel, the membrane compositionis 2:1 PC/PS and the ionic strength [KCI], is 100 mM. (A) The -25 mY electrostatic equipotentialcontour is essentially flat and located -10 A above the membrane surface. (B) A trivalent PIP2 lipid

(denoted by its yellow head group) greatly increases the negative potential of the membrane in itsvicinity. (C) A membrane-adsorbed basic peptide (heptalysine; blue) introduces a significant positivepotential at the membrane surface.

FIGURE 3 Molecular model of the N-tenninal region of Src interacting with the upper leafletof a 2: 1 PC/PS bilayer. The front portion of the bilayer has been removed to expose the myristate(green). In agreement with ESR and CD measurements, the 18-residue peptide (myristate-GSSKSKPKDPSQRRRSLE-NHz) is shown in an extended confonnation. The six basic residues areblue and the two acidic residues are red. In the bilayer, PS is identified by its exposed nitrogen (blue);oxygen is red, phosphorus is yellow, and carbon and hydrogen are gray. Reprinted from Murray etal., (1997), copyright 1997, with pennission from Elsevier Science.

MARCKS MARCKS(151-175)

~-c

FIGURE 5 Models for the membrane association of MARCKS. Left: A cartoon model of theinteraction of the intactMARCKS protein with a PC (white headgroups)/P1P2 (red head groups) mem-brane. The myristate is colored yellow, the location of basic and acidic residues throughout region125-200 are represented schematically by blue plus signs and red minus signs, respectively, and thearomatic residues in the basic effector domain are represented as green ovals penetrating the mem-

brane interface. Right: An atomic level model for the interfacial association of MARCKS( 151-175)based on the EPR measurements of spin-Iabeled peptides by Qin and Cafiso (1996). Reprinted fromArbuzova et at., (1998), copyright 1998, with permission from Elsevier Science.

FIGURE 8 FDPB calculations of the electrostatic properties of the effector region of MARCKS,MARCKS(151-175), adsorbed to the surface of a 2:1 PC/PS membrane in 100 mM KCl. The bluemesh and red line represent, respectively, +25-mV and -25-mV equipotential contours. Reprintedfrom Arbuzova et al... (1998), copyright 1998, with permission from Elsevier Science.

00

!l)T"

0~

ID(\I0 I!)

0~ ID.- OC\J

'Oc:.'

.Ubl)"'.'

.,.,.c:~.=

c:.,.c:N

""'..

~

t: .~

~

,..

~°,..:.§[Q

:g~\O

~.'",U

-tU.,O

\~

.C:.'C

=""

0~

...0\,"

,.. "'

.a tU

't

.tU6i».-5~

~

U

~

~

.~

U

"" '0

.::n'""'

~

U

"' ""

.~

-

~

] ~

u.S

,g

e- .~

~.""",

",c:",,0

>

c: ~

0

"' .=

0

c: ;=

: c:.£t:U

oa ooU

.

.., =

0

:>0. ..,

.=

"' ~

.~

., .,

""'. .~

U

.,

\0.0

~~

.~

"'

~

0\.~

""

>

0 c:

eO\

.Q

O..'"".,.~

C

:-~U

e""~

~.~

.S

~

tU

0

~

.

'O,,'0.,0.,tU

~.,

--"U.c:~

U~

..s..,~,..~

'.:"

..., "'..,

.~

...

'OtU

O.'",~

U.'

=

"' .,

V)

"'.~

=

tU

~

r--

~

tU

A.~

A~

.~

=~

V

) "'U

u. 1;=

:,-,r--

~

c: "

e""

'""' .~

V

) '0

I~~

~~

oo

"' '0-"~

e,;:

V).,O

tU..,cn

N>

.~",~

~c:

~>

.~-'O

UUo

cn -.~

~ .~

.~

~O

>o""~

bl)~U

~.£~

;=:<

.5.~~

E

tU

~

.~

~

:g

E

<>

..S=

:9>.tU

&~

~

"'

.8 ~

.0

~

-5-.~

~

~u-"~

c: ~

§'

:E

N

~

'""' ..,

.~,

...:a ., ..." ~

~

E

"' ,

e ~

0]

-~

.= 8.~

.., ~

0 ~

=

Q)

...c: ~

=

c:

."'

., otU

oo"'...

E>

0 ..,e

.., .£ "".,

"" ~

.~ bl) .,

'e"0 H

-.0

c: ~1-

~

'0 ~

~

.~

tU

V)

.., tU

§

'2 .E

:!Jr--~

UoU

.~

.B1

tU

~

§ .§

.s ]

~

V)]0.~

..:"'- .

~~

~.,.'C

=~

>.

~~

..."0~"'

~bl)

cn tU

0

'0 .~

,

c: 0

~

...: tU

~

~

~

.,

OU

U

""

., N

.,

e .~

~

~

;r;- -5 "0

~

:9 .5

~

<~

r--.J".'".§..S""a

~-51~

3'Oo~

R"'~0_.,~

0 P

. V)

~

.~

e A

-5

.£c:tU

~""t:~

0

.S

~

~

.~

tU

y .~

.~

~

~"~

u'OtU

>B

'O~

s

~

~

§ u

.£ c:

§..."'<

~

",~tU

o,g~

~c:~

~u~

b'""

=

~

., ~

.;

'""' \0

.~~

.U

"0

~

~

=

~

~

e.Q

'O

tU"'-.,

"" .,

=0~

:c,

.~

~

.Q--"

°e~

.U

U-..'..,

' 0

~c=

.~

"".,~~

.~~

~., :g P. "

e .~ '.: ~

tUlO

.'_~o

~

., ~

'0

," V

) .,

"'

$N-5~

etUer--[Q

>.

~

-[Q

~

~~

.,- I".,

r-- ~

-"

"' .."":;::

..u'.,~

., =

0~

0

~

"bJI ~

-5 ~

.S

cn

5 .~

~

E§ ] "0 ;Q

:E §

r.. >

. .,

."'u

:a U

'-' -"'

., 8

..., ..:

s: ~

] ]

~ ~

~

.;;

~E

.u ta ~

~

~

-5 <

300 Murray et at.

FIGURE 8 FDPB calculations of the electrostatic properties of the effector region of MARCKS,MARCKS(15l-175), adsorbed to the surface of a 2: 1 PC/PS membrane in 100 mM KC1. The bluemesh and red line represent, respectively, +25 mY and -25 mY equipotential contours. Reprintedfrom Arbuzova et al. (1998), copyright 1998, with permission from E1sevier Science. (See color plate.)

messengers, exocytosis, and endocytosis, the regulation of ion channels, and therecruitment and anchoring of specific proteins to membrane surfaces (Cockcroft,2000; Czech, 2000; Martin, 1998). Because they playa role in so many cellularfunctions, it would be surprising if their distribution and accessibility were notcarefully regulated by the cell. Experiments on model systems suggest mecha-nisms by which this could occur. For example, as described above and depictedin Fig. 7, MARCKS and MARCKS(151-175) inhibit the PLC-induced hydrolysisof PIP2 in phospholipid vesicles. Figure 7 also illustrates mechanisms by whichthe MARCKS-induced inhibition of PLC hydrolysis is relieved. Both PKC phos-phorylation of MARCKS(151-175) and the binding of Ca-CaM to MARCKS-(151-175) release the bound PIP2 by causing desorption of the effector domainfrom the membrane. It remains to be determined if these phenomena also occur to asignificant degree in the living cell and if the cell can thus produce localized burstsOfPIP2 in response to these two signals. FDPB calculations of the electrostatic prop-erties of MARCKS(151-175)/membrane systems help visualize the mechanismsof PIP2 sequestration and release. Figure 8 depicts the +25 m V (blue mesh) and-25 mV (red line) equipotential contours around a membrane-adsorbedMARCKS peptide. The peptide (net charge + 13) dramatically perturbs the neg-ative potential of the membrane and provides a strong positive potential at themembrane surface that could serve as a highly localized sink for the multivalentacidic PIP2.

v. FUTURE DIRECTIONS

A. Membrane Targeting

In summary, the function of MARCKS is not well understood and could in-volve both its ability to bind actin (Aderem, 1992; Hartwig et at., 1992; Naimand Aderem, 1992) and its ability to bind PIP2 as discussed here. In contrast,the biological functions of Src, K-ras4B, and HIV-1 Gag have been well estab-lished. However, the mechanisms by which any of these proteins are targeted to the

30110. Electrostatic Association of Peripheral Proteins

plasma membrane are not known. Recent work from the Hancock and Philips labshas shown that H- and N-ras are targeted to the plasma membrane by a mechanismthat involves their palmitoylation on internal membranes and subsequent transferto the plasma membrane by exocytosis (Apolloni et at., 2000; Choy et at., 1999).Silvius and co-workers (Leventis and Silvius, 1998; Roy et at., 2000) have pro-posed a simple biophysical mechanism by which K-ras4B may be targeted to theplasma membrane: Stronger electrostatic attraction drives the preferential bindingto plasma membrane. This proposal relies on the premise that the electrostatic po-tential of the cytoplasmic surface of the plasma membrane is more negative than thecytoplasmic surfaces of other intracellular membranes. Because the surface areaof internal membranes is about 10- fold greater than the surface area of the plasmamembrane in a typical cell, Silvius and co-workers estimate that the surface poten-tial of the plasma membrane would have to be about 20 m V more negative than thesurface potential of internal membranes in order for an electrostatic mechanism todrive plasma membrane targeting. This could occur if, for example, the "flippase"for PS is more active in the plasma membrane (Daleke and Lyles, 2000), resultingin a higher negative surface charge density. Although the magnitude of the surfacepotentials of the plasma and internal membranes is not well established, studies ofthe cellular distribution of K-ras4B/GFP constructs suggest a nonspecific electro-static mechanism is plausible. The fact that the fluorescent constructs are targetedto the plasma membrane even when the C-terrninal sequence of basic residuesis scrambled argues, at least, against a targeting mechanism that involves specificprotein-protein recognition. If the nonspecific electrostatic targeting mechanism isoperative, it could also act to target Src, MARCKS, HIV-l Gag, and other proteinswith basic residues to the plasma membrane.

It is important to note that even if electrostatics produces a bias for the plasmamembrane, other interactions can override this mechanism and direct a proteinto a different subcellular localization. For example, v- and c-Src have the sameN-terrninal sequences (myristate plus basic cluster, see Table I), but are targeted todifferent membranes. v-Src and activated c-Src localize to the plasma membrane,whereas the inactive form of c-Src is localized primarily to endosomal membranes.These observations suggest that other factors, for example, protein-protein inter-actions, contribute to their subcellular localization. In its inactive state, Src forms acompact folded structure through a series of intramolecular interactions mediatedby its SH2 and SH3 domains. Activation of the enzyme leads to the disruptionof these interactions and concomitant availability of the SH2 and SH3 domainsto interact with other ligands. Resh and co-workers found that the SH2 domainmediates cytoskeletal association of v-Src, demonstrating that domains other thanthe membrane-interacting motifs contribute to the subcellular localization of Src.Interestingly, a "Src-like" mutant construct of MARCKS, one in which the longstretch of residues between the myristate and basic effector domain is removed,is targeted to the nucleus rather than the plasma membrane as one might have

302 Murray et al.

expected (Seykora et at., 1996). The targeting of MARCKS is complicated by atleast two additional factors. First, it is possible that MARCKS is targeted to theplasma membrane through its interaction with PIP2. The plasma membrane hasa high relative concentration of PIP2 and this directs the plasma membrane target-ing ofPLC-8: Its PH domain anchors the protein to the plasma membrane throughits high-affinity interaction with PIP2 (Lemmon and Ferguson, 2000). The highaffinity of the effector domain of MARCKS for PIP2 could act in a similar man-ner. Second, MARCKS can bind actin, and protein-protein interactions could bothtarget it to different membranes and modify the lateral distribution of MARCKSin the plasma membrane. The role of phosphoinositides in regulating membrane-cytoskeletal interactions has long been recognized and was elegantly demonstratedby recent experiments using both GFP constructs and laser tweezers (Raucheret at., 2000). In conclusion, the cellular mechanisms for plasma membrane tar-geting of K-ras4B, Src, MARCKS, and HIV-I Gag remain to be elucidated, as dothe mechanisms by which the lateral distribution is modified; this area of investi-gation will surely continue to benefit from the new cell biology approaches being

developed.

B. Computationa[ Studies

The FDPB approach provides a framework for (1) describing in quantitativeterms the physical forces that drive nonspecific protein-lipid interactions, (2) de-veloping structure-based models for peptides and proteins at the membrane inter-face, and (3) identifying protein sequence and structure patterns that are predictiveof membrane binding potential. The methodology described here for calculat -

ing the membrane interaction of peptides can be applied to proteins of knownstructure as well. The next step toward gaining both a deeper and broader un-derstanding of the mechanisms proteins use to bind to membranes involves theintegration of detailed computer calculations of the underlying physical interac-tions, as described here, with tools for exploiting and analyzing the vast amountsof genomic data becoming available. These tools include facilities for sequenceanalysis, structure comparison and prediction, and surface property analysis. Thesynthesis of these computational approaches can be used to describe similaritiesand differences within and across whole protein families involved in membraneassociation (e.g., for the C2 domains; Murray and Honig, 2001). Representativestructures for peripheral proteins have recently become available, which allows forthis type of computational analysis. Calculation of the interaction of these proteinswith realistic models of membranes will provide information unobtainable throughstructural analysis of the proteins alone. Building on the examples described inthis chapter, computational results can be used in new ways to continue to aidin the design and interpretation of experiments, leading to new insights into themolecular basis of protein-membrane association.

30310. Electrostatic Association of Peripheral Proteins

AcknowledgmentsThis work was supported by National Institutes of Health Grant GM24971 and National Science

Foundation Grant MCB9729538 to S.M., National Science Foundation Grant MCB9808902 to B.H.and Helen Hay Whitney and Sloan/U.S. Department of Energy Computational Biology postdoctoralfellowships to D.M.

ReferencesAderem, A. (1992). Signal transduction and the actin cytoskeleton: The roles ofMARCKS and profilin.

Trends Biochem. Sci. 17, 438-443.Albert, K. A., Nairn, A. C., and Greengard, P. (1987). The 87-kDa protein, a major specific substrate

for protein kinase C: Purification from bovine brain and characterization. Proc. Natl. Acad. Sci.USA 84, 7046-7050.

Allen, L. A., and Aderem, A. (1995). Protein kinase C regulates MARCKS cycling between the plasmamembrane and Iysosomes in fibroblasts. EMBO I. 14, 1109-1120.

Allen, L. A., and Aderem, A. (1996). Mechanisms of phagocytosis. Curl: Opin. lmmunol. 8, 36--40.Apolloni, A., Prior, I. A., Lindsay, M., Parton, R. G., and Hancock, I. F. (2000). H-ras but not K-ras

traffics to the plasma membrane through the exocytic pathway. Mol. Cell. Biol. 20, 2475-2487.Arbuzova, A., Wang, I., Murray, D., Iacob, I., Cafiso, D. S., and McLaughlin, S. (1997). MARCKS,

membranes, and calmodulin: Kinetics of interaction. I. Biol. Chem. 272, 27167-27177.Arbuzova, A., Murray, D., and McLaughlin, S. (1998). MARCKS, membranes andcalmodulin: Kinetics

of interaction. Biochim. Biophys. Acta 1376, 369-379.

Arbuzova, A., Wang, L., Wang, I., Hangyas-Mihalyne, G., Murray, D., Honig, B., and McLaughlin, S.(2000). Membrane binding of peptides containing both basic and aromatic residues. Experimentalstudies with peptides corresponding to the scaffolding region of caveolin and the effector regionof MARCKS. Biochemistry 39, 10330-10339.

Ben-Tal, N., Honig, B., Peitzsch, R: M., Denisov, G., and McLaughlin, S. (1996). Binding of smallbasic peptides to membranes containing acidic lipids: Theoretical models and experimental results.Biophys. I. 71, 561-575.

Ben- Tal, N., Honig, B., Miller, C., and McLaughlin, S. (1997). Electrostatic binding of proteins tomembranes: Theoretical prediction and experimental results with charybdotoxin and phospholipidvesicles. Biophys. I. 73, 1717-1727.

Bhatuagar, R. S., and Gordon, I. I. (1997). Understanding covalent modifications of proteins by lipids:Where cell biology and biophysics mingle. Trends Cell Biol. 7, 14-21.

Brown, D. A., and London, E. (2000). Structure and function of sphingolipid- and cholesterol-richmembrane rafts. I. Biol. Chem. 275, 17221-17224.

Brown, M. T., and Cooper, I. A. (1996). Regulation, substrates and functions ofSrc. Biochim. Biophys.Acta 1287, 121-149.

Bryant, M., and Ratner, L. (1990). Myristoylation-dependent replication and assembly of human im-munodeficiency virus I. Proc. Natl. Acad. Sci. USA 87, 523-527.

Buser, C. A., Sigal, C. T., Resh, M. D., and McLaughlin, S. (1994). Membrane binding ofmyristy1atedpeptides corresponding to the NH2-terminus of Src. Biochemistry 33, 13093-13101.

Buss, I. E., Karnps, M. P., Gould, K., and Sefton, B. M. (1986). The absence of myristic acid decreasesmembrane binding of p60src but does not affect tyrosine protein kinase activity. I. Virol. 58,468-474.

Choy, E., Chiu, V. K., Silletti, I., Feoktistov, M., Morimoto, T., Michaelson, D., Ivanov, I. E., andPhilips, M. R. (1999). Endomembrane trafficking ofras: The CAAX motif targets proteins to theER and Golgi. Cell 98, 69-80.

Cockcroft, S. (2000). "Biology of Phosphoinositides." Oxford University Press, Oxford.Cooper, I. A. (1990). The Src family of protein-tyrosine kinases.ln "Peptides and Protein Phosphory-

lation" (B. E. Kemp, ed.), pp. 85-113. CRC Press, Boca Raton, FL.

304 Murray et at.

Czech, M. P. (2000). PIP2 and PIP3: Complex roles at the cell surface. Cell 100, 603-606.Daleke, D. L., and Lyles, J. V. (2000). Identification and purification of aminophospholipid flippases.

Biochim. Biophys. Acta 1486,108-127.Ferguson, K. M., Lemmon, M. A., Schlessinger, J., and Sigler, P. B. (1996). Structure of the high

affinity complex of inositol trisphosphate with a phospholipase C pleckstrin homology domain.

Cell 83, 1037-1046.Gabev, E., Kasianowicz, J., Abbott, T., and McLaughlin, S. (1989). Binding of neomycin to phos-

phatidylinositoI4,5-bisphosphate (PIP2). Biochim. Biophys. Acta 979,105-112.

Garcia, P., Gupta, R., Shah, S., Morris, A. J., Rudge, S. A., Scarlata, S., Petrova, V., McLaughlin, S.,and Rebecchi, M. J. ( 1995). The pleckstrin homology domain of phospholipase C-81 binds with

high affinity to phosphatidylinositol 4,5-bisphosphate in bilayer membranes. Biochemistry 34,

16228-16234.Gamier, L., Bowzard, B., and Wills, J. W. (1998). Recent advances and remaining problems in HIV

assembly. AIDS 12, S5-S16.Ghomashchi, F., Zhang, X., Liu, L., and Gelb, M. H. (1995). Binding of prenylated and polybasic

peptides to membranes: Affinities and intervesicle exchange. Biochemistry 34, 11910-11918.Gilson, M., and Honig, B. (1988). Calculation of the total electrostatic energy of a macromolecular

system: Solvation energies, binding energies, and conformational analysis. Proteins 4,7-18.

Glaser, M., Wanaski, S., Buser, C. A., Boguslavsky, V., Rashidzada, W., Morris, A., Rebecchi, M.,Scarlata, S. F., Runnels, L. W., Prestwich, G. D., Chen, J., Aderem, A., Ahn, J., and McLaughlin,S. (1996). MARCKS produces reversible inhibition of phospholipase C by sequestering phos-phatidy1inositoI4,5-bisphosphate in lateral domains. J. Bioi. Chem. 271,26187-26193.

Guadagno, S. N., BonIer, C., and Weinstein, I. B. (1992). Altered regulation of a major substrate ofprotein kinase C in rat 6 fibroblasts overproducing PKC beta I. J. Bioi. Chem. 267, 2697-2707.

Hancock, J. F., Paterson, H., and Marshall, C. J. (1990). A polybasic domain or palmitoylation isrequired in addition to the CAAX motif to localize p21ras to the plasma membrane. Cell 63,

133-139.Hartwig, J. H., Thelen, M., Rosen, A., Janmey, P. A., Nairn, A. C., and Aderem, A. (1992). MARCKS

is an actin filament crosslinking protein regulated by protein kinase C and calcium-calmodulin.

Nature 356,618-622.Heimburg, T., and Marsh, D. (1996). Thermodynamics of the interaction of proteins with lipid mem-

branes. In "Biological Membranes" (K. Merz, Jr., and B. Roux, eds.), pp. 405-464. Birkhauser,

Boston.Hill, C. P., Bancroft, D. P., Christensen, A. M., and Sundquist, W. I. (1996). Crystal structures of the

trimeric human immunodeficiency virus type I matrix protein: Implications for membrane binding

and assembly. Proc. Nati. Acad. Sci. USA 93,3099.HoIst, M. (1993). Numerical solutions to the finite-difference Poisson-Boltzmann equation.

Thesis/Dissertation, University of Illinois, Champaign-Urbana, IL.HoIst, M., and Saied, F. ( 1993). Multigrid solution of the Poisson-Boltzmann equation. J. Comp. Chem.

14,105-113.Honda, A., Nogami, M., Yokozeki, T., Yamazaki, M., Nakamura, H., Watanabe, H., Kawamoto, K.,

Nakayama, K., Morris, A. J., Frohman, M. A., and Kanaho, Y. (1999). Phosphatidylinositol4-phosphate 5-kinase alpha is a downstream effector of the small G protein ARF6 in membrane

ruffle formation. Cell 99, 521-532.Honig, B. H., and Nicholls, A. (1995). Classical electrostatics in biology and chemistry. Science 268,

1144-1149.Hurley, J. H., and Misra, S. (2000). Signaling and subcellular targeting by membrane-binding domains.

Annu. Rev. Biophys. Biophys. Chem. 29,49-79.Jakobsson, E. (1997). Computer simulation studies of biological membranes: Progress, promise and

pitfalls. Trends Biochem. Sci. 22, 339-344.

10. Electrostatic Association of Peripheral Proteins 305

Johnson, J. E., and Comell, R. B. (1999). Amphitropic proteins: Regulation by reversible membraneinteractions. Mol. Membrane Bioi. 16,217-235.

Kamps, M. P., Buss, J. E., and Sefton, B. M. (1985). Mutation of NH2-terminal glycine of p60srcprevents both myristoylation and morphological transformation. Proc. Natl. Acad. Sci. USA 82,4625-4628.

Kaplan, J. M., Varmus, H. E., and Bishop, J. M. ( 1990). The Src protein contains multiple domains forspecific attachment to membranes. Mol. Cell. Bioi. 10, 1000-1009.

Kholodenko, B. N., Hoek, J. B., and Westerhoff, H. V. (2000). Why cytoplasmic signaling proteinsshould be recruited to cell membranes. Trends Cell Bioi. 10,173-178.

Kim, J., Mosior, M., Chung, L. A., Wu, H., and McLaughlin, S. (1991). Binding of peptideswith basic residues to membranes containing acidic phospholipids. Biophys. J. 60, 135-148.

Kim, J., Blackshear, P. J., Johnson, J. D., and McLaughlin, S. (1994a). Phosphorylation reversesthe membrane association of peptides that correspond to the basic domains of MARCKS andneuromodulin. Biophys. J. 67,227-237.

Kim, J., Shishido, T., Jiang, X., Aderem, A., and McLaughlin, S. (1994b). Phosphorylation, high ionicstrength, and calmodulin reverse the binding ofMARCKS to phospholipid vesicles. J. Bioi. Chem.

269,28214-28219.Kleinschmidt, J. H., and Marsh, D. (1997). Spin-Iabel electron spin resonance studies on the interactions

of lysine peptides with phospholipid membranes. Biophys. J. 73, 2546-2555.Laux, T., Pukami, K., Thelen, M., Golub, T., Prey, D., and Caroni, P. (2000). GAP43, MARCKS,

CAP23 modulate PI( 4,5)P2 at plasmalemmal rafts, and regulate cell cortex actin dynamics througha common mechanism. J. Cell Bioi. 149,1455-1472.

Lemmon, M. A., and Perguson, K. M. (2000). Pleckstrin homology domains: Phosphoinositide-regulated membrane tether. In "Biology of Phosphoinositides." (S. Cockcroft, ed.), pp. 131-165.Oxford University Press, Oxford.

Lemmon, M. A., Perguson, K. M:, O'Brien, R., Sigler, P. B., and Schlessinger, J. (1995). Specificand high-affinity binding of inositol phosphates to an isolated pleckstrin homology domain. Proc.Natl. Acad. Sci. USA 92, 10472-10476.

Leventis, R., and Silvius, J. R. (1998). Lipid-binding characteristics of the polybasic carboxy-terminalsequence of K-ras4B. Biochemistry 37,7640-7648.

Magee, T., and Marshall, C. J. (1999). New insights into the interaction ofras with the plasma membrane.Cell 98, 9-12.

Martin, T. P. J. (1998). Phosphoinositide lipids as signaling molecules: Common themes for signaltransduction,cytoskeletal regulation, and membrane trafficking. Annu. Rev. Cell Dev. Bioi. 14,231-264.

May, S., Harries, D., and Ben-Shaul, A. (2000). Lipiddemixing and protein-protein interactions in theadsorption of charged proteins on mixed membranes. Biophys. J. 79,1747-1760.

McCloskey, M. A., and Poo, M. M. (1986). Rates of membrane-associated reactions: Reduction ofdimensionality revisited. J. Cell Bioi. 102, 88-96.

McIntosh, T. J., and Simon, S. A. (1986). Area per molecule and distribution of water in fully hydrated

di1auroylphosphatidylethanolamine bilayers. Biochemistry 25,4948-4952.McIntosh, T. J., and Simon, S. A. (1994). Hydration and steric pressures between phospholipid bilayers.

Annu. Rev. Biophys. Biomol. Struct. 23,27-51.McLaughlin, S. (1989). The electrostatic properties of membranes. Annu. Rev. Biophys. Biophys.

Chem. 18,113-136.McLaughlin, S., and Aderem, A. (1995). The myristoyl-electrostatic switch: A modulator of reversible

protein-membrane interactions. Trends Biochem. Sci. 20, 272-276.Merz, K., and Roux, B. (1996). "Biological Membranes: A Molecular Perspective from Computation

and Experiment." Birkhauser, Boston.

306 Murray et at.

Misra, V., and Honig, B. (1995). On the magnitude of the electrostatic contribution to ligand-DNA

interactions. Proc. Natl. Acad. Sci. USA 92,4691-4695.Misra, V., Sharp, K. A., Friedman, R. A., and Honig, B. (1994). Salt effects on ligand-DNA binding:

Minor groove binding antibiotics. I.. Mol. Bioi. 238, 245-263.Montich, G., Scarlata, S., McLaughlin, S., Lehnnann, R., and Seelig, J. (1993). Thennodynarnic

characterization of the association of small basic peptides with membranes containing acidic

lipids. Biochim. Biophys. Acta 1146,17-24.Morrison, D. K., and Cutler, R. E., Jr. (1997). The complexity of Raf-1 regulation. Curl: Opin. Cell

Bioi. 9,174-179.Mosior, M., and McLaughlin, S. (1991). Peptides that nrimic the pseudosubstrate region of protein

kinase C bind to acidic lipids in membranes. Biophys. 1.60, 149-159.Mosior, M., and McLaughlin, S. (1992). Binding of basic peptides to acidic lipids in membranes:

Effects of inserting alanine(s) between the basic residues. Biochemistry 31, 1767-1773.Murray, D., Ben-Tal, N., Honig, B., and McLaughlin, S. (1997). Electrostatic interaction of myristoy-

lated proteins with membranes: Simple physics, complicated biology. Structure 5,985-989.Murray, D., Hennida-Matsumoto, L., Buser, C. A., Tsang, J., Sigal, C., Ben- Tal, N., Honig, B., Resh,

M. D., and McLaughlin, S. (1998). Electrostatics and the membrane association of Src: Theory

and experiment. Biochemistry 37,2145-2159.Murray, D., Arbuzova,A., Mihaly, G., Ghambir,A., Ben-Tal, N., Honig, B., andMcLaughlin, S. (1999).

Electrostatic properties of membranes containing acidic lipids and adsorbed basic peptides: Theory

and experiment. Biophys. 1.77, 3176-3188.Murray, D., and Honig, B. (2001). Electrostatic control of the membrane targeting of C2 domains.

Molecular Cell. In press.Myat, M. M., Anderson, S., Allen, L. H., and Aderem, A. (1997). Identification of the basolateral

targeting determinant of a peripheral membrane protein, MacMARCKS, in polarized cells. Curl:

Bioi. 7,611-614.Naim, A. C., and Aderem, A. (1992). Calmodulin and protein kinase C cross-talk: The MARCKS

protein is an actin filament and plasma membrane cross-Iinking protein regulated by proteinkinase C phosphorylation and by calmodulin. Ciba Found. Symp. 164,145-154.

Oancea, E., and Meyer, T. (1998). Protein kinase Cas a molecular machine for decoding calcium and

diacylglycerol signals. Cell 95, 307-318.Parsons, J. T., and Parsons, S. J. (1997). Src family protein tyrosine kinases: Cooperating with growth

factor and adhesion signaling pathways. Curl: Opin. Cell Bioi. 9,187-192.Peitzsch,R. M., andMcLaughlin, S. (1993). Binding ofacylated peptidesand fatty acids to phospholipid

vesicles: Pertinence to myristoylated proteins. Biochemistry 32,10436-10443.Peitzsch, R. M., Eisenberg, M., Sharp, K. A., and McLaughlin, S. (1995). Calculations of the electro-

static potential adjacent to model phospholipid bilayers. Biophys. 1.68, 729-738.Qin, Z., and Cafiso, D. S. (1996). Membrane structure of the protein kinase C and calmodulin binding

domain of myristoylated alanine rich C kinase substrate determined by site-directed spin labeling.

Biochemistry 35,2917-2925.Raucher, D., Stauffer, T., Chen, W., Shen, K., Guo, S., York, J. D., Sheetz, M. P., and Meyer, T.

(2000). Phosphatidylinositol 4,5-bisphosphate functions as a second messenger that regulatescytoskeleton-plasma membrane adhesion. Cell 100, 221-228.

Rein, A., McClure, M. R., Rice, N. R., Luftig, R. B., and Schultz, A. M. (1986). Myristylation site inPr65gag is essential for virus particle fonnation by Moloney murine leukemia virus. Proc. Natl.

Acad. Sci. USA 83, 7246-7250.Resh, M. D. (1999). Fatty acylation ofproteins: New insights into membrane targeting ofmyristoylated

and palmitoylated proteins. Biochim. Biophys. Acta 1451,1-16.Rosen, A., Keenan, K. F., Thelen, M., Nairn, A. C., and Aderem, A. (1990). Activation of protein kinase

C results in the displacement of its myristoylated, alanine-rich substrate from punctate structuresin macrophage filopodia. I. Exp. Med. 172, 1211-1215.

30710. Electrostatic Association of Peripheral Proteins

Roux, M., Neumann, I. M., Bloom, M., and Devaux, P. F. (1988). 2H and 31p NMR study of pentalysineinteraction with head group deuterated phosphatidylcholine and phosphatidylserine. Eul: Biophys.

I. 16,267-273.Roy, M., Leventis, R., and Silvius, I. R. (2000). Mutational and biochemical analysis of plasma

membrane targeting mediated by the farnesylated, polybasic carboxy terminus of K-ras4B.

Biochemistry 39, 8298-8307.Seykora,I. T.,Myat,M.M.,Allen,L. A., Ravetch,I. V., and Aderem,A. (1996). Molecular determinants

of the myristoyl-electrostatic switch ofMARCKS. I. Bioi. Chem. 271,18797-18802.Sharp, K. A., and Honig, B. H. (1990). Calculating total electrostatic energies with the nonlinear

Poisson-Boltzmann equation. I. Phys. Chem. 94,7684-7692.Sigal, C. T., Zhou, W., Buser, C. A., McLaughlin, S., and Resh, M. D. (1994). The amino terminal

basic residues of Src mediate membrane binding through electrostatic interaction with acidic

phospholipids. Prac. Natl. Acad. Sci. USA 91,12253-12257.Silvius, I. R., andl'Heureux, F. (1994). Fluorimetric evaluation of the affinities ofisoprenylated peptides

for lipid bilayers. Biochemistry 33,3014-3022.Simons, K., and Ikonen, E. (1997). Functional rafts in cell membranes. Nature 387,569-572.Spearman, P., Horton, R., Ratner, L., and Kuli-zade, I. (1997). Membrane binding of human immunode-

ficiency virus type 1 matrix protein in vivo supports a conformational myristyl switch mechanism.

I. Viral. 71, 6582-6592.Swierczynski, S. L., and Blackshear, P. I. (1995). Membrane association of the myristoylated aIanine-

rich C kinase substrate (MARCKS) protein. Mutational analysis provides evidence for complex

interactions. I. Bioi. Chem. 270,13436-13445.Swierczynski, S. L., and Blackshear, P. I. (1996). Myristoylation-dependent and electrostatic inter-

actions exert independent effects on the membrane association of the myristoylated aIanine-rich

C-kinase substrate. I. Bioi. Chem. 271,23424-23430.Tall, E. G., Spector, I., Pentyala, S. N., Bitter, I., and Rebecchi, M. I. (2000). Dynamics of phos-

phatidylinositoI4,5-bisphospliate in actin-supported structures. Curl: Bioi. 10,743-746.Victor, K., and Cafiso, D. (1998). Structure and position of the N-terminal membrane-binding domain

ofpp60src at the membrane interface. Biochemistry 37,3402-3410.Vlachy, V. (1999). Ionic effects beyond Poisson-Boltzmann theory. Annu. Rev. Phys. Chem. 50,145-

165.Wang, I., Arbuzova, A., Hangyas-Mibalyne, G., and McLaughlin, S. (2001). The effector domain

of myristoylated aIanine-rich C kinase substrate binds strongly to phosphatidylinositol 4,5-

bisphosphate. I. Bioi. Chem. 276,5012-5019.White, S. H., and Wimley, W. C. (1999). Membrane protein folding and stability: Physical principles.

Annu. Rev. Biophys. Biomol. Struct. 28,319-365.White, S. H., Wimley, W. C., Ladokhin., A. S., and Hristova, K. (1998). Protein folding in membranes:

Determining the energetics of peptide-bilayer interactions. Meth. Enzymol. 295,62-87.Wills, I. W., and Craven, R. C. (1991). Form, function, and use of retroviral Gag proteins. AIDS 5,

639-654.Wimley, W. C., and White, S. H. (1996). Experimentally determined hydrophobicity scale for proteins

at membrane interfaces. Nat. Struct. Bioi. 3, 842-848.Zhou, W., and Resh, M. D. (1996). Differential membrane binding of the human immunodeficiency

virus type I matrix protein. I. Viral. 70, 8540-8548.Zhou, W., Parent, L. I., Wills, I. W., and Resh, M. D. (1994). Identification of a membrane-binding

domain within the amino-terminal region of human immunodeficiency virus type I Gag protein

which interacts with acidic phospholipids. I. Viral. 68,2556-2569.